Lysosomal iron liberation is responsible for the vulnerability of brain microglial cells to iron oxide nanoparticles: comparison with neurons and astrocytes.

Abstract Iron oxide nanoparticles (IONPs) are used for various biomedical and neurobiological applications. Thus, detailed knowledge on the accumulation and toxic potential of IONPs for the different types of brain cells is highly warranted. Literature data suggest that microglial cells are more vulnerable towards IONP exposure than other types of brain cells. To investigate the mechanisms involved in IONP-induced microglial toxicity, we applied fluorescent dimercaptosuccinate-coated IONPs to primary cultures of microglial cells. Exposure to IONPs for 6 h caused a strong concentration-dependent increase in the microglial iron content which was accompanied by a substantial generation of reactive oxygen species (ROS) and by cell toxicity. In contrast, hardly any ROS staining and no loss in cell viability were observed for cultured primary astrocytes and neurons although these cultures accumulated similar specific amounts of IONPs than microglia. Co-localization studies with lysotracker revealed that after 6 h of incubation in microglial cells, but not in astrocytes and neurons, most IONP fluorescence was localized in lysosomes. ROS formation and toxicity in IONP-treated microglial cultures were prevented by neutralizing lysosomal pH by the application of NH4Cl or Bafilomycin A1 and by the presence of the iron chelator 2,2′-bipyridyl. These data demonstrate that rapid iron liberation from IONPs at acidic pH and iron-catalyzed ROS generation are involved in the IONP-induced toxicity of microglia and suggest that the relative resistance of astrocytes and neurons against acute IONP toxicity is a consequence of a slow mobilization of iron from IONPs in the lysosomal degradation pathway.


Introduction
Iron oxide nanoparticles (IONPs) have gained substantial attention for biomedical applications and are already used as contrast agent in magnetic resonance imaging (MRI), for cancer treatment by hyperthermia or for targeted drug delivery (Li et al., 2013;Mahmoudi et al., 2011;Maier-Hauff et al., 2011). Nanoparticles can enter the brain via the intact or damaged blood-brain barrier (BBB), via the olfactory nerve or by direct injection into brain in the case of hyperthermia treatment (Petters et al., 2014b). By such applications all types of brain cells are likely to encounter IONPs in vivo.
In addition to neurons which are responsible for signal transduction, transmit information along their axon and release neurotransmitters to other neurons (Peters & Connor, 2014;Scalettar, 2006), the brain contains various types of glial cells in substantial numbers (Peters & Connor, 2014). Microglial cells are the immune competent cells of the brain and clear the brain from cell debris and intruders (Eyo & Dailey, 2013;Nayak et al., 2014). In case of incident, microglial cells become activated, transform from their resting ramified into their amoeboid form and migrate to the site of interference (Kettenmann et al., 2011;Nayak et al., 2014). Astrocytes cover with their processes most of the brain capillaries which form the BBB (De Bock et al., 2014), supply other brain cell types with nutrients (Bouzier-Sore &Pellerin, 2013; Stobart & Anderson, 2013), are responsible for the homeostasis of ions, water and metals Hohnholt & Dringen, 2013;Verkhratsky et al., 2014) and have important detoxifying functions in brain Fernandez-Fernandez et al., 2012).
Several studies have investigated the uptake and biocompatibility of IONPs in culture models for the different types of brain cells, including primary cultures of astrocytes, neurons and microglial cells (for overview, see Petters et al., 2014b). However, data on a quantitative comparison of the uptake and cytotoxicity of IONPs in different types of brain cells are scarce. In brain cell culture systems, microglial cells have been reported to take up fluorescent nanospheres and IONPs more efficiently than astrocytes and neurons (Fernandes & Chari, 2014;Jenkins et al., 2013;Pinkernelle et al., 2012). In addition, in co-cultures, IONP uptake into astrocytes and oligodendrocyte precursor cells was lowered in the presence of microglial cells (Pickard & Chari, 2010). In vivo, mainly macrophage-like cells were found IONP-positive after injection of IONPs into human glioma (van Landeghem et al., 2009), supporting the hypothesis that microglial cells have also in vivo a high potential to accumulate IONPs. Macropinocytosis and clathrin-mediated endocytosis have been identified as molecular mechanisms involved in the IONP uptake by microglial cells which direct the accumulated IONPs into the lysosomal degradation pathway (Luther et al., 2013).
