Biological Fate of Fe3O4 Core-Shell Mesoporous Silica Nanoparticles Depending on Particle Surface Chemistry

The biological fate of nanoparticles (NPs) for biomedical applications is highly dependent of their size and charge, their aggregation state and their surface chemistry. The chemical composition of the NPs surface influences their stability in biological fluids, their interaction with proteins, and their attraction to the cell membranes. In this work, core-shell magnetic mesoporous silica nanoparticles (Fe3O4@MSN), that are considered as potential theranostic candidates, are coated with polyethylene glycol (PEG) or 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) lipid bilayer. Their biological fate is studied in comparison to the native NPs. The physicochemical properties of these three types of NPs and their suspension behavior in different media are investigated. The attraction to a membrane model is also evaluated using a supported lipid bilayer. The surface composition of NPs strongly influences their dispersion in biological fluids mimics, protein binding and their interaction with cell membrane. While none of these types of NPs is found to be toxic on mice four days after intravenous injection of a dose of 40 mg kg−1 of NPs, their surface coating nature influences the in vivo biodistribution. Importantly, NP coated with DMPC exhibit a strong accumulation in liver and a very low accumulation in lung in comparison with nude or PEG ones.


Introduction
In the past two decades, nanoparticles (NPs) for medical applications have been investigated by numerous researchers. The NPs based theranostic agents combine, in a unique formulation, tracking,

Synthesis and Characterization of Fe 3 O 4 @MSN
Fe 3 O 4 @MSN core shell NPs were prepared by sol-gel reaction, following the optimized protocol previously described [22]. Fe 3 O 4 @MSN NPs present a magnetic core of 18 nm and a mesoporous silica shell of ca. 40 nm radius (Figure 1). Particles with an average primary diameter of ca. 100 nm were grafted in-situ by the addition of PEG 2000 groups. The primary diameter of PEG Fe 3 O 4 @MSN NPs observed on TEM images was the same than for the native ones. The coating with DMPC lipid bilayers was performed after the synthesis, washing and drying of the native NPs. TEM images allow for the measurement of the primary diameter and the observation of the shape of the NPs (Figure 1a,b), while cryoTEM images revealed also the presence of DMPC lipid coating on Fe 3 O 4 @MSN (Figure 1c).

