A Single-Cell Atlas of Radiation-Induced Lung Injury (RILI) at the Early Stage Post-FLASH and Conventional Dose Rate (CONV) Irradiation

Summary

Radiation therapy (RT) is essential for treating thoracic malignancies but often causes significant lung damage. FLASH-RT, an ultra-high dose rate irradiation technique, shows potential in reducing radiation-induced lung injury (RILI) while maintaining tumor control. However, the mechanisms underlying its benefits, particularly immune modulation, remain unclear. This study investigates the immune and cellular responses to FLASH-RT versus conventional dose rate (CONV) RT during the early phase of RILI. Using single-cell RNA sequencing (scRNA-seq), we pictured a dynamic landscape of the lung microenvironment during RILI within one week post-irradiation. Our analysis revealed that FLASH-RT induced a more immediate but transient cellular response, while CONV-RT caused sustained inflammation. FLASH irradiation significantly reduced neutrophil infiltration compared to CONV irradiation, particularly within the pro-inflammatory Ccrl2⁺ subset. FLASH also triggered stronger activation of Cd4⁺ Cd40lg⁺ Th cells, which are critical for regulating immune responses and balancing inflammation. Additionally, FLASH irradiation enhanced TGF-β signaling and epithelial-mesenchymal transition (EMT) in alveolar type 1 (AT1) cells, promoting tissue repair. These findings highlight FLASH-RT’s superior immune modulation and reparative potential, providing valuable insights into its clinical application for minimizing radiation damage and enhancing lung recovery.

Irradiation and dosimetry

CONV and FLASH whole-thorax irradiation was performed at 17.8 Gy using a 6 MeV electron line UHDR vertical test platform of the Department of Engineering Physics, Tsinghua University. An irradiation field of 2.5 cm (craniocaudal) × 6 cm (lateral) was used to cover the whole thorax. The irradiation parameters were listed in Table 1. Ashland’s GafchromicTM EBT3 films (Ashland Inc., Covington, Kentucky, USA) were used for dosimetry for both FLASH and CONV irradiation. Before the radiation experiment, the film was calibrated using 6 MV beam from an uRT-linac 506c linac (United Imaging, Shanghai, China). The films were scanned 24 hours after exposure with an Epson Perfection V850 Pro scanner. Entrance dose for every individual mouse irradiation was recorded by the films placed on the thorax of mice. The average entrance dose in the central 2.5 cm × 6 cm square of the field was recorded as the delivered dose.

Table 1. FLASH irradiation and CONV irradiation parameters

Single-Cell Sequencing Library Construction

Tissues were washed twice with 1× PBS and minced into approximately 1mm3 fragments with sharp scissors and digested with collagenase Ⅱ (sigma V900892, 2 mg/mL) and DNase I (sigma DN25-100MG10 μg/mL) at 37℃ for 1h. Then, the single cell suspensions were collected through a 40 μm strainer (MACS® SmartStrainer 40μm, Miltenyi Biotec, Germany), and centrifuged at 500 g for 5 min, the supernatant was completely aspirated and cells were resuspended with 1ml 1 × PBS containing 0.04% weight/volume BSA. Finally, the cell number and viability were assessed by an automatic cell counter. scRNA-seq libraries were constructed using the 10x Genomics Chromium platform 3′ library and Gel Bead Kit V3.1 according to manufacturer’s instructions, with an estimated 10,000 single cells per sample. Next-generation sequencing was performed on an Illumina Novaseq 6000 sequencer with a sequencing depth of at least 100,000 reads per cell with pair-end 150 bp (PE150) reading strategy.

Quality control, batch correction, and cell clustering

The Cell Ranger pipeline from 10x Genomics was employed to align reads to the mouse genome (mm10), quantify gene expression in single cells, and generate unique molecular identifier (UMI) count matrices for each sample. Further processing of the expression matrices was performed using the Seurat package (V4) in R. During quality-control (QC), genes expressed in fewer than three cells were excluded, and cells were filtered based on the following criteria: (i) ≤ 1,000 or ≥ 60,000 total UMIs, (ii) ≤ 500 or ≥ 8,000 transcribed genes, or (iii) ≥ 10% of UMIs mapped to mitochondrial genes. The expression matrices from different samples were then merged, log-normalized, and scaled in Seurat’s standard workflow. For principal component analysis (PCA), the union of the top 1,000 variable genes from each sample was selected, excluding genes related to ribosomal proteins, immunoglobulins, and the mitochondrial genome to minimize unwanted variation. Batch effects were corrected using the Harmony algorithm. Cell clusters were identified via the shared nearest neighbor (SNN) modularity optimization algorithm, and marker genes for each cluster were determined using the Wilcoxon rank-sum test. Clusters were visualized using Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction.

Additionally, Scrublet was used to detect potential doublets in the dataset. Doublet scores for single cells were calculated for each sample individually using default settings, with thresholds adjusted after visual inspection of the score distributions. Clusters with a doublet rate exceeding 10% were classified as doublet clusters and excluded from further analysis.