Different types of NPs (nanoparticles) are currently under development for diagnostic and therapeutic applications in the biomedical field, yet our knowledge about their possible effects and fate in living cells is still limited. In the present study, we examined the cellular response of human brain-derived endothelial cells to NPs of different size and structure: uncoated and oleic acid-coated iron oxide NPs (8–9 nm core), fluorescent 25 and 50 nm silica NPs, TiO2 NPs (21 nm mean core diameter) and PLGA [poly(lactic-co-glycolic acid)]-PEO [poly(ethylene oxide)] polymeric NPs (150 nm). We evaluated their uptake by the cells, and their localization, generation of oxidative stress and DNA-damaging effects in exposed cells. We show that NPs are internalized by human brain-derived endothelial cells; however, the extent of their intracellular uptake is dependent on the characteristics of the NPs. After their uptake by human brain-derived endothelial cells NPs are transported into the lysosomes of these cells, where they enhance the activation of lysosomal proteases. In brain-derived endothelial cells, NPs induce the production of an oxidative stress after exposure to iron oxide and TiO2 NPs, which is correlated with an increase in DNA strand breaks and defensive mechanisms that ultimately induce an autophagy process in the cells.
New diagnostic and therapeutic technologies have been developed in nanomedicine to improve the detection and treatment of human diseases. Therefore it is becoming necessary to better understand the mechanisms of interaction of nanomaterials, including NPs (nanoparticles), with living tissues in order to assess the biological consequences associated with nanotechnologies. The physicochemical and biochemical properties of the NPs, like surface properties, charge, size or the adsorption of biological components, are important factors mediating their interactions with cells, including cell-stress reactions and the biological characteristics of the particular cells [1,2]. Two types of NPs, either ‘solid core’ NPs, such as metal oxide-based NPs, or ‘soft core’ NPs, such as polymer-based NPs, are suitable for therapeutic use, chemically functionalized polymer-coated solid (metallic) core NPs and polymer–copolymer NPs encapsulating therapeutic agents. NPs may be internalized by cells and the solid core of these NPs or their polymer components may be found in different cellular localization.
The cellular reaction to NP uptake may result in the modification of cellular functions, such as repression/activation of genes and activation of specific cell organelles . Cellular stress, due to redox imbalance, is also an important expression of cytotoxicity, which will occur at early stages of interaction. Cellular oxidative stress and the production of ROS (reactive oxygen species) upon cell exposure to NPs have previously been shown to be a common property of NPs, relevant to assess their potential negative effects on cell functions [2,4–7]. Oxidative stress may induce DNA damage, lysosomal activation and autophagy, which may also represent a feedback mechanism to limit ROS-mediated cell activation by removing oxidatively damaged molecules and cell structure . Autophagy is the only cell process able to degrade large components, therefore it can be postulated that NPs may be processed by cells using autophagy. NPs-mediated autophagy may be an adaptive cellular response, aiding in the degradation and clearance of nanomaterials, but may also cause harmful cellular dysfunction. Some previous publications have described NPs-related autophagy processes [8–14], but none has compared the autophagy-inducing properties with the oxidative stress-inducing properties of several NPs, and none have evaluated these processes in brain vascular cells.
The BBB (blood–brain barrier) separating the brain from the bloodstream is represented by very specialized endothelial cells. The presence of tight junctions, of drug resistance mechanisms and of detoxification systems at the level of this vascular system protects the brain not only from the uncontrolled entry of blood-borne chemicals, but also from the entry of therapeutics, including NPs [16,17]. Thus, depending on the context, the potential of NPs to be taken up by brain-derived endothelial cells and the cellular consequences of such interactions may induce dysfunction of these cells important for brain functions and homoeostasis.
As the chemical composition, the solubility of the components and surface reactivity, such as surface charges and redox NP surface state are important for NP uptake by cells, cytotoxicity and oxidative stress induction, different types of NPs need to be compared with defined consequences of their interactions with cells. In the present work, we compared the cell response of human brain-derived endothelial cells as models of the BBB with exposure to iron-, silica- and titanium-oxide-based solid core NPs and one polymeric NP as models for NPs with different properties. We show that defined NPs induced an oxidative stress, DNA damage, the activation of lysosomal proteases and an autophagy reaction in these cells, which we hypothesize to be a defensive mechanism of the cerebral endothelial cells to protect the brain.
