Data acquisition
The bulk RNA-seq data download from the TCGA, ICGC and GEO database. The scRNA-seq data came from the GSE149614, which generated > 70,000 single-cell transcriptomes for 10 HCC patients from four relevant sites: primary tumor, portal vein tumor thrombus (PVTT), metastatic lymph node (MLN) and normal tissue.
The HCC cDNA microarry were purchased from Shanghai Outdo Biotech Co.,Ltd (Shanghai, China) with 87 samples (including 21 normal liver samples and 66 HCC samples, Ethics No.SHYJS-CP-1707015).
The matrix of HMMR expression were integrated in table S1 and the clinical characteristics of these data were integrated in table S2.
Data Processing
In order to make bulk RNA-seq data from different sources have good consistency, we conducted normality test on bulk RNA-seq data, and performed log2(x + 1) transformation on data that did not obey the normality test. The 10× scRNA-seq data were processed according to the following steps: 1) R software, “Seurat” package was adopted to convert 10× scRNA-seq data as a Seurat object; 2) quality control (QC) of the raw counts by calculating the percentage of mitochondrial (< 5%) or ribosomal genes (> 50) and excluding low-quality cells; 3) the “FindVariableFeatures” function was adopted to filter the top 1500 highly variable genes after QC; 4) principal component analysis (PCA) was performed based on the 1500 genes, and t-distributed stochastic neighbor embedding (t-SNE) was used for dimensionality reduction and cluster identification; 5) the “FindAllMarkers” function was exploited to identify significant marker genes; and 6) The cluster annotation was based the markers of different cell types which download from CellMarker 2.0 website (http://bio-bigdata.hrbmu.edu.cn/CellMarker/index.html).
InferCNV analysis
We isolated hepatocytes and construct a new gene-cell matrix. Somatic large-scale chromosomal CNV score of each hepatocytes were calculated using the R package inferCNV (v1.6.0). A raw counts matrix, annotation file, and gene/chromosome position file were prepared according to data requirements (https://github.com/broadinstitute/inferCNV). The hepatocytes came form normal tissue were selected as reference normal cells. The default parameters were applied (cutoff = 0; denoise = 0.1).
Cell-to-cell interaction analysis
The cell-to-cell interaction analysis was based on the expression of specifific ligands (Ls) and receptors (Rs). In current study, we used the R package “CellChat”, a tool that is able to quantitatively infer and analyze intercellular communication networks from single-cell RNA-sequencing (scRNA-seq) data. Through manifold learning and quantitative contrasts, CellChat classifies signaling pathways and delineates conserved and context-specific pathways across different datasets.
Functional Analysis
Gene set variation analysis (GSVA) was completed to estimate the biological functions and signaling pathways in bulk RNA-seq and scRNA-seq. The reference molecular signature was “h.all.v2023.1.Hs.symbols” (downloaded from https://www.gsea-msigdb.org/gsea/msigdb/).
Tumor microenvironment analysis
CIBERSORTx was utilized to measure the per sample levels of tumor-infltrating immune cell types. CIBERSORTx is an analytical tool to impute gene expression profiles and estimate the abundances of member cell types in a mixed cell population using gene expression data(https://cibersortx.stanford.edu/).
Genomic heterogeneity analysis
Mutation data were downloaded from the TCGA database, and the quantity and quality of gene mutations were analyzed by using the “Maftools” package of R. The tumor mutation burden (TMB) was defined as the total number of somatic mutations. The aneuploidy scores, cancer-testis antigen (CTA) scores and homologous recombination defificiency (HRD) scores were downloaded from the supplementary material of a previous publication.
Recursive Partitioning Analysis
Recursive Partitioning Analysis (RPA) is one of the most recognized methods for cancer prognosis staging. We performed RPA in the web of autoRPA (http://rpa.renlab.org/index.html)[18]. Using a permutation test, autoRPA can evaluate the contribution of each submitted factor and help singling out factors that significantly contributed to cancer staging.
