Gene expression datasets were downloaded from the Gene Expression Omnibus (GEO) database. For radiosensitivity analysis, GSE20549 (6 resistant vs. 6 sensitive), GSE185698 (3 vs. 3), and GSE197236 (3 vs. 3) were utilized. For RIBE analysis, GSE8993 (4 control vs. 12 bystander), GSE18760 (4 vs. 4), and GSE21059 (4 vs. 4) were employed. Differential expression analysis was performed in R using the limma package with a stringent threshold of |logFC| >0.5 and p-value <0.05. The RobustRankAggreg (RRA) method was used to integrate ranked lists and identify robust intersection genes. The protein–protein interaction (PPI) network was constructed using the STRING database (confidence > 0.4) and visualized with Cytoscape (v3.8.1). The MCODE plugin was used to identify significant modules (settings: Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 2, Max. Depth = 100).

The BMSC line was from ScienCell Corporation (USA, Item No. 7500), and both the human lung adenocarcinoma cell line A549 and the human non-small cell lung cancer cell line H1299 were provided by the Provincial Key Laboratory of Research on Major Diseases Molecular Medicine and Prevention of Traditional Chinese Medicine in Universities and Colleges of Gansu Province. The A549 cells were cultured in F12 medium (Gibco), to which 10% bovine serum (Gibco (Thermo Fisher Scientific), Grand Island, New York, USA) and 1% penicillin–streptomycin mixture (Solarbio, Beijing, China) were added. Cells were cultured in 95% air and 5% CO₂, and the temperature was kept at 37°C.

The irradiation was provided by the Heavy Ion Accelerator Cooled Storage Ring (HIRFL-CSR) at the Institute of Modern Physics, Chinese Academy of Sciences, with a dose rate of 2Gy/min (250 keV, 13.3 mA).

A schematic of the co-culture model is provided in Figure 1A. A549 cells were seeded in the upper chamber of a Transwell insert at 2.4 × 10⁴ cells/mL, while BMSCs were seeded in the lower chamber at 5 × 10⁴ cells/mL. After adherence, only the upper chamber containing A549 cells was irradiated with 2 Gy of X-ray at a dose rate of 2 Gy/min. Control groups included a co-culture without irradiation (CO) and monocultures of either A549 or BMSCs with and without irradiation to distinguish direct effects from bystander effects.

IFT88 small interfering RNA (siRNA) (sequence: 5′-GCCGAAGTTCTTTACCAGATA-3′) and a scrambled siRNA (used as a negative control, NC) were transfected into BMSCs using Lipofectamine 3000 according to the manufacturer’s protocol. Transfection efficiency was confirmed by Western blot after 48–72 h.

Cells in the logarithmic growth phase were inoculated into 6-well culture plates at a density of 5 × 10⁵ cells per well, added with antibiotic-free medium, and cultured in an incubator containing 5% CO₂ concentration for 24 h. Transfection was carried out when the degree of fusion reached 80%–90%. Instead of the culture medium, an Opti-MEM serum-free medium was used. Three microliters of transfection reagent was slowly added to 100 μL of Opti-MEM serum-free medium containing 1 μg of plasmid and incubated for 5–10 min at room temperature before being added dropwise to the 6-well plate and then gently mixed. After 24–72 h, the cells were observed under a microscope, and the transfection efficiency was detected by Western blot. The siRNA sequence was as follows (5′ to 3′): GCCGAAGTTCTTTACCAGATA.

The 18-mm cell crawls were pre-inoculated into 12-well plates at a density of 5 × 10⁴ cells per well. Preparation of fluorescent samples involved paraformaldehyde fixation for 10 min, 0.1% Triton X-100 punching for 10 min, and blocking with 5% goat serum on a shaker for 2 h. Anti-Arl13b (17711-1-AP; Proteintech, Wuhan, China) and anti-β-tublin antibodies (ab137822; Abcam, Cambridge, United Kingdom) were used to detect primary cilia, and anti-γ-H2AX (ab26350; Abcam, Cambridge, United Kingdom) and anti-53BP1 antibodies (ab36823; Abcam, Cambridge, United Kingdom) were used to detect DNA damage. Images were captured under an RVL-100-G (ECHO) fluorescence microscope. Ten fields were randomly selected in at least three independent experiments, and the number of ciliated cells, total cell number, and the number of lesions were counted. Mean fluorescence intensity was analyzed using ImageJ software (Version 1.52; National Institutes of Health (NIH), Bethesda, Maryland, USA).

