Metabolomic effects of CeO2, SiO2 and CuO metal oxide nanomaterials on HepG2 cells

To better assess potential hepatotoxicity of nanomaterials, human liver HepG2 cells were exposed for 3 days to five different CeO2 (either 30 or 100 μg/ml), 3 SiO2 based (30 μg/ml) or 1 CuO (3 μg/ml) nanomaterials with dry primary particle sizes ranging from 15 to 213 nm. Metabolomic assessment of exposed cells was then performed using four mass spectroscopy dependent platforms (LC and GC), finding 344 biochemicals. Four CeO2, 1 SiO2 and 1 CuO nanomaterials increased hepatocyte concentrations of many lipids, particularly free fatty acids and monoacylglycerols but only CuO elevated lysolipids and sphingolipids. In respect to structure-activity, we now know that five out of six tested CeO2, and both SiO2 and CuO, but zero out of four TiO2 nanomaterials have caused this elevated lipids effect in HepG2 cells. Observed decreases in UDP-glucuronate (by CeO2) and S-adenosylmethionine (by CeO2 and CuO) and increased S-adenosylhomocysteine (by CuO and some CeO2) suggest that a nanomaterial exposure increases transmethylation reactions and depletes hepatic methylation and glucuronidation capacity. Our metabolomics data suggests increased free radical attack on nucleotides. There was a clear pattern of nanomaterial-induced decreased nucleotide concentrations coupled with increased concentrations of nucleic acid degradation products. Purine and pyrimidine alterations included concentration increases for hypoxanthine, xanthine, allantoin, urate, inosine, adenosine 3′,5′-diphosphate, cytidine and thymidine while decreases were seen for uridine 5′-diphosphate, UDP-glucuronate, uridine 5′-monophosphate, adenosine 5′-diphosphate, adenosine 5′-monophophate, cytidine 5′-monophosphate and cytidine 3′-monophosphate. Observed depletions of both 6-phosphogluconate, NADPH and NADH (all by CeO2) suggest that the HepG2 cells may be deficient in reducing equivalents and thus in a state of oxidative stress. Metal oxide nanomaterial exposure may compromise the methylation, glucuronidation and reduced glutathione conjugation systems; thus Phase II conjugational capacity of hepatocytes may be decreased. This metabolomics study of the effects of nine different nanomaterials has not only confirmed some observations of the prior 2014 study (lipid elevations caused by one CeO2 nanomaterial) but also found some entirely new effects (both SiO2 and CuO nanomaterials also increased the concentrations of several lipid classes, nanomaterial induced decreases in S-adenosylmethionine, UDP-glucuronate, dipeptides, 6-phosphogluconate, NADPH and NADH).