Synthesis and Characterization of Fe3O4@MSN
Fe3O4@MSN core shell NPs were prepared by sol-gel reaction, following the optimized protocol previously described [22]. Fe3O4@MSN NPs present a magnetic core of 18 nm and a mesoporous silica shell of ca. 40 nm radius (Figure 1). Particles with an average primary diameter of ca. 100 nm were grafted in-situ by the addition of PEG2000 groups. The primary diameter of PEG Fe3O4@MSN NPs observed on TEM images was the same than for the native ones. The coating with DMPC lipid bilayers was performed after the synthesis, washing and drying of the native NPs. TEM images allow for the measurement of the primary diameter and the observation of the shape of the NPs (Figure 1a,b), while cryoTEM images revealed also the presence of DMPC lipid coating on Fe3O4@MSN (Figure 1c). CryoTEM images revealed that the lipid bilayers measured 5 nm in thickness, so the primary diameter of DMPC Fe3O4@MSN is 10 nm higher than the native and PEG Fe3O4@MSN. The dispersion and stability of the NPs in suspension in different media were investigated by the measurement of the hydrodynamic diameter using Nano ZS apparatus ( Figure 2). The NPs' surface chemistry influences their dispersion and stability when they were suspended in ethanol, HBS 5 (Hepes buffer 20 mM pH = 7.4 containing 5 mM NaCl), HBS 150 (Hepes buffer 20 mM pH = 7.4 containing 150 mM NaCl), complete cell culture medium Roswell Park Memorial Institute RPMI, and HBS 150 containing fetal calf serum (FCS) 10% ( Figure 2). In ethanol, native and PEG Fe3O4@MSN present a hydrodynamic diameter (HD) of 144.5 ± 2.7 nm and 169.5 ± 0.9 nm, respectively ( Figure 2a). These diameters are higher than the primary diameters observed by TEM, and the NPs stayed none aggregated in suspension after 2 and 15 h. Differently, DMPC Fe3O4@MSN suspended in ethanol presented a HD of 269.9 ± 15.9 nm, which increased to 643.5 ± 19.9 nm after 2 h and 1246 ± 48.2 nm after 15 h. The increase of HD of lipid coated Fe3O4@MSN in ethanol may be explained by a partial solubilization of the lipids in this solvent. In HBS 5, native Fe3O4@MSN presents a HD of 125.5 ± 1.2 nm at 0 h, 159.9 ± 1.9 nm after 2 h and 187.5 ± 3.3 nm after 15 h. PEG Fe3O4@MSN presents a HD of 141.8 ± 3.9 nm after 0 h, 202.1 ± 2.8 nm after 2 h and 232.5 ± 3.1 nm after 15 h. DMPC Fe3O4@MSN presents a HD of 167.1 ± 3.2 nm after 0 h, 171.6 ± 2.0 nm after 2 h and 177.1 ± 2.3 nm after 15 h. Thus, in HBS 5, native, PEG-grafted and lipid-coated NPs remained individually dispersed during 15 h after initial dispersion (Figure 2b). In HBS 150 buffer, native and PEG Fe3O4@MSN formed aggregates larger than 1 µm 2 h after initial dispersion ( Figure 2c). However, coating with DMPC lipid bilayers was found to stabilize NPs suspension in buffer with high ionic strength (HBS 150). In this buffer, DMPC Fe3O4@MSN presented a HD of 171.6 ± 0.1 nm after 0 h, 198.0 ± 0.2 nm after 2 h, and 270.9 ± 0.2 nm after 12 h. In cell culture medium (RPMI containing 10% FCS) and HBS 150 containing 10% FCS, lipid coated DMPC Fe3O4@MSN were found always stable 15 h after dispersion (Figure 2d,e). The good dispersion of lipid coated DMPC Fe3O4@MSN in HBS 150 buffer and HBS 150 containing 10% FCS was also observed on cryoTEM images (data not shown). In RPMI and HBS 150 containing 10% FCS, native NPs rapidly stabilized, presenting a HD of 128.2 ± 3.0 nm and 110.8 ± 2.0 nm at 0 h, in RPMI and HBS 150 10% FCS, respectively. However, the CryoTEM images revealed that the lipid bilayers measured 5 nm in thickness, so the primary diameter of DMPC Fe 3 O 4 @MSN is 10 nm higher than the native and PEG Fe 3 O 4 @MSN. The dispersion and stability of the NPs in suspension in different media were investigated by the measurement of the hydrodynamic diameter using Nano ZS apparatus ( Figure 2). The NPs' surface chemistry influences their dispersion and stability when they were suspended in ethanol, HBS 5 (Hepes buffer 20 mM pH = 7.4 containing 5 mM NaCl), HBS 150 (Hepes buffer 20 mM pH = 7.4 containing 150 mM NaCl), complete cell culture medium Roswell Park Memorial Institute RPMI, and HBS 150 containing fetal calf serum (FCS) 10% ( Figure 2). In ethanol, native and PEG Fe 3 O 4 @MSN present a hydrodynamic diameter (HD) of 144.5 ± 2.7 nm and 169.5 ± 0.9 nm, respectively ( Figure 2a). These diameters are higher than the primary diameters observed by TEM, and the NPs stayed none aggregated in suspension after 2 and 15 h. Differently, DMPC Fe 3 O 4 @MSN suspended in ethanol presented a HD of 269.9 ± 15.9 nm, which increased to 643. 5 Figure 2b). In HBS 150 buffer, native and PEG Fe 3 O 4 @MSN formed aggregates larger than 1 µm 2 h after initial dispersion ( Figure 2c). However, coating with DMPC lipid bilayers was found to stabilize NPs suspension in buffer with high ionic strength (HBS 150). In this buffer, DMPC Fe 3 O 4 @MSN presented a HD of 171.6 ± 0.1 nm after 0 h, 198.0 ± 0.2 nm after 2 h, and 270.9 ± 0.2 nm after 12 h. In cell culture medium (RPMI containing 10% FCS) and HBS 150 containing 10% FCS, lipid coated DMPC Fe 3 O 4 @MSN were found always stable 15 h after dispersion (Figure 2d,e). The good dispersion of lipid coated DMPC Fe 3 O 4 @MSN in HBS 150 buffer and HBS 150 containing 10% FCS was also observed on cryoTEM images (data not shown). In RPMI and HBS 150 containing 10% FCS, native NPs rapidly stabilized, presenting a HD of 128.2 ± 3.0 nm and 110.8 ± 2.0 nm at 0 h, in RPMI and HBS 150 10% FCS, respectively. However, the stabilization of PEG Fe 3 O 4 @MSN in protein-containing media took more time, presenting a HD close to the primary diameter only after 15 h. The zeta potentials (ZP) were measured in all media, except in ethanol (Figure 2f). Native Fe 3 O 4 @MSN presented a ZP of −18.9 ± 1.4 mV and −31.6 ± 0.8 mV, in HBS 150 and HBS 5, respectively. A decrease (in absolute value) of zeta potential to −16.1 ± 3.0 mV in HBS 150 and −28.4 ± 2.6 mV in HBS 5 by grafting by PEG groups was observed, and to −4.51 ± 1.1 mV in HBS 150 and −13.4 ± 0.3 mV in HBS 5, by coating by DMPC lipid bilayers. However, in protein-containing media, the zeta potential remained the same for native and lipid coated Fe 3 O 4 @MSN. The stabilization of native Fe 3 O 4 @MSN in RPMI and HBS 150 10% FCS is potentially due to the formation of a protein corona [23]. The stabilization of PEG Fe 3 O 4 @MSN in protein-containing media was longer due to the stealth effect of the PEG groups at the NPs surface impairing protein adsorption to the NPs surface.  [23]. The stabilization of PEG Fe3O4@MSN in protein-containing media was longer due to the stealth effect of the PEG groups at the NPs surface impairing protein adsorption to the NPs surface. To conclude, the presence of a strong ionic strength or proteins did not modify the physicochemical behavior of DMPC Fe3O4@MSN, which are therefore the only ones to remain dispersed and stable in biological media. To conclude, the presence of a strong ionic strength or proteins did not modify the physicochemical behavior of DMPC Fe 3 O 4 @MSN, which are therefore the only ones to remain dispersed and stable in biological media.