Total protein extraction of stem cells after co-culture radiation was performed, and 10% polyacrylamide gel was configured. An equal amount of samples per well was uploaded, and for the electrophoretic separation of protein, a stacking gel at 80 V for 30 min and a separating gel at 100 V for 90 min were used. The proteins were transferred to a PVDF membrane by the wet transfer method and blocked with 5% skimmed milk powder at room temperature for 2 h. Anti-53BP1 was incubated with the primary antibody (ab36823) or anti-GAPDH antibody (ab8245; Abcam, Cambridge, United Kingdom) overnight at 4°C or at room temperature for 3 h. The following day, the proteins were incubated with goat-anti-rabbit HRP-coupled secondary antibody (ab97051; Abcam, Cambridge, United Kingdom) or goat-anti-mouse HRP-coupled secondary antibody (ab97023; Abcam, Cambridge, United Kingdom) incubated at room temperature for 1.5 h. Imaging was developed with ECL luminol. The gray value was measured using ImageJ software, and the ratio of the gray value of the target protein bands of each group of samples to that of the internal reference bands was calculated as the relative expression level of the target protein. All antibodies used in this study are listed in Supplementary Table 1.

The BMSC cells of each group were made into a single-cell suspension, washed once with pre-cooled PBS, and centrifuged at 800 rpm for 5 min to collect the cell precipitate. A total of 300 μL of 1× binding buffer was added and the cells were resuspended by flicking and then filtered through a 300-mesh sieve into flow sampling tubes. Ten microliters of FITC-labeled Annexin-V solution and 5 μL of PI reagent were added to each tube and incubated for 5 min in the dark. A 200-μL 1× binding buffer was added, and the cells were sampled on a flow cytometer within 1 h for detection.

Cells in the logarithmic growth phase were collected and prepared as a single-cell suspension, inoculated into 96-well plates (5 × 10³ per well, 100 μL), and incubated at 37°C with 5% CO₂ until the cells were attached to the wall (a cell counter was used to measure the concentration of the cells before spreading the plates). Ten microliters of CCK-8 solution was added at 12, 24, 48, 72, 96, and 120 h, and the plates were further incubated for 4 h before termination. The absorbance (OD) value of each group of cells was detected at 450 nm with a full-wavelength enzyme labeler.

A diluent standard (1.0 mL) was added to the lyophilized standard and was allowed to stand for 15 min until it was fully dissolved and mixed, after which it was then diluted according to a certain concentration gradient. Then, it was sealed with a sealing film and was incubated for 30 min, washed five times with a washing solution and shaken dry, and 50 μL of enzyme reagent was added into each well, after which it was sealed again with a sealing film and was incubated at 37°C for 30 min. Each well was filled with a washing solution and left to stand for 30 s and then discarded, and this was repeated five times. Reagent A (50 μL) and reagent B (50 μL) were added to each well, shaken gently and mixed well, and color was developed for approximately 10 min at room temperature in the dark. A termination solution (50 μL) was added to each well to terminate the reaction. Zeroing was performed with a blank well, and the OD value was measured at 450 nm on an enzyme meter.

All data were analyzed using SPSS 23.0 software, and the results were expressed as mean ± standard deviation (± s). A t-test was used for two groups of data that conformed to normal distribution and for chi-square analysis, while one-way ANOVA was used for three or more groups of data. p < 0.05 indicated that the difference was statistically significant, and p < 0.05 indicated that the difference was significant.

A co-culture model of A549 cells and BMSCs was established using a Transwell system (Figure 1A). The results showed that radiation could induce injury to adjacent BMSCs. The number of BMSCs in the non-irradiated group was significantly lower than that in the irradiated A549 cells (p < 0.05) (Figure 1A). The expression of γ-H2AX was detected by immunofluorescence staining, and the DNA damage sites of BMSCs were significantly increased after co-culture. To more directly observe the number and status of cilia of adjacent BMSCs, we performed fluorescence staining experiments of cilia at 48 h of CO culture, and the results showed that the cilium-associated protein was elevated in the CO+IR group compared with the CO group, with abnormally increased cilia (p < 0.05) (Figure 1C).