Background
Metal oxide nanomaterials have many uses including: coatings, grinding, ceramics, catalysis, electronics, biomedical, energy and fuel additives (for CeO 2 ); biocides, sensor applications, catalysis and electronics (for CuO); and additives for rubber and plastics, composites for concrete and other construction materials and biomedical applications such as drug delivery and theranostics (for SiO 2 ). It is difficult to evaluate nanomaterials to determine their degree and type of toxicity [1]. For nanomaterials a major determinant of their biological action may be their surface properties, particularly their ability to donate or accept electrons [2] and/or to generate free radicals and to form reactive oxygen species (ROS) [3].
After the development of the genomics and proteomics technologies, metabolomics has more recently been developed and used as an analytical tool in general biological research [4] and toxicological studies (Kitchin et al. [5]). The analytical platforms most commonly used to determine cellular metabolites are liquid chromatography tandem mass spectroscopy (LC-MS/MS), LC-MS/ MS with hydrophilic interaction liquid chromatography (HILIC), gas chromatography-mass spectroscopy (GC-MS) and nuclear magnetic resonance (NMR). Metabolomics offers environmental and toxicological researchers the opportunity to determine the concentrations of many important cellular biochemicals in one experiment and provide complimentary information to traditional toxicological tests and other modern 'omics approaches to biological questions.
In the nanotoxicology world, functional assays have recently been proposed as a way to better predict and connect the physical-chemical properties of nanomaterials and their potential adverse health outcomes [6]. Metabolomics based determinations of the altered concentrations of many important cellular biochemicals offer many good possible functional assays as intermediates in the long causal chain between physical-chemical properties of nanomaterials and eventual toxicity.
This study partnered with the Metabolon Inc. (Durham, NC) which used four analytical platforms to measure as many HepG2 (human liver) metabolites as possibleliquid chromatography-tandem mass spectroscopy with positive ionization (LC-MS/MS+), liquid chromatography-tandem mass spectroscopy with negative ionization (LC-MS/MS−), HILIC LC-MS/MS with negative ionization and gas chromatography mass spectroscopy (GC-MS) (with positive ionization via electron impact ionization). With metabolomics tools such as these, cellular biochemicals from different metabolic classes can be determinedlipids, energy molecules, amino acids, peptides, carbohydrates, purines, pyrimidines and nucleotides etc. A prior metabolomics study had discovered several interesting biochemical changes in TiO 2 and CeO 2 exposed HepG2 cellsa large number of lipid increases, particularly of fatty acids and many decreases in glutathione-related biochemicals and increased asymmetric dimethylarginine by two CeO 2 nanomaterials [5]. Because of strong interest in the prior CeO 2 nanomaterial induced effects, five new CeO 2 nanomaterials were selected for the current study (labelled W4, X5, Y6, Z7 and Q) ( Table 1). CeO 2 based materials offer the possibility of Ce +4 <-> Ce +3 redox cycling [7] and the generation of ROS. Additionally, atomic layer deposition (ALD) using tris(isopropylcyclopentadienyl)cerium was attempted in an effort to produce a CeO 2 coated SiO 2 nanoparticle with a large amount of Ce +3 on the surface (nanomaterials labelled SiO 2 K1 and SiO 2 N2). Finally, a CuO nanomaterial was included because of interest in the toxicity of soluble copper ions and the oxidative stress theory of nanomaterial toxicity (all treatment nanomaterials are summarized in Table 1).
In vitro toxicity testing allows us to link molecular, biochemical and cellular functions to physicochemical properties of nanomaterials, adverse biological outcomes and better predict risk. The specific major goals of this metabolomics study was to replicate and/or further explore: 1) the findings of lipid elevations (e. g. fatty acids) caused by one CeO 2 nanomaterial, 2) the depletion of glutathione and gamma-glutamyl amino acids by several metal oxide nanomaterials (both CeO 2 and TiO 2 ), 3) elevations in asymmetric dimethylarginine found with 2 CeO 2 nanomaterials and 4) to explore the metabolomics effects of two new metal oxide nanomaterials based on SiO 2 and CuO and 5) to discover possible functional assays. Overall, functional assays can link individual experimental data with proposed mechanisms of action to inform adverse outcome pathway model development in support of regulatory decisions.
To assess potential hepatotoxicity issues from oral and/or inhalation exposure routes, 72 h exposures were conducted in human liver HepG2 cells. Thus, human liver HepG2 cells were exposed for 3 days to five different CeO 2 (either 30 or 100 μg/ml), 3 SiO 2 based (30 μg/ ml) or 1 CuO (3 μg/ml) nanomaterials with dry primary particle sizes ranging from 15 to 213 nm. Nanomaterialexposed cells were examined for their ability to cause cellular toxicity and effects on the concentrations of cellular metabolites in HepG2 cells (Table 1, from 15 to 213 nm dry size). In our study 344 cellular metabolites were found and relatively quantified. This metabolomics study included sufficient biochemicals to examine the biochemical components of several major cellular systemslipid homeostasis, cellular energetics, hepatic conjugation and excretion, urea cycle, polyamines, purines and pyrimidines. These metabolomics experimental results are discussed in the context of systems biology and the toxicology of nanomaterials.

Nanomaterials and their characterization and dispersion via ultrasound
The nine nanomaterials used in this study (Table 1) were selected to further determine the biological properties of various forms of CeO 2 nanomaterials as well as some other metal oxide based nanomaterials (SiO 2 and CuO). These nine nanomaterials are being used by three research laboratories at the US EPA in a coordinated research effort with many different scientific disciplines and experimental techniques.
Physical-chemical characterization of these nanomaterials was conducted by a variety of techniques for dry primary particle size, range of particle size, surface area and percent purity mostly by their manufacturer ( Table 1). The nanomaterials were obtained from six different vendors (Alfa Aesar, Aldrich, Sigma, Nanoxides, US Research Nanomaterials and Nanostructured and Amorphous Materials). When given, the chemical purity was high (>99.5%). The For dispersion prior to cell culture, measured amounts of bovine serum albumin (BSA, Sigma-Aldrich, product A7906) at 200 mg/ml and phosphate buffered saline (PBS) were added to the dry nanomaterials in a glass vial. The general protein coating recipe of Dale Porter [8] was followed with the mass ratio of the nanomaterial to BSA of 1/0.6. For example, in preparation of CeO 2 "Z7" for study, 16.04 mg nanomaterial CeO 2 Z7, 9.624 mg BSA and 4.95 ml of PBS were combined. Sonication occurred at a nanomaterial concentration of 3.21 mg/ml and 5.0 ml of volume. Sonication was done at room temperature with a S-4000 Misonix Ultrasonic Liquid Processor with a 2.5 in. cup horn (part #431-A, Farmington, NY) for two 10 min cycles of 13 s on, 7 s off with a total power of about 131 watts and a total energy of 166,120 joules. Excess unbound albumin was removed by pelleting (9300 × g for 5 min) the nanomaterials and resuspending them in cell culture media without any sonication of the cell culture media.
After nanomaterial dispersion, the degree of agglomeration was determined by dynamic light scattering at 35 o C at each treatment concentration used for metabolomics study and sometimes one lower concentration. Size and zeta potential measurements were made both just after sonication and 72 h later at the end of treatment period with a Malvern Model Zen3600 Zetasizer (data in Additional file 1: Table S1).

