5.1 Microarray data
OTSCC datasets were retrieved from the National Center of Biotechnology Information (NCBI) Gene Expression Omnibus. Original gene expression profiles were retrieved from GSE9844, GSE13601, GSE31056, GSE75538, GSE30784 and GSE78060 datasets. "Annotate," and "affy" packages in R were used to process the raw data, establish an expression matrix, and match probes to their gene symbols.
The GSE9844 dataset had 12 regular and 26 OTSCC tissues, and the GSE13601 had 26 normal and 31 OTSCC tissues, the GSE31056 had 24 normal and 22 OTSCC tissues, while GSE75538 had 14 normal and 14 OTSCC tissues.
5.2 Screening of DEGs
Linear models for the microarray data (LIMMA) package were used for screening DEGs between OTSCC and normal tissues. We use |logFC|>2/3 as the screening criteria.
5.3 OSCC Gene Expression Data
The TCGA cohort contains FPKM normalized gene expression data for 303 patients with oral squamous cell carcinoma (OSCC) from the "TCGA-HNSC" project. Their equivalent clinical data tables were available from Genome Data Commons (https:portal.gdc.cancer.gov) download. An independent cohort was used for external validation. The validation cohort contained 97 OSCC patients retrieved from the Gene Expression Omnibus (GEO, GSE41613).
For the TCGA dataset, a transformation of the RNA-sequencing data (FPKM values) into transcripts per kilobase million (TPM) values was done, after which they were log2(TPMs + 1) normalized. For the data set from GEO, the RMA algorithm in "Affy" was used to adjust the background and quantile normalization of the raw data. For all cohorts, only patients whose expression and survival data were available were included in the analyses.
5.4 Recognition of Tumor-Infiltrating Immune Cells
In the discovery cohort, ssGSEA was utilized to quantify enrichment levels of 28 immune signatures in each MIBC sample using ssGSEA scores. Immune cell type signatures were retrieved from earlier publications. Relative abundance for 28 immune gene sets in every tumor sample was calculated using this method. Univariate Cox regression analysis was performed for the identification of tumor-infiltrating immune cells with prognostic significance.
5.5 Establishment of the Dynamic Weighted Gene Co-Expression Network (WGCNA)
WGCNA, a systems biology method, is used to describe correlation patterns amongst genes in various microarray samples. It is used to identify clusters (modules) of highly related genes and summarize them using module eigengene or an intramodular hub gene. In addition, it is used for relating modules to one another and to external sample traits (using the eigengene network methodology) and evaluate module membership measures.
In this study, WGCNA matches the top 5000 genes in the variance of TCGA data, and modules were detected by a dynamic tree-cutting algorithm whose minimum module size was 50 genes, scale-free topology threshold was 0.9, and merged with a 0.5 MEDissThres parameter. After associating modules with clinical traits, modules that exhibited the highest Pearson's correlation coefficients were used in subsequent analyses.
5.6 Key Module Enrichment Analysis
To establish the key modules' biological roles, gene information was loaded into Metascape (http://metascape.org) for GO analysis. Pathway enrichment analyses were conducted using: KEGG, Pathway as well as Molecular Signatures Database (MSigDB) Hallmark Gene Sets. Terms with p < 0.05, > 1.5 enrichment factor, and a minimum count of 3 were obtained and clustered based on similarities in their membership (Kappa scores > 0.3). The most fantastic statistically significant term in a collection was selected to represent the cluster. In case > 20 terms for GO or pathway enrichments were established, visualization was done using the top 20 terms.
5.7 Establishment of Protein-Protein Interactions (PPIs) and Detection of Hub Genes
The PPI network was established using the Search Tool for the Retrieval of Interacting Genes (STRING) database. In the STRING database, 0.4 was the interaction score. Based on interaction data from the STRING database, Cytoscape was used for enhancing PPI network legibility. Then we evaluated the network to select the top ten hub genes in the network.
5.8 Cell Culture and Transfection
Human OSCC cell lines CAL-27 and SCC-4 were bought from the ShangCheng Beina ChuangLian Biology Technology Co., Ltd, Shanghai, China. They were cultured in DMEM (Gibco, Beijing, China), with 10% fetal bovine serum (CLARK Bioscience, Australia). Incubation was done at 37°C in an atmosphere of 5% CO2.
siRNA for KIF14 (si-KIF14 Cat.P202011110094, RiboBio, Guangzhou, China) as well as the negative control (si-NC) were provided by RiboBio (Guangzhou, China). They were transfected into CAL-27 and SCC-4 cells using RiboFECT CP Transfection Kit ( Cat.P202011110094, Guangzhou, China) for CCK-8, Flow cytometry, and western blotting.
5.9 Cell functional Experiment
At a density of 2×103 cells/well, the transfected CAL-27 and SCC-4 cells were seeded in a 96- well plate. At 0,24, 48,72 hrs, ten µL of the cell counting kit-8 (CCK-8) solution (Cat.MAO0218, Meilune, Dalian, China) was added to every well followed by four h of incubation at 37°C. Optical densities of cells were assessed at 450 nm by a microplate reader (ThermoFisher Multiskan FC, Shanghai, China).
Flow cytometry (Cat. BL100A, Biosharp, Hefei, China) was followed the manufacturer's instructions previously (18). The apoptosis rate was shown as follow:
$$\text{T}\text{a}\text{i}\text{n} \text{i}\text{n}\text{d}\text{e}\text{x}=\frac{\text{M}\text{e}\text{d}\text{i}\text{a}\text{n} \text{p}\text{o}\text{s}\text{i}\text{t}\text{i}\text{v}\text{e} \text{c}\text{e}\text{l}\text{l}\text{s}-\text{m}\text{e}\text{d}\text{i}\text{a}\text{n} \text{n}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e} \text{c}\text{e}\text{l}\text{l}\text{s}}{2\times \text{r}\text{S}\text{D} \text{n}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}}$$
5.10 Western Blotting
Total proteins were harvested from CAL-27 and SCC-4 cells with RIPA buffer (Cat. G2002, Servicebio, Wuhan, China) and protein inhibitor at the proportion for 100:1. And then, 10-12.5% SDS-PAGE (Cat.PG113, Epizyme Biotech, Shanghai, China) was employed to separate the protein (60 ~ 80 µg) and transferred it to NC membranes (Cat.66485, PALL, USA). 5% non-fat milk (Cat.G5002-100G, Servicebio, Wuhan, China)blocking at room temperature (RT) for two h. After blocking, membrane incubation was done with the antibody (anti-KIF14 (Cat.A10275, Abclonal, China), anti-BAX (Cat.GB11690), anti-Bcl-2 (Cat.WL01556), anti-cleaved-caspase-3 (Cat. WL02117), anti- ITGAMN (Cat.WL01193), anti-LCP2 (Cat.A2567) and anti-β-actin (Cat.WL01372) obtain from Wanleibio, Shenyang, China, at four °C overnight. After that, incubation was done with the secondary antibody (Beyotime, Shanghai, China) for one hour at RT. β-actin was the internal control.
5.11 Statistical Analysis
Maximally Selected Log-rank Statistic was used as classification selection. K-M analyses were based on the "survival" package, while the heatmap was based on the "heatmap" package. A log-rank test conducted survival analysis. Statistical analyses were performed by R software (version 3.6.1). Un-paired two-tailed Student’s T test was employed to analyze the two groups compared. One-way analysis of variance (Tukey’s post hoc) was used to test multiple comparisons. In addition, when we use P to evaluate the result, P < 0.05 indicated statistical significance.