Cell lines and cell culture conditions
The human hepatoellular carcinomas cell line HepG2, Hep3B, and Huh7 were obtained from ACTT with ACTTnumber: HB-8065, HB-8064, and PTA-4583. The mouse hepatoellular carcinoma cell line Hepa1-6 was obtained from ACTT with ACTTnumber: CRL-1830. The HEK293T cells were obtained from ACTT with ACTTnumber: CRL-3216. The Expi293F™ cells were obtained from Thermo Scientific with Cat#A14527. All of the cells were cultured in Dulbecco’s modified Eagle’s media (DMEM; Thermo Fisher Scientific) containing 10% fetal bovine serum (FBS; Thermo Fisher Scientific), 100 μg/mL streptomycin, and 100 units/mL penicillin (Thermo Fisher Scientific) at 37 °C with 5% CO2. Cells were passaged or harvested by incubation with 1 × TrypLE Express (Life Technologies) at 37 °C for 2 and 5 min.
Animal studies
The animal care and experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of National Center for Protein Sciences (Beijing), Ethical review number: IACUC-20210914-32MT. NOD-SCID and C57BL/6 J mice were purchased from Charles River, Inc (Beijing, Vital River Laboratory Animal Technology). All mice were female mice of 4–5 weeks. After tumor transplantation, the tumor grew to 100 mm3 and started to randomly divided into three groups: control (200 μL-30% PEG400 + 0.5% Tween 80 + 5% propylene glycol, intraperitoneal injection), sorafenib (20 mg kg−1 day−1, intraperitoneal injection), nilotinib (20 mg kg−1 day−1, intraperitoneal injection). After treatment for 28 days, the mice were killed by spinal dislocation. The investigator was blinded to group medication during the experiment and when assessing the outcome. Animals were housed in a specific pathogen-free mouse facility at the animal center National Center for Protein Sciences (Beijing). Maximal tumor size of CDX mouse model of experimental endpoint is less than 20 mm in any dimension, in accordance with the IACUC. None of the guidelines was exceeded in any experiment performed. Tumor size was measured using a caliper. Tumor volume in mm3 was calculated using the formula: tumor volume = 0.5 × length × (width)2.
Targeting SOAT1-based virtual screening approaches
The crystal structure of SOAT1 protein was obtained from the Research Collaboration for Structural Bioinformatics Protein Data Bank (RCSB PDB) (protein PDB code: 6VUM and 6L47) [12, 13]. The screening compound database contains 31,276 compounds recognized by the FDA obtained from Selleck (https://www.selleck.cn/screening-libraries). The virtual screening was performed using the previously reported method. Specifically, we used three widely accepted docking software, AutoDock 4.2 [14], sybyl 2.0 [15], and Glide [16] for high-throughput molecular docking. For molecular docking using autoDock, we first charged the protein 3D crystal structure, added polar hydrogens and dehydrated, and energy-optimized, and then created a lattice pocket to accommodate the entire SOAT1 catalytic site. The compounds are also subjected to polar hydrogenation and charge addition, and energy optimization to generate the molecule to be docked. The docking method adopts the flexible docking mode. Then the small molecules in the compound library were docked within the catalytic site pockets through a genetic algorithm and sorted according to the classical free energy function. These input files, including protein structure, active pocket box, and Gasteiger partial charge, were set to default parameters. Sybyl 2.0 docking was performed utilizing the Surflex-dock module. First, the protein and compound library molecules were polarized, hydrogenated, charged, and energy-optimized. The interface bag was centered on the 15-Å rectangle with the coordinates of the His460 residue at the SOAT1 catalytic site. The semi-flexible docking method used the integrated score as the scoring function. The Glide docking module was implemented in Schrödinger version 2015–4. We first used the protein preparation wizard module to add a polar hydrogen, charge, and remove water from the SOAT1 protein. Then, the binding site was defined as a 15-Å rectangle centered on the coordinates of the His460 residue of the SOAT1 catalytic site. At the same time, the LigPrep module with default parameters was utilized to process all screened compounds. For the Glide docking step, we selected standard precision (SP) and super precision (XP) for molecular docking. The ligand docking adopts flexible docking, and the energy is minimized. Finally, we utilized GlideScore as the scoring function. The Glide XP docking posture was utilized to display the final docking result, and the result images were completed by PyMOL 2.3 [17].
