Human TSCC cell lines SCC9 and CAL27, human gastric cancer cell line BGC823, and human hepatocellular carcinoma cell line HCC-LM3 were cultured in Dulbecco’s modified Eagle medium (DMEM; Gibco) or DMEM/F12, to which was added 10% fetal bovine serum (Gibco). SCC9 and CAL27 cells were treated with 40 μM of chrysin (Yuanye Bio-Technology, Shanghai) for 48 h.

The EVs were extracted from the culture medium of SCC9 cells that were treated with chrysin (EVs-chrysin) or PBS (EVs-Con). The culture medium of SCC9 cells with 70% confluency was immediately collected. Then, the cell culture was centrifuged at 300×g for 30 min, at 2000×g for 30 min, and at 12, 000×g for 45 min, and then the supernatant fluid was filtered. Next, the filter liquor was ultracentrifuged (Thermo Scientific) at 110, 000×g for 70 min. The EVs were collected and washed with PBS. The EVs were ultracentrifuged at 110, 000×g for 70 min again. For Au-EVs, the collected EVs that contained chrysin (EVs-chrysin) were incubated with HAuCl₄ (50 mM, Sigma) at 37°C overnight. The particle size and concentration of EVs-Con and EVs-chrysin were detected using dynamic light scattering (DLS) by nano-flow cytometry (NanoFCM, China). The shape of the EVs-Con, EVs-chrysin, and Au-EVs was determined by transmission electron microscopy (TEM, HITACHI, HT-7800).

The EVs (40 μl) were diluted and fluorescent-labeled antibodies were added (20 μl, CD9, CD63, CD81, and BGI) at 37°C for 30 min. Then, 1 ml of PBS was added and ultracentrifugated at 4°C, at 110,000×g for 70 min. IgG, CD9, CD63, and CD81 of EVs were detected by NanoFCM (BGI, China).

The EVs-Con, EVs-chrysin, and Au-EVs were labeled with PKH26. First, 1 μl PKH26 linker and 9 μl diluent C were premixed. Then, the EVs-Con, EVs-chrysin, and Au-EVs (10 and 30 μg) were added to the mixed solution and incubated for 10 min. Next, the EVs-Con, EVs-chrysin, and Au-EVs that were labeled with PKH26 were ultracentrifuged at 100, 000×g for 70 min. Finally, 200 μl PBS was added to resuspend the labeled EVs-Con, EVs-chrysin, and Au-EVs. A laser confocal microscope was used to capture the image.

5×10⁴ of SCC9, BGC823, and HCC-LM3 cells were cultured and incubated with 200 μl–labeled Au-EVs for 15 min or 24 h. The harvested cells were washed with PBS to remove non-incorporated AuEVs. Hoechst was used to label the nuclei. The image was observed using a laser confocal microscope.All experiments were performed in triplicate.

The total RNA from EVs-Con and EVs-chrysin was extracted. The EVs were treated with PBS as a con group (EVs-Con). Using DNBSEQ (Beijing Genomics Institute, BGI, China), 264 small RNAs were identified. Quality control was performed on the raw reads to get clean reads. Then, the clean reads (total clean reads, 32M) were aligned to the reference gene sequence. Small RNAs were counted and classified. The differential miRNA expression was determined (Fold Change >0.5, FDR <0.001). Using Dr.Tom software (BGI, China), the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and the GO enrichment analysis of differentially expressed miRNAs were determined. The data were submitted to the Gene Expression Omnibus (GEO) database (GSE185562).

Three patients underwent clinical surgery, and their samples were collected. The TSCC samples were immediately stored in liquid nitrogen for RNA isolation.

The let-7a-3p sequence is listed in Supplementary Table S1. The let-7a-3p mimics and inhibitor were purchased from GenePharma (Shanghai, China). The SCC9 cells were transfected with the mimic and inhibitor of let-7a-3p for 48 h. Nonspecific siRNA (Nc) was used as the control.

RNAs of TSCC cells and tumors were extracted and cDNAs (cDNA First-Strand Synthesis Kit, TIANGEN, China) were generated. QPCR (quantitative real-time PCR) was performed to analyze the miRNA expression pattern. The primers of miRNAs and apoptosis genes are listed in Supplementary Table S2. The qPCR conditions were as follows: 94°C for 3 min and then 94°C for 10 s after 35 cycles. The annealing was carried out at 59°C for 15 s. The products were extended at 72°C for 30 s. U6 or GAPDH was used as the control. All experiments were performed in triplicate.

