Silica-coated magnetic nanoparticles activate microglia and induce neurotoxic d-serine secretion

Nanoparticles have been studied for brain imaging, diagnosis, and drug delivery owing to their versatile properties due to their small sizes. However, there are growing concerns that nanoparticles may exert toxic effects in the brain. In this study, we assessed direct nanotoxicity on microglia, the resident macrophages of the central nervous system, and indirect toxicity on neuronal cells exerted by silica-coated magnetic nanoparticles containing rhodamine B isothiocyanate dye [MNPs@SiO2(RITC)]. We investigated MNPs@SiO2(RITC)-induced biological changes in BV2 murine microglial cells via RNA-sequencing-based transcriptome analysis and gas chromatography-mass spectrometry-based intracellular and extracellular amino acid profiling. Morphological changes were analyzed by transmission electron microscopy. Indirect effects of MNPs@SiO2(RITC) on neuronal cells were assessed by Transwell-based coculture with MNPs@SiO2(RITC)-treated microglia. MNPs@SiO2(RITC)-induced biological changes in the mouse brain in vivo were examined by immunohistochemical analysis. BV2 murine microglial cells were morphologically activated and the expression of Iba1, an activation marker protein, was increased after MNPs@SiO2(RITC) treatment. Transmission electron microscopy analysis revealed lysosomal accumulation of MNPs@SiO2(RITC) and the formation of vesicle-like structures in MNPs@SiO2(RITC)-treated BV2 cells. The expression of several genes related to metabolism and inflammation were altered in 100 µg/ml MNPs@SiO2(RITC)-treated microglia when compared with that in non-treated (control) and 10 µg/ml MNPs@SiO2(RITC)-treated microglia. Combined transcriptome and amino acid profiling analyses revealed that the transport of serine family amino acids, including glycine, cysteine, and serine, was enhanced. However, only serine was increased in the growth medium of activated microglia; especially, excitotoxic D-serine secretion from primary rat microglia was the most strongly enhanced. Activated primary microglia reduced intracellular ATP levels and proteasome activity in cocultured neuronal cells, especially in primary cortical neurons, via D-serine secretion. Moreover, ubiquitinated proteins accumulated and inclusion bodies were increased in primary dopaminergic and cortical neurons cocultured with activated primary microglia. In vivo, MNPs@SiO2(RITC), D-serine, and ubiquitin aggresomes were distributed in the MNPs@SiO2(RITC)-treated mouse brain. MNPs@SiO2(RITC)-induced activation of microglia triggers excitotoxicity in neurons via D-serine secretion, highlighting the importance of neurotoxicity mechanisms incurred by nanoparticle-induced microglial activation.


