Metabolic and lipidomic investigation of the antiproliferative effects of coronatine against human melanoma cells

Melanoma is the most aggressive form of skin cancer, with metastatic melanoma being refractory to currently available conventional therapies. In this study, we evaluated the inhibitory effect of coronatine (COR) on the proliferation of metastatic melanoma cells. COR inhibited the proliferation of melanoma cells but negligibly affected the proliferation of normal melanocytes. Comparative metabolic and lipidomic profiling using gas chromatography-mass spectrometry and direct infusion-mass spectrometry was performed to investigate COR-induced metabolic changes. These analyses identified 33 metabolites and 82 lipids. Of these, the levels of lactic acid and glutamic acid, which are involved in energy metabolism, significantly decreased in COR-treated melanoma cells. Lipidomic profiling indicated that ceramide levels increased in COR-treated melanoma cells, suggesting that ceramides could function as a suppressor of cancer cell proliferation. In contrast, the levels of phosphatidylinositol (PI) species, including PI 16:0/18:0, 16:0/18:1, 18:0/18:0, and 18:0/18:1, which were found to be potential biomarkers of melanoma metastasis in our previous study, were lower in the COR-treated cells than in control cells. The findings of metabolomic and lipidomic profiling performed in the present study provide new insights on the anticancer mechanisms of COR and can be used to apply COR in cancer treatment.


MJ or CoR inhibits the proliferation of human melanoma cells.
The effect of MJ and COR on the growth of melanoma cells or normal melanocytes was investigated by assessing the viability of HEMn-LP, A375, and A2058 cells treated with different doses of MJ and COR for 24 h by performing 3-(4,5-dimethylthiazol-2-yl)−2,5-diphenyltetrazolium bromide (MTT) assay. MJ or COR treatment decreased the viability of melanoma A375 and A2058 cells in a dose-dependent but negligibly affected the viability of normal human epidermal melanocytes HEMn-LP (Fig. 1A,B). The median inhibitory concentration (IC 50 ) values of MJ and COR were 13.18 ± 1.50 and 4.60 ± 0.51 μM, respectively, for A375 cells and 37.94 ± 1.18 and 7.93 ± 1.50 μM, respectively, for A2058 ( Table 1).

The combination of MJ or COR with 2-DG exerts synergistic antiproliferative effects on human melanoma cells.
Previous studies have shown that 2-DG-induced glycolysis inhibition enhances the effect of MJ on lung, colon, and breast cancer and sarcoma cells 42,43 . Therefore, we determined the effects of the combination of MJ and 2-DG and of COR and 2-DG on melanoma cells. MJ and COR concentrations were chosen such that cell viability was maintained at >60% in accordance with dose-response data (Fig. 1C,D). The synergistic effects of the combination treatments were determined by calculating combination index (CI) values, which depict synergism quantitatively. For both A375 and A2058 cells, the CI values of the combination treatment with 0.5 mM 2-DG and MJ or COR ranged from 0.12 to 0.39 (Fig. 1E,F). The combination of 2-DG with MJ and of 2-DG with COR exerted strong synergistic effects, as indicated by CI values of <1. Because a lower concentration of COR than that of MJ exerted synergistic antiproliferative effects along with 2-DG on the melanoma cells in the present study, we selected the combination treatment with 0.01 μM COR and 0.5 mM 2-DG for performing metabolomics and lipidomics analyses. At this concentration, COR negligibly affected the viability of normal melanocytes after 96 h of exposure ( Supplementary Fig. S1). To validate the antiproliferative effect of COR on melanoma cells, we examined the changes in A375 and A2058 cell proliferation after treatment with the combination of COR and 2-DG by performing 5′-bromo-2′-deoxyuridine (BrdU) incorporation enzyme-linked immunosorbent assay (ELISA). The percentage of melanoma cells showing the incorporation of BrdU, a marker of replicative DNA synthesis in S-phase cells, significantly decreased after treatment with the combination of COR and 2-DG (Fig. 2), which was consistent with the results of the MTT assay.
