Ethics
This study was reviewed and approved by the Ethics Committee of Shanghai Tenth People’s Hospital, and all patients signed written informed consent.
Data Collection
We collected 135 BCa samples confirmed histologically in Shanghai Tenth People’s Hospital between November 2019 and April 2021 according to the following criteria: (1) histologically confirmed BCa; (2) samples were available for RNA-sequence; (3) availability of clinical and follow-up data. The processes of total RNA extraction, paired-end libraries generation and RNA-sequence can be found in our previous publications (16).
RNA-sequence data (FPKM standardized data) for BCa patients were downloaded from The Cancer Genome Atlas (TCGA) (http://www.cancergenome.nih.gov). The microarray data sets GSE48075 and GSE13507 were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/). The transcriptomic data of IMvigor210 trial, which included metastatic urothelial cancer patients treated with atezolizumab (PD-L1 inhibitor), were downloaded from the http://research-pub.gene.com/IMvigor210CoreBiologies. The list of MAGs was extracted from the metabolic pathways Gene set enrichment analysis (GSEA, c2.cp.kegg.v7.1.symbols.gmt). We used the ComBat, a batch effect correcting tool in the sva R package v3.38, to adjust batch effects and combine these data sets.
Unsupervised clustering
We used the ‘ConsensusClusterPlus’ R package v.1.50.0 (Pearson correlation, 50 iterations and 80% resample rate) to divide BCa patients into different clusters based on the expression of MAGs. Gene Set Variation Analysis (GSVA) was performed to quantitatively investigate different biological pathways between clusters using ‘GSVA’ R package v.1.38.2. single-sample gene set enrichment analysis (ssGSEA) was used to evaluate the proportions of 28 TIICs in each BCa sample.
Construction of the metabolic-associated signature
The differentially expressed genes (DEGs, |Log2FC|>1.5 and adjusted P<0.05) between clusters were identified using ‘limma’ R package v3.46. The random forest along with Boruta feature selection was performed using the ‘Boruta’ R package v7.0 to obtain the most important DEGs involved in discriminating clusters, followed by the univariate cox regression. In this way, the selected genes can be divided into two groups based on the hazard ratio (HR) of the univariate cox regression: protective gene (HR<1) and risk gene (HR>1). The metabolic-associated signature was developed using the following formula:
metabolic score = scale ()
Where X is the expression of each selected gene with HR>1, and Y is the expression of each selected gene with HR<1. This formula was used to calculate the risk score of each patient. Then, patients were divided into low- and high-risk groups according to the optimal cutoff value of risk score calculated by the ‘survminer’ R package v0.4.8.
Chemotherapeutic Response Prediction
The chemotherapy response of each sample was predicted based on the Genomics of Drug Sensitivity in Cancer which is the largest public pharmacogenomics database. Two commonly used chemicals Cisplatin and Gemcitabine were selected to evaluated the IC50 of each sample using the ‘pRRophetic’ R package.
Gene mutation in Cancer Cell Line Encyclopedia (CCLE)
The gene expression data and mutation data of 592 carcinoma cell lines were down from CCLE (https://portals.broadinstitute.org/ccle). The risk scores of each carcinoma cell line were calculated using the formula mentioned above. Then, carcinoma cell lines were divided into low- and high-risk groups based on the median value of risk scores. The mutation counts were compared between low- and high-risk groups.
Hub genes screening
We used three algorithms, including support vector machine-based recursive feature elimination (SVM-RFE, ‘e1071’ R package v1.7) algorithm, MeanDecreaseGini-based random forest-feature selection (RFS-FS, ‘randomForest’ R package v4.6) and MeanDecreaseAccuracy-based RFS-FS, to rank genes with respect to the risk group. The top ten genes of each algorithm were put into a Venn online software (http://bioinformatics.psb.ugent.be/webtools/Venn/) to identify the overlapping hub genes. The GSEA (http://www.broadinstitute.org/GSEA) was used to investigate the enriched gene signatures.
Bioinformatics analysis
The Assistant for Clinical Bioinformatics is an online tool (www.aclbi.com) that contains gene expression data and survival information of various cancers in the TCGA database, and all data are analyzed by R v4.0.3. This data set can provide survival analysis, correlation analysis, prognostic nomogram and tumor tissue immune cell infiltration.
Kaplan-Meier Plotter is an online tool (www.kmplot.com) that contains gene expression data and survival information of various cancers for analyzing the prognostic value of TST and S100A16 in BCa. According to the median mRNA expression of BCa samples, they were divided into high expression group and low expression group, and then analyzed by Kaplan-Meier survival curve.
TIMER is an online tool (http://timer.cistrome.org) that contains a comprehensive analysis of immune infiltration in different types of cancer, and is used to analyze the association between TST and S100A16 and immune infiltration in BCa. The associations between genetic markers and various immune cells are also recorded.
GEPIA is an online tool (http://gepia2.cancer-pku.cn) that can provide tumor/normal differential expression analysis, survival analysis, correlation analysis. BCa patients were used to analyze the association between genes based on their mRNA expression.
