Machine learning developed an intratumor heterogeneity signature for predicting clinical outcome and immunotherapy benefit in bladder cancer

Abstract

Background: Bladder cancer is a common malignancy with high invasion and poor clinical outcome. Intratumor heterogeneity (ITH) is linked to cancer progression and metastasis and high ITH can accelerate tumor evolution. Our objective is to develop an ITH-related signature (IRS) for predicting clinical outcome and immunotherapy benefit in bladder cancer.

Methods: Integrative procedure containing ten machine learning methods was applied to develop an IRS with The Cancer Genome Atlas (TCGA), gene series expression (GSE)13507, GSE31684, GSE32984 and GSE48276 datasets. To evaluate the performance of IRS in predicting the immunotherapy benefit, we also used several predicting scores and three immunotherapy datasets, including GSE91061, GSE78220 and IMvigor210.

Results: The predicting model constructed with Enet (alpha =0.2) algorithm had a highest average C-index of 0.69, which was suggested as the optimal IRS. As an independent risk factor for bladder cancer, IRS had a powerful performance in predicting the overall survival (OS) rate of patients, with an area under curve of 1-, 3- and 5-year receiver operating characteristic (ROC) curve being 0.744, 0.791 and 0.816 in TCGA dataset. Bladder cancer patients with low IRS score presented with a higher level of immune-activated cells, cytolytic function and T cell co-stimulation. We also found a lower tumor immune dysfunction and exclusion (TIDE) score, lower immune escape score, higher programmed cell death protein 1 (PD-1) & cytotoxic T-lymphocyte associated protein 4 immunophenoscore, higher tumor mutation burden (TMB) score, higher response rate and better prognosis in bladder cancer with low IRS score. Bladder cancer cases with high IRS score had a higher half maximal inhibitory concentration value of common chemotherapy and targeted therapy regimens.

Conclusions: The current study developed an optimal IRS for bladder cancer patients, which acted as an indicator for predicting prognosis, stratifying risk and guiding treatment for bladder cancer patients. Further analysis should be focused on the exploration the differentially expressed genes (DEGs) and related underlying mechanism mediating the development of bladder cancer in different IRS score group.

Keywords: Bladder cancer; immunotherapy; intratumor heterogeneity (ITH); machine learning; prognostic signature.

Figure 1

The intratumor heterogeneity score of bladder cancer cases. (A) Association between intratumor heterogeneity score and the clinical characters of bladder cancer patients. (B) High intratumor heterogeneity score favors a poor overall survival rate in bladder cancer. (C) Differently expressed genes between high and low intratumor heterogeneity score group. (D) Potential prognostic biomarkers for bladder cancer based on univariate cox analysis. TCGA, The Cancer Genome Atlas; FC, fold change; IRG, intratumor heterogeneity related gene; BLCA, bladder urothelial carcinoma.

Figure 2

Integrative machine learning algorithms developing an intratumor heterogeneity related signature. (A) The C-index of 101 kinds prognostic models developed by 10 machine learning algorithms in TCGA and Gene Expression Omnibus datasets. (B-F) The survival curve of bladder cancer patients with different IRS score and their corresponding ROC curve in TCGA, GSE13507, GSE31684, GSE32984 and GSE48276 cohort. TCGA, The Cancer Genome Atlas; GSE, gene series expression; AUC, area under curve; NA, not available; IRS, intratumor heterogeneity-related signature; ROC, receiver operating characteristic.

Figure 3

The performance of IRS in predicting the clinical outcome of bladder cancer patients. (A) The IRS score of bladder cancer patients in different clinical stage. (B) The C-index of IRS and clinical characters in predicting the clinical outcome of bladder cancer patients in all datasets. (C,D) Risk factors for bladder cancer patients identified with univariate and multivariate cox regression analysis. (E) The C-index of IRS and other signatures that have developed for bladder cancer patients. (F,G) Predictive nomogram and calibration evaluating the overall survival rate of bladder cancer patients. ***, P<0.001. IRS, intratumor heterogeneity-related signature; TCGA, The Cancer Genome Atlas; GSE, gene series expression; OS, overall survival; HR, hazard ratio; CI, confidence interval.

Figure 4

The correlation between IRS and immune infiltration in bladder cancer. (A) Seven state-of-the-art algorithms evaluating correlation between IRS and immune cells in bladder cancer. (B-D) IRS score showed negative correlation with the abundance of CD8⁺ T cell, NK cells and macrophage M1. The level of immune cells (E), immune related functions (F), immune score (G), stromal score (H) and ESTIMAE score (I) in different IRS score group. *, P<0.05; **, P<0.01; ***, P<0.001. NK, natural killer; pDCs, plasmacytoid dendritic cells; aDCs, active dendritic cells; iDCs, inflammatory dendritic cells; TIL, tumor infiltrating lymphocyte; APC, antigen-presenting cell; CCR, C-C motif chemokine receptor; HLA, human leukocyte antigens; MHC, major histocompatibility complex; IFN, interferon; IRS, intratumor heterogeneity-related signature.

Figure 5

IRS acted as an indicator for predicting the immunotherapy benefits in bladder cancer. The PD1 & CTLA4 immunophenoscore (A), TMB score (B), TIDE score (C), immune escape score (D), HLA-related genes set score (E) and immune checkpoints gene set score (F) in bladder cancer patients with different IRS score. The immunotherapy response and overall rate in patients with high and low IRS score in IMvigor210 (G), GSE91061 (H) and GSE78220 (I) datasets. *, P<0.05; **, P<0.01; ***, P<0.001. PD-1, programmed cell death protein 1; CTLA4, cytotoxic T-lymphocyte associated protein 4; TMB, tumor mutation burden; TIDE, tumor immune dysfunction and exclusion; HLA, human leukocyte antigen; PR, partial response; CR, complete response; PD, progressive disease; SD, stable disease; GSE, gene series expression; IRS, intratumor heterogeneity-related signature.

Figure 6

The IC₅₀ value of common drugs in different IRS score group. Bladder cancer patients with low IRS score had a lower IC₅₀ value of common drugs correlated with chemotherapy (A) and targeted therapy (B). IC₅₀, half maximal inhibitory concentration; IRS, intratumor heterogeneity-related signature.

Figure 7

IRS-based function analysis in bladder cancer. (A) The gene set score correlated with cancer hallmarks in different IRS score group in bladder cancer. (B,C) The functional enrichment in different IRS score group in bladder cancer based on gene set enrichment analysis. NOD, nucleotide-binding oligomerization domain; IRS, intratumor heterogeneity-related signature.