Slc38a9 Deficiency Induces Apoptosis and Metabolic Dysregulation and Leads to Premature Death in Zebrafish

Abstract

Eukaryotic cells control nutritional homeostasis and determine cell metabolic fate through a series of nutrient transporters and metabolic regulation pathways. Lysosomal localized amino acid transporter member 9 of the solute carrier family 38 (SLC38A9) regulates essential amino acids’ efflux from lysosomes in an arginine-regulated fashion. To better understand the physiological role of SLC38A9, we first described the spatiotemporal expression pattern of the slc38a9 gene in zebrafish. A quarter of slc38a9^−/− mutant embryos developed pericardial edema and died prematurely, while the remaining mutants were viable and grew normally. By profiling the transcriptome of the abnormally developed embryos using RNA-seq, we identified increased apoptosis, dysregulated amino acid metabolism, and glycolysis/gluconeogenesis disorders that occurred in slc38a9^−/− mutant fish. slc38a9 deficiency increased whole-body free amino acid and lactate levels but reduced glucose and pyruvate levels. The change of glycolysis-related metabolites in viable slc38a9^−/− mutant fish was ameliorated. Moreover, loss of slc38a9 resulted in a significant reduction in hypoxia-inducible gene expression and hypoxia-inducible factor 1-alpha (Hif1α) protein levels. These results improved our understanding of the physiological functions of SLC38A9 and revealed its indispensable role in embryonic development, metabolic regulation, and stress adaption.

1. Introduction

The availability and quality of nutrients play a fundamental role in embryo development. Nutrients provide energy for cellular activity through catabolism and supply constituents for cellular biomass synthesis. Severe malnutrition may cause early embryonic mortality or affect human postnatal health [1]. Nutrient homeostasis is regulated by multiple mechanisms and signaling pathways [2]. For example, the mechanistic target of rapamycin complex 1 (mTORC1) dominates general protein synthesis through the phosphorylation of p70S6 kinase (S6K) and eIF4E binding protein (4EBP1). It integrates nutrient availability and growth factors to control cellular metabolism and growth. The full activation of mTORC1 requires its lysosomal localization. Amino acids, building blocks for protein synthesis, are one of the essential macronutrients for embryos’ development. The presence of nutrients, especially amino acids, induces mTORC1 to translocate to lysosomal membranes for activation [3]. How amino acids are sensed remained elusive until a series of nutrient sensors was recently discovered. Sestrin2, previously found as a stress-induced protein, binds intracellular leucine to transduce nutrient availability information [4,5]. Secretion associated Ras related GTPase 1B (SAR1B) is another novel leucine sensor but with a much higher affinity to leucine [6]. Cytosolic arginine sensor for mTORC1 subunit 1 (CASTOR1) is an intracellular arginine sensor that regulates mTORC1 signaling through a mechanism similar to sestrin2 and SAR1B [7]. In contrast to the above cytosolic amino acid sensors, SLC38A9, however, responds to lysosomal arginine levels specifically [8,9].

SLC38A9 was first identified as a potential amino acid transporter belonging to the slc38 family. It has a broad expression profile but a relatively low expression in rat tissues when compared with other slc38 proteins [10]. In contrast to SLC38A1-5, which localizes to the plasma membrane [11], and SLC38A10, which localizes to ER and Golgi [12], SLC38A9, as well as SLC38A7, are lysosomal membrane proteins [8,13]. SLC38A9 is a highly glycosylated transmembrane protein and consists of eleven transmembrane helices with a 120-residue N-terminal region towards the cytoplasm. Amino acid transport assay using proteoliposomes indicated that SLC38A9 is a bidirectional amino acid transporter [8]. SLC38A9 and Ragulator function together as noncanonical guanine exchange factors (GEF) to promote the activation of Rag GTPase heterodimers [14] and regulate mTORC1 signaling. In addition, SLC38A9 interacts with cholesterol through its conserved cholesterol-responsive motifs and mediates cholesterol-induced mTORC1 activation, independent of arginine sensing [15]. SLC38A9 also contributes to the sensing of lysine by the mTORC1 pathway [8].

