Endothelial cell activation, oxidative stress and inflammation induced by a panel of metal-based nanomaterials.

The importance of composition, size, crystal structure, charge and coating of metal-based nanomaterials (NMs) were evaluated in human umbilical vein endothelial cells (HUVECs) and/or THP-1 monocytic cells. Biomarkers of oxidative stress and inflammation were assessed because they are important in the development of cardiovascular diseases. The NMs used were five TiO(2) NMs with different charge, size and crystal structure, coated and uncoated ZnO NMs and Ag which were tested in a wide concentration range. There were major differences between the types of NMs; exposure to ZnO and Ag resulted in cytotoxicity and increased gene expression levels of HMOX1 and IL8. The intracellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1(VCAM-1) expression were highest in TiO(2) NM-exposed cells. There was increased adhesion of THP-1 monocytic cells onto HUVECs with Ag exposure. None of the NMs increased the intracellular ROS production. There were no major effects of the coating of ZnO NMs. The TiO(2) NMs data on ICAM-1 and VCAM-1 expression suggested that the anatase form was more potent than the rutile form. In addition, the larger TiO(2) NM was more potent than the smaller for gene expression and ICAM-1 and VCAM-1 expression. The toxicological profile of cardiovascular disease-relevant biomarkers depended on composition, size and crystal structure of TiO(2) NMs, whereas the charge on TiO(2) NMs and the coating of ZnO NMs were not associated with differences in toxicological profile.


Introduction
The research on toxicological properties of nanomaterials (NMs) is increasingly important due to the rapid expansion of occupational, scientific and commercial use (Nel, 2013). The cardiovascular system is considered to be affected by NM exposure because of the analogy with ambient air particles that are associated with an increased risk of acute morbidity and mortality of cardiovascular diseases (Brook & Rajagopalan, 2010;Delfino et al., 2005;Mills et al., 2009). The principal cause of cardiovascular mortality and hospitalizations is atherothrombosis, which can lead to acute myocardial infarction in patients with coronary heart disease (Mills et al., 2007). A recent review showed that pulmonary exposure to ambient air particles as well as engineered NMs was associated with the development of vasomotor dysfunction and atherosclerosis in animal experimental models and humans . It has been suggested that the size, crystal structure and charge of TiO 2 particles are important variables for oxidative stress and inflammation (Johnston et al., 2009), but the effects of these variables have not been thoroughly investigated in regard to vascular endpoints. Oxidative stress and inflammation are implicated in the initiation and development of endothelial dysfunction, which plays an important role in the pathogenesis of a wide range of cardiovascular diseases (Elahi & Matata, 2006;Elahi et al., 2009). Atherosclerosis is also an inflammatory disease where leukocytes generate both reactive oxygen species (ROS) and cytokines (Hansson & Libby, 2006). Endothelial cells express intracellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1 (VCAM-1) on the membrane during inflammation, which is believed to be a critical step in the initiation of atherosclerosis (Libby, 2002). VCAM-1 mediates the arrest and adhesion of inflammatory cells to the vascular endothelium (Carlos et al., 1990), whereas ICAM-1 is involved in the stable adhesion step and migration of monocytes across the endothelial barrier (Javaid et al., 2003). Thus, adverse effects of NM exposure on endothelial cells and monocytes could be relevant endpoints in hazard characterization with respect to atherothrombosis.
We hypothesized that adverse effects of NMs depend not only on the chemical composition and particle size, but also on the surface charge, crystal structure and coating. To this end, we investigated the effect of exposure to a panel of metal-based NMs on: (i) cytotoxicity; (ii) expression of adhesion molecules ICAM-1 and VCAM-1; (iii) ROS production; (iv) pro-inflammatory response (IL8, TNF, CCL2); (v) oxidative stress response (HMOX1); and (vi) cell adhesion of monocytes onto endothelial cells. These end points were measured in human umbilical vein endothelial cells (HUVECs) and/or THP-1 monocytic cells with a panel of eight engineered NMs, including Ag, five different TiO 2 with different sizes or crystal structure and/or surface charge, and coated and uncoated ZnO in a wide concentration range. Monocyte adhesion to HUVECs was only performed for four NMs, selected to address the importance of size for TiO 2 and effects related to the most cytotoxic NMs in terms of ZnO and Ag. The panel of NMs was part of the European research program, Engineered NanoParticle Risk Assessment (ENPRA). NMs were provided by the program of NMs health and safety research of the Organisation for Economic Co-operation and Development (OECD). TiO 2 NMs were used for the assessment of effects of size, structure and charge with negatively and positively charged NMs made specifically for the ENPRA study. TiO 2 is used as a food-coloring agent, white pigment in toothpaste and UVlight-blocking agent in sunscreens. TiO 2 is also commonly used in paints, where it typically contains particle surface modifications in order to provide special properties of the paints (Christensen et al. 2011). The ZnO NMs were selected for the assessment of the impact of coating because they are used as surface-coated preparations, for instance in sunscreens (Vandebriel & de Jong, 2012). Ag is used in many consumer and household products as an anti-microbial agent (Schluesener & Schluesener, 2013).