Chemicals and cell culture methods
The chemicals and suppliers used in this study were: BSA (Sigma) and fetal bovine serum, GlutaMAX™, sodium pyruvate, fetal bovine serum, Dulbecco's Phosphate-Buffered Saline and phosphate buffered saline (all from Invitrogen). Human Hepatocellular Carcinoma Cells, designation HepG2 (ATCC catalog number HB-8065), were obtained and expanded through passage seven using Basal Medium Eagle (Gibco) containing 2 mM GlutaMAX™, 1 mM sodium pyruvate and 10% fetal bovine serum and then frozen in liquid nitrogen. This combined cell culture media is called Eagle's mimimum essential medium (EMEM). Cells were subsequently carefully thawed and expanded before experimentation at passages 10 and 11. Cultures were maintained in a humidified incubator at 37°C and 95% air/ 5% CO 2 during the study. Cells were plated at 80,000 cells/ cm 2 in vented T-25 flasks (Corning) for 48 h prior to nanomaterial exposure. After sonifcation, centrifugation and resuspension, working stocks of each nanomaterial were prepared at 1.0 mg per mL and diluted using culture medium. Individual flasks were dosed with 200 uL per cm 2 of the appropriate nanomaterial dilution to achieve either 100 μg/ml (CeO 2 Q), 30 μg/ml (7 other nanomaterials) or 3 μg/ml (CuO) exposure concentrations. Cultures were then incubated for 72 h prior to harvesting. At 72 h, the media was vacuum aspirated and the flasks rinsed with warm Dulbecco's Phosphate-Buffered Saline (DPBS). The DPBS was aspirated and cells were scraped free of the flask and collected in labeled 15 mL tubes using 1 mL of warm DPBS by micropipette. The cells were then centrifuged at room temperature at 100 × g for 5 min. The supernatant was carefully removed via vacuum aspiration and the cellular pellet was flash frozen on dry ice before transfer to −80 o C freezer for storage prior to metabolomic analysis.

Cytotoxicity assays and kits
Many common cytotoxicity assays [MTT (3- , alamar blue (resazurin), neutral red (3-amino-7-dimethylamino-2 methylphenazine hydrochloride), ATP and simple visual examination of the cells] have been used by our laboratory seeking to avoid or minimize interferences from the nanomaterials themselves. After 72 h of culture with various nanomaterials, cytotoxicity assays based on MTT (Sigma-Aldrich, St Louis, MO), MTS (Promega, Madison, WI) and alamar blue (Cell Tier-Blue, Promega, Madison, WI) were performed in accordance with the enclosed kit directions. Alamar blue and MTS were used for all nanomaterial cytotoxicity experiments except for CeO 2 Q (MTT only was used). A PerkinElmer 1420 Multilabel Counter Victor 3 V plate reader was used for all cytotoxicity assays. Cytotoxicity assays results were always checked with each other and versus visual assessment of the cells to ensure the cytotoxicity assays were functioning properly.

Study design
For metabolomics study, three different exposure concentrations (3, 30 or 100 μg/ml) were used for the nanomaterials. Only CuO at 3 μg/ml and CeO 2 Q at 100 μg/ ml were not run at 30 μg/ml. The intent was (a) to give approximately equally cytotoxic concentrations of the nine different nanomaterials and (b) if feasible to compare CeO 2 nanomaterials at 30 μg/ml for better comparison to a prior study of our group that used this exposure dose for two prior CeO 2 nanomaterials [5]. The number of samples per group is either five for treatments or six for controls. Two different days were used for HepG2 culturing. On day 1 most of the CeO 2 (W4, X5, Z7 and Q) and the CuO treatment groups were run. On day 2 nanomaterials J0, K1 and N2 (the 3 SiO 2 based nanomaterials) and CeO 2 Y6 were run together.

Statistical analysis
Biochemical ion signals were processed by normalization to Bradford protein concentration, log transformation and imputation of missing values, if any, with the minimum observed value for each compound. Biochemicals that were detected in all samples from one or more groups, but not in samples from other groups, were assumed to be near the lower limit of detection in the groups in which they were not detected. In this case, the lowest detected level of these biochemicals was imputed for samples in which that biochemical was not detected. Then, Welch's two-sample t-test was used to identify biochemicals that differed significantly between experimental groups [9]. In modern gene array work, using the False Discovery Rate (FDR) is a common method of controlling false positive (Type I) error rates. Thus, to account for multiple comparisons in this metabolomics testing, false-discovery rates were computed for each comparison via the Q-value method [10]. P values and Q value false discovery rate-values for all statistical comparisons are reported in Additional file 2: Table S2.
Pathways were assigned for each metabolite, allowing examination of overrepresented pathways. The degree of statistical significance presented in this study is both the common P < 0.05 level used if this 0.05 criteria is met by both P and Q statistics and the more lenient standard of 0.10 if both P and Q are <0.10, because this more lenient standard is less likely to miss some true biological effects. Tables 3, 4, 5, 6 and 7 and Additional file 2: Table S2 have color high lighting to graphically display these P < 0.05 and <0.10 significance levels. The text of the paper uses the P < 0.05 level of claimed statistical significance with the P < 0.10 level mentioned only for NADPH.

