Public omics datasets
Public bulk-RNA seq datasets (TCGA-CHOL [6], TCGA-LIHC [26], GSE76297 [8], GSE26566 [27], E-MTAB-6389 [9], GSE32879 [28], E-MTAB-4171 [29], OEP001105 [11] and OEP000321 [30]), including gene expression profile of 565 intrahepatic cholangiocarcinoma (iCCA), 607 hepatocellular carcinoma (HCC), 7 combined hepatocellular carcinoma (CHC), 5 focal nodular hyperplasia (FNH) and 327 adjacent nontumor liver tissue specimens included in The Cancer Genome Atlas (TCGA), ArrayExpress, Gene Expression Omnibus (GEO), and National Omics Data Encyclopedia (NODE) were obtained and processed using R package TCGAbiolinks (version 2.20.1), oligo (version 1.56.0) and GEOquery (version 2.60.0), respectively [31,32,33]. Differentially expressed genes between nontumor liver tissue and iCCA or HCC samples were inferred using R package limma (version 3.50.3) based on |logFC| > 0.1 and p value < 0.05 [34].
Re-analyses of DNA methylation datasets GSE156299 [35] and TCGA-CHOL [6] were performed to decipher the methylation patterns of COL12A1 gene in iCCA and nontumor liver tissues. The raw data of DNA methylation were processed using minifi (version 1.38.0) as previously described [21, 36]. Weighted gene co-expression network analysis (WGCNA) was performed for weighted correlation network analysis of gene expression patterns in iCCA [37].
Single-sample gene set enrichment was performed using R package GSVA (version1.42.0) and GSEABase (version 1.56.0), and epithelia-to-mesenchymal transition (EMT) gene sets (GOBP_EPITHELIAL_TO_MESENCHYMAL_TRANSITION) in Molecular Signatures Database c5.all.v7.5.1.symbols.gmt was applied to calculate the EMT-gene set enrichment score in each sample [38].
Patients and clinical tissue specimen
A total of 10 iCCA, 10 HCC, and the paired nontumor liver tissue samples were prospectively collected from treatment-naive patients with iCCA or HCC post radical surgical treatment from May to September 2020 at our hospital. The clinicopathological characteristics of 10 iCCA/HCC patients are summarized in Additional file 8: Table S2. After snap-frozen in liquid nitrogen, the collected tissue samples were stored at − 80 refrigerator for subsequent molecular biology experiments as we previously described [21]. An external cohort (LVC1202, SUPERBIOTEK, Shanghai, CN), including 60 primary iCCA patients without receiving neoadjuvant treatment prior to primary surgery, was obtained for survival analysis. The TNM staging of each iCCA patient after surgical resection was performed in accordance with the 8th edition of the American Joint Committee on Cancer (AJCC) TNM classification system [39]. The pathological diagnosis of iCCA or HCC was performed according to World Health Organization criteria [40]. Written informed consent was collected from each patient enrolled in the present study. The experimental protocol (5826Z/20, 0531B/22) for using clinical tissue specimens was reviewed and approved by Institutional Ethics Committee in our hospital.
Cell lines, cell culture, stable cell line establishment and transient infection
RBE cell line was obtained from Yuchi Biological Technology Co., Ltd. (Shanghai, CN). HuCCT1 (JCRB0425) and HCCC9810 cell lines were kindly provided by the Key Laboratory of Carcinogenesis and Cancer Invasion (Fudan University), Ministry of Education, Shanghai, CN. CCLP1 and normal human intrahepatic biliary cell (HIBEpiC) lines were kindly provided by Key Laboratory of Combined Multiorgan Transplantation, National Health Commission, Hangzhou, Zhejiang, CN. HEK-293T cell line was purchased from the Cell Bank of Chinese Academy of Sciences (Shanghai, CN). RBE cell line came from a 37-year-old yellow race woman with primary iCCA. HCCC9810 cell line came from a 60-year-old yellow race woman with primary iCCA. CCLP1 cell line came from a 48-year-old Caucasian female with primary iCCA. HuCCT1 cell line came from a 56-year-old yellow race man with primary iCCA. HIBEpiC cells were isolated from normal liver tissue. All cell lines were tested by short tandem repeat typing. Furthermore, all cells were cultured in RPMI-1640 medium (Thermo Fisher Scientific, USA) containing 10% fetal bovine serum (FBS) (Thermo Fisher Scientific, USA), 100 U/mL penicillin (Cienry, Huzhou, CN) and 100 μg/mL streptomycin (Cienry, Huzhou, CN) at 37 °C in 5% CO2 atmosphere. Human hepatoma cell lines (including Hep3B, Bel-7402, HCC-LM3 and Huh7) and normal liver cell line HL7702 were cultured as we previously described [21]. Additionally, all of the cell lines had been tested regularly for mycoplasma contamination as previously described [21].