Background
Nanoparticles (NPs) for drug delivery, imaging, and diagnosis to the brain have been studied due to their ultra-small sizes. Some types of NPs, such as polymeric NPs [1], gold NPs [2], zinc oxide NPs [3], quantum dots [4], and silica-coated magnetic nanoparticles containing rhodamine B isothiocyanate dye [MNPs@SiO 2 (RITC)] [5] can cross the blood-brain barrier (BBB) through passive diffusion and endocytosis to access the brain. Notably, MNPs@SiO 2 (RITC) has been developed for cell labeling, separation, and magnetic resonance imaging contrast [6]. MNPs@SiO 2 (RITC) has a cobalt ferrite core and an RITC-containing silica shell; it can be detected using magnetic and fluorescence properties, thus enabling its broad application. Moreover, the silica shell is used for stability and biocompatibility owing to the highly toxic properties of cobalt ferrite [5,[7][8][9]. The toxicity of MNPs@SiO 2 (RITC) could not be detected by in vitro conventional methods, including fluorescenceactivated cell sorting (FACS) cell death assay, cell viability assay, and chromosome aberration assay, in the HEK293 cell line and human bone marrow-derived mesenchymal stem cells [9][10][11]. Moreover, the pathological symptoms could not be detected by conventional methods, such as histopathological, growth, behavior, and serum biochemical analyses, in an in vivo mouse model [5,11,12]. However, the subtle toxicity of MNPs@SiO 2 (RITC) has recently been detected using transcriptomics, metabolomics, and mechanobiology [7][8][9][13][14][15].
Accumulation of NPs in the brain is a serious concern owing to its toxic effects on CNS. Previous studies reported that NPs exposure and accumulation can induce neurodegeneration in brain by oxidative stress, deficits in motor and cognitive functions, neuronal deterioration, DNA damage by lipid peroxidation, neuroinflammatory response, microglia/astrocyte activation, and BBB damage/dysfunction [16]. In mice, MNPs@SiO 2 (RITC) crossed the BBB without disrupting it and accumulated in the brain for longer than 4 weeks, although toxicological symptoms were not detected using conventional methods [5].
Previous studies have found that protein fibrilization is observed in Alzheimer's disease by NPs (cerium oxide particles, copolymer particles, carbon nanotubes, quantum dots) in vitro [17], nuclear protein aggregation by silica NPs is observed in a cellular model of Huntington's disease, and formation of intracellular cytoplasmic inclusion bodies by MNPs@SiO 2 (RITC) in neuronal cells [8], may play a crucial role in neurodegenerative diseases. In addition, internalized metal-oxide nanoparticles in microglia-induced inflammation is a common factor in many neurodegenerative diseases [18]. However, studies related to neuro-nanotoxicity is very limited and its research area is still in infancy.
Microglia are resident macrophages which account for 5-20 % of the total brain cells. They are uniformly distributed [19] and activated by internalized NPs [20]. Upon activation by NPs, microglia secrete large amounts of ROS and nitric oxide (NO) and release various cytotoxic substances. The secreted substances can damage neurons and trigger neuronal death. For example, activated microglia impair neurons through N-methyl-D-aspartate (NMDA) receptor-mediated excitotoxicity by NMDA, polyamines, D-serine, and proteinogenic amino acid, such as glutamate and glycine [21,22]. However, the toxicological mechanisms of activated microglia by NPs are not clearly understood.
Previously, we analyzed the direct effect of MNPs@SiO 2 (RITC) on the SH-SY5Y cell line, primary rat cortical neurons, and dopaminergic (DAergic) neurons [8]. Owing to their neuronal characteristics, SH-SY5Y neuroblastoma cells have been widely used in neuroscience research, including research on Parkinson's disease, Alzheimer's disease, neurotoxicity, ischemia, and amyotrophic lateral sclerosis; however, SH-SY5Y cells also exhibit genetic aberrations of cancerous origin [23]. MNPs@SiO 2 (RITC)-induced intracellular ROS generation, reduction in proteasome activity, and formation of inclusion bodies were evaluated in SH-SY5Y cells, cortical neurons, and DAergic neurons, and the changes were ranked in the following order according to their sensitivity to oxidative stress: DAergic neurons > cortical neurons > SH-SY5Y cells. In this study, we analyzed the indirect effect of MNPs@SiO 2 (RITC) on SH-SY5Y cells, cortical neurons, and DAergic neurons with respect to excitotoxicity using a co-culture with MNPs@SiO 2 (R-ITC)-treated microglial cells.
There are also few limitations in the evaluation of nanotoxicity using conventional methods due to its delicate complexities. Therefore, omics analyses have been used in the assessment of nanotoxicity through the quantification of gene expression, proteins, and metabolites for interpreting complex biological events and finding relationships between molecules and phenomena [11,24]. Thus, integrated omics is indispensable to compensate for their related weaknesses and comprehensive analysis of nanotoxicity.
We investigated MNPs@SiO 2 (RITC)-induced biological changes in microglia by transcriptome analysis combined with amino acid profiling. Additionally, we examined the indirect effects of the MNPs@SiO 2 (RITC) on neuronal cells by co-culturing with MNPs@SiO 2 (R-ITC)-treated microglia. Finally, we validated MNPs@SiO 2 (RITC)-induced biological changes in the mouse brain in vivo through immunohistochemical analysis.

Analysis of morphological changes in NPs treated microglia
To analyze NPs-induced microglia activation, BV2 and primary rat microglia cells were treated with 10 and 100 µg/ml of MNPs@SiO 2 (RITC) and SiO 2 NPs or 10 and 100 ng/ml of LPS as activation control for 12 h. Images were acquired with an AxioVert 200 M fluorescence microscopy (Zeiss, Jena, Germany) at the Three-Dimensional Immune System Imaging Core Facility of Ajou University.

TEM
To analyze ultrastructural changes, Karnovsky's fixative (Sigma-Aldrich, MO, USA) was used for MNPs@SiO 2 (R-ITC)-treated BV2 cells fixation at 4°C for 12 h. After fixation, samples were washed with 0.1 M cacodylate buffer (pH 7.4). Cacodylate buffer (0.1 M) with osmium tetroxide [1 % (v/v), Polysciences, PA, USA] was used for post-fixation at RT for 2 h. The samples were dehydrated with graded ethanol solutions (50-100 %), infiltrated with propylene oxide, and embedded in Epon (Polysciences, PA, USA). The samples were incubated at 35°C for 6 h, 45°C for 12 h, and 60°C for 24 h. The blocks were sectioned using an ultramicrotome (Reichert-Jung, Bayreuth, Germany). The sections were double-stained with 6 % uranyl acetate (Electron Microscopy Sciences, PA, USA) for 20 min and lead citric acid for 10 min (Thermo Fisher Scientific, CA, USA) for contrast. Images were obtained using a SIGMA500 transmission electron microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) at the Three-Dimensional Immune System Imaging Core Facility of Ajou University. Particle number, vesicle size, mitochondria number, and size were analyzed using the Zen blue 2.3 (Carl Zeiss Microscopy GmbH, Jena, Germany) image analysis module.

RNA-seq
To analyze transcriptome of control BV2 cells and 10 or 100 µg/ml MNPs@SiO 2 (RITC) treated BV2 cells, total RNA was extracted using a TruSeq Stranded Total RNA Library Prep Kit (Illumina, CA, USA). An Agilent RNA 6000 Nano Kit (Agilent Technologies, Waldbronn, Germany) was used for analyzing RNA quality. mRNA was enriched using Magnetic beads conjugated with oligo(dT). Double-stranded cDNA was synthesized with purified mRNA after fragmentation. Sequencing adapters using a TruSeq RNA Sample Prep Kit (Illumina, CA, USA) was used for modification, including end-repair and poly(A) addition and connection, of the cDNA. The samples were isolated using Blue Pippin (Sage Science, MA, USA). cDNA library size and quality were determined using an Agilent High Sensitivity DNA Kit (Agilent Technologies, Waldbronn, Germany). The libraries were sequenced using an Illumina HiSeq2500 sequencer (Illumina, CA, USA).

