Selective Targeting of Neurons with Inorganic Nanoparticles: Revealing the Crucial Role of Nanoparticle Surface Charge.

Nanoparticles (NPs) are increasingly used in biomedical applications, but the factors that influence their interactions with living cells need to be elucidated. Here, we reveal the role of NP surface charge in determining their neuronal interactions and electrical responses. We discovered that negatively charged NPs administered at low concentration (10 nM) interact with the neuronal membrane and at the synaptic cleft, whereas positively and neutrally charged NPs never localize on neurons. This effect is shape and material independent. The presence of negatively charged NPs on neuronal cell membranes influences the excitability of neurons by causing an increase in the amplitude and frequency of spontaneous postsynaptic currents at the single cell level and an increase of both the spiking activity and synchronous firing at neural network level. The negatively charged NPs exclusively bind to excitable neuronal cells, and never to nonexcitable glial cells. This specific interaction was also confirmed by manipulating the electrophysiological activity of neuronal cells. Indeed, the interaction of negatively charged NPs with neurons is either promoted or hindered by pharmacological suppression or enhancement of the neuronal activity with tetrodotoxin or bicuculline, respectively. We further support our main experimental conclusions by using numerical simulations. This study demonstrates that negatively charged NPs modulate the excitability of neurons, revealing the potential use of NPs for controlling neuron activity.

Inorganic nanoparticles (NPs) are original tools to treat brain diseases. Given their high surface to volume ratio, NPs enable efficient loading of therapeutic molecules, while their material-specific intrinsic properties offer alternative therapeutic opportunities.[1−3] For instance, for patients affected by glioblastoma multiforme, cisplatin-tethered gold NPs act as drug carriers and as radiosensitizers, emitting ionizing photoelectrons and Auger electrons in radiotherapy.[4] Similarly, the heat generated by iron oxide NPs under an oscillating magnetic field has been exploited (i) to perform hyperthermia treatment on human glioblastoma,[5] (ii) to temporally damage the brain barrier thus facilitating NP crossing,[6] and (iii) to promote NP uptake by opening heat-sensitive ion channels.[7,8] However, the use of NPs as drug carriers or therapeutic materials requires a deeper understanding of the principles governing their interaction and functional effects in neuronal circuits. Given the increasing number of NPs available and the increasing interest in studies pertaining to NP–cell interactions,[9−12] there has been a substantial effort to correlate the NP material features (e.g., composition, size, shape, coatings, etc.) to their neuronal toxicity and activity.[13−18] Since aberrant neuronal electrical activity is associated with most neurological diseases[19,20] and can be an early marker of neurotoxicity,[21] it is crucial to understand how NPs can modulate brain electrical function.

Controlled NP-induced bioelectric activity could be exploited to design nanotools that can regulate the imbalanced excitation/inhibition phenomena observed in many brain diseases. For instance, both single-[22] and multiwall[23] carbon nanotubes (CNTs) boost electrical activity of neurons cultured on CNT-coated substrates. Since functional changes were also observed with “non-targeted” CNTs, the authors proposed that the intrinsic electrical conductivity of CNTs might determine neuronal electrophysiological effects. More recent studies performed with NPs of various compositions appear to contradict the initial hypothesis that only the electrical properties of the materials determine functional effects on neuronal activity. For example, neurons cultured on multi-electrode arrays (MEAs) exposed to carbon black (CB), hematite (Fe2O3), and titanium dioxide (TiO2) NPs display an acute, concentration-dependent alteration of spontaneous electrical activity.[15] An increased electrical excitability also occurs in mouse hippocampal slices exposed to star-shaped Au NPs,[18] an observation confirmed at the single neuron level by patch clamp recordings.[24] Moreover, as shown in another study, quantum dots (QDs) with a negative coating were preferentially internalized by neurons with respect to astrocytes, oligodendrocytes, or microglia, suggesting that the surface charge is critical for neuron-specific uptake.[25] In this study, we have elucidated the mechanisms through which NPs modulate bioelectric activity from single neuron to neuronal networks.

We conducted our study on primary cultures of neurons since this experimental model facilitates access both to single-cell and large neuronal networks,[15,26,27] and it allows complementary measurements at the subcellular, cellular, and network scale. We have tested the effects of NPs that differ in shape (spherical and cylindrical), dimensions (inorganic core material ranging from 5 to 75 nm), composition (cadmium- or iron-based), functionalization, and charge. By combining confocal microscopy and transmission electronic microscopy with single neuron and network-wide electrophysiology techniques on primary hippocampal neuronal cultures, we found that soon after administration of the NPs, the neuron–NP interactions are exclusively determined by the surface charge of the NPs. We observed that negatively charged particles were selectively localized on the neuronal membrane and induced electrophysiological alterations at both single neuron and network levels. Further, results obtained through the pharmacological manipulation of the electrical neuronal activity suggest that the neuron–NP interaction is mediated by neuronal activity.

