First, Fe₃O₄ (0.1 g) was dispersed in ethanol (60 mL), deionized water (10 mL) and ammonium hydroxide solution (1.0 mL) and sonicated for 40 min. Tetraethyl orthosilicate (2.0 mL) was added, and the mixture was stirred for 24 h to obtain Fe₃O₄@SiO₂. Fe₃O₄@SiO₂ (20 mg) was dispersed in ethanol solution and functionalized with 3-aminopropyltriethoxysilane (APTES, 1.5 mL) for 24 h. The precipitate was collected with a magnet and washed several times to obtain Fe₃O₄@SiO₂-NH₂. For Ce6 coupling, Ce6 (4.0 mg), EDC (23.0 mg) and NHS (13.8 mg) we. re dissolved in dimethyl sulfoxide (4 mL), and Fe₃O₄@SiO₂-NH₂ (16 mL, 50 mg mL^-1) was added to the mixed solution and stirred for 24 h. The reaction products were separated with a magnet and washed several times to obtain Fe₃O₄@SiO₂-Ce6 (FSC) for further use.

The microstructure of the FSC nanorobots was observed by transmission electron microscopy (TEM) on a JEM-2100F instrument. The valence states and elemental composition of the FSC nanorobots were characterized by X-ray photoelectron spectroscopy (XPS, ESCAlab250 instrument) and Fourier transform infrared spectroscopy (FTIR), respectively. The potential and hydrodynamic particle size were measured by dynamic laser scattering on a Zeta Sizer system (Nano ZS90, Malvern Instruments Ltd.). The magnetic properties were evaluated by a vibrating sample magnetometer. The UV‒vis–NIR absorption spectra of the Fe₃O₄@SiO₂-NH₂, Ce6 and Fe₃O₄@SiO₂-Ce6 nanoparticles were measured with a Shimadzu UV-3600 ultraviolet‒visible spectrophotometer.

DPBF experiment: A DMF solution (8 mM, 40 μL) containing DPBF with an FSC nanorobot dispersion (150 μg mL^-1, 2.96 mL) was prepared and irradiated with low-intensity ultrasound (1.0 MHz, 1.0 W cm^-2, 50% duty cycle), and the change in the absorption value at 419 nm was measured every 2 min with an ultraviolet spectrophotometer for 10 min in total. ESR experiment: TEMP (200 mM, 20 μL) was added to the FSC nanorobot dispersion (150 μg mL^-1, 20 μL) and Ce6 solution (10 μg mL^-1, 20 μL). After ultrasound irradiation (1.0 MHz, 1.0 W cm^-2, 50% duty cycle, 2 min), the ¹O₂ signal was immediately detected by ESR. For control purposes, the Ce6+TEMP, Ce6+TEMP + US, FSC + TEMP, and FSC + TEMP + US groups were tested simultaneously.

The magnetic nanorobot dispersion (2 mg mL^-1, 1 mL) was placed in a confocal dish on the platform of the magnetic actuation system. Then, a rotating magnetic field (2 Hz) was applied to make the nanorobots aggregate at one point. For the locomotion of the nanorobot collectives, different biological fluids were deployed in the custom-designed ‘R’, ‘O’, and ‘S’ channels of acrylic plates. The magnetic nanorobot dispersion (2 mg mL^-1, 20 μL) was added to the channels separately. By adjusting the moving platform, the magnetic nanorobot could be directed along a trajectory. The images were recorded by an operating microscope (DOM-1001, RWD Life Science, Shenzhen, China).

