Nanoparticles Induce Changes of the Electrical Activity of Neuronal Networks on Microelectrode Array Neurochips

Background Nanomaterials are extensively used in industry and daily life, but little is known about possible health effects. An intensified research regarding toxicity of nanomaterials is urgently needed. Several studies have demonstrated that nanoparticles (NPs; diameter < 100 nm) can be transported to the central nervous system; however, interference of NPs with the electrical activity of neurons has not yet been shown. Objectives/methods We investigated the acute electrophysiological effects of carbon black (CB), hematite (Fe2O3), and titanium dioxide (TiO2) NPs in primary murine cortical networks on microelectrode array (MEA) neurochips. Uptake of NPs was studied by transmission electron microscopy (TEM), and intracellular formation of reactive oxygen species (ROS) was studied by flow cytometry. Results The multiparametric assessment of electrical activity changes caused by the NPs revealed an NP-specific and concentration-dependent inhibition of the firing patterns. The number of action potentials and the frequency of their patterns (spike and burst rates) showed a significant particle-dependent decrease and significant differences in potency. Further, we detected the uptake of CB, Fe2O3, and TiO2 into glial cells and neurons by TEM. Additionally, 24 hr exposure to TiO2 NPs caused intracellular formation of ROS in neuronal and glial cells, whereas exposure to CB and Fe2O3 NPs up to a concentration of 10 μg/cm2 did not induce significant changes in free radical levels. Conclusion NPs at low particle concentrations are able to exhibit a neurotoxic effect by disturbing the electrical activity of neuronal networks, but the underlying mechanisms depend on the particle type.

The extensive use and development of nanomaterials demand intensified research efforts regarding toxicity and possible health effects of nanomaterials [e.g., nano particles (NPs), nanotubes]. The small size of NPs provides them with special properties, such as high surfaceto-volume ratio and high surface charge (Linse et al. 2007). In daily life, NPs can be found almost everywhere (e.g., in cosmetics, dusting powder, paints). Because of their small size, NPs can trans locate from the inter stitial region of the lung after respiratory exposure and can penetrate through the skin after application in cosmetics, reaching different organs of the body, such as liver, spleen, kidney, and brain, through blood circulation (L'Azou et al. 2008).
The potential translocation of different types of NPs into the central nervous system (CNS) via the olfactory pathway has been reported by Oberdörster and Utell (2002) and Nel et al. (2006; for reviews, see Oberdörster et al. 2004Oberdörster et al. , 2009Papp et al. 2008). Intranasally instilled anatase titanium dioxide (TiO 2 ) NPs in female mice translocated into the CNS and caused a potential lesion in the hippocampus region of the brain (Wang et al. 2008). Nanoscale TiO 2 particles stimulate brain microglia to produce reactive oxygen species (ROS) through oxidative burst and interfere with mitochondrial energy production in vitro (Long et al. 2006). To this point, there is no evidence regarding interference of NPs with electrical activity of neurons. High levels of ROS in the brain have been related to the degeneration of neurons and their mitochondria, leading to the development of Alzheimer's (Sompol et al. 2008) and Parkinson's (Smith and Wayne 2007) diseases.
Spontaneously active networks in culture have been proposed as a sensitive and efficient model system to study the neuroactive and toxic properties of chemicals (Gramowski et al. 2006b;Gross et al. 1997;Stett et al. 2003). In vitro neuronal networks represent a simplified model to study neural electrophysiology (i.e., the functional output of cellular activity in the CNS). Cultured networks coupled to micro electrode array (MEA) neurochips constitute a valuable tool to investigate changes in the electrophysiological activity of neurons in response to chemi cal exposure. Neuronal networks respond to transmitters, their blockers, and many other pharmacological substances in a histio typic manner; responses are substance specific and similar to those observed in vivo (Bal-Price et al. 2008;Gramowski et al. 2000Gramowski et al. , 2004Gramowski et al. , 2006bKeefer et al. 2001b; Morefield et al. 2000).
