First, we extracted exosomes secreted by BMSCs to prepare HEs by fusing them with liposome membranes and then prepared HEs loaded with DOX (HE/DOX) and liposomes loaded with DOX (LP/DOX) using the ammonium sulfate gradient method ( Figure 1 ). Transmission electron microscopy revealed the morphology of the exosomes, HE/DOX, and LP/DOX, all of which showed a double-layer membrane structure ( Figure 2A ). Nanoparticle tracking analysis demonstrated that the peak diameter of HE/DOX was 151.1 ± 10.2 nm ( Figure 2B ), similar in size to exosomes and LP/DOX. Western blot analysis showed that HEs expressed the exosomal markers ALG-2-interacting protein X (ALIX), CD63, and tumor susceptibility 101 (TSG101) but did not express the negative marker calnexin ( Figure 2C ), indicating the successful fusion of liposomes and exosomes. In addition, we measured the surface charge of the exosomes, LP/DOX, and HE/DOX, which had zeta potentials of −11.7 ± 0.8, −30.2 ± 2.0, and −19.7 ± 0.9 mV, respectively. The difference between LP/DOX and HE/DOX was statistically significant when compared to the exosome group (exosome vs. LP/DOX, P = 0.0003; exosome vs. HE/DOX, P = 0.0008) ( Figure 2D ). We used PKH26 to label HEs (red), and fluorescence overlap was observed using the intrinsic fluorescence of DOX (green) under confocal microscopy, which demonstrated that DOX was encapsulated within HEs ( Figure 2E ). To test the stability of the exosomes, LP/DOX, and HE/DOX, they were stored at 4°C and their stability in terms of particle size and zeta potential was measured. Compared with the zeta potential on the first day, the average changes in zeta potential in the exosome, LP/DOX, and HE/DOX groups on the last day were 4.5 ± 0.6, 3.8 ± 0.3, and 1.2 ± 0.5 mV, respectively. The differences between the groups were statistically significant (P = 0.0004). The results showed that HE/DOX was more stable than the exosomes ( Figures 2F, G ).

We studied the drug loading and release profiles of HE/DOX. The results showed that even at a DOX concentration of 0.8 mg/mL, the drug encapsulation efficiency was high (approximately 90%), whereas the loading efficiency continued to increase and could reach approximately 70% ( Figure 3B ). LP/DOX had a drug-loading profile similar to that of HE/DOX ( Figure 3A ), indicating that the extracellular vesicle membrane did not change the drug-loading properties of the liposomes. Leakage rates of HE/DOX and LP/DOX were also measured. Supplementary Figure S1 shows that at 4°C, no significant leakage of HE/DOX and LP/DOX was observed within 7 d, indicating that the HE/DOX membrane was highly stable.

In addition, we studied the drug release profiles of HE/DOX and LP/DOX under simulated physiological conditions (pH 7.4) and in a tumor endolysosomal compartment interior environment (pH 5.5). As shown in Figure 3 , HE/DOX and LP/DOX were rapidly released in the acidic environment; however, their release was slower under physiological conditions ( Figures 3C, D ).

To determine whether BMSC-derived exosomes enter osteosarcoma cells in vitro, we isolated exosomes from BMSCs and used liposomes as a control. PKH26-labeled BMSC-derived exosomes and liposomes were co-cultured with the osteosarcoma cell lines 143B and MG63 for 12 h, and fluorescence was quantified using confocal microscopy. The results showed that the uptake of BMSC-derived exosomes by 143B and MG63 cells was significantly higher than that of liposomes ( Figures 4A, B ). In addition, we co-cultured DiD-labeled exosomes with osteosarcoma cells and used flow cytometry to quantify exosome uptake by detecting fluorescence. The results also indicated that BMSC-derived exosomes were taken up by osteosarcoma cells more easily than were liposomes ( Figures 4C, D ).

To investigate the uptake of HE/DOX by tumor cells, we co-cultured PKH26-labeled HE/DOX with 143B and MG63. To study the temporal pattern of osteosarcoma cell internalization of HE/DOX, we observed internalization at 1, 3, and 6 h using confocal microscopy ( Figures 5A, B ). Osteosarcoma cell internalization of HE/DOX was significantly time-dependent, showing a distinct increase in internalization over time.

We used the Cell-Counting Kit- 8 (CCK-8) assay to evaluate the growth inhibition/cytotoxic effects of free DOX and HE/DOX on 143B and MG63 cells ( Figures 5C, D ). The dose-response curves showed that HE/DOX had a significantly stronger inhibitory effect on 143B and MG63 cells than free DOX. The half-maximal inhibitory concentrations (IC₅₀) of HE/DOX on 143B and MG63 cells were 1.5 and 1.97 μg/mL, respectively, which were significantly lower than the IC₅₀ of free DOX at 4.81 and 4.89 μg/mL, respectively. These results indicate that HE/DOX has an inhibitory effect on osteosarcoma and enhances the toxicity of free DOX. Additionally, we investigated the cytotoxic effects of BMSC-derived exosomes, liposomes, and HEs on cells ( Supplementary Figure S2 ). The dose-response curves indicated that BMSC-derived exosomes, liposomes, and HEs did not show significant cytotoxicity.