Treatment with the combination of COR and 2-DG alters the relative abundance of metabolites and lipids in melanoma cells. The metabolic and lipidomic changes induced by COR alone or in combination with 2-DG were investigated using the human melanoma A375 and A2058 cells, which have different metastatic potentials, and primary HEMn-LP melanocytes by performing the GC-MS and DI-MS analyses. The www.nature.com/scientificreports www.nature.com/scientificreports/ GC-MS analysis detected the following 33 metabolites: two alcohols, fifteen amino acids, four organic acids, three purines, two pyrimidines, five sugars, and two other metabolites (Supplementary Table S1). The relative changes in metabolite levels are shown as fold changes, which are log 2 ratios of normalized intensities of the identified The bars indicate mean values, and the error bars indicate the standard deviation of triplicate experiments. The asterisk (*) denotes significant difference compared with the control group, as determined using the Student's t-test (p < 0.05). CI, combination index; COR, coronatine; 2-DG, 2-Deoxy-d-glucose; MJ, methyl jasmonate; MTT, 3-(4,5-dimethylthiazol-2-yl)−2,5-diphenyltetrazolium bromide.
www.nature.com/scientificreports www.nature.com/scientificreports/ metabolites in treated cells compared with those in control cells (Supplementary Table S2). Notably, treatment with 1.0 μM COR alone or with the combination of 0.01 μM COR and 0.5 mM 2-DG induced similar changes in the levels of organic acids; sugars; and amino acids, including alanine, aspartic acid, glutamic acid, proline, and threonine, in the two melanoma cell lines. The levels of these metabolites significantly decreased after COR treatment. Particularly, the levels of sugars showed a higher decrease in A375 cells treated with the combination of COR with 2-DG than in those treated with COR alone. The levels of metabolites involved in energy metabolism, such as glucose, glucose-6-phosphate, lactic acid, fumaric acid, and malic acid, were lower in the melanoma cells treated with COR alone and with COR + 2-DG than in control cells. However, the levels of the glycolytic metabolites, including glucose, glucose-6-phosphate, and lactic acid, were not significantly altered and the levels of fumaric acid and malic acid were increased in COR-treated normal melanocytes. Moreover, COR treatment decreased the levels of purines and pyrimidines such as hypoxanthine, inosine, guanine, uracil, and uridine in A375 cells. However, no significant differences were noted in the levels of purines between the treated and control A2058 cells. Significant changes in metabolite levels were also observed in the normal melanocytes; however, these changes were lower than those observed in the melanoma cells.
Various lipid molecules regulate cellular processes, including cell growth and proliferation, membrane homeostasis, metastasis, apoptosis, and drug response 44 . COR-induced alterations in lipid metabolism were investigated by performing lipidomic profiling with nano-electrospray ionization-mass spectrometry (nanoESI-MS). In all, 17 lipid species, including 11 phosphatidylcholine (PC) and 6 triacylglycerol (TG) species, were detected in the positive ion mode and 65 lipid species, including 9 ceramide (Cer), 15 phosphatidylethanolamine (PE), 3 phosphatidylglycerol (PG), 25 phosphatidylserine (PS), and 13 phosphatidylinositol (PI) species, were detected in the negative ion mode. Detailed information on the proposed compositions of fatty acyl chins, ion species, mass-to-charge ratio (m/z) values, MS/MS fragment ion, and fold changes (log 2 ratio) of intact lipid species in the control and treated cells is listed in Supplementary Tables S3 and S4. The total relative abundance of each lipid class in the melanocytes and melanoma cells was determined as the sum of individual lipid species of that class (Fig. 3). The total levels of Cer, PS, PC, and TG species were higher in the A375 cells in the COR and 2-DG + COR treated groups than in those in the control group. Moreover, the total Cer, PS, and PC levels increased in the the A2058 cells in the 2-DG + COR treatment group. In contrast, the total PI level significantly decreased only in the A2058 cells in both the treatment groups.