The property of being able to sense a wide range of nutrients implies a potential role for SLC38A9 in maintaining cellular homeostasis. SLC38A9 was downregulated in drug-resistant ovarian cancer cell lines [16] and was a potential clinical target for mTORC1 deregulation-associated disease [8]. A recent study found that SLC38A9 also mediated endomembrane damage-promoted autophagy to maintain endomembrane homeostasis through the GATOR system [17]. As a newly discovered protein that orchestrates nutrient availability, cytoplasmic endomembrane homeostasis, and mTORC1 activity, the in vivo physiological role of SLC38A9 is rarely reported. To explore the intrinsic function of SLC38A9 during embryonic development, we took advantage of zebrafish embryos, which are feasible to image and manipulate due to their external development and transparency. The constant steady supply of nutrients from yolk cells makes the study of gene function free from external nutritional fluctuations. Furthermore, the crystal structure of zebrafish SLC38A9 is already solved [18]. In this study, we revealed the expression profile of slc38a9 and generated slc38a9^−/− mutant zebrafish using CRISPR/Cas9. slc38a9 deficiency caused pericardial edema and premature death. Transcriptional profiling revealed that SLC38A9 plays a critical role in amino acid and glucose metabolism. In addition, loss of slc38a9 resulted in a significant reduction in hypoxia-inducible gene expression and Hif1α protein levels. It improved our understanding of the stress response function of SLC38A9.

2. Results

2.1. Spatiotemporal Expression Pattern of slc38a9 in Zebrafish

To evaluate the function of SLC38A9 in early development, we first examined the spatiotemporal expression pattern of slc38a9 using qRT-PCR and whole-mount in situ hybridization. slc38a9 mRNA was detectable in the first 4 h post-fertilization (hpf), indicating that slc38a9 is a maternal mRNA. Zygotically expressed slc38a9 reached its peak at 24 hpf and remained highly expressed after that (Figure 1A). Spatially, the expression of slc38a9 was diffused across the whole embryo before 18 hpf. Interestingly, we found slc38a9 mRNA was then highly expressed in tissues including the optic primordium, midbrain, hindbrain, and heart at 48 hpf (Figure 1B–H). Specific expression in the heart, optic primordium, brain, and liver were observed at 48–72 hpf (Figure 1I,J). Strong expression in the gut was also detected at 4–5 dpf (Figure 1L,M).

2.2. Knockout slc38a9 Causes Developmental Defects and Premature Death in Zebrafish

The specific expression pattern of slc38a9 implies that it may play an indispensable role during development. We then generated the slc38a9 mutant using CRISPR/Cas9 (Figure 2A). Two independent slc38a9^−/− alleles were obtained. The slc38a9 Δ5 and Δ2 alleles harbor five and two base pair (bp) deletions, respectively (Figure 2B). Both mutants introduce premature stop codons at the N-terminal domain and lack all transmembrane domains (Figure 2C). The qRT-PCR assay indicated clear RNA decay (Figure 2D). No morphological difference was observed at 12 and 24 hpf between wide type and slc38a9 Δ5 alleles. However, 24 out of 98 mutant zebrafish exhibited pericardial edema and small eyes at 48 hpf; the abnormal phenotypes became severe at 72 hpf (Figure 2E). Similar results were also observed in the slc38a9 Δ2 allele (Figure S1). The slc38a9 Δ5 line was referred to as slc38a9^−/− mutant fish hereafter, and was used in subsequent studies unless otherwise stated. The overall body length of slc38a9^−/− mutant was decreased at 24 and 72 hpf (Figure 2F,G). While slc38a9^−/− mutant larvae with development defects died at 4–7 dpf, those normally developed slc38a9^−/− mutants (defined as slc38a9^−/− viable) have the same viability as wide type and exhibit normal fertility (Figure 2H).