All NPs used in the present study, with the exception of aminoPVA [poly(vinyl alcohol/vinylamine)]-coated USPIO (ultrasmall superparamagnetic iron oxide) NPs , are commercially available and were provided with physicochemical characterization by the provider (Table 1). Some further characterization of the commercial NPs was performed (D. Bilanicova and G. Pojana, University of Venice, Italy; and M. Dusinska and the NanoTEST Consortium), when requested for the present study. AminoPVA-coated USPIO NPs were prepared and characterized according to the protocol previously described . Briefly, ferrofluid was prepared by alkaline co-precipitation of ferric and ferrous chlorides, refluxed in nitric oxide-ferrous nitrate and dialysed, providing iron oxide NPs (ferrofluid) of 9 nm. To obtain aminoPVA-USPIO NPs, the ferrofluid was mixed with PVA [poly(vinyl alcohol)] and aminoPVA at a ratio of polymer to iron of 10 and a ratio of PVA to aminoPVA copolymer of 45 (mass ratios) , resulting in USPIO NPs with a hydrodynamic diameter of 25–30 nm and a positive ζ-potential of +25 mV. Uncoated USPIO NPs (Fe3O4, uncoated USPIO NPs) were obtained from PlasmaChem as a ~3% nanosuspension in water, average NP size 8±3 nm [as determined by DLS (dynamic light scattering), intensity (Z) average], ζ-potential +15 mV in 10 mM NaCl and −23 mV in DMEM (Dulbecco's modified Eagle's medium). The iron content was determined to be 18 mg iron/ml by quantitative Prussian Blue reaction according to a previously described protocol . Oleic acid-coated USPIO NPs (Fe3O4, 3% oleic acid coating) were obtained from PlasmaChem as an ~7% nanosuspension in water, average particle iron oxide core size 8±3 nm, hydrodynamic size 14–15 nm (determined by DLS) and ζ-potential −30 mV at pH 7. The iron content was determined to be 206 mg iron/ml by quantitative Prussian Blue reaction. TiO2 NPs (Aeroxide® TiO2 P25) was obtained from Evonik Degussa and were provided by M. Whelan, JRC, Ispra, average particle size 21 nm, large dispersion size and ζ-potential −30 mV. The size distribution of the solid core NPs was controlled by TEM (transmission electron microscopy). Polymeric NPs consisting of a mixture of PLGA [poly(lactic-co-glycolic acid)] and PEO [poly(ethylene oxide)] (PLGA-PEO NPs) were obtained from Advancell Nanosystems as a 10 mg/ml suspension in water, average particle size 143 nm and ζ-potential −39 mV. Fluorescent rhodamine-labelled silica nanospheres (25 nm silica NPs and 50 nm silica NPs) were obtained from Corpuscular as a 25 mg/ml suspension in water [17 and 12 mg/ml respectively as measured by thermogravimetric analysis by H. Hofmann and P. Bowen (EPFL, Lausanne, Switzerland] with a ζ-potential of −40 mV. The size of the two coated aminoPVA-USPIO NPs  and of oleic acid-coated USPIO NPs are of the order of 35 nm in cell culture medium and the size of uncoated USPIO NPs and TiO2 NPs was determined in cell culture medium to be 910 and 650 nm respectively by DLS, with large size dispersion. The size of other NPs did not notably change in the cell culture medium.
|NPs||Core size*† (nm)||Hydrodynamic size† (nm)||ζ-potential‡ (mV)||Initial concentration (mg/ml)|
|Uncoated USPIO||8||–||+15 (10 mM NaCl); −23 (DMEM)||18§|
|Oleic acid-coated USPIO||8||14–15||−30||206§|
|NPs||Core size*† (nm)||Hydrodynamic size† (nm)||ζ-potential‡ (mV)||Initial concentration (mg/ml)|
|Uncoated USPIO||8||–||+15 (10 mM NaCl); −23 (DMEM)||18§|
|Oleic acid-coated USPIO||8||14–15||−30||206§|
Preparation of NP suspensions
Uncoated and oleic acid-coated USPIO NPs in suspension in water were diluted in PBS to obtain a final concentration of 2 mg of iron/ml. Immediately before use aliquots were vortex-mixed for 1 min and added to cell culture medium to obtain a 75 μg of iron/cm2 working solution, then serially diluted in the cell culture medium. Stock suspensions of TiO2 NPs were made as a suspension of 2 mg/ml in culture media without FBS (fetal bovine serum) containing 15 mM Hepes, then sonicated for a total of 3 min (9×20 s) at 4°C and 60 W, vortex-mixed for 10 s, then stored at −20°C, according to a previously described protocol . Immediately before use the NP suspension was thawed, vortex-mixed for 10 s, sonicated for a total of 1 min (3×20 s) at 4°C and 60 W , added to the cell culture medium to achieve a 75 μg/cm2 working solution and then serially diluted. PLGA-PEO NPs in suspension in water were diluted in the appropriate cell culture medium to obtain a stock solution of 75 μg/cm2, then serially diluted in the cell culture medium. Fluorescent 25 nm and 50 nm silica-NPs in suspension in water were vortex-mixed for a few minutes, diluted in the appropriate cell culture medium to obtain a stock solution of 75 μg/cm2, then serially diluted in the cell culture medium.
