Ultrasmall iron oxide nanoparticles cause significant toxicity by specifically inducing acute oxidative stress to multiple organs

Iron oxide nanoparticles have been approved by food and drug administration for clinical application as magnetic resonance imaging (MRI) and are considered to be a biocompatible material. Large iron oxide nanoparticles are usually used as transversal (T2) contrast agents to exhibit dark contrast in MRI. In contrast, ultrasmall iron oxide nanoparticles (USPIONs) (several nanometers) showed remarkable advantage in longitudinal (T1)-weighted MRI due to the brighten effect. The study of the toxicity mainly focuses on particles with size of tens to hundreds of nanometers, while little is known about the toxicity of USPIONs. We fabricated Fe3O4 nanoparticles with diameters of 2.3, 4.2, and 9.3 nm and evaluated their toxicity in mice by intravenous injection. The results indicate that ultrasmall iron oxide nanoparticles with small size (2.3 and 4.2 nm) were highly toxic and were lethal at a dosage of 100 mg/kg. In contrast, no obvious toxicity was observed for iron oxide nanoparticles with size of 9.3 nm. The toxicity of small nanoparticles (2.3 and 4.2 nm) could be reduced when the total dose was split into 4 doses with each interval for 5 min. To study the toxicology, we synthesized different-sized SiO2 and gold nanoparticles. No significant toxicity was observed for ultrasmall SiO2 and gold nanoparticles in the mice. Hence, the toxicity of the ultrasmall Fe3O4 nanoparticles should be attributed to both the iron element and size. In the in vitro experiments, all the ultrasmall nanoparticles (< 5 nm) of Fe3O4, SiO2, and gold induced the generation of the reactive oxygen species (ROS) efficiently, while no obvious ROS was observed in larger nanoparticles groups. However, the ·OH was only detected in Fe3O4 group instead of SiO2 and gold groups. After intravenous injection, significantly elevated ·OH level was observed in heart, serum, and multiple organs. Among these organs, heart showed highest ·OH level due to the high distribution of ultrasmall Fe3O4 nanoparticles, leading to the acute cardiac failure and death. Ultrasmall Fe3O4 nanoparticles (2.3 and 4.2 nm) showed high toxicity in vivo due to the distinctive capability in inducing the generation of ·OH in multiple organs, especially in heart. The toxicity was related to both the iron element and size. These findings provide novel insight into the toxicology of ultrasmall Fe3O4 nanoparticles, and also highlight the need of comprehensive evaluation for their clinic application.

Ultrasmall iron oxide nanoparticles (USIONPs) exhibit high longitudinal relaxivity (r 1 ), presenting a good candidate for positive T 1 -weighted imaging [17]. In addition, the NPs could be efficiently cleared from kidney after i.v. administration once the size was controlled below 5.5 nm [18]. Although USPIONs have attracted great attention, the bio-safety is not completely understood and it is quite necessary to study the toxicity, since various kinds of nanoparticles demonstrate size-dependent toxicity in vitro and vivo.
Gold nanoparticles with size smaller than 5 nm are much more toxic to human embryonic stem cells [19] and mouse fibroblasts [20]. Silica also indicated size-dependent toxicity [21,22] and size-dependent activity that only ultrasmall NPs (< 5 nm) could activate the mononuclear cells [23]. The toxicity might be related to the large surface, quick dissolution, and good penetration [24]. To date, the toxicity of USPIONs is debatable and needs to be further clarified.
We evaluated the toxicity of USPIONs with three sizes (2.3, 4.2, and 9.3 nm) in vitro and in vivo using gold and silica NPs as controls. The result was pretty impressive that small-sized USPIONs (2.3 and 4.2 nm) were highly toxic and showed lethal dose of 100 mg/kg, while the mice treated with gold NPs, silica NPs, or large-sized iron oxide NPs (9.3 nm) were alive. It is reasonable to deduce that the strong toxicity of USPIONs is related to both the ultrasmall size and the iron element. Here, we investigated the possible lethal factors associated with the size and iron element, including the biodistribution of the NPs, USPIONs induced generation of ROS, and ferroptosis based on Fe 2+ and radicals (·OH).