Differentially Expressed Gene (DEG) analysis
Low-quality reads (reads containing more than 10 % skipped bases, reads containing more than 40 % of bases with quality scores < 20, and reads with an average quality score < 20) were filtered from the RNA-seq data using in-house scripts. Filtered reads were identified using the aligner TopHat [26]. Expression levels were calculated using Cufflinks v2.1.1 [26] and Ensembl, release 77. Multi-read-correction and fragbias-correct were used for increasing the measurement accuracy. DEGs were singled out using Cuffdiff with default parameter settings, based on p < 0.05.

Gene Ontology (GO) and pathway analyses
The GO database classifies genes into functional categories and can be used to predict gene function based on Mouse Genome Informatics (MGI) data [27]. Biological changes, related with canonical pathways and functions, were analyzed using an ingenuity pathway (IPA) web-based software (Qiagen, CA, USA). A 1.5-fold change in gene expression and a 1.2-fold change in amino acid levels were used as cut-offs.

Quantitative reverse transcription PCR (RT-qPCR)
To quantify gene expression levels of MNPs@SiO 2 (R-ITC) BV2 cells, RNA extraction and cDNA construction for RT-qPCR were performed as previously study [8]. The cells were lysed using RNAzol B (Tel-Test, TX, USA). Chloroform (Sigma-Aldrich, MO, USA) and isopropyl alcohol (Sigma-Aldrich, MO, USA) were used for total RNA precipitation. Pellets were washed with 70 % ethanol and the samples were dissolved in RNase-free water. The purities of samples were measured by 260 nm/230 nm and 260 nm/280 nm absorbance ratio using spectrophotometry (Eppendorf, Hamburg, Germany). cDNA was synthesized using the iScript Advanced cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). The thermal cycle was as follows: 46°C for 20 min followed by 95°C for 1 min.
Expression levels of genes were quantified by qPCR using SsoAdvanced™ Universal SYBR® Green Supermix (Bio-Rad) and gene-specific primers (Supplementary  Tables 1 and 2) on a Rotor Gene-Q system (Qiagen, CA, USA). Thermal cycles were as follows: 95°C for 5 min, followed by 50 cycles of 95°C for 5 s and 60°C for 30 s. Target gene expressions were calculated with melting curves by the 2 −ΔΔCt method.

Gas chromatography (GC)-MS
The amino acid content in the MNPs@SiO 2 (RITC)treated BV2 cells and media was analyzed using GC-MS as reported previously [9]. Norvaline is used as internal standard. GC-MS was conducted in both the scan and selected-ion monitoring (SIM) modes on a 6890 N gas chromatograph (Agilent Technologies, Santa Clara, CA, USA) interfaced with a 5975B mass-selective detector (70 eV, electron impact ionization mode; Agilent Technologies).

Determination of D-serine secreted from microglia
D-Serine concentrations in the medium were determined using a D-serine colorimetric assay (Cosmo Bio, Tokyo, Japan) per the manufacturer's instructions. Briefly, medium of the MNPs@SiO 2 (RITC) or lipopolysaccharide (LPS)-treated BV2 cells and primary rat microglia were mixed with NADH and D-serine dehydratase from Saccharomyces cerevisiae (DsdSC), D-serine to pyruvate converting enzyme, at 37°C for 45 min. Dserine samples incubated with nicotinamide adenine dinucleotide (NADH), but without DsdSC were used as blank. The absorbance at 340 nm was measured using a microplate reader (Molecular Devices, CA, USA). Lactate dehydrogenase (LDH) was added to the mixed samples and incubated at 37°C for 45 min. The pyruvate to lactate conversion by LDH leaded to reduction in optical density at 340 nm by NADH oxidation. The D-serine concentration was quantified by calculating the reduction absorbance at 340 nm and a D-serine standard curve.