Mouse ovarian epithelial cancer cells (ID8) (Shanghai Institute of Cells, Chinese Academy of Sciences) were precultured in 96-well plates (1 × 10⁴ cells/well) for 12 h and allowed to adhere. The medium was then removed from each well, and the cells were washed twice with phosphate buffer (PBS). Subsequently, Fe₃O₄@SiO₂-Ce6 nanoparticles at different concentrations (0, 100, 200, 400 and 800 μg mL^-1) were dispersed in high-glucose DMEM containing 10% foetal bovine serum (FBS), added to each well, and incubated with the cells for 24 h and 48 h, respectively. The cells were then washed with PBS, and a standard CCK-8 assay (100 μL, V_CCK8: V_DMEM = 1: 9) was performed to detect cell viability, which was measured on a microplate reader at a wavelength of 450 nm after incubation for 1–2 h.

ID8 cells (1 × 10³ cells/well) were seeded into confocal-specific plates overnight. After the cells had completely adhered to the wall, the cell medium was replaced, and the cells were treated with Fe₃O₄@SiO₂ and FSC for 12 h. Subsequently, the cells were further exposed to US irradiation (1.0 MHz, 1.0 W cm^-2, 50% duty cycle, 5 min) and incubated with DCFH-DA (1:1,000 dilution) for another 30 min. Then, the treated cells were washed 3 times with PBS and imaged by fluorescence microscopy. Correspondingly, the cells were collected and analysed quantitatively for intracellular green fluorescence intensity by flow cytometry (FACSCalibur; BD Biosciences).

ID8 cells were seeded in several 6-well plates for overnight. After that, the cells were incubated with Fe₃O₄@SiO₂ and FSC (400 μg mL^-1) for 12 h. Next, the cells were stimulated with or without US (1.0 Wcm^-2, 1.0 MHz, 50% duty cycle) for 5 min. After that, the cells were washed with PBS, and were frozen and thawed 3 times. The liquid supernatants were collected for GSH/GSSG detection using the assay kit.

ID8 cells were seeded in culture dishes for overnight and randomly divided into four groups (control, Fe₃O₄@SiO₂, FSC, FSC + US). The cells were cultured for 12 h and then subjected to US irradiation for 5 min. Subsequently, the cells were rinsed with PBS. Cell samples treated for quantitative protein analysis with a BCA protein assay kit. Then, the MDA assay kit reagents were added according to the manufacturer’s instructions, and the mixtures were heated in a boiling water bath at 100 °C for 15 min. After centrifugation at 12,000 rpm for 10 min, the absorbance of the supernatant was measured with a microplate reader at a wavelength of 532 nm. The MDA content in each sample was calculated in terms of the unit weight of protein content, and the results are presented as μmol MDA/mg protein.

ID8 cells were seeded in 6-well plates (1 × 10⁵/well) and cultured for 24 h. The cells were divided into the following experimental groups: ① simple diffusion + US group; ② magnetic regulation + US group. For the simple diffusion + US group, the FSC nanorobot dispersion was applied to the nontumor cell area in the middle of the round dish and allowed to freely disperse to the tumour cell area. In the magnetic regulation + US group, the FSC nanorobot dispersion was also applied to the nontumor cell area in the middle of the dish, and the dispersion was then aggregated by a magnetic actuation system and directed to the tumour cell area. US irradiation (1.0 MHz, 1.0 W cm^-2, 50% duty cycle, 5 min) was applied to the tumour cell area, and the cells were incubated for 12 h. All treated cells were digested with trypsin and stained with the Calcein-AM/PI Kit for fluorescence microscopy.

For all the experiments conducted, we ensured that there were at least three replicates included. The data obtained from these experiments were presented as the mean value along with the standard error of the mean (SEM). To assess the differences between two groups, we employed the independent sample t-test. Additionally, for comparisons among multiple groups, we used one-way analysis of variance (ANOVA) followed by the Tukey post-test. In our analysis, statistical significance was denoted as p < 0.05, and this was further specified as *p < 0.05, **p < 0.01, ***p < 0.001.

