Cell-free RNA changes precede symptomatic radiation pneumonitis in cancer patients receiving thoracic radiotherapy

Cancer therapies are limited by normal tissue toxicity. Radiation pneumonitis (RP) is a dose-limiting toxicity that commonly occurs in cancer patients who receive thoracic radiation therapy. RP is caused by inflammation and increased vascular permeability in the lungs and can lead to severe pulmonary symptoms and sometimes death. Currently, the strongest predictors of RP are radiation dose metrics, such as the volume of normal lung receiving ≥20 Gy (V20) or mean lung dose (Dmean); however, these have only modest predictive accuracy. Therefore, it is currently not possible to accurately identify patients who might benefit from RP-directed therapy prior to developing symptoms and there remains an unmet need for sensitive, non-invasive biomarkers to identify patients at risk for RP before symptoms develop. Plasma cell-free RNA (cfRNA) is a promising analyte that enables non-invasive profiling of gene expression in diverse tissues. Our group recently developed an ultrasensitive method called RARE-Seq to detect low abundance transcriptional signatures in cfRNA (Nesselbush et al. Nature 2025). In this study, we applied RARE-Seq to analyze cfRNA from 160 plasma samples from 56 lung cancer patients, 34 of whom developed RP, collected before, during, and after radiation therapy. We observed enrichment of lung- and airway-specific gene transcripts in cfRNA from patients with symptomatic RP compared to those who received radiation but did not develop RP. Additionally, pre-symptomatic samples from patients who later developed RP were enriched for cfRNA signatures of lung pneumocytes, suggesting early molecular changes preceding clinical symptoms. Lastly, we identified cfRNA signatures that distinguished symptomatic RP from non-RP samples with an AUC of 0.85 and pre-symptomatic RP from non-RP samples with an AUC of 0.77, significantly outperforming the classic dosimetric predictors lung V20 and Dmean (V20 P = 4.9e-5, Dmean P = 7.4e-5, paired DeLong test). Collectively, these findings demonstrate proof of concept that cfRNA profiling can predict radiation-induced toxicities before onset of symptoms, establishing it as a potentially transformative biomarker for improving monitoring and personalized management of patients treated with radiation therapy.