Back

Proton FLASH radiotherapy enhances control of triple-negative breast cancer through STING-IRF3 and CD8+ T-cell immunity

Loap, P.; Paraskevaidis, I.; Cheptea, C.; Kolker, K.; Kim, M.; Setianegara, J.; Singh, K.; Berianu, I. M.; Metz, J.; Greenberg, R. A.; Bilker, W. B.; Berlin, E.; Koumenis, C.; Assenmacher, C.-A.; Diffenderfer, E.; Taunk, N. K.; Verginadis, I. I.

2026-09-22 cancer biology
10.64898/2026.09.19.751623 bioRxiv
Show abstract

FLASH radiotherapy delivers radiation at ultra-high dose rates and has been demonstrated to spare normal tissue compared to standard radiotherapy, but it is not known if dose rate also modifies tumor response. Here we compare a single 13.5 Gy fraction of proton irradiation delivered at FLASH (F-PRT) or Standard (S-PRT) dose rate in immunocompetent C57BL/6 mice bearing EO771 or AT3 triple-negative mammary tumors. At this identical physical dose, F-PRT delays tumor growth more than S-PRT at both heterotopic and orthotopic sites. The effect is largest in EO771, where F-PRT also prolongs tumor-volume endpoint-free survival and reduces the emergence of lung metastases relative to S-PRT. F-PRT induces earlier intratumoral STING expression and IRF3 nuclear translocation, higher type I interferon levels and greater CD8+ T-cell infiltration. CD8+ T-cell depletion or systemic STING inhibition abolishes the F-PRT advantage. Combined with anti-PD-1 and agonistic anti-CD40, both modalities produce durable complete responses that reject contralateral rechallenge, but F-PRT limits tumor progression before response and accelerates regression. FLASH proton radiotherapy not only improves normal-tissue tolerance, but also antitumor immunity, suggesting that ultra-high dose rate could widen the therapeutic window from both sides.

Matching journals

The top 6 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.