All the identified Cer species accumulated in both A375 and A2058 melanoma cells treated with COR alone and with the combination of COR and 2-DG. Particularly, Cer (d18:1/18:0) showed the highest accumulation compared with other fatty acyl chain-containing species. In contrast, the levels of all PG species (PG 16  www.nature.com/scientificreports www.nature.com/scientificreports/ with higher metastatic potential and suggested that these PI species were potential novel biomarkers of melanoma metastasis 45 . Interestingly, in the present study, treatment with COR alone decreased the levels of these PI species in the two melanoma cell lines compared with those in the control cells. Treatment with the combination of COR and 2-DG decreased the levels of PI 16:0/18:1 and PI 18:0/18:1 but did not affect the levels of PI 16:0/18:0 and PI 18:0/18:0. In normal melanocyte, COR treatment decreased the levels of most Cer species. In addition, no significant differences were observed in the levels of most PI species between the COR-treated and control cells.

Multivariate statistical analysis of the metabolic and lipidomic profiles of melanoma cells treated with the combination of COR and 2-DG.
The COR-induced changes in the metabolite and lipid profiles of the normal melanocytes and two melanoma cell lines were compared by analyzing the GC-MS and DI-MS spectral datasets through principle component analysis (PCA) (Fig. 4). In score plots, the first three principal components accounted for 94.6% total variance and reflected most of the spectral information. The two melanoma groups showed obvious separation from the normal group. Moreover, a clear separation was observed between the two melanoma groups. Interestingly, the control samples and the samples treated with COR alone or with the combination of COR and 2-DG were closer to one another in the normal group. However, the COR and COR + 2-DG treated samples were remote from the untreated control samples in the melanoma groups. These results implied that the COR and 2-DG combination treatment induced significant perturbations in the metabolite and lipid profiles of the melanoma cells.
Next, the metabolites and lipids that showed maximum changes after COR treatment were identified by obtaining S-plots by comparing untreated groups with the COR or COR + 2-DG treated groups in A375 and www.nature.com/scientificreports www.nature.com/scientificreports/ A2058 melanoma cells through orthogonal projection to latent structure-discriminant analysis (OPLS-DA). Model quality is indicated by parameters R 2 Y and Q 2 Y, which represent the goodness of fit and predictability of the model, respectively. Validity was assessed using the analysis of variance of cross-validated predictive residuals; p < 0.05 46 . The parameters for each model are listed in Supplementary Table S5. The S-plots combine a scatter plot p [1], which describes the magnitude of each variable within a model, and p(corr) [1], which represents the reliability of each variable (modeled correlation). The most significantly altered variables are plotted in the lower left or upper right of the S-plot and are highly correlated with group separation 47 . In the present study, 10 compounds at the outermost bottom and top of the S-plots were selected as potentially relevant metabolites altered by COR treatment based on a p(corr) cutoff value of |0.8|. Candidate predictive markers of response to COR treatment were selected from these compounds based on the principle VIP value of >1.0 and on compounds whose levels commonly increased or decreased in more than two groups treated with COR (Supplementary Table S6). These compounds are highlighted using red filled circles in . Among these selected candidate markers, glutamic acid, lactic acid, Cer d18:1/18:0, Cer d18:1/16:0, and Cer d18:0/18:0 showed the same variation tendency in all the COR treated groups. Therefore, these compounds were indicated as potential biomarkers for evaluating the therapeutic effect of COR on melanoma cells. The main altered metabolic and lipidomic pathways in the COR treated melanoma cells are shown in Figs 6 and 7. These pathways included glycerophospholipid and sphingolipid metabolism, and energy metabolism such as glycolysis, TCA cycle, and glutaminolysis.

