Introduction

Nanoparticles (NPs) are increasingly used for diagnostic and treatment purposes in biomedicine (Jain and Stylianopoulos 2010; Qi et al. 2017; Stark 2011). Magnetic nanoparticles (MNPs) and MNPs coated with biocompatible compounds such as silica, polyethylene glycol, and polysaccharides are used in vivo as tracers, contrast agents for magnetic resonance imaging (MRI)-based cell labelling, and transfection agents (Delyagina et al. 2014; Ding et al. 2018; Ito et al. 2006; Silva et al. 2017; Yoon et al. 2005). However, the nature and underlying cellular mechanisms of NP toxicity remain unclear.

Previous studies have demonstrated that internalised NPs induce cytotoxicity by stimulating ROS production and oxidative protein modification, alterations in redox-regulation, and mitochondrial dysfunction (Krug and Wick 2011; Meng et al. 2009; Phukan et al. 2016). Under physiological conditions, these disturbances are associated with functional alterations in cellular metabolism. Indeed, glucose and energy metabolic dysfunction after NP exposure can potentially contribute to disease occurrence (Lai et al. 2015) and renal cytotoxicity (Iavicoli et al. 2016). The kidneys play an important role in glucose homeostasis by filtering and reabsorbing ~ 180 g of glucose per day in humans (DeFronzo et al. 2012). Impaired glucose uptake in the kidneys leads to glycosuria, a general state of imbalanced glucose homeostasis with symptoms similar to those of diabetes mellitus and Fanconi–Bickel syndrome (Bahillo-Curieses et al. 2017; Cersosimo et al. 2014). However, the effects of NPs on glucose metabolism are poorly understood on a cellular level.

We previously demonstrated that exposure to MNPs@SiO2(RITC) altered the expression of cellular metabolism-related genes and produced metabolic disturbances in organic acids (OAs) and amino acids (AAs) in HEK293 cells using gas chromatography mass spectrometry (Shim et al. 2012; Yoon et al. 2005). Importantly, the metabolic profiling of OAs, AAs, and fatty acids (FAs) as an end-point biological phenotype (Gibney et al. 2005) can reflect important changes in cellular processes. However, there are few methods available for the amplification of very low-abundance metabolites, and quantitative analyses of targeted methods provide only a partial representation of metabolism in the cell (Van Assche et al. 2015). New approaches integrating transcriptomics and metabolomics provide a more powerful and comprehensive analysis of treatment outcomes after NP exposure compared to classical methods for the analysis of nanotoxicity (Shin et al. 2018).

In this study, we performed a comprehensive evaluation of MNPs@SiO2(RITC)-induced toxicity in vitro. We analysed changes in OAs, AAs, and FAs using gas chromatography–tandem mass spectrometry (GC-MS/MS) and disturbances in the transcriptome using a microarray in MNPs@SiO2(RITC)-treated cells. Finally, we applied a metabotranscriptomics approach to better inform the nature and mechanisms of MNPs@SiO2(RITC) nanotoxicity.

Materials and methods

Chemicals and reagents

Standards for 14 OAs (3-hydroxybutyric acid, pyruvic acid, acetoacetic acid, lactic acid, glycolic acid, 2-hydroxybutyric acid, malonic acid, succinic acid, fumaric acid, oxalacetic acid, 2-ketoglutaric acid, malic acid, 2-hydroxyglutaric acid, and citric acid), 20 AAs (alanine, glycine, valine, leucine, isoleucine, proline, γ-aminobutyric acid, pyroglutamic acid, methionine, serine, phenylalanine, cysteine, aspartic acid, glutamic acid, asparagine, ornithine, glutamine, lysine, tyrosine, and tryptophan), and 13 FAs (myristoleic acid, myristic acid, palmitoleic acid, palmitic acid, linoleic acid, oleic acid, stearic acid, arachidonic acid, arachidic acid, behenic acid, nervonic acid, lignoceric acid, and cerotic acid); ethyl chloroformate (ECF); methoxyamine hydrochloride; and trimethylamine (TEA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). N-methyl-N-(tert-butyldimethylsilyl) trifluoroacetamide (MTBSTFA) was obtained from Pierce (Rockford, IL, USA). Toluene, diethyl ether, ethyl acetate, dichloromethane, and sodium chloride were supplied by Kanto Chemical (Tokyo, Japan).

