The B7 family subgroup reflects tumor cell heterogeneity and patient post-operative prognosis in gallbladder cancer.

PURPOSE: To elucidate the biological heterogeneity of gallbladder cancer (GBC) cells and refine post-operative risk stratification by investigating the expression and downstream signaling of the B7-family immune checkpoint molecules CD276 (B7-H3), VTCN1 (B7-H4), and HHLA2 (B7-H7) at single-cell resolution.

METHODS

Single-cell RNA sequencing (scRNA-seq) data from seven primary tumors delineated five epithelial subgroups via non-negative matrix factorization (NMF). The functional association between HHLA2 and RAC1/CDC42-PAK1-Cofilin signaling was validated by lentiviral manipulation, pharmacologic inhibition, and subcutaneous xenografts. A total of 188 surgically treated GBC patients were enrolled for survival modeling. Seven machine-learning survival algorithms were trained on five variables (CD276, VTCN1, HHLA2 expression, tumor size, and differentiation) and compared by C-index, ROC-AUC, and calibration curves.

RESULTS

CD276 + and VTCN1 + epithelial cells displayed pro-proliferative profiles, whereas HHLA2 + cells exhibited high EMT and migration signatures. HHLA2 overexpression led to increased RAC1/CDC42-PAK1-Cofilin signaling activity, enhanced proliferation, invasion, and EMT in vitro, and accelerated tumor growth in vivo. These effects were reversed by RAC1, CDC42, or PAK1 inhibitors, as well as by CFL1 knockdown. NMF classified tumor epithelial cells into five functionally distinct subgroups. A gradient-boosting machine (GBM) model integrating expression of the three B7 molecules with tumor size and differentiation achieved superior discrimination and accurate calibration on both the training and validation sets.

CONCLUSION

B7-family expression delineates biologically distinct GBC subpopulations; HHLA2 promotes EMT via RAC1/CDC42-PAK1-Cofilin signaling. The GBM model incorporating CD276, VTCN1, and HHLA2 expression with tumor size and differentiation demonstrates potential for enhanced post-operative risk stratification, offering promising candidates for biomarker development and targeted therapeutic interventions.