All animal procedures were approved by the Ethics Committee of Qilu Hospital of Shandong University (permit number: DWLL-2021-095). This study was conducted in full compliance with relevant 100 regulations and guidelines.

Primary GBM#P3 cells and BG7 glioma stem cells (GSCs) were kindly gifted by Prof. Rolf Bjerkvig (University of Bergen, Bergen, Norway). The cells were cultured in NeurobasalTM Medium (Gibco/Thermo Fisher Scientific; Waltham, MA) supplemented with 2% B27 supplement (Invitrogen; Carlsbad, CA), 2% L-glutamine (BE17-605E, BioNordika; Oslo, Norway), EGF (20 ng/mL; Thermo Fisher Scientific), and basic FGF (10 ng/mL; Thermo Fisher Scientific).

Safranal (purity ≥98.0%, CAS# 116-26-7) was purchased from MedChemExpress (MCE; Monmouth, NJ,120 USA). It was dissolved in dimethyl sulfoxide (DMSO; Sigma–Aldrich; St. Louis, MO, USA) as a stock solution (1 M) and stored at -20°C. Temozolomide was purchased from Topscience (Shanghai, China). It was dissolved in DMSO as a stock solution (100 mM) and stored at -20°C.

The following antibodies were used: GAPDH (#5174S, CST; Boston, MA, USA), CDK1 (#28439S, CST), CCNB1 (67686-1-Ig, ProteinTech; Wuhan, China), CDK6 (66278-1-Ig, ProteinTech), PARP1 (66520-1-Ig, ProteinTech), Caspase3 (# 14220S, CST), Caspase7 (# 12827S, CST), AKT (# 4691S, CST), mTOR (66888-1-Ig, ProteinTech), phospho-mTOR (Ser2448) (# 5536S, ProteinTech), beta-actin (66009-1-Ig, ProteinTech), Ki67 (GB111499, Servicebio; Wuhan, China)

We used Cell Counting Kit-8 (CCK-8; Dojindo, Kumamoto, Japan) to assess cell viability. First, we pretreated the 96-well plate with 1× polylysine (Solarbio) for 15 min. Then, we discarded polylysine, washed the 96-well plate with sterile phosphate buffered saline (PBS) three times and seeded 1×10⁴ cells into each well. After 24 h, the supernatant was replaced with different concentrations of safranal, temozolomide, or vehicle control (dimethyl sulfoxide (DMSO)). At 12, 24, or 48 hours after treatment, GBM cells were incubated with 10 μl of CCK-8 reagent in 100 μl of NeurobasalTM Medium supplemented with 2% B27 supplement, 2% L-glutamine, EGF (20 ng/mL), and basic FGF (10 ng/mL) at 37°C for 1 hour. Then, we used an EnSight Multimode Plate Reader (PerkinElmer; Singapore) to measure the absorbance at 450 nm.

GBM#P3 or BG7 cells were seeded into 3D Culture Qualified 96-well spheroid formation plates (Trevigen) with 50 μl of medium, with 3000 cells per well. Then, they were cultured at 37°C and 5% CO2 for 72 hours in a humidified incubator for 72 h to form tumour spheroids. Then, we placed the plate on ice for 15 minutes and added 50 μl of invasion matrix (Trevigen, 3500-096-03) into each well. The plates were centrifuged at 300×g at 4°C for 15 minutes and incubated at 37°C for 1 hour. Then, 100 μl medium with 100 μM safranal or DMSO was added to each well. After 48 h (GBM#P3) or 72 h (BG7), we used a 4× objective to capture the spheroids under a bright field microscope. The images were analysed with ImageJ software.

The GBM-brain organoid coculture invasion ex vivo system was performed as described in a previous study (27). GFP-transfected GBM#P3 cells (3000 cells per well) or BG7 cells (5000 cells per well) were cultured to generate glioma spheroids. Then, mature brain organoids were cocultured for 36 h with GBM#P3 cells or 48 h with BG7 cells. Confocal microscopy (Leica TCS SP8; Wetzlar, Germany) was used to capture GBM cell invasion images. The images were analysed with ImageJ software.