MNPs@SiO2(RITC)

MNPs@SiO2(RITC) consisted of a ∼9-nm cobalt ferrite core (CoFe2O3) that was chemically bonded to rhodamine isothiocyanate dye (RITC) and coated with a silica shell (Phukan et al. 2016; Shim et al. 2012; Yoon et al. 2005) and was provided from BITERIALS (Seoul, South Korea). The size of the NP was 50 nm in diameter and MNPs@SiO2(RITC) was previously reported to have a zeta potential between − 40 and − 30 mV (Beck et al. 2012; Yoon et al. 2005). A previous study determined that the MCF-7 cell line internalised ~ 105 particles of MNPs@SiO2(RITC) per cell using inductively coupled plasma atomic emission spectrometry (Yoon et al. 2005). The dosage used in this study was determined by treating HEK293 cells with MNPs@SiO2(RITC) at concentrations ranging from 0.01 to 2.0 µg/µl for 12 h and calculating the uptake efficiency at each concentration (Shim et al. 2012). The optimal concentration of MNPs@SiO2(RITC) for in vitro use was 0.1 µg/µl, consistent with its safe use for MRI contrast in human cord blood-derived mesenchymal stem cells (Park et al. 2010). Disturbances of gene expression and metabolic profiles at this concentration were similar to those in control HEK293 cells (Shim et al. 2012). The uptake efficiency of MNPs@SiO2(RITC) plateaued at 1.0 µg/µl. Therefore, we used 0.1 µg/µl to represent a normal or low dose and 1.0 µg/µl to represent an overdose of NPs in the present study. Additionally, we verified that the off-target biological effects of MNPs@SiO2(RITC) after treatment of HEK293 cells for 48 h were induced by the shell rather than the cobalt ferrite core (Phukan et al. 2016; Shin et al. 2018).

Cell culture

Human embryo kidney 293 (HEK293) cells were used to analyze the nanotoxicity of MNPs@SiO2(RITC) as HEK293 has been well studied for its relation to silica nanoparticle-induced cytotoxicity and the renal toxicity models (Wang et al. 2009; Zhang et al. 2015). HEK293 cells were obtained from the American Type Culture Collection. Cells were cultured in Dulbecco’s high-glucose modified Eagle’s medium (DMEM, Gibco, USA) supplemented with 10% foetal bovine serum (Gibco), 100 units/ml penicillin, and 100 µg/ml streptomycin (Gibco) and incubated in a humidified atmosphere of 5% CO2 at 37 °C.

Preparation of standard solutions

Standard stock solutions (10 µg/µl) were individually prepared for OAs including 3,4-dimethoxybenzoic acid as an internal standard, AAs including norvaline as an internal standard in 0.1 M HCl, and FAs including pentadecanoic acid as internal standard in methanol. Working solutions of 0.01 and 0.1 µg/µl were then prepared by diluting each AA stock solution with 0.1 M HCl and each OA and FA stock solution with methanol. All standard solutions were stored at 4 °C until use.

GC-MS/MS

The GC-MS/MS analysis was performed with a Shimadzu 2010 Plus gas chromatograph interfaced with a Shimadzu TQ 8040 triple quadruple mass spectrometer (Shimadzu, Kyoto, Japan) equipped with an Ultra-2 (5% phenyl-95% methylpolysiloxane bonded phase; 25 m × 0.20 mm ID, 0.11 µm film thickness) cross-linked capillary column (Agilent Technologies, Atlanta, GA, USA). Samples were introduced in split-injection mode (10:1). The oven temperature was initially set to 60 °C for 2 min, increased to 255 °C at a rate of 25 °C/min, and further increased to 300 °C at a rate of 7 °C/min with a holding time of 2.5 min. The temperatures of the injector, interface and ion source were 260 °C, 300 °C, and 230 °C, respectively. Helium (0.5 ml/min, constant flow mode) and argon were used as carrier and collision gases, respectively. Ionization used the electron ionization (EI) mode set to 70 eV.