Results and Discussion

NP Interaction with Neurons Depends on NP Surface Charge and Is Shape-, Size-, and Material Independent

In a first series of experiments, rod-shaped semiconductor NPs made of a core/shell CdSe/CdS were added to the cell culture media of in vitro grown primary hippocampal neuronal cultures. These NPs, also known as quantum rods (QRs), with different length and diameter sizes (see Tables S1.1 and S1.2 and Figure S1.1 of the Supporting Information, SI) were initially chosen because of their bright fluorescent signal, enabling a straightforward investigation of interactions by confocal microscopy. The hydrophobic QRs, synthesized in nonpolar solvent with organic ligands, were transferred in water using an established amphiphilic polymer coating protocol; the binding of amino-modified polyethylene glycol (PEG) molecules provides increased stability to these NPs in physiological media.[28,29] In a typical experiment, upon QR addition, neurons became fluorescent within 10 min (QRs at 1 nM, see Figure A and the time-lapse movie Movie M2.1). This suggests a fast localization of these NPs to the neuronal cell membrane, through the somatic and the dendritic neuronal subregions. A similar fluorescent signal was observed with different QR samples of the same composition and of different longitudinal lengths or diameters (see for instance Figure S2.1). However, in some experiments, when testing QR batches with similar sizes and surface coatings, QRs did not produce the same fluorescent signal (Figure B). These divergent results prompted us to carry out a systematic study on a wide array of QRs samples, with the aim of disentangling their physicochemical properties and their ability to interact with neurons.

Figure 1

The effect of NP surface charge on the interaction with neurons: Confocal microscope images (left panel; excitation wavelengths of 488, 560, and 647 nm) of primary hippocampal neural cells incubated with 1 nM of negatively charged fluorescent QRs (sample no.7 in Table S1.2) after 10 min of incubation at RT. Neurites and dendrites are covered by negatively charged QRs (A, green signal). Yellow represents the combination of NPs (green signal) and a neuronal marker (VGAT, red signal) and highlights the healthy condition of the entire neural network and the colocalization between QRs and synapses (MCC = 0.45 ± 0.06). (B) The same neuronal culture incubated with QRs identical in shape and size (sample no. 9 in Table S1.2), but with a positive zeta potential; note the absence of QR fluorescence. These results are independent of QR size. Functionalization effect on the neuron–QRs interaction (right panel; excitation wavelengths of 405 and 488 nm); the chemistry used to tune NP surface charge was varied. In functionalization type I (C), to the same polymer-coated (PC) QRs, amino-PEG derivatives bearing carboxyl (COOH) or methoxy (OCH3) or amino (NH2) as the other ending moieties were bound to the QR surface. In functionalization type II (D), to PC QRs, a fixed amount of amino-PEG-OCH3 (via EDC chemistry) was attached, and different amounts of tertiary amine (N,N-dimethylethylenediamine) were covalently linked to the polymer shell. Negatively charged QRs always interacted with neurons independently of the functionalization protocol (E and G, sample nos. 3 and 4, respectively, in Table S1.2), while positive QRs did not interact (F and H, sample nos. 10 and 9, respectively, in Table S1.2). The blue signal in panels e–h indicates the cell nuclear staining with DAPI. Scale bars: 100 μm.

We found that the NP surface charge, which is measured as zeta potential value, was crucial in determining their interaction with neurons. QR samples with zeta potential more negative than −20 mV adhered on the neuronal membrane and gave rise to the fluorescent signal, whereas QRs with a zeta potential value higher than −20 mV did not show this effect and resulted in no detectable neuronal fluorescence signal. To confirm this dependency on zeta potential, we designed an experiment that allowed tuning of the final zeta potential (Figure C, functionalization type I). An aliquot of the same pristine hydrophobic QRs was first transferred in water by the polymer wrapping procedure and later, the QR charge was tuned from positive to negative by attaching different amino-PEG derivatives (amino-PEG-metoxy, amino-PEG-carboxyl, or diamino-PEG molecules) to the polymer-coated QRs. Contrary to the negatively charged QRs (−35 mV), which maintained their interaction with neurons, neither the positively charged (zeta potential of +25 mV) nor the neutral QRs interacted with neurons, as confirmed by confocal microscopy (compare Figure E and 1F). Almost neutral or slightly positive QRs also failed to interact with neurons; QR fluorescence was either visible on the substrate as tiny aggregates or not detectable (Figure S2.2, even in absence of neurons, positive QDs interact with the substrate resulting in tiny fluorescent spots; data not shown).

Interestingly, while the charge was tuned by changing the type of functionalization, we found that the final interaction of QRs with neurons primarily depends on the QRs’ zeta potential. We used, for instance, an alternative strategy to adjust the charge of the QRs from −50 mV to +35 mV by fixing the amount of amino-PEG-OCH3 molecules and varying the amount of amino-alkyl-tertiary amine molecules bound to the QRs (Figure D, functionalization type II). Even in this case, the negative QRs still interacted with neurons, while the positively charged QRs did not (compare Figure G,H).

In addition to rod-shaped NPs, we also investigated spherical NPs (QDs) of similar composition (CdSe/ZnS) and with similar fluorescent properties to QRs, thus enabling confocal imaging. When comparing the interaction with neurons of negative, neutral, and positive QDs, the results are consistent with those obtained with QR (see Figure S2.3 and an example of a time-lapse recorded with neutral QDs, Movie M2.2). This supports the hypothesis that only negatively charged NPs interact with neurons, independent of their size and shape. These findings are in agreement with a previous study that reported the selective uptake of negatively charged QDs in brain slices after an overnight incubation.[30]

To directly image the QRs on neurons at higher magnification, we performed transmission electron microscopy (TEM) and scanning transmission (STEM). Negatively charged QRs localized close to the plasma membrane of cell bodies (Figure A). They were also frequently located at synapses, inside the synaptic clefts (Figure B,C), and also along neurites (Figure S3.1A,B). In contrast, positive QRs were never found on neurites, neuronal somata, or synapses (Figure S3.2A–C). Negative QRs were only occasionally observed in endocytic pits or enclosed in endocytic vesicles in the cell cytoplasm (Figure A). In fact at a short incubation time (10 min), endocytosis is not expected to yet occur. EM tomography enabled us to reconstruct the 3D distribution of QRs on neurons. QRs were mostly detected either contacting or in the close vicinity of the neuronal membrane. EM tomography unambiguously confirmed the presence of QRs at synapses, inside the synaptic cleft (Figure C and Movie M3.1).