COL12A1 CRISPR/Cas9 KO plasmid (sc-402361) was purchased from Santa Cruz Biotechnology (USA). 2 × 105 HuCCT1 or RBE cells (per well) in the logarithmic phase were seeded in a 6-well tissue culture plate and incubated with antibiotic-free standard growth medium (3 mL per well). Upon cells growing to a 50–70% confluency after initial seeding, 2 μg (0.1 μg/μL) of plasmid DNA diluted in 130 μL of Plasmid Transfection Medium (sc-108062, Santa Cruz Biotechnology, USA) and 10 μL of UltraCruz® Transfection Reagent (sc-395739, Santa Cruz Biotechnology, USA) diluted in 140 μL of Plasmid Transfection Medium, were transfected into HuCCT1 and RBE cells following the manufacturer’s protocol. After incubation for 48 h, cells with successful transfection of COL12A1 KO plasmid were determined by detecting the green fluorescent protein (GFP) using fluorescent microscopy (Leica DMi8, Germany). Then, after cell sorting for GFP by FACS (MoFloAstrios EQ, Beckman), the harvested cells were cultured in a standard growth medium for 3–4 weeks and the selected single-cell colony with COL12A1 knockout was then used for the downstream experiments.
The mimic of miRNAs including miR-424-5p (miR10001341-1-5), miR-424-3p (miR10004749-1-5), miR-497-5p (miR10002820-1-5), miR-195-5p (miR10000461-1-5) and the negative control (miR-NC, miR1N0000001-1-10) were purchased from RIBOBIO Co. Ltd. (Guangzhou, CN). Transient transfection with the mimic of miRNAs was performed using riboFECT CP Transfection Kit (C10511-05, RiboBio Co. Ltd., Guangzhou, CN) in accordance with the manufacturer's protocol.
DNA methylation inhibitor decitabine (S1200) was purchased from Selleck (Shanghai, CN). 3 × 105 HuCCT1 or CCLP1 cells (per well) were seeded in a 6-well cell culture plate and incubated for 48 h in RPMI-1640 containing 10% FBS and 10/20 μmol/mL decitabine.
Protein extraction and immunoblotting
Protein extraction from fresh clinical tissue samples or human cell lines and immunoblotting procedures were performed as previously described [21]. Samples containing 20–40 μg proteins were subjected to SDS-PAGE (8–10% polyacrylamide) and blotted on polyvinylidene difluoride (PVDF) membranes. After blocking with 5% skim milk, the membranes were first incubated with COL12A1 antibody (1:2000, Sigma-Aldrich, #HPA009143), Vimentin antibody (1:1000, Abcam, #ab92547), E-cadherin antibody (1:1000, Cell Signaling Technology, #3195), or N-cadherin antibody (1:1000, Cell Signaling Technology, #13116), and then incubated with Goat Anti-Rabbit IgG H&L (1:5000, Abcam, #ab205718). GAPDH (1:5000, Abcam, #ab181602) was used as a loading control.
RNA isolation and real-time PCR
Total RNA isolation from clinical tissue samples or cell lines was performed using TRIzol™ Plus RNA Purification Kit (12183555, Invitrogen™, Thermo Fisher Scientific, USA) following the instruction for the user. Real-time PCR was performed as previously described [21]. The primers of the human COL12A1 and β-actin gene (GCD0259363) were purchased from GeneChem (Shanghai, CN). The primer sequences of COL12A1 and β-actin are as follows: COL12A1, forward 5′-AGGTCGGATGACGGGAAGA-3′ and reverse 5′-GCGGACATTCAAGGTGCTG-3′; β-actin, forward 5′-CGACAGATGCAGAACGAGA-3′ and reverse 5′-GACCCTGGATGTGACAGCTC-3′.
MicroRNAs (miRNAs) in clinical tissue samples and human cell lines were extracted using an ultrapure miRNA isolation kit (5080576001, Roche) following the instructions for the user. The cDNA template was produced using Bulge-Loop miRNA qRT-PCR Starter Kit (C10211-2, RIBOBIO, Guangzhou, CN) following the instructions for the user. U6 snRNA was used as the endogenous control, and miRNAs primers (MQPS0001272-1-200, MQPS0001271-1-200, MQPS0000002-1-200) were purchased from RiboBio Co. Ltd. (Guangzhou, CN). A final volume of 20 μL mixture containing Bulge-Loop miRNA qRT-PCR Starter Kit (C10211-1, RIBOBIO, Guangzhou, CN) reagents, RT-PCR Grade Water (AM9935, Applied Biosystems, Thermo Fisher Scientific, USA) and cDNA template was amplified using ABI QuantStudio-5 Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific, USA) in accordance with the manual for users. Reaction conditions were set as 95 °C for 10 min, 40 cycles of denaturation at 95 °C for 2 s, annealing at 60 °C for 20 s and extension at 70 °C for 10 s. The ΔΔCT method was used to determine the relative miRNA expression level in each sample.
