PLACE method
The PPI network was constructed using the HuRI-Union dataset. The PPIs in HuRI were identified by yeast two-hybrid (Y2H) assay or curated literature. For ease of use, we redefined 3 relationships (between any two proteins A and B). Level 1: Proteins A and B in direct contact and interaction—protein A-protein B; level 2: Proteins A and B in indirect contact with an interval of protein X—protein A-protein X-protein B; level 3: Proteins A and B in indirect contact with an interval of two proteins X1 and X2—protein A-protein X1-protein X2-protein B. We calculated the level 1 counts, level 2 counts and level 3 counts for each DEG. Apart from this, we then examined the relationship between each DEG, and Pearson’s coefficient was calculated for all genes. We retained the level 1 counts, level 2 counts and level 3 counts based on correlation values r >0.5 and p<0.05, and the network was visualized with Cytoscape software. The genes were arranged in descending order by the number of level 1, level 2 and level 3 genes.
Data processing
We downloaded GSE149614 scRNA-seq submitted by Yiming Lu et al from the Gene Expression Omnibus database 48. A total of 13736 cells (10672 cancer tissue-derived cells and 3064 paired adjacent noncancerous tissue-derived cells) were selected from the scRNA-seq.
We downloaded TCGA-LIHC-FPKM data from The Cancer Genome Atlas Program. We subsequently converted FPKM values to TPM (transcripts per million) using TPM = [FPKM/FPKMsum]*10^6. We also downloaded survival data from The Cancer Genome Atlas Program (https://xena.ucsc.edu/public/).
We downloaded the HuRI-union dataset submitted by Luck et al. (64006 PPIs involving 9094 proteins were identified) 34.
Single-cell RNA sequencing data analysis: dimensionality reduction and clustering
After preliminary screening of 13736 cells (10672 cancer tissue-derived cells and 3064 paired adjacent noncancerous tissue-derived cells), the expression matrix of cells was processed using R software (Seurat package). Following data normalization (NormalizeData Function) and scaling (ScaleData Function), principal component analysis (PCA) was conducted using genes with highly variable expression. Seurat graph-based clustering was then applied to visualize the identified clusters in tSNE plots (RunTSNE Function).
Single-cell RNA sequencing data analysis: cell type annotation
The cell types were annotated according to the SingleR prediction function and confirmed according to the list of marker genes (Table S1). We visualized the marker genes in clustering plots by the FeaturePlot function.
Single-cell RNA sequencing data analysis: Biomarker genes that showed differential expression between cancer cell-derived hepatocytes and paired adjacent noncancerous cell-derived hepatocytes.
Hepatocytes were selected from the pool of single cells (subset Function). We performed differential gene expression analyses on cancer tissue-derived cells and paired adjacent noncancerous tissue-derived cells. Differentially expressed genes (DEGs) were then identified by differential gene expression analysis. The Wilcoxon test (adjusted P value <0.05) and a loge (FC) greater than 0.25 were used to test for significance 39.
Gene enrichment analysis
With the help of the clusterProfiler package and GSEA dataset, hallmark enrichment and KEGG pathway enrichment were performed using the hallmark gene set (http://www.gsea-msigdb.org/gsea/msigdb/index.jsp) and KEGG database (https://www.genome.jp/kegg/).
Validation using TCGA RNA-seq data
To determine the value of the prognostic gene signature in prognosis at the RNA level, TCGA-LIHC TPM data and survival data were used for validation. Survival was analyzed using Kaplan–Meier survival analysis. Overall survival (OS) and disease-specific survival (DSS) of HCC patients with the gene of interest were assessed and compared between the long-survival and short-survival groups.
Validation in human HCC cell lines
To determine the functions of TMEM14B, we knocked down TMEM14B expression using siRNA in human HCC cell lines (LM3 and HepG2).
siTMEM14B#1: sense5’-3’ GUGCUUACCAGCUGUAUCATT,
siTMEM14B#2 sense5’-3 GCCUGUAGGUUUAAUUGCATT.
Cell counting kit 8 (CCK8) assay
For the Cell Counting Kit-8 (CCK-8) assay, LM3 and HepG2 cells in DMEM containing 10% FBS were seeded into 96-well plates at a concentration of 1 × 104 cells per well and incubated for 24 h, 48 h and 72 h. CCK-8 solution (10 μl/well) was added to the 96-well plates and incubated for 1 h to detect the viability of LM3 and HepG2 cells. The light absorbance values at 450 nm were measured in a microplate reader (Bio-Rad, Hercules, CA, United States), and cell viability was determined.
Wound-Healing Assay
A culture insert (Ibidi, Munich, Germany) was used to generate a wound of 500 μm. The insert was placed on 24-well plates; then, 3 × 105 cells were seeded in each culture insert and incubated for 24 h. After removing the culture insert, the cells were allowed to grow in medium without FBS for 24 h. The original area and migration area were measured using ImageJ software, and the wound closure rates are shown according to the ratio of the migration area to the original area. Each treatment was performed in triplicate wells, and three independent experiments were repeated.
Transwell assay
Transwell migration assays were performed using a 6.5-mm transwell insert with an 8.0-μm pore polycarbonate membrane (Merck Millipore, Burlington, MA, United States). A total of 300 μl of cell suspension containing 3 × 105 cells without FBS was added to the upper chamber, and 800 μl of medium containing 10% FBS was added to the lower chamber. After incubation for 24 h, cells in the lower chamber were fixed with 4% paraformaldehyde for 15 min and stained with crystal violet for 15 min. Images of each chamber were captured randomly for cell counting. Three independent experiments were repeated.