Handling and preparation of NMs
The NMs with the prefix ''NM'' were obtained from the European Commission Joint Research Centre (Ispra, Italy): NM 101 (Hombikat UV100; TiO 2 , anatase, 7 nm), NM 110 (BASF Z-Cote; zinkite, uncoated, 100 nm), NM 111 (BASF Z-Cote; zinkite coated with triethoxycaprylylsilane, 130 nm) and NM 300 (RAS GmbH; Ag capped with polyoxylaurat Tween-20 520 nm). NRCWE 004 (TiO 2 , rutile, 94 nm) was purchased from NaBond Technologies Co., Hong Kong, China. NRCWE 001 (TiO 2 , rutile, 10 nm) was purchased from NanoAmor (Houston, TX) and used for the production of NRCWE 002 (amino-TiO 2 with positive charge) and NRCWE 003 (carboxy-TiO 2 with negative charge) as described previously (Kermanizadeh et al., 2012a). The handling and the preparation, including dispersion of the NMs was standardized among the laboratories participating in ENPRA (Jacobsen et al., 2010). The preparation of suspensions is described in detail in section S1 in the supplementary material. In brief, the NMs (2.56 mg/ml) were dispersed in 0.22 mm filtered MiliQ water with 2% serum using a Branson Sonifier mounted with a disruptor horn (Model S-450D, Branson Ultrasonics Corp., Danbury, CT). The sonicator was operated for a total of 16 min with an amplitude of 10%, and the samples were continuously cooled on ice. The stock suspensions of NMs were freshly used and diluted with cell culture medium to the final concentrations. The endotoxin content was measured by the Limulus Amebocyte Lysate (LAL) pyrogent assay (Lonza, Walkersville, MD). None of the NMs (256 mg/ml) displayed a positive reaction of endotoxin in the LAL assay, which had a sensitivity of 0.06 EU/ml.
The main physical and chemical properties of the NMs have previously been published (Kermanizadeh et al., 2012a). TEM size and TEM characteristics are reproduced in Table 1. The size of the NMs in suspensions with or without heat-inactivated fetal bovine serum (FBS) was evaluated by the Nanoparticle Tracking Analysis. The NMs were suspended in either 0.22 mm filtered Milli-Q water or in RPMI cell culture medium with 10% FBS. The results are shown in Table 1. The particles were sonicated as described above and diluted to a final concentration of approximately 1 mg/ml. The samples were analyzed on a Nanosight LM20 (Nanosight Ltd, Malvern Instruments, Malvern, UK) with a blue (405 nm) laser.

Cell cultures and NM exposure
THP-1 cells (American Type Culture Collection, Manassas, VA) were used as surrogate cells for human circulating monocytes in the cardiovascular system and kept in an undifferentiated state (Qin, 2012). The cells were cultured in RPMI cell medium with 10% FBS (EU origin) from Gibco Õ , Life Technologies Europe BV (Naerum, Denmark), as previously described (Danielsen et al., 2009). THP-1 cells are cultured in suspension. Thus, the exposure concentrations are given per volume of medium rather than per culture area.
The HUVECs and growth medium were purchased from Cell Applications (San Diego, CA). The cells were cultured in Endothelial Cell Growth Medium, with 2% serum as previously described . The cells were used at passages 2 to 5 to maintain morphologic and phenotypic characteristics of endothelial cells.
The single cultures were exposed to the same 2-fold concentration range between 2 and 256 mg/ml (corresponding to 0.64-80 mg/cm 2 ) in most experiments. For gene expression measurements the concentrations of 2, 8 and 32 mg/ml were excluded. For the cell adhesion assay the highest concentrations of 128 and 256 mg/ml were excluded and NMs included in the assay were the small and large TiO 2 (NRCWE 001 and NRCWE 004, respectively), uncoated ZnO (NM 110) and Ag (NM 300). Cellular uptake of the NMs, assessed by flow cytometry, is described in section S2 in the supplementary material. All measurements were repeated on three different days of analysis and they were analyzed in triplicate on each day of analysis.