The strong accumulation of IONPs by microglia has been connected with a high vulnerability of these cells towards IONPs (Jenkins et al., 2013;Luther et al., 2013) while astrocytes (Geppert et al., 2011(Geppert et al., , 2012Jenkins et al., 2013) and neurons (Petters & Dringen, 2015;Sun et al., 2013) appear to be relative resistant against acute IONP toxicity. However, the mechanisms involved in the toxicity of IONPs for microglial cells and the reason for the apparent resistance of astrocytes and neurons against IONPs are currently unknown. To address such questions, we have exposed primary cultures of microglia, neurons and astrocytes under identical experimental conditions to fluorescent IONPs and directly compared the potential of the different neural cells for IONP accumulation as well as the potential of IONPs to compromise cell viability. Here, we report that cultured microglial cells accumulated low concentrations of fluorescent IONPs more efficiently than astrocytes and neurons, while high concentrations of IONPs caused severe ROS production and toxicity in microglial cells but not in astrocytes and neurons. IONP-induced damage of microglial cells was prevented by neutralization of the acidic lysosomal pH or by chelation of iron ions, suggesting that in microglial cells the rapid lysosomal mobilization of iron from internalized IONPs causes accelerated ROS formation and oxidative cell damage.
Samples for transmission electron microscopy (TEM) were prepared by dropping 10 mL of 1 mM IONPs dispersed in water or glia-conditioned medium (GCM) onto a carbon-coated copper grid and subsequent air drying at room temperature, followed by washing the grid with distilled water in case of IONPs in GCM. Images were taken by a FEI Tecnai F20 S-TWIN (Hillsboro, OR) operated at 200 kV equipped with a GATAN GIF2001 SSC-CCD camera. Energy dispersive X-ray analysis was used for elemental analysis in the scanning mode of the microscope (STEM) with an EDAX r-TEM-EDX-detector with an energy resolution of 136 eV measured at Mn-Ka.
The hydrodynamic diameter and the z-potential of IONPs in suspension were determined by dynamic and electrophoretic light scattering in a Beckman Coulter (Krefeld, Germany) DelsaTM Nano Particle analyzer at 25 C at scattering angle of 165 and 15 , respectively. Concentrations of IONPs refer to the total concentration of iron in dispersion and not to the concentration of nanoparticles.

Cell culture
Microglial cultures were generated by tryptic removal of the astrocyte layer of astrocyte-rich primary cultures in 6-well plates, one day prior to experiments as described previously (Luther et al., 2013). The remaining attached microglia were washed once with culture medium (90% DMEM, 10% fetal calf serum, 25 mM glucose, 1 mM pyruvate, 18 units/mL penicillin G and 18 mg/mL streptomycin sulfate) and incubated in 2 mL glia-conditioned medium (GCM) before incubations with IONPs were started. Immunocytochemical characterization of the microglial cultures revealed that more than 98% of the cells in these cultures are positive for the microglial marker protein CD11b (Luther et al., 2013).
Astrocyte-rich primary cultures were prepared from newborn Wistar rats as described recently in detail (Tulpule et al., 2014). Cells were seeded at a density of 375 000 cells/well in 6-well plates and used at day five in culture for experiments. Astrocyterich cultures are strongly enriched in cells positive for the astrocyte marker protein glial fibrillary acidic protein, contain some microglial and oligodendroglial cells but no neurons (Petters & Dringen, 2014;Tulpule et al., 2014).
Cerebellar granule neuron cultures were prepared from brains of 7-day-old Wistar rats as described recently in detail (Tulpule et al., 2014). The cells were seeded in neuron culture medium (90% MEM, 10% heat-inactivated FCS, 25 mM KCl, 30 mM glucose and 2 mM glutamine, 100 U/mL penicillin and 100 mg/mL streptomycin) in a density of 1 875 000 cells/well in 6-well plates. Neurons were used for experiments after seven days in culture. Immunocytochemical characterization of the neuron cultures revealed that 99% of the cells expressed the neuronal marker protein microtubule-associated protein-2 (Tulpule et al., 2014).

Experimental incubations
Glia-conditioned medium is required to avoid toxicity of microglial cells in cultures (Saura et al., 2003). In order to establish comparable incubation conditions for primary microglial cells, neurons and astrocytes, all primary cultures were incubated in GCM. GCM was prepared as previously described (Luther et al., 2013) by incubating astrocyte-rich primary cultures in 175 cm 2 flasks for 1 d with 50 mL culture medium. The medium was harvested, filtered through a 0.2 mm sterile filter and stored at 4 C for not more than three weeks (Luther et al., 2013). In case of experiments on cerebellar granule neurons, the GCM had to be supplemented with 25 mM KCl to prevent neuronal apoptosis (Contestabile, 2002).