We used the Traditional Chinese Medicine Systems Pharmacology (TCMSP) (https://old.tcmsp-e.com/tcmsp.php) database, SwissTargetPrediction (http://www.swisstargetprediction.ch/) database, and PharmMapper (http://www.lilab-ecust.cn/pharmmapper/) database to predict potential targets of safranal. A Venn diagram was made on the Venny website (https://bioinfogp.cnb.csic.es/tools/venny/).

We used the keyword “glioma” to determine the potential glioma-related targets from the GeneCards (https://www.genecards.org/) and OMIM (https://omim.org/) websites.

The “venndiagram” package of R software (version 4.1.3 (2022-03-10)) was used to obtain 72 intersecting genes among glioma-related genes and the safranal target genes, and a Venn diagram was drawn. We used Cytoscape 3.9.1 software to visualize them as a “Drug-Targets-Disease” network.

The STRING database (https://cn.string-db.org/) was used to obtain the relationships between the 72 intersecting genes. PPIs with a combined score greater than 0.4 were selected and used to construct a PPI network. Then, we used the MNC algorithm in the cytohubba plug-in in Cytoscape 3.9.1 software to find the key proteins in the network.

The “clusterProfiler” and “org.hs.eg.db” packages in R software were used to analyse the enrichment of drug-disease genes in GO terms and KEGG pathways. Then, we visualized the results with R software.

We treated GBM#P3 cells with safranal (100 μM) or vehicle (DMSO) for 24 h and performed RNA sequencing (LC-Bio; Shanghai, China). Full data accompanying this experiment are available in the Sequence Read Archive (SRA). The accession number is PRJNA991989. Differentially expressed genes were analysed and visualized by R software. The processes for PPI network analysis and GO and KEGG pathway enrichment analysis were the same as above.

We first searched the UniProt ID of the target protein from the UniProt database (https://www.UniProt.org/), and then searched the structure of target proteins in the PDB database (https://www.rcsb.org/) using the UniProt ID of the protein. Protein structures were downloaded. PyMOL software were used to remove water and organic.

The 2D structures of small-molecule ligands were acquired with PubChem (https://pubchem.ncbi.nlm.nih.gov/).The pubchem CID of safranal was 61041, and the molecular weight was 150.22g/mol.

We used AutoDockTools-1.5.7 to add hydrogens to the protein and simulate potential targets for molecular docking. Then, AutoDockVina were used to perform molecular docking.

Total protein of GBM#P3 or BG7 cells was extracted with cold RIPA Lysis Buffer (Beyotime; Shanghai, China) supplemented with phenylmethanesulfonyl fluoride (PMSF, Beyotime). Western blotting was performed according to previously described protocols (28). We used a chemiluminescence imager (Bio-Rad; Hercules, CA, USA) to detect the chemiluminescent signals according to the manufacturer’s protocol.

Seventy percent ethanol was used to dehydrate the cells overnight at 4°C. Then, the cells were stained with PI/RNase (550825, BD Biosciences; San Jose, CA). After 15 min in the dark, the cells were analysed by flow cytometry (Accuri C6, BD Biosciences).

We resuspended cells in 300 μL binding buffer and stained them with Annexin V-FITC (BD Biosciences) according to the manufacturer’s instructions. Then, the cells were analysed by flow cytometry.

For statistical purposes, three nude mice were used in each group. A total of 1×10⁶ luciferase-stable GBM#P3 cells were injected into the brains of 15 4-week-old male nude mice. One week later, tumours were examined using bioluminescence imaging (PerkinElmer IVIS Spectrum; Waltham, MA, USA). We found that in the 15 mice, one had no intracranial tumor formation and two had spinal metastases, so we excluded them. Our inclusion criteria included successful tumor implantation, uniform tumor size, and no spinal metastasis. Exclusion criteria included that there was no intracranial tumor formation, the tumor volume was significantly larger or smaller than that in other mice, and spinal metastasis. Then, we classified the remaining 12 mice as numbers 1-12, used excel software to generate random numbers, arranged the random numbers in reverse order, and then divided the mice into four groups(each group has 3 mice) from top to bottom: control group (n=3), safranal group (n=3), TMZ group (n=3) and TMZ+safranal group (n=3). Tumour volume was monitored using bioluminescence imaging. Mice were euthanized after 2 weeks of treatment.Cages were placed in the same house, keeping the external environment consistent.