Cytotoxicity
The cytotoxicity was measured with the colorimetric WST-1 assay from Roche Diagnostics GmbH (Mannheim, Germany). The amount of formazan dye formed in the assay directly correlates with the number of metabolically active cells. The assay was carried out as described in the manufacturer's instruction. In brief, cells (5 Â 10 4 THP-1 or 2 Â 10 4 HUVECs) were seeded in 96-well plates and exposed to NMs for 24 h. The cells were subsequently washed and incubated in fresh media with 10% WST-1 reagent for 2 h. The absorbance was measured at 450 nm, with 630 nm as reference, using an ELISA reader (Multiskan Ascent, Thermo Labsystems, Thermo Fisher Scientific Inc, Waltham, MA). The results are shown as percent cytotoxicity, where 100% corresponds to the absorbance measured in non-exposed cells. The concentration, which is cytotoxic to 50% of the cells (LC 50 ), was extrapolated from the concentration-response curve. The assay included two tests for interference between WST-1 and the NMs (2.56 and 256 mg/ml): (1) Cell medium + NMs and (2) Cell medium + WST-1 + NMs. The tests did not show any interference between NM exposure and WST-1 absorbance (results are shown in Table S1 in the supplementary material).

ROS production
The intracellular ROS production was measured in THP-1 cells and HUVECs as well as in a cell-free experiment by the 2 0 -7 0dichlorodihydrofluorescein diacetate (DCFH-DA) assay as previously described (Danielsen et al., 2011). In brief, 2 Â 10 4 HUVECs were seeded in black 96-well plates and incubated in cell culture medium for 24 h before the incubation with DCFH-DA (Cayman Chemical, Ann Arbor, MI). The HUVECs were incubated with 2 mM DCFH-DA for 15 min and washed twice with Hank's balanced saline solution. The THP-1 cells (5 Â 10 4 cells) were incubated with 2 mM DCFH-DA for 15 min and spun down and washed twice before they were transferred into black 96-well plates. The fluorescence in the cells was measured continuously on a fluorescence spectrophotometer at 485 and 538 nm (Fluoreskan Ascent FL, Thermo Labsystems) at 37 C for 3 h. The continuous measurement for 3 h was chosen because (1) the probe might be toxic to the cells during a longer exposure, (2) ROS production reaches a plateau after 3 h incubation that maybe due to the toxicity of the probe or that the probe is utilized, (3) a longer time outside the incubator can affect the viability of the cells. Carbon black particles (Printex 90, 10 mg/ml) were used as positive control because it has been shown that this concentration increases the ROS production by 4-fold in both THP-1 cells and HUVECs (Danielsen et al., 2011;Vesterdal et al., 2012). All experiments were repeated three times and the accumulated ROS production over time was calculated as the area under the curve. The results are reported as fold-increase compared with the control. The concentrations with statistically significant cytotoxicity, indicated by the WST-1 assay, are excluded from the figures and the statistics.

ICAM-1 and VCAM-1 surface expression
The surface expression of ICAM-1 and VCAM-1 was measured with a modified ELISA procedure (Frikke-Schmidt et al., 2011;Hemmingsen et al., 2011). HUVEC cells were seeded in transparent 96-well plates (2 Â 10 4 cells/well) and allowed to attach for 24 h. The NMs were added to the cells in triplicate and incubated for 24 h. The cell culture medium was discarded and the cells were incubated with primary ICAM-1 (cat. no. BBA17, R&D Systems, Abingdon, UK) or VCAM-1 (cat. no. BBA19, R&D Systems, Abingdon, UK) antibody. The cells were washed three times with warm cell culture medium containing 1% bovine serum albumin, followed by incubation with the secondary rabbit anti-goat IgG antibody (cat. no. A5420, Sigma, St. Louis, MO) that was diluted 1:1000 in phosphate buffered saline (PBS) with 0.1% Tween. The plates were then washed five times with icecold PBS/0.1% Tween. To measure the membrane-bound antibodies, 100 ml phosphatecitrate/o-phenylenediamine solution with 20 ml of 30% H 2 O 2 was added to each well, and the plate was incubated for 30 min in the dark. The absorbance was measured in an ELISA reader (Multiscan Ascent, Thermo Labsystems) at 450 nm. The proinflammatory cytokine TNF (100 ng/ml) was used as a positive control. The concentrations with statistically significant cytotoxicity, indicated by the WST-1 assay, are excluded from the figures and the statistics.
using High Capacity cDNA Reverse Transcription Kit, and the syntheses were carried out on the GeneAmpPCR system 2700 (both from Applied Biosystems Õ , Life Technologies Europe BV, Naerum, Denmark). Quantitative real-time PCR reactions were carried out in ABI PRISM 7900HT (Applied Biosystems Õ , Life Technologies Europe BV, Naerum, Denmark), using TaqManGene Expression Assays from Applied Biosystems Õ , Life Technologies Europe BV, Naerum, Denmark. The assay IDs of the genes were as follows: TNF, Hs00174128_m1; IL8, Hs00174103_m1; CCL2, Hs00234140_m1; HMOX1, Hs00157965_m1. We used 18S rRNA as reference gene (Eukaryotic 18S rRNA Endogenous Control, 4352930E, Applied Biosystems Õ , Life Technologies Europe BV, Naerum, Denmark). The gene expression level is reported as the ratio between the mRNA level of the target gene and the 18 S rRNA reference gene using the comparative 2 ÀDCt method. The concentrations with statistically significant cytotoxicity, indicated by the WST-1 assay, are excluded from the figures and the statistics.