Discussion
The metabolic status of a cell is closely associated with its growth, survival, and proliferation. In this study, we investigated the effect of COR and 2-DG on the normal melanocytes and two metastatic melanoma cell lines by performing comprehensive metabolic and lipidomic profiling and identified major metabolic alterations induced by COR treatment. First, we found that COR was more effective than MJ, a structural COR analog that exerts anticancer effects on metastatic melanoma cells, in inhibiting the growth of melanoma cells and exerted negligible effects on the growth of normal cells 15 . Treatment with the combination of COR and 2-DG significantly decreased the levels glycolysis and TCA cycle metabolites in the melanoma cells. Most cancer cells highly depend on glycolysis to meet their increased energy and biosynthesis demands for proliferation and differentiation. Therefore, COR treatment-induced decrease in glycolysis in melanoma cells may be effective for inhibiting the growth of and inducing cell death in cancer cells 35,[48][49][50] . Inhibition of glucose metabolism markedly decreases ATP levels in cancer cells and sensitizes myeloid leukemia and cervical, breast, and prostate cancer cells to death receptor-induced www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ apoptosis 51,52 . In the present study, the COR and 2-DG combination treatment exerted an enhanced inhibitory effect on the growth of melanoma cells. 2-DG reduces glycolytic enzyme activity and decreases intracellular ATP level [53][54][55] . A previous study showed that treatment with the combination of 2-DG with cisplatin and staurosporine, which are proapoptotic drugs, increased cell death in human metastatic melanoma cells by inducing apoptosis 41 .
In the present study, decrease in lactic and glutamic acid levels was the most relevant COR-induced metabolic alteration in the metastatic melanoma cells. Because lactic acid accumulation through glycolysis induces acidosis in cells, tumor cells promote lactic acid efflux through monocarboxylate transporters to maintain pH homeostasis 56,57 . This reduces extracellular pH and enhances the migration and invasion of human melanoma cells [58][59][60] . Therefore, the COR-induced decrease in lactic acid levels observed in the present study suggests the inhibition of glucose metabolism in and suppression of the invasion and metastasis of melanoma cells. The initial step of glutaminolysis, a major metabolic pathway contributing to tumor growth, is the conversion of glutamine to glutamic acid, which is an important source for biosynthesis pathways and energy in proliferating cancer cells 61,62 . Glutamic acid, a precursor of α-ketoglutarate, serves as a carbon source to replenish the TCA cycle and is degraded to citrate, pyruvate, aspartic acid, alanine, serine, and proline to provide metabolic intermediates for cell maintenance 63,64 . Thus, the decreased glutamic acid level in the COR-treated cells relative to that in the control cells may reflect metabolic changes that suppress the growth of melanoma cells. Moreover, the decreased alanine and aspartate levels in the COR-treated cells may be closely associated with disturbances in alanine, aspartate, and glutamate metabolism, which is the most relevant pathway in melanoma development and had the highest impact score in pathway analysis performed in our previous study 45 .
The melanoma cells treated with COR showed significant accumulation of various Cer species, particularly Cer with an 18-carbon chain, such as Cer d18:1/18:1, Cer d18:1/18:0, and Cer d18:0/18:0. Cer, a powerful tumor suppressor, limits cancer cell proliferation by inducing apoptosis, cell cycle arrest, and autophagic cell death. Moreover, Cer forms Cer-enriched membrane platforms that cluster death receptors such as TNF-related apoptosis-inducing ligand receptor and CD95 to activate apoptotic signaling pathways 65,66 . Thus, the role of Cer in cell death suggests that Cer metabolism is an attractive therapeutic target. Several studies have examined Cer-based therapies involving agents that increase Cer level for cancer treatment. For example, taxol induces cellular Cer generation in breast cancer cells (MCF-7 and MDA-MB-468), thus increasing their apoptosis 67 . In addition, treatment with 1-phenyl-2-decanoylamino-3-morpholino-1-propanol, a Cer glucosylation inhibitor, results in the concomitant accumulation of Cer and enhances curcumin-induced apoptosis of B16 and WM-115 www.nature.com/scientificreports www.nature.com/scientificreports/ melanoma cells 68 . Exogenously added cell-permeable short-chain Cer (C6) exerts cytotoxic effects on human melanoma cell lines 69,70 . Therefore, elevation of intracellular Cer species in the COR-treated melanoma cells may be important for suppressing the proliferation of these cells.
