We purchased Pae from MedChemExpress (MCE, USA) with a purity level of at least 98%. We dissolved it in dimethyl sulfoxide (DMSO; Solarbio, Beijing) at 200 mmol/L and kept it at −20 °C. In addition, 3-methyladenine (3-MA) and the Cell Counting Kit-8 (CCK-8; Multisciences Biotech, China) were utilized as reagents (MedChemExpress, USA). The protein phosphatase inhibitor and phenylmethylsulfonyl fluoride were purchased from Solarbio in Beijing, China, for the purpose of protein extraction. Affinity provided the PI3K and phosphorylated PI3K antibodies at a 1:1000 dilution, HuaBio provided the AKT and phosphorylated AKT antibodies at a 1:500 dilution, and Abcam provided the Bax, Caspase-3, and Bcl-2 antibodies at 1:2000 dilutions. In addition, MedChemExpress (USA) supplied the cell-permeable PI3K activator 740Y-P (HY-P0175).

Short tandem repeat (STR) profiling was used for verification of OSCC cell lines CAL-27 and HSC-3, as well as normal oral keratinocytes (NOK). The cells were grown in a humidified environment at 37 °C with 5% CO₂ in Dulbecco’s Modified Eagle Medium (DMEM) that contained 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. In every experiment, cells were employed while they were in the logarithmic growth phase.

A density of 5 × 10³ cells per well of CAL-27, HSC-3, and NOK cells was used in the seeding of 96-well plates, which were then incubated at 37 °C for 24 h. Cells were then treated for an additional 24 h with several doses of Pae (0.5, 1, 1.5, 2, 2.5, and 3 mM) or a vehicle control (DMSO). The maximum concentration of Pae was mimicked in the DMSO concentration given to control cells. Three times, the experiments were repeated. After incubating with a 10% CCK-8 solution for 2 h, the absorbance at 450 nm was measured using a microplate reader to test cell viability. Thus, half-maximal inhibitory concentration (IC₅₀) values were computed, and Pae concentrations of 0, 0.4, 0.8, and 1.6 mmol/L were chosen for the following experiments according to these findings.

Standard conditions were used to incubate CAL-27 and HSC-3 cells in 6-well plates with 2 mL of DMEM supplemented with 10% FBS until around 90% confluence was achieved. A sterile pipette tip was used to create uniform wounds. Cells were washed with phosphate-buffered saline (PBS) and then cultured in new media with Pae (0, 0.4, 0.8, 1.6 mM) after first imaging under an inverted microscope. The wound healing rates were determined by taking images again at 24 h and dividing the difference between the scratch areas at 0 and 24 h by the scratch areas at 0 h, and then multiplying the result by 100%.

The upper chambers of the transwell inserts were coated with 200 μL of Matrigel that had been diluted in serum-free DMEM at a 1:8 ratio. The coated inserts were planted with CAL-27 and HSC-3 cells that were suspended in serum-free DMEM with Pae (0, 0.4, 0.8, 1.6 mM). The chemoattractant in the lower chambers was 500 μL of DMEM containing 10% FBS. The invasive cells were preserved with 4% paraformaldehyde after 24 h of incubation, stained with 0.1% crystal violet, and gently washed with PBS. Using cotton swabs, non-invasive cells were extracted. To measure the number of invasive cells, images were taken from five fields at a magnification of ×200, which were chosen at random.

CAL-27 and HSC-3 cells were placed in 6-well plates at 1 × 10³ cells/well and left to adhere for a full day. The culture was maintained for a total of 14 days after exposure to the stated treatments, with three duplicates per condition. Media was replaced every 3 days thereafter. To assess their clonogenic capacity, cells were washed with PBS, fixed with 4% formaldehyde for 20 min, stained with 0.1% crystal violet for 30 min, and finally counted.