2.3. Transcriptional Profiling of slc38a9^−/− Mutant

To identify affected networks that cause developmental defects in slc38a9^−/− mutants, we particularly sampled slc38a9^−/− mutants with abnormal phenotypes at 60 hpf for high-throughput RNA-seq. A number of 41.14–47.91 million (M) raw reads of each sample were collected (Table S1). An amount of 92.72–93.49% clean reads were mapped to the zebrafish genome (GRCz11) using HISAT2 with a unique map between 78.21 and 79.1%. Finally, 22, 859 protein codon genes were identified and used in the following analysis. Principal component analysis (PCA) showed that the first (PC1) and second (PC2) principal components explained 53.26 and 9.9% of the variance, respectively (Figure 3A). By comparing two genotypes, a total of 2338 differentially expressed genes (DEG) were identified with 1427 being upregulated and 911 being downregulated in slc38a9 mutants (Figure 3B). The DEGs were assigned to biological process (BP), cellular component (CC), and molecular function (MF) gene ontology (GO) categories. Totals of 289 BP, 25 CC, and 77 MF terms were significantly affected (Figure 3C, left panel). The top 10 affected GO terms are shown in Figure 3C. DEGs annotated as GO terms related to “cellular amino acid metabolic process”, “carboxylic acid metabolic process”, “organic acid metabolic process”, “oxoacid metabolic process”, and “amino acid activation” were the most significantly enriched in biological processes (Figure 3C). We found “amino acid activation” to be the most specific biological process GO term by DAG analysis (Figure 3D). DEGs annotated as GO terms related to “Z disc”, “I band”, “apical plasma membrane”, “sarcomere”, and “apical part of cell” were the most significantly enriched in the biological process (Figure 3C). These GO terms were highly related to muscle contraction. We then asked if the cardiac rhythm was affected. The results showed that knockout of slc38a9 significantly reduced heart rate (Figure 3E). DEGs annotated as GO terms related to “aminoacyl-tRNA ligase activity”, “ligase activity, forming carbon-oxygen bonds”, “catalytic activity, acting on a tRNA”, “ligase activity”, and “tetrapyrrole binding” were the most significantly enriched in biological process (Figure 3C). The linear regression analysis between RNA-seq and qRT-PCR was performed to validate the RNA-seq data (r = 0.9899, p < 0.0001) (Figure 3F).

2.4. Slc38a9 Deficiency Induces Apoptosis and Alters Multiple Metabolic Pathways

To further reveal the mechanisms by which slc38a9 deficiency leads to embryonic lethality, the annotated DEGs were applied to Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. The results showed that although 149 KEGG pathways were enriched using DEGs, only 9 KEGG pathways were significantly affected when using padj <0.05 as a threshold (Table S2). These KEGG pathways were categorized into five classes, including “Genetic Information Processing” (dre00970: Aminoacyl-tRNA biosynthesis), “Cellular Processes” (dre04115: p53 signaling pathway, dre04210: Apoptosis), “Metabolism” (dre01230: Biosynthesis of amino acids, dre00010: Glycolysis / Gluconeogenesis, dre00670: One carbon pool by folate, dre00260: Glycine, serine, and threonine metabolism), “Organismal Systems” (dre04744: Phototransduction), and “Organismal Systems” (dre04920: Adipocytokine signaling pathway) (Figure 4A). In the apoptosis-related DEGs, both caspase8 (casp8) and caspase9 (casp9) were upregulated (Figure 4B). Unexpectedly, BCL-2 family gene bcl2b was upregulated, whereas bax was downregulated. The expression of pmaip1 (orthologous to human PMAIP1/NOXA), another pro-apoptotic BCL-2 family gene, was upregulated. The elevated levels of eif2ak3, eif2s1b, and ddit3 indicated that ER stress occurred. Death receptor tnfrsf1a and its downstream effectors were also upregulated, such as jun and fosaa. In addition, a decrease in the pro-survival gene pik3r3b was observed (Figure 3B). The expression of multiple p53 target genes and p53 itself was enhanced, whereas the cell cycle regulator ccne1, a regulatory subunit of CDK2, was downregulated (Figure 3C). AO staining results showed more apoptosis cells in slc38a9^−/− mutants (Figure 3D). Apoptosis was further confirmed and quantified by a flow cytometer. The proportion of apoptosis cells was increased from 1.53 to 3.18% in slc38a9^−/− mutants (Figure 3E,F).