Changes at the Interface between NPs and a Supported Lipid Bilayer
The measurement of the interactions between NPs and a supported membrane model has been performed using a quartz crystal microbalance with dissipation monitoring (QCM-D). A SiO 2 coated 5 MHz quartz crystal was used. After injection of EPC (egg phosphatidyl choline) SUV (small unilamellar vesicles) suspension, a SLB (supported lipid bilayer) was formed, presenting a frequency shift of −26 Hz and a dissipation shift less than 0.5 × 10 −6 ; typically recorded for a supported bilayer [28]. The surface coated with the SLB was considered as the reference surface, and frequency and dissipation shifts induced by NPs deposition were measured from this surface. NPs were put in contact by flowing a NPs suspension on the top of the SLB in HBS 150 10% FCS medium ( Figure 3). In this protein-containing medium, only very little NPs deposition on the lipid bilayer surface for PEG-Fe 3 O 4 @MSN and nativeFe 3

Changes at the Interface between NPs and a Supported Lipid Bilayer
The measurement of the interactions between NPs and a supported membrane model has been performed using a quartz crystal microbalance with dissipation monitoring (QCM-D). A SiO2 coated 5 MHz quartz crystal was used. After injection of EPC (egg phosphatidyl choline) SUV (small unilamellar vesicles) suspension, a SLB (supported lipid bilayer) was formed, presenting a frequency shift of −26 Hz and a dissipation shift less than 0.5 × 10 −6 ; typically recorded for a supported bilayer [28]. The surface coated with the SLB was considered as the reference surface, and frequency and dissipation shifts induced by NPs deposition were measured from this surface. NPs were put in contact by flowing a NPs suspension on the top of the SLB in HBS 150 10% FCS medium ( Figure 3). In this protein-containing medium, only very little NPs deposition on the lipid bilayer surface for PEG-Fe3O4@MSN and nativeFe3O4@MSN was observed. DMPC-Fe3O4@MSN interaction with the SLB surface resulted in more important frequency and dissipation shifts than those obtained for PEG-Fe3O4@MSN and native-Fe3O4@MSN. The kinetics of deposition were very different depending of the surface coating. The deposition of DMPC-Fe3O4@MSN occurred rapidly during the NPs flowing, inducing a frequency shift of −7.55 ± 1.49 Hz and a dissipation shift of 3.37 ± 0.71 × 10 −6 after 1 h deposition. The maximum frequency shift observed was −9.81 ± 3.11 Hz and the maximum dissipation shift was 4.46 ± 0.43 × 10 −6 after 10 h deposition. PEG-Fe3O4@MSN began to deposit at 4 h 07 ± 27 min, with a maximum frequency shift of −6.88 ± 0.75 Hz and a maximum dissipation shift of 1.49 ± 0.30 × 10 −6 was obtained after 10 h deposition. NativeFe3O4@MSN began to deposit at 6 h 16 ± 33 min, with a maximum frequency shift of −3.69 ± 3.03 Hz and a maximum dissipation shift of 0.84 ± 0.30 × 10 −6 was obtained after 10 h deposition. The deposition rates of the NPs were dependent of the aggregation state and the presence of a protein corona. DMPC-Fe3O4@MSN, which were well-dispersed and not influenced by the presence of proteins (

