Back

Distinct Sarcoma Microenvironments Predict Benefit from Addition of Pembrolizumab to Preoperative Radiotherapy and Surgery in SU2C-SARC032

Testa, S.; Himes, J. E.; Subramanian, A.; Nouth, S. C. L.; Ballman, K. V.; Heise, R. S.; Pierpoint, M.; Nemat-Gorgani, N.; Sears, T. J.; Binkley, M. S.; Kalbasi, A.; Corcoran, D. L.; Hong, A. M.; Brigman, B. E.; Riedel, R. F.; van de Rijn, M.; Mowery, Y. M.; Weinhold, K. J.; Kirsch, D. G.; Moding, E. J.

2025-11-04 oncology
10.1101/2025.11.01.25339299 medRxiv
Show abstract

The addition of pembrolizumab to preoperative radiotherapy (RT) improved disease-free survival (DFS) for patients with stage III undifferentiated pleomorphic sarcoma (UPS) and dedifferentiated/pleomorphic liposarcoma (LPS) in the randomized SU2C-SARC032 trial. To precisely identify patients who benefit from pembrolizumab and RT, we performed comprehensive multi-omics profiling of pre- and post-treatment tumor and blood samples, including bulk RNA-seq, flow cytometry, and cytometry by time of flight. Additionally, we built a single-cell RNA-seq atlas spanning 65,786 cells from UPS and LPS to recover single-cell states in bulk tumor samples using digital cytometry. Two opposing tumor microenvironments (TMEs), immune-cold sarcoma ecotype 1 (SE1) and immune-hot sarcoma immune class E (SIC E), benefited from pembrolizumab. Pembrolizumab combined with RT caused an overall increase in activated CD8+ T cells, CD56low NK cells, and T cell receptor diversity, while diminishing matrix-remodeling stromal cells and sarcoma cells. Our findings identify different mechanisms of response to pembrolizumab in localized, high-risk UPS/LPS and suggest that sarcoma TME signatures may identify patients most likely to benefit from adding pembrolizumab to preoperative RT.

Matching journals

The top 4 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.