A structural diagram of an FSC nanorobot is shown in Figure 1A. TEM images illustrate that the FSC nanorobots had robust spherical morphology, uniform size, and good dispersibility (Figure 1B). The average hydrodynamic size was approximately 460 nm (Figure 1C), with no significant change observed after half a month in a physiological environment, indicating excellent dispersion stability (Figure 1D). Fourier transform infrared spectroscopy revealed a characteristic peak at 1,697.533 cm^-1 for the -C=O of Ce6 in FSC (Figure 1E). X-ray photoelectron spectroscopy (XPS) analysis was conducted to examine the chemical composition and state of FSC. As shown in Figure 1F, obvious diffraction peaks were observed at 711.3, 532.5, 399.3, 284.6 and 103.3 eV, which can be attributed to Fe2p, O1s, N1s, C1s and Si2p, providing further confirmation that the FSC consist of Fe, O, N, C, and Si. The UV absorption spectrum of FSC exhibited characteristic absorption peaks of Ce6 at 401 nm and 656 nm, confirming successful coupling (Figure 1G). The potential of Fe₃O₄@SiO₂ was 24.30 ± 2.2 mV. Upon coupling with Ce6, the potential of FSC changed to 1.40 ± 0.39 mV (Figure 1H), indicating that Ce6 was coupled with Fe₃O₄@SiO₂. The vibrating magnetization curve results demonstrated that FSC exhibited a saturation magnetization of 38.45 emu g-¹ in the high-field region of 2000 Oe, facilitating effective magnetic separation under the influence of an applied magnetic field and confirming its desirable superparamagnetic properties (Figure 1I).

We further examined the in vitro ROS production of Fe₃O₄@SiO₂-Ce6 (FSC nanorobot). The quantity of ROS in the four groups was analyzed. The ESR spectra (Figure 2A) showed significant ¹O₂ generation under US irradiation in the Ce6 + US group and the FSC nanorobot + US group compared to the groups not subjected to US irradiation. Moreover, within a specific concentration range, the production efficiency of ¹O₂ was found to be correlated with the power intensity of US (Figure 2B). Simultaneously, DPBF acts as a ROS probe: the generated ROS can react with DPBF, resulting in a decrease in the characteristic absorption at 410 nm in the UV‒Vis spectrum. The absorption peak of DPBF decreased with time under US irradiation (1.0 MHz, 1.0 W cm^-2, 50% duty cycle), indicating favorable ROS generation performance (Figure 2C).

Next, to verify that magnetic FSC nanorobots can achieve precise target locomotion, we conducted experiments with a specifically designed permanent magnetic actuation system. The magnetic actuation system consists of four permanent magnets and a rotating motor, which can generate a rotating magnetic field and gradient magnetic field simultaneously (Figure 3A). As shown in Figure 3B, the nanorobots were first arranged into small strips under the action of a static magnetic field after being placed in the magnetic actuation system. When the permanent magnets were rotated, under the combined action of the rotating magnetic field and gradient magnetic field, the magnetic poles of the nanorobot chains attracted each other to aggregate into collectives. In addition, we designed a three-axis movable platform for positioning the samples. By changing the relative position between the platform and the magnetic actuation system, targeted locomotion of the magnetic nanoparticle collectives was achieved.

We further conducted locomotion experiments on magnetic FSC nanorobots in different biological solutions. As shown in Figure 3C, the nanorobot solution was deployed in the ‘R’, ‘O’, and ‘S’ channels containing normal saline, PBS and DMEM, respectively. Under the control of a magnetic field, magnetic FSC nanorobot collectives can achieve trajectory locomotion. All these results indicated that the magnetic FSC nanorobots could achieve target locomotion in biological fluids, providing a foundation for subsequent targeted therapy of tumor cells.