Construction and characterization of the Fe 3 O 4 , SiO 2 , and Au NPs
Different-sized Fe 3 O 4 NPs were synthesized using a modified solvothermal reaction with diethylene glycol, sodium acetate, and trisodium citrate as solvent, alkali source, and stabilizer, respectively. The particle size could be controlled by adjusting the reaction temperature and time. TEM and DLS were used to characterize the morphology and size of synthesized Fe 3 O 4 NPs. As shown in Fig. 1a was observed in the acidic medium due to the pH sensitive dissolution of the iron oxide (Fig. 1d). The release rate was also influenced by the size of the NPs that smaller Fe 3 O 4 NPs resulted in quicker release because of the relatively higher specific surface (Fig. 1d).
We then studied the Fenton reaction of H 2 O 2 catalyzed by Fe 3 O 4 NPs. The generated ·OH group was determined using coumarin-3-carboxylic acid (3-CCA) as an indicator, which was non-fluorescent and could be hydroxylated by ·OH to generate fluorescent 7-hydroxycoumarin [25]. Different-sized Fe 3 O 4 NPs were incubated with H 2 O 2 at pH 5.0 for 2 h and then the fluorescence was determined. As shown in Fig. 1e , and H 2 O 2 , which were subsequently catalyzed to highly toxic ·OH to destroy the cells.

The uptake of the NPs in cells
Most nanoparticles are unstable in ion containing media, readily forming aggregates due to the change of surface charge. We determined the size change of the NPs by ultrafiltration and elemental analysis. As shown in Fig. 2a, b, after incubation in the DMEM media for 2 h, the NPs showed significant aggregation that the amount of NPs with size > 10 nm was over 40%. The aggregation was also found in 9.3 nm Fe 3 O 4 NPs (Fig. 2c). Although the aggregation happened, we could also observe the presence of NPs with size < 2.9 nm. Fe 3 O 4 , SiO 2 and Au NPs showed size-dependent internalization in MCF-7 cells (Fig. 2d-f ). After incubation at a fixed concentration of 50 μg/mL for 2 h, higher Fe/Si/Au conctents were observed in larger NPs treated cells. This phonomenon has been reported before [26].
The intracellular uptake of the different-sized NPs was characterized by LSCM. As presented in Additional file 1: Fig. S4, the 2.3 nm and 4.2 nm Fe 3 O 4 NPs could enter the nuclei efficiently after incubation with cells for 0.5 h, while the 9.3 nm Fe 3 O 4 NPs was observed outside the nuclei. Similar phenomenon was also found in SiO 2 and Au NPs groups (Additional file 1: Figs. S5, S6). These results indicates only NPs with size < 5 nm can enter the nuclei since the nuclei pore is ~ 5 nm.

The ultrasmall NPs induced generation of ROS in cells
We investigated the NPs induced ROS in MCF-7 cells using DCFH-DA as a probe, which was non-fluorescent and could be oxidized by ROS to produce fluorescent DCF [27,28]. In Fe 3 O 4 groups, we could observe bright fluorescence in 2.3 nm and 4.2 nm Fe 3 O 4 NPs treated MCF-7 cells (Fig. 3a, b). The strength of the fluorescence increased as the incubation time extended. By contrast, only extremely weak signal was detected in 9.3 nm Fe 3 O 4 NPs group. Considering the relatively lower internalization of ultrasmall NPs, the elevated ROS should be attributed to the stimulation of the ultrasmall NPs. As for SiO 2 group, size-and time-dependent fluorescence was also observed in the cells (Fig. 3c, d). ROS was hard to be detected in 7.3 nm SiO 2 group. Similar phenomenon was observed in Au NPs treated cells that fluorescence was only observed in 4.1 nm NPs groups, while no obvious fluorescence was detected in 7.2 nm and 11 nm Au NPs treated cells (Fig. 3e, f ). For the three different NPs, there was significant difference of the fluorescence intensity between the NPs of < 5 nm and > 5 nm groups, indicating the remarkable ROS induction ability of the ultrasmall NPs (< 5 nm). We also noticed that the DCF fluorescence in 2.3 nm and 4.2 nm Fe 3 O 4 NPs was located within the whole cells including the nucleus. This is probably because only ultrasmall NPs could enter the nucleus efficiently.