Cell lines and culture conditions
HCECs (human cerebral endothelial cells) were a gift from Professor D. Stanimirovic, Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada, and were grown in DMEM, containing 4.5 g/l glucose, 10% FBS and penicillin/streptomycin. All cell culture reagents were purchased from Gibco and Invitrogen. Cells were exposed to NPs in 250 μl of complete culture medium in 48-well plates (Costar) at decreasing concentrations, from 235 μg/ml equivalent to 75 μg/cm2 NPs to 0.4 μg/ml equivalent to 0.12 μg/cm2 NPs. Experiments were performed in three wells at least twice in independent experiments.
Evaluation of cell viability
Cells were grown in 48-well cell culture plates (Costar) until 75% confluent, then exposed to NPs for the concentration and time indicated, then washed in saline. Cell viability was evaluated using the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H tetrazolium bromide] assay (Sigma–Aldrich) added to the cells in fresh complete culture medium at a 250 μg/ml final concentration. After 2 h the supernatant was removed and the precipitated formazan was dissolved in 0.1 M HCl in propan-2-ol and quantified at 540 nm in a multiwell plate reader (iEMS Labsystems, BioConcepts).
Evaluation of DNA synthesis
Briefly, following exposure of cells to NPs for 48 h, [3H]thymidine (GE Healthcare, 400 nCi/ml final concentration) was added to the cells for 2 h. Then the cell layers were precipitated with 10% trichloroacetic acid and dissolved in 0.1% SDS in 0.1 M NaOH and scintillation cocktail (Optiphase HI-Safe, PerkinElmer) was added. Radioactivity was counted with a β-counter (WinSpectra). The radioactivity counts of treated cells were compared with the radioactivity counts of untreated cells. Experiments were conducted in triplicate wells and repeated twice. Means±S.D. were calculated.
Evaluation of ROS production
Free radical formation was determined with 5/6-carboxy-2,7-dichlorodihydrofluorescein. Cells were grown until confluent in 48-well plates (Costar), then preincubated for 40 min with carboxy-H2DCFDA (5/6-carboxy-2,7-dichloro-dihydrofluorescein diacetate, Molecular Probes, Invitrogen, stock solution 25 mg/ml in DMSO, final concentration 20 μM), washed and incubated at 37°C with the NPs diluted in enriched HBSS (Hanks balanced salt solution; containing 1.3 mM Ca2+, 1.1 mM Mg2+ and 5 mM glucose, from Gibco). Then 5 mM t-butyl-H2O2 (Sigma) was added as a positive control for free radical formation. Then the plate was read at λex/λem=485/530 nm in a thermostatically controlled fluorescence plate reader (CytoFluor Series 4000, PerSeptive Biosystems) at t=0 h, then every hour for 4 h.
Cells were grown for 24 h on glass slides (Menzel-Gläser). Then the medium was changed, and the NPs diluted in complete culture medium were added at the desired concentrations for a further 24 h (uptake of silica NPs) or only 4 h (oxidative stress evaluation). At the end of the treatment, the cell layers were washed in PBS, fixed in 4% PFA (paraformaldehyde) for 20 min at 4°C, washed in PBS and incubated with DAPI (4′,6′-diamidino-2-phenylindole; Roche Diagnostics; 1 μg/ml in PBS) for 30 min at 37°C. For oxidative stress evaluation, the incubation with DAPI was preceded by carboxy-H2DCFDA staining (20 μM) for 30 min at 37°C. The slides were washed with PBS and mounted in 20% glycerol in PBS, then examined by fluorescence microscopy.
Determination of cellular thiols
The monobromobimane assay was used to measure cellular thiol levels. Cells were grown until half-confluent in 48-well plates and exposed for 4, 24 or 48 h to the NPs, washed with HBSS, then 250 μl/well of 100 μM monobromobimane (Sigma–Aldrich) in HBSS was added at room temperature (22°C) for 5 min in the dark. The cell layers were washed with HBSS and lysed with 0.1% Triton X-100 (Sigma–Aldrich) in HBSS. For a positive control, cells were exposed to 100 μM NEM (N-ethylmaleimide, Sigma–Aldrich) for 1 min before the assay. The fluorescence was immediately read in a thermostatically controlled fluorescence plate reader (CytoFluor) at λex/λem=485/580 nm.