Sample preparation for OA, AA, and FA profiling in vitro

We performed profiling analyses of 14 OAs, 20 AAs, and 13 FAs as EOC/MO/TBDMS derivatives on GC-MS/MS (Paik and Kim 2004; Paik et al. 2005). Briefly, cells were homogenised by freeze-thawing and 1.0 ml of distilled water containing 0.1 µg of 3,4-dimethoxybenzoic acid, norvaline, and pentadecanoic acid was added to each sample. Each aliquot solution was adjusted to a pH ≥ 12 with 5.0 M sodium hydroxide in dichloromethane (2.0 ml) containing ECF (40 µl), which was converted to the EOC derivative and subsequently the MO derivative by reaction with methoxyamine hydrochloride at 60 °C for 60 min. The aqueous phase was acidified (pH ≤ 2.0 with 10% sulphuric acid), saturated with sodium chloride, and extracted twice with diethyl ether and ethyl acetate (3 ml × 2). The extracts were evaporated to dryness using a gentle nitrogen stream. Dry residues containing OAs, AAs, and FAs were reacted with TEA (5 µl), toluene (20 µl), and MTBSTFA (20 µl) at 60 °C for 30 min to form TBDMS derivatives.

Metabotranscriptomic data analysis

Differences in gene expression were examined using the Affymetrix system (ISTECH, South Korea) in conjunction with the Human U133 Plus 2.0 50K microarray, which contains 54,675 probes. Between-group differences in the data distributions were analysed with GenPlex 3.0 software as previously descried (Shim et al. 2012) and probe signals were quantile normalised. Profiles of OAs, AAs, and FAs were imported from GC-MS/MS data. Biological pathways and functions were identified using web-based bioinformatics software (IPA; Ingenuity Systems, USA). We used ± 3.0-fold change for genes and ± 1.2-fold change for metabolites as cut-offs to generate data sets of genes and metabolites that were significantly different between the untreated control and MNPs@SiO2(RITC)-treated groups.

Determination of glucose uptake efficiency and concentration in cells and media

Uptake efficiency and amounts of glucose were quantified using a luciferase-based kit in accordance with manufacturer specifications (Promega, USA). Briefly, cells were treated with 0.1 or 1.0 µg/µl MNPs@SiO2(RITC) for 12 h in a 384-well white plate (Corning, CA). After incubation, cells were washed twice with phosphate buffered saline (PBS). For the determination of glucose uptake, cells were incubated with 1 mM 2-deoxyglucose (2DG) for 10 min. The uptake of 2DG was stopped with an acid detergent solution (Stop buffer, described in the manufacturer’s protocol) and cells were subjected to lysis. The pH was neutralised with a high-pH buffer solution (Neutralization buffer, described in the manufacturer’s protocol) for enzymatic reaction. Then, lysates were mixed with G6PDH, nicotinamide adenine dinucleotide phosphate (NADP+), reductase for proluciferin, ATP, and luciferase. G6PDH oxidized 2DG6P to 6-phosphodeoxygluconate and simultaneously reduced NADP+ to NADPH, and the reductase converted proluciferin to luciferin using NADPH. Thus, the assay end-point was luciferase luminescence. The end-point luminescence was recorded using a 0.3-1 s integration on a Synergy 2 luminometer (BioTek, CA) and subsequently captured using a ChemiDoc™ Touch Gel Imaging System (Bio-Rad).

To visualize the efficiency of glucose uptake into cells, they were treated with 0.1 or 1.0 µg/µl MNPs@SiO2(RITC) for 12 h in cover slips. After treatment, cells were treated with fluorescent d-glucose analogue 2-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]-2-deoxy-d-glucose (2-NBDG) at 37 °C for 30 min and the fluorescence associated with 2-deoxy-d-glucose was observed by fluorescence microscopy (Axiovert 200M, Carl Zeiss, Jena, Germany).

To determine the amount of glucose in cells, cells were treated with 0.1 or 1.0 µg/µl MNPs@SiO2(RITC) and subsequently washed twice with PBS. Cellular activities were stopped with a stop buffer and neutralised with neutralization buffer. The lysates were mixed with glucose dehydrogenase, NADP+, reductase for proluciferin, ATP, and luciferase. Glucose in the media was measured after 400-fold dilution with PBS. Glucose dehydrogenase oxidized glucose to glucono-1,5-lactone and simultaneously reduced NADP+ to NADPH, and the reductase converted proluciferin to luciferin using NADPH. The produced luciferin was oxidized by luciferase, generating luminescent. End-point luminescence was recorded using a 0.3–1-s integration on a Synergy 2 luminometer (BioTek, CA) and subsequently captured using a ChemiDoc™ Touch Gel Imaging System (Bio-Rad).