Results
Dispersion and agglomeration of nanomaterials (size and zeta potential) By dynamic light scattering, these sonicated nanomaterial samples displayed a fairly large hydrodynamic diameter in both water based cell culture media (EMEM with 10% fetal bovine serum) and PBS (Additional file 1: Table S1). In cell culture media the mean sizes by peak intensity ranged between 154 to 540 nm for CeO 2 , 312 to 554 nm for SiO 2 and 148 to 188 nm for CuO (Additional file 1: Table S1). These hydrodynamic sizes are much larger than the dry primary particle sizes of 15, 22.5, 25, 200 and 213 nm for the five different forms of CeO 2 studied. In cell culture media the mean zeta potentials ranged between −4.4 to −10.3 mV for CeO 2 , −4.7 to −10.5 for CuO and −4.7 to −8.7 for SiO 2 (Additional file 1: Table S1).

The coating of SiO 2 K1 and SiO 2 N2 and ICP-MS results
Our attempt to use atomic layer deposition to put a thin layer of CeO 2 on the J0 SiO 2 based particles failed. By ICP-OES analysis performed at both Missouri University of Science and Technology and the US EPA, almost zero Ce was found in nanomaterials SiO 2 K1 and SiO 2 N2 (Additional file 3: Table S3).

Cytotoxicity results
The exposure concentrations used in this metabolomics study (3,30 or 100 μg/ml) were below concentrations which produced a full degree of cytotoxicity in HepG2 cells via common colorimetric and fluorimetric assays ( Table 2). At the administered dose, no sign of cytotoxicity was observed for CeO 2 W4, CeO 2 X5 and CeO 2 Y6; a low degree of cytotoxicity for CeO 2 Z7, CeO 2 Q, SiO 2 K1 and SiO 2 N2; and a medium degree of cytotoxicity for SiO 2 J0 and CuO ( Table 2).

Altered lipids
In Tables 3, 4, 5, 6 and 7, the displayed numbers are the ratio of the treatment metabolite concentration mean divided by the concurrent control metabolite concentration mean. Increased concentrations of medium and long chain fatty acids, polyunsaturated fatty acid (n3 and n6), fatty acid branched, fatty acid dicarboxylate and monoacylglycerols were observed after treatment with several CeO 2 (W4, X5, Z7 and Q), SiO 2 (J0 only) and CuO nanomaterials (Tables 3 and 4). In this study far fewer increases were noted with fatty acid metabolites, lysolipids, carnitine, inositol metabolites, phospholipid metabolites, phospholipidserine, diacylglycerol and sphingolipid metabolites, showing the selectivity of this lipid effect (Tables 3 and 4). CuO was the only nanomaterial to induce many increases in these classes of less responsive lipids (Tables 3 and 4). The most active lipidelevating nanomaterials were W4, X5, Z7 (all are CeO 2 ), SiO 2 J0 and CuO. CeO 2 Y6 and the two ALD coated SiO 2 based nanoparticles (K1 and N2) did not elevate as many lipid metabolite concentrations. P and Q numbers are tabulated for all 344 biochemicals for every nanomaterial treatment comparison with concurrent controls in Additional file 2: Table S2.

Hepatic conjugation systems (methylation, glucuronidation and glutathione)
Treatment of HepG2 cells with nanoparticles from the day-1 set (CeO 2 X5, CeO 2 Z7, CeO 2 Q and CuO) resulted in declines in S-adenosylmethionine (SAM) and several increases in S-adenosylhomocysteine (SAH) (by CeO 2 X5 and CeO 2 Z7) ( Table 5), though methionine levels were largely unchanged. In the liver methylation capacity is required to support Phase II methylation of xenobiotics to facilitate clearance. The lower SAM levels were accompanied by a sharp decline in serine (by CeO 2 X5, CeO 2 Z7, CeO 2 Q and CuO), in day-1 nanomaterial treated cells. Serine is consumed in the regeneration of methionine from homocysteine, in the one-carbon metabolism pathway. Most of the day-1 nanoparticletreated samples had SAM below the limit of detection, however 5 of 6 day-1 control cell samples had SAM levels above the lower limit of measurement. SAM levels were relatively unchanged with exposure to the day-2 nanoparticles (CeO 2 Y6, SiO 2 J0, SiO 2 K1 and SiO 2 N2) and declines in serine were also limited and not statistically significant.
It is quite surprising that reduced glutathione (GSH) levels were below detection limit in most control and treated samples in this study (some GSH was detected in three of our samples). Similar to prior results with 4 TiO 2 and 2 CeO 2 nanomaterials [5], there were decreases observed in gamma-glutamyl amino acids with several CeO 2 and SiO 2 based nanomaterials (Table 5). Most effected were gamma-glutamylthreonine, gammaglutamylvaline and gamma-glutamylgluatamate. In contrast, the CuO nanomaterial caused large fold increases in four gamma-glutamyl-amino acid compounds -leucine (9.0 fold increase), -isoleucine (10.2), -threonine (7.1) and -valine (9.2) but not -glutamine (0.66) orglutamate (1.07) ( Table 5).