In the present study we investigated the influence of NPs on the electrical activity changes in neuronal networks on MEA neurochips. For this study we used three different kinds of NP: carbon black (CB), TiO 2 , and hematite (Fe 2 O 3 ). We exposed primary cortical networks cultured on MEA neurochips to the different NPs in an acute concentration-dependent manner. The action potential patterns of the electrical activity of the neuronal networks were recorded and changes in four activity categories-general activity, burst structure, synchronicity, and oscillatory behavior-were quantified by multiparametric pattern analysis. Additionally, deposition and trans location of NPs in the neuronal network were observed using scanning electron microscopy and transmission electron microscopy (TEM), and acute functional toxicity, cytotoxicity, and radical formation were evaluated.
Background: Nanomaterials are extensively used in industry and daily life, but little is known about possible health effects. An intensified research regarding toxicity of nanomaterials is urgently needed. Several studies have demonstrated that nanoparticles (NPs; diameter < 100 nm) can be transported to the central nervous system; however, interference of NPs with the electrical activity of neurons has not yet been shown. oBjectives/methods: We investigated the acute electrophysiological effects of carbon black (CB), hematite (Fe 2 O 3 ), and titanium dioxide (TiO 2 ) NPs in primary murine cortical networks on microelectrode array (MEA) neurochips. Uptake of NPs was studied by transmission electron microscopy (TEM), and intra cellular formation of reactive oxygen species (ROS) was studied by flow cytometry. results: The multiparametric assessment of electrical activity changes caused by the NPs revealed an NP-specific and concentration-dependent inhibition of the firing patterns. The number of action potentials and the frequency of their patterns (spike and burst rates) showed a significant particledependent decrease and significant differences in potency. Further, we detected the uptake of CB, Fe 2 O 3 , and TiO 2 into glial cells and neurons by TEM. Additionally, 24 hr exposure to TiO 2 NPs caused intra cellular formation of ROS in neuronal and glial cells, whereas exposure to CB and Fe 2 O 3 NPs up to a concentration of 10 µg/cm 2 did not induce significant changes in free radical levels. conclusion: NPs at low particle concentrations are able to exhibit a neurotoxic effect by disturbing the electrical activity of neuronal networks, but the under lying mechanisms depend on the particle type. The surface area of all the NPs was measured using the Brunauer-Emmett-Teller (BET) technique at the Nanolab of Focas Institute (Dublin Institute of Technology, Dublin, Ireland). For BET, the samples were analyzed in a dry state by nitrogen adsorption after degassing at 300°C (Gemini 2360 analyzer; Micromeritics Instrument Corp., Norcross, GA, USA). Zeta potential values were obtained using the Smoluchowski model for analysis, and particle size distribution was measured using the dynamic light-scattering technique (Nano Zetasizer ZS; Malvern Instruments, Worcestershire, UK) after suspending the NPs in fetal bovine serum-free cell culture medium by ultra sonication in an ultrasonic water bath.
Primary frontal cortex cell culture. Frontal cortex tissue was harvested from embryonic day 15 crl:NMRI mice. After ethyl ether anesthesia, mice were sacrificed by cervical dislocation according to the German Animal Protection Act (1998). Animals were treated humanely and with regard for alleviation of suffering. Frontal cortex tissue was cultured according to the method of Ransom et al. (1977) with minor modifications: DNase I (8,000 U/mL) and papain (10 U/mL) were used for tissue dissociation (Huettner and Baughman 1986). This method has been described in detail previously (Gramowski et al. 2006a).
Primary frontal cortex networks cultured on MEA neurochips routinely develop spontaneous electrical activity, which starts after approximately 3-4 days in vitro (Gramowski et al. 2004) in the form of random spiking. This stabilizes after 4 weeks in culture into a synchronized activity pattern composed of a coordinated burst pattern and interburst spiking [see Supplemental Material, Figure 1 (doi:10.1289/ehp.0901661)]. Such networks can remain spontaneously active and pharmaco logically responsive for > 6 months and show high inter culture repeatability (Gramowski et al. 2006b;Gross 1994).