Expression and purification of SOAT1 protein
The SOAT1 protein expression and purification refer to the previously reported method [12, 13]. Specifically, we cloned the cDNA of human SOAT1 (NCBI reference sequence NM_003101.6) into a pCAG vector with a 6 × His-tag at the carboxyl end. The HEK293F suspension cells were cultured in Freestyle 293 medium (Thermo Fisher Scientific) in a constant temperature incubator at 37 °C and provided with 5% CO2 and 80% humidity. When the cell density reached 2 × 106 cells per milliliter, the cells were transiently transfected with the expression plasmid and polyethyleneimine. Approximately 1 mg of expression plasmid and 3 mg of polyethyleneimine (Sigma-Aldrich) are pre-mixed in 50 ml of fresh medium and incubated for 15–30 min before transfection. Then 50 ml of the mixture was add to 1 l of cell culture and incubate for 15–30 min. The transfected cells were collected after incubating for 48 h. To purify SOAT1, the collected HEK293F cells were resuspended in a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl (Buffer A), and a protease inhibitor mixture (approximately 1 g of cells plus 5–10 mL of Buffer A). After sonication on ice, the cell membrane pellet was collected and weighed by centrifugation at 120,000 g in an ultra-high-speed centrifuge at 4 °C for 2 h. The membrane fraction was subjected to 25 mM Tris pH 8.0, 150 mM NaCl, and 1% (w/ v) glyco-diosgenin (GDN, Anatrace) (Buffer B) dissolved for 2 h. After centrifugation at 120,000 g for 1 h, the supernatant was sought and applied to the nickel column for affinity purification and washed with a washing buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, 40 mM imidazole, and 0.02% GDN (Buffer C). The protein was eluted with a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, 400 mM imidazole, and 0.02% GDN (Buffer D). Then the eluate was concentrated and purified by size exclusion chromatography (SuperdexTM200 10/300 gL, GE Healthcare) in a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, and 0.02% GDN (Buffer E), and then purified. The protein concentration and SDS-PAGE detection, the protein concentration, and purity detection through the standard (purity > 95%) can be divided and stored at − 80 ℃ for future use.
Surface plasmon resonance for affinity screening and affinity determination
A Biacore T200 (GE Healthcare) optical biosensor was used for affinity screening of target proteins and the determination of equilibrium dissociation constant (KD) value measurements for protein–ligand interactions. The CM5 chip (GE Healthcare) was utilized to couple the target protein, and the running buffer was phosphate buffer saline (PBS, Sigma-Aldrich) containing 5% dimethyl sulfoxide (DMSO; Sigma-Aldrich). Before coupling, the chip channel was activated with 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC, GE Healthcare) and N-hydroxysuccinimide (NHS, GE Healthcare) at a flow rate of 10 μL/min for 60 s. After that, the recombinantly expressed and purified SOAT1 protein was diluted with 10 mmol/L sodium acetate buffer (pH 5.0) to about 50 mg/mL. The flow rate for SOAT1 protein solution was 10 μL/min, and the injection time lasted for 120 s. The protein coupling volume of the channel was about 8000 RU, and the channel was sealed with ethanolamine at a flow rate of 10 μL/min for 120 s. For molecular fragment library screening, the concentration of each compound was diluted to 100 nmol/L with running buffer in a 96-well measuring plate, and then passed through the CM5 chip coupled to the target protein at a flow rate of 30 μL/min for 180 s. After specifying a sample, the chip was regenerated with NaOH at a flow rate of 30 μL/min for 60 s. All raw datum were recorded and stored. When determining the protein–ligand interaction affinity KD value, each compound was diluted 11 times from 20 μM to 0.0195 nM, and the small molecules were passed through the chip coupled to the target protein from low concentration to high concentration. The flow rate was 30 μL/min and the duration was 180 s. After each concentration point flowed through, the chip was regenerated with NaOH at a flow rate of 30 μL/min for 60 s, and then recorded and saved the data in real time. Molecular weight adjustment and solvent correction were used simultaneously to remove non-specific binding and signal drift molecular effects. Finally, all data processing was performed in Biacore T200 analysis software (GE Healthcare).