Figure 2

EM imaging of mouse primary hippocampal neurons treated with negatively charged CdS-CdSe QRs (A–C), Fe2O3 nanorods (D), and Fe2O3 dots (E, F). (A) TEM image of a cluster of −21 mV, 75 nm long QRs (sample no. 6 in Table S1.2), close to a neuron cell body plasma membrane (arrowheads). (B) TEM image of −35 mV, 32 nm long QRs (sample no. 2 in Table S1.2), inside the synaptic cleft and among neurites (arrowheads). (C) 3D model of −35 mV, 32 nm long QRs inside a synaptic cleft, representing the reconstruction of a 250 nm-thick tomogram acquired in HAADF STEM). The 3D model is set on a single tomographic section. (Right, C1) Single tomographic slices corresponding to a median section in the 3D reconstruction. Arrowheads point to QRs. (D) 3D model of nanorods with zeta potential of −21 mV, 65 nm long Fe2O3 (sample no. 11 in Table S1.2), close to a group of neurites, representing the reconstruction of a 100 nm-thick tomogram acquired in HAADF STEM (see Movie M3.2). The 3D model is set on a single tomographic section; n1–n5 are the five neurites present in the reconstruction. (Left, D1) Single tomographic slices corresponding to sections D1 and D2 in the 3D reconstruction. Arrowheads point to nanorods. (E) Projection image through dots close to neurite membranes (arrowheads). (F) STEM projection image showing the presence of Fe2O3 dots (sample no. 14 in Table S1.2), close to neurite membranes and inside the synaptic cleft (arrowheads). Note a cluster of dots entering a phagocytic pit (double arrowheads). Asterisks indicate presynaptic vesicles. Scale bars are 0.2 μm, except in A and B where they are, respectively, 0.5 and 0.1 μm. Abbreviations: cyt, cytoplasm; n, neurites; prs, presynaptic terminal; ps, postsynaptic terminal. Color code for 3D reconstructions: neurite, green; synaptic vesicles, cyan; microtubules, orange; electron-dense bodies, red; cytoplasmic vesicles, blue; nanorods, yellow.

All of these experiments were performed with cadmium-based NP material. Therefore, we could not exclude that the material composition might also affect the NP–neuron interaction. To eliminate this possibility and demonstrate the same phenomenon with a second material, we tested magnetic NPs made of iron oxide NPs, γ-Fe2O3, with either a spherical (12 nm in diameter) or rod (65 nm × 6 nm) shape. Since the magnetic NPs do not fluoresce, the imaging of NP interaction with neuronal cultures was performed exclusively by TEM and STEM. Similar to the QRs, both spherical and rod-shaped NPs with negative zeta potential interacted with neurons, independent of their length and shape (Figures D,F and S3.1C,D). Electron tomography further confirmed that negative Fe2O3 rods were randomly oriented on the neuron surface, lying parallel, obliquely, or perpendicular with respect to the cell membrane (Figure D and Movie M3.2). They were arranged in a more parallel fashion mainly in the synaptic cleft, likely due to steric hindrance effects (Figures B,D and S3.1C,D). Also in this case, positively charged iron oxide NPs were never found on neurites or neuron cell bodies (Figure S3.2D–F).

Interaction of Negatively Charged NPs with Neurons Is Driven by Neuronal Spiking Activity

Next, we studied the localization of NPs in cultures expressing both neuronal and glial cell types, the latter being nonexcitable cells that modulate neuronal activity.[31] Ten minutes after the application of the negative fluorescent QRs (L = 30 nm, d = 5 nm, Zp −25 mV), the fluorescence of negative QRs was selectively detected on neuronal cells (Figure ). This occurred even though the glial cells were closely associated with neurons (identified in red by using glial fibrillary acidic protein, GFAP).

Figure 3

Negative NPs interact exclusively with excitable neurons. Neurons labeled with NeuN (blue) and glial cells labeled with GFAP antibody (red) were treated with 1 nM negatively charged QRs (green, −35 mV, sample no. 2 in Table S1.2). Note the QRs attach exclusively to the membrane of neural cells. Scale bar: 50 μm; excitation wavelengths: 488, 560, and 647 nm.

An explanation for such selectivity of negative NPs for neuronal cell types could be related to the electrical activity of neuronal cells, which might specifically attract NPs with a “charge-driven effect”. To test this hypothesis, we pharmacologically manipulated electrophysiological activity in neuronal networks, by increasing or reducing spiking activity using Bicuculline (BICU) or Tetrodotoxin (TTX), respectively, and characterized the NP–neural interaction with confocal microscopy. These experiments confirmed that the level of neuronal spiking activity modulates the attraction of negatively charged QRs to the neural cell membranes (Figure S2.5). An increase in network activity with BICU resulted in enhanced neuronal fluorescence, while suppression of the neuron network activity by TTX impaired the association of negative QRs with neurons. In contrast, positive QRs were never localized on neurons in the presence of either BICU or TTX. Therefore, negative NPs interact with neuronal membranes, and this interaction is mediated by neuronal spiking activity.