RNA isolation and quantitative real-time PCR

Total RNA was isolated from cells using RNAzol B (Tel-Test, USA). Briefly, cells treated with 0.1 or 1.0 µg/µl MNPs@SiO2(RITC) for 12 h and untreated control cells (2 × 106 cells) were harvested in RNAzol B solution followed by the addition of chloroform and incubation for 5 min on ice. Cells were then treated with isopropyl alcohol to precipitate total RNA. Pellets were washed in 70% ethanol followed by air drying, and total RNA was dissolved with RNase-free water. RNA purity was determined with optical density values of 1.8–2.0 at wavelength ratios of 260/230 and 260/280 using spectrophotometry (Eppendorf, Hamburg, Germany). A cDNA library was synthesised using the iScript Advanced cDNA Synthesis Kit (Bio-Rad). The reaction conditions were as follows: 46 °C for 20 min followed by 95 °C for 1 min.

The expression levels of metabotranscriptomic network-related genes were detected by qPCR using the SsoAdvanced™ Universal SYBR® Green Supermix real-time PCR kit (Bio-Rad) with gene-specific primer pairs (Supplementary Table 1) on a Rotor Gene-Q system (Qiagen, CA). The reaction conditions were as follows: 95 °C for 5 min followed by 50 cycles of 95 °C for 5 s and 60 °C for 30 s. The threshold/quantification cycle (Ct/Cq) value was determined at the point where the detected fluorescence was statistically higher than the background level. PCR products were analysed based on a melting curve constructed using Rotor-Gene 1.7 software (Qiagen). PCR was prepared as independent triplicate samples. The relative quantification of target gene expression was performed using the 2−ΔΔCt method.

Statistical analysis

The results were analysed with a one-way analysis of variance (ANOVA) and Tukey’s honestly significant difference (HSD) post hoc tests using IBM-SPSS software (IBM Corp., USA). Differences were considered to be statistically significant when p < 0.05. The composition levels of AA, OA, and FA in each group were compared using a supervised multivariate PCA using R statistical software (https://www.R-project.org).

Results

Metabolic disturbance in MNPs@SiO2(RITC) treated cells analysed by GC-MS/MS

Metabolic changes in AAs, OAs, and polyamines in MNPs@SiO2(RITC)-treated cells are significantly associated with the generation of ROS (Phukan et al. 2016; Shim et al. 2012). In this study, we performed profiling analyses of 14 OAs, 20 AAs, and 13 FAs using EOC/MO/TBDMS derivatisation and GC-MS/MS analyses of cells treated with 0.1 or 1.0 µg/µl MNPs@SiO2(RITC) (Supplementary Table 2). Eleven of 14 OAs were positively detected at baseline. Seven OAs, 8 AAs, and 6 FAs were significantly affected by treatment with 1.0 µg/µl MNPs@SiO2(RITC), and 5 OAs, 6 AAs, and 4 FAs were significantly affected by treatment with 0.1 µg/µl MNPs@SiO2(RITC) compared to an untreated control condition.

Metabolite ratio values were calculated as the change in metabolite level in MNPs@SiO2(RITC)-treated cells divided to the mean corresponding control value. Moreover, we performed a principal component analysis (PCA) of all 44 metabolites in the untreated control, 0.1, and 1.0 µg/µl MNPs@SiO2(RITC) conditions and used 3 principal components, i.e., PC1, PC2, and PC3, to construct three-dimensional score plots (Supplementary Fig. 1). Three distinct clusters and 3 PCs accounted for 83.2% of the total variance in the raw data. Untreated control cell clusters were visibly different from those of MNPs@SiO2(RITC)-treated cells, and this difference was most evident in a comparison of the untreated control and 1.0 µg/µl MNPs@SiO2(RITC)-treated cell clusters.

A visual star symbol plot was drawn using the metabolite ratio values and rays of the plot based on Supplementary Table 2 (Fig. 1). Star shapes representing the MNPs@SiO2(RITC) treatment groups were readily distinguishable from that of the untreated control group mean. With regard to OA profiles, 1.0 µg/µl MNPs@SiO2(RITC) treatment was associated with decreased levels of 3-hydroxybutyric, acetoacetic, succinic, oxaloacetic, 2-hydroxyglutaric, and citric acids and increased levels of pyruvic acid (Fig. 1a). Similarly, 0.1 µg/µl MNPs@SiO2(RITC) treatment was associated with decreased levels of pyruvic, acetoacetic, oxaloacetic, and citric acids and increased level of malic acid. Common differences were more pronounced in the 1.0 µg/µl MNPs@SiO2(RITC) treatment group.