and CuO nanomaterials in HepG2 cells
Both the number and degree of response was considered for each of the eight parameters germane to "cytotoxicity" The eight cytotoxicity parameters are visual microscopic cellular appearance, alamar blue, MTS, cellular protein and microalbumin concentrations and release of lactate dehydrogenase, alanine aminotransferase and aspartate aminotransferase Cellular energetics, reducing capacity and oxidative stress (maltotriose, 6-phosphogluconate, NADPH, NADH and NAD + and dipeptides) Seven out of nine nanomaterial treatment groups (only CeO 2 Y6 and SiO 2 N2 did not) increased maltotriose concentrations ranging from 3.45 to 24.4 fold of concurrent control values. Three increases were above 10 fold increases (13.4 by CeO 2 W4, 14.8 by CeO 2 X5 and 24.4 by CuO). Maltotriose levels can represent a measure of glycogen degradation, from which maltotriose is derived. The first step in conversion of glucose 6-phosphate to 6-phosphogluconate generates NADPH. 6-phosphogluconate was significantly depleted by four of the 5 day-1 set of nanoparticles (Table 6). NADPH concentrations were numerically decreased in all nine nanoparticle treatments (range 0.34 to 0.81) ( Table 6), achieving statistical significance for nanoparticle CeO 2 Y6 at the P < 0.05 level, while the CeO 2 Z7, CeO 2 Q and SiO 2 J0 particles were statistically significant at the lower P < 0.10 level, relative to controls. NADH concentration was significantly decreased (P < 0.05) by CeO 2 Y6 (0.45). No significant elevations were seen for NADH or NADPH. Both nicotinamide (2 decreases) and NAD + were significantly decreased by three nano CeO 2 treatments (CeO 2 Z7, CeO 2 Q and CeO 2 Y6) ( Table 6). Nicotinamide riboside (a NAD + precursor) was significantly elevated in all three cases where NAD + was depleted (CeO 2 Z7, CeO 2 Q and CeO 2 Y6) ( Table 6).
CuO nanomaterial exposure decreased (P < 0.05) the concentrations of all 16 dipeptides ranging from 0.07 to Table 3 Nanomaterial effects on responsive lipids Darker shading (red for increases, green for decreases) means P and Q are both <0.05; Lighter shading means P and Q are both <0. 10 The numbers are the ratio of the treated mean divided by the control mean 0.55 fold change. With the exception of CeO 2 W4, CeO 2 X5 and CeO 2 Z7 induced decreases in the dipeptide glycylleucine, few other dipeptides were decreased by CeO 2 , or SiO 2 based nanomaterials. CuO was also the only nanomaterial that caused a large decrease in the concentration of cysteine (0.07) while elevating cystine (2.26) ( Table 5). This cysteine-cystine redox perturbation suggests oxidative stress caused by CuO exposure.

Cellular effects (urea cycle, polyamines, purine and pyrimidine metabolism, nucleotide sugars)
Several urea cycle, creatinine and polyamine pathway biochemicals were significantly increased by nanomaterial treatment, such as creatine (4 increases), creatinine (5 increases), creatine phosphate (4 increases), putrescine (4 increases) and 5-methylthioadenosine (5 changes with 3 increases) ( Table 7). Levels of putrescine, spermidine Table 4 Nanomaterial effects on less responsive lipids Darker shading (red for increases, green for decreases) means P and Q are both <0.05; Lighter shading means P and Q are both <0. 10 The numbers are the ratio of the treated mean divided by the control mean Table 5 Nanomaterial effects on SAM, SAH, glutathione-related and nucleotide sugar metabolites Darker shading (red for increases, green for decreases) means P and Q are both <0.05; Lighter shading means P and Q are both <0. 10 The numbers are the ratio of the treated mean divided by the control mean Table 6 Nanomaterial effects on maltotriose, 6-phosphogluconate, nicotinamide metabolites and dipeptides Darker shading (red for increases, green for decreases) means P and Q are both <0.05; Lighter shading means P and Q are both <0. 10 The numbers are the ratio of the treated mean divided by the control mean and 5-methylthioadenosine were significantly elevated for many of the CeO 2 nanoparticles in the day-1 set, but these biochemical were not elevated in the day-2 nanomaterials (Table 7). CuO exposure increased putrescine 22.7 fold and N-acetylputrescine 63.3 fold, among the highest elevations observed in this data set. Following CuO exposure, high putrescine concentration (22.7 fold) coupled with low ornithine concentration (0.27 fold) suggest that the enzyme activity of the rate limiting step of polyamine synthesis, ornithine decarboxylase, may have been increased. To a much lesser extent this pattern also occurred with CeO 2 X5 (putrescine (3.58) and ornithine (0.38)) CeO 2 exposures.
However, there were many examples of increased nucleic acid degradation products: inosine (4 changes with 3 increases), hypoxanthine (4 increases), xanthine (5 increases), urate (5 increases) and allantoin (4 changes, 3 increases). Thus, the overall purine and pyrimidine pattern is one of decreased nucleotides and increased nucleic acid degradation products.
In the six component nucleotide sugar biochemical sub pathway (Table 5), all six members of the group showed statistically significant (P < 0.05) decreases in 3 or more of the nine treatment groups (often following CeO 2 Z7, CeO 2 Q, CeO 2 Y6, SiO 2 K1 and SiO 2 N2 exposure). The nucleotide sugars are important in Phase II glucuronidation and glycation reactions. Most active nanomaterials were CeO 2 Z7, CeO 2 Q and CeO 2 Y6; least active were CeO 2 X5, SiO 2 J0, SiO 2 K1, SiO 2 N2 and CuO. There is a major data imbalance here with no significant increases and 19 significant decreases observed in 54 nucleotide sugar observations (Table 5). Moreover, some of the treated-to-control ratios were quite low for three nucleotide sugarsbetween 0.09 and 0.13 for UDP-glucuronate (by CeO 2 Z7, CeO 2 Q and CeO 2 Y6), UDP-N-acetylglucosamine (by CeO 2 Z7 and CeO 2 Q) and UDP-N-acetylgalactosamine (by CeO 2 Z7 and CeO 2 Q).