Following 24 h of culture, the cells were subjected to total protein extraction using ice-cold RIPA buffer that had been treated with PMSF and protein phosphatase inhibitors (Solarbio, Beijing). A BCA assay kit (Solarbio) was used to quantify protein quantities. Proteins in equal concentrations were denaturated at 100 °C, separated using 10% SDS-PAGE, and then transferred to PVDF membranes. After being coated with Rapid Closure Solution for 20 min, the membranes were incubated overnight with primary antibodies (with β-actin or GAPDH as loading controls; for EMT-related blots, total-protein staining was used as an additional loading verification). The next day, they were incubated under dark conditions with secondary antibodies. Using an ECL detection system, proteins were seen, and band intensities were evaluated using ImageJ software.

Structural data of Pae (SDF format) were acquired from PubChem and subsequently uploaded to the Pharmmapper platform to predict potential molecular targets. Resultant protein targets were standardized into gene symbols utilizing the UniProt database.

OSCC-related molecular targets were identified by querying the DisGeNET (https://www.disgenet.org/), GeneCards (https://www.genecards.org/), and OMIM (https://omim.org/) databases using the keyword “OSCC.” After consolidating the results, duplicate entries were removed to compile a comprehensive list of OSCC-associated targets.

The intersection between Pae and OSCC-related targets was established, and overlapping targets were visualized through a Venn diagram constructed using the online Microbiotics tool (http://www.bioinformatics.com.cn/?p=1).

Overlapping targets were submitted to the STRING database (https://cn.string-db.org/) with the organism set to “Homo sapiens” and a minimum required interaction score of 0.4 (medium confidence). The resulting protein-protein interaction (PPI) network was exported and imported into Cytoscape (v3.7.2) for visualization and topological analysis. Hub targets were ranked using the CytoHubba plugin based on the maximal clique centrality (MCC) algorithm; nodes were sorted by MCC score, and the top 10 targets were defined as hub genes.

The DAVID database (https://david.ncifcrf.gov/) was used to conduct enrichment analysis on overlapping targets. The significance criterion was set at p < 0.05, and the target species chosen was “H. sapiens”. Categories that were enhanced comprised GO Biological Processes (BP), Molecular Functions (MF), Cellular Components (CC), and KEGG pathways. With the help of the Weixin web (http://www.bioinformatics.com.cn/?p=1) application, bubble charts were created to show the twenty most enriched KEGG pathways, and bar charts were used to display the ten most important GO keywords (BP, MF, CC).

Before being treated with different doses of Pae (0, 0.4, 0.8, and 1.6 mM) for 24 h, CAL-27 and HSC-3 cells were plated in 6 cm culture dishes until they reached about 50%–70% confluence. Using an Annexin V-FITC/PI Apoptosis Detection Kit (KGA1102, KeyGenBio, China), cells were detached using EDTA-free trypsin and subsequently stained upon collection. Flow cytometry (FC) was used to quantify apoptotic cells, and the data were processed with the help of FlowJo software.

Centrifugation was used to extract CAL-27 and HSC-3 cells after they had been treated with Pae (1.6 mM) for 24 h. The cells were then fixed overnight with 2.5% glutaraldehyde. The cells were first immersed in 1% agarose, then fixed for 2 h at room temperature in darkness using 1% osmium tetroxide. After that, they were dehydrated using an ethanol gradient and embedded in acetone. Representative micrographs were taken after staining and examining ultrathin slices (60–80 nm) under a transmission electron microscope.

Under typical incubation settings (37 °C, 5% CO₂, and 90% humidity), CAL-27 cells were transfected with mTagRFP-senseGFP-LC3-lentivirus (KL103481-jlV, KeyGenBio, China). After that, they were treated with Pae (1.6 mM) for 24 h. The autophagic flux was tracked and recorded with the help of a laser confocal microscope (FV3000-RS, Olympus).