Identification of the generated ROS in cells
We further investigated the type of ROS induced by the different kinds of nanoparticles. We firstly detected H 2 O 2 using dihydrorhodamine 123 (DHR123) as indicator [29]. As shown in Fig. 4a, fluorescence signal could be seen in all the groups, indicating H 2 O 2 could be induced by Fe 3 O 4 , SiO 2 , and Au NPs. We also noticed that the fluorescence intensity of Fe 3 O 4 NPs was slightly stronger than the other two groups (Fig. 4b), which might be due to the superior capability in inducing oxidative stress. We then measured the O 2 − in cells using a WST-1 assay, which could be oxidized to form orange formazan [30]. As shown in Fig. 4c NPs was endocytosed via an endolysosomal pathway (Fig. 4e). Although the Fe 3 O 4 NPs were found in the nucleus (Additional file 1: Fig. S4), no fluorescence could be detected in the nucleus. The possible reason for this phenomenon might be the acidic environment of the lysosomes, which promoted the dissolution of the Fe 3 O 4 NPs and accelerated the generation of ·OH.

In vivo toxicity of the Fe 3 O 4 NPs
The in vivo toxicity of the NPs was studied by intravenous injection of SiO 2 (2.4 nm) and Au (4.1 nm) and To explore the possible in vivo toxic mechanism, we tried to understand the distribution of the NPs when the mice were dead. We elevated the dosage to 500 mg/ kg. At such a dosage, the mice will be killed 5 min after the intravenous injection. We could observe the extremely high content of iron element in heart in 2.3 nm Fe 3 O 4 NPs group, which was 2.35 and 3.33 times to that of 4.2 and 9.3 nm groups, respectively (Fig. 5c). In comparison, the distributions of SiO 2 and Au NPs in heart were relatively low (Additional file 1: Figs. S8, S9, and S10). Hence, the damage to heart is probably the leading cause of the sudden death. When the injection rate was slow down by injection for 4 times with each interval for 5 min (at total dosage of 100 mg/kg), only one mouse out of five was killed, indicating the  Table S1).