Mice in each group were harvested, and the brains were removed, fixed in 4% formalin, and prepared as paraffin-embedded sections. Immunohistochemistry was performed as previously reported (28).

First, we performed in vitro experiments to test whether safranal had any effect on the proliferation and invasion of primary GBM cells. The CCK-8 assay showed that the viability of GBM cells decreased significantly with increasing safranal concentration ( Figure 2A ). GBM-brain organoid coculture experiments ( Figures 2B, C ) and 3D tumour spheroid invasion assays ( Figures 2D, E ) demonstrated that safranal could inhibit the invasion of GBM cells.

Then, candidate targets of safranal were searched from the SwissTargetPrediction database (http://www.swisstargetprediction.ch/), PharmMapper (http://www.lilab-ecust.cn/pharmmapper/), and TCMSP (https://old.tcmsp-e.com/tcmsp.php) ( Figure 3A ).

A total of 1932 glioma-related targets were obtained from the GeneCards (https://www.genecards.org/) and OMIM (https://omim.org/) databases.

We used 308 targets of safranal and 1932 glioma-related targets for Venn analysis ( Figure 3B ) to obtain 72 intersecting targets and a drug–targets–disease network ( Figure 4 ). The hexagon in violet represents glioma, the “V” in blue represents safranal, and the ellipses in green represent the intersection targets.

We used the STRING database to find the protein interaction relationship of 72 intersection targets ( Figure 5A ) and counted the top 20 nodes connected most in the network ( Figure 5B ). Then, we imported the results of the STRING database into Cytoscape software and used the MNC algorithm in the CytoHubba plug-in to obtain the top 20 core targets ( Figure 5C ). We found that the top 20 core targets obtained by the MNC algorithm had 19 of the same targets as the top 20 targets calculated in Figure 5B ( Figure 5D ), which illustrated that these 19 targets played a key role in the network.

To explore the possible mechanism of safranal in treating glioma, KEGG pathway enrichment analysis was performed on 19 core targets ( Figures 5E, F ). We found that these 19 targets were enriched in some critical tumour-related pathways, such as apoptosis and the PI3K-Akt signaling pathway. This suggests that safranal may inhibit the growth of glioma cells through these two signaling pathways.

Then, primary GBM cells (GBM#P3) were treated with DMSO or 100 μM safranal, and mRNA sequencing was performed. We obtained 1751 differentially expressed genes between the treatment groups ( Figures 6A, B ). We used the STRING database to obtain the interaction network between the proteins encoded by these genes. The MNC algorithm in the CytoHubba plug-in of Cytoscape software was used to find the core protein of the network ( Figure 6C ). Subsequently, KEGG pathway enrichment analysis was performed based on these proteins. We found that these targets were enriched in the cell cycle and PI3K-Akt signaling pathways ( Figures 6D, E ). We continued to perform Gene Ontology (GO) functional analysis. The results showed that the core proteins of the differentially expressed genes were enriched in the cyclin-dependent protein kinase holoenzyme complex, cyclin-dependent protein serine/threonine kinase regulator activity and other cell cycle-related pathways ( Figures 6F–H ).

Based on the above analysis, we found that the inhibitory effect of safranal on glioma was most likely achieved through apoptosis, the PI3K-Akt signaling pathway and the cell cycle pathway. Therefore, we selected several key proteins in these pathways to perform molecular docking with safranal. As previously reported, safranal can inhibit the PI3K/Akt/mTOR axis, and we docked safranal with the proteins in this axis. The results showed that the binding energies of safranal with AKT1, PI3K and mTOR in the PI3K/AKT/mTOR axis; PARP1, CASP3, and CASP7 in the apoptosis pathway; and CCNB1, CDK1 and CDK6 in the cell cycle pathway were lower than -5 kcal/mol ( Figure 7 , Table 1 ). This indicates that safranal’s inhibitory effect on glioma is most likely to be achieved by inhibiting the PI3K/Akt/mTOR axis, acting on apoptosis-related proteins to increase tumour cell apoptosis, and acting on cycle-related proteins to regulate the tumour cell cycle.