Fig. 1
figure 1

Metabolic disturbance in MNPs@SiO2(RITC)-treated HEK293 cells. Star patterns for 11 organic acids (OAs; a), 20 amino acids (AAs; b), and 13 fatty acids (FAs; c) in HEK293 cells treated with 0.1 or 1.0 µg/µl MNPs@SiO2(RITC) and untreated control cells. *p < 0.05 untreated control vs. 0.1 µg/µl MNPs@SiO2(RITC), §p < 0.05 untreated control vs. 1.0 µg/µl MNPs@SiO2(RITC), and #p < 0.05 0.1 µg/µl vs. 1.0 µg/µl MNPs@SiO2(RITC)

Full size image

With regard to AA profiles, 1.0 µg/µl MNPs@SiO2(RITC) treatment was associated with decreased levels of glycine, serine, cysteine, and glutamine and increased levels of leucine, glutamic acid, lysine, and tyrosine (Fig. 1b). Similarly, 0.1 µg/µl MNPs@SiO2(RITC) treatment was associated with decreased levels of glutamine and tryptophan and increased levels of leucine, serine, glutamic acid, and lysine. Common differences were more pronounced in the 1.0 µg/µl MNPs@SiO2(RITC) treatment group.

With regard to FA profiles, 1.0 µg/µl MNPs@SiO2(RITC) treatment was associated with decreased levels of myristoleic, palmitoleic, and linoleic acids and increased levels of arachidonic, behenic, and lignoceric acids (Fig. 1c). Similarly, 0.1 µg/µl MNPs@SiO2(RITC) treatment was associated with decreased levels of myristoleic, palmitoleic, and linoleic acids and increased levels of arachidonic acid. Changes were similar between the 0.1 and 1.0 µg/µl MNPs@SiO2(RITC) groups and more pronounced in the 1.0 µg/µl MNPs@SiO2(RITC) group, except for palmitoleic acid.

Metabolomic network associated with ROS generation and glucose metabolic dysfunction after MNPs@SiO2(RITC) treatment

Based on the metabolic profiles, we generated a metabolomics network using a bioinformatics tool, the ingenuity pathway analysis (IPA). Considering the physiological state and most significant changes in metabolic profiles after MNPs@SiO2(RITC) treatment, we postulated that a ± 1.2-fold change would be an adequate cut-off to reflect abnormal conditions. There was a clear network of metabolic relationships among OAs, AAs, and FAs in the 1.0 µg/µl MNPs@SiO2(RITC)-treated group compared to the 0.1 µg/µl MNPs@SiO2(RITC)-treated and untreated control groups (Fig. 2 and Supplementary Fig. 2). Biological changes including ROS generation and glucose metabolic dysfunction were deduced from the metabolic profile. Increased ROS generation and glucose metabolic dysfunction were predicted after treatment with 1.0 µg/µl MNPs@SiO2(RITC) (Supplementary Fig. 3).

Fig. 2
figure 2

Metabolomic network of 1.0 µg/µl MNPs@SiO2(RITC)-treated HEK293 cells. The analysis employed a fold-change cut-off value of ± 1.2. Red and green areas indicate metabolite concentrations that were increased and decreased compared to the untreated control group, respectively. (Colour figure online)

Full size image

Representative selected-ion monitoring (SIM) chromatograms of 3 influential OAs (pyruvic acid, acetoacetic acid, and succinic acid), 4 influential AAs (glycine, cysteine, aspartic acid, and glutamic acid), and 2 influential FAs (linoleic acid and arachidonic acid) in the metabolomic network are shown in Fig. 3.