Negatively Charged NPs Affect Neuronal Bioelectric Activity in Networks

We next investigated whether these NPs might also interfere with electrical activity at the scale of single neurons or neuronal networks. First, we characterized the effects of NPs on network-wide electrical activity by exploiting high-density multi-electrode arrays (HD-MEAs) and by computing mean activity parameters. These devices (Figure A) provide 4096 simultaneously extracellular recording electrodes (i.e., 42 μm pitch, array arranged in a 64 × 64 layout) and allow computing with high statistical significance mean activity parameters to characterize network activity.[27] After culturing dissociated hippocampal cultures on HD-MEA chips for 21 days in vitro (DIV), the typical experimental protocol consisted of recording the spontaneous activity for 10 min, followed by administration of 1 nM NPs, and then recording the activity after treatment for additional 10 min. Representative temporal traces of the detected spiking activity on single electrodes as a function of time, before and after addition of NPs, are shown in Figure B. Network activity changed upon administration of negative NPs (Figure B), while there was no clear variation for positive NPs. The effects on the network activity of the different NPs were quantified by computing four mean activity parameters: (i) the number of active channels (channels with >0.1 spike/s), (ii) the mean firing rate (MFR, the firing rate averaged on all the active channels), (iii) the mean burst rate (MBR, the average number of bursts per minute), and (iv) the interspike burst interval (ISBI, the average firing rate of spikes during the burst events). While the MBR and ISBI characterize the spiking synchronicity of the network, the MFR and the number of active channels define the global spiking activity of the network. Since in these measures, functional effects were also exclusively dependent on the surface charge of the NPs, the statistical analysis was performed by grouping all NPs belonging to the same zeta potential (Zp, X-axis) interval (i.e., Zp < −40 mV, −40 mV < Zp < −20 mV, −20 mV < Zp < 0 mV, Zp > 0 mV) independent of their shape, material, or type of functionalization (Figure c). Each data point represents the normalized increase of the plotted parameters after NP administration with respect to basal conditions averaged from at least three independent measurements for each NP type. For all the parameters, there is a significant increase with respect to the basal network activity when Zp < −20 mV, thus demonstrating a clear relationship between charge of the NPs and modulation of the network activity. This indicates that negatively charged NPs not only increase the global spiking activity of the network but also the spiking synchronicity of the network, up to a plateau of about −20 mV. These effects disappear when Zp is close to 0 mV and when positive NPs are administered.

Figure 4

Characterization of electrophysiological effects of NPs at single-neuron and network scales. (A) CMOS-MEA device used for large-scale neuronal network activity recordings from 4096 electrodes simultaneously. Scale bar = 20 μm. (B) Representative raw traces of four electrode channels before and 10 min after administration of negatively charged NPs (Zp < −40 mV). (C) Quantification of the normalized changes in MBR, ISBI, MFR, and number of active channels, as a function of Zp of the NPs administered. All NPs employed in these experiments are indicated in Table S1.2. All four parameters are affected by NPs with Zp < −20 mV. (D) DIC image of cultured hippocampal neurons showing a patched neuron during an electrophysiological recording. Scale bar 10 μm. (E) Top: Representative traces of IPSC recorded before and 10 min after the application of NPs with Zp < −40 mV. Bottom: Averaged IPSC traces corresponding to the recording shown above, before, and after the application of NPs with Zp < −40 mV. (F) Quantification of the normalized changes in IPSC frequency, amplitude, and charge transfer upon the application of NPs with different Zp. (Zp < −40 mV: n = 13; Zp −40 mV ÷ −20 mV: n = 18; Zp −20 mV ÷ 0 mV: n = 8; Zp > 0 mV, n = 25; one way ANOVA followed by Tukey’s post test; * p < 0.05; ** p < 0.01; *** p < 0.001).

Second, we investigated the effects of NPs on single-neuron activity with the patch clamp technique in the voltage clamp mode. In this experimental configuration, it was possible to record postsynaptic currents (i.e., electric signals below the action potential threshold) from an individual neuron within the neuronal network. This electrophysiological approach is complementary to HD-MEA recordings and provides high-quality intracellular signals from single neurons. Interestingly, these patch clamp experiments revealed that negatively charged NPs induced a significant increase in the amplitude and frequency of spontaneous postsynaptic currents, thus resulting in an enhanced charge transfer to the postsynaptic element (Figure D–F). In contrast, neutral or positive NPs did not significantly alter synaptic parameters (Figure F). These results corroborate the HD-MEA data, indicating that negative NPs are able to trigger an overall increase in neuronal and synaptic activity.