Drug sensitivity analysis
The sensitivity of each drugs was evaluated by IC50 calculation using the “pRRophetic” package, and the corresponding data were obtained from the Genomics of Drug Sensitivity in Cancer (GDSC) database.
Quantitative Reverse Transcription PCR
Relative quantitation was determined by quantitative reverse transcription polymerase chain reaction (qRT-PCR; SuperScript IV Reverse Transcriptase 18090010; Thermo Fisher, United States). The amplifification reactions were performed as described previously (Bustin and Mueller, 2005). HMMR-specifific primers were: forward primer, 5′-TGACCAGGACTAATGAA-3′ and reverse primer, 5′-AGACTCCTTTGGGTGAC-3’.
Cell culture
Human HCC cells of SNU-449, SMMC7721, HepG2, Huh7, LM3, H22, and Hepa 1–6 were purchased from the American Type Culture Collection (ATCC). Cells of SUN-449, HepG2, Huh7, and LM3were cultured with DMEM medium (Gibco, USA), and SMMC7721, H22, and Hepa 1–6 were cultured with 1640 medium (Gibco, USA), both of which were supplemented by 10% fetal bovine serum (FBS) (Gibco, USA). All cells were cultured in an incubator at 37℃ with 5% CO2.
Cellular transfection
Lipofectamine 3000 (Invitrogen, USA) was used to transfect cells plated in 6-well plates with an siRNA specific for HMMR or a control construct purchased from GeneChem (Shanghai, China). Cells were utilized for downstream assays at 48 h post-transfection. Analyses were conducted in triplicate. siRNA for HMMR was customized from Zaigene (Fuzhou, China).
Edu assay
According to the super Proliferation Kit (RiboBio, China), cells were first seeded into 24-well plates at a density of 5 ×104/well and cultured for 24h. Then, cells were fixed with 4% paraformaldehyde after 2h incubation with 5-ethynyl-2′- deoxyuridine (EdU). EdU cells were counted under an Olympus FSX100 microscope (Olympus, Japan) to determine the cell proliferation.
Colony formation assay
Firstly, 500 cells/well were plated into 6-well plates and cultured for ~ 10 days. When the colony reached a sufficient size, it was gently washed with PBS, fixed by formalin, and stained with 0.1% crystal violet. Stained colonies were imaged and counted via ImageJ software (version 2.0.0) to evaluate the cloning efficiency.
Wound Healing Assay
Cells from each group were plated into 6-well plates at around 95% confluence. Then, we used a 200ul pipette tip to make symmetrical wounds. After being washed by PBS twice, cells were incubated with a non-serum medium for 24h (or 48h). Migration pictures were taken at 0h, 24h and 48h after drawing the wound. The wound distance of each group at 40x magnification was measured by Image J software. Each experiment was performed in triplicate.
Western Blot Analysis
Antibodies against HMMR (1:1000, DF4809) were purchaed from Affinity. Briefly, cells were lysed by RIPA buffer with protease and phosphatase inhibitor cocktail following the manufacturer’s specification, and then the concentrations were measured and normalized by BCA assay. Western blotting was performed according to the standard methods as depicted in the manufacturer’s specification and previous studies.
Cell Cycle Detection
The cell cycle was detected by flow cytometry (FCM). Briefly, Cells at the logarithmic growth phase were stained with propidium iodide (PI) according to the manufacturer’s protocol and then detected by the flow Cytometer (Accuri C6 Plus; BD Pharmingen, Shanghai, China) and analyzed by FlowJo-V10 software (Tree Star Inc, Ashland, OR, USA).
Statistical Analyses
Distributed data were compared by performing the Student’s t-test and Wilcoxon test, whereas proportion differences were calculated by the chi-square test. Additionally, component analysis in subgroups were compared by the Fisher’s test. While survival differences between different groups were assessed via the log-rank test, prognostic factors were identifified by the Cox regression analyses. All statistical analyses were performed using RStudio version 4.0.3, and two-sided p < 0.05 was considered as statistically signifificant.