Cell migration was measured by applying a wound healing assay in the SCC9 cells. The SCC9 cells (5 × 10⁵) were treated with let-7a-3p mimics or inhibitors for 48 h. Then, the cells were scratched and cultured with FBS medium. The scratched area was analyzed at intervals of 12, 24, and 48 h.

The SCC9 cells (3 × 10⁴) were used for cell invasion assays. The cells were transfected with the let-7a-3p mimics or inhibitor. Matrigel matrix (20 μl, BD Biosciences, United States) was added and incubated along with the medium overnight. Crystal violet dye (0.2%, Solarbio, China) was used to stain the SCC9 cells. The stained cells were analyzed using ImageJ software. Each group of experiments was performed in triplicate.

The SCC9 cells (1 × 10³) were used for the colony formation assay. The cell colonies were fixed with paraformaldehyde and stained with crystal violet after 8 days. The number of cells were counted using ImageJ. The experiments of the colony formation assay were performed in triplicate.

The let-7a-3p mimics or inhibitor were transfected in SCC9 cells. An annexin V-FITC/PI reagent was used for cell apoptosis analysis. Flow cytometry (BD Biosciences, Franklin Lakes, NJ, United States) was performed to detect apoptosis cells. All the apoptosis experiments were performed in triplicate.

The SCC9 cells were treated with Au-EVs and Au-EVs with NIR. The Con group was treated with PBS. Then, the cells were fixed with 4% paraformaldehyde. These cells were blocked with PBS containing 1% BSA in the dark for 1 h with TdT and fluorescein-conjugated dUTPs (In Situ Cell Death Detection kit; Roche, Germany). DAPI was used to stain the nuclei. A fluorescence microscope was used to capture the images. The TUNEL assay was performed in triplicate.

The protein was collected using a protein extraction buffer (Beyotime, China). The BCA assay was used to quantify the protein (Tiangen, Beijing, China). The protein separation process used sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The polyvinylidene difluoride (PVDF) membrane was incubated with a p53 antibody (Abcam, ab131442, United States), BAX (Cell Signaling Technology, D2E11, United States), BCL-2 (Cell Signaling Technology, D55G8, United States), caspase-3 (Wanleibio, WL02117, China), and GAPDH (Bioworld, AP0066, United States). HRP-conjugated AffiniPure goat antibodies IgG (Boster, China) were used as a secondary antibody. The ECL Super Signal kit (Pierce, United States) was used to analyze the bands.

The lungs, liver, spleen, kidneys, and heart tissues from the Con, Au-EVs, and Au-EVs + NIR groups were fixed in 4% paraformaldehyde for 48 h. Then, the samples were embedded in paraffin wax and cut into 5-μm sections. The slides were stained with H and E and observed under a light microscope.

The female nude mice (N = 20, 6–8 weeks old) were utilized to determine tumor growth in vivo. The nude mice were injected with SCC9 cells (8×10⁶) into the left flank and then divided into four groups: the Con, chrysin, Au-EVs, and Au-EVs + NIR groups. The tumors were observed after 7 days. The Con group was treated with saline water. The chrysin group was treated with chrysin (20 mg/kg) each day by intragastric administration. The PKH26-labeled Au-EVs were subcutaneously injected below the tumor on day 8 and day 15. The Au-EVs + NIR group was exposed to NIR (808 nm) after injection (twice, day 8 and day 15), and the tumor growth was observed in vivo with a fluorescence imaging system (λex = 530 nm, λem = 600 nm, AniView600, Guangzhou Biolight Biotechnology, China).

An unpaired Student’s t-test was utilized in the present study. Statistical analysis was conducted using GraphPad Prism 5.0 (GraphPad Software, Inc,.). All data were expressed as mean ± SD. A p-value <0.05 was considered statistically significant.

To obtain EVs, chrysin was treated in SCC9 cells. Our results showed that the shape of the EVs-Con and EVs-chrysin was round (Figures 1A,B). The DLS analysis determined that the size of the EVs-Con and EVs-chrysin was 50–150 nm (Figures 1C–F). The NanoFCM results indicated that CD9, CD63, and CD81 appeared in the EVs-chrysin (Figure 1G). Furthermore, the SCC9 cells were treated with the EVs-Con and the EVs-chrysin. The results showed that the EVs-Con and EVs-chrysin were successfully absorbed by SCC9 cells (Figures 1H,I). These results indicated that the EVs were isolated from SCC9 cells that contained chrysin and are absorbed by SCC9 cells. To improve the antitumor effect of EVs-chrysin, HAuCl₄ was used to incubate EVs-chrysin and synthesize Au-EVs (Figure 2A). To find the optimal protocol for EVs-chrysin to synthesize Au-EVs, 10 and 30 μg of EVs-chrysin were analyzed. The results showed that AuNPs were self-grown on the surface of EVs-chrysin (both 10 and 30 μg) and formed a new nanomaterial, which was Au-EV (Figure 2B). Similar to EVs-chrysin, Au-EVs were also effectively absorbed by SCC9 cells (Figure 2C). These results indicated that Au-EVs were stably absorbed in SCC9 cells.