Cell viability assay
In order to determine the effect of the compound on the viability of liver cancer cells, the CCK-8 kit (Sigma-Aldrich) was used to detect the cell viability. HepG2, Hep3B, and Huh7 liver cancer cells were cultured in a 96-well plate, using DMEM containing 10% FBS, 100 μg/mL penicillin, and 100 μg/mL streptomycin, in a 37 °C, 5% CO2 incubator 12 h, the concentration of a single compound was diluted 10 times from 20 μM, and each concentration point was repeated 3 times. The diluted compound was added to the serum-free medium to treat the cells for 48 h, and then the cells were evaluated with the CCK-8 kit for activity, referring to the instruction manual for the experimental method. After the Tecan microplate reader detects all the absorbance values, the IC50 of the drug to the cells is calculated according to the concentration-absorbance value in GraphPad (GraphPad Prism 8.2.1, GraphPad).
Cell proliferation assay
The cell proliferation experiment was carried out with xCELLigence RTCA DP (Roche), and the E-plate-16 cell detection plate (Roche) was used for cell culture and detection. Specifically, the HepG2, Hep3B, and Huh7 cells are first made into a cell suspension and counted, and the cell concentration is diluted to 8 × 104 cells/L for later use. After adding 50 µL of medium to the wells of E-plate-16 for baseline detection, 100 μL of reconciled cell suspension was added to the wells to make the number of cells per well 8000 cells/100 μL. After standing on the clean bench for 30 min at room temperature, the E-plate-16 is put on the RTCA station again for testing. The working program is set to detect the cell index once every 15 min, record the cell growth process, pause the program after culturing for 3–4 h, add the prepared drugs to it (the concentration of each drug refers to the IC50 value of the cell viability determination), and continue the cultivation and observation. After adding the drug, the program is set to detect the cell index once in 15 min and continuously observe and record for 120 h. The cell growth index-time curve was finally drawn in GraphPad Prism 8.2.1.
Cell migration assay
The effect of the drug on the migration of liver cancer cells was performed using a classic transwell migration assay. HepG2, Hep3B, and Huh7 cells in logarithmic growth phase were taken, and the number of cells was adjusted to 1 × 106 with serum-free DMEM medium. Two hundred microliters of cells was inoculated into the upper chamber of transwell (Millipore). Six hundred microliters of DMEM medium containing 20% FBS was added in the chamber. After culturing in a 37 °C, 5% CO2 incubator for 12 h, the drug was added to be tested in the upper and lower chambers of the transwell. The drug concentration shall consult the IC50 or twice the concentration determined above. After culturing for 24 h, the transwell was removed with a cotton swab. Cells that did not penetrate the chamber were fixed with methanol for 15 min, washed with PBS 3 times, stained with crystal violet for 30 min, and photographed in five different fields of view from the top, bottom, left, and right under a × 100 optical microscope, and the average number of cells was calculated. Inhibition rate of the drug on cell migration = (1-addition group/control group) × 100%.
Cell cycle assay
In the cell cycle assay, flow cytometry was used to detect the propidium iodide (PI)-stained cells. The cells were fixed (approximately 2 × 106) in ice-cold buffer with 70% ethanol. Then, the cells were washed and resuspended with PBS. Finally, a reaction solution with 25 mg/mL PI (CWBIO, Beijing, China) and 50 mg/mL RNase A was added to the samples and incubated in the dark for 30 min at 37 °C to stain DNA. The fluorescence emitted from the PI-DNA of individual cells was measured using a BD LSRFortessa (BD, USA) FACS flow cytometer.
Colony formation assay
HepG2 and Hep3B cells were seeded in a 6-well plate at a density of 1 × 103 cells/well and incubated for 24 h. Cells were treated with nevanimibe (5 μM) and nilotinib (500 nM), and DMSO as a control. Cells were incubated for 8 days. The medium was carefully removed, and cells were washed with PBS and fixed with 100% methanol for 30 min. After removing methanol, cells were stained with 0.5% (w/v) crystal violet for 30 min and washed with tap water. The plate was dried at 25 °C and images were captured. The stained area was automatically measured using ImageJ. This experiment was independently performed 3 times. The area of the colony was calculated as a percentage of the total area of the well.