Liquid Chromatography (LC)-MS
L-and D-Serine were purchased from Tokyo Chemical Industry (Tokyo, Japan). HPLC-grade acetonitrile, ethanol, and trifluoroacetic acid were obtained from Thermo Fisher Scientific (CA, USA). Stock solutions of L-and D-serine were prepared at 1 mg/mL in water. A standard stock solution of amino acids was diluted to 0.01-100 µg/ml for calibration. Calibration curves were plotted with seven points using the peak areas obtained for D-and L-serine versus the concentration by linear regression. The samples were centrifuged at 12,000 ⋅ g for 15 min. The supernatant was filtered through a 3-kDa filter (Millipore, Darmstadt, Germany) by centrifugation at 14,000 ⋅ g at 4°C for 30 min. The filtered aliquot was used for analysis. Chiral DL-serine separations were performed using an Agilent 1290 infinitely LC (Agilent Technologies, Walbronn, Germany) using a CROWNPAK CR-I(+) (3.0 × 150 mm, 5 μm, Chiral Technologies) at 30°C. Mobile phase comprising acetonitrile: ethanol: water: trifluoroacetic acid (80: 15: 5: 0.5) was employed at a flow rate of 0.4 ml/min. Two microliters of sample were injected into the column using a thermostated HiP-ALS autosampler. The HPLC system was interfaced with an Agilent 6495 Q-TOF (Agilent Technologies, Walbronn, Germany) mass spectrometer. The source conditions were set to a Vcap of 3.5 kV in positive ESI mode with a sheath gas temperature of 200°C and a sheath gas flow rate of 11 ml/min. The DL-serine was detected by multiple reaction monitoring. A product ion at m/z 60 fragmented by a collision energy of 10 V from the precursor ion at m/z 106 was monitored for 200 ms.
Evaluation of neuronal cell viability in a microglia coculture system The coculture system was set up as reported previously [28]. Briefly, neuronal cells or SH-SY5Y cells were cultured on the bottom side of a Costar Transwell plate (0.4-µm pore size; Corning, NY, USA) in a humidified 5 % CO 2 incubator at 37°C for 24 h. Microglia in the culture inserts were treated with MNPs@SiO 2 (RITC) or LPS for 12 h. Then, the inserts were placed in the plate and incubated for another 12 h. To measure the viability of the neuronal cells, the upper inserts were removed, and 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2 H-tetrazolium solution (CellTiter 96® AQueous One; Promega, WI, USA) was added to each well containing the microglia. The assay plate was incubated at 37°C for 1 h. The amount of soluble formazan produced via cellular reduction was measured using a plate reader (Molecular Devices, CA, USA) at 490 nm. Values were normalized to the corresponding total protein. Images of cell morphology and density were acquired before the cell viability assay under an Axiovert 200 M fluorescence microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) at the Three-Dimensional Immune System Imaging Core Facility of Ajou University. The excitation wavelength for the MNPs@SiO 2 (RITC) was 530 nm.

Measurement of the ATP concentration
The ATP concentration of the microglia cocultured neurons and D-serine treated neurons were determined using an ATP quantification kit (Promega, WI, USA) per the manufacturer's protocol. Briefly, the neuronal cells were detached from plate and washed with PBS. Intracellular ATP was extracted with a 1.0 % trichloroacetic acid (Sigma Aldrich, MO, USA) at RT for 30 min. The extractant were mixed with a luciferin reagent/buffer mixture and then split in white 384-well plates. After 10 min of incubation at 25°C, the luminescence in each well was measured using a luminometer (LMaxII 384 ; Molecular Devices, CA, USA).

Proteasome activity assay
Proteasome activity was measured using a Proteasome-Glo™ Chymotrypsin-Like Cell-Based Assay (Promega, WI, USA) per the manufacturer's protocol. Briefly, microglia cocultured neurons and D-serine treated neurons were washed three times with PBS, replenished with 50 µl of serum-free medium, and left at RT. Proteasome-Glo™ cell-based buffer was mixed with luciferin detection agent and appropriate Proteasome-Glo™ substrate and incubated at RT for 30 min. Fiftymicroliter aliquots of this mixture were then added to the medium in each well and incubated at RT for 10 to 15 min. Next, 90 µl of the medium containing a proteasome assay solution was transferred to corresponding wells in a white plate. Proteasome activity was determined by measuring the chymotrypsin-like activity of cellular proteasomes using the luminogenic substrate Suc-Leu-Leu-Val-Tyr-aminoluciferin in a luminometer (LMaxII 384 ; Molecular Devices, CA, USA) per the manufacturer's instructions.

Evaluation of D-serine distribution and inclusion body formation in vivo
All animal experimental protocols were approved by the Laboratory Animal Research Center of Ajou University Medical Center (approval no. 2020-0033) and complied with the institutional ethical use protocols (NIH Guide for Care and Use of Laboratory Animals). Male ICR mice (8 weeks old, Orientbio, Seongnam, Korea) were maintained under 12-h light/dark cycles with free access to food and water. Four mice per group were used in this study. The biodistribution and toxicological effects of the MNPs@SiO 2 (RITC) at 25, 50, and 100 mg/kg were previously reported [5]. The study showed the broad-range tissue distribution of the MNPs@SiO 2 (R-ITC) and the brain localization of the MNPs@SiO 2 (R-ITC) without blood-brain barrier disruption. The particles did not induce significant toxicological symptoms in terms of growth, behaviors, biochemical changes in serum, and histopathology, even at 100 mg/kg. However, based on our in vitro findings, we postulated that there would be subtle toxicity in the brain at 100 mg/kg, which was the maximum concentration evaluated in the previous study. Thus, MNPs@SiO 2 (RITC) were injected intraperitoneally with sterile saline at 100 mg/kg per mouse. Control mice were injected with sterile saline only. The endpoint was determined at 5 days, based on a previous biodistribution study [5]. Five days after the injection, the mice were anaesthetized with urethane (1.2- After the wash steps, the sections were mounted with Vectashield mounting medium containing DAPI (Vector Laboratories, CA, USA), and coverslips were applied. Immunostained sections were scanned using a slide scanner (Axio Scan Z1, Carl Zeiss Microscopy GmbH, Jena, Germany), and regions of interest were viewed under a 40× objective lens on an A1R HD25 confocal microscope (Nikon, Tokyo, Japan) at the Three-Dimensional Immune System Imaging Core Facility of Ajou University. Three-dimensional reconstructions of branch structures were acquired and quantified for number and length, using Imaris 9.2 software (Bitplane, Zurich, Switzerland) at the Three-Dimensional Immune System Imaging Core Facility of Ajou University.