MEA neurochips recording. MEA neurochips were provided by the Center for Network Neuroscience (University of North Texas, Denton, TX, USA). These 5 × 5 cm 2 glass chips have a central 1-mm 2 recording matrix with 64 passive gold electrodes and indium tin oxide conductors. Fabrication techniques and culture methods have been described previously (Gross 1994;Gross et al. 1985).
The particle concentration range for the three NPs investigated in the electrophysiological studies was chosen based on the effective range on the frontal cortex activity.
In preliminary experiments, we determined the complete acute (12 hr) effective range for each NP and determined the accumulating concentrations to optimally cover the effective range. Before application to the MEA neurochip chamber or 12-well chamber, the NPs were dispersed by ultrasound sonification and vortexed.
Multichannel recording and data analysis. For extra cellular recording, MEA neurochips were placed into sterilized constant-bath recording chambers (Gross 1994) and maintained at 37°C. Recordings were made in Dulbecco's modified Eagle's medium/10% horse serum at 1 mL chamber volume. The pH was maintained at 7.4 with a continuous stream of filtered, humidified air with 10% CO 2 . Sets of 32 pre amplifiers were positioned to either side of the recording chamber, and recording was performed with the multichannel acquisition processor system and a computer-controlled 64-channel amplifier system (Plexon, Inc., Dallas, TX, USA). The total system gain used was 10 K, with a simultaneous 40-kHz sampling rate. Spike identification and separation were accomplished with a template-matching algorithm in real time [see Supplemental Material, Figure 1B High content analysis of the network activity patterns provides a multi parametric description characterizing the changes in four categories: general activity, burst structure, synchronicity, and oscillatory behavior. We quantified the substance-specific activity changes by extracting a total of 35 activitydescribing spike train parameters as previously described (Gramowski et al. 2006b). Four of these parameters quantify the concentration-response kinetics in their course, number of phases, slope (Hill coefficient), and 10%, 50%, and 90% effective concentrations (EC 10 , EC 50 , and EC 90 , respectively) of the maximum inhibitory effect induced by the NPs for the frontal cortex spike rate (SR). The remaining 31 parameters [listed in Supplemental Material, Table 1 (doi:10.1289/ ehp.0901661)] include parameters representing general activity [e.g., SR and burst rate (BR)] and burst structure (e.g., burst duration and number of spikes within a burst).
For direct comparability, all parameters were normalized for each experiment and each experimental treatment with respect to the corresponding values of the native reference activity. Values were derived from 60-sec bin data from 30 min after stabilization of activity. From one network, between 24 and 101 separate neurons were simultaneously reordered.
The changing BR as a function of the concentration was fitted to a sigmoidal doseresponse curve in order to obtain dose-response curves for each experiment. Where the activity changes showed a biphasic or multi phasic behavior in the dose-response curve, the phases were fitted and quantified separately.
Raster scanning electron microscopy (REM). Neuronal networks on MEA neuro chips were incubated for 24 hr with the three NP types, fixed in 4% glutaraldehyde for 1 hr, washed with phosphate-buffered saline (PBS), postfixed with 1% osmium tetroxide (OsO 4 ), dehydrated in a graded series of acetone, and dried with a critical point dryer. The sample was coated with gold in a sputter coater and studied in the REM (DSM 960A; Carl Zeiss AG, Oberkochen, Germany).