Fig. 3
figure 3

Representative selected-ion monitoring chromatograms of influential metabolites. a Three organic acids (OAs; pyruvic acid, acetoacetic acid, and succinic acid), b 4 amino acids (AAs; glycine, cysteine, aspartic acid, and glutamic acid), and c 2 fatty acids (FAs; linoleic acid and arachidonic acid) in the metabolomic network. IS, internal standard (3,4-dimethoxybenzoic acid for OAs, norvaline for AAs, and pentadecanoic acid for FAs)

Full size image

Generation of the metabotranscriptomic network and evaluation of glucose homeostasis in MNPs@SiO2(RITC)-treated cells

The proposed metabolic network and subsequent predictions were only based on an end-point phenotype with limited capacity to reflect actual biological processes (Evans 2015; Rehrauer et al. 2013). To compensate for potential weakness in the proposed metabolomic network, we generated a transcriptomic network associated with ROS generation and glucose metabolic dysfunction and integrated this information to yield a metabotranscriptomic network (Fig. 4a, Supplementary Fig. 4, and Supplementary Table 3). The transcriptomic network predicted increased ROS generation and glucose metabolic dysfunction after treatment with 1.0 µg/µl MNPs@SiO2(RITC) (Supplementary Fig. 5). In the 1.0 µg/µl MNPs@SiO2(RITC)-treated group, expression levels of 36 genes were changed compared to control cells; 21 genes were upregulated and 15 genes were downregulated. In the 0.1 µg/µl MNPs@SiO2(RITC)-treated group, only 14 genes were upregulated and 7 genes were downregulated (Fig. 4b).

Fig. 4
figure 4

Transcriptomic disturbance in 1.0 µg/µl MNPs@SiO2(RITC)-treated HEK293 cells. a Functional analysis of the transcriptomic network in 1.0 µg/µl MNPs@SiO2(RITC)-treated HEK293 cells. The analysis used a fold-change cut-off value of ± 3. b Heat map of 36 genes related to reactive oxygen species (ROS) generation and glucose metabolic dysfunction with altered expression in a microarray analysis. Red and green areas indicate genes that were upregulated and downregulated compared to the untreated control group, respectively. (Colour figure online)

Full size image

Next, we integrated the metabolomic and transcriptomic networks based on biological functions, ROS generation, and glucose metabolic dysfunction. The metabolome and transcriptome were interconnected with strong relationships between biological functions in the 0.1 and 1.0 µg/µl MNPs@SiO2(RITC)-treated groups (Fig. 5a and Supplementary Fig. 6). The metabotranscriptomic network predicted increased ROS generation and glucose metabolic dysfunction in a manner that was more pronounced than in any single omics network (Supplementary Fig. 7). Thus, we hypothesised that treatment with 1.0 µg/µl MNPs@SiO2(RITC) would alter glucose metabolism in vitro. In an experimental analysis, treatment with 1.0 µg/µl MNPs@SiO2(RITC) decreased the efficiency of glucose uptake by 40% compared to treatment with 0.1 µg/µl MNPs@SiO2(RITC) or untreated control cells (Fig. 5b). Similarly, intracellular glucose decreased by 30% after treatment with 1.0 µg/µl MNPs@SiO2(RITC) compared to treatment with 0.1 µg/µl MNPs@SiO2(RITC) or untreated control cells (Fig. 5c). However, there were no changes in the residual amount of glucose in media (Fig. 5d).

Fig. 5
figure 5

Metabotranscriptomic network, evaluation of glucose uptake efficiency, and determination of glucose concentration in MNPs@SiO2(RITC)-treated cells. a The analysis used fold-change cut-off values of ± 1.2 for metabolites and ± 3 for genes. Red and green areas indicate upregulated and downregulated factors compared to the untreated control group, respectively. Glucose uptake efficiency (b) and glucose determination in MNPs@SiO2(RITC)-treated cells (c) and media (d) were determined on luminescent images. Data represent the mean ± standard deviation of 3 independent experiments. *p < 0.05 vs. untreated control. (Colour figure online)

Full size image

Glucose uptake-related metabolites and genes in the metabotranscriptomic network

Given the observation that reductions in the amount of intracellular glucose were caused by decreased uptake efficiency, we reanalysed the proposed metabotranscriptomic network with the cellular function “glucose uptake”. This query identified 11 directly related factors: 3 AAs and 8 genes (Fig. 6a and Supplementary Fig. 8). The network predicted high suppression of glucose uptake as a result of up- and downregulated metabolites and genes after treatment with 1.0 µg/µl MNPs@SiO2(RITC) (Supplementary Fig. 9). The metabolites cysteine, lysine, and tyrosine were more closely related to glucose uptake than others.