Statistical analysis
Data were analyzed by analysis of variance (ANOVA) with Bonferroni's multiple-comparison tests, using IBM-SPSS software (IBM, NY, USA). Differences were considered significant at p < 0.05.

Results
Phenotype changes and NP uptake efficiency of MNPs@SiO 2 (RITC)-treated microglia To evaluate the effect of MNPs@SiO 2 (RITC) on microglia, we first analyzed morphological changes in MNPs@SiO 2 (RITC)-treated BV2 immortalized murine microglial cells and primary rat microglia. The cells were treated for 12 h with 10 or 100 µg/ml MNPs@SiO 2 (R-ITC) (Fig. 1a) or with 0 (non-treated control), 10, or 100 ng/ml of LPS as a positive control for morphological activation. Morphological activation (i.e., swollen and round cells) was observed in MNPs@SiO 2 (RITC)-and LPS-treated microglia (Supplementary Fig. 4). Compared to non-treated control cells, BV2 cells and primary rat microglia showed dose-dependent morphological changes following MNPs@SiO 2 (RITC) or LPS treatment, but there was no significant difference in cell density.
We previously suggested that the biological changes in HEK293 cells were induced by the shell (silica) rather than the cobalt ferrite core of MNPs@SiO 2 (RITC) [7,8,11,[13][14][15]. Moreover, our previous studies revealed that these biological changes, which included cell viability, intracellular ROS generation, and proteasome activity, were similar between MNPs@SiO 2 (RITC)-and silica NP-treated HEK293 cells. In this study, we also analyzed morphological changes in silica NP-treated microglia, which exhibited similar trends between MNPs@SiO 2 (R-ITC)-and silica NP-treated microglia ( Supplementary  Fig. 5).
Iba1 is a microglial marker whose expression is elevated upon microglial activation [29].
To evaluate detailed changes at the organelle level, the MNPs@SiO 2 (RITC)-treated BV2 cell ultrastructure was visualized using TEM. MNPs@SiO 2 (RITC) were clearly visible as black dots throughout the cytoplasm, and the mitochondria showed the most pronounced changes in size and number among the organelles (Fig. 1c). MNPs@SiO 2 (RITC) were not distributed in the nucleus. The lysosomes, containing MNPs@SiO 2 (RITC), were approximately 800 nm in diameter. The number of granular structures was increased by~3-fold in 100 µg/ml MNPs@SiO 2 (RITC)-treated BV2 cells compared to 10 µg/ml MNPs@SiO 2 (RITC)-treated and control cells.

D-Serine secretion in MNPs@SiO 2 (RITC)-treated microglia in vitro
To validate the above findings, we analyzed the extracellular secretion of the above 19 amino acids, including serine family amino acids, in the medium of MNPs@SiO 2 (RITC)-treated BV2 cells using EOC/ TBDMS derivatization and GC-MS (Supplementary Table 6). The increase in serine in the medium was significantly higher than the increases in glycine and cysteine.
The GC-MS platform used in this study could not distinguish enantiomers of serine; therefore, we measured Dserine in the media of MNPs@SiO 2 (RITC)-treated BV2 cells and primary rat microglia using a D-serine-specific enzyme-based method. D-Serine secretion was increased after LPS and MNPs@SiO 2 (RITC) treatment in BV2 cells and primary microglia (Fig. 3b). Specifically, the increase in D-serine secretion was the most pronounced in 100 µg/ ml MNPs@SiO 2 (RITC)-treated primary microglia. To validate the aforementioned results of the indirect method, we reanalyzed D-serine by liquid chromatography mass spectrometry, which is a direct detection method for pure D-serine level, using an enantio-separation column. Land D-serine were clearly separated, and the extracellular levels of D-serine were increased in 10 ng/ml LPS-and 10 µg/ml MNPs@SiO 2 (RITC)-treated primary rat microglia; the changed levels exhibited a similar trend and degree (difference within ± 10 %) in both the indirect and direct methods (Fig. 3c, d). L-Serine levels were rather decreased in LPS-and MNPs@SiO 2 (RITC)-treated cells compared to those in control cells (Fig. 3e).