TEM. Before application of NPs to the cortical networks on MEA neurochips, the NPs were dispersed by ultra sound sonification and vortexed. For incubation, the networks were washed twice with serum-free medium and incubated for 24 hr with the three types of NPs (CB, 10 µg/cm 2 ; TiO 2 , 10 µg/cm 2 ; Fe 2 O 3 , 5 µg/cm 2 ). At the end of incubation, the mono layers were washed with serum-free medium and fixed with 4% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.4) for 1 hr at 4°C. The fixed mono layers were removed from the culture dishes with a rubber policeman, washed in 0.1 sodium phosphate buffer, and post fixed with 1% OsO 4 in the same buffer for 1 hr at 4°C. The sample pellets were then washed in PBS and dehydrated. After postfi xation, the samples were washed in PBS, dehydrated in a graded series of acetone, and embedded in the epoxy resin Araldite (Fluka, Buchs, Switzerland). Ultrathin sections were cut with an Ultracut S ultra microtome (Leica, Wetzlar, Germany), mounted on copper grids, stained with uranyl acetate and lead citrate, and studied in an EM 902 A TEM with electron energy loss spectroscopy (EELS) or Libra 120 TEM with EELS and electron-dispersive X-ray spectroscopy (Carl Zeiss AG). Digital pictures were taken with a 2K CCD camera (Proscan, Lagerlechfeld, Germany). The particle size was measured using the EFTEM software package (Zeiss-SIS, Jena, Germany) at a magnification of 15,000×. Measurement of ROS. ROS production was detected using the dihydro rhodamine 123 (DHR), which is oxidized by hydrogen peroxide (H 2 O 2 ), hypochlorous acid (HClO • ), and peroxy nitrite anion (ONOO • ) to the fluorescent dye rhodamine 123. We measured the fluorescence intensity of reduced rhodamine 123 at 525-nm emission wavelength by a flow cytometer (EPICS XL-MCL4; Beckman Coulter, Krefeld, Germany) as previously described by Lantow et al. (2006). After exposure to the different NPs at concentrations ranging from 0.5 to 10 µg/cm 2 for 24 hr, neuronal cultures were incubated with 1 µM DHR (end concentration) in Hank's balanced salt solution (0.9% NaCl, 14 mM HEPES, pH 7.4) for 25 min at 37°C in the dark. ROS production was immediately measured by flow cytometry. For data analysis, we used EXPO32 MultiCOMP analysis software (version 1.2; Beckman Coulter). Results are expressed as mean ± SD from three independent experiments, in which all measure ments were carried out in triplicate.
Statistical analysis. The electro physio logical results are expressed as series means ± SE. The absolute parameters' distributions were tested for normality. For the electrophysiological studies, we assessed the level of significance after compound application using SPSS statistical software, version 17.0 (SPSS, Chicago, IL, USA). Significant changes induced by substance application were tested by analysis of variance (ANOVA) followed by Dunnett's multiple comparison post hoc test with native activity as the common control.
To determine the EC 50 , standard logistic concentration-response curves (either one or the sum of two, depending on the data) were fitted to the data points using the non linear regression algorithm of the Solver module in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA). The end point was set to the maximum effect of the SR and BR network changes induced by the respective NP.
For the ROS measurements, we determined statistical differences between the means using the ANOVA test. Values of p < 0.05 were considered significantly different from control.