Altered lipids
Comparison of the results of this study with prior results from one CeO 2 nanomaterial (M from Nanoamour, dry size 8 nm) [5] shows that the results of the two studies are similar in respect to CeO 2 nanomaterial-induced elevations in fatty acids and monoacylglycerols. There were additional elevations in lysolipids, diacylglycerols and sphingolipids caused by CuO (this study) and by CeO 2 M [5], but in the current study the other five CeO 2 nanomaterials did not cause these particular lipid elevations. Possible explanations of the lipid increases seen with 3 CeO 2 , 1 SiO 2 and 1 CuO nanomaterial include: a) increases in lipolysis of complex lipids, b) increased synthesis of fatty acids, c) decreased utilization in βoxidation or complex lipid assembly or d) greater uptake of lipids from the cell culture media containing 10% fetal bovine serum because of nanoparticle uptake through endocytosis or nanomaterial induced cell membrane leakage. The major fatty acids of fetal bovine serum are palmitic, stearic and oleic [11]. However, these fatty acids were not particularly elevated over other fatty acids, thus arguing somewhat against the "greater uptake of lipids" interpretation.
A literature search showed elevated free fatty acids mentioned as a biomarker in ozone toxicity studies and ethanol-induced liver injury. Free fatty acids have been proposed as an "emerging biomarker" of nonalcoholic steatohepatitis [12]. From 1 to 48 h after exposure to hepatic irradiation, rat hepatic fatty acid concentrations were elevated [13]. Ozone exposures to both rats [14] and humans [15] elevated serum fatty acid concentrations. In addition, rat serum, brain and liver fatty acid concentrations were elevated by ethanol-induced liver injury [16]. In one in vitro study, exposure to quantum dots caused the down-regulation of beta-oxidation of fatty acids in PC12 cells (rat pheochromocytoma) [17]. In both PC12 cells and primary mouse hypothalamic cell culture, Zn-S coated quantum dots induced the accumulation of lipid droplets [17].
Glycerol levels were higher in several of nanoparticletreated cells relative to controls (Tables 3 and 4). Reduced glycerol 3-phosphate concentration was observed with each of the day-1 nanoparticles that elevated lipid concentrations (Tables 3 and 4). Glycerol 3-phosphate is utilized in the assembly of free fatty acids into triacylglycerides. A decline in glycerol 3-phosphate concentrations may be an indication of increased complex lipid assembly for storage [18]. Alternatively, a partial blockage in the transformation of glycerol into glycerol 3phosphate might reduce the synthesis of triglycerides and thus elevated free fatty acids, exactly what is observed in many cases (Tables 3 and 4).

