Journal of Nanoparticle Research

, Volume 12, Issue 1, pp 61–73

Toxicological consequences of TiO2, SiC nanoparticles and multi-walled carbon nanotubes exposure in several mammalian cell types: an in vitro study


  • Sabrina Barillet
    • Laboratoire Pierre SüeCEA-CNRS UMR9956, IRAMIS, CEA Saclay
  • Angélique Simon-Deckers
    • Laboratoire Pierre SüeCEA-CNRS UMR9956, IRAMIS, CEA Saclay
  • Nathalie Herlin-Boime
    • Laboratoire Francis PerrinCEA-CNRS URA2453, IRAMIS, CEA Saclay
  • Martine Mayne-L’Hermite
    • Laboratoire Francis PerrinCEA-CNRS URA2453, IRAMIS, CEA Saclay
  • Cécile Reynaud
    • Laboratoire Francis PerrinCEA-CNRS URA2453, IRAMIS, CEA Saclay
  • Doris Cassio
    • INSERM UMR-S 757, Centre Universitaire
  • Barbara Gouget
    • Laboratoire Pierre SüeCEA-CNRS UMR9956, IRAMIS, CEA Saclay
    • Scientific DepartmentAFSSA
    • Laboratoire Pierre SüeCEA-CNRS UMR9956, IRAMIS, CEA Saclay
Special focus: Safety of Nanoparticles

DOI: 10.1007/s11051-009-9694-y

Cite this article as:
Barillet, S., Simon-Deckers, A., Herlin-Boime, N. et al. J Nanopart Res (2010) 12: 61. doi:10.1007/s11051-009-9694-y


The development of nanotechnologies may lead to dissemination of potentially toxic nanoparticles in the environment. Toxicology of these nano-sized particles is thus attracting attention of public and governments worldwide. Our research is focused on the in vitro response of eukaryotic cells to nanoparticles exposure. For this purpose, we used cellular models of primary target organs (lung: A549 alveolar epithelial cells), or secondary target organs (liver: WIF-B9, Can-10 and kidneys: NRK-52E, LLC-PK1 proximal cells), i.e., organs exposed if nanoparticles are translocated through epithelial barriers. These cells were exposed to TiO2, SiC nanoparticles or multi-walled carbon nanotubes (MWCNT). The influence of nanoparticles physico-chemical characteristics on various toxicological endpoints (cytotoxicity, reactive oxygen species generation, genotoxicity) was specified. Our data demonstrate that nanoparticles toxicity depend on their size, morphology, and chemical composition, the finest, spherical shaped, and anatase TiO2 nanoparticles being the more cytotoxic to NRK-52E cells, while SiC nanoparticles exert almost no cytotoxicity. MWCNT cytotoxicity neither depended on their length, nor on the presence of metal impurities. Nanoparticles cytotoxicity also depended on the exposed cell line. All the tested nanoparticles were uptaken by cells and caused intracellular reactive oxygen species generation. Relative to genotoxic effects, DNA strand breaks were detected in NRK-52E cells via the alkaline comet assay after exposure of cells to TiO2 nanoparticles and to a lesser extent after exposure to MWCNT, but no double strand breaks were detected. The originality of this study lies on the panel of nanomaterials which were tested on a variety of cell lines. All these data may lead to a better understanding of nanomaterial toxicity and hazards for health.


NanotoxicologyTitanium oxideSilicon carbideMulti-walled carbon nanotubesCytotoxicityAccumulationOxidative stressGenotoxicityEHSNanomedicine


The increasing production and use of nanoparticles for industrial applications leads to a considerable questioning about their potential impact on the environment and for human health. While a large literature describes the impact of nanoparticles and carbon nanotubes on primary target organs (particularly on the pulmonary tract), little is known about the potential effects of such materials on other organs if the hypothesis of a translocation of these nanoparticles through epithelial barriers is assumed. Such translocation phenomenon has already been demonstrated in both respiratory and digestive tracts. Indeed, for the respiratory route of exposure, translocation was noticed with radioactive technetium-99m-labeled carbon nanoparticles, after intratracheal instillation of hamsters (Nemmar et al. 2001) or inhalation by humans (Nemmar et al. 2002), with 192-Ir radiolabeled nanoparticles (Semmler et al. 2004), and with Ag nanoparticles (Takenaka et al. 2001). Secondary target organs have then been identified as blood vasculature, heart, liver, spleen, kidneys, and brain (Takenaka et al. 2001; Nemmar et al. 2001, 2002; Semmler et al. 2004). In the specific case of TiO2 nanoparticles, this translocation phenomenon from the respiratory system does exist too, since TiO2 nanoparticles were observed within lung microvasculature after inhalation exposure of rats (Geiser et al. 2005). For the digestive route of exposure, similar results were described in a recent in vivo biodistribution study in which the authors showed that nano-sized TiO2 particles (administered to mice by a single oral gavage) are mainly retained in the liver, spleen, kidneys, and lung tissues, indicating that TiO2 particles could be transported to other tissues and organs after uptake via the gastrointestinal tract (Wang et al. 2007). Furthermore, considering carbon nanotubes toxicokinetics, some authors have demonstrated that both intraperitoneal and intravenous administration lead to a rapid distribution of these nano-sized particles in mice. These authors identified liver, kidneys, stomach, and lung as the target organs of the carbon nanotubes (Wang et al. 2008a). Thus, a knowledge gap has to be filled with new nanotoxicity data on such organs.