Evaluation of DNA damage
Induction of DNA strand breaks by the NPs in HCECs was determined using the alkaline single-cell gel electrophoresis (Comet assay) . Briefly, cells were grown for 24 h in a 24-well plate (Costar), then exposed to the NPs for the concentration and time indicated. Cells were detached using trypsin/EDTA (TrypLE Express, Invitrogen) and the cell suspension (0.8×106 cells/ml) in 1% low-melting-point agarose (Sigma–Aldrich) was deposited on glass slides coated with 1% normal-melting-point agarose (Eurobio). After 5 min at 4°C, to allow solidification of the cell layer, slides were immersed in lysis buffer (2.5 M NaCl, 100 mM EDTA, 10 mM Tris-base and 1% Triton X-100, pH 10.0) for 1 h at 4°C, then a 40 min unwinding process at 4°C and electrophoresis at 25 V for 30 min were performed, both under alkaline conditions (300 mM NaOH and 1 mM EDTA, pH 13.0). Following washing with ice-cold PBS and ice-cold nanopure water, slides were stained with 1 μg/ml DAPI (Roche) in PBS. Cells treated with 100 μM H2O2 for 5 min at 4°C were used as a positive control. Analysis of Comet appearance was performed with a Zeiss Axioplan 2 Imaging microscope at a ×200 magnification and λex/λem=365/420 nm. For each experiment, 100 cells were analysed twice randomly per treatment using the Comet assay image analysis software (Comet Visual, NILU, Health Effects Laboratory). Cells were scored visually and assigned to one of five classes according to the Comet tail size (from undamaged 0 to maximally damaged 4) giving expression in arbitrary units from 0 (completely undamaged) to 400 (with maximal damage) .
Cells were grown in 10-cm-diameter Petri dishes and exposed to the NPs for the indicated time and concentration. After treatment, the cell layers were washed with ice-cold PBS and lysed in 200 μl of lysis buffer (150 mM NaCl, 2 mM EDTA, 0.5% Triton X-100, 50 mM Tris/HCl, 2 mM sodium orthovanadate and 50 mM NaF, pH 7.2) and 10 μl of protease inhibitor cocktail (Sigma–Aldrich), scraped with a cell scraper, extracted by three cycles of freeze/thawing and centrifuged at 10000 rev./min at 4°C for 10 min on an Eppendorf 5415R centrifuge. Supernatants were applied to SDS/PAGE and transferred on to a nitrocellulose membrane (Whatman). The membranes were blocked with 5% non-fat dried skimmed milk powder in PBS, washed in 0.05% Tween 20 (Sigma–Aldrich) in PBS and incubated overnight at 4°C with anti-human LC3 (Novus Biologicals; diluted 1:2000 in 1% non-fat dried skimmed milk powder and 0.05% Tween 20 in PBS), anti-cathepsin B (BioVision; diluted 1:5000), anti-cathepsin D (H-75) (Santa Cruz Biotechnology; diluted 1:2000) or anti-SQSTM1/p62 (Cell Signaling Technologies; diluted 1:3000) rabbit polyclonal antibodies, and then exposed for 60 min to horseradish peroxidase-conjugated anti-rabbit antibody (Promega; diluted 1:5000) and visualized using chemoluminescence (ECL®, GE Healthcare). Transferrin receptor/CD71 was determined with the goat polyclonal anti-CD71 antibody (Santa Cruz Biotechnology; diluted 1:3000), and then exposed for 60 min to horseradish peroxidase-conjugated anti-goat antibody (Sigma; diluted 1:10000) and visualized using chemoluminescence. To control for loading, the membranes were stripped by successive incubation in 0.1 M glycine, pH 2.3, 1 M NaCl in PBS and 0.05% Tween 20 in PBS, blocked for 1 h with 5% non-fat dried skimmed milk powder in PBS and exposed to a monoclonal anti-chicken α-tubulin mouse antibody (Abcam; diluted 1:5000) for 1 h at room temperature followed by 1 h incubation with horseradish peroxidase-conjugated anti-mouse antibody (Sigma) and treated as described above. For densitometric analysis, the band areas of cathepsin B, cathepsin D and α-tubulin were evaluated using ImageJ software (NIH), then the ratio of the area of the cathepsin band to the α-tubulin band was calculated.