Negatively Charged NPs Depolarize Neuronal Membrane: A Numerical Simulation

In order to estimate the effect of negative NPs on the membrane potential, numerical simulations were performed with the finite element method (Comsol Multiphysics). Figure A,B shows the results for CdSe/CdS QR and CdSe/ZnS QD, respectively. To build the model, zeta potential and the hydrodynamic radius (RHy) of the NPs were used as input parameters to determine the charge distribution around each NP. Furthermore, the static permittivity for both CdSe/CdS and for CdSe/ZnS was considered.[32,33] We first considered the simplest case of spherical QDs. Starting with the experimental measurements of the zeta potential of −22 mV for the QD (−25 mV for QR) and knowing the hydrodynamic radius measured for QD samples (23 nm), we assigned the same hydrodynamic radius (RHy of 23 nm) to the rod-shaped QR, which is less obvious to extract from the DLS data. We assumed indeed that the polymer layer (polymer coating + PEG molecules) occupies the same space around each NP. For the QR, we considered a longitudinal arrangement of the rod (with length of 30 nm and diameter of 6 nm) on the surface, since from the TEM analysis, most of the rods were parallel to the membrane. With these parameters and assuming a membrane resting potential of −70 mV and a membrane thickness of 5 nm, we computed a depolarization effect around 5 mV for both QDs and QRs when the NPs are in close contact to the membrane. This simulation supports the depolarization effect measured with MEA and patch clamp measurements.

Figure 5

Numerical finite element method calculations resolving the effect of a CdSe/ZnS QD (A) and CdSe/CdS QR (B) on the membrane potential. For both types of particles, the hydrodynamic radius RHy is equal to 23 nm, as suggested by our experimental estimation. For both the QD (A2, A3) and QR (B2, B3), two situations are shown: zeta potential (Zp) equal to a negative or a neutral value (0 V), respectively. In (A2), for QD with Zp = −22 mV, there is a depolarization effect corresponding to ΔV = 65 mV. The same result occurs in case (B2) for Qr with Zp = −25 mV. For both structures (A3, B3), upon Zp = 0 V, there is no variation of the membrane potential. The 2D electric potential distribution is also shown for both structures. It is interesting to note that the effect of the charges is noticeably stronger along the axis passing through the center of QR/QD. Finally, the colored mapping images (under A2, A3, B2, and B3) represent the charge distribution along the zeta axis on the membrane once in contact with negatively charged QD or QR.

Indeed, in the plots of Figure A (similarly for B), three curves are shown: (i) the blue curve associated with an isolated membrane with static resting potential of −70 mV; (ii) a green curve describing the potential along the axis of an isolated QR (or QD); and (iii) a red curve resulting from the presence of a QR (or QD) on top of the cell membrane. For all three cases, the electric potential is calculated along the Z-axis of the NPs, moving from the point of contact of the NP at the membrane to Z positions that are further from the membrane (corresponding to the white dotted lines in the 2D electric potential distributions below).

For both QR and QD, we observed a clear shift of the membrane potential from −70 mV to −65 mV. This suggests an increased neuronal action potential firing once the membrane is in contact with negatively charged NPs. Such a potential reduction is not uniform over the contact area between the QR/QD and the membrane, as shown by the 2D distributions of the electric potential. In particular, for the zeta potentials of −22 mV (QD) and −25 mV (QR), the 2D plots describe the depolarization effect with the brightest colors associated with the strongest occurring depolarization. Conversely, when the zeta potential is equal to zero, there is no reduction of the membrane potential. Finally, the two images showing the charge distributions around the membrane upon contact with negatively charged particles explain why, experimentally, our negative QD and QR could easily adhere to the cell membrane: A net attractive Coulomb interaction has pushed the negative NPs to move toward the membrane. A positively charged NP would not promote the attraction toward an outer positive membrane.

Discussion

The present study identifies the key determinants of NP interactions with neuronal functional activity at both synaptic and network levels. The NP surface charge is crucial for the modulation of neuronal electrical activity, while NP size, shape, and material play a negligible role in the association and functional interaction with neurons. These results are in agreement with the literature on single material studies. For instance, similar NP–neuron interactions were observed for negative carbon black, hematite iron oxide, titanium dioxide,[15] and gold NPs.[18,24] However, the interactions in those studies, unlike those reported in this work, were investigated over prolonged time exposure (from few hours to days) and at different concentration ranges within them. In this work, we have conducted a systematic study on different NP materials, of different size, shape composition, and coating, and we have investigated the immediate effects of NPs just after their administration, within the first few minutes and at a rather low-concentration range. Our experimental data demonstrate four main results: (i) only negative NPs (zeta potential < −20 mV) adhere to neurons (bodies, neurites and synaptic clefts); (ii) negative NPs selectively interact with neuronal excitable membranes; (iii) negative NPs on neurons cause a depolarization of neurons and measurable changes in neuron and neuronal network electrical activity; (iv) the suppression of neuronal activity by TTX hinders the interaction of negative NPs with neurons.

Our theoretical model explains how a Coulombic potential on the outer neuron surface retains only negative NPs. Importantly, these simulations predict membrane depolarization that supports the experimental evidence of increased neuronal activity, as measured at the single neuron and at the whole network levels. However, this simulation model does not account for the selective interactions of NPs with neurons and not with nonexcitable glial cells. Based on our data, three key factors appear to determine the selective localization of negatively charged NPs on neurons and the electrophysiological activity change. First, neuronal spiking activity attracts NPs, as demonstrated by experiments with TTX and by the lack of localization on glial cells. Second, Coulombic attraction retains only negatively charged NPs at the cell membrane, as observed in our simulations. Third, the presence of negatively charged NPs on neurons induces a depolarization, as measured in our electrophysiological experiments and also predicted by our model, which in turn facilitates the crossing of the threshold for firing an action potential.