Xenograft model
In this study, human HepG2 and murine Hepa1-6 hepatocelluar carcinoma cells were used to in our murine xenotransplant tumor model, and immunodeficiency NOD-SCID mice and normal C57BL/6 J mice (Beijing Vital River Laboratory Animal Technology) were used to model cell-derived xenografts. We resuspended 5 × 106 cells in 200 μl sterile PBS and injected the resuspended cells into the right flank of each NOD-SCID or C57BL/6 J (female, 4–5 weeks old). Once the tumor reached 100 mm3, the mice were randomized to groups. In this experiment, the efficacy of nilotinib and sorafenib was evaluated at the same time, and a placebo control group (n = 6 for each model) was set up. The dosage of the three drugs was 20 mg/kg/day, and the drugs were administered to the mice via intrapulmonary injection. The administration was expected to continue for 4 weeks, and the tumor size was measured every other day. The mice were killed by spinal dislocation. During the experiment and when assessing the impact, the investigator was unaware of the group medication. The animals were housed in the pathogen-free mouse facility at the National Protein Science Center (Beijing, China) of the Animal Center. According to our Institutional Animal Care and Use Committee (IACUC) protocol, the maximum tumor size of the CDX mouse model is not to exceed 20 mm in any size. All guidelines were followed in this study. Vernier calipers were used to measure tumor size. The following formula was used to calculate the tumor volume in mm3: tumor volume = 0.5 × length × (width)2.
Preparation of the proteome and metabolome samples
For omics samples, HepG2 cells grown in log phase were used for experiments. After the cells were cultured at 37 °C and 5% CO2 for 12 h, two compounds of nevanimibe and nilotinib were added to each dish (dosing refer to the aforementioned IC50), and the control group was equal to 1/1000 One DMSO, and then continue to incubate for 12 h. For proteome analysis, the cells were cultured in 75-cm2 plates. Each sample provided 5 × 10e6 cells, and each group was repeated three times. In the metabolism analysis, the cells were cultured in 150-cm2 plates. The number of cells in each sample was 1 × 10e7, and each group was repeated six times. Bio-Rad TC20TM (Bio-Rad) automatic cell counter was used to count all cells.
Proteome experimental procedures
According to the following experimental conditions, HepG2 cells were divided into three groups: (1) nevanimibe treatment, (2) nilotinib treatment, and (3) DMSO treatment. Each group contains three biological copies. Protein extraction, digestion, peptide separation, and liquid chromatography tandem mass spectrometry (LC–MS/MS) detection methods were roughly the same as our previous research methods. Briefly, the sample was lysed with lysis buffer ultrasonically and then centrifuged at 12,000 × g for 10 min at 4 °C to remove cell debris. The supernatant was transferred to a new centrifuge tube, and the protein concentration was determined via a bicinchoninic acid (BCA) Protein Assay Kit (Thermo Fisher Scientific). The same amount of protein in each sample was enzymatically digested, and the peptides digested by trypsin were desalted with Phenomenex and then freeze-dried in vacuo. The peptides were dissolved in the mobile phase A of liquid chromatography (0.1% (v/v) formic acid aqueous solution) and then separated using the nanoflow high-performance liquid chromatograph (HPLC) instrument (Easy nLC1000 System, Thermo Fisher) coupled to an Orbitrap Fusion mass spectrometer (Thermo Fisher) with a nanoelectrospray ion source (Thermo Fisher). Mobile phase A is an aqueous solution containing 0.1% formic acid and 2% acetonitrile; buffer B is an aqueous solution containing 0.1% formic acid and 90% acetonitrile. The mass spectra data were searched using ProteomeDiscover 2.0 and MaxQuant 1.6.1.0 and matched in the SwissProt human proteome database. The matching also used the anti-database to eliminate the false positive rate (FDR) caused by random matching. The protease was set to Trypsin/P, the minimum length of the peptide was set to 7 amino acid residues, the maximum missing cleavage sites was set to 2, and the maximum number of charges was set to 5. The maximum tolerable mass error of the primary precursor ion was set to 10 ppm, and the maximum tolerable mass error of the secondary product ion was set to 0.02 Da. The fixed modification was set to cysteine alkylation, and the variable modification was set to methionine oxidation. The peptide score was greater than 20 points. We also used 1% glyco-diosgenin (GDN) or 5% CHAPS for better extraction of membrane protein during protein extraction.
