Datasets
We downloaded datasets corresponding to 50 normal liver and 374 LIHC tissue samples from The Cancer Genome Atlas Program (TCGA; https://cancergenome.nih.gov/). We also downloaded the Liver Cancer-FRANCE (LICA-FR) LIHC dataset from the International Cancer Genome Consortium database (ICGC; https://icgc.org); this dataset contains data pertaining to 150 LIHC tissue samples and 11 paracancerous samples.
Data pre-processing and DEGs identification
The Affy package in R (Gautier et al., 2004) was used to analyse original data. After correcting for the inter-batch difference, original data were subjected to perform background correction, quality control and standardisation processing. The limma package in R (Ritchie et al., 2015) was applied to filter DEGs from the data set. The interception criteria for DEGs in the above database were |log fold change (FC)| ≥1 and p < 0.05 after adjustment. The results were visualised as a Venn diagram.
Gene Ontology (GO) analysis, Kyoto Encyclopaedia of Genes and Genomes (KEGG) analysis, and Gene Set Enrichment Analysis (GSEA) of DEGs
GO is one of the most frequently used tools for gene annotation, often employed for large-scale gene annotation. KEGG was used to identify pathways associated with DEGs. The distribution trends of genes and phenotypes were evaluated using GSEA. GO analysis, KEGG analysis, and GSEA were conducted using the clusterProfiler package in R (Yu et al., 2012).
Construction of protein-protein interaction (PPI) network and hub gene identification
The interaction between proteins encoded by the DEGs was evaluated by STRING tool (Szklarczyk et al., 2015). Then Cytoscape software (v3.7.1) (Smoot et al., 2011) was a platform for visualizing the PPI information. The top 20 genes were selected as hub genes using the cytoHubba plug-in (Chin et al., 2014).
Prognostic validation of the hub genes
Hub genes strongly correlated with prognosis were screened out by univariate and LASSO Cox regression analysis. The Survminer package (https://github.com/kassambara/survminer) and the survival package in R (https://github.com/therneau/surviva) were then used for mapping analysis. The association between genes and clinicopathological characteristics was analysed by logistic regression and receiver operating characteristic curve analyses. The selection criterion was P<0.05.
Expression andverification of hub genes
The Gene Expression Profiling Interactive Analysis database (Tang et al., 2017) was used to analyse hub gene expression and related transforming growth factor β (TGFβ) signalling. The online tool TIMER (Li et al., 2017) was used to explore the correlation between hub genes and tumour-infiltrating immune cells.
Tissue specimens
A total of 42 paraffin-embedded and formalin-fixed LIHC tissue samples were collected from Lanzhou University Second Hospital. All patients had undergone curative resection surgeries. The study protocol was approved by the Institutional Ethics Committee of Lanzhou University Second Hospital. Written informed consent was obtained from all patients. Disease-free survival (DFS) was the period between the day of surgery and the date of recurrence.
Immunohistochemistry (IHC)
Tissues were serially sectioned into 4 μm‒thick sections, heated in an oven at 67 ºC for 20 min, and then dewaxed in alcohol and xylene. After antigen retrieval, the samples were treated with 3% H2O2. Subsequently, the primary antibodies (BUB1, 1:50, Abcam, United States; CD3, CD4, CD8, CD20, CD56, CD68, CD163, ready-to-use, Maixin, Fujian) along with goat globulin (3 mg/mL, Sigma, Aldrich) were used to incubate the sections at 4 ºC overnight. Sections were then probed with anti-mouse/rabbit antibodies (1:100, Solarbio, Beijing) at 37 ºC for 30 min. Finally, 3, 3′-diaminobenzidine (Solarbio, Beijing) was used to stain the sections.
Wherein a score of 0 indicated < 5% positive cells; a score of 1 indicated 6–20% positive cells; a score of 2 indicated 21–50% positive cells; and a score of 3 indicated > 50% positive cells.
Western blotting (WB)
Radioimmunoprecipitation assay buffer supplemented with phenyl methane sulfonyl fluoride was used to lyse tumour cells. After the lysate was centrifuged, the supernatant was collected for further analysis. After the protein concentration was measured, equal amounts of protein were loaded onto 10% SDS-PAGE gels and transferred onto polyvinylidene fluoride membranes. Then, the membranes were blocked using 5% skimmed milk at 37 ºC for 2 h and were incubated with the primary antibodies (anti-SMAD2/SMAD3, anti-BUB1, anti-PSMAD2/PSMAD3, 1:1,000, Abcam; anti-PDL1, 1:1,000, eBioscience, California; anti-GAPDH, 1:1,000, Proteintech, Chicago) at 4 ºC overnight. Then, the anti-mouse/rabbit antibodies (1:3,000, Beyotime, Shanghai) were used to incubate the membranes at 37 ºC for 1h. The protein bands were detected by the chemiluminescence kit (Thermo Fisher, Shanghai).
Transfection
To silence the expression of endogenous BUB1 in human LIHC cell lines, 2.5 µg short hairpin RNA (shRNA) plasmids (Genechem, Shanghai) and 5 μL Lipofectamine 6000 (Beyotime) were mixed in 500 μL DMEM (Gibco, United States) medium for 5 min. After incubation for 5 min, the mixture was added to cells that had reached a confluence of 60 to 70% of 6-well plate. The cells were used for further experiment after transfection for 24 h. shRNA sequences were as follows: shBUB1, forward 5'-CCGG CCTGGGTCAGAGTATAGATATCTCGAGATATCTATACTCTGACCCAGGTTTTTG-3'; reverse, 5'-AATTCAAAAACCTGGGTCAGAGTATAGATATCTCGAGATATCTATACTCTGACCCAGG-3'.
Cell Counting Kit-8 (CCK-8) assay
For the cell counting assay, 1,000 cells (HuH7 and SKP1 cells)/well were cultured in five replicate wells of a 96-well plate. The CCK-8 solution was sequentially added to every well and incubated for 2 h before detected. Absorbance450 was detected by an instrument (TECAN, Infinite M200 Pro, Shanghai) at 12 h, 24 h, 48 h, and 72 h.
Migration assays
For migration assays, 1 × 105 HuH7 and SKP1 cells were independently resuspended in the medium (1% foetal bovine serum) and seeded in the upper chamber of a Transwell chamber (Corning, Shanghai) placed in 24-well culture plates. Five hundred microlitres of medium (10% foetal bovine serum) was added to the bottom chamber. After incubation at 37 ºC for 24 h, a cotton swab was used to remove the non-migrating cells on the membrane located at interface of the upper and lower chambers, and the migrated cells were fixed, stained, counted, and imaged. Six areas from each membrane were imaged from three independent wells.
Statistical analysis
The clusterProfiler package in R (Yu et al., 2012) and GraphPad Prism 8 (GraphPad Software Inc., Santiago) were used for data analysis. Student’s t-tests were used to compare data between groups. Results are expressed as the mean ± standard deviation. Significance was defined as p < 0.05.
Statement in the methods
All methods were performed in accordance with the relevant guidelines and regulations.