According to previous studies, safranal could cause G2/M phase arrest of HepG2 cells (23). In our molecular docking prediction, we found that safranal may have a direct interaction with the key proteins CCNB1 and CDK1 of the G2/M transition. We suspect that safranal may cause G2/M phase arrest in GBM cells, so we used the primary GBM cells (GBM#P3 and BG7 cells) for experimental verification. Figures 8A, B shows that after treatment with safranal for 24 h, GBM#P3 and BG7 cells showed G2/M phase arrest. In addition, the western blotting results showed that safranal could significantly reduce the protein levels of CDK1, CCNB1, and CDK6 in GBM#P3 and BG7 cells ( Figures 8E, F ).

We further assessed the effect of safranal on the apoptosis pathway of GBM cells. In Figures 8C, D we used flow cytometry to detect the effect of safranal on the apoptosis rate of GBM#P3 and BG7 cells. As we previously predicted, the apoptosis rate of GBM cells was significantly increased after safranal treatment. The Western blotting results also showed that cleaved caspase3, cleaved caspase7, and cleaved PARP1 were significantly upregulated in GBM#P3 and BG7 treated with safranal ( Figures 8G, H ), which further confirmed that safranal could promote GBM cell apoptosis.

In our molecular docking studies, we found that the binding energies (BEs) of safranal with AKT1, PI3K and mTOR were less than -5 kcal/mol, indicating that safranal may be involved in regulating the PI3K/AKT/mTOR axis. This prediction was tested experimentally. The Western blotting results showed that the levels of PI3K, AKT, mTOR and p-mTOR in GBM#P3 and BG7 cells decreased significantly after safranal treatment ( Figures 8I, J ). This confirms that safranal inhibits the PI3K/AKT/mTOR axis in GBM.

TMZ is a first-line treatment for newly diagnosed glioblastoma (29, 30). However, not all patients respond to TMZ (30, 31). It was previously reported that PI3K/AKT axis inhibitors could inhibit temozolomide resistance in glioblastoma cells (32–34), and activation of the PI3K/AKT axis could promote TMZ resistance in GBM cells (35–37). Therefore, we suspect that safranal and TMZ may have a synergistic effect against GBM. We used a CCK-8 experiment for verification. The results in Figure 8K and Figure 8L show that the combination of safranal and TMZ had a better effect than either single agent alone, which verified that safranal and TMZ could synergistically inhibit the growth of GBM cells in vitro.

To further explore whether systemic administration of safranal is effective for intracranial tumours, we implanted 1×10⁶ luciferase-stable GBM#P3 cells into the brains of 4-week-old nude mice. One week later, we randomly divided them into four treatment groups and started treatment. In the control group, the vehicle was injected intraperitoneally (twice a day); in safranal group, intraperitoneal injection of safranal (15 mg/kg, twice daily) was used; the TMZ group received intraperitoneal injection of temozolomide(25 mg/kg) every morning and vehicle every afternoon; TMZ+safranal group received intraperitoneal injection of temozolomide(25 mg/kg) with safranal (15 mg/kg) every morning and safranal 15 mg/kg every afternoon. Two weeks later, tumour growth was evaluated using luciferase bioluminescence. We found that the tumours in the safranal-treated group were significantly smaller than those in the control group, while the intracranial tumours in the safranal combined with temozolomide group were significantly smaller than those in the temozolomide monotherapy group ( Figures 9A, B ).The immunohistochemical results showed that the expression of Ki-67 in tumour cells of safranal-treated mice was significantly lower than that tumour cells of the control mice ( Figures 9C, D ). This indicated that safranal could inhibit the growth of GBM in vivo.