Time-Dependent Internalization of NPs in Hep-G2 cells
The uptake of native, PEG and DMPC Fe3O4@MSN by Hep-G2 cells (human hepatocyte carcinoma) after different incubation times were followed by TEM from ultrathin sections (Figure 4). After 3 h incubation period, native Fe3O4@MSN were observed aggregated near the cell membrane (Figure 4a). PEG Fe3O4@MSN were less aggregated than the native Fe3O4@MSN but were not observed in the cells (Figure 4d).
Conversely, DMPC Fe3O4@MSN were observed inside Hep-G2 cells after 3 h of exposure (Figure 4g). After 6 h incubation period, native Fe3O4@MSN were observed individually in vesicular structures in the cytoplasm (Figure 4b) and some PEG Fe3O4@MSN were internalized also ( Figure  4e). DMPC Fe3O4@MSN were observed as small groups of NPs in vesicular structures (Figure 4h). These small groups were then observed for natives, PEG and DMPC Fe3O4@MSN after 24 h of NPs exposure with Hep-G2 cells (Figure 4c,f,i, respectively). This suggests a faster internalization of DMPC Fe3O4@MSN in human hepatic cells.

In Vivo Biocompatibility
To estimate the biocompatibility of MSN, mice were injected intravenously with nanoparticles at a concentration of 40 mg kg −1 . This concentration is high and corresponds to concentration already used to evaluate the acute toxicity of NPs in mice [36]. Four days after the injection of NPs, mice were sacrificed and organs, urines and blood were collected. Histological analyses were performed. No noticeable structural modifications on liver, kidney and spleen were observed (Figure 5a). Moreover, as shown in Figure 5b-e, no significant differences between control and treated mice for the biomarkers such as creatinine (kidney), interleukine-6 (IL-6) and tumor necrosis factor alpha (TNF-α) (systemic inflammation), and alanine aminotransferase (ALT) (liver) were observed, confirming the functional integrity of organs. These results demonstrated the biocompatibility of the nanoparticles under consideration in this work.

In Vivo Biocompatibility
To estimate the biocompatibility of MSN, mice were injected intravenously with nanoparticles at a concentration of 40 mg kg −1 . This concentration is high and corresponds to concentration already used to evaluate the acute toxicity of NPs in mice [36]. Four days after the injection of NPs, mice were sacrificed and organs, urines and blood were collected. Histological analyses were performed. No noticeable structural modifications on liver, kidney and spleen were observed (Figure 5a). Moreover, as shown in Figure 5b-e, no significant differences between control and treated mice for the biomarkers such as creatinine (kidney), interleukine-6 (IL-6) and tumor necrosis factor alpha (TNF-α) (systemic inflammation), and alanine aminotransferase (ALT) (liver) were observed, confirming the functional integrity of organs. These results demonstrated the biocompatibility of the nanoparticles under consideration in this work.

In Vivo Biodistribution
The amount of silicium has been quantified using inductively coupled plasma-mass spectrometry (ICP-MS) analysis. No significant elevation of silicium amount was detected in the kidneys and urines 4 days after injection. The quantities of silicium in the blood, the liver, the spleen, and the lungs were different depending of the NPs coating. These differences suggested changes in biodistribution and in pharmacokinetics of the NPs, due to the presence of the coating. Silicium amount in the liver 4 days after injection of DMPC Fe3O4@MSN is elevated compared to the level in liver of mice treated by native and PEG Fe3O4@MSN (Figure 6a). This shows a strong capture of