2.5. Biosynthesis of Amino Acids and Aminoacyl-tRNA Were Enhanced in slc38a9^−/− Embryos

SLC38A9 was originally identified as an amino acid transporter; its deficiency may lead to dysregulation of amino acid metabolism. KEGG analysis revealed that amino acid biosynthesis pathways were significantly affected in slc38a9^−/− mutant (Figure 5A). Furthermore, enhanced aminoacyl-tRNA synthetases were also observed (Figure 5B). Branched-chain amino-acid transaminase (bcat1), which facilitates the reversible transamination of BCAAs, was upregulated. Accordingly, Ile, Leu, and Val cellular levels were significantly increased (Figure 5C). Both mtr (5-methyltetrahydrofolate-homocysteine methyltransferase) and its product Met were upregulated. However, the expression of methionine adenosyltransferase mat1a and mat2ab were also increased, indicating a high turnover rate of Met. Serine hydroxymethyltransferase shmt1 and shmt2, which regulate the serine–glycine conversion, were upregulated. Cth catalyzes the last step in the transsulfuration pathway from methionine to cysteine, and Asns catalyzes the synthesis of the non-essential amino acid Asn from aspartate. Asp and Gln are two critical downstream targets of amino acid response (AAR) signaling pathway. We found the expression of both cth and asns was enhanced. Arginase-1 (Arg1) and Arginase-2 (Arg2) convert Arg to urea and ornithine. Our result showed while arg1 is downregulated, arg2 and free Arg levels were increased. Among the 17 free amino acids detected, 14 amino acids were upregulated in slc38a9^−/− mutants, whereas the level of Glu, Asp, and Pro were identical to wild-type. Since the total free amino acids level were modified, we then asked if amino acid catabolism were affected by investigating the expression of major aminotransferases and glutamate dehydrogenases. As shown in Figure 5D, glutamic–oxaloacetic transaminase 2b (got2b) was downregulated, whereas alanine aminotransferase 2 (gpt2) was upregulated. These results indicate that knockout of slc38a9 leads to amino acid metabolism disorders.

2.6. Slc38a9 Regulates Glycolysis and Gluconeogenesis in Zebrafish Embryo

Twenty-two DEGs related to glycolysis and gluconeogenesis metabolism pathways, with six DEGs being downregulated and sixteen DEGS being upregulated (Figure 6A). The hexokinases gck and hk2 were oppositely regulated in slc38a9^−/− mutants (Figure 6A). Whole-body glucose levels, detected using LC/MS in slc38a9 mutants with severe phenotype, was dramatically lower than in wild type; however, the viable slc38a9^−/− mutant had a higher glucose level than abnormal slc38a9^−/− mutants. Similar results were also observed in glucose-6-phosphate levels. Fructose-bisphosphate aldolase B (aldob), glyceraldehyde-3-phosphate dehydrogenase (gapdh), enolase 1a (eno1a), and enolase 3 (eno3) are enzymes that can catalyze both glycolysis and gluconeogenesis. Their expression in slc38a9^−/− mutants was significantly increased (Figure 6A). Glucose-6-phosphatase catalytic subunits (g6pca.1, g6pca.2, g6pcb), fructose-1,6-bisphosphatase (fbp1b, fbp2), and phosphoenolpyruvate carboxykinase (pck1) are rate-limiting enzymes that are involved in gluconeogenesis. Most of these genes were upregulated except for fbp2, which was downregulated (Figure 6A). Pyruvate kinase (pklr) is the enzyme involved in the last step of glycolysis to produce pyruvate, and its expression was upregulated. Nevertheless, pyruvate content was reduced in slc38a9^−/− mutants. However, this reduction was ameliorated in viable slc38a9^−/− mutants (Figure 6D). In addition, the level of lactate, the final product of anaerobic glycolysis, was significantly elevated in abnormal slc38a9^−/− mutants but not in viable slc38a9^−/− mutants (Figure 6E).

2.7. Slc38a9 Deficiency Leads to Dysregulated Hypoxia Response

Increased lactate is common during stress conditions [19]. RNA-seq results showed a significant increase in the expression of vegfaa and hmox1a, two stress-induced genes (Figure 7A). We then asked if slc38a9^−/− mutant is less tolerant to environmental stresses such as hypoxia, which is a typical stress often faced by aquatic animals. Treating embryos with hypoxic conditions could further increase the phenotypic penetrance of slc38a9^−/− mutation (Figure 7B). Organisms respond to oxygen deficiency by initiating the expression of a series of genes through the key transcription factors Hif (hypoxia-inducible factors). By detecting the expression of Hif target genes, we found both redd1 and igfbp1a transcriptional levels were elevated under normoxic conditions in slc38a9^−/− mutant. While hypoxia significantly increased redd1 and igfbp1a expression in wild type, this hypoxia-induced increase was completely blocked by slc38a9^−/− knockout (Figure 7C). The expression of gadd34 was also reduced, although to a lesser extent (Figure 7C). We then measured the protein levels of Hif1α under different conditions. By immunoprecipitation, we were able to enrich and detect Hif1α protein. We found that the basal protein level of Hif1α in slc38a9^−/− mutant is lower than in wild type. While hypoxia increased Hif1α protein level in wild type embryos, this increase was attenuated in slc38a9^−/− mutants (Figure 7D). These results indicated that loss of slc38a9 inhibited the response to hypoxia.