ATP and proteasome activity are reduced in neuronal cells cocultured with microglia
To evaluate the effect of activated microglia and secreted D-serine on neuronal cells, we used a Transwell coculture system (Fig. 4a). Microglia, activated with 10 or 100 µg/ml MNPs@SiO 2 (RITC) or 10 or 100 ng/ml LPS for 12 h, were cocultured with SH-SY5Y, cortical, and DAergic neurons. After 12 h, the morphology and density of the neuronal cells were observed by microscopy, and cell viability was analyzed ( Supplementary  Fig. 9). The viability of cortical neurons was decreased, and aggresome-like structures were detected after coculture with 100 µg/ml MNPs@SiO 2 (RITC)-activated microglia. However, cell viability and density were not affected in MNPs@SiO 2 (RITC)-activated microgliacocultured SH-SY5Y cells and DAergic neurons.
To evaluate the effect of D-serine secreted from the activated microglia on neuronal cells, we first focused on the mitochondria, because calcium ions from the excitotoxic cascade mainly impair calcium channel-containing organelles such as the mitochondria and the endoplasmic reticulum [30], and the apoptosis cascade is highly associated with the mitochondria [31]. The mitochondrial membrane potential (Δψ m ) was analyzed in Red and green areas indicate the up-and downregulated genes, respectively. As cut-offs, fold changes > 1.5 or <-1.5 for genes and of > 1.2 or <-1.2 for amino acids were used. d RT-qPCR analysis of gene expression in each group, using GAPDH as an internal control. Data represent the mean ± standard deviation of three independent experiments. *p < 0.05 vs. non-treated control Decreases in Δψ m are highly correlated with decreases in mitochondrial function, specifically ATP production [32]. Intracellular ATP was measured in neuronal cells cocultured with MNPs@SiO 2 (RITC)-activated microglia. Intracellular ATP levels were reduced in neuronal cells cocultured with 100 µg/ml MNPs@SiO 2 (RITC)-activated microglia, and the most pronounced decrease was observed in cortical neuronal cells (Fig. 4b). The decrease in ATP was also alleviated in the presence of DCKA.
Proteasome activity is highly dependent on the intracellular ATP level [33][34][35]. Thus, we also analyzed the proteasome activity of neuronal cells cocultured with MNPs@SiO 2 (RITC)-activated microglia (Fig. 4c). In line with the decrease in intracellular ATP (Fig. 4d-f), proteasome activities were decreased in neuronal cells cocultured with 100 µg/ml MNPs@SiO 2 (RITC)-activated microglia and the decrements were also alleviated in the presence of DCKA (Fig. 4g-i). In addition, we evaluated the effects of D-serine on the intracellular ATP level and proteasome activity in D-serine-treated SH-SY5Y cells, dopaminergic neurons, and cortical neurons (Supplementary Fig. 11). Decreases in the intracellular ATP level and proteasome activity were more pronounced in cortical neurons than in dopaminergic neurons; these decreases were alleviated in the presence of DCKA. However, there were no significant changes in D-serinetreated SH-SY5Y cells. The impairments of mitochondria and proteasome activity in the neuronal cells were highly correlated with exposure to D-serine secreted from the MNPs@SiO 2 (RITC)-activated microglia.

Inclusion body formation in neuronal cells cocultured with microglia
We assessed the accumulation of ubiquitinated proteins following the reduction in proteasome activity.
Ubiquitinated protein levels were significantly higher in neuronal cells cocultured with MNPs@SiO 2 (RITC)-activated microglia (Fig. 5a). When neuronal cells were cocultured with the 100 µg/ml MNPs@SiO 2 (RITC)-activated microglia, the increase in ubiquitinated proteins was the most pronounced in the cortical neurons. DCKA suppressed the increases in ubiquitinated proteins.
Immunocytochemical analysis revealed a dramatic increase in inclusion bodies in cortical neuronal cells cocultured with 100 µg/ml MNPs@SiO 2 (RITC)-activated microglia (Fig. 5b). However, inclusion body formation was not detected in SH-SY5Y cells ( Supplementary  Fig. 12). We observed a dose-dependent increase in inclusion bodies in the DAergic neuronal cells cocultured with MNPs@SiO 2 (RITC)-activated microglia (Supplementary Fig. 13), but not as strong as that observed in the cortical neuronal cells (Fig. 5c).