To systematically investigate the common molecular basis of radiosensitivity and the RIBE in NSCLC, we first conducted a bioinformatics analysis. Gene expression datasets related to NSCLC radiosensitivity and RIBE were obtained from the GEO database.

For the radiosensitivity analysis, three datasets were included: GSE20549 (radioresistant vs. radiosensitive, n = 6 each), GSE185698 (n = 3 each), and GSE197236 (n = 3 each). Differential expression analysis was performed using the limma package (criteria: |logFC| > 0.5, p < 0.05), identifying 6231, 686, and 689 differentially expressed genes (DEGs), respectively (Figure 2A, the upper panel shows representative heatmaps and volcano plots, e.g., for GSE20549). To obtain robust common gene signatures, the RobustRankAggreg (RRA) method was employed to integrate and rank the results from the three datasets, yielding 1,606 intersection DEGs associated with radiosensitivity (Figure 2A).

For the RIBE analysis, three datasets were also included: GSE8993 (control vs. bystander, n = 4 vs. 12), GSE18760 (n = 4 vs. 4), and GSE21059 (n = 4 vs. 4). Differential analysis (same criteria) identified 717, 787, and 1,119 DEGs, respectively (Figure 2A). Integration via RRA yielded 1,433 intersection DEGs associated with RIBE (Figure 2A).

To further focus on key genes potentially involved in regulating both radiosensitivity and RIBE, the two sets of intersection DEGs (1,606 and 1,433) were overlapped, resulting in 136 common genes (Figure 2B). To decipher the PPI relationships among these 136 genes, they were imported into the STRING database to construct a PPI network (confidence > 0.4), which was visualized using Cytoscape. The preliminary network comprised 52 nodes and 163 interaction edges. Subsequently, network topology analysis was performed using the MCODE plugin to identify the most densely connected functional module. Parameters were set as follows: Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 2, and Max Depth = 100. This analysis identified a highly interconnected subnetwork consisting of 14 core genes (Figure 2C). This subnetwork contained 74 edges, suggesting close functional collaboration among these genes. The 14 core genes were as follows: IEL3, BIRC3, BMP2, CXCL5, CXCL2, SPP1, LIF, CCL20, FOS, MMP1, PTGS2, CXCL8, IL1B, and IFT88. Notably, IFT88 encodes a key protein involved in intraflagellar transport and primary cilium formation. The specific fold changes (logFC) for these 14 core genes are provided in Supplementary Table S1.

Based on this analysis, we hypothesized that the primary cilium might play a significant role in RIBE signal transduction and cellular radiation response. To test this hypothesis and elucidate the function of IFT88, we proceeded with the following experiments.

The results showed that co-culture with irradiated A549 cells increased the expression of cilium and cilium-associated proteins, and the expression of IFT88 and ARL13B in the CO+IR group was significantly higher than that in the CO group at 24 and 48 h, and protein elevation was more pronounced at 48 h (p < 0.05) (Figure 2D; Ps1 A-B). The results showed that, compared with non-irradiation, the expression of TGF-βR1 and p-Smad3 in BMSCs increased at 24 and 48 h after irradiation, while the expression of RAD51 increased significantly at 48 h after irradiation (p < 0.05) (Figure 2E; Ps1C-D). Next, we knocked down IFT88, an siRNA knockdown of IFT88 (si-IFT88), for the elimination of primary cilia. Compared with the CO+IR group, the expression of TGF-βR1, p-Smad3, and RAD51 in the adjacent BMSCs was significantly decreased by Western blot (WB). Compared with the SI group, the expression of RAD51 in the CO+IR+Si group was significantly decreased (p < 0.05) (Figure 2F; Ps1E-F). To validate the critical role of primary cilia in the repair of DNA damage by radiation-induced paracentric BMSCs, colocalization of TGF-βR1 with Arl13b in paracentric BMSCs was observed using immunofluorescence staining after 48 h of co-culture (p < 0.05) (Figure 2G). Furthermore, qPCR analysis confirmed that radiation also upregulates IFT88 and ARL13B mRNA levels in BMSCs (Figure Ps4).