3. Discussion

The specific subcellular localization and structural characteristics make SLC38A9 a unique lysosomal arginine sensor. Accumulated evidence indicates that SLC38A9 plays important roles in amino acid efflux from lysosomes, mTORC1 activation, and tumor growth using cultured cells and the orthotopic allograft model [8,9,20,21]. To further reveal the physiological function of SLC38A9, the slc38a9 mutant zebrafish was generated in the present study. We found slc38a9^−/− mutant embryos presented with small eye size, pericardial edema, and metabolic disorder. The pericardial edema became much more severe at 4–7 dpf. Moreover, swim bladder inflation was disrupted at 5 dpf in slc38a9^−/− mutants (data not shown), indicating that the swimming activity may be reduced. Knockdown of lat3, a Na⁺-independent neutral l-amino acid transporter, also resulted in pericardial edema by 96 hpf in zebrafish [22]. Pericardial edema was generally associated with heart failure [23,24]. We indeed found reduced heartbeat rate in slc38a9^−/− mutants, and GO enrichment indicated affected heme binding and iron ion binding, hinting that SLC38A9 might be involved in normal heart development. However, pericardial edema was considered as a non-specific phenotype due to its high incidence in genetic mutants and pharmacological studies [23]. The known function of SLC38A9 is regulating mTOR signaling and transporting amino acids. The I68A variant of SLC38A9 mutants cannot bind Rag–Ragulator to control the mTOR signaling pathway but can transport arginine similarly to wild-type SLC38A9 [8], whereas the T133W mutant loses the ability to interact and transport amino acids [21]. Studying both mutants in slc38a9-null fish needs to be done to further reveal the distinct function of SLC38A9 in vivo.

KEGG analysis revealed that the expression of multiple genes which are involved in apoptosis and p53 signaling pathways were affected in slc38a9^−/− mutants. Paradoxically, the anti-apoptotic BCL-2 family gene bcl2b was upregulated, whereas the pro-apoptotic gene bax was downregulated. One possible explanation is that stresses induced by gene mutation imposed selective pressure for elevated levels of pro-survival proteins, such as BCL-2. Cells with elevated BCL-2 can be more susceptible to apoptosis than normal cells [25]. The lack of Sesn2, an intracellular leucine sensor, in mice also results in apoptosis in dendritic cells, although Sesn2-null mice are fully viable [4]. Apoptosis is readily induced by inhibition of mTOR [26], which is intrinsically downstream of amino acid sensors, including SLC38A9. Unexpectedly, the mTORC1 signaling was not enriched by KEGG analysis. mTOR is essential for embryonic development. mTOR-null is lethal in mice, and their development is arrested at E5.5 [27]. Whether lack of slc38a9 in zebrafish causes dysregulation of mTOR signaling need to be further tested.

A study in HEK293 cells found that loss of SLC38A9 had no effect on the whole-cell amino acid level but enormously boosted the lysosomal concentrations of several amino acids, including most non-polar, essential amino acids [21]. Using an in vivo model, we were able to quantify the effect of SLC38A9 on whole-body free amino acid homeostasis. According to our results, while most amino acid levels were increased, Glu and Asp levels weren’t changed in the slc38a9 mutant. This might be due to the reduced expression of got, which catalyzes the convention of Glu and Asp, and enhanced expression of gpt, which facilitates the Glu to Ala conversion. Fish uptake and degrade yolk proteins or free amino acids to meet their amino acid requirements during development [28]. Autophagy plays an essential role in yolk nutrition absorption in zebrafish [29]. Although overexpression of SLC38A9 inhibits autophagy in HEK293T, the SLC38A9-null cells did not have a defect in autophagy activation [9,21]. slc38a9 deficiency elevating free amino acid levels in zebrafish may also be due to reducing protein synthesis caused by impaired mTOR activation [9,21]. Thus, whether the elevated whole-body free amino acid levels in slc38a9 mutant fish are caused by increased absorption and/or reduced utilization need to be further determined.