We further examined the production of intracellular ROS after different treatments by fluorescence microscopy and flow cytometry. The fluorescence signal derived from 2′,7′-dichlorofluorescin diacetate (DCFH-DA) was used as a probe of cellular oxidative stress levels. Figure 4A shows minimal fluorescence in the control group, a modest increase in green fluorescence in the Fe₃O₄@SiO₂ and FSC nanorobot groups, which may be due to the role of Fe₃O₄ in the Fenton reaction in tumor microenvironment. And green fluorescence was significantly increased in the FSC nanorobot group irradiated with US. This observation was substantiated by the flow cytometry analyses shown in Figure 4B. The fluorescence intensity trend was consistent with that observed by fluorescence microscopy, suggesting the essential role of Ce6 in sonodynamic therapy. Our findings indicate that FSC nanorobots can modulate the tumour microenvironment, elevate ROS levels, and thereby increase therapeutic efficacy.

Glutathione (GSH) and oxidized glutathione (GSSG) are major components of the redox homeostasis system in cells (Lei et al., 2022; Zhu et al., 2021). As displayed in Figure 4C, the GSH levels of the cells in the Fe₃O₄@SiO₂ and FSC groups exhibited a slight decrease, which may be due to the low ROS production deplete the GSH. Moreover, FSC + US significantly reduced the GSH compared with other groups, revealing the efficient GSH depletion ability. The ratio of GSH/GSSG is considered a powerful index of oxidative stress (Wang D. et al., 2023; Liu et al., 2023). The results showed that in the Fe₃O₄@SiO₂ and FSC treatment groups, the GSH/GSSG ratio was slightly varied, while it was greatly decreased in FSC + US group (Figure 4D), altogether confirming the superiority of FSC sonodynamic therapy. Malondialdehyde (MDA) can also be used as a stable marker to evaluate lipid peroxidation (Zhu et al., 2021). The MDA level in each group was opposite to GSH, FSC + US increased significantly (Figure 4E), which illustrates that FSC nanorobots induce sonodynamic therapy associated with lipid peroxidation.

Inspired by the excellent ability of the FSC nanorobot to generate ROS under low-intensity ultrasound excitation, we next evaluated the biosafety and the in vitro SDT efficacy of the FSC nanorobots in ID8 cells. Initially, the cytotoxicity of the FSC nanorobots was evaluated using the CCK-8 assay. At a concentration of 800 μg mL^-1, cell viability remained approximately at 95%, highlighting the favorable biocompatibility of FSC nanorobot within the range of 0–800 μg mL^-1 in ID8 cells (Figures 5A,B). Subsequently, we investigated the efficacy of SDT in vitro. The experimental findings demonstrated a decrease in cell viability in the presence of the FSC nanorobot upon US stimulation (Figure 5C). Consistent results were corroborated through calcein-AM/PI staining. Fluorescence microscopy revealed a noteworthy increase in the red signal intensity following US stimulation (Figure 5D). Collectively, these results suggest that after incubation with FSC nanorobots for 24 h, tumour cells can be effectively killed through US stimulation (1.0 MHz, 1.0 Wcm^-2, 50% duty cycle, 5 min).

We further studied the efficacy of magnetic FSC nanorobots for targeted tumour cell therapy. Two experimental groups were treated with or without magnetic actuation. As shown in Figure 6A, the cell dishes were divided into tumour cell areas and blank areas. The FSC nanorobot dispersion was added to the blank area. In the control group, no magnetic field was applied, and the nanorobots relied on passive diffusion to reach the tumour cell area, simulating the traditional systemic drug administration strategy. For the magnetic actuation group, after the FSC nanorobot dispersion was added to the blank area, a magnetic field was applied to aggregate the nanorobots to one point and then direct the nanorobot collectives to the tumour region. Both groups were subsequently subjected to low-intensity ultrasound of the tumour cell area. The results indicated that only a few FSC nanorobots arrived at the tumour region in the control group, while in the magnetic actuation group, nearly all the FSC nanorobots reached the tumour region (Figure 6B). The SDT results further confirmed the enrichment of Ce6 in the tumour area by the magnetic actuation strategy. As shown in Figure 6C, more tumour cell death occurred under US irradiation. The enhanced sonodynamic effect highlights the ability of FSC nanorobots to facilitate targeted drug delivery and increase the therapeutic efficacy of SDT.