Discussion
The present study combined conventional biomolecular and integrated omics-based analyses to evaluate the MNPs@SiO 2 (RITC)-induced activation of microglia and (See figure on previous page.) Fig. 3 Analysis of the changes in serine family amino acids and quantification of L-and D-serine secretion by BV2 cells treated with MNPs@SiO 2 (RITC). a SIM chromatograms of glycine, cysteine, and serine in (i) control (ii) 10 µg/ml, and (iii) 100 µg/ml MNPs@SiO 2 (RITC)-treated microglia. The internal standard was norvaline. b Quantitative analysis of D-serine in the medium of MNPs@SiO 2 (RITC)-treated microglial cells (top panel: BV2 cells, bottom panel: primary rat microglia) using an enzyme-based system. Data were normalized to the control and represent the mean ± SD of three independent experiments. *p < 0.05 vs. non-treated control. c Multiple reaction monitoring peaks of L-and D-serine in the control, 10 µg/ml, and 100 µg/ml MNPs@SiO 2 (RITC)-and 10 and 100 ng/ml LPS-treated primary rat microglia. Quantification and statistical analysis of L-d and D-serine e in control, 10 µg/ml, and 100 µg/ml MNPs@SiO 2 (RITC)-and 10 and 100 ng/ml LPS-treated primary rat microglia. *p < 0.05 vs. non-treated control. # p < 0.05 vs. 10 µg/ml MNPs@SiO 2 (RITC)-treated cells the neuronal toxicity of the activated microglia. Especially, we profiled the transcriptome and amino acids in the MNPs@SiO 2 (RITC)-treated microglia and subjected significantly changed factors to functional analysis to deduce relationships between microglial activation and Dserine-related mechanisms after NP exposure. We demonstrated that high-concentration MNPs@SiO 2 (RITC) induced microglial activation and D-serine secretion. The high concentration of D-serine secreted from the activated microglia induced neuronal cell death, especially in cortical neurons, via excitotoxicity, including mitochondrial depolarization, a reduction in intracellular ATP, a decrease in proteasomal activity, and the accumulation of ubiquitinated proteins. Our results suggested that the secondary effect of nanoparticle exposure should also be considered in toxicological assessments. Moreover, minimizing NP usage for drug delivery, imaging, and diagnosis purposes is important for preventing nanotoxicity in the brain with respect to microglial activation and neuronal excitotoxicity. The direct effect of MNPs@SiO 2 (RITC) on microglia should also be considered. ROS generation by NPs triggered mitochondrial dysfunction and decreased ATP production [9,11,36]. Owing to mitochondrial dysfunction, microglia exposed to NPs might exhibit limited regulation of brain homeostasis, which is a decisive role of the microglia in the brain. Therefore, toxicological assessment of microglia exposed to MNPs@SiO 2 (RITC) is also necessary for understanding the neuronanotoxicity of MNPs@SiO 2 (RITC).
In this study, MNPs@SiO 2 (RITC) was used in a monodispersed state, as shown in Supplementary Fig. 1. However, the parameters of NPs reaching the brain depend on the form of the primary particle, agglomerates, and aggregates; therefore, these factors should be considered [37,38]. Thus, further studies are required to determine the level of NP agglomeration and aggregation for assessing neuronanotoxicity in the brain.
The leaching of MNPs@SiO 2 (RITC) components, including RITC and cobalt ferrite, has previously been considered during long-term exposure. In a previous study, the fluorescence intensities of RITC were decreased by 10 % at the third passage and by 60 % at the seventh passage in human cord blood-derived mesenchymal stem cells; this phenomenon was considered a dilution effect by cell proliferation [10]. In addition, cobalt ferrite is highly toxic [8,9]; however, there were no significant decrease in the viability of MNPs@SiO 2 (R-ITC)-treated HEK293 cells for seven days. Moreover, MNPs@SiO 2 (RITC)-treated mice did not show any pathological symptoms until four weeks [5]. Thus, we postulated that the MNPs@SiO 2 (RITC) remained intact for at least seven days and that the component leaching effect may be minor in MNPs@SiO 2 (RITC)-treated cells. However, the subtle toxicity caused by leaching of MNPs@SiO 2 (RITC) components should be considered during the long-term exposure of MNPs@SiO 2 (RITC) and requires further analysis.
We observed intracellular granular structure formation and increment of D-serine secretion in MNPs@SiO 2 (R-ITC)-treated microglia. D-Serine secretion in microglia responds to amyloid beta, LPS, and secreted amyloid precursor protein [39,40], but D-serine secretion in NPtreated microglia had not been reported. Concomitant with the increase in D-serine secretion, gene expression related to the transport of serine family amino acids was increased as indicated by combined transcriptome and amino acid profiling analyses. In astrocytes, D-serine secretion mechanisms include synaptosomal-associated proteins receptor (SNARE)-dependent exocytosis as well as non-exocytic mechanisms such as volume-regulated anion channel, alanine-serine-cysteine transporter, and gap junction hemichannels. The mechanism of D-serine release by microglia is still unclear [40,41]. Moreover, although we have previously observed excitotoxic glutamate production in MNPs@SiO 2 (RITC)-treated HEK293 cells based on metabotranscriptome analysis [9], the mechanisms of D-serine production and secretion in MNPs@SiO 2 (RITC)-treated microglia have not been unraveled. In MNPs@SiO 2 (RITC)-treated microglia, D-serine did not colocalize with VAMP2 nor NPY, which are exocytic vesicle markers in astrocytes ( Supplementary Fig. 15) [42,43], suggesting that microglia utilize a pathway for D-serine secretion different from that of astrocytes and detailed future analysis is required to reveal the specific mechanism. In a previous study, the MNPs@SiO 2 (RITC)-uptake efficiency of MCF-7 mammary gland adenocarcinoma cells was determined