In the present study, COR treatment significantly altered the levels of glycerophospholipids, which are major lipid components of biological membranes, and inhibited the growth of melanoma cells. The COR-treated cells showed increased levels of TG species, which was consistent with that reported previously in various cell types in response to apoptotic stimuli. Early TG accumulation after Fas mAb treatment is associated with apoptosis onset in Jurkat T-cells. Shotgun lipidomics profiling showed that TG species were accumulated in apoptotic keratinocytes treated with narrow-band UVB irradiation 71,72 . Although the exact role of TG accumulation is unknown at present, an increase in fatty acid level after TG accumulation is suggested to contribute to Cer generation, which eventually induces cell death 73 . The levels of most PS species increased in the melanoma cells treated with the combination of COR and 2-DG. This increase in the levels of PS species may be associated with altered PS metabolism during apoptosis. PS externalization to the outer plasma membrane is an early apoptotic marker, and PS biosynthesis is stimulated in different apoptotic cells 74 . In camptothecin-induced apoptotic human leukemia U937 cells, the total synthesis of PS is stimulated greater than other major membrane phospholipids such as PC, PE, and sphingomyelin. Jurkat cells undergoing CD95-induced apoptosis show enhanced PS synthesis by inhibiting the formation of PE through decarboxylation of the PS. Particularly, newly synthesized PS is preferentially translocated to the outer leaflet of the plasma membrane; thus, increased PS synthesis may contribute to the phagocytosis of apoptotic cells 75,76 . In the PI class, the levels of PI species with saturated and monounsaturated fatty acyl chains decreased in the COR-treated melanoma cells. Interestingly, the levels of PI 16:0/18:0, PI 16:0/18:1, PI 18:0/18:0, and PI 18:0/18:1, whose accumulation is suggested to be a potential biomarker of melanoma metastasis in our previous study, decreased after COR treatment 45 . PI and its metabolites play an important role in various cellular responses. Particularly, phosphatidylinositol-3-kinase (PI3K)/Akt pathway is critical for melanoma initiation and therefore is a therapeutic target [77][78][79] . Previous studies showed that PI3K inhibitors prevented Akt activation, induced apoptosis of various melanoma cell lines, and inhibited tumor growth in an in vivo mouse model of melanoma brain metastasis 78,80,81 . Although the association between the levels of PI and inhibition of cancer cell proliferation remains unclear, inhibitors of PI synthesis suppress the growth of cancer cells, including small cell lung carcinoma and oral squamous carcinoma cells 82,83 . Together, these findings suggest that the COR treatment-induced reduction in the levels of signaling lipids such as PI phosphates and PI suppresses the growth and proliferation of melanoma cells by inactivating the PI3K/Akt pathway.
Thus, the present study investigated the effects of COR on the proliferation of metastatic melanoma cells by performing metabolite and lipid profiling combined with multivariate statistical analysis. To our knowledge, this is the first study to show the metabolic and lipidomic alterations in COR-treated melanoma cells. The results of the present study showed that COR treatment suppressed the growth of melanoma cells by inhibiting glycolysis, TCA cycle, and glutaminolysis, which play essential roles in energy production during cancer cell proliferation. In addition, COR treatment induced the accumulation of Cer, a tumor suppressor, and increased the levels of TG and PS. Interestingly, the levels of PI species, including PI 16:0/18:0, PI 16:0/18:1, PI 18:0/18:0, and PI 18:0/18:1, which are suggested to be the potential biomarkers of melanoma metastasis, decreased in the COR-treated melanoma cells. This decrease in the levels of PI species was suggested to inhibit cancer cell growth. Together, these results provide new insights on the major metabolic and lipidomic alterations induced by COR in melanoma cells and offer a basis for using COR in melanoma treatment. Future in vivo studies assessing the toxicity, antiproliferative activity, and mechanisms under the antiproliferative activity of COR will help in determining its potential for treating metastatic melanoma.

Methods
Chemicals and reagents. High-performance liquid chromatography-grade chloroform, methanol, and water were purchased from Fisher Scientific (Pittsburg, PA). Ammonium acetate, butylated hydroxytoluene, COR, 2-DG, dimethyl sulfoxide (DMSO), MJ, methoxyamine hydrochloride, myristic-d 27  Cell culture. The primary human epidermal melanocytes HEMn-LP were cultured in medium 254 supplemented with human melanocyte growth supplement (Cascade Biologics, Portland, OR). The human metastatic melanoma cell lines A375 and A2058 (American Type Culture Collection, Manassas, VA) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Hyclone Labs, Logan, UT). All the cells were cultured at 37 °C in a humidified incubator with an atmosphere of 5% CO 2 and were subcultured when they reached 90% confluency.