Another finding in the present study is only a quarter of the offspring of slc38a9 mutants showed abnormal phenotype. The penetrance of the mutant is modulated by inherited factors and, in some cases, environmental factors [30]. Reduced penetrance was found in human hereditary diseases and zebrafish mutants [31,32]. Loss of igfbp5a, for example, induces premature death of zebrafish, whereas this mutant doesn’t show overt phenotypes when raised in calcium-rich solutions [32]. We found acute hypoxia treatment increased the proportion of abnormal phenotypes to 100% and further aggravated the abnormal phenotype, implying that SLC38A9 plays a critical role in acute hypoxia tolerance. Interestingly, the expression of redd1 and igfbp1a were increased under normoxic culture conditions. Given the fact that different stresses can induce both genes’ expression, we proposed that the slc38a9^−/− mutant was facing chronic stress. Redd1 can be induced by DNA damage, excess ROS, and ER stress [32]. In addition, Redd1 can be induced by energetic stresses, including glucose deprivation [33]. Our data indeed showed that the glucose levels in slc38a9^−/− mutant decreased, which might be due to elevated glycolysis.

The expression of glycolysis and gluconeogenesis-related genes are differently regulated in slc38a9^−/− mutants. Interestingly, some genes belonging to the same protein family showed opposite expression patterns. For example, while the hexokinase gene hk2 mRNA level was increased, the glucokinase (gck) mRNA level was reduced. Different hexokinases are expressed in a tissue-specific manner, and only HK2 is overexpressed in cancer cells, which might contribute to the high glycolytic rate in tumors [34]. We also found the expression of fbp1b was increased, and that of fbp2 decreased. Mammalian FBP1 and FBP2 encode the liver and muscle isoforms of FBPase, respectively, and FBP1 is mainly expressed in gluconeogenic organs [35]. One primary source for gluconeogenesis is glycogenic amino acids. Whether elevated free amino acid levels increased in slc38a9^−/− mutant-induced gluconeogenesis is unclear.

The metabolite assay found dramatic differences between viable and lethal slc38a9^−/− mutants. The metabolite pattern of the viable mutant was more similar to that of the wild type, indicating that targeting metabolism may rescue the lethal phenotype of slc38a9^−/− mutants. In drosophila, methionine supply rescued the fecundity defect in aci-reductone dioxygenase 1 mutants [36]. Dietary supplementation of unsaturated fatty acids (UFAs) could rescue the lethal phenotype of desat1 mutants, a gene involved in drosophila UFA metabolism [37]. These results indicate that targeting metabolism is an efficient way to ameliorate phenotype caused by a metabolism gene mutant.

In summary, we revealed the spatiotemporal expression pattern of slc38a9 in zebrafish and generated slc38a9^−/− mutant fish. We found that slc38a9 deficiency results in severe pericardial edema and premature death. By performing whole-organism RNA-seq of wild type and slc38a9^−/− mutant fish, we identified increased apoptosis, dysregulated amino acid metabolism, and glycolysis/gluconeogenesis disorder in slc38a9^−/− mutant fish. We experimentally confirmed that free amino acids and lactate were accumulated in zebrafish. Moreover, elevated expression of stress-related genes and attenuated hypoxia tolerance were observed in slc38a9^−/− mutant fish. The results of this study have revealed the indispensable function of slc38a9 on embryonic development. Nevertheless, whether the phenotype observed in slc38a9^−/− mutant fish was caused by the direct or indirect effects of slc38a9 deficiency is unclear. Future studies are needed to determine the mechanism by which SLC38A9 regulates embryonic development.

4. Materials and Methods

4.1. Experimental Animals

Zebrafish (Danio rerio) were maintained in a circulation water system on a 14:10 h darkness cycle at 28 °C and fed twice daily. Fertilized eggs were obtained by natural breeding. They were raised at 28.5 °C in an embryo rearing solution (ERS, 5 mmol/L NaCl, 0.33 mmol/L MgSO₄, 0.33 mmol/L CaCl₂, 0.17 mmol/L KCl, pH 7.2) and staged according to the standard method [38]. All experiments were conducted following the guidelines approved by the Animal Care Committee of Ocean University of China (Permit Number: 11001).

