In brief, Fe₃O₄ nanoparticles were prepared by a high temperature approach through the thermal decomposition method. Oleic acid (OA) was chosen as both particle surface capping agent and co-solvent used together with 1-octadecene. The particle size shown in Figure 1A and size distribution shown in Supplementary Figure S1 in Supplementary Material reveal that the Fe₃O₄ particles were successfully synthesized with the average size of 11.3 ± 1.0 nm and a narrow size distribution. To render the Fe₃O₄ nanoparticles stabilized by OA water-soluble, asymmetric polyethylene glycols (PEGs) carrying a methoxyl group at one end and two phosphate groups at the other end, denoted as CH₃O-PEG-dp, respectively, were used to replace the OA ligand of the as prepared Fe₃O₄ nanoparticles based on the fact that the phosphate group has a higher binding affinity to Fe³⁺ than the carboxyl group from OA. The TEM image of water soluble Fe₃O₄ nanoparticles were shown in Figure 1B, the particle morphology and careful statistical studies (Supplementary Figure S2) reveal that the ligand exchange did not alter the size or the size distribution profiles of the OA-capped Fe₃O₄ nanoparticles.

The impact of the above ligand exchange on the properties of the Fe₃O₄ nanoparticles was further investigated with dynamic light scattering (DLS). As shown in Figure 1C, the OA-capped Fe₃O₄ nanoparticles in cyclohexane present single scattering peak locating at 13.2 nm, before converting into the aqueous phase. After ligand exchange, the hydrodynamic diameter (D_h) of Fe₃O₄ particles stabilized by CH₃O-PEG-dp in water apparently increased, reaching 16.2 nm. The DLS result on the one hand suggests that PEG-phosphate can effectively replace oleate ligand. On the other hand, the size distribution profiles remain nearly unchanged in comparison with those of the hydrophobic counterparts, suggesting the surface modification occurred without undesired aggregation.

The Ultraviolet−visible absorption spectrum of aqueous solution of Fe₃O₄ nanoparticles was also carried out. As shown in Figure 1D, the Fe₃O₄ nanoparticles exhibited a broad featureless absorption covering almost the entire visible region. However, the Fe³⁺ concentration dependent absorbance spectra perfectly followed the Lambert-Beer law with the correlation coefficient at any designated single wavelength, and no significant baseline drift was observed (Supplementary Figure S3), indicating the excellent aqueous stability of the particles.

The colloidal stability of Fe₃O₄ nanoparticles in aqueous solution is one of the prerequisites for in vivo bioapplications. As shown in Figure 1E, the temporal evolutions of the D_h and zeta potential reveal that, the current nanoparticles can remain stable in water for at least one week. It should be mentioned that the weak positive surface potential of particles might be attributed to the unsaturated bonded Fe³⁺. The satisfying colloidal stability provides the prerequisite for the biomedical applications of the nanoparticles.

As we know that MRI works based on computer-assisted imaging of relaxation signals of proton spins within the human body excited by radiofrequency waves in a strong magnetic field (Wang et al., 2018). And the relaxation of proton spins to their equilibrium states via two processes, namely longitudinal relaxation, characterization by a relaxation time T ₁, and transverse relaxation, characterized by a relaxation time T ₂. The MRI contrast agents principally work by shortening the T ₁ or T ₂ relaxation times of protons located nearby. The Fe₃O₄ nanoparticles like what we used in the experiment, can effectively reduce the T ₂ relaxation time and consequently produced negative enhancement effects on T ₂-weighted images.

Therefore, the MRI performance of Fe₃O₄ nanoparticles was evaluated on a 7T MRI scanner. As shown in the Supplementary Figure S4, Fe₃O₄ nanoparticles exhibited stronger T ₂ contrast enhancement effect with the increased concentration of Fe³⁺, which showed darker color in the T ₂-weighted imaging. By linear regression fitting of the experimental data, the transverse molar relaxivity r ₂ of nanoparticles was extracted as 85.0 mM⁻¹ s⁻¹ (Figure 2A). On the other hand, the T ₁ contrast enhancement ability of Fe₃O₄ nanoparticles was also measured. As shown in Figure 2B and Supplementary Figure S5, the longitudinal molar relaxivity r ₁ of nanoparticles was 0.71 mM⁻¹ s⁻¹ according to the slope of regression curve. Therefore, the r ₂/r ₁ ratio of the particles can be calculated as high as 119.7. The high r ₂ as well as high r ₂/r ₁ ratio makes them an ideal candidate for T ₂ contrast agents.