To explain the attraction of NPs, which appears to be induced by neuronal spiking activity, we speculate that the strong depolarization during the firing of an action potential (up to +40 mV) could transiently remove the counterions surrounding the negatively charged NPs, thus allowing a tight and stable electrostatic interaction between the negatively charged NP surface and the positively charged outer side of the neuronal membrane. This mechanism could explain the selective, activity-dependent interaction of NPs with neurons, as we observed with imaging, EM, and electrophysiology. Besides the Coulombic interactions at the cellular membrane, the presence of NPs at the synaptic cleft could also be due to the matching of NP size with the cleft space. NPs at synapses are indeed expected to be strategically located to operate a further “fine modulation” of neuronal excitability.

The selective NP–neuron interaction shown here paves the way to innovative applications of NPs in neuroscience. The key role of NP surface charge for neuronal interactions is a fundamental finding for the future design of NPs for neuron targeting. The specific interaction with neurons, but not with glial cells, could represent an excellent opportunity to discriminate between these two cell types for selective drug delivery. In addition, the higher sensitivity to neuronal activity suggests that negatively charged NPs could be used as markers of active neurons for long-term imaging, easily measured by NP fluorescence. This approach could be exploited, for instance, to visualize aberrant increased neuronal activity in neurological disorders. In the same conceptual framework, negatively charged NPs could also reveal alterations of excitability in different subregions of individual neurons (e.g., soma vs distal dendrites), an important factor for the emergence of specific brain states, including network oscillations and selection of cell assemblies.[34] Along the same lines, the increased neuronal activity could be exploited to increase the activity of inhibitory neurons with reduced excitability that are responsible, for example, for severe forms of epilepsy.[35] More generally, given an efficient neuronal subtype-specific targeting, in principle, NPs could be exploited to finely modulate the balance between excitation and inhibition in the brain, a major determinant for most neurological diseases.

HD-MEA experiments have revealed that the increase in the overall spiking activity by NPs is accompanied by a change in the network synchronicity. This could be induced by the uniform increase of the neurons’ activity, reaching a high level of network stability. Alternatively, NPs could differentially affect neurons in the network by, for example, preferentially activating neurons with high connectivity (hub neurons) that may be more likely to interact with NPs due to longer dendritic/axonal length and more complex arborization. Such activity-dependent behavior of negative NPs might allow the design of nanomedicine tools for the control of neuronal activity or for the selective delivery of drugs to firing neurons.

Conclusions

NPs with negatively charged surfaces rapidly localize on neuronal membranes and induce electrophysiological alterations. In contrast, the same type of NPs with positive or neutral zeta-potential shows no or low nonspecific interaction with the cellular membrane and no effects on bioelectric activity. These effects are material independent because similar results occurred with magnetic NPs as well as Cd-based semiconductors. Importantly, the interaction of negative NPs is not only selective for excitable neuronal cells but also dependent on neuronal activity. These observations suggest that electric activity most likely plays a role in the specificity of the NP–neuron interaction. Finally, numerical simulation supports the proposed mechanism for the selectivity of negative NPs and the modulation of neural electric activity.

Materials and Methods

Materials and Cell Method Protocols

Materials for the NP Synthesis

Poly(maleic anhydride-alt-1-octadecene) MW 20,000–25,000, poly(isobutylene-alt-maleic anhydride), poly(styrene-co-maleic anhydride) cumene terminated, dodecylamine, oleylamine, bis(6-aminohexyl)amine, N-(3-(dimethylamino)propyl)-N′-ethylcarbodiimide hydrochloride (EDC), methoxypolyethylene glycol amine MW 750 (NH2-PEG-OCH3), O-(2-aminoethyl)-O′-(2-carboxyethyl)polyethylene glycol hydrochloride MW 3000 (NH2-PEG-COOH), poly(ethylene glycol) bis(amine) MW 2000 (NH2-PEG-NH2), N,N′-dimethylethylendiamine (DMEDA), poly-d-lysine hydrobromide (PDL), tetrodotoxin (TTX), bicuculline (BICU), diethylzinc, 1.0 M in hexane, hexamethyldisilathiane, toluene, and methanol anhydrous were purchased from Sigma-Aldrich. Trioctylphosphine oxide 99% (TOPO), cadmium oxide 99.999% (CdO), tri-n-octylphosphine 97% (TOP), sulfur 99%, selenium powder 99.99%, Fe(CO)5, and trioctylphosphine were purchased from Strem Chemicals. n-Octadecylphosphonic acid (ODPA) and hexylphosphonic acid (HPA) were purchased from DPI synthesis.

NP Synthesis

All QDs, QRs, and magnetic NPs used were prepared by the seeded growth approach as already reported.[36−39] The surfactant-coated colloidal NPs were transferred from organic solvent into water by implementing a previously developed polymer-coating procedure.[29,40] To tune the surface charge of the NPs, further surface coating with amino derivate PEG molecules (henceforward referred as functionalization type I) was achieved by means of EDC cross-linking chemistry.[28,41] Three different PEG molecules derivates were used: monoamino PEG, diamino PEG, and bifunctional carboxy-amino PEG. As alternative functionalization, the tertiary amine N,N-dimethylethylenediamine was covalently attached to the polymer shell (hence referred as functionalization type II). In both types of functionalization, proper variation of the molar ratios of PEG/EDC or N,N-dimethylethylenediamine/EDC, respectively, allowed to adjust the average surface charge of the NPs from negative to positive values (see Table S1.1 for functionalization conditions chosen). In all cases, the surface charge of the NPs was varied in the range from −50 mV to +35 mV as measured after the functionalization in aqueous media. In Table S1.2, a summary of the type of functionalizatrion, zeta potential (Zp), NP size parameters (length (L), rod diameter (d), dot radius (R), and emission wavelength for fluorescent NPs are reported. Table S1.2 shows also which NPs were employed for confocal microscopy, TEM, and electrophysiological investigations. In Figure S1.1, a collection of TEM images of NPs used in this study is shown Figure S1.1.