In Vivo Biodistribution
The amount of silicium has been quantified using inductively coupled plasma-mass spectrometry (ICP-MS) analysis. No significant elevation of silicium amount was detected in the kidneys and urines 4 days after injection. The quantities of silicium in the blood, the liver, the spleen, and the lungs were different depending of the NPs coating. These differences suggested changes in biodistribution and in pharmacokinetics of the NPs, due to the presence of the coating. Silicium amount in the liver 4 days after injection of DMPC Fe 3 O 4 @MSN is elevated compared to the level in liver of mice treated by native and PEG Fe 3 O 4 @MSN (Figure 6a). This shows a strong capture of DMPC Fe 3 O 4 @MSN by the liver.
However the accumulation in the spleen is almost similar between the 3 batches of MSN. Importantly, the lung is a biological barrier that is crucial to avoid and we can see that the amount of silicium in the lung is as low as the background level when mice were treated with DMPC Fe 3 O 4 @MSN. In contrast, this level is higher after injection of native and PEG Fe 3 O 4 @MSN.
In the blood, 2 h after injection, the levels of silicium are equally elevated between the different batches ( Figure 6b). In contrast, 6 h after injection, we can see that the amount of circulating nanoparticles for natives and PEG Fe 3 O 4 @MSN decrease slowly, while DMPC Fe 3 O 4 @MSN do not circulate anymore (Figure 6b). This could be correlated to a rapid and strong uptake by the liver. We note that 24 h after the injection, PEG Fe 3 O 4 @MSN are still well present in blood, suggesting a better circulation time due to the coating. DMPC Fe3O4@MSN by the liver. However the accumulation in the spleen is almost similar between the 3 batches of MSN. Importantly, the lung is a biological barrier that is crucial to avoid and we can see that the amount of silicium in the lung is as low as the background level when mice were treated with DMPC Fe3O4@MSN. In contrast, this level is higher after injection of native and PEG Fe3O4@MSN.
In the blood, 2 h after injection, the levels of silicium are equally elevated between the different batches ( Figure 6b). In contrast, 6 h after injection, we can see that the amount of circulating nanoparticles for natives and PEG Fe3O4@MSN decrease slowly, while DMPC Fe3O4@MSN do not circulate anymore (Figure 6b). This could be correlated to a rapid and strong uptake by the liver. We note that 24 h after the injection, PEG Fe3O4@MSN are still well present in blood, suggesting a better circulation time due to the coating.