THP-1 cell adhesion assay
The adhesion was assessed by measurement of 5-bromo-2deoxyuridine (BrdU)-labeled THP-1 cells onto HUVECs as previously described (Forchhammer et al., 2012). Briefly, 2 Â 10 4 HUVECs were seeded in 96-well plates and incubated overnight. The cell medium was removed and 5 Â 10 3 BrdUlabeled THP-1 cells were added to each well. The co-cultures were then exposed to NMs or the positive control TNF (100 ng/ ml) for another 24 h. After exposure, the cell medium was removed and the cells were washed twice with PBS. The supernatants (cell medium and PBS) were transferred to another 96-well plate. After centrifugation, the BrdU content both in the co-culture and supernatant was determined according to the manufacture's instruction (Roche Diagnostics GmbH, Mannheim, Germany). The cell medium was removed and the cells were fixed. The antibody solution (monoclonal antibody from mousemouse hybrid cells conjugated with peroxidase) was added to the wells. The antibody binds to the BrdU-labeled DNA and unbound antibodies were removed by three washing steps. The immune complexes were then detected by peroxidase-catalyzed conversion of the colorless substrate tert-methylbenzidine to a blue product. The reaction was stopped by the addition of 2 M HCl (50 ml/well) and the absorbance was determined with an ELISA-reader at 450 nm and with 690 nm as reference. The percentage of THP-1 cells remaining in the co-culture was calculated with the following equation:

Statistical analysis
The effects on cytotoxicity, ROS production and gene expression were assessed by ANOVA with the concentration nested in the type of NM as categorical variable. VCAM-1 and ICAM-1 surface expression data for TiO 2 NMs had non-normal distribution of residuals and inhomogeneity of variance between groups, but these data were also assessed by nested ANOVA because it tests for differences between types of particles. Non-parametric Kruskal-Wallis test on each of the five datasets of TiO 2 showed statistically significant effect of the same samples as the nested ANOVA test. Consequently, we have reported statistical differences in the nested ANOVA on ICAM-1 and VCAM-1 results in TiO 2 exposed cells, which have been verified by non-parametric tests. The validity of the nested ANOVA analysis was accepted on the basis of normal distribution of the residuals. For concentrations with statistically significant cytotoxicity, assessed as decreased WST-1 activity, we excluded results for other endpoints from the figures and the statistical analysis because they could be secondary effects. We kept concentrations of NM 101 (32 and 64 mg/ml) in the analysis of ICAM-1 expression and NM 300 (16 mg/ml) in the analysis of gene expression despite cytotoxicity in order to keep a balanced design in the statistical analysis. Statistical significant effects were accepted at 5% level in the overall nested ANOVA and in the post-hoc least significant difference (LSD) tests. The statistical analysis was performed in Statistica 5.5 (StatSoft, Inc., Tulsa, OK).