Cells were grown in 10-cm-diameter Petri dishes (Nunclon) until 70–80% confluent, then exposed to NPs for the concentration and time indicated. At the end of the incubation, the cells were washed twice with PBS, detached from the Petri dishes in trypsin/EDTA and centrifuged, then the cell pellets were fixed in 2.3% cacodylate-buffered glutaraldehyde (Sigma) for up to 24 h, washed in cacodylate buffer and mixed with 4% low-gelling agarose (Sigma). Samples were post-fixed in 1.3% osmium tetroxide in 0.2 M cacodylate buffer, pH 7.4, for 1 h and dehydrated in graded ethanol, then in propylene oxide and embedded in 50% (w/w) epoxy embedding medium, 26% (w/w) DDSA (dodecenylsuccinic anhydride), 23% (w/w) MNA (methyl nadic anhydride), 1% (w/w) DMP-30 [2,4,6-tris(dimethylaminomethyl)phenol] (all from Fluka and Sigma–Aldrich). Blocks were cured 48 h at 60°C, thin sections (70–80 nm) were cut using an ultramicrotome (Ultracut E, Reichert-Jung Optische Werke AG) and mounted on 3-mm 200-mesh copper grids. Grids were stained for 75 min in saturated uranyl acetate solution (Fluka), then for 100 s in lead citrate (Ultrostain 2, Laurylab), examined and photographed with a Philips CM10 transmission electron microscope combined with a MegaView III Soft Imaging system to evaluate USPIO NPs associated to cells.
Calculation of results
Each experiment was repeated in three wells at least twice. Means±S.D. were calculated and statistical significance was evaluated using a Student's t test.
RESULTS AND DISCUSSION
Previously published studies have suggested that following intravenous injection NPs can be trapped in brain vessels, possibly being biodegraded inside brain endothelial cells [17,21–24]. We have previously shown that biocompatible cationic USPIO NPs were internalized by cells , even when functionalized with a fluorescent reporter  or with anti-cancer drugs [26,27]. Their final cellular localization in human melanoma cells was the lysosomes, where they started to dissolve, inhibiting the expression of the transferrin receptor and activating the lysosomal protease cathepsin D . We have also previously shown that human brain-derived endothelial cells internalized these cationic USPIO NPs, but neither released nor transported them , suggesting an intracellular processing of such NPs and their retention in the cells. Previous studies using cells from a different origin and iron oxide-based NPs have demonstrated lysosomal localization, reorganization of cell cytoskeleton and intracellular iron-mediated increase in the production of ROS [29,30]. Therefore the present study focused on the mechanisms of interactions between NPs with different physicochemical properties and human cells representative of the BBB, analysing the cellular consequences of these interactions, focusing on the generation of oxidative stress and cellular degradative pathways.
TiO2 NPs and uncoated USPIO NPs very rapidly agglomerate in culture medium containing FBS, resulting in agglomerates of large dispersion size, average 850 nm  and 910 nm respectively as determined by DLS (M. Dusinska and the NanoTEST Consortium) and electron microscopy (results not shown), providing information for agglomerated NPs with two different physicochemical characteristics. Uncoated and oleic acid-coated USPIO NPs with similar cores allowed comparison of NPs with different surface properties. Silica NPs with different sizes, 25 and 50 nm, allowed evaluation of the size effects of NPs with similar composition, whereas polymeric NPs were considered as control NPs not inducing a cellular reaction.
Internalization of NPs by HCECs
After 24 h incubation of the cells with the NPs their uptake was studied with methods adapted to the NP type: fluorescence microscopy for fluorescent silica NPs, the histological Prussian Blue  reaction for USPIO NPs and TEM for inorganic solid core (iron oxide, TiO2 and silica) NPs to analyse the intracellular localization of solid core NPs inside cell organelles. The cell uptake of silica NPs was observed only at the highest concentration tested (75 μg/cm2) and the uptake of 25 nm NPs was slightly higher than the uptake of 50 nm NPs (Figure 1A). Uncoated USPIO NPs agglomerated in the cellular environment which resulted in their rapid deposition on the cell surface. The internalization of these NPs as determined by Prussian Blue reaction (results not shown) was much higher than the uptake of oleic-acid-coated USPIO NPs, which were stable in cell culture medium and were only poorly internalized by endothelial cells, as previously shown . The intracellular presence of the solid cores of iron oxide, TiO2 and silica NPs was confirmed by TEM (Figures 1Ba–1Bf). The NPs were localized in small or large endosomes in the cytoplasm and in lysosomes. Therefore TEM and fluorescence microscopy demonstrated the intracellular localization of solid core NPs in cell organelles, most probably the endosomal–lysosomal compartment.