Discussion
The presented data demonstrate that native, DMPC and PEG Fe 3 O 4 @MSN stay dispersed and stable during 15 h when diluted in a medium having a low ionic strength (HBS 5 mM NaCl, pH 7.4).
When the ionic strength was increased to a more physiological amount of NaCl (HBS 150 mM NaCl, pH 7.4), native and PEG Fe 3 O 4 @MSN rapidly aggregated. The aggregation may be explained by a compression of the electrical double layer [23]. The coating with PEG is supposed to reduce the aggregation of NPs in suspension, by adding a steric repulsion between the NPs surfaces [37]. If the polymer grafting is heterogeneous, there is low electrostatic repulsion between the NPs, and the proteins are adsorbed on the non-coated surfaces leading to a progressive dispersion of the NPs. So the stabilizing effect of polymer grafting is not always obtained, depending of the polymer chain length, of the grafting density and the conformation of the chains at the NPs surface [6]. These different parameters influence the formation of a water shell [38] and the adhesion of some proteins at the surface [39]. Native Fe 3 O 4 @MSN rapidly dispersed in the presence of proteins (SVF 10%, in HBS 150 mM NaCl, pH 7.4 or RPMI). The electrostatic repulsion between negatively charged NPs surface and the proteins is reduced by the high ionic strength [40]. The protein corona forms a stabilizing shell around native Fe 3 O 4 @MSN. On the contrary, PEG Fe 3 O 4 @MSN are aggregated when diluted in HBS 150 mM NaCl, pH 7.4, in the presence or in the absence of proteins, or in RPMI containing SVF 10% and then slowly dispersed in the presence of proteins. The stealth effect expected by NPs coating with PEG is due to a significant reduction of the protein corona [41]. Yet, some proteins are still adsorbed because, as it has been previously described, the adsorption of proteins is a prerequisite to induce a stealth effect [8]. In our study, the adsorption of proteins onto the PEG Fe 3 O 4 @MSN surface was progressive, inducing a reduction of the aggregation state in suspension. DMPC Fe 3 O 4 @MSN stay stable in suspension during 15 h in HBS 150 mM NaCl, pH 7.4 or in cell culture medium (RPMI), containing or not SVF 10% proteins. The reduction of aggregation state of NPs by coating with a lipid bilayer was previously described [42]. The influence of proteins on the aggregation state of NPs has consequences on their interaction with the cell membrane, and potentially on their toxicity [4]. From QCM-D experiments performed in HBS 150 mM NaCl, pH 7.4, containing FCS 10%, DMPC-Fe 3 O 4 @MSN were found to rapidly deposit during the NPs flowing. Conversely, native and PEG Fe 3 O 4 @MSN were found to deposit more slowly. The stabilization of native Fe 3 O 4 @MSN and less importantly PEG Fe 3 O 4 @MSN by the presence of proteins in the medium seemed to reduce the deposition of the NPs on the membrane surface. On the contrary, the DMPC Fe 3 O 4 @MSN stability in suspension was not influenced by the presence of the proteins, and they were rapidly deposited on the membrane model.
The internalization of DMPC Fe 3 O 4 @MSN by Hep-G2 cells was also faster than the internalization observed for native and PEG Fe 3 O 4 @MSN. DMPC Fe 3 O 4 @MSN were observed inside intracellular vesicles after 3 h while native and PEG Fe 3 O 4 @MSN were observed inside cells after 6 h of incubation. These results suggest an early interaction between DMPC Fe 3 O 4 @MSN and Hep-G2 cell membrane than for the two other types of NPs. This is in accordance with the results obtained by real-time cell impedance measurements reported previously [22]. Using this indirect technique, an early variation of the cell index was observed when the Hep-G2 cell line was exposed to DMPC Fe 3 O 4 @MSN in comparison to native and PEG Fe 3 O 4 @MSN. To conclude on the impact of proteins on NPs cell membrane interactions, the more protein corona is important, the less NPs interact with the cellular membrane. Moreover, native Fe 3 O 4 @MSN, rapidly coated with serum proteins were slowly internalized by Hep-G2 cells, but remained the more cytotoxic according to the MTT assay and real-time cell impedance measurements. In this previous work, native Fe 3 O 4 @MSN induced 40% of cell death at a concentration of 100 µg mL −1 , when in the same conditions NPs recovered by PEG or DMPC induced 5% or 30% of cell death, respectively.
From in vivo experiments using mice, no toxic effect was observed four days after intravenous injection of native, PEG, or DMPC Fe 3 O 4 @MSN for a particle concentration of 40 mg kg −1 on the renal or hepatic function, neither on inflammation factors. The histological observations show no difference between the liver, the spleen or kidneys of treated and non-treated mice. These results are in accordance with other published works on native MSN [19,36].
However, different biodistribution profiles were observed depending of the particle coating. Firstly, the DMPC Fe 3 O 4 @MSN were quickly cleared from the blood circulation, because no more silicium was detected in comparison to control mice 6 h after injection of the NPs. This could be due, at least in part, to an efficient targeting of the liver. In addition, a large amount of in vivo data indicates that nanoparticles have the capacity to exert adverse pulmonary effects after different ways of exposure and systemic also, and it is a real challenge to avoid lung penetration [43]. Here, the very low level of silicium in the lungs of mice treated with DMPC Fe 3 O 4 @MSN translates the poor accumulation of these NPs in lungs, which is of major interest for a medical application.
Conversely, the amount of silicium 24 h after injection is not significantly different to 6 h for PEG Fe 3 O 4 @MSN, suggesting that the PEG coated NPs are slowly distributed in the organs. The presence of native and PEG Fe 3 O 4 @MSN in the lung may be due to the easier formation of aggregates than with DMPC Fe 3 O 4 @MSN. The prolonged circulation time observed for PEG Fe 3 O 4 @MSN is associated to a poor and slow cell internalization of the NPs. On the contrary, DMPC Fe 3 O 4 @MSN are rapidly internalized by cells; plus the absence of aggregation of DMPC Fe 3 O 4 @MSN at high ionic strength or in the presence of proteins, is a considerable advantage in the challenge of a translational objective. This is the first study on well-characterized monodisperse core-shell MSN comparing native or coated with PEG polymer or DMPC lipid bilayer by in vitro, in vivo experiments, and using cell membrane models.

Materials
All reagents were commercially available and used without any further purification. Hydrated Concerning cryo-electron microscopy, three microliters of suspension were applied to glow discharged Quantifoil R 2/2 grids (Quantifoil Micro Tools GmbH, Jena, Germany) or Lacey grid (Ted Pella Inc., Redding, CA, USA), blotted for 1s and then flash frozen in liquid ethane using a CP3 cryo-plunge (Gatan Inc., Pleasanton, CA, USA). Before freezing, the humidity rate was stabilized at about 95%. Cryo-EM was carried out on a JEOL 2200FS FEG operating at 200 kV under low-dose conditions (total dose of 20 electrons/Å 2 ) in the zero-energy-loss mode with a slit width of 20 eV. Images were taken at a nominal magnification of 50,000× corresponding to a calibrated magnification of 45,591× with defocus ranging from 1.4 to 2.5 µm.