To investigate acute consequences of IONPs on cultured primary neural cells, the cultures were incubated at 37 C in a humidified atmosphere with 10% CO 2 in 1 mL GCM without or with IONPs and/or other compounds as indicated in the legends of the figures and tables. IONPs were applied in concentrations of up to 3 mM iron. After incubation, cells were washed once with 1 mL phosphate-buffered saline (PBS, 10 mM potassium phosphate buffer, pH 7.4, containing 150 mM NaCl) and lysed in 1 mL 1% (w/v) Triton X-100 in PBS to measure cellular lactate dehydrogenase (LDH) activity or in 700 mL 50 mM NaOH to determine cellular protein and iron contents.
For cellular localization of fluorescent IONPs, cultures were incubated with 0.1 mM IONPs for 6 h or with 0.3 mM for 24 h followed by 90-min incubation with 100 nM lysotracker DND-99 as previously described (Luther et al., 2013). Cells were fixed with 3.5% (w/v) paraformaldehyde in PBS and washed twice prior to the analysis of cellular fluorescence by microscopy.

Determination of cell viability and ROS production
The cell viability was investigated by determining the cellular LDH activity and the cellular MTT reduction capacity as well as by propidium iodide (PI) staining. The cellular LDH activity was determined as previously described (Dringen et al., 1998;Tulpule et al., 2014). Aliquots of the Triton X-100 lysates were mixed with pyruvate and NADH and the decline in absorbance at 340 nm was measured using a Sunrise-Basic microtiter plate reader (Tecan, Grödig, Austria). As the presence of millimolar concentrations of IONPs in the incubation medium strongly disturbed the photometric LDH measurement (data not shown), cellular LDH activity and not LDH release from damaged cells was used to investigate loss of membrane integrity. Incubation of cells with up to 3 mM IONPs did not interfere with the measurement of the LDH activity in cell lysates, while higher concentrations did (data not shown).
The MTT reduction capacity was determined after a given incubation as described previously (Scheiber et al., 2010). Briefly, the cells were incubated with 1 mL of 1 mg/mL MTT in GCM for 90 min, the formazan crystals dissolved in 1 mL dimethyl sulfoxide and the absorbance at 540 nm was determined. Under the conditions used, the presence of IONPs in cells did not affect the analysis of MTT reduction capacity (data not shown).
For PI staining, the cells were incubated with 1 mL GCM containing 5 mM PI and 10 mM H33342 (Tulpule et al., 2014) for 15 min, washed once with PBS and monitored immediately by fluorescence microscopy using the settings stated below in the section ''Fluorescence microscopy''.
Presence of cellular ROS was investigated by rhodamine staining. The fluorescent rhodamine 123 is generated in cells by the ROS-dependent oxidation of dihydrorhodamine 123 (Geppert et al., 2012). The cultures were washed with 1 mL incubation buffer (20 mM HEPES, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 5.4 mM KCl, 145 mM NaCl, pH 7.4) and stained for 15 min with 1 mL of 5 mg/mL dihydrorhodamine 123 and 10 mM H33342 in incubation buffer. After the incubation cells were washed once with PBS and monitored immediately for fluorescence using the settings stated in the section ''Fluorescence microscopy''.

Determination of protein and iron content
The protein content of the cultures was determined by the Lowry method (Lowry et al., 1951) with bovine serum albumin as a standard. The iron content of IONP-treated cultures was measured by a modified ferrozine-based iron assay (Geppert et al., 2009). Briefly, 100 mL cell lysate was mixed with 100 mL of 10 mM HCl and 100 mL of a solution containing 2.25% (w/v) KMnO 4 and 0.7 M HCl. After incubation at 60 C for 16 h, 30 mL of a mixture of 1 M ascorbate, 2.5 M ammonium acetate, 6.5 mM ferrozine and 6.5 mM neocuproine in water was added and the absorbance at 540 nm was measured after 1 h. The iron content determined for a sample was normalized on the protein content of the respective sample.

Fluorescence microscopy
Stained cells were monitored for fluorescence using an Eclipse TE-2000U fluorescence microscope with a DS-QilMc camera (Nikon, Düsseldorf, Germany). The following filter sets were used for monitoring the BP-signal of fluorescent IONPs (excitation at 465-495 nm, emission at 505-515 nm, dichromatic mirror at 505 nm), for the H33342 staining (excitation at 330-380 nm, emission at 420 nm, dichromatic mirror at 400 nm) and for visualizing the staining by PI, rhodamine 123 and lysotracker DND-99 (excitation at 510-560 nm, emission at 575 nm, dichromatic mirror at 590 nm).