Hepatic conjugation systems (methylation, glucuronidation and glutathione)
An important role of the liver is to conjugate various molecules with methyl, glucuronic acid or glutathione groups often as part of Phase II "drug metabolism" pathways [19]. Nanoparticle exposure may result in an increase in trans-methylation reactions and thus explain the observed SAM depletion.
One potentially important consequence of an insufficient supply of hepatocyte UDP-glucuronate would be a lack of glucuronidation capacity for Phase II metabolism of xenobiotic substances. Thus, even if nanoparticle clearance does not require glucuronidation per se, nanoparticle-induced UDP-glucuronate depletion may impair glucuronidation and clearance of other medicinal or toxic substances. Thus, with declines in both UDPglucuronate (Table 5) and SAM (Table 5), hepatocytes may have a diminished capacity to methylate, glucuronidate and excrete xenobiotics. In many animals, but not humans or guinea pigs, UDP-glucuronate is also a synthetic intermediate in the biosynthesis of ascorbic acid, an important cellular antioxidant. Gulonic acid, another biochemical intermediate in ascorbic acid biosynthesis was also decreased by prior administration of nanomaterials CeO 2 Z7, CeO 2 Q, and CuO (Table 5).
In this study, no useful GSH concentrations information was obtained because the measured GSH concentrations were often below the quantitation limit. In the sample preparation for metabolomics profiling, there was no added acid, chelators or deoxygenation of solutions, all well established factors that preserve GSH in the reduced oxidation state [20]. The size of the cell pellet was about 1/3 of that in our previous study so the factor of small cell pellet size also probably contributed to GSH being below the lower limit of measurement in most samples. It seems that the LC-MS/MS parts of the analytical procedure were working properly because other cell based studies run the following day and 2 days previous to our study measured GSH at typical levels for a cell based assay.
Cellular energetics, reducing capacity and oxidative stress (maltotriose, 6-phosphogluconate, NADPH, NADH and NAD + and dipeptides) Maltotriose, a trisaccharide consisting of three glucose moieties with alpha 1->4 glycosidic bonds between them is not known to be connected to toxicology or environmental health in any major way. However, maltotriose might be valuable as a biomarker of exposure for some metal oxide nanomaterials (e.g. 24.4 fold elevation by CuO). In yeast, exposure to either H 2 O 2 or CuSO 4 leads to increased maltotriose concentrations (https:// www.wikipathways.org/index.php/Pathway:WP478).
Most nano forms of copper give off Cu + and/or Cu ++ ions [21]. The single peptide bond of all dipeptides is capable of reducing Cu ++ to Cu + (the biuret reaction). In the presence of H 2 O 2 and Cu + , hydroxyl radical can be generated (the Fenton reaction) [22]. Such hydroxyl radicals are capable of destroying molecules within a short diffusional distance, such as the dipeptides binding site at which the Cu + may have been generated. This could explain why all 16 dipeptide concentrations were decreased (0.07 to 0.55 fold) by CuO nanomaterial administration. Neither CeO 2 , SiO 2 ( Table 6) or TiO 2 [5] nanoparticles caused large numbers of decreases in the dipeptide concentrations. After CuO exposure, 17 out of 20 single amino acids also exhibited decreases in concentration but not to as large an extent as observed for dipeptides (Additional file 2: Table S2). It does not seem as if CuO administration causes selective reductions of primary amine or carboxy group containing biochemical concentrations as there is substantial evidence against this possibility. For example, two primary amines containing biochemicals are significantly increased by CuO nanomaterial administration, namely putrescine (22.7 fold) and N-acetyl putrescine (63.3) (Additional file 2: Table S2). Three carboxy group containing biochemicals were also significantly increased by CuO nanomaterial treatment namely trans-4-hydroxyproline (1.8 fold), 4acetamidobutanoate (3.6) and pro-hydroxy-pro (proline-hydroxyproline, CAS 18684-24-7) (3.1 fold) (Additional file 2: Table S2).
Thus, CuO nanomaterials produced three effects at very high frequency of occurrenceelevation of certain lipids (Tables 3 and 4), decrease of most dipeptides ( Table 6) and decreases in many single amino acids (Additional file 2: Table S2). Thus, even if dissolution of CuO to copper ions produces hydroxy radicals, dipeptides and single amino acids are showing the large, consistently decreased cellular concentrations while other similar biochemicals are not showing decreases. An alternative explanation of the observed dipeptide decreases would be that protein breakdown was decreased.
Cellular effects (urea cycle, polyamines, purine and pyrimidine metabolism, nucleotide sugar) Among the CeO 2 nanoparticles from the day-1 set, CeO 2 Z7 stood out for its elevation of citrulline, ornithine and dimethylarginine, relative to controls and the other CeO 2 nanoparticles in the set. The higher levels of citrulline and ornithine in CeO 2 Z7-treated cells were not accompanied by a decrease in arginine, relative to control or the other CeO 2 nanoparticles. Dimethylarginine (both asymmetric and symmetric dimethylarginine were quantified together) were highest in CeO 2 Z7 treated cells and, given the inhibitory properties of asymmetric dimethylarginine towards iNOS, it is possible that less arginine gets converted directly to citrulline through iNOS and instead is converted to ornithine. There were fewer dimethylarginine increases observed in this data set than in the preceding metabolomics study in which 2 CeO 2 nanomaterials increased asymmetric dimethylarginine [5]. In addition, this study determined asymmetric and symmetric dimethylarginine together (Table 7) so this might have masked some asymmetric dimethylarginine increases.
Changes in urea cycle metabolites were also observed in the prior study with two forms of CeO 2 [5], with changes being more pronounced in the current study. The levels of creatine were correlated with creatinine and creatine phosphate (Table 7). Glycine is consumed in the synthesis of creatine. Glycine levels are decreased with several nanoparticle exposures (CeO 2 Z7, CeO 2 Q, and CuO) ( Table 5).
Among the day-1 nanomaterials, CuO caused the greatest amount of purine nucleotide degradation, as judged by the urate and allantoin levels. Metabolites connected with pyrimidine nucleotide degradation, such as thymidine and cytidine were elevated with several day-1 nanoparticle treatments (Table 7). Other purine nucleotide degradation metabolites were also increased. Hypoxanthine (4 increases) oxidation to xanthine (5 increases) and subsequent xanthine oxidation to urate (5 increases) by the enzyme xanthine oxidoreductase can produce superoxide or hydrogen peroxide, under some conditions. This can result in redox stress if sufficient anti-oxidants such as glutathione are not present.
Our first study with TiO 2 and CeO 2 and this current study with CeO 2 and SiO 2 agree in respect to the metabolite identity and direction of changes (increase or decrease) for several biochemicals notably NAD + , 6-phosphogluconate, UDPglucuronate, UDP-acetylglucosamine, UDP-galactosamine and gamma-glutamlyglutamate. In summarizing the results, there does not appear to be a single, obvious cause of some of the metabolomics effects observed (Additional file 5: Table  S5). The single CuO nanomaterial studied was quite different in number and some types of metabolomics effects it caused. This could be because of the different nanomaterial elemental composition (Cu rather than Ce or Si), higher degree of cytotoxicity observed with 3 μg/ml of CuO and the ability to form toxic copper ions via dissolution. Table 8 presents a summary of the treatment effects of the CeO 2 , SiO 2 and CuO particles for 13 of the more important altered biochemical pathways. Table 8 shows the direction of significant changes (up or down) for some of the altered biochemicals in each pathway. The number of significant changes observed per biochemical pathway was one in the glycogen pathway (maltotriose), two in the ascorbic acid synthesis pathway (gulonic acid and UDP-glucuronate), six in the glucuronidation-related pathway (glucoronate, UDP-N-acetylgalactosamine, UDP-N-acetylglucosamine, UDP-glucuronate, uridine 5′-diphosphate (UDP), and uridine 5′-monophosphate (UMP)) and over 40 in the lipid pathways (e. g. oleate, sterate and palmitate).