GraphPad Prism software (version 10) was utilized for statistical evaluation and graphical representation. Data from experiments repeated independently three times were presented as mean ± standard deviation (SD). Appropriate statistical tests, including Student’s t-test, one-way ANOVA, two-way ANOVA, or three-way ANOVA, were employed to determine group differences. Statistical significance was defined at p < 0.05, with significance levels marked as follows: ns (p > 0.05), *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.000.

The chemical structure of Pae is illustrated in Figure 1A. To ascertain optimal dosages, CAL-27 and HSC-3 cells were treated with increasing concentrations of Pae for 24 h. IC₅₀ values, assessed using the CCK-8 assay, were identified as 1.853 mmol/L for HSC-3 and 1.316 mmol/L for CAL-27 cells (Figures 1B,C). Subsequent experiments utilized Pae concentrations representing approximately ¼ IC₅₀, ½ IC₅₀, and full IC₅₀ values (0.4, 0.8, and 1.6 mM). A dose-dependent inhibitory effect of Pae on cell migration was observed via scratch wound assays in both OSCC cell lines (Figures 1D,E). Complementary Transwell assays further validated these results, demonstrating significant suppression of cell migration and invasion upon Pae treatment (Figures 1F,G). Since EMT significantly contributes to cancer cell invasiveness and migration (Patra et al., 2020), EMT-related markers (Du and Shim, 2016), and Vimentin were examined using WB. Pae treatment notably increased E-cadherin while reducing N-cadherin and Vimentin, suggesting that Pae effectively suppresses OSCC cell invasiveness and migration.

To further investigate the impact of Pae on OSCC cell viability, CAL-27, HSC-3, and NOK were incubated with Pae concentrations ranging from 0 to 3 mM for 24 h. The CCK-8 assay indicated that OSCC cell viability diminished with increasing Pae concentration, while NOK viability remained largely unaffected at concentrations below 2 mmol/L, indicating limited cytotoxic effects on normal epithelial cells (Figure 2A). Colony formation assays further revealed that higher Pae concentrations notably suppressed clonogenic potential in both OSCC cell lines (Figure 2B). WB analysis of proliferating cell nuclear antigen (PCNA) protein showed concentration-dependent decreases after Pae treatment, confirming Pae’s anti-proliferative effects (Figure 2C). These collective results demonstrate the inhibitory capacity of Pae on OSCC proliferation, with minimal cytotoxicity toward normal oral epithelial cells.

Based on the PharmMapper database, 95 potential target proteins of Pae were identified. From the GeneCards database, 6107 OSCC-related targets were retrieved, and an additional 1019 targets were sourced from the OMIM database. Following merging and eliminating duplicates, 639 targets related to OSCC were identified. The potential targets of Pae were compared with the OSCC disease targets, revealing 40 common targets. A Venn diagram illustrating this overlap was created utilizing an online application (Figure 3A).

Forty intersecting targets between Pae and OSCC were analyzed further by generating a PPI network via the STRING database (Figure 3B). The generated network comprised 40 nodes (with two proteins showing no interactions) connected by 146 edges, with an average interaction degree of 7.3. Hub targets were ranked using the MCC algorithm in CytoHubba, and the top 10 targets were selected as hub genes (Figure 3C), including CDC42, AKT1, CCL5, STAT1, MMP9, LCK, SRC, JAK2, KIT, and IL2. These results suggest that these proteins may serve as core targets through which Pae exerts its anti-OSCC effects.

The top ten enriched GO terms within BP, CC, and MF categories are presented in Figure 4A. The predominant BP terms included protein phosphorylation, regulation of protein kinase B signaling, stimulation of cell proliferation, and inhibition of apoptosis. Primary CC terms involved cytosolic regions and macromolecular complexes, while notable MF terms encompassed ATP binding, enzyme binding, protein kinase binding, and phospholipase activator activity. Additionally, the twenty most significant KEGG pathways are depicted in a bubble plot (Figure 4B). Of these, pathways such as PI3K/AKT and chemokine signaling were markedly enriched. Given its significant relevance, the PI3K/AKT pathway was chosen for further exploration.