Hematoxylin and eosin staining and immunohistochemistry
Immunohistochemical (IHC) or Hematoxylin and eosin (H&E) staining was performed as previously described [2, 21]. Paraffin-embedded liver tissue sections (3 μm thickness) were incubated with anti-COL12A1 (1:200, Sigma #HPA009143) or anti-Ki67 (1:300, Abcam, #ab16667), respectively. The COL12A1 staining score was used to divide iCCA patients into COL12A1-low and COL12A1-high expression groups based on an optimal threshold.
Immunofluorescence
A total of 2 × 103 iCCA cells, seeded on coverslips coated in a 24-well cell culture plate, were incubated in RPMI-1640 medium containing 10% FBS and allowed to adhere for 6–8 h. After 3 times washing using phosphate-buffered saline (PBS), cells were fixed with 4% paraformaldehyde for 10 min at room temperature, then washed 3 times with ice-cold PBS, permeabilized with 0.1% Triton-X100 in PBS for 10 min, blocked with 1% bovine serum albumin (BSA) and 22.52 mg/mL glycine in PBS supplemented with 0.1% Tween 20 (PBST) at room temperature for 30 min, and finally incubated with anti-COL12A1 antibody (1:50) overnight at 4 °C in the dark. After 3 times washing using PBS, cells were then incubated with fluorescence-labeled secondary antibody (1:200, Abcam, #ab150079) for 1 h in the dark, washed 3 times with PBS, followed by incubation with 0.1 μg/mL 4′,6-diamidino-2-phenylindole (DAPI) at room temperature for 1 min in the dark, and finally rinsed 3 times with PBS in the dark. Coverslips were mounted with UltraCruz® Aqueous Mounting Medium (sc-24941, Santa Cruz) and sealed with nail oil. Finally, cells on the coverslip were microphotographed using a Leica fluorescence microscope (TCS SP8, Germany).
Cell viability assay
We assessed iCCA cell viability using CCK8 assays (Beyotime, Shanghai, CN) as previously described [21]. Briefly, 1 × 103 iCCA cells (per well) were seeded into 96-well plates and incubated for 24 h and 48 h, respectively. Then, CCK8 (1:10) combined with fresh growth medium was added to each well. After incubation for 1 h, the absorbance at 450 nm was tested as previously described [21].
Cell colony-forming assay
1 × 103 iCCA cells in the logarithmic phase were seeded into a 6-cm cell culture dish and incubated for 2 weeks. After 3 times washing with PBS, the cell colonies were stained with 0.05% crystal violet (Sigma-Aldrich, USA) for 30 min, then photographed and counted using ImageJ (version.1.8.0), which were performed at room temperature.
Dual-luciferase reporter assays
The sense and antisense fragments of the human COL12A1 3'UTR including either the wild-type (GOSE0310921) or the mutated sites (GOSE0310922) binding to miR-424-5p were synthesized separately by GeneChem (Shanghai, CN). Then the respective sense and corresponding antisense strands were hybridized and then cloned into the GV272-Firefly-Luciferase vector at the XhoI and KpnI sites. 2 × 104 HEK-293 T cells (per well) were seeded into a 96-well cell culture plate, and co-transfected with either the GV272 vector (Control), GV272-COL12A1-3'UTR (wild-type), or the GV272-COL12A1-Mut-3'UTR, and the miR-424-5p mimic and CV045-TK promoter-Renilla-Luciferase plasmid (GeneChem, Shanghai, CN) using LipofectamineTM3000 transfection reagent (Invitrogen, Thermo Fisher Scientific, USA) following the instructions for the user. After incubation for 48 h at 37 °C in 5% CO2, the luciferase activity of cell lysate was tested using a Dual-Glo luciferase assay kit (E1910, Promega). Luminescence was measured by a SpectraMax i3x Multi-model Microplate reader and the SoftMax Pro7 (Molecular Devices, USA).