To evaluate the inhibitory effect of HE/DOX on tumors, we injected phosphate-buffered saline (PBS), HEs, free DOX, LP/DOX, or HE/DOX into orthotopic tumor-bearing mice every other day for a total of five injections. Figure 6A shows the treatment of the nude mouse tumor model. The dose of DOX administered to each mouse in the treatment group was 5 mg/kg. The results showed that mice in the PBS and HE groups had the fastest increase in tumor volume, whereas those in the free DOX, LP/DOX, and HE/DOX groups showed slow tumor growth ( Figures 6B–D ). Compared to the free DOX group, the LP/DOX group showed better tumor growth inhibition. Furthermore, HE/DOX showed a more successful therapeutic effect than LP/DOX. Additionally, no significant difference in body weight was observed between the groups during treatment ( Figure 6E ), indicating that the drugs had no significant side effect on mice.

The human BMSCs and osteosarcoma 143B and MG63 cell lines were purchased from the Peking Union Cell Bank, China. All cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 1% streptomycin, and 100 IU/mL penicillin. The BMSCs were cultured in exosome-free medium; specifically, the added fetal bovine serum was centrifuged at 100,000 × g for 16 h to remove the exosomes in the serum. All cells were cultured at 37°C in a 5% CO₂ incubator.

The exosomes were isolated using ultracentrifugation. Specifically, BMSCs were cultured in DMEM without exosomes for 48 h and the cell supernatant was collected. Debris or dead cells were removed by gradient centrifugation at 300 × g for 10 min, at 2000 × g for 20 min, and at 10,000 × g for 30 min. The supernatant was then placed in an ultracentrifuge tube and centrifuged at 100,000 × g for 70 min to obtain uniform exosomes. The concentrated exosomes were resuspended in PBS and stored at −80°C for further study. All centrifugation procedures were performed at 4°C.

The liposomes were prepared via film hydration and membrane extrusion. Specifically, L-α-phosphatidylcholine (Egg, Chicken) (EggPC) and cholesterol were uniformly mixed in chloroform at a ratio of 66:34. The mixture was dried overnight. The next day, the dried film was hydrated with PBS and fully dispersed using ultrasound. An extruder (Avanti Mini-Extruder) was used to extrude the dispersed post-solution subsequently through 400- and 200-nanometer polycarbonate membranes to obtain uniform nanoliposomes.

HEs were prepared using a simple extrusion method. Specifically, the isolated exosomes and prepared liposomes were fully mixed in PBS solution at a mass ratio of 1:5 and ultrasonically treated. The mixed solution was extruded through 400- and 200-nanometer polycarbonate membranes using an extruder to obtain monodisperse HEs.

DOX was encapsulated within HEs using the ammonium sulfate concentration gradient method. Briefly, exosomes were mixed with liposomes and resuspended in 240 mM ammonium sulfate solution. After sufficient vortexing and ultrasonic treatment, HEs encapsulated in an ammonium sulfate solution were obtained through 400- and 200-nanometer polycarbonate membranes. The solution was injected into a Slide-A-Lyzer Nutritional Cassette (MWCO 20 kDa) and dialyzed overnight at room temperature in PBS solution (pH 7.4) to remove ammonium sulfate outside the HE membrane and obtain a solution with a concentration gradient. The next day, a final concentration of 0.2–0.8 mg/mL DOX was added to the HE solution and then incubated at room temperature for 6 h. Subsequently, the solution was injected into the Slide-A-Lyzer Nutritional Cassette (MWCO 20 kDa). The obtained HE solution containing DOX was named HE/DOX and stored at −80°C for subsequent research. Liposomes loaded with DOX were prepared in the same manner without exosomes and were named LP/DOX.

The surface morphologies of LP/DOX and HE/DOX were observed using transmission electron microscopy (Hitachi, HT7700). The concentration and size distribution of LP/DOX and HE/DOX were measured using nanoparticle tracking analysis (Malvern, NanoSight NS300). The surface charges on LP/DOX and HE/DOX were characterized using dynamic light scattering (DLS, Malvern ZSP). Confocal microscopy (A1, Nikon) was used to confirm the successful encapsulation of DOX within HEs.

Positive exosomal markers (ALIX, CD63, and TSG101) and a negative marker (calnexin) in HEs and exosomes were analyzed using western blotting. RIPA lysis buffer (R0010, Solarbio) was used to extract proteins from the cells, HEs, and exosomes, and the total amount of protein was determined by BCA colorimetry. The extracted protein samples were then analyzed using polyacrylamide gel electrophoresis and western blotting. The following antibodies were used: ALIX (1:1000 dilution; Proteintech, USA), CD63 (1:1000 dilution; ABclonal Technology, China), TSG101 (1:1000 dilution; Proteintech), and calnexin (1:1000 dilution; ABclonal Technology).

