Hepatic steatosis commonly occurs in intensively farmed tilapia. This disease is harmful to fish growth and health, but knowledge of the metabolic changes in tilapia with fatty liver is limited. In the present study, we compared genetically improved farmed tilapia (GIFT, Oreochromis niloticus) fed a high-fat diet (HFD) with those fed a normal-fat diet (NFD) for 8 weeks using LC-MS-based hepatic metabolomic assays and traditional nutritional assessments. Juvenile GIFT fed a HFD displayed higher fat disposition in the liver than did those fed a NFD. The metabolomic analyses revealed 61 differentially accumulated metabolites between the groups, and these metabolites were involved in 37 signaling pathways. Our results illustrate the development of metabolic disorders related to various hepatic biological processes, including protein metabolism, lipid metabolism, carbohydrate metabolism, and nucleotide metabolism in HFD-fed GIFT. These physiological changes may be related to the lower growth rate of GIFT. Overall, our study reveals the metabolic disorders in GIFT fed HFD, and enhance our knowledge of the mechanism of fatty liver formation in GIFT.

Sample collection

    GIFT were fasted for one day after feeding trial. Before sampling, all GIFT were anesthetized with 100 mg L⁻¹ MS-222 and weighed, three fish were chosen from each container and discrete liver tissues for further LC-MS analysis (n=9). All liver samples were frozen immediately in liquid nitrogen and stored at −80°C until analysis.

 LC-MS analysis

    50mg of liver samples were homogenized in 120 μL pre-cooled 50% methanol buffer using a high-throughput tissue lyser (Scientz-192, Xinzhi, NingBo, China). The solutions were incubated on ice for 10 min and then centrifuged at 4000 g for 20 min. The supernatants were collected and used for metabolomic analyses. An ultra performance liquid chromatography system (SCIEX, UK) equipped with an ACQUITY UPLC BEH Amide column (100 mm × 2.1 mm, 1.7 µm, Waters, UK) was used for metabolite analysis. The flow rate was 0.4 ml/min and the sample injection volume was 4 µL. The elution gradient was as follows: solvent A (25 mM ammonium acetate + 25 mM NH4H2O) and solvent B (IPA:ACN=9:1 + 0.1% formic acid): hold at 95% B for 0.5 min, decrease linearly from 95% to 65% B (0.5–9.5 min), decrease linearly from 65% to 40% B (9.5–10.5 min), hold at 40% B (10.5–12min), increase from 40% to 95% B (12–12.2min) and hold at 95% B until 15 min. Mass spectrometry data acquisition was performed using a high-resolution tandem mass spectrometer TripleTOF5600plus (SCIEX, UK). The Q-TOF was operated in both positive (POS) and negative (NEG) modes using the following parameters: curtain gas was set to 30 PSI, nebulizer pressure was set to 60 PSI, and the interface heater temperature was 650°C. The detector voltage was 5 kV in positive ion mode (−4.5 kV in negative ion mode). Data were collected in the IDA mode and the mass spectra scan range was set from m/z 60 to 1200. Quality control (QC) samples were prepared by mixing the supernatants from all 18 individual liver samples and these QC samples were injected at constant intervals to ensure the stability of the LC-MS analysis.

2.5 Multivariate data analysis

    ProteoWizard was employed to convert LC−MS raw data into mzXML format. Then, XCMS software (http://xcmsonline.scripps.edu) was used for peak alignment, extraction, and quantification. The peak data were further processed by metaX toolbox implemented with R software (version 3.5.1). Detectable features were first processed to remove low-quality features (detected in less than 50% of QC samples or 80% of biological samples). The remaining features were regarded as high-quality features and were used for further analysis. PCA was used to visualize the dataset of the liver samples in the HFD and NFD groups. PLSDA was then used for cluster analysis and to build a discriminant model. Permutation tests (n = 200) were utilized to evaluate the validity of the PLSDA models. In this study, significantly different features were first screened based on VIP values generated from the PLSDA model. The VIP cut-off value was set at 1.0 and VIP values > 1.0 were selected as significant discriminatory features between the HFD and NFD groups. Then, these features were confirmed by a student’s t test. The adjusted P value was adjusted by Benjamini and Hochberg's approach. The fold change (FC) of each feature between these two groups was calculated as follows: FC = normalized mean value obtained from the HFD group/ normalized mean value obtained from the NFD group. Features with a VIP value >1, FC > 2 or < 0.5, and an adjusted P-value < 0.05, were selected as significantly different features and were used for further analyses. Significantly different features were further identified by comparing their m/z data and retention time (RT) with those in an in-house fragment spectrum library of metabolites. Some online databases, including ChEBI (http://www.ebi.ac.uk/chebi/init.do), KEGG (http://www.genome.jp/kegg/), and HMDB (http://www.hmdb/ca), were used to further identify and annotate each metabolite. To visualize the metabolic changes under a HFD, significant different metabolites were clustered by hierarchical clustering analysis using TigrMev 4.2 software. Afterwards, these metabolites were subjected to a pathway analysis against the zebra fish (Danio rerio) KEGG library in MetaboAnalyst 4.0 (http://www.metaboanalyst.ca/MetaboAnalyst/).