Immune Spatial Organization Predicts Distant Metastasis Risk in Aggressive Localized Prostate Cancer

Area of Science:

• Computational pathology
• Cancer research
• Immunology

Background:

• Computational pathology offers potential for prostate cancer (PCa) risk stratification.
• Existing methods often lack interpretability or focus solely on tumor morphology.

Purpose of the Study:

• To identify an interpretable computational pathology biomarker derived from the immune microenvironment for PCa.
• To assess the association between immune spatial clustering and distant metastasis in PCa patients.

Main Methods:

• Utilized two cohorts (n=490) with M0 PCa and whole-slide images (WSIs) for discovery and validation.
• Employed deep learning (CellViT) for immune cell identification and DBSCAN for spatial clustering.
• Performed immune cell deconvolution (CIBERSORTx) and immune receptor repertoire reconstruction (TRUST4).

Main Results:

• Increased immune spatial clustering was associated with a decreased risk of distant metastasis in high-grade (Gleason 8-10) PCa.
• This association was not observed in lower-grade (Gleason 6-7) PCa.
• High-cluster samples in Gleason 8-10 PCa were enriched for CD8+ T cells, Tregs, and clonal T cell populations.

Conclusions:

• Immune spatial clustering represents a novel, interpretable computational pathology biomarker for PCa.
• This biomarker provides insights into the distinct immune microenvironment of high-grade PCa.
• Findings support the utility of immune-based computational pathology for improved PCa risk assessment.