Presentation of data
Quantitative data shown in figures and tables are means ± SD of values that were obtained in experiments, which had been performed on n independently prepared cultures. Microscopic images are from representative experiments that were reproduced twice on independently prepared cultures with similar results. Signal intensities were increased using Adobe Photoshop 7.0 for better contrast. Significance of differences between groups of data was analyzed by ANOVA followed by the Bonferroni post-hoc test. Analysis of significance of differences between two groups of data was done by the unpaired t-test. p40.05 was considered as not significant.

Characterization of iron oxide nanoparticles
Fluorescent BP-DMSA-coated IONPs were synthesized and characterized as previously reported (Petters et al., 2014a). Dispersed in water or GCM the IONPs displayed spherical morphology in TEM images with a particle diameter of 5-20 nm as seen by transmission electron microscopy (Supplementary Figure S.1a,b) as previously reported (Petters et al., 2014a). Energy dispersive X-ray spectroscopy detected iron from the nanoparticle core, sulfur from the DMSA in the coat and fluorine from the BP in the coat (Supplementary Figure S.1c,d).
Dispersions of BP-DMSA-coated IONPs in GCM which contained 10% FCS contain the IONPs in the form of nanosized agglomerates. Compared to water dispersions, the IONPs dispersed in GCM had an increased hydrodynamic diameter, a broader size distribution as demonstrated by the increased polydispersity index, and a reduced negative surface charge as demonstrated by the less negative z-potential (Table 1). This was expected as high concentrations of salts in the medium have been reported to facilitate agglomeration at least for carboxylate-coated silver nanoparticles (Caballero-Díaz et al., 2013). Moreover, serum proteins are considered to form a protein corona around DMSA-coated IONPs, which is accompanied by an increase in diameter and a lowering of the negative z-potential (Geppert et al., 2013;Petters et al., 2014a). However, the protein corona itself is unlikely to contribute substantially to the strongly increased diameter as adsorbed proteins have been reported to form only a shell of a few nanometers (Hühn et al., 2013). 1 mM 90 ± 15*** 0.297 ± 0.029** À10 ± 4** 3 mM 158 ± 28**,## 0.263 ± 0.025* À10 ± 2*** Asterisks indicate significant differences between dispersion of IONPs in glia conditioned medium (GCM) and water (*p50.05, **p50.01, ***p50.001) and hashes between 1 and 3 mM IONPs (##p50.01). The data were obtained by measurements on three different batches of IONPs.
The average hydrodynamic diameter of dispersions of 3 mM IONPs in GCM was around 160 nm. A severe shift in size distribution (Supplementary Figure S.1f) caused for this condition a significant increase in the average hydrodynamic diameter compared to GCM dispersion of 1 mM IONPs (90 nm), which is consistent with the concentration-dependent alteration in particle size reported for citrate-coated IONPs in serum-containing culture medium (Safi et al., 2011). In contrast, polydispersity index and z-potential did not differ for IONP dispersions in GCM containing 1 mM or 3 mM iron (Table 1). Hydrodynamic diameter, polydispersity and z-potential of IONPs were not affected if GCM was supplemented with additional 25 mM KCl for incubation of neurons (data not shown). Thus, as the IONPs applied in GCM to cell cultures were partially agglomerated, the effects observed for cultured brain cells after exposure to IONPs were due to the presence of nanosized agglomerates of the IONPs applied.
Uptake and toxicity of iron oxide nanoparticles in microglial cultures Previously, it was shown that exposure of cultured microglial cells to IONPs for up to 6 h caused a time-dependent increase in cellular iron content and compromised cell viability (Luther et al., 2013). In order to investigate the mechanisms involved in this toxicity, we applied fluorescent IONPs in concentrations between 0.1 mM and 3 mM to cultured microglial cells and determined the cell viability and specific cellular iron contents after 6 h of incubation. As expected, the viability of the IONP-treated cells was compromised in a concentration-dependent manner ( Figure  1a). While microglial cells incubated without IONPs maintained a high cellular LDH activity and a high MTT reduction capacity during the 6 h of incubation, the presence of 3 mM iron in form of IONPs significantly lowered the cellular LDH activity (Figure 1a, Table 2) and the cellular MTT reduction capacity (Table 2) to around 50% of the values determined for control cells. In addition, microglial cultures exposed for 6 h in the absence of IONPs did not contain cells which were stained by the membrane-impermeable fluorescent dye PI ( Figure 2b) and hardly any cells characterized by elevated ROS formation (Figures 3b and 4b).
In contrast, most cells in microglial cultures that had been incubated with 3 mM IONPs were positive for PI ( Figure 2l) and ROS (Figures 3l and 4d). The incubation of microglial cultures with IONPs was accompanied by a concentration-dependent increase in the specific cellular iron content from 55 ± 44 nmol/mg protein of untreated cultures (n ¼ 4) to around 1000 and 2500 nmol/mg for cells that had been exposed to 1 mM and 3 mM IONPs, respectively (Figure 1b).