Cell viability and combination treatment analysis. The cells were seeded in 96-well microtiter plates at a density of 2 × 10 4 cells/well and were incubated at 37 °C for 24 h. Next, the cells were treated with MJ, COR, or 2-DG alone or their combinations at different concentrations for 24 h. 2-DG was added 1 h before adding MJ or COR. Stock samples of MJ and COR were dissolved in ethanol and DMSO, respectively, and were diluted with the respective culture media immediately before performing the experiments. Final concentrations of ethanol and DMSO in the cell culture medium were always <0.1% (v/v) and had no effects on cell proliferation. Cell viability was determined at 24 h after the treatment by adding 10 μL MTT solution, followed by incubation for 1 h at 37 °C. Formazan crystals in viable cells were dissolved using 100 μL DMSO. Optical density of the dissolved formazan was measured using a microplate spectrophotometer (xMark; Bio-Rad, Berkeley, CA) at 570 nm. Cell viability data were then analyzed using CompuSyn software (ComboSyn Inc., Paramus, NJ) to calculate the IC 50 and CI (2019) 9:3140 | https://doi.org/10.1038/s41598-019-39990-w www.nature.com/scientificreports www.nature.com/scientificreports/ values. CI values provide a quantitative assessment of the synergism between drugs being examined based on Chou and Talalay method. CI values were calculated using the following equation: CI = (D) 1 /(D x ) 1 + (D) 2 /(D x ) 2 , where (D) 1 and (D) 2 are the doses of drugs 1 and 2, respectively, used in combination for a given effect and (D x ) 1 and (D x ) 2 are the doses of the drugs 1 and 2 alone, respectively, inducing the given effect. CI values of 1, < 1, and >1 indicate that the drugs 1 and 2 exert additive, synergistic, and antagonistic effects, respectively 84 .
Cell proliferation assay. The proliferation of A375 and A2058 cells (density, 2 × 10 4 cells/well) was measured using Cell Proliferation ELISA, BrdU (colorimetric) kit (Sigma-Aldrich), according to the manufacturer's instructions. sample preparation. The cells (seeding density, 4 × 10 5 cells/well) were detached by treating with trypsin-EDTA solution and the cells from four wells of a six-well plate were pooled. Next, the cells were pelleted by centrifugation at 1000 × g and 4 °C for 2 min, washed twice with ice-cold PBS to remove extracellular metabolites and residual medium, and immediately frozen with liquid nitrogen to quench cellular metabolism. The deep-frozen cells were suspended in PBS and were lysed using two freeze-thaw cycles. Briefly, the cells were thawed in a 4 °C water bath, vortexed, and sonicated on ice for 20 min. Next, the cells were transferred to liquid nitrogen for 60 min, thawed in a 37 °C water bath for 30 min, and vortexed briefly. This freeze-thaw cycle was repeated once for complete cell lysis, followed by additional sonication for 10 min. Protein content in the cell lysates was estimated using BCA Protein Assay Kit (Thermo Scientific, Rockford, IL), with bovine serum albumin as a standard for normalization. Before the analysis, the samples were freeze-dried and stored at −80 °C. Metabolites and lipids were extracted by using a modified Folch procedure, as described previously 45 . Briefly, an extraction mixture (chloroform/methanol, 2:1, v/v) was added to the freeze-dried cells. The mixture along with the cells was vortexed for 20 s, followed by sonication in ice-water for 30 min. The mixture was then incubated with shaking on ice for 40 min, followed by the addition of ice-cold water to induce phase separation. The mixture was incubated again with shaking for 10 min and was centrifuged at 18,500 × g and 4 °C for 10 min. Both the upper phase and lower phase were collected separately for performing metabolite and lipid analyses.