In vivo generation of ROS
We then evaluated the in vivo ROS levels after intravenous injection of different-sized Fe 3 O 4 , SiO 2 and Au NPs. The serum and main organs were collected 2 h post injection and the ROS was determined using DCFH-DA as a probe. As shown in Fig. 5d-i, there were remarkable increases of ROS in serum, heart, liver, and spleen in 2.3 nm Fe 3 O 4 group, which were 2.16, 1.61, 4.03, and 1.9 times as high as that of control. The repeated injection would cause the further increase of the ROS (Additional file 1: Fig. S11). We also notice that the ROS level in 2.3 nm Fe 3 O 4 group was much higher than that of 9.3 nm Fe 3 O 4 group. For SiO 2 and Au NPs, the ROS level of ultrasmall group in liver, heart, and spleen was also higher than larger counterparts. However, the ROS levels in SiO 2 and Au NPs groups were notably lower than that in Fe 3 O 4 group. This result might be due to the quick elimination of ROS by ROS-scavenging enzymes, including SOD, catalase, and glutathione peroxidase. These enzymes help the cells to regulate the O 2 − and H 2 O 2 at a relatively low level and prevent them from being damaged. However, some kinds of ROS, such as ·OH and peroxynitrite (ONOO -), cannot be scavenged by any known enzymes in vivo, resulting in high toxicity. Considering the quick scavenge of O 2 − and H 2 O 2 , it is reasonable to deduce that the ROS in 2.3 nm Fe 3 O 4 group shall be the ·OH.
To explore the type of ROS in vivo after administration of the NPs, we investigated the concentration of ·OH in serum and main organs using 3-CCA as indicator. After administration of 2.3 nm Fe 3 O 4 NPs for 2 h, the ·OH concentrations in serum, heart, liver, spleen and lung increased significantly (Fig. 6). In particular, the ·OH concentration in heart increased by 2.56 times. The extremely high ·OH level should be attributed to the high accumulation of 2.3 nm Fe 3 O 4 NPs in heart and resulting high Fe 2+ concentration. It has been reported that , liver (f), spleen (g), lung (h) and kidney (i) after intravenous injection of different-sized Fe 3 O 4 , SiO 2 and Au NPs (100 mg/kg) for 2 h. The ROS was determined using DCFH as an indicator. All of the data were presented as the mean ± SD (n = 3). P values: *P < 0.05, **P < 0.01 and ***P < 0.001 compared with the corresponding control group the heart was sensitive to the ROS, since the oxidative stress can drive heart failure, cardiac arrhythmias, and sudden cardiac death by regulating the ion channels [33,34]. Hence, the ultrasmall NPs caused death is probably due to the cardiac oxidative stress. We could observe the prominent cardiotoxicity and also the pulmonary toxicity in the H&E stained images as illustrated by the necrotic regions and acute inflammatory cells (Additional file 1: Fig. S12). No noticeable increase of ·OH concentration was observed for the SiO 2 and Au NPs groups, indicating that the generation of ·OH was induced by the Fe 3 O 4 NPs. As for 9.3 nm Fe 3 O 4 NPs, the generated ·OH was negligible because of the low distribution in heart, limited capability in inducing ROS, and the slow release of Fe 2+ .

Inhibition of the ROS
GSH is an antioxidant that reduces free radicals with the help of glutathione peroxidase 4 (GPX 4). To prove the toxicity was mainly caused by ROS, the mice were pretreated with GSH. As shown in Additional file 1: Fig. S13, the levels of malondialdehyde (MDA) in heart, liver and serum increased by 1.01, 3.41 and 0.63 times, respectively, after the exposure of 2.3 nm Fe 3 O 4 NPs, indicating the serious lipid peroxidation. Upon the pretreatment of GSH, the malondialdehyde (MDA) levels in heart, liver and serum decreased by 29.3%, 26.4, and 12.5%, respectively. These results suggest the protective effect of GSH to Fe 3 O 4 NPs induced lipid peroxidation. The GSH could also decrease the levels of ALT and AST (Additional file 1: Fig. S7) and increase the survival rate of the Fe 3 O 4 -2.3 NPs treated mice (Additional file 1: Table S1). These results imply that the toxicity of 2.3 nm Fe 3 O 4 NPs is caused by the oxidative stress and can be attenuated by antioxidant.

Generation of the ROS
Ulrtasmall NPs with size < 5 nm, including Fe 3 O 4 , SiO 2 and Au NPs, can stimulate the cells to generate ROS, while NPs with size over 5 nm do not show remarkable ROS induction effect. Hence, the generation of ROS is dependent on the size of the NPs. It is reported that ultrasmall Au NPs (4.5 nm) can directly penetrate into cell cytoplasm to promote robust ROS production and to activate the NLRP3 inflammasome for Caspase-1 maturation and interleukin-1β production, while only low level of ROS is induced by large nanoparticles [35]. Ultrasmall (< 10 nm in diameter) silica NPs can induce the generation of ROS, and the ROS level increases significantly several hours before cell death [36]. The mitochondrial respiratory chain is a potential source of ROS, and reduced mitochondria membrane induced by the ultrasmall NPs potential leads to increased generation of ROS All of the data were presented as the mean ± SD (n = 3). P values: *P < 0.05, **P < 0.01 and ***P < 0.001 compared with the corresponding control group [37]. As such, various ultrasmall NPs can stimulate the cells to generate the ROS.