Comparison of the effects of iron oxide nanoparticles on different types of brain cell cultures To compare microglial cells with other types of brain cells regarding the consequences of an exposure to IONPs, primary cultures of astrocytes and neurons were incubated with IONPs under experimental conditions that were identical (with the exception of the essential presence of additional 25 mM KCl in the GCM for neurons) to those applied for microglial cells. In contrast to microglial cells, the presence of IONPs in concentrations of up to 3 mM did neither lower the cellular LDH activity of cultured astrocytes and neurons (Figure 1a and c) nor did it cause any severe increase in the cellular ROS production (Figure 4h and l). However, a delayed concentration-dependent toxicity was observed for cultured neurons after 24 h exposure to IONPs (Supplementary Figure S.2a,b), while the viability of cultured astrocytes was not compromised even after 24 h incubation with 3 mM IONPs (Supplementary Figure S.2c,d).
Incubation of cultured astrocytes or neurons with IONPs caused a concentration-dependent increase in the cellular iron contents (Figure 1b). After 6 h incubation with IONPs in concentrations of 0.1 mM or 0.3 mM, the specific cellular iron contents were significantly lower in astrocytes and neurons compared to the respective values determined for microglial cultures, while the iron contents did not significantly differ between the three types of neural cell cultures after incubations with 1 mM or 3 mM IONPs (Figure 1b). Direct correlation of the cellular LDH activity to the specific cellular iron content confirmed that the loss in microglial viability was almost proportional to the increase in specific cellular iron content while no cellular LDH loss was observed from IONP-treated astrocytes or neurons, despite of similar high specific iron Figure 1. Cell viability and iron accumulation in IONP-treated cultured microglial cells, astrocytes and neurons. The cultures were incubated with the indicated concentrations of IONPs for 6 h and the cellular LDH activity (a) and the specific cellular iron content (b) were determined. In panel c, the cellular LDH activity is plotted against the respective cellular iron content. The data represent means ± SD of results obtained in five to 12 (microglia) or three (astrocytes, neurons) experiments on individually prepared cultures. The 100% values for the cellular LDH activity of microglial cells, astrocytes and neurons are 505 ± 163 nmol/ (min Â mg) (n ¼ 20), 402 ± 81 nmol/(min Â mg) (n ¼ 3) and 718 ± 215 nmol/(min Â mg) (n ¼ 3), respectively. Asterisks indicate significant differences of data compared to the respective control (incubation without IONPs) for one type of neural culture (*p50.05, ***p50.001). Significant differences of data obtained for astrocytes or neurons compared to those recorded for microglial cells are indicated by hashes (#p50.05, ##p50.01, ###p50.001). contents in all three types of neural cell cultures after exposure to millimolar concentrations of IONPs (Figure 1c).

Cellular localization of fluorescent IONPs in cultured brain cells
Fluorescence microscopy was used to investigate the cellular localization of accumulated fluorescent IONPs. As application of millimolar concentration of fluorescent IONPs caused toxicity in microglia, the neural cell cultures were incubated for 6 h with only 0.1 mM IONPs and subsequently stained with lysotracker. Fluorescence microscopy revealed for all three types of IONPtreated neural cultures dotted fluorescence patterns for both BP-labeled IONPs and lysotracker ( Figure 5). However, the cellular distribution of fluorescence differed strongly between the culture types. In cultured microglial cells the majority of IONP fluorescence was found co-localized with lysotracker and displayed a mainly perinuclear staining (Figure 5a  Microglial cultures were incubated without (control) or with 3 mM IONPs in the absence (none) or presence of Bafilomycin A1 (100 nM), NH 4 Cl (10 mM), 2,2'-bipyridyl (100 mM) or 4,4 0 -bipyridyl (100 mM) for 6 h before the cellular LDH activity (100% ¼ 569 ± 184 nmol/(min*mg)), the cellular MTT reduction capacity (100% ¼ absorbance of 1.23 ± 0.2/well) and the specific cellular iron content were determined. The iron content of control cells that had been incubated for 6 h without IONPs was 55 ± 44 (n ¼ 4) nmol/mg protein. Asterisks indicate significant differences between cells that had been exposed to the indicated compound and the respective control cells (none) (*p50.05, **p50.01, ***p50.001). Significant differences between data obtained for incubations in the absence and presence of IONPs are indicated by hashes (#p50.05, ##p50.01, ###p50.001). Figure 3. Staining of reactive oxygen species in IONP-treated microglial cultures. The cells were incubated for 6 h without (a-j) or with 3 mM IONPs (k-t) in the absence (a,b,k,l) or the presence of 100 nM Bafilomycin A1 (c,d,m,n), 10 mM NH 4 Cl (e,f,o,p), 100 mM 2,2 0 -bipyridyl (g,h,q,r) or 100 mM 4,4 0 -bipyridyl (i,j,s,t) before the cells were incubated for 15 min with dihydrorhodamine to test for ROS generation. The nuclei of all cells present were visualized by H33342. The scale bar in panel t refers to 100 mm and applies to all panels.

