Integrative analysis of semaphorins family genes in colorectal cancer: implications for prognosis and immunotherapy

• Published in: Frontiers in immunology Vol. 16; p. 1536545
• Main Authors: Zhu, Jiahao, Xu, Benjie, Wu, Zhixing, Yu, Zhiwei, Ji, Shengjun, Lian, Jie, Lu, Haibo
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Media S.A 04.03.2025
• Subjects: Biomarkers, Tumor - genetics; Cell Line, Tumor; colorectal cancer; Colorectal Neoplasms - genetics; Colorectal Neoplasms - metabolism; Colorectal Neoplasms - mortality; Colorectal Neoplasms - pathology; Colorectal Neoplasms - therapy; Computational Biology - methods; Gene Expression Profiling; Gene Expression Regulation, Neoplastic; Humans; Immunology; immunotherapy; Immunotherapy - methods; Machine Learning; Prognosis; semaphorins; Semaphorins - genetics; Semaphorins - metabolism; single-cell sequencing; spatial transcriptome sequencing; Transcriptome; single-cell sequencing; colorectal cancer; semaphorins; spatial transcriptome sequencing; immunotherapy
• ISSN: 1664-3224; 1664-3224
• DOI: 10.3389/fimmu.2025.1536545

Semaphorins (SEMAs), originally identified as axon guidance factors, have been found to play crucial roles in tumor growth, invasiveness, neoangiogenesis, and the modulation of immune responses. However, the prognostic value of SEMA-related genes in colorectal cancer (CRC) remains unclear. We applied a novel machine learning framework that incorporated 10 machine learning algorithms and their 101 combinations to construct a SEMAs-related score (SRS). Multi-omics analysis was performed, including single-cell RNA sequencing (scRNA-seq), and spatial transcriptome (ST) to gain a more comprehensive understanding of the SRS. A series of cell experiments were conducted to prove the impact of key genes on CRC biological behavior. A consensus SRS was finally constructed based on a 101-combination machine learning computational framework, demonstrating outstanding performance in predicting overall survival. Moreover, distinct biological functions, mutation burden, immune cell infiltration, and immunotherapy response were observed between the high- and low-SRS groups. scRNA-seq and ST demonstrated unique cellular heterogeneity in CRC. We observed that SRS-high and SRS-low malignant epithelial cells exhibit different biological characteristics. High SRS malignant epithelial cells interact with myeloid and endothelial cells via SPP1 and COL4A2-ITGAV-ITGB8 pathways, respectively. Low SRS cells engage with myeloid and endothelial cells through MIF and JAG1-NOTCH4 pathways. Additionally, knocking down SEMA4C significantly inhibits the proliferation and invasion of CRC cells, while promoting apoptosis . SRS could serve as an effective tool to predict survival and identify potential patients benefiting from immunotherapy in CRC. It also reveals tumor heterogeneity and provides valuable biological insights in CRC.