Cytotoxicity
We initially screened all NMs in the concentration range of 2-256 mg/ml for cytotoxicity by means of the WST-1 assay (concentration-response curves are shown in the supplementary material, Figures S1 and S2 for HUVECs and THP-1 cells, respectively). All the TiO 2 NMs, except NRCWE 003, generated statistically significant cytotoxicity at the highest concentrations (128 and 256 mg/ml, p50.05) in HUVEC cells, whereas NRCWE 002 and NM 101 also showed increased cytotoxicity at lower concentrations (64 and 32 mg/ml, respectively, p50.05). The ZnO NMs (NM 110 and NM 111) were both cytotoxic from the concentration of 32 mg/ml (p50.05) and Ag (NM 300) from the concentration of 64 mg/ml (p50.05). LC 50 was reached following exposure to ZnO at 29 mg/ml (NM 110) and 28 mg/ml (NM 111), and Ag at 57 mg/ml (NM 300). The cytotoxicity pattern was slightly different in THP-1 cells. All the TiO 2 NMs were noncytotoxic for all concentrations. Similar to the HUVECs, both of the ZnO were cytotoxic at the concentration of 32 mg/ml (p50.05). However, the Ag (NM 300) was cytotoxic at a lower concentration (16 mg/ml, p50.05) in the THP-1 cells as compared to HUVECs. LC 50 was reached following exposure to ZnO at 43 mg/ml (NM 110) and 46 mg/ml (NM 111) and Ag at 24 mg/ml (NM 300). The increased WST-1 formation at low concentration for some of the NMs could be due to an increased metabolic activity because the exposure triggered cellular stress. Thereby, the conversion of tetrazolium salts to formazan is increased until the concentration is so high that the cells are overwhelmed and consequently the metabolic activity drops.

Assessment of effect of size, crystal structure and charge (TiO 2 NMs)
The TiO 2 NMs had in general a low ability to generate ROS in cellfree dispersions and within cells during a 3 h exposure period (concentration-response curves for each of the NMs are shown in the supplementary material, Figure S3). The exposure to NRCWE 002 was associated with increased ROS production at 256 mg/ml in cell-free conditions (1.22-fold, p50.05) and within THP-1 cells (1.29-fold, p50.05). The NRCWE 004 was also associated with a statistically significant increase in the ROS production at the highest concentration in cell-free conditions (2.30-fold, p50.05).
The surface expression of ICAM-1 and VCAM-1 on HUVECs after exposure to TiO 2 NMs is shown in Figure 1. The expression level of ICAM-1 on the membrane of the HUVECs was increased in cells exposed to TiO 2 (NRCWE 004 and NM 101) at the highest concentrations (32 and 64 mg/ml, p50.05). These exposure-concentrations showed some cytotoxicity for NM 101 in the WST-1 assay (29 and 47% reduced viability, respectively). NRCWE 004, NM 101 and NRCWE 002 increased the levels of VCAM-1 at the highest concentrations (32 or 64 mg/ml, p50.05).
The mRNA expression levels of CCL2, IL8, TNF and HMOX1 in THP-1 cells after exposure to TiO 2 particles for 3 h are shown in Figure 2. The assessments of gene expression were carried out in the full concentration range up to 256 mg/ml because the exposure to TiO 2 was shown to be non-cytotoxic for THP-1 cells. The mRNA levels of CCL2 and IL8 were significantly increased compared to the control at the highest concentrations (64-256 mg/ml depending on the type of TiO 2 ). The level of TNF was only significantly increased in NRCWE 004 (128 mg/ml, p ¼ 0.01) exposed cells, but the highest concentration of 256 mg/ml showed borderline statistical significance (p ¼ 0.07). The mRNA level of HMOX1 was significantly increased in cells exposed to NRCWE 003 (256 mg/ml) and NRCWE 004 (from 64 mg/ml).
NRCWE 001 and NRCWE 004 were selected for the assessment of adhesion of THP-1 cells to HUVECs because they had shown the widest difference in potency for ICAM-1 and VCAM-1 expression on HUVECs and expression of inflammation genes in THP-1 cells (CCL2, IL8 and TNF). Nevertheless, neither NRCWE 001 nor NRCWE 004 increased the attachment of THP-1 cells onto HUVECs in co-cultures that were exposed for 24 h, whereas there was increased adhesion of THP-1 cells after exposure to the positive control TNF (p50.05, Figure 3).
The overall summary of comparison of the effect of size, crystal structure and surface charge of TiO 2 NMs is outlined in Table 2. In general, there was a somewhat stronger response for the anatase (NM 101) as compared to the rutile (NRCWE 001) TiO 2 NMs, assessed mainly by the expression of ICAM-1 and VCAM-1 on HUVECs. In addition, the size of rutile TiO 2 particles was an important variable in as much as the larger particle (NRCWE 004) was more potent than the smaller rutile NM (NRCWE 001) in generating increased expression of cell adhesion molecules on HUVECs and gene expression in THP-1 cells. The positively charged amino-TiO 2 (NRCWE 002) and negatively charged carboxy-TiO 2 (NRCWE 003) were generally associated with low or statistically non-significant effects and it therefore indicates that the charge had limited importance for the endpoints in this investigation.