Mechanism of IONP-mediated toxicity to microglia
Lysosomal localization of IONPs (Figure 5a-d) as well as the increased ROS formation (Figures 3l and 4d) and compromised viability (Figures 1b and 2l) of IONP-treated microglial cells suggests that rapid iron liberation from IONPs in lysosomes and subsequent iron-mediated ROS formation are involved in the observed IONP-mediated microglial toxicity. To test for this hypothesis, microglia were treated with NH 4 Cl or the vacuolar H + -ATPase inhibitor Bafilomycin A1 to neutralize lysosomes (Poole & Ohkuma, 1981;Teplova et al., 2007) and thereby preventing the release of iron ions from the nanoparticles. Alternatively, the cells were incubated with 2,2 0 -bipyridyl to chelate the ferrous iron ions released from IONPs (Hohnholt et al., 2010). As a control for potential iron-independent effects of the chelator, also the non-chelating isomer 4,4 0 -bipyridyl was applied as control substance. Microglia were treated in the presence and absence of 3 mM IONPs for 6 h with the above-mentioned compounds and the cell viability was investigated by determination of cellular LDH activity and MTT reduction capacity (Table 2) as well as by PI ( Figure 2) and ROS staining (Figure 3). None of the compounds applied to modulate IONP-induced toxicity altered significantly the specific cellular iron content of microglial cells after incubation for 6 h with 3 mM IONPs (Table 2). In the absence of IONPs, incubation for 6 h with NH 4 Cl, Bafilomycin A1, 2,2 0 -bipyridyl or 4,4 0 -bipyridyl did neither cause any loss in cellular LDH or MTT reduction capacity (Table 2) nor was any staining of PI-permeable (Figure 2d, f, h and j) or ROS-positive (Figure 3d, f and h, j) microglial cells observed. The H33342 staining demonstrated that none of the conditions used lowered obviously the number of cells attached to the cell culture dish (Figures 2 and 3).
Exposure of microglial cells to 3 mM IONPs caused a loss in cellular LDH activity by around 50%, which was completely prevented by the presence of NH 4 Cl, Bafilomycin A1 or the iron chelator 2,2 0 -bipyridyl, while the non-chelating 4,4 0 -bipyridyl did not rescue microglial cells from IONP-induced loss in cellular LDH activity (Table 2). Also the substantial IONP-induced loss in microglial MTT reduction capacity was prevented by the presence of NH 4 Cl and at least partially prevented by the presence of Bafilomycin A1 or 2,2 0 -biypyridyl, whereas application of 4,4 0bipyridyl did not rescue the MTT reduction capacity of IONPtreated microglial cells (Table 2). PI-staining of IONP-treated microglial cells (Figure 2) confirmed the results obtained on LDH loss and MTT reduction capacity of IONP-treated microglial cells (Table 2). While nearly all nuclei in microglial cultures were PIpositive after incubation with 3 mM IONPs (Figure 2l), hardly any PI-positive cells were found if the cells were incubated with IONPs in the presence of NH 4 Cl, Bafilomycin A1 or 2,2 0 -biypyridyl (Figure 2n, p and r). In contrast, presence of 4,4 0 -bipyridyl did not lower the number of PI-positive cells in IONP-treated microglial cultures (Figure 3t). Also, the strong ROS-staining observed for IONP-treated microglial cells ( Figure  3l) was completely prevented by the presence of Bafilomycin A1 (Figure 3n) and at least partially prevented by NH 4 Cl and 2,2 0bipyridyl (Figure 3p and r), while the presence of 4,4 0 -bipyridyl hardly affected ROS-staining of IONP-treated cells (Figure 3t).