Uptake of the NPs by HCECs
Evaluation of cell viability and DNA synthesis by HCECs exposed to the NPs
To evaluate for potential cytotoxicity, the metabolic activity of HCECs after 72 h exposure to the NPs was evaluated using the MTT assay (Figure 2). At the highest concentration, oleic acid-coated USPIO NPs and 25 nm fluorescent silica NPs (Figures 2A and 2B) induced a decrease in the metabolic activity of HCECs, whereas uncoated USPIO NPs and 50 nm silica NPs were less cytotoxic. Acute cytotoxicity was already observed after 24 h exposure to high doses (75 μg/cm2) of oleic acid-coated USPIO NPs (results not shown). Exposure of cells to TiO2 NPs induced a significant cytotoxicity for HCECs (Figure 2C). PLGA-PEO NPs were not cytotoxic at any of the concentrations tested.
Cytotoxicity of the NPs for HCECs
The synthesis of DNA by HCECs was also determined after 48 h exposure of the cells to NPs. A significant decrease in DNA synthesis was observed after cell exposure to high doses of oleic acid-coated USPIO NPs and 25 nm silica NPs (Figure 3). Exposure to uncoated USPIO NPs and TiO2 NPs caused a slight decrease in DNA synthesis in a dose-dependent manner. PLGA-PEO NPs and 50 nm silica NPs had no effect on DNA synthesis by exposed cells (results not shown).
Effect of NPs on DNA synthesis by HCECs
Effect of NPs on oxidative stress generation and DNA damage in HCECs exposed to the NPs
The mitochondria content of endothelial cells at the BBB is higher than for endothelial cells of another location, rendering them more susceptible to generation of oxidative stress. Several studies have demonstrated that the generation of ROS and oxidative stress are the key mechanism by which NPs exert deleterious effects , including oxidative-stress-mediated entry of NPs into oxidation-damaged endothelial cells of cerebral microvessels . NPs may induce oxidative stress and/or directly interact with the endosomal–lysosomal pathway, two possibilities that we evaluated in the experiments reported in the present paper. The production of ROS by HCECs exposed to the NPs was first measured using carboxy-H2DCFDA, detecting the intracellular presence of H2O2 (Figure 4). After 4 h of cell exposure [which we had determined as the optimal time for the determination of ROS production (results not shown)], to uncoated USPIO NPs, oleic acid-coated USPIO NPs and TiO2 NPs, a significant increase in ROS production was detected by quantitative fluorescence measurements (Figure 4A), which was confirmed by fluorescence microscopy (Figure 4B).
ROS production by HCECs exposed to NPs
Then, the content in thiols of HCECs was determined after 4, 24 or 48 h exposure of the cells to the NPs using the bromobimane assay (Figure 5). A more pronounced dose-dependent decrease in thiol levels was observed after cell exposure to uncoated USPIO NPs, already apparent after 4 h, than to oleic acid-coated USPIO NPs (results not shown). As exposure of cells to 75 μg/cm2 oleic acid-coated USPIO NPs induced almost complete cytotoxicity (Figure 2A) we limited evaluation of this particular NP to the highest non-cytotoxic concentration of 30 μg/cm2. PLGA-PEO and 50 nm silica NPs did not affect cellular thiol content after 4 h (results not shown), 24 or 48 h, whereas TiO2 and 25 nm silica NPs induced a slight decrease in cellular thiols after 24 and 48 h at the highest concentration tested (Figure 5).
Determination of the thiol content of HCECs exposed to NPs
A potential DNA-damaging effect of NPs in HCECs was measured by the alkaline single-cell gel electrophoresis (Comet assay)  detecting single and double DNA strand breaks and alkali-labile sites (Figure 6). Significant DNA damage was found after 24 and 48 h of cell exposure to both uncoated and oleic acid-coated USPIO NPs and this effect was even more pronounced for TiO2 NPs. Thus we show that defined NPs induced ROS production by brain-derived endothelial cells, decreasing the levels of thiols in the cells and increasing DNA damage.