Synthesis and Characterization of NPs
The native, PEG-grafted and lipid-coated Fe 3 O 4 @MSN were synthesized following the procedure recently described [22]. The same characterization methods were used. Hydrodynamic diameters and zeta potentials were determined using a Nano ZS apparatus (Malvern Instruments, Malvern, UK). Data were collected from the He-Ne laser light source (λ = 633 nm) at 173 • from the transmitted light beam. Stock suspensions of the Fe 3 O 4 @MSN were prepared in stable conditions at a concentration of 10 mg mL −1 . Pristines Fe 3 O 4 @MSN were dispersed in ethanol 95%, DMPC Fe 3 O 4 @MSN in HBS 150 mM NaCl and PEG Fe 3 O 4 @MSN in HBS 5 mM NaCl. The different suspensions were then diluted in different media at a concentration of 50 µg mL −1 . The dispersion behavior of the different Fe 3 O 4 @MSN was compared in 5 different media: ethanol, HBS 5 mM NaCl, HBS 150 mM NaCl, HBS 150 mM NaCl containing 10% FCS and complete RPMI (FCS 10%). The same suspensions were analyzed directly after dispersion, 2 h and 15 h later. Results are presented as Z-average obtained in intensity mode, associated to the polydispersity index (PDI).

Interaction of NPs with Membrane Models
The interactions of Fe 3 O 4 @MSN with model membranes were investigated using a QCM-D E1 setup (Biolin Scientific, Västra Frölunda, Sweden). The system is composed of a measurement chamber containing a 5 MHz quartz crystal sensor with a silica-coated surface. After each measurement, the chamber, the flow module, and the quartz were rinsed with 2 mL 2% Hellmanex and with 5 mL MilliQ water at a flow of 0.5 mL min −1 . The silica-coated quartz sensor was stored in a SDS 2% solution between two analyses. All measurements were performed at 25 • C. Firstly, the silica-coated surface was rinsed with MilliQ water, dried and placed during 15 min in UV-ozone chamber. Then, HBS 150 mM NaCl was flushed at 100 µL min −1 , and the quartz sensor was equilibrated for each overtone (1st to 13th). The baseline was obtained using HBS 150 mM NaCl both for the frequency and dissipation recordings. A supported lipid bilayer (SLB) is formed on a silica-coated surface of the QCM-D sensor after fusion of EPC small vesicles SUVs). For SUVs formation, a EPC lipid film was resuspended in Hepes 150 mM NaCl, ultrasonicated using a microtip (Digital Sonifier 250 ® , Branson Ultrasonics Corporation, Danbury, CT, USA), and then centrifuged at 20,000× g for 15 min (to remove titanium particles). The supernatant containing SUVs was collected and filtrated through 50 nm porous polycarbonate membrane before coating the SiO 2 QCM-D sensor. The SLB was formed by adding Egg PC SUVs (50 nm) in HBS 150 mM NaCl buffer at a lipid concentration of 0.1 mg mL −1 using a flow of 100 µL min −1 for 10 min. The formation of a continuous lipid bilayer is characterized by a frequency shift ∆f of −26 Hz and a dissipation shift ∆D < 0.5 × 10 −6 followed on the 7th overtone. The membrane was washed during 1 h with HBS 150 mM NaCl under a flow of 100 µL min −1 for stabilization. Before adding NPs on top of the SLB, the NPs dispersant media was injected to recording ∆f and ∆D changes due to the solvent composition. The NPs were dispersed at 0.25 mg mL −1 in HBS 150 mM NaCl containing 10% FCS, and then added on the membrane for 15 min (100 µL min −1 , 1.5 mL), and the flow was stopped for 10 h. Results were presented after offset of the SLB frequency and dissipation recordings, in order to analyze only the NPs effect.

Cell Culture
The Human hepatocyte carcinoma (Hep-G2) cell line was obtained from Sigma-Aldrich. Cells were cultured in RPMI 1640 (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal calf serum (FCS) and 1% penicillin/streptomycin (100 U mL −1 , 100 µg mL −1 ) and incubated in a cell incubator at 37 • C and 5% CO 2 . Cells were used between passages 20 to 40. Cells were passed once a week and the medium was changed twice a week, keeping cells confluence below 80%.