Discussion
Application of IONPs has previously been reported to compromise the viability of cultured microglial cells (Jenkins et al., 2013;Luther et al., 2013;Neubert et al., 2015;Pickard & Chari, 2010). This vulnerability was confirmed for microglial cells treated with fluorescent DMSA-coated IONPs as demonstrated by a loss of cellular LDH activity and MTT reduction capacity as well as by increased permeability of the cell membrane for PI. The increase in cellular ROS-staining suggests that iron-catalyzed radical formation is involved in the damage caused by IONPs to microglial cells as previously also shown for peripheral macrophages and other cell types (Dwivedi et al., 2014;Hanini et al., 2011;Lee et al., 2014). The most likely pathway involved in the ROS generation and toxicity in IONP-treated microglial cells involves endocytotic uptake of IONPs, lysosomal degradation of IONPs to iron ions, iron-catalyzed ROS formation by the Fenton reaction and oxidative cell damage, as previously suggested (Jenkins et al., 2013;Luther et al., 2013).
Endocytotic uptake of IONPs and trafficking of the internalized IONPs to the lysosomal compartment have previously been demonstrate for cultured microglial cells (Luther et al., 2013) and was confirmed for the conditions used here by the colocalization of fluorescent IONPs with lysotracker in microglial cells. The ability of NH 4 Cl and Bafilomycin A1, which neutralize the lysosomal pH by diffusion of ammonia to the acidic lysosome (Poole & Ohkuma, 1981) and by inhibition of the lysosomal proton pump (Teplova et al., 2007), respectively, to lower or even prevent ROS formation and cell toxicity in IONP-treated microglial cells demonstrates that the low pH of lysosomes is involved in the IONP-induced microglial toxicity. This is consistent with literature data from cell-free experiments showing that iron ions are liberated from IONPs at lysosomal pH (Levy et al., 2010;Malvindi et al., 2014). Also, the presence of reducing substances such as glutathione in lysosomes (Kurz et al., 2010) may accelerate the liberation of iron ions from the IONPs and also helps to maintain iron in the ferrous state, which is required for export from the acidic compartment into the cytosol by the divalent metal transporter 1 (Lawen & Lane, 2013). This transporter has been reported to be expressed in microglial cells (Rathore et al., 2012;Urrutia et al., 2013). Thus, rapid lysosomal liberation of internalized IONPs in microglial cells is likely to generate large amounts of ferrous iron, which causes ROS formation and cell damage. This view is strongly supported by the ability of the membrane permeable ferrous iron chelator 2,2 0bipyridyl, but not its non-chelating analog 4,4 0 -bipyridyl (Kinnunen et al., 2002;Ma et al., 2006), to lower ROS generation and prevent toxicity in IONP-treated microglial cells. The residual ROS formation in microglial cells in the presence of 2,2 0bipyridyl might result from ROS production on the IONP-surface (Voinov et al., 2011) as the Fe 2+ -2,2 0 -bipyridyl complex is not Fenton reactive (Pierre & Fontecave, 1999). However, microglial cells appear to cope with a low rate of ROS production found under such conditions as their high antioxidative potential (Dringen, 2005;Hirrlinger et al., 2000) is likely to protect the cells against toxicity by a moderate amount of ROS.
Cultured astrocytes and neurons accumulated similar amounts of IONPs under the conditions used. The specific iron contents determined were in the range of values previously reported for these cultures in serum-containing media (Geppert et al., 2011(Geppert et al., , 2013Petters & Dringen, 2015). Comparison of microglial cultures with astrocytes and neurons revealed that microglial cells accumulated IONPs more efficiently when the different types of cultures were treated with low concentrations of 0.1 mM or 0.3 mM IONPs as demonstrated by the 2 -to 5-fold higher specific cellular iron contents in microglial cells. This stronger IONP accumulation by microglial cells is consistent with literature data for other types of IONPs (Jenkins et al., 2013;Pinkernelle et al., 2012). Different types of endocytotic pathways involved in IONP uptake are unlikely to explain the observed differences in IONP accumulation between the three types of neural cells investigated as in serum-containing media clathrinmediated endocytosis and macropinocytosis have been reported to mediate IONP uptake into both cultured microglial cells and astrocytes (Geppert et al., 2013;Luther et al., 2013), while neurons appear to take up IONPs predominantly by clathrinmediated endocytosis (Petters & Dringen, 2015). However, as microglia are the ''phagocytotic cells'' in the brain and as microglial cells, but not astrocytes and neurons, have the prominent function to incorporate and digest cell debris from their environment (Eyo & Dailey, 2013;Nayak et al., 2014), microglial cells are likely to be equipped with a strong endocytotic transport capacity, which could explain the higher efficiency of microglial cells to accumulate iron from low concentrations of IONPs. For incubations with 3 mM IONPs similar specific iron contents were determined for all three types of neural cultures. However, this apparent inability of microglial cells to accumulate more IONPs than astrocytes and neurons under these conditions is likely to be a direct consequence of the developing microglial toxicity which will impair IONP accumulation while astrocytes and neurons remained viable during the incubation and continued to accumulate IONPs.