The toxicity of the ultrasmall Fe 3 O 4
The toxicity of the ultrasmall Fe 3 O 4 is believed to be due to the ferroptosis, a type of programmed cell death accompanied by a large amount of iron accumulation and characterized by the accumulation of lipid peroxides. In both in vitro and in vivo experiment, the ·OH is detected in Fe 3 O 4 group instead of SiO 2 and gold groups. ·OH is highly reactive and toxic, which cannot be decomposed by any known intrinsic enzymes.  (Fig. 7). The quick release of Fe 2+ from ultrasmall Fe 3 O 4 NPs facilitates the catalytic reaction. The toxicity is inhibited by the pretreatment of GSH, indicating the toxicity was mainly caused by ROS. To prove the role of released iron in ultrasmall Fe 3 O 4 NPs induced ferroptosis, deferoxamine can be used as a chelating agent to remove excess iron.
The toxicity is associated with the injection rate. When the total dose was split into 4 doses with 5 min for each interval, the toxicity is significantly reduced. This result may be because when the Fe 3 O 4 NPs are injected slowly, the released iron will be bonded by the ferritin to regulate the cellular iron concentration at a reasonable level and avoid ferroptosis. However, when the Fe 3 O 4 NPs are burst injected, the released iron remarkably increases and overloads, causing serious peroxidation. As such, the iron oxide is administrated at a slow rate in clinic. For example, administer Feraheme as an undiluted intravenous injection delivered at a rate of up to 1 mL/s (30 mg/s).

Conclusions
We found that the ultrasmall Fe 3 O 4 with sizes of 2.3 and 4.2 nm was highly toxic, while no obvious toxicity was detected for the larger Fe 3 O 4 (9.3 nm). To explore the toxicological mechanism, different-sized SiO 2 and Au NPs we synthesized as control and the toxicity was studied. The in vitro results indicated that intracellular ROS was only induced by NPs could increase the in vivo ROS and ·OH levels efficiently due to the special capability in inducing ROS and the subsequent Fenton reaction, resulting in the oxidative stress to multiple organs. Among these, heart exhibited highest ·OH level because of the high cardiac distribution of ultrasmall Fe 3 O 4 NPs, which might be the principal cause of the mice death, since the cardiac oxidative stress can drive cardiac arrhythmias and sudden cardiac death. This study not only illustrates the potential toxicological mechanism of ultrasmall Fe 3 O 4 Fig. 7 Possible causes of toxicity of USPIONs (< 5 nm): (1) high biodistribution of the ultrasmall NPs in heart; (2) ultrasmall NPs (< 5 nm) trigger the generation of ROS; (3) quick release of Fe 2+ from ultrasmall NPs induces the ferroptosis based on hydroxyl radicals (·OH) nanoparticles, but also presents a potential application of ultrasmall Fe 3 O 4 in cancer therapy owing to the high toxicity.

MCF-7 (Human breast carcinoma cancer cell line) cells
were purchased from the cell bank of Chinese Academy of Sciences (Shanghai, China). All the cells were cultured in RPMI-1640 medium (Gibco, USA) supplemented with 10% FBS (Sijiqing Biologic, Zhejiang, China), streptomycin (100 U mL −1 ) and penicillin (100 U mL −1 ) at 37 °C in a humidified atmosphere with 5% CO 2 . Male ICR mice (4-5 weeks, 20 ± 1 g) were provided by Laboratory Animal Center of Jiangsu University and all animals were used in accordance with the protocols approved by Guide for the Care and Use of Laboratory Animals.