4.2. slc38a9 Knockout by CRISPR/Cas9 System

The zebrafish slc38a9 mutants were generated by the CRISPR/Cas 9 system. The dual NLS-tagged Cas9 mRNA was synthesized in vitro using the linearized pSP6-2sNLS-spCas9 as templates with SP6/T7 mMESSAGE mMACHINE Kit (Ambion, Carlsbad, CA, USA) according to the manufacturer’s instruction. For making slc38a9 sgRNA (5′-AGGACAGTAAACCGTTACTG-3′), the transcription template DNA was amplified from the plasmid pT7-gRNA, a plasmid containing the T7 promoter, gRNA sequence, and the tracrRNA sequences. The primers used for PCR amplified are listed in Table S3. To obtain injected F0 embryos, we collected the wild-type embryos and microinjected them with 1nl of Cas9 mRNA (500 ng/μL) and slc38a9 sgRNA (100 ng/μL) at one-cell stage embryos. PCR was performed for genotyping using the primers in Table S3, and the products were analyzed by DNA sequencing. Finally, two independent slc38a9 mutant lines were obtained. Most of the assays were performed with the zebrafish of the slc38a9Δ5 unless specifically indicated.

4.3. Hypoxia Treatment

We pumped N₂ gas into ERS until the dissolved oxygen (DO) = 1 mg/L to generate hypoxic ERS (DO = 1 mg/L). The dissolved oxygen of ERS was detected by the DO sensor (Xylem, Munich, Germany). Then, we transferred fifteen 24 hpf embryos and 2 mL hypoxic ERS into a 35 mm culture dish, which was placed into a modular hypoxic chamber (MIC-101, Billups-Rothenberg, San Diego, CA, USA). The chamber was sealed and placed into a thermo-static incubator at 28.5 °C.

4.4. Quantitative RT-PCR (qRT-PCR)

Total RNA was isolated from fifteen embryos using TRIzol reagent and reverse-transcribed to cDNA using M-MLV reverse transcriptase (Takara, Shiga, Japan). qRT-PCR was carried out using TB Green Premix Taq (Takara, Shiga, Japan) with a real-time PCR detection system (Bio-Rad, Hercules, CA, USA). All primer sequences of target genes are listed in the Table S3. The expression level of a particular gene transcript was calculated based on the standard curve and normalized by the β-actin levels, as no expression changes of β-actin were observed in zebrafish among different treatments. The gene expression levels were calculated by a 2^−ΔΔCT method. The data were reported as fold increases of the control.

4.5. Whole-Mount In Situ Hybridization

In situ hybridization was performed as previously reported [39]. Briefly, slc38a9 mRNA was detected using a digoxigenin (DIG)-labeled antisense riboprobe. After incubation in a byan alkaline phosphatase-conjugated anti-DIG antibody (Roche, Switzerland), larvae were stained in nitroblue tetrazolium (NBT) /5-Bromo-4-chloro-3-indolyl phosphate (BCIP) (Roche, Basel, Switzerland) substrate. After staining, they were washed in 1 × PBST. Images were acquired using Nikon stereomicroscopes (Nikon Inc., Melville, NY, USA).

4.6. High-Throughput RNA Sequencing

Total RNA was isolated from pools of 10 larvae (60 hpf) by standard extraction methods. The quality of each sample was determined by an Agilent2100 bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA) and agarose gel electrophoresis. The RNA quality score (RIN/RQN) was 10 for all of the samples.

The mRNA with polyA tail was enriched by Oligo (dT) magnetic beads, and cDNA libraries were constructed using the NEBNext^® Ultra™ RNA Library Prep Kit for Illumina^® (NEB, Beverly, MA, USA). The sequencing of these cDNA libraries was performed on the Novogene HiSeq4000 sequencing platform (Novogene, Beijing, China) according to the manufacturer’s stand protocol.

4.7. Bioinformatics Analysis of RNA Sequence Data

Raw reads were tested for quality control and filtered into clean reads. The clean reads were aligned with the zebrafish genome Danio rerio, GRCz11) using HISAT (v2.0.4) to obtain the localization information of reads on the reference genome. The gene expression was quantified using the FPKM method (fragment per kilobase of transcript per million fragments sequenced) [40]. Differentially expressed genes were identified by DESeq2 software [41]. Genes with fold change >2 and adjusted p-value ≤ 0.05 are statistically significant. The GO (gene ontology) functional enrichment analysis and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis were performed using clusterProfile software. The threshold of significant enrichment was padj < 0.05.