Genomic DNA isolation and targeted bisulfite sequencing
The extraction of genomic DNA in fresh-frozen iCCA and the paired nontumor liver tissue samples (n = 10), as well as human iCCA cells, was performed as we previously described [21]. The concentration of each extracted DNA sample was fluorometrically tested using Qubit 4.0 (Invitrogen, Thermo Fisher Scientific). Bisulfite conversion and next-generation sequencing (NGS) library preparation were performed as previously described [41]. According to the miRNA promoter recognition methods described by Annalisa Marsico et al. [42], we identified miR-424 promoter locus (chrX: 133683002–133684302) and designed primers for miR-424 promoter region using MethPrimer (version 2.0.0) [43, 44]. The putative primers for miR-424 promoter region (Additional file 9: Table S3) were validated and synthesized by BGI genomics (Shenzhen, CN). A total of 1 µg genomic DNA was subjected to bisulfite DNA conversion with ZYMO EZ DNA Methylation-Gold Kit (#D5006, Zymo Research, Irvine, CA, USA) accordingly to the manufacturer’s instructions. Then, a total of 500 ng bisulfite-converted DNA sample was used for PCR amplification using KAPA HiFi HotStart Uracil+ ReadyMix PCR Kit (#KK2801, Kapa Biosystems, Wilmington, MA, USA) combined with bisulfite-specific primers according to the standard conditions.
After the quality control using Labchip GX Touch (PerkinElmer, Germany), one-twentieth of the elution PCR reaction products were subjected to second round PCR for ligation of barcoded sequencing adapter with T4 DNA ligase. Thereafter, all samples were quantified on an Agilent 2100 (Agilent, USA) and adjusted for the same copy number for subsequent sequencing, and the sequencing was performed on a NovaSeq6000 system (Illumina, USA) using the NovaSeq 6000 S4 Reagent Kit v1.5 (300 cycles) in collaboration with Yunbios (Shanghai, CN).
The basic statistics on the quality of the raw reads was performed using The FastQC tool (http://www.bioinformatics.babraham.ac.uk/projects/fastq). The adapters and low-quality reads were removed using Trimmomatic v0.36 with default parameter [44]. The alignment of the bisulfite sequence to the human reference genome (hg19) was performed using the BSMAP v2.7.3 with parameters ‘-n 0 -g 0 -v 0.08 -m 50 -× 1000’ [45]. The unique mapped reads were kept for the subsequent analyses and the sequence depths of methylated cytosines were greater than 20X.
Experimental animal model
The animal experiments were performed as previously described [21]. The experimental protocol (1615/22) for all animal experiments was reviewed and approved by the Institutional Animal Care & Use Committee (IACUC). Five-week-old male NSG mice used in the study were purchased from Model Animal Research Centre of Nanjing University, Nanjing, CN.
To determine the direct impact of COL12A1 on iCCA progression, 4 × 106 HuCCT1 cells with or without COL12A1 knockout, resuspended in 200 μL PBS, were subcutaneously implanted into the flanks of 5-week-old male NSG mice. After observations for 8 weeks for tumor formation, mice were killed according to IACUC protocol. The collected tumor lesions were measured and weighted by a digital caliper and an electronic balance, respectively, and then fixed tumor tissue with 4% paraformaldehyde for use in downstream experiments. The tumor volume was calculated based on the formula as previously described [21].
Decitabine drug treatment model: HuCCT1 cells (3 × 106) were implanted into the flanks of 5-week-old male NSG mice. Upon tumor reaching 25 to 50 mm3, mice were randomly divided into two groups with each group including 5 mice. Then, mice were treated with decitabine (2 mg/kg, 5 times weekly) or PBS by tail vein injection.
MiR-424-5p agonist (modified miR-424-5p mimic) administration model: After implantation of 3 × 106 HuCCT1 cells into the flanks of 5-week-old male NSG mice, tumor development was monitored 3 times weekly at 3wks after implantation. Upon tumor reaching 130–200 mm3, mice were randomly divided into AgoMIR (20 nmol, i.v. 2 times weekly) group or control (100ul PBS, i.v. 2 times weekly) group, with each group consisting of 3 mice. AgoMIR (miR40001341-4-5) was purchased from RIBOBIO Co. Ltd. (Guangzhou, CN).
Statistical analysis
All statistical analyses were performed with R v.4.1.0 (R Foundation for Statistical Computing, Vienna, Austria) or with SPSS v.22.0 (IBM Corp., Armonk, NY, USA). Unless otherwise indicated, data are expressed as means ± standard error of the mean (SEM). A comparison of two groups was made using a two-tailed Student’s t test or Wilcoxon test. A comparison of multiple groups was made using a one-way analysis of variance (ANOVA). Correlation analysis between two variables was assessed by Pearson correlation analysis. The association between COL12A1 expression and the clinicopathological features of iCCA patients was assessed by chi-square test or Fisher's exact test. Survival analysis was performed with the Kaplan–Meier method, and the statistical difference was determined by log-rank test. The “surv_cutpoint ()” function in R package survminer (https://cran.r-project.org/web/packages/survminer/index.html) was applied to determine the optimal cutoff point of COL12A1 mRNA/protein expression level corresponding to the survival of iCCA patients. Cox proportional hazards regression model was applied for multivariate survival analysis.