Results
Physicochemical analysis. Table 1 shows the surface area, zeta potential, and average hydro dynamic diameter (particle size distribution) of the NPs used. Electron-dispersive X-ray analysis for detecting the elemental composition of the NPs revealed a carbon peak for CB NPs, an iron peak for Fe 2 O 3 NPs, and a titanium peak for TiO 2 NPs ( Table 1) Influence of CB NPs on electrical activity of cortical networks. We tested CB in a concentration range of 0.001-300 µg/cm 2 (n = 8). The high interculture repeatability of the electro physiological activity response induced by the three NPs is demon strated in Supplemental Material, Figure 3 (doi:10.1289/ ehp.0901661). CB induced biphasic concentration-dependent activity changes in the electrical activity of cortical networks. At low concentrations (0.03-100 µg/cm 2 ), CB evoked a reduction of the general activity (phase 1). This initial activity drop was followed by an increase of the cortical network SR activity at higher CB concentrations (10-300 µg/ cm 2 ) (phase 2; Figure 1). In phase 1 the significant decrease in the general activity started at 0.03 µg/cm 2 SR, 92.0 ± 1.3%; BR, 93.9 ± 1.4%). We observed a maximum decrease in the general activity at 30 µg/cm 2 (SR, 67.2 ± 4.3%; BR, 66.1 ± 4.4%). To determine the EC 50 , we fitted the dose-response curves to the maximum inducible effect of the NPs. The initial value was set to 100% and the final value was not fixed. Fitted EC 10 , EC 50 , and EC 90 values for SR in this activity decline (phase 1) were 0.001, 0.915, and 851 µg/cm 2 , respectively ( Table 2). The Hill coefficient for the dose-response curve was 0.32 for the phase 1 slope. During phase 1, activity changes decreased with rising CB concentrations beginning at 10 µg/cm 2 , which is reflected by an increasing coefficient of variation (CV) averaged over the network (CV net ) for the BR (174.0 ± 24.1%; see Supplemental Material, Figures 4 and 5). These changes were accompanied by a decomposition of the network oscillation at 20 µg/cm 2 and upward, which is reflected by the increasing CV averaged over time (CV time ) for the BR (203.0 ± 26.3%). The number of bursting units remained unaltered.
The activity decrease of cortical networks was followed by a CB-induced switch in activity changes characterized by an enhancement of activity at higher concentrations (phase 2). This phase 2 activity induction occurred at 100-300 µg/cm 2 CB (Figure 1, Table 2). SR and BR showed a maximum increase of 16.3% and 13.1%, respectively. These activity changes were not accompanied by changes in the burst structure. Within this activity enhancement, network synchronization and network oscillation behavior became more coordinated. Influence of CB NPs on cortical network morphology. Studies of ultra thin sections of CB-treated neuronal networks in TEM demon strated the presence of the NPs in the cells (Figure 2). A few CB particles with a diameter of about 55 nm were bound to the cell surface. We detected no obvious cell damage or injury after 24-hr exposure.

Influence of CB NPs on intracellular ROS formation.
We investigated the effects of the different NPs at concentrations of 0.5, 5, and 10 µg/cm 2 after 24-hr exposure on the formation of ROS (Figure 3). Results show that exposure to CB NPs did not induce changes in the ROS level in neuronal networks.
Influence of Fe 2 O 3 NPs on electrical activity of cortical networks. We tested Fe 2 O 3 NPs for neuro toxic potential in concentrations of 0.1-100 µg/cm 2 (n = 16 per treatment). These NPs caused less pronounced changes in the cortical activity network patterns than did CB NPs. The general activity showed a steady reduction with rising Fe 2 O 3 concentrations, with a maximum decline in SR and BR to 71.0 ± 3.1%, and 76 ± 3.7%, respectively, at 100 µg/cm 2 (the highest concentration tested) (Figure 4). At a concentration of 0.5 µg/cm 2 , Fe 2 O 3 induced a significant decrease in activity. We calculated the EC 10 , EC 50 , and EC 90 values for the SR as 0.025, 6.6, and 1,760 µg/cm 2 , respectively (Table 2). This effect on SR was accompanied by changes in the burst structure (specifically, a decrease in the number of spikes in burst to 88.1 ± 2.4%) (Figure 4) and by a reduction in synchronicity and oscillatory behavior [see Supplemental Material, Figures 6 and 7 (doi:10.1289/ehp.0901661)]. However, even at the highest concentration tested, all neurons in the network remained actively bursting. In 16 of the 31 activity parameters, Fe 2 O 3 induced significant changes.

Influence of Fe 2 O 3 NPs on cortical network morphology.
Binding of Fe 2 O 3 to the cell surface of neuronal network cells and particle uptake by endo cytosis was visible by electron microscopy ( Figure 5). In about 1% of investigated cell preparations, we found NPs in what looked like a loose neuropil also containing synapses ( Figure 5B). A differentiation of glial cells or the special subtypes of neurons was not possible in TEM.

