Data collection and processing
The RNA-seq transcriptome data of 36 primary ICC samples and 9 non-tumor samples (33 ICC samples and 8 non-tumor samples have complete clinical information) were obtained from the TCGA database (http://cancergenome.nih.gov/). These data were current as of February 17, 2021. Next, the Ensembl database (http://asia.ensembl.org/index.html) was used to convert gene names from Ensembl IDs to a matrix of gene symbols. In addition, differentially expressed genes (DEGs) were obtained by using R package “edgeR.” The selection criteria were as follows: false discovery rate (FDR) < 0.05 and |log2 fold change|> 1.0. The level of mRNAsi (mRNA expression-based stemness index) and EREG-mRNAsi (epigenetically regulated mRNAsi) were detected in normal and tumor samples. The corrected mRNAsi (mRNAsi/tumor purity) and corrected EREG-mRNAsi (EREG-mRNAsi/tumor purity) were measured in tumor samples. The prognostic value of corrected mRNAsi or corrected EREG-mRNAsi was estimated by Kaplan–Meier analysis and log-rank test.
Confirmation of significant modules and key genes
Weighted gene co-expression network analysis (WGCNA) was performed using the WGCNA R package [20, 21]. EREG-mRNAsi and mRNAsi were chosen as the sample traits to locate CSCs-related modules and genes. Modules with the highest module significance were considered as the most sample trait-related modules. After detecting modules of interest, we calculated gene significance (GS) and module membership (MM) for each gene. The thresholds for screening target genes in a certain module were defined as cor.MM > 0.8 and cor.GS > 0.5.
Cell lines and cell culture
The cholangiocarcinoma cell line RBE and HCCC9810 were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The primary ICC cells were gathered from fresh clinical tissue specimens. The culture medium for primary ICC cells and cholangiocarcinoma cell lines was RPMI1640 (contained 20% fetal bovine serum, 100 μg/ml penicillin G and 100 U/ml streptomycin) in a humidified cell incubator under an atmosphere of 5% CO2 at 37 °C.
qRT-PCR and western blot
A total of 40 pairs of ICC tumor and paratumor collected at Renji Hospital from 2011 to 2013 were used for qRT-PCR analysis and 4 pairs of ICC specimen were collected for western blot analysis. Total RNA was isolated by TRIzol reagent (Invitrogen) and reverse-transcribed to cDNA with a cDNA Synthesis Kit (Thermo Scientific). qRT-PCR was performed on the CFX96 Real-Time PCR system with SYBR Green PCR Master Mix (Applied Bio-Rad). Primers sequences are listed in Additional file 2: Table S1. Protein extracts of ICC tissues and cholangiocarcinoma cells were analyzed by western blot according to the standard protocol provided by manufacturer.
Target gene overexpression and shRNA-mediated interference
For stable cell line construction, lentivirus and shRNAs productions were completed by Hanbio Biotechnology Co., Ltd. (Shanghai, China) and used according to the manufacturer’s protocols. Briefly, cells were transfected with concentrated virus at a multiplicity of infection of 20 with polybrene for 6 h and then 1:1 fresh medium was added for the following 18 h. Expression of target genes in the infected cells was validated by qRT-PCR and western blot.
Flow cytometric analysis
Cholangiocarcinoma cells were stained with PE-conjugated antihuman CD133 and APC-conjugated antihuman CD44 antibodies (eBioscience). The isotype-matched IgG was served as a control (eBioscience). The fluorescence intensity was measured on a flow cytometer (BD Biosciences). For cell sorting, PE-conjugated antihuman CD133 antibodies and APC-conjugated antihuman CD44 antibodies were incubated with cholangiocarcinoma cells, followed by sorting with a flow cytometer (BD Biosciences). In our study, CD133+CD44+ subgroup was defined as CSC, while the rest cells were defined as non-CSC.
Spheroid formation assay
A total of 1000 single cholangiocarcinoma cell lines were plated onto 12-well ultra-low attachment culture dishes (Corning) in serum-free DMEM/F12 medium for 7 days. DMEM/F12 was supplemented with 20 ng/ml epidermal growth factor (EGF, Invitrogen), 10 ng/ml basic fibroblast growth factor (bFGF, Invitrogen), B27 (1:50; Invitrogen), N2 (1:100; Invitrogen), 1% sodium pyruvate, 100 μg/ml penicillin G and 100 U/ml streptomycin. The primary spheres were derived from fresh clinical tissue specimen and subsequently dissociated with trypsin and resuspended in DMEM/F12 medium with the above supplements. Tumorsphere-derived adherent cells were maintained in the same medium without EGF, bFGF, B27, N2 and were instead supplemented with 10% fetal bovine serum. The number of spheroids formed (≥ 50 μm) was counted under an inverted microscope (Leica).
Cell invasion assay
The transwell chamber was coated with 1: 8 diluted Matrigel (100 μl, Corning, USA) and incubated for 5 h. Then, a total of 1 × 105 cells were suspended in serum-free RPMI1640, seeded into the upper chamber and allowed to migrate toward RPMI1640 (with 10% FBS) in the lower side of the chamber for 48 h. The migrated cells were fixed in formaldehyde and stained with crystal violet. The invaded cells were counted from three random fields of microscope.