Internalization Assay
Hep-G2 cells were seeded on glass coverslips for 24 h. After controlling their adherence and growing, cells were exposed to 50 µg mL −1 of NPs in RPMI for 3, 6 or 24 h at 37 • C and 5% CO 2 . The medium was removed and cells were rinsed twice with DPBS. Cells were fixed by incubation with 2.5% (v/v) glutaraldehyde in DPBS buffer, for 1 h at room temperature (RT). Then, cells were extensively washed with DPBS. The staining of samples was obtained upon incubation with 1% osmium tetroxide. Samples were dehydrated by ascending grades of EtOH; for impregnation, the samples were firstly treated with a mix EtOH/EPON TM resin (1:1, v/v) for 1 h, and twice in EPON for 2 h. The polymerization was performed by embedding cells in EPON resin for 12 h at 60 • C, plunged in liquid nitrogen at −195 • C to detach the coverslip, and placed for two days at 60 • C for completing polymerization. The ultrathin sections (70 nm) were obtained using an ultramicrotome (Leica Ultracut, Wetzlar, Germany) and disposed on the copper grids. The grids were incubated in uranyl acetate for 2 min, rinsed in water, and then incubated in lead citrate for 2 min, and finally rinsed with water. For each condition (native, PEG and DMPC Fe 3 O 4 @MSN at 3, 6 or 24 h exposure time of Hep-G2 cells), almost 20 cells were observed, and then the observations were focused on the areas were NPs were observed.

Animals
Female C57BL/6 mice (25 g) used for this study were procured from Charles River (Wilmington, MA, USA) and housed in the Institutional animal house under standard environmental conditions (23 ± 1 • C, 55 ± 5% humidity and 12/12 h, light/dark cycles) and maintained with free access to standard diet and water. To establish the preliminary toxicological analyses, 4 groups of

Preliminary Toxicological Assessment
Four days after treatment, mice were sacrificed and organs, blood and urine were collected for histological analysis and biochemical assays. The blood samples with heparin were centrifuged at 1300 rpm for 10 min. The plasma and urine were stored at −20 • C up to analysis. We measured plasma cytokines (TNF-α and IL-6) to assess the inflammatory reaction or systemic toxicity. Plasma TNF-α and IL-6 levels were quantified using commercial ELISA kits as described in the manufacturer's protocol (R&D systems, Minneapolis, MN, USA). The evaluation of renal function was determined by measuring creatinine levels in plasma and in urine using colorimetric assay at 495 nm with alkaline picrate. Liver function was determined from alanine aminotransferase (ALT) activity. Plasma ALT activity was measured according to standard protocol (Infinity, Thermo Scientific, Waltham, MA, USA). Moreover, liver, spleen and kidney were fixed in 10% paraformaldehyde, embedded in paraffin and cut 5 µm tick sections in a microtome. Sections were mounted on glass slides. After staining with hematoxylin-eosin, the sections were examined and imaged under a light microscope.

In Vivo Distribution and Degradation of MSN
Inductively coupled plasma-mass spectrometry (ICP-MS) was used to quantify silicium distribution in the digested tissue samples (liver, kidney and spleen). 2-100 mg of biological samples (urine, blood, or lyophilized and crushed kidney, spleen or liver) were digested in HNO 3 (1 mL, 67%) for 1 h, and then HCl (1 mL, 34%) and of H 2 O 2 (0.5 mL, 30%) for 1 h. The digestion was completed by microwave in Teflon vials, by using UltraWAVE single reaction chamber (Milestone, Shelton, AL, USA). Silicium content was further analysed using a NexION™ 300 ICP-MS instrument (PerkinElmer, Waltham, MA, USA), from the analytical platform facilities of CEREGE (Aix en Provence, France).

Statistical Analysis
Statistical analysis was performed using the Student's t-test to compare paired groups of data. A p value < 0.05 was considered as statistically significant. in particular Franck Godiard and Véronique Richard. Patrick Bron et Joséphine Lai-Kee-Him would like to thank the French Infrastructure for Integrated Structural Biology (FRISBI) for its support, a national infrastructure supported by the French National Research Agency (ANR-10-INBS-05). We also thank Laure Lichon for technical assistance in the experiments.