Back

B7-H4 represents a site-specific immunotherapy target in small bowel gastrointestinal stromal tumor

Singer, H.; Morris, M. T.; Maestro, R.; Paolo Dei Tos, A.; DeMatteo, R. P.; Vitiello, G. A.

2026-08-09 immunology
10.64898/2026.08.04.742601 bioRxiv
Show abstract

Small bowel gastrointestinal stromal tumors (GISTs) are more aggressive than gastric GISTs, yet the biologic basis for this difference remains poorly understood. We hypothesized that differential expression of immune checkpoints contributes to this site-specific behavior. Bulk RNA sequencing of 42 primary GISTs (36 gastric, 6 small bowel) revealed marked upregulation of VTCN1, which encodes the inhibitory checkpoint B7-H4, in small bowel tumors (log2FC = 7.95, adjusted P < 0.001). In contrast, expression of the therapeutically targeted checkpoints PD-L1, PD-1, and CTLA-4 was comparable between sites. Concordantly, B7-H4 enrichment was accompanied by an immunosuppressive tumor microenvironment, characterized by reduced antigen-presenting cells, fewer effector-memory CD8+ T cells, lower granzyme B expression, and suppression of interferon and inflammatory signaling pathways. Notably, the differences in B7-H4 expression were independent of imatinib-treatment status. These findings were corroborated in an external cohort of 77 untreated GISTs, in which VTCN1 was similarly enriched in small bowel tumors. Independent immunohistochemical analysis of a tissue microarray comprising 68 untreated primary GISTs confirmed the pattern, showing median B7-H4 positivity of 78.6% in duodenal, 20.5% in jejunal/ileal, and 0% in gastric tumors, with staining localized to tumor cells rather than stroma. Collectively, these data identify B7-H4 as a site-specific feature of small bowel GISTs and a potential therapeutic target for tumors that have not responded to conventional checkpoint blockade.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.