Metabolome experimental procedures
Total metabolites were extracted from each group using approximately 1 × 107 cells in 1 ml of MTBE (methyl tert-butyl ether, Sigma-Aldrich) methanol aqueous solution (7:2:1, v/v). First, 100 μL of water was added to the washed cells to resuspend the cells. After mixing, 200 μL of methanol solution was added. The samples were then placed in a shaker and vortexed for 3 min. Then 700 μL of MTBE was added, and the samples were vortexed for 3 min. The cells were then disrupted via ultrasound at 4 °C, 200 W power for 5 min. The program was set to ultrasound for 1 s and pause for 2 s. Finally, the samples were incubated at 25 ℃ for 1 h, centrifuged at 13,000 × g for 15 min, and then the supernatant was transferred to a new EP tube. The samples were vacuum-dried at 45 °C to remove organic solvents, and then placed in a freeze dryer to remove the remaining moisture. For LC–MS detection, the samples were dissolved with MTBE methanol aqueous solution (7:2:1, v/v). Samples were analyzed by LC–MS (a UPLC I-class/VION IMS QTOF Mass spectrometry system, Waters). The samples were separated using a C18 reverse chromatographic column (3 μm, Durashell, Agela Technologies). The column temperature was 20 °C, and the flow rate was 0.3 mL/min. The mobile phase of composition A was 10 mM ammonium acetate acetonitrile aqueous solution (acetonitrile: water = 6:4, v/v), and B was 10 mM ammonium acetate acetonitrile isopropanol solution (acetonitrile: isopropanol = 1:9, v/v). The gradient elution procedure was as follows: 0–7 min, B was maintained at 30%; 7–25 min, B changed linearly from 30 to 100%; 25.1–30 min, B was maintained at 30%. The sample was placed in the autosampler at 10 °C during the entire analysis. In order to avoid the influence caused by the fluctuation of the detection signal of the instrument, a random sequence was adopted to carry out continuous analysis of the sample. In the sample queue, one QC sample was set every eight experimental samples in order to monitor and evaluate the stability of the system and the reliability of experimental data. Mass spectrometry conditions were detected using electrospray ionization (ESI) positive ion and negative ion modes. The ESI positive source conditions were as follows: heater temperature 300 °C; spray voltage 3.0 kV; capillary temperature 350 °C; S-lens RF level 50%; MS1 scan ranges 50–1500. The ESI negative source conditions were as follows: heater temperature 300 °C; spray voltage 2.5 kV; capillary temp 350 °C; S-lens RF level 60%; MS1 scan ranges 50–1500. The mass-to-charge ratios of lipid molecules and lipid fragments were collected according to the following method: 10 slice patterns (MS2 scan, HCD) were collected after each full scan. MS1 had a resolution of 70,000 at m/z 200, and MS2 had a resolution of 17,500 at m/z 200. The raw data were collected by MassLynx software (MassLynx 4.2, Watres), and then Progenesis QI (Progenesis QI 2.4, Waters), MS-DIAL (Free down load on http://prime.psc.riken.jp/), and Unifi (Unifi 2.0, Waters) analysis software were used for non-targeting and targeted identification. The raw data of LC–MS/MS were for peak alignments, retention time correction, and peak area extraction. The alignment of MS1 and MS2 tolerance was set to 0.001 Da, and the retention time tolerance was set to 0.02 min. Metabolite structure identification used accurate mass matching (< 10 ppm) and secondary spectrum matching methods to retrieve self-built databases from the laboratory. SIMCA-P (SIMCA 14.1, Umetrics) and GraphPad software were used for data statistics and analysis. R_studio (Free down load on https://www.rstudio.com/) was used to perform correlation analysis and mapping.
SDS-PAGE and western blot analysis
Proteins were extracted in 25 mM Tris, pH 8.0, 150 mM NaCl, and 3% (w/v) CHAPS buffer. The cells were lysed by sonication on ice for 5 min (sonication for 3 s, off for 2 s; power 60 W), centrifuged at 12,000 × g for 15 min at 4 °C, and the supernatant was collected. The protein lysates were quantified with a bicinchoninic acid (BCA) assay and then resolved via SDS-PAGE gel electrophoresis. SDS-PAGE was performed on a 10% SDS-PAGE gel 90 min at 160 V, and then bromophenol blue staining was performed. Photographs were recorded after destaining with acetic acid–methanol-water solution. Western blot analysis was performed on unstained SDS-PAGE gels transferred to nitrocellulose membranes. Membranes were then blocked with 5% non-fat milk in Tris-buffered saline for 1 h at 25 ℃. After blocking, membranes were incubated 12 h at 4 °C with the following primary antibodies: rabbit anti-SOAT1 polyclonal antibody (ABN66, Merck) and rabbit anti-β-actin monoclonal antibody (6609, Proteintech). Membranes were then washed and incubated with secondary antibodies for 1 h at 25 ℃. Protein bands were visualized using enhanced chemiluminescent western blotting substrate.