A standard curve was drawn using a Multimode Microplate Reader (Thermo Fisher Technologies, USA) at excitation and emission wavelengths of 480 and 594 nm, respectively, and the concentration of DOX was determined.

As mentioned above, DOX was loaded into HEs using the ammonium sulfate concentration gradient method, and the concentration of DOX added was 0.2–0.8 mg/mL (0.1, 0.2, 0.4, 0.6, and 0.8 mg/mL). The following DOX loading efficiency (LE) and encapsulation efficiency (EE) formulas were used: LE = DOX_loaded/(HE_initial + DOX_loaded) × 100% and EE = DOX_loaded/DOX_initial × 100%, where DOX_loade is the mass of DOX loaded into HEs, HE_initial is the mass of HEs, and DOX_initial is the mass of DOX initially added to the solution.

We measured the drug release rates for HE/DOX and LP/DOX in different pH environments (pH 7.4 or 5.5). The concentrations of HE/DOX or LP/DOX were adjusted to 100 μg/mL. Next, 1 mL of HE/DOX or LP/DOX solution was added to a Slide-A Lyzer Dialysis Cassette (MWCO 20 kDa). The dialysis cassette was placed in 150 mL of PBS at pH 7.4 or 5.5. The samples were placed on a shaker at 37°C and shaken at a constant speed. At 3, 6, 12, 24, 48, and 72 h, 1 mL of the PBS dialysate was removed to determine the DOX content, and 1 mL of fresh PBS was added to replenish the sample.

Exosomes were labeled with PKH26 (Sigma-Aldrich). Specifically, 3 μL of PKH26 reagent and 500 μL of diluent C were added to 100 μg of exosome suspension and incubated at 37°C for 30 min. Free PKH26 was then removed using ultracentrifugation.

143B and MG63 cells were seeded onto 24-well culture slides at a density of 2 × 10⁴ cells/well. After 24 h of culture in an incubator, PKH26-labeled human BMSC-derived exosomes or liposomes were added. After 12 h of re-culture, all the cells were washed three times with PBS, fixed in 4% paraformaldehyde for 30 min, and then washed three times with PBS. Subsequently, 100 μL of DAPI staining solution (Beyotime, China) were added to each well for nuclear staining. Cell fluorescence was measured using an Axio Vert A1 fluorescence microscope (Carl Zeiss, Germany), and fluorescence intensity was calculated using the ImageJ software.

DiD (Thermo Fisher) was used to label exosomes. Brielfy, 10 μL of DiD reagent was added to 100 μg of exosomes and incubated at 37°C for 30 min. Free DiD was removed using ultracentrifugation. 143B and MG63 cells were seeded in 6-well plates and cultured for 24 h. DiD-labeled liposomes or BMSC-derived exosomes were added and co-cultured for 12 h. The cells were then digested with trypsin, the supernatant was removed by centrifugation, and the cells were washed twice with PBS. The cells were resuspended and analyzed using flow cytometry on a DxFLEX system (Beckman Coulter).

PKH26-labeled HE/DOX and LP/DOX were prepared following the same method.

Osteosarcoma 143B and MG63 cells were seeded onto glass slides in 24-well plates at a density of 2 × 10⁴ cells/well. PKH26-labeled HE/DOX were co-cultured with 143B and MG63 cells for 12 h. After washing three times with PBS, the cells were fixed with 4% paraformaldehyde for 30 min, washed three times with PBS, and stained with DAPI. Images were captured using a confocal microscope and quantitatively analyzed using the ImageJ software.

A CCK-8 (Sigma-Aldrich) assay was conducted to analyze the toxicity of free DOXand HE/DOX in the osteosarcoma cell lines. First, 2,000 cells were seeded in 96-well plates and cultured in DMEM medium containing 10% fetal bovine serum for 24 h. The next day, the medium was replaced with fresh medium containing BMSC-derived exosomes, liposomes, HEs, free DOX, or HE/DOX and cultured for 24 h. The CCK-8 reagent was added and incubated at 37°C for 2 h. Absorbance was measured at 450 nm using a Multimode Microplate Reader (Thermo Fisher Technologies).

Osteosarcoma 143B cells were inoculated into the right tibia of nude mice. After 10 days, the mice were randomly divided into four groups (N = 5) and injected with 150 μL of PBS, HEs, free DOX, LP/DOX, or HE/DOX every 2 d for five times. Free DOX, LP/DOX, and HE/DOX were administered at a dose of 5 mg/kg. The body weight of the mice was measured every 2 d, and the tumor volume was measured using a Vernier caliper. The formula for calculating tumor volume (V) is as follows:

where AP is the distance between the tumors on both sides of the kneecap and L is the length of the tumor in front of the tibia.

All statistical analyses were performed using GraphPad Prism 8 (GraphPad software). All experimental data were described as mean ± standard deviation. Differences between groups were analyzed using two-way analysis of variance. P values less than 0.05 were considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001). All the results were obtained from three replications of the experiments.