Single-cell and bulk transcriptome analysis identifies B-cell subpopulations and associated cancer subtypes with distinct clinical and molecular characteristics

Abstract

Backgroud: Previous studies have identified B cell subpopulations with pro- and anti-tumoral activities, while the clinical relevance of B cell subpopulations-specific markers in pan-cancer remains understudied. Methods: We integrated 14 scRNA-seq datasets (102,504 cells from 424 patients, 15 cancer types) to identify B cell subpopulations via unsupervised clustering. We characterized their functional dynamics and prognostic relevance through analyzing single-cell, bulk and spatial transcriptomic data. Moreover, using B cell subpopulations-specific gene signatures, we constructed models for predicting cancer prognosis and immunotherapy response. Results: We identified eight B cell subpopulations (b00-b07) which were classified into naive, plasma, memory, germinal center (GC), and cycling B cells. Trajectory analysis revealed b02-naive and b04-GC cells in early phases, evolving into b01- and b03-plasma/b05- and b06-memory/b07-cycling and b05-memory subpopulations. Anti-tumor responses were activated in early pseudotime, complement/immunoglobulin pathways peaked in mid-pseudotime, and energy metabolism increased in late-pseudotime. The enrichment of b07-cycling and b04-GC was negatively correlated with cancer prognosis, while b02-naive had a positive correlation. Spatial transcriptomic analysis showed clustered b00-b06 versus dispersed b07 cells, with b04-GC and b07-cycling cells distant from tertiary lymphoid structure cores. Based on the expression profiles of 1,047 B cell subpopulations-specific signatures, we identified three pan-cancer subtypes with distinct clinical and molecular characteristics. Using 13 B cell subpopulations-specific signatures, we constructed models to accurately predict cancer survival outcomes and immunotherapy response. Conclusions: Our study delineates eight B cell subpopulations with distinct prognostic relevance. Signature-based stratification and models underscore their clinical relevance in cancer outcomes and therapy response, advancing understanding of B cell heterogeneity in cancer.