Synthesis of Fe 3 O 4 , SiO 2 , and gold NPs
The 9.3 nm Fe 3 O 4 nanoparticles were synthesized using a DEG-mediated solvothermal reaction [38]. Briefly, anhydrous FeCl 3 (0.648 g, 4 mmol) and trisodium citrate dihydrate (0.48 g, 1.6 mmol) were dissolved in DEG (40 mL). Afterwards, the mixture was heated to 80 °C under vigorously stirring to form a clear solution. With the addition of anhydrous NaAc (1.0 g, 12.0 mmol), the mixture was transferred into a Teflon-lined stainless-steel autoclave (100 mL capacity). Then, the autoclave was heated at 230 °C for 6 h. After cooling down to room temperature, the sample was centrifuged and washed with ethanol for several times. The resultant black sample was dispersed in water for further use. The 2.3 nm and 4.2 nm Fe 3 O 4 nanoparticles was obtained at 200 °C for 6 h and 10 h, respectively.
To synthesize 4.2 nm silica nanoparticles, 0.43 mmol of TMOS was added into 10 mL of 0.002 M ammonia aqueous solution [39]. After stirring at room temperature for 20 h, 0.21 mmol of PEG-silane was added and the mixture was stirred at room temperature for another 12 h. After reaction at 80 °C overnight, the solution was cooled down to room temperature and dialyzed (Pierce, molecular weight cutoff = 10,000 Da) in 2000 mL of DI water. The dialysis media was replaced with fresh water every day for total 6 days. As for 2.4 nm silica nanoparticles, the concentration of TMOS was reduced to 0.011 mmol. To prepare 7.3 nm nanoparticles, 0.43 mmol of TMOS was added into 10 mL of 0.002 M ammonia aqueous solution, then the mixture was stirred at 80 °C instead of room temperature for 20 h, while other conditions were identically same as that of 4.2 nm.
Gold nanoparticles (Au NPs) were prepared according to the method developed by Frens [40]. Briefly, 4 mL of 12.7 mM chloroauric acid was added to 49 mL of boiling water, followed by the addition of trisodium citrate solution (5 mL, 38.8 mM). 15 s later, the solution turned to be red and the reaction was quenched in ice bath to obtain 11 nm Au NPs. 4.1 nm Au NPs were prepared by borohydride reduction of the gold salt. Briefly, 1.968 mL of 12.7 mM chloroauric acid was added to 96.432 mL of H 2 O. The solution was kept in ice bath for 10 min. Then, 1 mL of 50 mM trisodium citrate solution was added. 1 min later, 0.6 mL of 0.5 M iced sodium borohydride solution was added to reduce the chloroauric acid to generate 4.1 nm Au NPs. To synthesize 7.2 nm Au NPs, 20 mL of 5 mM hydrogen tetrachloroaurate (III) trihydrate was added into a 3-neck flask and diluted to 90 mL with water. After adding β-cyclodextrin at final concentration of 25 mM, the solution was heated to boiling followed by the rapid addition of sodium citrate (40 mM, 10 mL). 10-40 s later, the solution changed to a murky gray color, suggesting the nucleation of the gold. Within 2 min, Au NPs were formed as indicated by the color of burgundy red. All Au NPs were washed with H 2 O and concentrated by centrifugation. To observe the morphology and structure, the NPs stock solutions were diluted and dropped on TEM grid. After air-drying, the samples were characterized by JEOL transmission electron microscope (JEM-2100) at an acceleration voltage of 200 kV. The size distributions of the nanoparticles were calculated by measuring 150 particles randomly selected from the TEM images.

Particle aggregation in culture medium
@@@The aggregation of the nanoparticles was determined in the cell-free growth medium at 37 °C for 2 h. Different sized NPs were separated by ultrafiltration centrifuge tube (< 10 kDa, with pore of 2.9 nm) and 10 nm filter membrane (GVS polycarbonate membrane). Total Fe/Si/Au analysis of dispersions and their particle-sizing fractions was conducted by inductively coupled plasma optical emission spectroscopy (ICP-AES). The element analysis thus revealed the percentages in the dispersion below 2.9 nm, between 2.9 and 10 nm, and above 10 nm (considered aggregation) in size.