Neuronal Cultures

Culturing media and additional compounds were acquired from Gibco-Invitrogen. Primary hippocampal neurons were prepared from E18 rat as previously described.[42] Briefly, hippocampi were manually isolated in Hanks’ balanced salt solutions (HBSS) and incubated for 15 min with 0.125% trypsin at 37 °C. After trypsinization, tissues were mechanically dissociated through a fire-polished Pasteur pipet.

The cells were then plated onto coverslips (diameters of 18 and 25 mm) previously coated with 0.1 mg/mL PDL to promote cell adhesion and sterilized by UV light exposure for 30 min at a density of 120 and 160 cells/mm2 in 12 and 6 multiwell plates (Corning, Lowell, MA, USA).

Neurons were let to adhere 2 h in an incubator, and then 2 mL of NB complete medium were added in each well and incubated at 37 °C and 5% CO2. The culture medium was replaced every 7 days. Neurons were cultured until 29 DIV.

NP–Neuron Interactions

Most of the experiments were performed at 24−25 DIV, unless differently specified. NPs were administered in the NB complete medium to reach a final concentration of 1 nM and let to incubate for 10 min. For TEM experiment, a higher NPs concentration (10 nM) was used.

NPs interaction with neural networks: Confocal Microscopy Characterization

Immunofluorescence and Confocal Microscopy

Cells were washed once in 1× phosphate buffered saline (PBS) solution, fixed with 4% (w/v) paraformaldehyde in PBS at room temperature (RT) for 30 min and washed three times in 1× PBS. Fixed samples were permeabilized with 0.1% (v/v) Triton X-100 in 1× PBS for 15 min. Blocking solution (PBS 1×, 1% BSA, 5% FBS) was added for 45 min at RT to block nonspecific reactions. The following primary antibodies diluted in blocking solution were used: mouse anti-VGAT (Synaptic System, Goettingen, Germany, cat. no. 13101) or mouse anti-PSD95 Thermo Fisher Scientific, mouse monoclonal MA1–045) for synapsis labeling, rabbit anti-GFAP (Millipore, Billerica, MA, USA) as glial cell marker, and mouse anti-NeuN (Millipore, Billerica, MA, USA) for neuron nuclei. After incubation for 3 h at RT, samples were washed and incubated with secondary Alexa647-labeled goat antimouse or antirabbit (Invitrogen) for 45 min at RT. In the case of nuclei staining, secondary Alexa488 goat antimouse (Invitrogen) was used. Samples were mounted using DAKO antifade reagent, in some cases containing DAPI (Invitrogen), and stored at 4 °C. Samples were imaged using a Nikon A1 confocal microscope. Imaging was performed using 405 nm excitation wavelength and emission bandpass filter 458/10 nm for DAPI; 488 nm excitation wavelength for cell autofluorescence, with emission bandpass filter 525/40 nm; 488 nm wavelength to excite NPs, with emission bandpass filter 595/50 nm; 638 nm wavelength to excite Alexa-647-labeled VGAT and PSD95, with emission bandpass filter 700/40 nm. Manders’ correlation coefficient (MCC) was used to quantify merging of two fluorescence channels. Quantification of NP adsorption was computed as the mean pixel intensity of NP fluorescence normalized over the number of cells detected in each image (to take into account the variability of cell density).

The acquisition of the videos was performed in liquid, by injecting NPs at a final concentration 1 nM directly in the well containing the coverslips in PBS buffer and imaging at an acquisition rate of 1 frame every 30 s.

Confocal Imaging of Neural Networks Activity Alteration

Neural network activity was chemically altered by using TTX and BICU. Neurons at 21 DIV were incubated for 15 min with TTX and BICU dissolved in NB serum complete solution at concentration 1 μM and 30 μM respectively. After incubation time, the solution was removed and replaced with NPs dispersed at 1 nM concentration in NB serum complete solution containing either 1 μM TTX or 30 μM BICU at concentration. Cells were then washed in PBS, fixed in PFA, mounted in coverslip using Prolong Antifade reagent containing DAPI, and analyzed by confocal microscopy as described in paragraph Immunofluorescence and Confocal microscopy.

Electron Microscopy (EM) Analysis

EM Methods

TEM and high-angular annular dark-field (HAADF) STEM were performed on mouse primary hippocampal neurons doped for 10 min with Fe2O3 nanorods and dots differing exclusively in their overall surface charge in a range from −40 mV to +25 mV). Similar experiments have been conducted on a plethora of nanorods differing in composition, overall surface charge and dimensions (see Tables S1.1 and S1.2). The neurons were plated on glass coverslips and fixed with 0.66 M cacodylate-buffered 1.25% glutaraldehyde solution for 1 h at room temperature. After extensive rinsing in the same buffer, the samples were postfixed in a solution of 1.5% potassium ferrocyanide and 1% osmium tetroxide in 0.1 M cacodylate buffer, stained overnight in the dark in a 0.5% aqueous solution of uranyl acetate, dehydrated in a graded ethanol series, and embedded in EPON resin. Sections of 70 nm were cut with a diamond knife (Diatome) on a Leica EM UC6 ultramicrotome and stained with lead citrate and uranyl acetate. EM tomography was performed in bright-field TEM and HAADF STEM on thick sections (ranging in thickness from 100 to 250 nm). Computation weighted back-projection (WBP) of tomograms was done with the IMOD software package.[43,44] Segmentation and three-dimensional visualization were performed using the Amira software package (FEI Visualization Science Group, Bordeaux, France). EM projection images and HAADF STEM tilted series were collected using a FEI Tecnai G2 F20 equipped with a Schottky field-emission gun and recorded with a 2k × 2k Gatan BM UltraScan charge-coupled device (CCD) camera.

