Main Session
Sep
29
PQA 06 - Radiation and Cancer Biology, Health Care Access and Engagement
3096 - Tumor Microenvironment Characterization and Outcome Prediction in Concurrent Radiation Therapy and Immunotherapy Using CODEX
Presenter(s)
Hao Peng, PhD - UT Southwestern Medical Center, Dallas, TX
J. Liu1, H. Zhang1, S. Diegeler2, B. Dawod1, E. Elghonaimy2, A. P. Rodriguez1, N. N. Sanford3, P. Gopal2, M. Wachsmann2, R. D. Timmerman4, H. Peng5, and T. A. Aguilera6; 1UT Southwestern Medical Center, Dallas, TX, 2UT Southwestern, Dallas, TX, 3Harvard Radiation Oncology Program, Boston, MA, 4Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, 5University of Texas Southwestern Medical Center, Dallas, TX, 6Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, TX
Purpose/Objective(s):
This study aims to combine CODEX (co-detection by indexing)-based spatial profiling and magnetic resonance image (MRI)-based radiomics for tumor microenvironment (TME) characterization and treatment response prediction in locally advanced rectal cancer (LARC). The ultimate goal is to develop a framework for identifying both spatial biomarkers and radiomics features that predict response or resistance to combined radiation and immunotherapy, uncovering potential synergies between the two treatments.Materials/Methods:
The study is based on 21 patients from the INNATE trial (NCT04130854) (Arm1: NCRT + anti-CD40, n=12; Arm2: NCRT alone, n=9). The treatment included short-course radiotherapy (SCRT), the CD40 agonist sotigalimab (Sotiga), and FOLFOX chemotherapy. To develop a spatially resolved profile of immune cells with phenotypic information at a single-cell resolution, we conducted multiplexed spatial imaging using the Akoya’s CODEX Fusion 2.0 system. Pairwise cell-cell interactions were quantified using the cKDTree algorithm to identify each cell’s five nearest neighbors, with statistical enrichment or depletion assessed via Z-scores. Spatial clustering between different cell types in tumor regions was analyzed using normalized cross-K (nK) functions implemented with the SpatStat package in R, incorporating Ripley’s isotropic correction to mitigate edge effects. Confidence intervals were derived from random spatial permutations to assess significance relative to complete spatial randomness.Results:
The developed framework was successfully implemented on pre- and post-treatment CODEX results for both Arms. The framework is able to effectively capture changes in cell-cell interactions within the TME, reflecting both anti-tumor and immunosuppressive activities. Differences in CD8? T cell and macrophage clustering patterns are consistent with previous studies, strongly correlating with treatment outcome.Conclusion:
Our approach offers a framework leveraging spatial and interaction-based cellular features with CODEX, making it a valuable tool for investigating the biological evidence of synergistic effects between radiation and immunotherapy, as well as for biologically informed outcome prediction. Future work will aim to expand this methodology to larger datasets and integrate CODEX-based features with standard image-based radiomics.