Fig. 6
figure 6

Functional analysis of the metabotranscriptomic network for glucose uptake. a A second metabotranscriptomic network analysis was performed for glucose uptake. The analysis used fold-change cut-off values of ± 1.2 for metabolites and ± 3 for genes. Red and green areas indicate upregulated and downregulated factors compared to the untreated control group, respectively. Factors directly related to glucose uptake are highlighted in magenta. b A qPCR analysis was performed to determine glucose uptake-related gene expression in each group using GAPDH as an internal control. Data represent the mean ± standard deviation of 3 independent experiments. *p < 0.05 vs. untreated control and #p < 0.05 vs. 0.1 µg/µl MNPs@SiO2(RITC). c A visual analysis for glucose uptake efficiency of cells was performed using 2-NBDG. Scale bar = 20 µm. (Colour figure online)

Full size image

Based on the prediction results, we quantified changes in the expression of 4 genes using quantitative real-time PCR (Fig. 6b). Gene expression of HNF1 homeobox A (HNF1A) was significantly downregulated in 1.0 µg/µl MNPs@SiO2(RITC)-treated cells compared to 0.1 µg/µl MNPs@SiO2(RITC)-treated and untreated control cells. Alternatively, the gene expression levels of nuclear receptor subfamily 4 group A member 1 (NR4A1), protein kinase C alpha (PRKCA), and solute carrier family 2 member 2 (SLC2A2) significantly increased in 1.0 µg/µl MNPs@SiO2(RITC)-treated cells compared to 0.1 µg/µl MNPs@SiO2(RITC)-treated and untreated control cells.

To analyse the efficiency of glucose uptake visually, we used fluorescent d-glucose analogue 2-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]-2-deoxy-d-glucose (2-NBDG) as a tracer (Yamada et al. 2000). Since 2-NBDG uptake is mediated through glucose transporter, its uptake has been used for monitoring glucose uptake into cells. We found that glucose uptake efficiency of cells was decreased in 1.0 µg/µl MNPs@SiO2(RITC)-treated cells compared to 0.1 µg/µl MNPs@SiO2(RITC)-treated and untreated control cells (Fig. 6c).

Discussion

The present study used an integrated omics-based approach to investigate NP-induced cellular and glucose metabolic dysfunction. Specifically, we investigated the metabolome and transcriptome in NP-treated cells to identify important mechanisms of nanotoxicity and performed a functional metabotranscriptomics analysis to deduce the relationship between ROS generation and glucose metabolic dysfunction after NP exposure. Additionally, we demonstrated that “overdose” (high-concentration) treatment with MNPs@SiO2(RITC) deregulated glucose metabolism by preventing glucose uptake. Our results indicate that minimizing NP usage or exposure is important for preventing nanotoxicity related to glucose metabolic dysfunction and transport.

In the present study, we simultaneously profiled OAs, AAs, and FAs as EOC/MO/TBDMS derivatives using GC-MS/MS, which has better sensitivity and accuracy than GC-MS used in previous studies (Phukan et al. 2016; Shim et al. 2012). This allowed shorter analytical times and the simultaneous analysis of 44 metabolites in a single run, which was useful for the rapid and exact monitoring of metabolic alterations after MNPs@SiO2(RITC) treatment.

Changes in cell metabolic profiles can reflect the impairment of redox status, energy metabolism, and biogenesis (Johnson et al. 2016). Previous studies suggest that MNPs@SiO2(RITC) exposure induces ROS generation and mitochondrial damage as well as related metabolic disturbances in OAs, AAs, and polyamine (Phukan et al. 2016; Shim et al. 2012). OA profiles are more vulnerable to ROS and mitochondrial damage than other metabolites because the Krebs cycle occurs in mitochondria and utilises redox reactions for electron transfer and the generation of adenosine triphosphate (ATP) (Mailloux et al. 2014). We previously reported that the metabolic disturbance of OAs contributed to the accumulation of glutamic acid in MNPs@SiO2(RITC)-treated cells (Shim et al. 2012). In this study, we found that pyruvate showed the largest alterations among all OAs after MNPs@SiO2(RITC) exposure; importantly, pyruvate is the final product of glycolysis in the cytosol. In subsequent steps of energy metabolism, pyruvate is transported into mitochondria via pyruvate translocase to interact with pyruvate dehydrogenase, which converts pyruvate into acetyl-CoA on the inner membrane of the mitochondrial matrix (Stacpoole et al. 2003; Sutendra et al. 2014). We previously demonstrated structural disintegration and depolarization in mitochondria following MNPs@SiO2(RITC) treatment (Shim et al. 2012). These mitochondrial effects may contribute to the accumulation of pyruvate and thus glucose metabolic dysfunction in MNPs@SiO2(RITC)-treated cells.