Evaluation of DNA-damage induced by NPs in HCECs
Autophagy induction and lysosomal activation by the NPs in HCECs
Autophagy is a process by which cells consume their own components in lysosomes and is the only cell process able to degrade large components, therefore it can be postulated that NPs may be processed by autophagy. There are several forms of autophagy, but all involve the delivery of cell components to lysosomes in response to sublethal cell stress, such as oxidative stress or DNA damage, protein aggregates, particulate structures, mitochondrial damage or pathogens [33,34]. ROS activation of autophagy may represent a feedback mechanism to limit ROS-mediated cell activation by removing oxidatively damaged molecules and cell structure . NP-mediated autophagy may be an adaptive cellular response, aiding in the degradation and clearance of nanomaterials, but may also cause harmful cellular dysfunction. Some previous publications have described NP-related autophagy processes [8,14,15,34–36], but none has compared the autophagy-inducing properties with the oxidative stress-inducing properties of several NPs. For example, gold NPs induced oxidative stress and autophagy in human fibroblasts . The reports of NP-mediated induction of autophagy in the endothelium, in particular in cerebral endothelial cells, are very scarce. Pollution NPs have been associated with disruption of the BBB  and fullerene NPs were shown to induce autophagic vacuolization of HUVECs (human umbilical vein endothelial cells) .
During autophagy, a small flattened vesicle forms in the cytoplasm and grows to a cup-shaped isolation membrane, engulfing the materials to be degraded, then closing to form a double-membrane autophagosome that fuses with lysosomes. Cellular structures presenting the characteristics of an autophagic response of the cells to the NPs as well as lysosomal involvement were observed by TEM in HCECs following exposure to the NPs (Figure 7). The presence of characteristic double membrane vacuoles containing cellular material indicated the formation of autophagosomes. After the fusion with lysosomes they become autolysosomes able to degrade cellular material. Autolysosomes were observed in NP-exposed cells (Figures 7a–7d and 7f), some of them containing NPs (Figure 7f). Generalized high vacuolization of cells was another autophagy characteristic that could be observed in cells exposed to the NPs (Figure 7e), in particular in cells exposed to 50 nm silica NPs. Thus we demonstrated in cells exposed to NPs the appearance of cellular structures presenting the characteristics of an autophagic response of the cells as well as lysosomal involvement.
TEM images of autophagic vacuoles in HCECs exposed to NPs
The formation of autophagosomes involves complex protein pathways, one of which includes the conjugation of an ubiquitin-like protein, LC3, that is cleaved at its C-terminal by a ROS-sensitive thiol protease to release LC3-I exposing a glycine residue that is then conjugated to the lipid PE (phosphatidylethanolamine) in an ubiquitin-like reaction to produce LC3-II. LC3-II is then attached to the double membranes of mature autophagosomes and maintained after their fusion with lysosomes. LC3-I into LC3-II conversion is commonly used to monitor autophagy . Both NP exposure  and autophagy induction also involves the activation of the enzymes of lysosomes, including the cathepsins. The expression of LC3-I and LC3-II, and of the (pro)cathepsins B and D were determined in extracts of HCECs exposed to the NPs (Figure 8A). We demonstrated NP-specific induction of LC3-II expression in HCECs, supporting our TEM observations. NP-specific activation of procathepsins B and D was also demonstrated (Figures 8A and 8B). In the HCECs, LC3-II levels together with the activation of procathepsin D were particularly enhanced in cells exposed to uncoated USPIO NPs and TiO2 NPs. It has to be emphasized that these two NPs are particularly prone to agglomeration in cell culture medium. (Pro)cathepsin B is expressed at only very low levels in HCECs, and we showed that NP-dependent induction or activation of procathepsin B was very low for all NPs, possibly being slightly more marked with the non-agglomerating solid-core NPs (Figure 8B). In a time-course study to evaluate the autophagic flux, the rate at which intracellular material is transported to the lysosomes , in HCECs exposed or not exposed for 6–72 h to uncoated USPIO NPs, we demonstrated that the appearance of activated cathepsin D and LC3-II conversion were correlated, starting after 48 h exposure, whereas the iron-mediated repression of the transferrin receptor/CD71, being apparent already after 24 h exposure, preceded cathepsin D activation and LC3-II conversions (Figure 8C). However, NP-mediated autophagy induction seems to be different from aggregated protein-related degradation by autophagy, since the expression of sequestosome 1/SQTM1/p62 [38–40] was not decreased in HCECs exposed to uncoated USPIO NPs (Figure 8C). The polyubiquitin-binding protein sequestosome SQSTM/p62 has been shown to be degraded in some autophagic processes, and inhibition of autophagy increased its expression . Thus we showed that defined NPs have the potential to induce ROS generation, DNA damage or to selectively activate procathepsins B and D to their active forms and to induce autophagy. However, different NPs were responsible for these effects. These effects were particularly obvious for the two NPs that agglomerate in the culture conditions and induced oxidative stress, i.e. uncoated USPIO NPs and TiO2 NPs. Therefore we hypothesize that NP aggregation may be important in ROS generation, activation of defined lysosomal cathepsins and the NP-mediated autophagic cell reaction. However, NP-mediated DNA damage seems to play a less relevant role, since oleic acid-coated USPIO NPs that induce DNA damage at the same level as uncoated USPIO NPs do not induce such an autophagic reaction and generate much lower oxidative stress. NP size has also been suggested to be relevant for the induction of the autophagic process. PEG [poly(ethylene glycol)]-coated quantum dots induction of autophagy was size-dependent in human mesenchymal stem cells [12,42]. Rare earth oxide nanocrystals induced autophagy and vacuolization in human cancer cells. Vacuolization was size-dependent and actually caused by large aggregates of NPs . All of these results are in agreement with these of the present study.