Immunohistochemistry (IHC)
The Human Protein Atlas (http://www.proteinatlas.org) online database can obtain IHC slides and information to check the expression levels of TST and S100A16 in normal bladder tissues and BCa specimens. The score of the positive staining ratio of tumor cells is defined as: 0 (none); 1 (<25%); 2 (25%-75%); 4 (>75%). The staining intensity score is determined as: 0 (Negative); 1 (Weak); 2 (Moderate) and 3 (Strong). The immune response score is evaluated by multiplying the staining intensity and the positive cell ratio score.
Cell lines and culture
Human HEK293T, human BCa cell line T24 were obtained from the American Type Culture Collection (ATCC, Rockville, USA). Human HEK293T is maintained in DMEM medium (Gibco, USA); T24 cells are cultured in RPMI-1640 (Gibco, USA). All cell lines were cultured in a humidified incubator containing 5% CO 2 at 37°C. All cell culture media contain 10% fetal bovine serum (FBS; Gibco) and 1% penicillin/streptomycin (Hyclone; GE Healthcare Life Sciences, Logan, UT, USA).
Generation of TST and S100A16 knockout cell lines
TST and S100A16 gene knockout T24 cell lines were produced using lenti CRISPR technology (42). In short, the oligonucleotides encoding gRNA (TST sgRNA1: 5'-TCCGGAACTGGCTGAAGGA-3'; TST sgRNA2: 5'-CATCAGGACTGGCAAGCTG-3'; S100A16 sgRNA1: 5'-AGGCCTTACCGACAGCATG-3'; S100A16 sgRNA3: 5'-CAGGACACAGGGAACCGGA-3') was constructed into the lentiviral CRISPR-puro vector. Then, the plasmid was co-transfected into HEK293T cells with packaging vectors including pSPAX2 and pMD2G. After 48 hours, the culture supernatant was harvested to infect the T24 cell line. Puromycin (1μg/ml) was used to select lentivirus-transduced cells for 3 days and verified by Western blot.
Real-time quantitative PCR (RT-qPCR)
According to the manufacturer's instructions, total RNA was extracted from the cells using TRIzol reagent (Invitrogen, USA), and then reverse transcription was performed using HiScript®Q Select RT SuperMix (Vazyme Biotechnology Company, China) to synthesize cDNA samples. RT-qPCR was performed using SYBR qPCR Master Mix (Vazyme Biotech, China), and quantified by CFX real-time PCR detection system (BIO-RAD, USA). The primers used in this study are as follows: The primers were provided in Supplementary Table 1.
Western Blot
The cells were collected and lysed in lysis buffer (150 mM NaCl, 0.5% NP-40, 1 mM EDTA, 10% glycerophosphate, 50 mM Tris-Cl, pH 7.4 and protease inhibitor cocktail). After 30 minutes, the cell lysate was separated by centrifugation at 12,000 rpm 4°C for 15 minutes. The protein concentration was quantified by BCA protein analysis kit (Beyotime, Shanghai, China). Separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE), and analyzed by immunoblotting. The western blot was detected using an electrochemiluminescence (ECL) imaging system (Tanon, Shanghai, China). Anti-TST (ab166625), CD47 (ab218810), PD-L1 (ab136845), E-cadherin (ab1416), N-cadherin (ab12221), Vimentin (ab92547), Slug (ab85936) antibodies were obtained from Abcam; anti-Tubullin (2148S) antibody was obtained from Cell Signaling Technology; anti-S100A16 (11456-1-AP) antibody was obtained from Proteintech.
Flow cytometry
Cells were washed in PBS and stained for cell surface CD47 or PD-L1 on ice for 30 min in PBS plus 2% FACS. After washing in PBS samples were analysed on a BD LSR Fortessa and analysed in FlowJo. APC anti-human CD274 Antibody Obtained from Biolegend, Clone 29E.2A3, Cat# 329708; PE anti-human CD47 Antibody was obtained from Biolegend, Clone: CC2C6, Cat# 323108.
Cell migration assay
According to the manufacturer's instructions, cell migration was analyzed using the Transwell system (Corning, NY, USA). Resuspend the T24 cell line in serum-free medium, and then add 200μl of cell suspension (containing 5×104 cells) to the upper Transwell chamber. The lower chamber contains 600 μl of medium with 10% FBS. After 16 hours of incubation at 37°C, the cells that migrated on the lower surface of the chamber were fixed with 4% paraformaldehyde and stained with crystal violet.
Matrigel droplet invasion assay
5×104 T24 cells were suspended in 10 μl Matrigel, seeded in a 24-well plate in the form of droplets, and the droplets were imaged using a microscope (Olympus Corporation) on day 0 and day 4, as described previously (43, 44). The added medium was changed once a day. The migration of cells extending to the outside of the droplet was measured by the ImageJ software.
Statistical analysis
Statistical analysis was conducted with SPSS 23.0 (IBM Corp.) and R statistical software v3.6.1 (https://www.r-project.org/). The factors between clusters or risk groups were compared using the Chi-square test, the t-test or the Mann-Whitney U test, as appropriate. Survival analyses were performed using the Kaplan-Meier method and Cox proportional hazard models. All tests were 2-tailed, and P values<0.05 were considered statistically significant.