Assessment of the effect of surface coating on ZnO NMs
The ZnO particles did not induce ROS production in the cell-free assay or within cells ( Figure S4). HUVECs had increased levels of both ICAM-1 and VCAM-1 at the highest and non-cytotoxic concentrations of ZnO NMs (8 or 16 mg/ml, p50.05, Figure 4). The mRNA levels of CCL2 were unaltered in THP-1 cells exposed to either ZnO NMs, whereas the expression of TNF was increased at 4 and 16 mg/ml for the coated ZnO (NM 111) (p50.05, Figure 5). The expressions of IL8 and HMOX1 were increased at 16 mg/ml for both ZnO NMs (p50.05, Figure 5).   (NM 110), the adhesion of THP-1 cells was not statistically increased, although the average adhesion at 8 and 16 mg/ml was nominally higher than that achieved by TNF exposure (Figure 3).

Assessment of effects of Ag NM
The Ag NM did not induce ROS production in the cell-free assay or cultured cells ( Figure S5). The expression of ICAM-1 and VCAM-1 on HUVECs was increased at 16 and 32 mg/ml (p50.05, Figure  S6), which were non-cytotoxic concentrations assessed by the WST-1 assay. The mRNA expression in THP-1 cells was unaltered for CCL2 and TNF, increased at 4 and 16 mg/ml for IL8 (p50.05) and at 16 mg/ml for HMOX1 (p50.05, Figure 5). There was also an increased adhesion of THP-1 cells onto HUVECs at 32 mg/ml (p50.05, Figure 3), although it should be noted that this concentration was associated with cytotoxicity in THP-1 cells (only 35% cell viability as assessed by the WST-1 assay). Figure S7 in the supplementary material shows a representative image of the co-culture exposed to Ag NM, examined by combined Differential Interference Contrast (DIC) and fluorescence microscopy in a Leica AF6000 inverted widefield microscope (Leica Microsystems GmbH, Wetzlar, Germany).