Influence of Fe 2 O 3 NPs on intracellular ROS formation.
Similar to results with CB, exposure of neuronal networks to Fe 2 O 3 NPs did not induce a statistically significant increase of ROS formation up to a tested concentration of 10 µg/cm 2 and an exposure time of 24 hr ( Figure 3).

Influence of TiO 2 NPs on electrical activity of cortical networks.
We tested TiO 2 NPs at concentrations ranging from 0.01 to 300 µg/ cm 2 (n = 10). TiO 2 NPs caused severe inhibition of the general electrical network activity, with 1 µg/cm 2 TiO 2 NPs. At 300 µg/cm 2 , SR and BR dropped to 53.3 ± 5.5%, and 60.3 ± 5.7%, respectively ( Figure 6). The EC 10 , EC 50 , and EC 90 values for the SR in this activity decrease were 0.11, 665, and 3,840 µg/cm 2 , respectively (Table 2). With rising TiO 2 concentrations, the synchronicity and the oscillatory regularity of the network activity pattern were reduced. This was observed at 300 µg/cm 2 in the increasing BR CV time (216.9 ± 27.9%) and CV net (231.6 ± 40.1%) [see Supplemental Material, Figures 8 and 9 (doi:10.1289/ehp.0901661)]. In addition, the burst structure changed with increasing concentrations. The burst duration was markedly reduced to 73.9 ± 9.2% and the number of bursts to 74.0 ± 6.5%. However, all neurons remained bursting over the time of the concentration-response experiments. In 17 of the 31 activity parameters, TiO 2 induced significant changes. Figure 7 demonstrates the uptake of TiO 2 into neuronal network cells as well as an accumulation of particles or particle agglomerates in a space near the cell surface.

Influence of TiO 2 NPs on cortical network morphology.
Influence of TiO 2 NPs on intra cellular ROS formation. TiO 2 NPs induced a statistically significant and concentration-dependent increase in ROS formation in neuronal cells compared with the untreated control. They induced a substantial increase of ROSproducing cells (76.07%) compared with CB (31.02%) and Fe 2 O 3 (37.64%) NPs ( Figure 3A) and considerably higher ROS levels than did Fe 2 O 3 NPs at the same concentration ( Figure 3B).

Discussion
Properly anticipating the potential adverse effects caused by NPs requires a fundamental understanding of their physi cal properties and distribution, as well as a detailed multiparametric analysis of their cellular effects. Particle properties. The dynamic lightscattering analysis demonstrated that NPs suspended in serum-free liquid medium formed agglomerates; because of that, the hydrodynamic diameter increased. This can be due to the presence of weak van der Waals forces or strong chemical bonds between the individual NPs. Agglomeration of suspended NPs depends on their zeta potential, which in turn relies on several factors such as ionic strength, pH, surface charge, and surface coating (Jiang et al. 2009). Stable dispersions of NPs in solution occur only at zeta potentials greater than 30 mV (positive or negative) (Wagner et al. 2007). In the present study, we observed a clear correlation between zeta potential and agglomeration state of the NPs by determining the hydrodynamic diameter of the NPs. CB NPs with a low zeta potential of 14 mV (positive or negative) indeed showed a high degree of agglomeration, whereas the other types with zeta potentials of 29 mV and 49 mV (positive or negative) remained monodisperse to a large extent.
Effects on the electrical network activity in vitro. For the activity changes in frontal cortex networks, SR and BR were the predomi nant parameters quantifying the specific concentration-dependent responses.
The results indicate that even very low particle concentrations (1-10 ng/cm 2 ) can disturb the electrical activity of neuronal networks in vitro. All three types of NPs induced significant changes of SRs and BRs, whereby CB had the highest efficiency at the lowest particle concentrations, followed by Fe 2 O 3 and TiO 2 NPs. In contrast, TiO 2 induced the strongest inhibition in the network activity. CB NPs induced a signifi cant activity drop at lower concentrations (1-100 µg/cm 2 ), whereas higher concentrations (100-300 µg/ cm 2 ) induced an increase in the general activity, synchronicity, and oscillatory regularity. Although concentrations > 100 µg/cm 2 will rarely occur, this condition would resemble the electrical activity pattern of epileptic seizures (Keefer et al. 2001a). In the present study we did not observe this biphasic pattern for TiO 2 and Fe 2 O 3 NPs.
