M2-LIKE POLARIZED TAMS SIGNATURE DEFINED BY SCRNA: IMPLICATIONS FOR POOR PROGNOSIS IN BLADDER CANCER AND CONSTRUCTION OF A PROGNOSTIC MODEL BASED ON M2-LIKE GENES

Abstract

Introduction

Bladder cancer (BCa) is among the most common malignancies worldwide and overall survival (OS) remains poor despite treatment advances. Tumor-associated macrophages (TAMs) are monocyte-derived cells that invade the BCa tumor microenvironment (TME) and display a range of functional pro-tumor and anti-tumor phenotypes dependent on distinct environmental factors and stimuli. Specifically, TAMs polarized toward M2-like states are associated with angiogenesis, immunosuppression, and metastasis. Despite this knowledge, the impact of M2-like subsets on BCa prognosis remains largely unexplored partly due to nonspecific transcriptomic signatures hindering identification. Single-cell (scRNA) sequencing technology has transformed our understanding of functional diversity at the single-cell level in multiple cancers. In this study, we harnessed scRNA data from bladder tumor patients and characterized a polarized M2-like gene signature to explore the association of M2-like TAMs with OS in BCa. We then constructed a prognostic risk model based on M2-like gene signatures to identify biomarkers predictive of poor prognosis.

Methods

The scRNA data of 15 BCa patients were collected and analyzed (Seurat R package, v5.0.1). Samples were from the NCBI Gene expression Omnibus and collaborators. Seurat objects were created, merged, and normalized as a single dataset (Seurat v5 integration pipeline). We identified the M2-like gene cluster based on canonical markers in the literature. Strict criteria were used to define the M2-like cluster and M2-like associate genes (|logFC|>3.0, p<0.05). EnrichR database was used to functionally characterize the M2-like cluster. Infiltration levels of M2-like gene signatures in individual samples were estimated with single-sample Gene Set Enrichment Analysis. To construct a prognostic risk score based on M2-like TAMs, we used M2-like differentially expressed genes as candidate genes to explore those significantly associated with BCa prognosis. We integrated survival event and OS time with LASSO Cox regression model to find survival-related genes and form the prognostic model and risk score.

Results

The scRNA sequencing data of 15 BCa samples yielded 46,968 cells for further analysis after quality control and batch-effect correction. We identified 34 clusters and further annotated these 34 clusters into three major categories, keeping the M2-like cluster intact for simpler visualization (Figure 1). We selected 79 genes identified from the M2-like TAMs cluster from scRNA and used them as candidate genes to construct the prognostic risk model. LASSO Cox regression obtained nine key prognostic genes that had the highest impact on OS [RGS2 (coef= 0.105918), CCL3L3 (coef= -0.000413), CXCL14 (coef= 0.006135), AMICA1 (coef= -0.084312), SPP1 (coef= 0.007112), CPVL (coef= 0.054197), EMP3 (coef= 0.082911), CD74 (coef= -0.101872), CD52 (coef= -0.007678)]. The prognostic model was constructed from these genes and risk score from this model was significantly better than age, tumor stage, nodal involvement, metastasis, and smoking status for predicting prognosis in BCa patients (p< 0.001) (Figure 2).

Conclusions

In this study, M2-like polarized TAMs signature, as defined by scRNA sequencing, was associated with significantly reduced overall survival in few cohorts, which suggests that M2-like polarization may be associated with poor OS in BCa. Future investigations will focus on validation within a larger cohort to establish M2-like TAMs gene signature as a predictive marker for OS in BCa patients which can guide decision-making and treatment options for BCa patients.