To determine whether Au-EVs were specific to the cell type, SCC9, BGC823, and LM3 cells were used. The results showed that, compared to BGC823 and LM3 cells, the uptake of Au-EVs was specific in SCC9 cells (Figures 3A–D). Considering that Au-EVs contain chrysin, the TUNEL assay was used to analyze the cell apoptosis. The results suggested that Au-EVs and Au-EVs combined with NIR induced apoptosis in SCC9 cells (Figures 3E,F). Moreover, irradiation by NIR enhanced apoptosis in SCC9 cells (Figures 3E,F). These results indicated that Au-EVs combined with NIR promote significant apoptosis compared with that of Au-EVs.

Considering that EVs contain lots of miRNAs that are involved in cell apoptosis, RNA-seq was used to screen differentially expressed miRNAs between EVs-Con and EVs-chrysin. Overall, 158 genes were identified in EVs (Figure 4A). A total of 12 miRNAs were differentially expressed between EVs-Con and EVs-chrysin. Compared to the EVs-Con, 8 miRNAs were significantly upregulated, while 4 were significantly downregulated in EVs-chrysin (Figure 4B). The KEGG pathway and heatmap data indicated that the differentially expressed miRNAs have a role in cell growth and death (Figures 4C,D). Further, we analyzed the log2 fold change in the 8 upregulated miRNAs (let-7a-3p, miR-122, miR-199b, miR-26, miR-410, miR-451a, miR-6529, and miR-148) and the 4 downregulated miRNAs (miR-247, miR-264, miR-619, and miR302) (Figure 4E). To confirm these data, 4 miRNAs (miR-26, miR-122, miR-199b, and let-7a-3p) that were associated with cell growth and death were investigated. The qPCR results indicated that only let-7a-3p was upregulated after chrysin treatment in SSC9 and CAL27 cells (Figures 4F,G). These results suggested that the expression of let-7a-3p that might be involved in cell growth and death is regulated by chrysin.

To analyze the expression pattern of let-7a in TSCC patients, qPCR was performed. The results revealed that let-7a-3p showed less expression in the tumor than the paracancerous tissue (Figure 5A). To determine whether let-7a has a role in cell apoptosis (Figure 5B), overexpression and knockdown expression of let-7a-3p were used. The qPCR and Western blot results indicated that mimics of let-7a-3p increased the expression of the p53 protein, which was a key factor in the cell apoptosis pathway (Figures 5C,D; Supplementary Figure S1). Moreover, increased expression of let-7a-3p induced apoptosis in SCC9 cells (Figures 5E,F). In addition, the overexpression of let-7a-3p inhibited cell invasion (Figures 6A–D). Reduced expression of let-7a-3p promotes migration in SCC9 cells (Figures 6E,F). To confirm the let-7a-3p expression pattern, the SCC9 cells were treated with chrysin. The results showed that chrysin induced cell apoptosis and inhibited invasion (Figure 7). These results indicated that chrysin induced apoptosis and suppressed invasion via let-7a-3p in SCC9 cells.

To investigate the antitumor effect of Au-EVs in vivo, SCC9 cells were injected into nude mice. After 7 days, Au-EVs were injected below the tumor and irradiated with NIR in the nude mice for tumor growth analysis at day 8 and day 15 (Figure 8A). The results suggest that the Au-EVs could move toward the tumor. Moreover, the fluorescence intensity of the Au-EVs increased in a time-dependent manner. In addition, NIR irradiation could quench the fluorescence of the Au-EVs (Figure 8B). At day 21, the tumors were collected. As expected, Au-EVs combined with NIR significantly inhibited tumor growth and did not alter others organs in vivo (Figures 8C,D; Supplementary Figure S2). Moreover, the qPCR results indicated that, compared to the Con group, the expression of let-7a-3p was increased in the chrysin and Au-EV groups (Figure 8E). These results demonstrated that Au-EVs mediated PPT effectively and inhibited tumor growth in vivo.