Particle concentrations. Although in vivo to in vitro comparisons are difficult in general, our electrophysiological in vitro findings correlate to some degree with in vivo studies. Exposure concentrations used for CB and TiO 2 studies are in the range of 6-10 mg/m 3 (Gallagher et al. 1994), which corresponds to 6-10 ng/cm 2 in our in vitro experi ments. The recommended exposure limit suggested by the U.S. National Institute for Occupational Safety and Health (1988) is 3.5 mg/m 3 for CB and < 5 mg/m 3 (corresponding to 5 ng/cm 3 ) for personal inhalable dust exposure. The lowest effective CB concentrations in the present study were in the range of a few nanograms per square centi meter. However, concentrations of NPs or metals that reach the brain are much lower than exposure concentrations. For example, in mice exposed intra nasally to 500 µg TiO 2 NPs (rutile, 80 nm/animal; anatase, 155 nm/animal) for 30 days, Wang et al. (2008) found 280 ng TiO 2 NPs per gram of brain tissue. Several studies regarding sizedependent accumulation and distribution were carried out with gold NPs. Sonavane et al. (2008) concluded from their studies that gold NPs < 50 nm are able to cross the blood-brain barrier (BBB). The TiO 2 NPs used by Wang et al. (2008) were > 50 nm and were found in the brains of exposed mice. Terentyuk et al. (2009) found 0.6 µg/mL gold NPs (15 nm) in rat brain after intra venous injection of 57 µg/ mL gold NPs. Smaller NPs cross the BBB more easily than do larger NPs. The NPs used in our own experiments were relatively large (50-100 nm); thus, it might be that the effects we observed would be even stronger using NPs < 50 nm. Also, NPs seem to bio accumulate in inner organs (including brain) after exposure. In our study, we observed effects after shortterm exposure, but long-term exposure might even cause stronger effects.
The concentrations at which neuronal networks respond to neuro pharmaceuticals in vitro correlate well with concentrations known to be active in vivo or measured in treated patients (Keefer et al. 2001b). This has also been shown for the anti fouling agent trimethyltin chloride (Gramowski et al. 2000); for anandamide, the endogenous ligand of the CNS cannabinoid receptors (Morefield et al. ; and for pyrethroid pesticides (Meyer et al. 2008). Therefore, networks of neurons on MEA neurochips can be considered to represent robust, fault-tolerant, spontaneously active dynamic systems with high sensitivity to their chemical environment (Gross et al. 1997) and have a predictive value that may be useful for regu latory testing (Bal-Price et al. 2008). BBB and entry into the CNS. The in vitro MEA neurochip model system lacks the BBB, and it does not take into consideration the bioavailability of compounds. However, we conducted the measurements in this study under serum media conditions. Therefore, the potential adsorption of NPs to plasma proteins needs to be considered (Wasdo et al. 2008), which means that the freely available particle concentration may be even lower. As mentioned above, the olfactory nerve pathway can be considered another critical portal of NPs entry to the CNS, especially under conditions of elevated environmental or occupational exposure (Fechter et al. 2002;Gianutsos et al. 1997;Henriksson and Tjälve 2000).