Electrophysiological and Neural Network Activity Recording

Electrophysiological Recordings

Inhibitory spontaneous postsynaptic currents (sIPSCs) were recorded in the whole-cell configuration of the patch-clamp technique. External recording solution contained (in mM): 145 NaCl, 2 KCl, 2 CaCl2, 2 MgCl2, 10 glucose, and 10 HEPES, pH 7.4. Patch pipettes, pulled from borosilicate glass capillaries (Hilgenberg, Malsfeld, Germany), had a 4–5 MΩ resistance when filled with intracellular recording solution containig (in mM): 150 KCl, 1 CaCl2, 2 MgCl2, 1 EGTA, 10 HEPES, and 2 Na2ATP (300 mOsm and pH 7.2 with KOH). Currents were acquired using Clampex 10.0 software (Molecular Devices, Sunnyvale, CA). sIPSCs were recorded at room temperature from a holding potential of −60 mV in the presence of CNQX (10 mM) to isolate GABAergic events. Currents were sampled at 20 kHz and digitally filtered at 3 kHz using the 700B Axopatch amplifier (Molecular Devices). sIPSCs were detected by using the “scaled sliding template” detection algorithm implemented in pClamp10, first described by Clements and Bekkers[45] and by setting the “detection criterion value” to 5. QRs and QDs were bath applied in the recording chamber following the recording of a baseline trace of at least 5 min. The effects of QRs and QDs were examined 10 min after application. The stability of the patch was checked by repetitively monitoring the input and series resistance during the experiments. Cells exhibiting 10–15% changes in the duration of the whole experiment were excluded from the analysis.

Neural Network Activity Recording with High-Density Multi-Electrode Array (HD-MEA)

Extracellular electrophysiological in vitro experiments were carried out by means of a recently developed HD-MEA. These devices, thoroughly reviewed in ref (46), provide a matrix of 4096 electrodes (sensing area 21 × 21 μm2, pitch 42 × 42 μm2) arranged in a 64 × 64 layout and covering an area of 2.7 by 2.7 mm2. After seeding and growing dissociated neuronal cultures on top of the active area, extracellular recordings of the spontaneous network activity were performed simultaneously from all the 4096 electrodes at 7 kHz sampling rate.

Before seeding, HD-MEA chips were sterilized 20 min in 70% ethanol, rinsed abundantly in double deionized sterile water (DDW), filled with neurobasal complete medium (NB complete medium) and kept overnight in incubator to increase hydrophilicity. Then the active area of the devices were coated with promoting adhesion factors as poly-d-lysine (PDL) 0.1 mg/mL or Polyethylenimine (PEI) 1:500 for at least 12 h. Successively a drop of 20–30 μL containing dissociated hippocampal neurons (see Neuronal Cultures) at a nominal concentration of 1000 cell/μL was seeded on the recording area of the chips, and after 2 h about 1.6 mL of NB complete medium was added to fill the reservoir. Chips were maintained in an incubator at 37 °C, 5% CO2, and about 50% of the volume of the culture medium was replaced every 7 days.

Experiments were performed after 21 DIV once cultures already provide strong spontaneous activity expressing both spikes as well as network bursts (i.e., synchronous tonic spiking events spreading on all the network). The experimental protocol consisted of 10 min recording of the basal activity at room temperature 20 min after removing the chip from the incubator. Then NPs were administered to the NB complete medium to a final concentration of 1 nM (i.e., 5 μL in a total volume of 0.5 mL). A few seconds after administration, the activity of the neural network was recorded for 10–15 min. Control experiments administering to the neural networks using the same volume of PBS buffer, without NPs, were also carried out.

Four main parameters were analyzed to evaluate the effect of NPs on network activity: (i) the number of active channels (>0.1 spike/sec), (ii) the MFR, the firing rate averaged on all the active channels, (iii) the MBR, the average number of burst per minute, and (iv) the ISBI, the average firing rate of spikes during the burst events. To obtain them, spike events were detected by using the precise time spike detection algorithm already presented in ref (47), while the identification of the network bursts was performed following the algorithm described in ref (48).

Numerical Calculations

The FEM model was setup by using COMSOL Multiphysics v5.2. The simulation included two electrostatic calculations. The first calculation served as reference and solved the Poisson equation by fixing a 0 V potential on the external surface of the membrane and −70 mV on the internal one. No charge was placed in the QD/QR shell at this stage. The second calculation solved the Poisson equation by placing a charge density (in order to provide a surface potential matching the measured zeta potential) within the QD/QR shell. No membrane potential was set at this stage. The total potential was then plot by adding potential values from the two calculations at each point of the domain.

For both electrostatic calculations, the designed elements were surrounded by a 200 nm × 200 nm × 200 nm box terminated with 10 nm of infinite elements layer to truncate the calculation and to mimic an effectively more extended simulation domain. The outer surfaces of the box were set as ground for both calculations.