At present, molecular docking has become an effective way for computer-aided screening of pharmacodynamic molecules. In order to find effective small molecules of A. membranaceus, the two targets TGF-βR1 and IFT88 were predicted by molecular docking and other network information technology. Previous studies by our research group have screened the small molecule vanillic acid for TGF-βR1 docking and determined the optimal concentration for the intervention of paracellular stem cells after irradiation (18). In order to screen the best small molecules to bind to IFT88, a series of explorations was carried out in this chapter. First, we predicted the protein structure of IFT88 by homology modeling, and then 87 kinds of small molecules of A. membranaceus were linked to IFT88. Some small molecules with high scores were screened out, and these were isorhamnetin, rhamnoside, Alpinia officinalis, isorhamnetin, kaempferol, coumarin, calycosin isoflavone, larch resinol, 3-hydroxy-9,10-dimethoxy rosewood, trans-isoferulic acid, and vanillic acid. Combined with the rest of the pre-group screen for Astragalus small molecules for stem cells, six kinds of ideal small molecules were selected, which were vanillic acid (VA), isoglycyrrhizin (II), onosanthin (On), daidzein (Da), isosanthin (Im), and rosewood (Me). The molecular docking binding energies of IFT88 with the six small molecules are presented in Supplementary Table S3.

Using these six small molecules to interfere with paracellular stem cells, WB detected the primary cilium-associated proteins, and the results showed that the protein expression levels of IFT88 and ARL13B in paracellular stem cells after irradiation could be significantly reduced by zetalane (p < 0.05) (Figure 3A; Ps 1G); therefore, this subject selected rosewood and vanillic acid as IFT88 and TGF-βR1 double targets of small molecules for follow-up experiments (Figure 3B).

In conclusion, we selected vanillic acid (25 μmol/L) and 3-hydroxy-9,10-dimethoxy rosewood (6.25 μmol/L) from A. membranaceus small molecules as drugs for the treatment of radiation side effects and carried out follow-up experiments.

Based on the above-screened small molecules targeting the two key targets, we next explored the effect of small molecules alone or in combination on the paradiotropic effect of BMSCs after radiotherapy for lung cancer. It was found that the two small molecules could effectively improve the cell state of irradiated BMSCs (Figure 3C) and reduce the apoptosis and DNA damage of BMSCs (Figures 3D, E; Ps 1H). Then, we used WB experiment and immunofluorescence experiment to find the small molecules that can reduce the effect of primary cilia, especially after superposition of the drugs, and compared with the single use of small molecules, the number of primary cilia in stem cells was significantly reduced after irradiation (Figure 3F; Ps1 I).

Based on the above-selected small molecules for two key targets, we used WB experiments to find the small molecules that can reduce related proteins of the primary cilium/TGF-βR1 after 24/48 h. Especially after superimposed administration, the primary cilium protein/TGF-βR1 of radiation stem cells was significantly reduced compared with the small molecule alone (Figures 4A, B; Ps 2A-B). We found that compared with no drug, both small molecules, either alone or superimposed, inhibited the TGF-βR1 pathway-related protein and the DNA damage repair protein RAD51, with the best effect. In order to determine the connection between the selected small molecules and the selected key targets, we also conducted molecular dynamics simulation experiments, in which we found that vanilloid acid and TGF-βR1 and IFT88 have a stable binding relationship (Figures 4C, D). Then, in order to verify the mechanism of action of the small molecules, we found that the coincidence rate of TGF-βR1 and Arl13b proteins was significantly decreased in the drug-added group, which was consistent with previous experiments (Figure 4E).

In order to determine whether the Pscs were affected by TGF-β1, we examined the TGF-β1 content in the stem cell culture media of each group after radiation dosing. The results showed that radiation could cause the increase in TGF-β1, whereas there was no significant decrease after dosing (Ps 3A).

To verify whether these two small molecules also play a role in enhancing the radiosensitivity of tumor cells, we selected two lung cancer cell lines—A549 and H1299—which were subjected to the same radiation (2 Gy) as the paracellular model with drug intervention. In this part, we observed cell morphology and used a flow cytometry CCK-8 experiment to detect apoptosis and proliferation (Figures 5A–E; Ps2 C-D), and the results showed that neither the single drug nor the superimposed drug had a killing effect on lung cancer cells, although the killing effect on tumor cells was stronger after the superimposed drug was given.