Multicolor flow cytometry analysis
To analyze the effect of the SOAT1-targeting compounds on the tumor immunophenotype, the tumors from mice 28 days after dosing were taken down, ground into a single-cell suspension, and resuspended in a centrifuge tube with Hanks’ balanced salt solution (HBSS, Gibco), centrifuged at 2000 rpm for 4 min, and the supernatant was removed. Cells were then digested with collagenases 2 and collagenase 4 to remove connective tissue and collagen components. Finally, the immune cell population was enriched with 35% percoll. We used an anti-CD45-APC/cy7 antibody (103,115, Biolegend), anti-CD11b-APC antibody (K009928M, Solarbio), anti-CD3-BV510 antibody (100,233, BD Biosciences), anti-CD19b-FITC antibody (53,343, Cell Signaling Technology), anti-F4/80-PE antibody (64763S, Cell Signaling Technology), anti-Ly6G-Alexa fluor 700 antibody (127,622, Biolegend), anti-CD8-PerPC/cy5.5 antibody (100,733, Biolegend), and anti-CD49- BV421 antibody (740,030, BD Biosciences) for staining. The samples were collected by the BD LSR Fortessa flow cytometer (BD Biosciences). We then utilized FlowJo software (Free download on https://www.flowjo.com/) to analyze the data. The fluorescence-activated cell sorter (FACS) gating/sequencing strategy is shown in Supplementary Figure S6.
Bioinformatics analysis
In this study, the UniProt-GOA database (http://www.ebi.ac.uk/GOA/) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.kegg.jp/) Gene Ontology (GO) annotations and metabolic pathway annotations are provided. Volcano and Venn diagrams were used to identify and classify differential proteins or metabolites. All differentially expressed proteins or metabolites were imported into R_studio to search for relevant pathways. The outputs of the pathways were automatically classified into rank categories and were considered significant when the adjusted P value was < 0.05, and a P < 0.01 was considered very significant. GO analyzes the biological processes, cellular components, and molecular functions of dysregulated proteins or metabolites. Differential proteins or metabolites of each signaling pathway were marked using the online website KEGG, and the final images were visualized in Cytoscape 7.1 [18]. R_studio was used to draw the various volcano maps, Venn diagrams, scatter plots, and correlation heat maps.
Qantification and statistical analysis
For detailed analysis of statistical results, please refer to each method. The P values appearing in this report were calculated by two-tailed Student’s t test. A P < 0.05 was considered significant, and a P < 0.01 was considered very significant. Relative standard deviation (RSD) was utilized to express the accuracy of the analytical test results. The smaller the value, the better the repeatability. The false discovery rate (FDR) is the expected value of the number of false rejections, as a percentage of all rejected hypotheses. As a control index for the test hypothesis error rate, the control value was selected according to the need, and the value of the traditional hypothesis test was usually set to 0.05. In this study, all biochemical analyses were performed independently at least three times. GraphPad Prism 8.21 was utilized to analyze the data. A Pearson correlation analysis was used to measure the linear correlation of the data. A volcano diagram was used to display the results of differential expression analysis, and a Venn diagram was used to display the logical connection between different groups (sets). GO enrichment analysis and KEGG pathway analysis are based on online tools (such as KEGG mapper and string) and offline tools (such as Cytoscape 3.7). Microsoft Excel was used for other non-computational analyses (Microsoft Excel 2019).
Data and code availability
The protein MS data were deposited in iProX [19] (an official member of ProteomeXchange consortium) as an attachment file (http://www.iprox.org) with the project ID IPX0003944000. The metabolomics data reported in this paper were deposited in MetaboLights [20] as an attachment file (https://www.ebi.ac.uk/metabolights/MTBLS4088) with the Project ID MTBLS4088.