Any ideal contrast agent should be non- or low-toxicity to normal tissues. Therefore, the cytotoxicity of the current nanoparticles was evaluated through Cell Counting Kit-8 (CCK-8) cell proliferation assay on NIH 3T3 mouse embryonic fibroblast cells (Figure 2C). As a result, the Fe₃O₄ nanoparticles presented a comparable cell security, which did not show significant cytotoxicity when Fe³⁺ concentration reach 1 mM that was five orders of magnitude higher than the dose adopted for the following in vivo studies (10 mg/kg body weight). This lower cytotoxicity could partly be attributed to the satisfied biocompatibility of the current nanoparticles, which further highlighted the high biosafety of them.

Apart from the cytotoxicity of the nanoparticles, the hemolysis rate was investigated before the following in vivo experiments. As shown in Figure 2D, the pure water and the 1 × PBS solutions were set as the positive control and negative controls with 100 and 0% hemolysis rate, respectively. With respect to the Fe₃O₄ nanoparticles 1 × PBS solution, the calculated hemolysis rates were less than 1% at the Fe concentration of particles ranging from 0.01 to 10 mM, which significantly lower than the threshold that the hemolysis rate should be lower than 5% specified in the International Organization for Standardization (ISO) and the American Society for Testing and Materials (ASTM) (Rondon et al., 2020). These results further showed the biosafety and the blood compatibility of Fe₃O₄ nanoparticles in the blood stream, providing a prerequisite for the intravenous injection of the nanoparticles.

On the basis of the outstanding MRI performance and biosafety of the Fe₃O₄ nanoparticles in the in vitro experiments, it was employed to visualize the central nervous system damage caused by PMB in vivo. Six BALB/c mice were randomly divided into two groups (n = 3), and were intravenously (i.v.) injected with PMB (6 mg per kg weight) for one group or the same volume of 1 × PBS for anther group once a day for three days, and then one mouse from each group was used as a representative and subjected to the MR imaging, as shown in Figure 3A. Before injection of Fe₃O₄ nanoparticles, T ₂-weighted MR images of brain section was obtained and provided in the Figure 3B. It seems that the brain anatomic structures of mice injected with either PBS or PMB did not show significant damage.

In the previous studies, MR susceptibility-weighted imaging (SWI) is a high-resolution and 3D gradient-echo T ₂* MR technique, which is very sensitive to the magnetic susceptibility difference, particularly suitable for blood, hemorrhage, and iron storage sensitive imaging. Therefore, for clearly visualize the potential central nervous system damages caused by PMB injection, the brains of mice were further studied with a susceptibility-weighted imaging (SWI) sequence with the help of superparamagnetic Fe₃O₄ nanoparticles. The susceptibility-weighted images were acquired pre- and post-injection of the Fe₃O₄ nanoparticles. The nanoparticle–enhanced SWI were shown in Figures 3C,D, after the intravenous administration, the contrast of the cerebral vessels in both PMB- and PBS-treated mice were strongly enhanced. However, in comparison with the PBS-treated mouse, the nanoparticles extravasated into the brain parenchyma in the PMB-treated mouse, which resulted in a darker color surrounding the striatum in the image, identifying the potential BBB damaged sites. To further highlight these SWI signals, the post-contrast SWI was handled by pseudo-color processing, as shown in Figure 3E. Contrasting to the PBS-treated mouse, several discrete punctate signals distributed asymmetrically across brain regions of PMB-treated mouse, as indicated by the red arrows, which suggested that the structure of lenticulostriate artery may have been impaired by continuous administration of PMB.

For further confirming the pathological changes in the brains, the PMB- and PBS-treated mice were sacrificed after SWI. The brain tissues of them were extracted, cut into slices, and subjected to hematoxylin-eosin (H&E) staining for the histopathological analysis. The representative field of view of slices are shown in Figures 4A,B. Accordingly, in the brain slice of PMB-treated mouse, several neurons in dentate gyrus (DG) region (black frame) and cornu ammonus area 1 (CA1) (red frame) were atrophied with deeper stained color, and the cytoplasm and nucleus of them were poorly demarcated, as indicated by the black arrows. In contrast, the neurons in the same regions of PBS-treated mice were arranged regularly with the normal morphological structure. The boundary between cytoplasm and nucleus of the cells can be clearly differentiated. This PMB-caused brain impairment can also be quantitatively characterized by the abnormal cells in the hippocampal region of brain slices (Supplementary Figure S6).