Transport and distribution in the CNS. The high endocytic activity of nerve cells at synapses makes the CNS prone to increased NP incorporation, distribution, and subsequent interference in neuronal functions. Axoplasmic transport provides the mechanism to bidirectionally translocate not only organelles and viruses but also injected NPs, which are widely used neuro anatomists to trace nerve fiber connectivity (Weiss and Gorio 1982). Oberdörster et al. (2009) summarized the possible translocation pathways of NPs and evaluated the hazardous potential of NPs. De Lorenzo (1970) demon strated that intranasally instilled silver-coated colloidal gold particles (50 nm) translocate antero gradely in the axons of the olfactory nerves to the olfactory bulbs. The NPs even crossed synapses in the olfactory glomeruli and reached mitral cell dendrites within 1 hr after intra nasal instillation. Their velocity was 2.5 mm/hr, and they were preferentially located in mitochondria, raising a major concern regarding their toxicity (Oberdörster et al. 2004(Oberdörster et al. , 2005. Yu et al. (2007) showed that inhaled 20-nm nano gold particles accumulate in the olfactory bulb of rats. After 15 days, significant accumulations of gold particles were detected in the septum and ento rhinal cortex, both brain structures receiving direct neuronal projections from the olfactory bulb and important in attention and new memory formation. Particles with diameters > 1 µm do not enter the olfactory nerve (Oberdörster et al. 2005).
Intracellular localization. In the present study, our REM data show NPs bound to the cell surface, and the TEM results indicate that NPs are taken up into neuronal network cells. These data are only qualitative and preliminary, so more detailed studies are needed to clarify how the NPs enter the cells, what the neuron-to-glia ratio is for NP entry, and whether only internalized NPs or also those attached to the neuronal or axonal membranes influence the action potential propagation and general electrical activity.
Significance of the zeta potential. The surface charge is one of the major physical properties of the NPs and can be measured only indirectly through the zeta potential, which is a function of the surface charge of the particle or any adsorbed layer at the interface and the nature and composition of the surrounding medium in which the particle is suspended. In our study using NPs suspended in serum-free media, we found a high positive zeta potential for TiO 2 NPs compared with the negative zeta potential of Fe 2 O 3 and CB NPs. The activity drop of the neuronal networks at low particle concentrations was most pronounced for CB, followed by Fe 2 O 3 and TiO 2 NPs.
Using engineered NPs with different surface charges, Lockman et al. (2004) and Kreuter (2004) studied the influence of the zeta potential and the applied NP concentration on BBB integrity and evaluated BBB integrity and NP brain permeability by in situ rat brain perfusion. Neutral NPs and low concentrations of anionic NPs had no effect on BBB integrity, whereas high concentrations of anionic NPs and low concentrations of cationic ones disrupted the BBB function, indicating the importance to consider particle charges.

Role of ROS.
For ROS production in the present study, the different NP species exhibited a potency sequence opposite that for disturbing electrical activity. TiO 2 was most effective, followed by Fe 2 O 3 and CB. Therefore, ROS production cannot explain the functional neuro toxicity (e.g., electrophysiological changes) that we observed. Results for ROS production in this study are in good agreement with previous investigations in which significant differences were reported in intra cellular radi cal formation in human lung cells exposed to TiO 2 and Fe 2 O 3 NPs (Bhattacharya et al. 2009). Although TiO 2 NPs induced a concentration-dependent increase in ROS formation, the effect was delayed in Fe 2 O 3 -exposed BEAS-2B cells and was significant only for the particle concentration of 5 µg/ cm 2 (Bhattacharya et al. 2009). Interestingly, in the present study we found that CB NPs did not increase ROS levels in neuronal networks after 24 hr exposure. Similar observations were reported by Yang et al. (2009), who compared different nanomaterials and reported a low potential for ROS formation in CB NPs.

Conclusion
Using co-cultures of primary neurons and glial cells, which form stable and electrically active neuronal networks on MEA neurochips, we found that electro physiological studies are suitable to measure changes caused by NP exposure. NPs are able to induce acute functional neuro toxicity at low particle concentrations, whereas formation of intra cellular free radicals may require higher particle concentrations and longer exposure times. To the best of our knowledge, this is the first study demonstrating acute functional neurotoxic effects of NPs. The results justify further studies with MEA neurochips to analyze the under lying mechanism of action and the longterm consequences of low-dose NP exposure.