Dosimetry
In in vitro nanomaterial toxicology there are large numbers of complex factors involved in the pharmacokinetics and dosimetry between administered dose (expressed as μg/ml in this study) and internalized dose to the cultured HepG2 cell. Some of the major factors that determine in vitro intracellular dose of nanomaterials include particle dose, shape, surface chemistry, size, charge, density, binding of molecules to the particle surface (protein corona), agglomeration, diffusion and gravitational settling [23][24][25]. In our nanomaterial studies we have collected ICP-OES data on Ce and Cu cellular concentrations from CeO 2 and CuO exposed HepG2 cells. Eventually this cellular Ce and Cu dosimetry data may be useful in more deeply understanding the complex relationship between administered dose, internal cellular dose and various biological effects.

Altered lipids
This study confirms and extends the prior observation that a single CeO 2 nanomaterial (M) caused concentration increases in large numbers of several classes of lipids in HepG2 cells (most notably fatty acids and monoacylglycerols) [5]. In this study 4 CeO 2 , 1 SiO 2 and 1CuO nanomaterials were also shown to have this property of increasing lipid concentrations (Tables 3 and 4). In respect to structure-activity, we know that five out of six tested CeO 2 , and both SiO 2 and CuO, but zero out of 4 TiO 2 nanomaterials have caused this elevated concentration of lipids effect (Tables 3 and 4 and [5]). Thus, cellular lipid concentration increases may be a general property of exposure to many metal oxide nanomaterials and may impact hepatocyte and systemic lipid homeostatis.

Hepatic conjugation systems (methylation, glucuronidation and glutathione)
Metal oxide nanomaterial exposure may compromise the methylation, glucuronidation (Table 5) and glutathione conjugation systems (GSH data of [5]). The large number of metabolomics findings of decreased SAM coupled with increased SAH suggest an increase in transmethylation reactions and a depletion of SAM capacity. This shortage of methyl groups could have profound and adverse effects on cells in respect to DNA methylation and drug metabolism. From gammaglutamyl amino acid decreases data (Table 5), there was a degree of indirect confirmation of glutathione depletion and oxidative stress observed in our prior study with TiO 2 and CeO 2 nanomaterials [5].
Cellular energetics, reducing capacity and oxidative stress (maltotriose, 6-phosphogluconate, NADPH, NADH and NAD + and dipeptides) Increases in the concentration of maltotriose occurred in the prior metabolomics study (1.76 fold increase by CeO 2 M) [5] and also in this current study where the observed increases were much larger (a range of from 3.45 to 24.4-fold). To date, maltotriose concentrations have been significantly elevated by four out of six tested CeO 2 , along with both CuO and SiO 2, but zero out of 4 TiO 2 nanomaterials (Table 6 and [5]).
Observed depletions of both 6-phosphogluconate, NADPH and NADH suggest that the HepG2 cells may be out of redox equilibrium (not enough reducing equivalents) and thus in a state of oxidative stress. The unexpected pattern of CuO nanomaterial decreasing all 16 quantified dipeptides (Table 6) can be explained by the dissolution of CuO to ionic copper, peptide bond   Table  number  Tables 3 and 4 Tables 3 and   4   Table 5 Table 5   Table 5   Table 5   Table 6   Table 6   Table 6   Table 7   Table 7   Table 7   Table 7 Key: Up = several increases observed; Down = several decreases observed; − = No obvious pattern observed