Expression of autophagy and lysosomal markers by HCECs exposed to NPs
Cathepsin D (EC 126.96.36.199) is an aspartyl lysosomal protease expressed in all cells, synthesized in the ER (endoplasmic reticulum) as an inactive proenzyme of 52 kDa, then activated by intramolecular autocatalytic splitting in the acidic environment of lysosomes into active cathepsin D, constituted of two subunits of 34 and 14 kDa. Cathepsin B (EC 188.8.131.52) is a thiol lysosomal protease, biosynthesized as an inactive proenzyme of 39 kDa, which is transported into the lysosomes where acidification triggers the activation of the enzyme. The 30 kDa single chain active enzyme can be further processed into an active two-chain form (25 and 5 kDa), linked by a disulfide bridge, by removal of an internal dipeptide. Using a different model of induction of apoptosis and autophagy by serum deprivation in PC12 cells, it was previously shown that cell death was mediated by mitochondria and involved caspase-3 activation. However, during this process, the functions of the lysosomal cathepsins, particularly cathepsin B and cathepsin D, were also altered, cathepsin B decreased whereas cathepsin D increased in serum-deprived PC12 cells . It was also shown that overexpression of cathepsin D accelerated apoptosis, whereas overexpression of cathepsin B increased cell viability. In another report of nutrient deprivation of breast cancer cells, it was shown that inhibition of autophagy stimulated cathepsin D expression and induced apoptosis via cathepsin D accumulation; apoptosis was reduced by knockdown of cathepsin D . It was also previously shown that increased cathepsin D expression may be responsible for switching cell pathways from apoptosis to autophagy, possibly by depleting caspase 3  and salvaging the cells from cell death.
In conclusion, we have explored the mechanisms of interactions between NPs with different biophysical characteristics and human brain-derived endothelial cells. Taken together, our previous results  and those of the present study indicate that depending on their physicochemical characteristics, NPs may be found in brain vascular cells, where they induce an NP-specific cell reaction, suggesting that they may modulate brain functions from this location. We have shown that the NPs induced the production of ROS, were transported into lysosomes, interfering with the lysosomal hydrolases, cathepsins D and B, and induced an autophagy process. The observation that NPs induced autophagy in these cells probably in response to oxidative stress and as being recognized as agglomerated materials, suggests a defensive mechanism of the cerebral endothelial cells to protect the brain. These observations have important biomedical consequences. Owing to the suspected risks of distributing deleterious chemicals, such as metals and cations to the CNS (central nervous system), the technologies based on NPs are designed in such a way that they cannot get access to the CNS. Therefore the response of cells forming the interface between the brain and the blood, the cells of the BBB, needs to be evaluated in detail for NPs that will be designed in such a way.
central nervous system
dynamic light scattering
Dulbecco's modified Eagle's medium
fetal bovine serum
Hanks balanced salt solution
human cerebral endothelial cell
reactive oxygen species
transmission electron microscopy
ultrasmall superparamagnetic iron oxide
Blanka Halamoda Kenzaoui and Lucienne Juillerat-Jeanneret designed the experiments, analysed and interpreted the results and wrote the paper; Blanka Halamoda Kenzaoui, Catherine Chapuis Bernasconi and Seher Guney-Ayra performed the experiments and participated in the evaluation of the results.
We thank P. Bowen and H. Hofmann from EPFL, Lausanne, for providing the thermogravimetric analyses of the silica NPs; M. Dusinska and her team at NILU, Oslo, for their help in the Comet assay and helpful comments; and D. Bilanicova, G. Pojana and A. Marcomini, University of Venice, for providing physicochemical characterization data of the NPs.
This work was supported by the European Community 7th Framework Programme [project number 2007-201335 ‘NanoTEST’]; the Switzerland-France InterReg programme, project ‘NAOMI’; and the Swiss Government-University of Bern BNF programme.