using ICP-atomic emission spectroscopy [6]. In addition, MCF-7 cells have been thoroughly characterized in nanoparticle uptake studies with various composites of nanoparticles and are thought to be a candidate for model cell lines [44,45]. The efficiency was approximately 10 5 particles/cell at 2.0 mg/ml MNPs@SiO 2 (RITC) in MNPs@SiO 2 (RITC)-treated MCF-7 cells. In the case of BV2 cells in this study, the b Immunohistochemical analysis of the cerebellum regions of mice. Low-magnification images are merged with the florescence images of Hoechst 33,342 (blue), MNPs@SiO 2 (RITC) (red), Iba1 (green), and D-serine (violet) to show region-specific structures and the distribution of MNPs@SiO 2 (RITC) and D-serine. Scale bar = 100 μm. Magnified images are separated into Hoechst 33,342, MNPs@SiO 2 (RITC), Iba1, and D-serine and Iba1-based 3D-rendered images. White scale bar = 10 μm. c Filament length determined from the 3D-rendered images. d Quantification of nuclei in brain sections. *p < 0.05 vs. control efficiency of internalized MNPs@SiO 2 (RITC) was approximately 2.3 × 10 4 particles/cell at 10 µg/ml MNPs@SiO 2 (RITC) and 2.0 × 10 5 particles/cell at 100 µg/ml MNPs@SiO 2 (RITC). Thus, the MNPs@SiO 2 (RITC)-uptake efficiency of BV2 cells was approximately 40-fold higher than that of MCF7 cells.
The MNPs@SiO 2 (RITC)-induced D-serine secretion reduced the intracellular ATP level, mitochondrial function, and proteasomal activity in neuronal cells via excitotoxicity. Excitotoxicity is highly correlated with the pathological phenotype of neurodegenerative disease [46,47], and it is triggered by the opening of calcium channels, including the NMDA receptor, by agonists. Physiologically, extracellular and intracellular calcium ion concentrations differ by 20,000-fold (mM vs. 100 nM) [48]. The excessive influx of calcium ions is regulated by storage in the endoplasmic reticulum and mitochondria in the physiological state [49]. However, an overwhelming influx of calcium ions against the physiological regulation induces a calcium ion imbalance [50] and activates calcium ion-cofactor catabolic enzyme activities, which are involved in cell death mechanisms [49]. Moreover, overloaded calcium ions induce mitochondrial membrane depolarization, mitochondrial dysfunction, and reduced ATP production [30,51]. ATP is essential for ubiquitination, de-ubiquitination, and proteasome complex assembly [33][34][35]. When ATP strongly decreases, the ubiquitin-proteasome system is corrupted, resulting in the aggregation and accumulation of defective proteins in cells. In neurons, such protein aggregations are histopathological biomarkers for neurodegenerative diseases [52]. Thus, the D-serine level in the brain may be monitored after NP exposure as a secondary (indirect) phenotype of neuronanotoxicity.
Cortical neuronal cells reacted more sensitively to Dserine-induced excitotoxicity than DAergic neurons. Overactivation of the NMDA receptor by D-serine induces neuronal death and protein aggregation [53,54]. In this study, cortical Lewy bodies, which are a pathological feature of Lewy body dementia closely related to Parkinson's disease [55], were detected in in vitro MNPs@SiO 2 (RITC)-activated microglia-cocultured cortical neurons and in vivo in the mouse brain. Neuronal cells in the MNPs@SiO 2 (RITC)-activated microglia coculture system showed decreased viability, along with decreases in intracellular ATP and proteasome activity, and the effects were the most pronounced in cortical neurons. Our previous study showed that DAergic neurons are more vulnerable to MNPs@SiO 2 (RITC)-induced nanotoxicity (direct effect) than cortical neuronal cells [8]. In this study, cortical neurons responded more sensitively to D-serine secreted from MNPs@SiO 2 (R-ITC)-activated microglia than DAergic neurons. This is consistent with the finding that cortical neurons are more vulnerable to excitotoxic conditions than DAergic neurons, because endogenous dopamine from DAergic neurons can block excitotoxicity-induced calcium signals and plays a major role in conserving calcium homeostasis in cells [56]. Moreover, molecules abundant in DAergic neurons, including kynurenic acid, brain-derived neurotrophic factor, and prostaglandin A1, protect against excitotoxicity in DAergic neurons [57][58][59]. Collectively, we showed the induction of a series of dysfunctions in neuronal cells by D-serine secreted from MNPs@SiO 2 (RITC)-activated microglia, suggesting the secondary toxicity of NPs.
In a previous study, the mouse whole-body biodistribution of MNPs@SiO 2 (RITC) was investigated at three different concentrations (25 mg/kg, 50 mg/kg, and 100 mg/kg) using a confocal microscope [5]. Intraperitoneally administered MNPs@SiO 2 (RITC) accumulated in the brain, spleen, heart, liver, lungs, kidneys, and uterus. Although quantification of MNPs@SiO 2 (RITC) in the brain was limited, the fluorescence intensity of cells was similar between brain slices of mice receiving 100 mg/kg and microglia treated with 10 µg/ml MNPs@SiO 2 (R-ITC). To compare the relevance of NPs in vitro versus that in vivo in the brain, further research is required to calculate the number of NPs in vivo using highresolution techniques such as X-ray-based analytics, magnetic resonance imaging, or laser ablation ICP-MS [60][61][62]. In addition, further studies are required to investigate the interspecies comparisons and the possibility of using cell lines versus primary cultures in NP-treated cells.

Conclusions
Based on transcriptomics and amino acid profiling, this study revealed that NP-induced microglial activation is associated with excitotoxic D-serine secretion. The present findings suggest that exposure to NPs can directly activate microglia and indirectly be deleterious to neurons. These results highlight the importance of the safety evaluation of NPs and the assessment of secondary nanotoxicity effects.