4.8. AO Staining

Zebrafish were incubated in a culture medium containing 2 µg/mL acridine orange (AO) (Sigma-Aldrich, St. Louis, MO, USA) for 30 min. Before observation, embryos were washed in fresh ERS three times and anesthetized by 0.08% MS-222. Images were captured using a microscope equipped with a Nikon DS-U3 camera (Nikon Inc., Melville, NY, USA). NIS-Elements software (Nikon Inc., Melville, NY, USA) was used.

4.9. Apoptosis Analysis by Annexin V-FITC/PI Staining

Quantitative analysis of apoptosis in the single cells of zebrafish embryos was done using the Annexin V-FITC/PI kit (Beyotime, Shanghai, China). Using recombinant Annexin V conjugated to green-fluorescent FITC dye, the kit detects the internalization of phosphatidyl-serine in the apoptotic cells, and dead cells are detected using propidium iodide (PI) that binds to exposed DNA. Briefly, zebrafish larvae (3 dpf, n = 30/group) were digested with Liberase TM (Roche, Basel, Switzerland) for 30 min at 28.5 °C into single-cell isolation. Then, the cell supernatant was filtered through 40 μm mesh and stained using FITC-Annexin V/PI and analyzed in a flow cytometer. All the experiments were performed in triplicates. Data were analyzed using FlowJo (v7.6) software (Tree Star Inc. Ashland, OR, USA).

4.10. Free Amino Acid Analysis

An L-8900 automated amino acid analyzer (Hitachi, Tokyo, Japan) analyzed zebrafish’s free amino acid concentration with a high-performance lithium column. Seventy embryos of 60 hpf for each group were homogenized in 200 μL 10% sulfosalicylic acid solution. After centrifugation at 13,000 rpm for 15 min, the supernatant was filtered through 0.22 μm filters for free amino acid measurement.

4.11. Mass Spectrometry

Sixty zebrafish embryos were transferred to a 2 mL quenching solution (40:40:20 Methanol:Acetonitrile:H₂O, 0.1 M Formic acid, 15% NH₄HCO₃) homogenized at −20 °C. After centrifugation at 13,000 rpm for 15 min, the supernatants were analyzed using a QTRAP5500 LC-MS/MS mass spectrometer (AB Sciex, Framingham, MA, USA) in positive/negative mode and with multiple reaction monitoring (MRM) scan type. An amount of 5 μL of metabolite extracts were injected on an amide column (100 × 2.1 mm, 1.7 μm; Waters, Milford, MA, Ireland) using a mobile phase A (water, 5% acetonitrile, 20 mM ammonium acetate, 20 mM NH₄OH, pH 9.4) and mobile phase B (100% acetonitrile) at a constant flow rate of 300 μL/min; initial conditions: 95%B, 1–12 min: B goes from 95% linearly to 55%, 12–13 min: B goes from 55% linearly to 40%, 13–15 min: 40%B, 15–15.1 min: B goes from 40% linearly to 95%, 15.1–18 min: 95%B. The raw data were processed and analyzed using SCIX OS (AB Sciex, Framingham, MA, USA) and Markerview1.3.1 software (AB Sciex, Framingham, MA, USA).

4.12. Western Blot and Immunoprecipitation

Sixty zebrafish embryos were sonicated in Precooled Lysis Buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% NP-40, 1 mM EGTA, five tablets of phosphatase inhibitor, and two tablets of EDTA-free protease inhibitor (Roche, Basel, Switzerland) per 50 mL)). Lysates were cleared by centrifugation in a microcentrifuge (13,000 rpm, 10 min, 4 °C). Proteins were quantified with BCA (Beyotime, Shanghai, China). For immunoprecipitations, lysates were incubated with HIF1α primary antibody (#NB100-479, Novus Biologicals, Littleton, CO, USA) and protein A/G-sepharose (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) (overnight with rotation, 4 °C). Beads were recovered and washed four times with PBS before analysis by SDS–PAGE and immunoblotting.

4.13. Statistical Analysis

Values are presented as mean ± standard error of the mean (SEM). Statistical significance among experimental groups was determined using one-way ANOVA followed by Tukey’s multiple comparison test. Statistical tests were performed using GraphPad Prism 8 software (GraphPad Software, San Diego, CA, USA), and significance was defined as p < 0.05 or greater. Correlation analysis was carried out by fitting the data to the equation Y = X to calculate r² using GraphPad Prism 8 (GraphPad Software, San Diego, CA, USA).