Changes in AA profiles can signal homeostatic deregulation (Chaveroux et al. 2009). We previously demonstrated that increases in glutamic acid after MNPs@SiO2(RITC) treatment were caused by changes in metabolic pathway-related genes and associated OA profiles. Additionally, free AAs such as tryptophan, tyrosine, histidine, and cysteine are directly oxidized by ROS (Droge 2002). Glycine attenuates superoxide anion radical release in the presence of nicotinamide adenine dinucleotide phosphate and decreases protein carbonyl and lipid peroxidation by promoting levels of glutathione synthetase and, consequently, glutathione (Ruiz-Ramirez et al. 2014). ROS generation is increased by the accumulation of basic AAs such as arginine, ornithine, and lysine in the mitochondrial membrane and induces mitochondria-dependent cell death via aberrant ubiquitination (Braun et al. 2015). In this study, we found that 1.0 µg/µl MNPs@SiO2(RITC) significantly increased levels of leucine, glutamic acid, lysine, and tyrosine and decreased those of glycine, serine, cysteine, and glutamine in HEK293 cells. These changes may be closely related to MNPs@SiO2(RITC)-induced ROS generation.

Our study had a limited ability to assess changes in FAs, although some MNPs@SiO2(RITC)-associated changes were detected compared to the control group. FA levels are highly related to ROS generation and have important roles in cellular function. ROS can oxidize polyunsaturated phospholipids, glycolipids, and cholesterol membrane components to impair plasma membrane function (Girotti 1998). Malondialdehyde, acrolein, and 4-hydroxynonenal are generated from the oxidative cleavage of polyunsaturated phospholipids (Negre-Salvayre et al. 2008) and deplete unsaturated phospholipids and cholesterol. In this study, we were unable to identify consistent changes in gene patterns due to the limitations of our bioinformatics database.

In the present study, we made deductions and connections based on the datasets of differentially expressed genes and metabolites to generate a metabotranscriptomic network. Cysteine and lysine have been associated with glucose transporter protein expression (Gazit et al. 2003; Katsumata 2011), and tyrosine reversibly inhibits glucose transporter function (Widmer et al. 1990). Moreover, ROS directly oxidize cysteine to generate cystine, a dimer form of cysteine (Droge 2002). In our study, the contribution of decreased cysteine availability to the metabotranscriptomic network predictions in the 1.0 µg/µl MNPs@SiO2(RITC) treatment condition was potentially related to its involvement in glucose uptake and the sensitivity of cysteine to ROS generation. In the gene expression analysis, HNF1A, a transcriptional regulator of SLC2A2 (Bae et al. 2010), was significantly downregulated after treatment with 1.0 µg/µl MNPs@SiO2(RITC). Additionally, there were significant increases in the expression of NR4A1, which controls the expression of key metabolic genes for glucose transportation, insulin signalling, glycolysis, and glycogenolysis (Corrocher et al. 2017). PRKCA is a kinase involved in glucose transporter activation (Lee et al. 2015) and SLC2A2, also known as glucose transporter 2 (GLUT2), is a bidirectional glucose transporter that is responsible for the bulk of glucose transportation in the cell (Mueckler and Thorens 2013). SLC2A2 and its regulator gene, NR4A1, are transcriptionally and translationally sensitive to oxidative stress (Lee et al. 2014; Shimizu et al. 2015; Shirakawa and Terauchi 2014). We presume that these disturbances in metabolic profiles and the transcriptome converge to explain increased ROS generation and glucose metabolic dysfunction after MNPs@SiO2(RITC) exposure in vitro.

Even though kidneys are known to particularly susceptible to nanotoxicity due to nanoparticles accumulation and these are one of the main elimination routes of nanoparticles in vivo (Iavicoli et al. 2016), our investigation is limited in kidney-derived cell lines treated with nanoparticle. Therefore, further studies are needed concerning the effect of nanoparticles on primary kidney cells and in vivo before biomedical applications.

In conclusion, our findings indicate that exposure to high concentrations of NPs can be deleterious to glucose metabolism and cellular glucose uptake. These results highlight important safety considerations for nanotoxicity when using NPs for therapeutic or diagnostic purposes.