In addition, the terminal deoxynucleotidyl transferase (TdT) mediated dUTP Nick End Labeling (TUNEL) assay was also carried out to check whether the nervous cells underwent extensive DNA degradation, which is the striking feature of apoptosis. Almost no nuclei were stained by TUNEL in the brain of PBS-treated mice. However, the TUNEL-positive cells were appeared surrounding the CA1, CA2, and DG regions of the brain extracted from PMB-treat mouse, shown in the left panel of Figures 4C,D, respectively, indicating that PMB may impair the activity of neurons within the hippocampus of the brain. The quantitative statistical results of TUNEL-positive cells in hippocampus also confirmed this conclusion (Supplementary Figure S7). Moreover, the blood vessels surrounding the caudate putamen (CPu) area of the brain were impaired in the PMB-treated mice. This result suggested that PMB may lead to the destruction of the neurons and BBB in brain, which is consistent with the imaging results in Figure 3, indicating that the Fe₃O₄ enhanced SWI in the current work is precise and reliable.

Ferric trichloride hexahydrate (FeCl₃•6H₂O, 99.0%) was purchased from Shandong Xiya Chemical Industry Company (Shandong, China). OA and NaOH were purchased from Sigma-Aldrich. Tetrahydrofuran (THF) was purchased from Fuchen (Tianjin) Chemical Reagent, Co., Ltd (Tianjin, China). Cyclohexane was purchased from Tianjin Damao Chemical Reagent Factory (Tianjin, China). Analytical grade chemicals such as acetone and 1-octadecene were purchased from Beijing Chemical Reagent, Co., Ltd (Beijing, China). All the above chemicals were used without further purification. Polymyxin B (≥6,000 u/mg) sulfate was purchased from Beijing Solarbio Science & Technology Co., Ltd. The BALB/c mice were purchased from Beijing Vital River Laboratories (Beijing, China). The sodium pentobarbital was purchased from Baxter Healthcare Corporation.

OA-capped Fe₃O₄ nanoparticles were synthesized through thermal decomposition method. In brief, 3.6 g (4 mmol) of prepared iron oleate and 3.39 g (4 mmol) of OA were dissolved in 25 ml of 1-octadecene. The resultant mixture was heated to 310°C with a rate of 3.3°C min⁻¹, and then maintained at 310°C for 30 min under nitrogen protection. The preparation was terminated by cooling the reaction mixture down to room temperature. The resultant OA-capped Fe₃O₄ nanoparticles were precipitated by acetone, collected by magnetic separation, washed with acetone several times, and finally re-dispersed in THF or cyclohexane for further experiments.

Typically, 20 mg of dp-PEG-OCH₃ was dissolved in 5 ml of THF containing 2 mg OA-capped Fe₃O₄ nanoparticles. Then, the reaction mixture was heated to 40°C and kept at this temperature overnight under stirring. After that, the resultant NP-PEG nanoparticles were precipitated by cyclohexane, washed with cyclohexane for three times, and then dried under vacuum at room temperature. To remove the excess free PEG ligand, the NP-PEG nanoparticles dissolved in MilliQ water were further purified through ultrafiltration with 30 kDa MWCO centrifugal filter (Millipore YM-100) for 5 cycles at 3,000 g.

TEM images of the nanoparticles were taken on a JEM-2100 transmission electron microscope at an acceleration voltage of 200 kV. The particle size was determined by averaging at least 25 nanoparticles per sample. DLS measurements were carried out at 298 K with Nano ZS (Malvern) equipped with a solid state He-Ne laser (λ = 632.8 nm) for determining the hydrodynamic size of the different particles. The iron contents in different systems were determined with 1,10-phenanthroline through absorption spectroscopy after the particles were eroded with concentrated hydrochloric acid. The absorption of water solution of nanoparticles was also analyzed through absorption spectroscopy at room temperature on the UV-Vis spectrophotometer (Thermo, MULTISKAN GO).

