Identification and validation of an immune signature associated with EMT and metabolic reprogramming for predicting prognosis and drug response in bladder cancer

Abstract

Background: Epithelial-mesenchymal transition (EMT), one leading reason of the dismal prognosis of bladder cancer (BLCA), is closely associated with tumor invasion and metastasis. We aimed to develop a novel immune-related gene signature based on different EMT and metabolic status to predict the prognosis of BLCA.

Methods: Gene expression and clinical data were obtained from TCGA and GEO databases. Patients were clustered based on EMT and metabolism scores calculated by ssGSEA. The immune-related differentially expressed genes (DEGs) between the two clusters with the most obvious differences were used to construct the signature by LASSO and Cox analysis. Time-dependent receiver operating characteristic (ROC) curves and Kaplan-Meier curves were utilized to evaluate the gene signature in training and validation cohorts. Finally, the function of the signature genes AHNAK and NFATC1 in BLCA cell lines were explored by cytological experiments.

Results: Based on the results of ssGSEA, TCGA patients were divided into three clusters, among which cluster 1 and cluster 3 had completely opposite EMT and metabolic status. Patients in cluster 3 had a significantly worse clinical prognosis than cluster 1. Immune-related DEGs were selected between the two clusters to construct the predictive signature based on 14 genes. High-risk patients had poorer prognosis, lower proportions of CD8⁺ T cells, higher EMT and carbohydrate metabolism, and less sensitivity to chemotherapy and immunotherapy. Overexpression of AHNAK or NFATC1 promoted the proliferation, migration and invasion of T24 and UMUC3 cells. Silencing ANHAK or NFATC1 could effectively inhibit EMT and metabolism in T24 and UMUC3 cells.

Conclusion: The established immune signature may act as a promising model for generating accurate prognosis for patients and predicting their EMT and metabolic status, thus guiding the treatment of BLCA patients.

Keywords: bladder cancer; epithelial-mesenchymal transition; gene signature; metabolic reprogramming; prognosis.

Figure 1

EMT score shows important clinical prognostic value and biological implication in TCGA cohort. (A-E) The correlation between EMT score and grade, clinical stage and AJCC TNM stage. (F) Comparison of overall survival between high and low EMT score groups. (G) Multiple metabolic regulatory pathways were significantly enriched in the high EMT score group. * p<0.05, **** p<0.0001.

Figure 2

Identification of molecular subtypes based on EMT and metabolism scores. (A) The TCGA cohort was divided into three clusters based on cluster analysis of the ssGSEA scores. (B) Overall survival of the three clusters was compared by KM survival analysis. (C-F) Comparison of clinical characteristics between the three clusters. (G) Comparison of the proportion of metastasis and local recurrence after receiving chemotherapy. (H) Comparison of the expression levels of EMT marker proteins. (I) Comparison of the ESTIMATE scores. (J) Comparison of the infiltration of 22 leukocyte types among three clusters. *p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, ns: no significance.

Figure 3

Screening and functional enrichment analysis of immune-related DEGs. (A) Volcano map showed the DEGs between cluster 3 and cluster 1. (B) Venn diagram displayed the amount of immune-related DEGs. (C) Top 8 items of molecular function (MF), cellular component (CC) and biological process (BP) of GO analysis. (D) The result of KEGG pathway analysis.

Figure 4

Construction of the novel immune-related gene signature based on EMT and metabolic status. (A, B) The LASSO regression was used to reduce the dimensionality of survival related genes after univariable Cox survival analysis. (C) Stepwise multivariate Cox regression analysis to construct the 14-gene prognostic signature. (D) The heatmap of the 14-gene expression and the distribution of patient survival status ranked by corresponding RS. (E) The time-dependent ROC curve for predicting 1-, 3-, and 5-year overall survival in TCGA cohort. (F) The KM survival analysis between high and low RS groups. (G-K) The correlation between RS and clusters, clinical stage and AJCC TNM stage. *p<0.05, **p<0.01, ***p<0.001.

Figure 5

Subgroup analysis, external validation and nomogram proved the predictive value of the 14-gene signature. (A-H) The KM survival curves of OS between high and low RS groups in different subgroups. (I-L) The KM survival curves of OS between high and low RS groups and the time-dependent ROC curves of validation cohorts. I and J, GSE31684. K and L, GSE32894. (M) The nomogram created by integrating clinical information and Riskscore. (N) Calibration curve of the nomogram. (O) Time-dependent AUC curves of different prognostic models.

Figure 6

The immune infiltration, metabolic status and pathway enrichment between high and low risk groups in TCGA cohort. (A) Comparison of 22 types of immune cell infiltration. (B) Comparison of various metabolic scores of ssGSEA result. (C) Multiple malignant regulatory pathways were significantly enriched in the high risk group. (D) The KM survival curve of OS between high and low TMB groups. (E) Comparison of the TMB value between risk groups. *p<0.05, ***p<0.001.

Figure 7

Prognostic value of the gene signature to chemotherapy and immunotherapy. (A) Comparison of RS among patients with complete response and progressive disease after chemotherapy. (B) The KM survival curve of OS between high and low RS groups in patients after chemotherapy. (C) Distribution of TIDE value in TCGA cohort. (D) Comparison of the proportion of responders and non-responders in different RS groups of TCGA cohort. (E) The association between IPS and RS groups. (F) The KM survival curve of OS between high and low RS groups in the IMvigor210 cohort. (G) Comparison of outcomes after receiving immunotherapy in different RS groups in the IMgivor210 cohort. ****p<0.0001.

Figure 8

Screening of the key model genes for further experimental validation. (A) Mutation frequencies of the 14 genes in the TCGA cohort. (B) The correlation between the 14 genes and clinical stage of TCGA BLCA patients. (C) The results of univariate Cox and KM survival analyses for the 14 genes. (D) Expression levels of AHNAK and NFATC1 in cancerous and normal tissues. N, normal; T, tumor. *p<0.05, **p<0.01, ***p<0.001.

Figure 9

qPCR (A) and WB (B, C) were used to detect the effects of AHNAK and NFATC1 knockdown with siRNAs on key glycolysis enzymes, amino acid metabolism enzymes, EMT and PD-L1 immune checkpoints of two bladder cancer cell lines T24 and UMUC3. Glycolysis enzymes involve PFKFB3 and LDHA. Glutamine metabolic enzymes involve GLS and GLUD1. EMT involves E-cadherin and vimentin. β-actin and α-actin were used as internal references. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. ns, no significance.

Figure 10

AHNAK and NFATC1 increases cell migration, invasion and proliferation in BLCA. (A, B) Transwell migration and invasion assays of T24 and UMUC3 cells transfected with siRNAs against AHNAK and NFATC1 (100× magnification). (C, D) MTT assay of T24 and UMUC3 cells transfected with siRNAs against AHNAK and NFATC1. **p<0.01, ***p<0.001, ****p<0.0001.