Discussion
This study focused on the impacts of a panel of metal-based NMs with different size, crystal structure and coating on oxidative stress, inflammation and cardiovascular-related biomarkers in human cell cultures of monocytes and endothelial cells. Our data show that the acute cytotoxic effects of the NMs differ in the HUVECs and monocytic THP-1 cells. Such differences in cytotoxicity could be due to the different cell type, type of medium or whether the cells are adherent or growing in suspension. For both cell types Ag and coated and uncoated ZnO were highly cytotoxic, assessed by the WST-1 assay. The ZnO NMs were the most toxic NMs for HUVECs, whereas Ag was the most toxic NM for the THP-1 cells. Parallel studies conducted as part of the ENPRA project with the same panel of NMs showed similar results in hepatocytic and renal cell lines, where Ag was more toxic than the ZnO NMs (Kermanizadeh et al., 2012b. In contrast, the TiO 2 NMs showed no cytotoxicity in THP-1 cells and only moderate cytotoxicity in HUVECs, which did not reach LC 50 values. The FBS content in the cell culture medium may also play a major role for NMinduced cell cytotoxicity, which has been suggested to be due to better bioavailability . The HUVEC and THP-1 cell culture medium contains 2 and 10% FBS, respectively. This difference in serum content could have an impact on the NM size and affect cellular uptake due to differences in protein coronas (Walkey & Chan, 2012). Nevertheless, the NM sizes in HUVEC and THP-1 cell culture medium revealed no clear pattern due to the FBS content (Table 1). THP-1 cells and HUVECs appeared to have similar uptake of NMs with NRCWE 004 showing particularly high uptake as described in Table S1. The uptake was evaluated by the Flow Cytometric Light Scatter analysis where cell granularity is used as a marker of particle uptake (Suzuki et al., 2007). It has been argued that this method does not distinguish between particles attached to the cell membrane and internalized particles . However, we have documented excellent agreement between flow cytometry and confocal microscopy for uptake of gold nanoparticles in HUVECs (Klingberg et al., 2014). The uptake of TiO 2 and silver nanoparticles has in several studies been evaluated using flow cytometry in combination with confocal or dark field microscopy to verify a dose-dependent internalization (Suzuki et al., 2007;Vranic et al., 2013;Zucker et al., 2010Zucker et al., , 2013. In relation to oxidative stress, TiO 2 NMs did not generate intracellular ROS during a continuous measurement for 3 h in any of the cell types, which is comparable with a previous study, where only a large TiO 2 (220 nm) induced ROS production in HUVECs and two other TiO 2 with smaller sizes (95 and 17 nm) did not (Mikkelsen et al., 2013). Similarly, neither ZnO NMs nor Ag increased the intracellular ROS production in our experiments. Interestingly, only the TiO 2 NMs induced ROS production in hepatocytes, whereas the two ZnO NMs and Ag induced ROS production in a renal cell line (Kermanizadeh et al., 2012a. The ROS production was measured by means of DCFH once after 6 or 24 h exposure in the study on hepatocytes (Kermanizadeh et al., 2012a), whereas we used continuous measurements for 3 h and expressed the results as a fold-change in ''area under the curve''. The study on renal cells used dihydroethidium and flow cytometry for the measurement of ROS production . Therefore, the discrepancy in ROS production between the studies using the same NMs and dispersion procedure could be due to differences in cell type and cell culture medium or due to other differences in the experimental method, such as the fluorescent probe and exposure time. Two types of TiO 2 NMs (NRCWE 004 and NM 101), both ZnO NMs and the Ag were able to induce ICAM-1 and VCAM-1 expression in a concentration-dependent manner. Overall, our results indicate that the upregulation of ICAM-1 and VCAM-1 does not depend on cellular oxidative stress in terms of intracellular ROS production for the presently tested NMs. However, it should be emphasized that the DCFH assay detects a variety of hydroxyl-, superoxide-, NO radicals, peroxynitrite and other ROS (Wardman, 2007).
It has been suggested that the proinflammatory cytokine TNF plays a key role in particle-elicited inflammation by functioning as a mediator for expression and secretion of chemokines (Driscoll et al., 1997). Other proinflammatory cytokines include IL8 and monocyte chemoattractant protein 1 (CCL2 alias MCP-1). IL8 exhibits mostly neutrophil chemotactic activity (Roebuck, 1999), whereas CCL2 is a chemoattractant for Figure 3. Adhesion of THP-1 cells onto HUVECs after 24 h exposure to TiO 2 NMs (NRCWE 001 and NRCWE 004), ZnO (NM 110,uncoated) and Ag (NM 300). TNF is included as a positive control. Data are expressed as percentage of adherent THP-1 cells to the total (cell culture + supernatant) and the bars are means ± SEM of three independent experiments. *p50.05 compared to unexposed controls.   At these concentrations (NM 300, 4 and 16 mg/ml) significant levels of cytotoxicity (WST-1 assay) were observed in THP-1 cells, but the data are included in the statistical analysis to keep a balanced design. Note that cytotoxicity was measured after 24 h exposure and gene expression after 3 h exposure.