The relaxivity measurements were carried out on a 7.0 T Bruker Biospec animal MRI instrument (BioSpec70/20USR). A series of aqueous solutions containing nanoparticles with different Fe concentration were prepared in 200 μl Eppendorf tube for MR studies. The detailed parameters for T ₁ and T ₂ measurements were set as follows: TE = 5.01 ms, TR = 300 ms, MTX = 200 × 200, and FOV = 40 × 40 mm² for T ₁-weighted imaging; TE = 40 ms, TR = 3,000 ms, MTX = 200 × 200, and FOV = 40 × 40 mm² for T ₂-weighted imaging.

Mouse embryonic fibroblast cell line NIH 3T3 was cultured in a medium of Dulbecco’s Modified Eagle Medium (DMEM) high glucose and F-12K nutrient mixture (1: 1) supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin solution (100×) at 37°C under a 5% CO₂ atmosphere.

Cell viability was determined by CCK-8 assay. 3T3 Cells were seeded into a 96-well cell culture plate by 5 × 10³ per well under 100% humidity, and then cultured at 37°C for 24 h. The Fe₃O₄ nanoparticles with a series of Fe³⁺ concentration were added to the wells respectively and incubated with the cells for another 24 h at 37°C. Subsequently, the supernatants containing the free formulations were decanted. CCK-8 (100 μl, 10 mg/ml) was added to each well and incubated for 3 h at 37°C. The optical density of each well at 450 nm was recorded on a microplate reader (Thermo, Multiskan GO).

Hemolysis test was conducted according to the previous report (Zhang et al., 2007; Love et al., 2012). Briefly, a 2 ml blood sample was mixed with 6 ml of phosphate-buffered saline (PBS) and centrifuged at 1,500 rpm for 15 min. The supernatant was drawn off and five subsequent washes were carried out before diluting the sample to a final volume of 1 ml in PBS. The red blood cells were then diluted 1: 4 into 1 × PBS solutions (negative control); water (positive control); nanoparticle solutions (in PBS) at varying concentrations (tested samples).

Due to the broad absorption of nanoparticles, the absorption background was composed of PBS containing the nanoparticle at corresponding concentrations without red blood cells. The samples were left at room temperature in the dark for 4 h at 37°C and then centrifuged at 3,000 rpm for 5 min. After centrifuging, the supernatant was transferred to a 96-well plate and the absorbance at 541 nm was recorded.

The hemolysis rate was calculated by the following equation: Hemolysis rate =   D t - D b D pc - D nc × 100 % where D_t, D_b, D_nc, and D_pc were the absorbance of the tested sample, the background sample, the negative control and the positive control, respectively.

6-week-old male BALB/c mice were used in this study. All mice were maintained in ventilated cages at a temperature of 20–25°C and a relative humidity of 30–50% under a 12 h light/dark cycle and were given free access to food and water.

Three mice were intravenously injected with PBS solution of PMB (1 mg/ml) at a dosage of 6 mg per kg weight once a day for 3 days, while the other three mice were intravenously injected with pure PBS solution with the same volume. The mice were subjected to MR scanning for brain imaging in the 4^th day.

The brain MR imaging was carried out on a 7.0 T MRI system (Bruker BioSpec70/20 USR) equipped with a rat head surface coil to receive signals. The MR images pre- and post-contrasted with nanoparticles were acquired using a gradient echo SWI sequence. The imaging parameters are set as follows: repetition time (TR) = 469.8 ms, echo time (TE) = 8.07 ms, field of view = 20 × 20 mm, matrix = 200 × 200, and slice thickness = 1 mm T ₂WI sequence were also obtained pre-contrast. The detailed parameters for T ₂WI were set as follows: TR = 2,366 ms, TE = 35 ms, field of view = 20 × 20 mm, and matrix = 200 × 200.

During the MRI experiments, the animals were anesthetized with 2% isoflurane in oxygen-mixed air via a facemask. Rectal temperature was maintained at 37 ± 1°C. The nanoparticles were injected into the tail vain at the injection dose of 10 mg Fe per kg body weight, and the final concentration of Fe was 1.5 mg Fe mL⁻¹ in PBS.

After MR imaging, the mice were sacrificed, and the brain tissues were extracted, fixed with formalin, and embedded in paraffin. After that, the brain tissues were sliced with an ultrathin semiautomatic microtome to obtain coronal sections of 3 µm. The adjacent slices were selected for H&E and TUNEL staining, respectively.

All animal experiments reported herein were performed according to a protocol approved by the Peking University Institutional Animal Care and Use Committee.