Colony formation assay
In total, 1000 cells were seeded in 6-well plates for 14 days. After different treatment, the colonies were fixed with 10% PBS-buffered formaldehyde and stained with crystal violet to visualize the colonies.
Cell viability was determined by the Cell Counting Kit-8 (CCK8) assay (DOJINDO). CCK8 assay was performed according to manufacturer’s protocols. Briefly, cells were cultured in 96-well plates at 1 × 104 cells/well to attach overnight, followed by additional treatment with 5-fluorouracil (5-FU; Sigma) or cisplatin (CDDP; Sigma) at different concentrations. All experiments were based on at least three parallel measurements and each measurement contained triplicates.
TACE therapy after surgery
ICC tissues and paired paratumor tissues of primary cohort were obtained from 71 ICC patients treated with adjuvant TACE (1–2 months after surgery) at Renji Hospital (Shanghai, China); the validation cohort contained 34 cases gathered from the same institution. The detailed clinical pathological characteristics of the patients are presented in Additional file 2: Table S2. The regimen for the adjuvant TACE consisted of 5-FU 0.75 g, mitomycin C 16 mg, CDDP 60 mg and lipiodol 5 ml. In our study, all clinical samples and patient information were obtained following informed consent and protocols which were approved by the ethical review committee of the WHO Collaborating Center for Research in Human Production (authorized by Shanghai Municipal Government).
Tissue microarray (TMA) and immunohistochemistry (IHC)
A total of 71 samples of ICC specimens (primary cohort) collected at Renji Hospital from 2011 to 2015 were used to construct the TMA slice I. Another 34 ICC tissue samples (validation cohort) gathered in Renji Hospital were used to build the TMA slice II (2015–2017). Then, the TMA slices were subject to immunohistochemical staining according to standard protocols using specific antibodies against MRPL11 (Cell Signaling Technology); SDHAF2, MRPS34 and COX8A (Invitrogen). The staining score was assessed by the percentage of positively stained area and staining intensity. Two experienced pathologists calculated the IHC score independently, and the final IHC score was averaged from both pathologists.
Metabolomic analysis and sample preparation
For suspension samples, cells were collected and quenched with four volumes of sodium chloride solution (150 mM), vortex mixing 1 min, 3000 g centrifugation for 5 min at 4 °C, and then discarded the supernatant. For adherent samples, after discarding the culture medium, the surface of the plates was washed by sodium chloride solution (150 mM) and then released adherent cells from the plate surface by a cell scraper. Cell pellets were collected as described before. Polar metabolites (aqueous fraction) and lipid species (organic fraction) in cell pellets were separated by a two-phase liquid–liquid extraction as described previously [14, 22]. LC–MS analysis was conducted by using a UPLC system (Waters Corp) and a mass spectrometer (Thermo Fisher Scientific).
Methionine cycle inhibition assay
Vector cells and lentivirus transfected cells (Lv-SDHAF2, Lv-MRPS34, Lv-MRPL11 and Lv-COX8A) were treated with FIDAS-5 (5 μM final concentration, Merck Millipore), and the mRNA level of pluripotent transcription factors was further analyzed by qRT-PCR.
Methionine-restricted diet nude mice model
For in vivo methionine-restricted diet nude mice model, 3-week-old mice (male BALB/c nude mice) were subjected to either the control or the methionine-restricted diet. Two weeks after that, bearing tumor cells (derived from either control RBE cells or RBE cells with overexpression of 4-key-genes) were implanted subcutaneously into the left flank of nude mice. Then, the mice were intraperitoneally injected with CDDP (5 mg/kg body weight), twice a week for 3 weeks. Tumor growth was calculated by the following formula: V = 0.5 × W2 × L. All procedures involving animals were approved and performed in accordance with the Animal Care and Use Committee of Shanghai Jiaotong University.
Methionine dietary and methionine supplementary therapy
Methionine intake was gathered by analyzing daily diet of ICC patients. Low-methionine diet was defined as less than 8.6 mg kg−1 day−1, about 50% reduction in daily methionine intake. High-methionine diet was defined as more than 8.6 mg kg−1 day−1. Methionine supplementary therapy was defined as ≥ 3 days amino acids supplement following adjuvant TACE, containing 2.25 g/day methionine supplement (intravenous drip). Informed consent was obtained from each patient, and protocols were approved by the ethical review committee mentioned before.
The statistical analysis was carried out using SPSS 20.0 and GraphPad Prism 7. The Student’s t test or Pearson’s chi-square test was used to compare the difference of clinicopathologic features between two groups. The Kruskal–Wallis test was conducted as a nonparametric test when appropriate. Survival and tumor recurrence rates were analyzed using Kaplan–Meier analysis and log-rank test. Univariate and multivariate analyses were performed based on a Cox proportional hazard regression. A nomogram was built based on the independent factors for OS. The nomogram accuracy was evaluated by C-index and calibration curves. The cutoff point of nomogram was set as 135 (3-year overall survival = 50%). Bootstraps with 2,000 resamples were used to stabilize our results. All statistical tests were two-sided, and a p value of < 0.05 (*), < 0.01 (**), or < 0.001 (***) was considered significant.