Metabolite analysis using GC-MS. Derivatization was performed by transferring 0.4 mL cell extracts into GC vials, followed by evaporation under nitrogen gas. The dried extracts were dissolved in 30 μL 20,000 μg/ mL methoxyamine hydrochloride in pyridine, 50 μL BSTFA containing 1% TMCS, and 10 μL myristic-d 27 acid in pyridine (500 μg/mL as an internal standard) and were incubated at 65 °C for 60 min before performing the GC-MS analysis. Each derivatized sample (1 μL) was injected into Agilent 7890 A GC system equipped with a 7683B series autosampler and 5975 C mass selective detector (Agilent Technologies, Santa Clara, CA), with a split ratio of 1:10. DB5-MS column (length, 30 m; inner diameter, 0.25 mm; and film thickness, 0.25 μm; Agilent Technologies) was used with a constant flow rate of 1.0 mL/min, with helium as carrier gas. In electron impact ionization mode, electron energy was 70 eV. Temperatures of the ion source, quadrupole, and auxiliary were set to 230 °C, 150 °C, and 280 °C, respectively. A full-scan mode was used in a mass range of 50-700 Da. The GC analysis of metabolites in the cell extracts was initiated at an initial oven temperature of 70 °C, followed by an increase in the temperature to 190 °C (at 5 °C/min), 240 °C (at 6 °C/min), and 280 °C (at 5 °C/min). All the metabolites were identified by comparing the obtained mass spectra with NIST-Wiley Mass Spectra Library, and peaks with a matching quality of >70% were assigned compound names. The spectra were also matched with the spectra in the Human Metabolome Database (HMDB; http://www.hmdb.ca/) and Golm Metabolome Database (GMD; gmd.mpimp-golm.mpg.de/). Leucine and isoleucine were identified by using chemical standards to compare the retention times as well as MS/MS fragmentation.
Lipid analysis using nanoESI-MS. Lipid extracts were passed through a 0.2-μm PTFE syringe filter (Whatman, Maidstone, UK), and 0.9 mL filtrate taken and evaporated under nitrogen gas. The dried extracts were reconstituted in 150 μL methanol/chloroform (9:1, v/v) containing 7.5 mM ammonium acetate buffer solution. 1,2-Diheptadecanoyl-sn-glycero-3-phosphoethanolamine (PE 17:0/17:0: m/z of 720.5 in the positive ion mode and 718.5 in the negative ion mode) was used as an internal standard. Automated shotgun experiments were performed using linear-ion-trap mass spectrometer (LTQ-XL; Thermo Fisher Scientific, San Jose, CA) equipped with a robotic nanoflow ion source (TriVersa NanoMate System; Advion Biosciences, Ithaca, NY) in the positive and negative ion modes. Next, 10 μL sample was infused into the MS system through a nanoelectrospray chip with 5.5-μm-diameter spray nozzles. The ion source was controlled using Chipsoft 8.3.1 software (Advion Biosciences). Ionization voltage and backpressure were 1.4 kV and 0.4 psi, respectively, in the positive ion mode and −1.7 kV and 0.6 psi, respectively, in the negative ion mode. Data were acquired in the profile mode for 2 min, and scan range was set at m/z 400-1,200. Capillary and tube lens voltages were set to 49 and 145 V, respectively, in the positive ion mode and −9 and −72.67 V, respectively, in the negative ion mode. A data-dependent MS/MS scan was performed under the collision energy offset of 35 eV. Dynamic exclusion parameters were set at a repeat duration of 60 s, exclusion duration of 60 s, and exclusion list size of 50. All spectra were recorded using Xcalibur software (version 2.2.; Thermo Fisher Scientific), and lipid species were identified using LipidMAPS (http://www. lipidmaps.org/), LipidBlast, and an in-house MS/MS library. Data processing. To relatively quantify the metabolites and lipids, the GC-MS spectrum data were processed using Expressionist ® MSX software (version 2013.0.39; GeneData, Basel, Switzerland). A list of retention time, base peak intensity, and m/z values was obtained for each chromatogram. Raw data files (*.raw) of lipids were converted to *.mzXML files by using ProteoWizard msConvert software, and the spectrum data were further processed using Expressionist ® MSX software. Data matrices, including m/z and peak intensity, were exported as Excel files (version 2010; Microsoft, Redmond, WA). Normalization was performed by dividing the peak intensity of each compound by that of the internal standard and by the total protein content.