Development of the Pseudotargeted Metabolomics Method. The pseudotargeted metabolomics method was proposed based on LC/MS system operated in the MRM mode. Metabolomics allows the analysis of as many metabolites as possible, it is impossible to obtain a standard for each metabolite. In this work, the ~500 MRM ion pairs for pseudotargeted analysis were acquired by some metabolite standards or published literatures 16–19, 22, the ~350 MRM ion pairs were selected for the quantitative MRM transition, along with a second transition for identity confirmation. The mass spectometry parameters and retention time (RT) are summarized in supplementary Table S1. The coverage of detected metabolites using the developed pseudotargeted method is involved in amino acid metabolism, nucleotidemetabolism, glycolysis, tricarboxylic acid (TCA) cycle, etc.
A total of ~278 ion peaks were determined by the use of the above analytical method. Representative MRM chromatograms of cell samples in the positive and negative detection modes are shown in Figs. 1(A, B). The pseudotargeted method was validated for repeatability, inter- and intra-day precision. The within-day precision was evaluated on the same day by QC samples, 13.7, 35.3, 69.4 and 92.5 % of the peaks had the relative standard deviation (RSDs) below 5, 10, 15 and 25 %, respectively (Fig. 2(A)). The inter-day precision was evaluated for three consecutive days, where 15.1, 33.8, 62.2 and 90.3 % of the peaks showed the RSDs below 5, 10, 15 and 25 %, respectively (Fig. 2(B)). The QC samples were also used to evaluate the repeatability of the method, and one QC sample was inserted after every six test cells. 13.3, 36.7, 61.9 and 91.4 % of the peaks had the RSDs below 5, 10, 15 and 25 %, respectively (Fig. 2(C)). These results showed that the developed method were suitable for metabolomics analysis in complex cell matrices.
Multivariate Analysis of the Metabolite Profile. The developed method was used to investigate the differential metabolites between erastin-treated MGC-803 cells and the controls. In order to screen for the metabolic difference between the controls and treated cells at different time points (6 h, 12 h, 24 h and 48 h), multivariate analysis was applied to the corresponding spectra (n = 48). Firstly, PCA and OPLS-DA models were built to show the difference of metabolites between erastin-treated cells and the controls. The PCA-X score plots are shown in Fig. 3. As shown in Fig. 3, in the PCA-X score plot, erastin-treated cells were obviously separated from the controls at different time points. It can therefore be concluded that erastin induced a great metabolic disorder in MGC-803 cells. The OPLS-DA score plots are shown in Fig. S1.
Heatmap Visualization. To further study the relative levels of differential metabolites between erastin-treated cells and the controls, a hierarchical cluster analysis (HCA) was performed on MEV software. The metabolites screened for the heatmap analysis were based on the principles of Student’s t-test p-value <0.05 and the variable importance in projection (VIP) value >1 in OPLS-DA. The heatmap visualization of these differential metabolites showed a great difference between erastin-treated cells and the controls, which is consistent with OPLS-DA results. As shown in Fig. 4(A), 19 differential metabolites were discovered between erastin-treated cells and the controls with two metabolites increased and 17 metabolites decreased in treated cells at 6 h. In Fig. 4(B), 34 differential metabolites were discovered between erastin-treated cells and the controls with 18 metabolites increased and 16 metabolites decreased in treated cells at12 h. In Fig. 4(C), 23 differential metabolites were discovered between erastin- treated cells and the controls with one metabolite decreased and 22 metabolites increased. In Fig. 4(D), 29 differential metabolites were discovered between erastin- treated cells and the controls with two metabolites decreased and 27 metabolites increased.
Integrated Pathway Analysis. To better understand the differences of metabolic disorders between the erastin- treated cells and the controls, an integrated pathway analysis was carried out. MetaMapp is an R based script whose outputs are well compatible with the open-source platform CytoScape 23. It can be used to visualize metabolomics datasets that can conclude all identified compounds while maintaining the modular organization of metabolites in biochemical pathways 24. The metabolic relation network was generated through MetaMapp and drew by CytoScape (Fig. 5). The red lines represent the relationship between the two metabolites contained in KEGG. The blue lines represent the relationship between the two metabolites without contained in KEGG, but the related metabolites had a similar chemical structure, judged by PubChem. The diameter was corrected with the fold change value and t-test p value. The raw data is shown in Table S2.
Of the differential metabolites (p<0.05), 21% (18/87) and 43% (56/129) were up-regulated in erastin-treated cells at 6 h and 12 h, respectively. With the prolongation of drug action time, the up-regulated metabolites gradually increased. More than 90% of the differential metabolites were up-regulated in treated cells at 24 h or 48 h. As shown in Fig. 5(A), most of differential metabolites showed a decreasing trend in treated cells at 6 h. The metabolites with fold change values >3 were N6-methyladenosine, inosine, 6-O-methylguanosine, 1-methyladenosine, and carbamoyl phosphate. The metabolites with fold change values <0.3 were reduced glutathione, melatonin, dCMP, cyclic di-AMP and sorbose. As shown in Fig. 5(B), the metabolites with fold change values >3, including 5-methylthioadenosine, N6-methyladenosine, S-adenosyl-L-methionine, carbamoyl phosphate, glucose 1-phosphate, adenine, inosine, 6-O-methylguanosine, uridine and guanosine were mainly incorporated in nucleotide metabolism and glycolysis. The metabolites with fold change values < 0.3 were S-adenosyl-L-homocysteine, reduced glutathione, oxidized glutathione, dCMP, and melatonin. As shown in Figs. 5(C)~(D), most of the differential metabolites were up-regulated including amino acid metabolism, nucleotide metabolism, glycolysis and TCA cycle. Only a few metabolites with fold change values <0.3, including reduced glutathione, oxidized glutathione, S-adenosyl-L-homocysteine were mainly involved in cysteine and methionine metabolism and glutathione metabolism.