The relative resistance of astrocytes and neurons towards high concentrations of IONPs is most likely caused by a slow transfer of internalized IONPs into the lysosomal compartment, which is required for liberation of iron ions from IONPs. This view is supported by the apparent absence of a co-localization of DOI: 10.3109/17435390.2015.1071445 fluorescent IONPs with lysotracker in cultured astrocytes and neurons that had been exposed for 6 h to IONPs. In contrast, IONP-treated microglial cultures showed under identical conditions strong co-localization of fluorescent IONPs with lysotracker. Thus, a more rapid internalization and lysosomal degradation of IONPs to ROS-generating iron ions in microglial cells is likely to cause the severe acute toxicity observed for this cell type after exposure to IONPs. Such processes appear to be slower in astrocytes and neurons, explaining the relative resistance of these cell types against acute IONP-induced toxicity (Geppert et al., 2011(Geppert et al., , 2012(Geppert et al., , 2013Petters & Dringen, 2015). However, astrocytes and neurons also have the potential to slowly liberate iron from accumulated IONPs. At least for other experimental conditions some ROS formation has been reported for IONP-treated astrocytes and neurons (Geppert et al., 2012;Petters et al., 2014b) and depending on the type of IONPs applied and the incubation conditions used also delayed toxicity was observed for these cell types (Petters & Dringen, 2015;Rivet et al., 2012;Sun et al., 2013). Indeed, treatment of cultured neurons for 24 h with IONPs resulted in a concentration-dependent loss of cell viability, although at best little lysosomal localization of IONPs and ROS production was observed for IONP-treated neurons. This suggests that under the conditions used, cultured neurons may suffer from a particle-induced toxicity as previously suggested for various types of nanomaterials (Nel et al., 2006;Yang et al., 2009) and that lysosomal iron liberation and subsequent iron-mediated ROS production is not predominately responsible for IONP-induced neuronal toxicity. In contrast to microglial cells and neurons, astrocytes appear to be highly resistant against IONP-induced toxicity. Although after a 24 h incubation, substantial amounts of IONPs were found in the lysosomes and substantial ROS production was detected, the viability of astrocytes was not compromised. Likely reason for this apparent discrepancy is the well known strong antioxidative potential of astrocytes Fernandez-Fernandez et al., 2012). In addition, the known upregulation of the iron storage protein ferritin in IONP-treated astrocytes (Geppert et al., 2012) will help to lower the cellular concentration of free iron ions and will thereby lower the rate of iron-catalyzed ROS formation in astrocytes. Further studies are now required to investigate for sub-toxic IONP concentrations the fate of IONPs and of IONP-derived iron in the different types of brain cells, including a potential storage of IONP-derived iron ions in ferritin as well as iron export by ferroportin.

Conclusions
In the present study, we show that microglia are more susceptible to acute IONP exposure as astrocytes and neurons, which is likely to be caused by fast lysosomal iron liberation from IONPs in microglial cells which causes severe iron-mediated ROS production and oxidative cell damage. This suggests that also in vivo microglia may be harmed more strongly by IONPs than other brain cells. However, microglial toxicity in brain as a consequence of an IONP exposure is likely to depend on the route, which mediates entry of IONPs into the brain. Little harm to microglial cells is expected if IONPs are applied to the blood, e.g. as MRI contrast agent, as IONPs which pass the intact BBB via transcytosis (Yan et al., 2013) encounter astrocytes. These cells cover almost completely the brain capillaries (De Bock et al., 2014) and have at least in culture the capacity to accumulate large amounts of IONPs without being damaged by the accumulated IONPs (Geppert et al., 2012;Jenkins et al., 2013;present study). In contrast microglial cells are very likely to encounter bloodderived IONPs, if the BBB is damaged by injury or tumors (Nayak et al., 2014). For such conditions, predominantly microglial cells have been reported to take up IONPs (Neuwelt et al., 2004;Taschner et al., 2005). Also intranasal application of IONPs (Wang et al., 2011) as well as direct injection of IONPs into brain tumors for magnetic fluid hyperthermia (van Landeghem et al., 2009) establish direct contact of microglial cells with IONPs, which causes microglial IONP accumulation and microglial activation (Wang et al., 2011). Such an IONPinduced activation of microglial cells might result in neurotoxicity (Correale, 2014). However, IONP-induced microglial toxicity will subsequently impair the neuroprotective functions of microglia in brain including the defense against pathogens and the support in repair processes (Kettenmann et al., 2011;Nau et al., 2014). For such conditions, strategies to prevent the fast IONP internalization, the liberation of iron ions from internalized IONPs and/or iron-mediated ROS production should be considered in order to prevent IONP-induced microglial toxicity.