monocytes (Ueda et al., 1997). The data on the proinflammatory cytokines CCL2 and IL8 showed that all the TiO 2 NMs increased the mRNA expression after 3 h exposure, whereas TNF expression showed no response. The expression of the oxidative stress response gene HMOX1 was increased in cells exposed to negatively charged TiO 2 (NRCWE 003) and the larger rutile TiO 2 (NRCWE 004). Both ZnO NMs significantly increased the expression of IL8 and HMOX1, but only the coated ZnO (NM 111) significantly increased the expression of TNF in the THP-1 cells. Hepatocytes, exposed to the same panel of NMs as in this study, showed a concentration-dependent increase in the production of IL8, when exposed 24 h to TiO 2 , whereas the production of IL8 peaked around the LC 50 levels, when exposed to ZnO NMs and Ag (Kermanizadeh et al., 2012b). It was primarily the larger rutile TiO 2 (NRCWE 004) and the anatase TiO 2 (NM 101) that were able to induce VCAM-1 and ICAM-1 expression on the surface of HUVECs in a concentration-dependent manner compared to the less potent smaller rutile TiO 2 (NRCWE 001). The same pattern was observed for the gene expression of IL8, TNF and HMOX1. There were statistically significant differences in gene expression between NRCWE 001, NRCWE 004 and NM 101, which could relate to either differences in particle size (NTA size measure: 108 nm (NRCWE 001) and 202 nm (NRCWE 004)) or crystal structure (anatase (NM 101) versus rutile (NRCWE 001)). A previous study showed that TiO 2 with different sizes (12, 21 and 288 nm) increased the levels of ICAM-1 and VCAM-1 on the membrane of HUVECs, but without clear size or crystal structure dependency (Mikkelsen et al., 2013). A recent study of polymorph-and sizedependent uptake and toxicity of TiO 2 in lung epithelial cells concluded that the induction of IL-8 and MCP-1 might be a sizerelated effect (Andersson et al., 2011). The crystal structure has been shown to be important for the biological effects in some studies. Nano-scale anatase TiO 2 was shown in vitro to release more IL8 than nano-scale rutile TiO 2 (Sayes et al., 2006), whereas in vivo studies have shown that rutile TiO 2 was more inflammogenic than anatase TiO 2 (Lu et al., 2009;Roursgaard et al., 2011). In acellular conditions, anatase TiO 2 exhibited higher ROS production than similar sized rutile TiO 2 (Jiang et al., 2008). Also surface properties, such as surface charge can have an impact on the measured biological outcome. In our study, one of the TiO 2 (NRCWE 001) were functionalized using 3-aminopropyltriethoxysilane and sucinic anhydride/tetrahydrofuran. The recovery products were a positively charged amino-TiO 2 (NRCWE 002) and a negatively charged carboxy-TiO 2 (NRCWE 003). We observed similar increases in ICAM-1 (approximately 1.5-fold) and VCAM-1 (approximately 1.6-fold) expression on HUVECs after exposure to the negatively charged (NRCWE003), positively charged (NRCWE 002) and neutral TiO 2 (NRCWE 001). It has been shown that amino-functionalized polystyrene nanoparticles, but not carboxyl-functionalized, induced inflammasome activation leading to mitochondrial damage, ROS and IL-1b production (Lunov et al., 2011). It is clear from these results that particle characteristics can play a significant role in toxicological responses. Still there is a paucity of toxicological studies that systematically examine the role of particle size or surface properties, such as crystal structure. The co-culture experiment showed increased adhesion of THP-1 monocyte cells to HUVECs after exposure to Ag NMs. This interaction is an early event in atherosclerosis, although it is also a physiological response to tissue inflammation in the vicinity of blood vessels for recruitment of inflammatory cells. The Ag also significantly increased the levels of ICAM-1 and VCAM-1, but not the intracellular ROS production. It has been reported that dissolution of Ag ions from Ag NMs is an important mechanism for cytotoxicity, but it has also been shown that induction of ROS production and cell membrane damage was higher for Ag NMs than for Ag ions (Ivask et al., 2014). In a recent study, it was shown that Ag ions released from AgNO 3 induced higher levels of cell death compared to Ag NMs, which showed low toxicity despite a higher cellular concentration of silver from the latter (Cronholm et al., 2013). In contrast, another study found that Ag NMs increased the cytotoxicity to a greater extent than AgNO 3 (Piao et al., 2011). The Ag (NM 300) used in the present study showed less than 1% dissolution of Ag ions measured in deionized water or cell culture medium after 24 h incubation . Consequently, it is assumable that the effect of ions is negligible in this study.
Our data suggests major differences in cytotoxicity between the three overall types of NMs. The non-cytotoxic concentrations of ZnO NMs dramatically increased the expression levels of HMOX1 and IL8 ($80-to 175-fold) and this at a lower exposure concentration than for TiO 2 NMs, where the increase was not higher than 5.5-fold. Also the Ag increased the expression levels of HMOX1 and IL8 ($92-and 22-fold). In contrast, the ICAM-1 and VCAM-1 expressions were higher in TiO 2 NM-exposed HUVECs as compared to ZnO NMs and Ag NM at non-cytotoxic concentrations. The relationship between NM characteristics (coating, crystal structure, size and charge) and markers of toxicity are summarized in Table 2.

Conclusions
We found that there were no major effects of the coating of ZnO NMs. The TiO 2 NMs data suggest that the anatase form (NM 101) was more potent than the rutile (NRCWE 001) in terms of ICAM- Figure 5. Relative mRNA expression of CCL2, HMOX1, IL8 and TNF in THP-1 cells after exposure to ZnO NMs (NM110,uncoated;NM 111,coated) and Ag (NM 300) for 3 h. The bars are means ± SEM of three independent experiments. *p50.05 compared to unexposed cells.
1 and VCAM-1 expression. Related to size we found that the larger TiO 2 NM (NRCWE 004) were more potent than the smaller (NRCWE 001) for gene expression and ICAM-1 and VCAM-1 expression. The effect on the cardiovascular disease-relevant biomarkers depended on NM composition, size and crystal structure of TiO 2 NMs, whereas the charge of TiO 2 NMs and coating of ZnO NMs was not associated with differences in the toxicological profile.