Cellular uptake of the NPs
Three kinds of NPs with different sizes were incubated with MCF-7 cells for 2 h at a concentration of 50 μg/ mL. Then, the cells were collected with the help of trypsin and homogenized. After digestion with aqua regia, the concentrations of Fe/Si/Au were measured by ICP-AES.
The uptake of the NPs was also characterized by laser scanning confocal microscopy (LSCM). The NPs were labeled with doxorubicin (DOX), a fluorescent drug that can enter into the nucleus. DOX labeled NPs were prepared by mixing 1 mL of DOX (1 mg/mL) with 2 mL of NPs (2 mg/mL) overnight. Unbounded DOX was removed by centrifugation and washed with water for three times. Then, the DOX labeled NPs were cultured with MCF-7 cells at DOX concentration of 10 μg/mL. After incubation for 0.5 h or 1 h, the culture media was removed and the cells were washed with PBS for three times. The cells were stained with LysoGreen (1 μM)and Hoechst 33342 (10 µg/mL) and observed with LSCM.

Cytotoxicity of the Fe 3 O 4 NPs
Cytotoxicity of the Fe 3 O 4 NPs was assessed using 3-[4,5-dimethylthialzol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) method. MCF-7 cells were seed into 96-well plate at a density of 5 × 10 3 cells/well and cultured overnight. Then, Fe 3 O 4 NPs with various concentrations were added and incubated for another 24 h. After removing the NPs, the cells were washed with PBS and the cell viability was determined by MTT using an ELISA plate reader (Thermo Multiskan Spectrum, USA). For GSH pretreatment groups, the cells were incubated with GSH for 4 h at a concentration of 300 μg/mL. Then, the culture media was removed and the cells were treated with H 2 O 2 and Fe 3 O 4 NPs.

Determination ROS in MCF-7 cells
The intracellular ROS levels were determined using 2′,7′-dichlorofluorescein diacetate (DCFH-DA) as indicator, which was non-fluorescent and could be oxidized to generate fluorescence. MCF-7 cells (5 × 10 4 ) were seeded in single-well plates with glass bottom. After 24 h incubation, the culture medium was replaced with 1 mL of culture medium containing NPs at a concentration of 500 μg/mL and the cells were incubated for another 2 h and 6 h. Then, the medium was removed and the cells were rinsed with PBS for three times. After staining with DCFH-DA (10 μM, 30 min) and Hoechst 33,342 (10 μg/mL, 15 min), the cells were fixed with 4% paraformaldehyde (30 min) and imaged using a laser confocal microscope (Delta Vision TM Elite, GE).

The in vivo toxicity of NPs
To evaluate the toxicity, the male ICR mice (4-5 weeks, 20 ± 2 g) were randomly divided the into five groups (with 5 mice for each group, n = 5) to receive SiO 2 (2.4 nm), Au (4.1 nm) and Fe 3 O 4 (2.3, 4.2, and 9.3 nm) NPs. The NPs were intravenously injected at a dosage of 100 mg/kg. The number of mice surviving over a period of 24 h was recorded. The toxicity of NPs is usually related with the injection rate. We also investigated the toxicity of Fe 3 O 4 2.3 nm by intravenously injecting for 4 times with total dosage of 100 mg/kg. The interval for each injection was 5 min. For GSH pretreatment group, the GSH was orally administrated at a dosage of 800 mg/kg. 2 h later, the Fe 3 O 4 NPs were injected.
To study the damage of the Fe 3 O 4 NPs to the organs, the 2.3, 4.2, and 9.3 nm Fe 3 O 4 NPs were injected for 7 continuous days with the dosage of 80 mg/kg for the 1st day and 40 mg/kg for 2ed-7th day. At the 8th and 14th day, mice were sacrificed and the organs were harvested, sectioned, and stained with H&E for histological analyses.

Serum biochemical analysis
After intravenous injection of Fe 3 O 4 NPs (100 mg/kg) for 6 h, the mice were anesthetized and the whole blood was sampled from mice eyes. 30 min later, the samples were centrifuged at 3000 rpm for 15 min and the serum was obtained. The enzymes, including ALT and AST ALP were determined using clinical chemistry analyzers (Beckman Coulter, AU5800). For GSH pretreatment group, the GSH was orally administrated at a dosage of 800 mg/kg. 2 h later, the Fe 3 O 4 NPs were injected.