To clarify whether apoptosis mediated the inhibitory effects of Pae on OSCC proliferation, apoptosis was analyzed via Annexin V/PI staining and FC. Pae exposure increased the proportions of early and late apoptotic cells in a dose-dependent manner (Figure 5A). Additionally, WB results demonstrated marked downregulation of Bcl-2 and upregulation of Bax and cleaved-caspase-3, accompanied by reduced pro-caspase-3 levels in Pae-treated cells compared to untreated controls (Figure 5B). Network pharmacology analyses identified the PI3K/AKT pathway as a potential apoptotic mechanism targeted by Pae. Consistently, Pae reduced the phosphorylation of PI3K and AKT (p-PI3K, p-AKT) without altering their total protein expression levels (Figure 5C). Further validation through PI3K activation demonstrated reversal of Pae-induced changes in PI3K/AKT phosphorylation and apoptosis-associated proteins upon activator co-treatment (Figure 5D).

Autophagy induction by Pae was analyzed through transmission electron microscopy (TEM) in OSCC cells after 1.6 mM Pae exposure for 24 h. TEM images revealed increased autolysosomal structures in treated cells compared to controls (Figures 6A,B). Autophagic activity was further examined by assessing autolysosomal structures, which were notably increased in Pae-treated cells compared to controls (Liu et al., 2023). WB analysis of autophagy markers p62 and LC3 showed decreased p62 and elevated LC3-I / LC3-II levels dose-dependently after Pae treatment, reinforcing the induction of autophagy in OSCC cells (Figures 6C,D).

During autophagosome formation, LC3-I is converted to its lipidated form LC3-II, which serves as an established marker for autophagy (Zhou et al., 2012). To investigate Pae-induced autophagic flux comprehensively, CAL-27 cells were transduced with mTagRFP-senseGFP-LC3 lentivirus. The SensGFP fluorescent signal diminishes under the acidic environment of autolysosomes, whereas StubRFP fluorescence remains stable irrespective of pH (Kaizuka et al., 2016; Singh et al., 2014). This dual-fluorescence system was used to assess autophagic flux (Figures 7A,B). Compared to untreated controls, CAL-27 cells treated with 1.6 mM Pae for 24 h showed numerous yellow puncta (StubRFP⁺/SensGFP⁺), indicating the presence of autophagosomes. Notably, red puncta (StubRFP⁺) increased significantly, while green fluorescence (SensGFP⁺) decreased after Pae treatment, reflecting the formation of autolysosomes. Confocal microscopy revealed elevated accumulation of autophagosomes and autolysosomes following Pae treatment, confirming enhanced autophagic flux. These observations align closely with the WB results, providing additional support for Pae’s autophagy-promoting effects in OSCC cells.

To examine whether pharmacological modulation of autophagy/PI3K signaling influences the cellular response to Pae, OSCC cells were treated with 3-MA, an inhibitor of class III PI3K commonly used to suppress early-stage autophagy. The CCK-8 assay showed that co-treatment with Pae (1.6 mM) and 3-MA (2.5 mM) reduced cell viability more than Pae alone (Figure 7C). In WB analyses, 3-MA alone increased p62 and reduced LC3-I/LC3-II, whereas co-treatment altered LC3-I/LC3-II and p62 levels relative to Pae alone (Figure 8A), consistent with modulation of the autophagy pathway. Co-treatment also decreased p-PI3K and p-AKT (Figure 8B) and shifted apoptosis-related proteins toward a pro-apoptotic pattern (decreased Bcl-2 and increased Bax; Figure 8C). Collectively, these data indicate that 3-MA co-treatment potentiates Pae-associated growth inhibition and apoptosis-related changes, although the relative contributions of autophagy inhibition versus PI3K pathway effects require further clarification.