Furthermore, in the global framework of nanotoxicology, in vitro studies sometimes lead to controversial conclusions. One explanation of this controversy is that different production processes, dispersion methods, and life cycles of nanoparticles may confer changes in their surface characteristics and reactivity. For example, in some studies TiO2 nanoparticles do not induce toxic events (Singh et al. 2007; Park et al. 2007), while other studies reveal toxic effects in which nanoparticle size is a key determinant (Park et al. 2007). In some studies, TiO2 toxicity rather depends on crystalline phase than on particle size (Sayes et al. 2006). This controversy also exists in the context of carbon nanotube toxicity evaluation: in some cases the presence of metal trace impurities explains toxicity (Pulskamp et al. 2007), in some cases it does not (Simon-Deckers et al. 2008). Considering these confounding factors, it appears essential to accurately control the chemical and physical parameters of each tested nanomaterial.

Main toxic events observed after exposure to ambient ultrafine particles are pulmonary inflammation, oxidative stress, and fibrosis (Nel et al. 2006). In vitro, nanoparticles toxicity studies support these physiological responses, implicating oxidative stress in the release of proinflammatory cytokines and cytotoxic cellular responses. TiO2 adverse effects on cells are generally attributed to reactive oxygen species (ROS) production (Singh et al. 2007; Jin et al. 2008). In the case of carbon nanotubes, ROS production has been demonstrated, and attributed to metal-based impurities (Pulskamp et al. 2007). Silicium carbide (SiC) nanocrystals are assumed to be highly biocompatible, particularly to blood, and are envisaged as biological labels for cell imaging applications, due to their photoluminescence properties. A recent study shows their luminescence in both cell cytoplasm and cell nucleus (Fan et al. 2008), suggesting that they may distribute through the entire cell, potentially interact with DNA, and consequently exert both cyto- and genotoxic effects. These toxicological effects still remain to be studied in cells where SiC nanocrystals are accumulated.

In a view to understand the potential implications of nanoparticles exposure both on directly exposed organs (lung) and secondary organs (liver, kidneys), we screened a panel of cell lines, representative of these organs, for their response to nanoparticles or multi-walled carbon nanotubes (CNT). The chosen nanoparticles differ in terms of chemical composition (TiO2, Al2O3, SiC), shape (spherical, elongated), crystalline phase (anatase, rutile) and size (mean diameter ranging from 12–140 nm). CNT differ in terms of length and purity, i.e., the presence of iron-based phases in the non-purified CNT sample. On the chosen cell lines, we addressed the cytotoxic effects of these nanoparticles, and their intracellular accumulation was observed by transmission electron microscopy (TEM). Intracytoplasmic ROS production was measured. The influence of nanoparticles physico-chemical characteristics was evaluated. The originality of this study lies on the panel of tested nanoparticles, having different size, shape, crystalline structure, chemical composition and properties, and different purity. The results presented herein enrich our previously acquired data on A549 cells (Simon-Deckers et al. 2008), enlarging the model organs to liver and kidney. Moreover, preliminary results of oxidative stress generation and DNA damage are presented, allowing to give a first answers on the mechanisms underlying nanoparticles toxicological effects.

Materials and methods

Chemicals and nanomaterials

Chemicals and cell culture media were obtained from Sigma-Aldrich. Multi-walled carbon nanotubes (CNT), SiC, and TiO2-CEA nanopowders were synthesized by aerosol-assisted catalytic chemical vapor deposition using Fe as catalyst (CNT; Pinault et al. 2005) or by laser pyrolysis (SiC and TiO2-CEA nanoparticles; Pinault et al. 2004). For their purification, CNT were heated under argon atmosphere at 2,000 °C during 2 h in order to remove iron-based phases (Pignon et al. 2008) and TiO2-CEA nanoparticles were annealed under air at 400 °C during 3 h to remove free carbon impurities. Toxicological impact of these nanotubes and nanoparticles were compared to the effect of commercial nanoparticles: TiO2–P25 (Aeroxide® P25, 75% anatase), and Al2O3 (Aeroxide® AluC) from Degussa AG, TiO2-Sigma (ref. T8141, 100% anatase) and TiO2-Sigma-R (ref. 637262, 100% rutile) from Sigma-Aldrich. SiC nanoparticles had a chosen Si/C ratio of 0.83 in order to minimize surface oxidation.

Nanomaterial suspensions

Nanoparticles were dispersed by sonication (Autotune 750 W, Bioblock Scientific) in ultrapure sterile water (pH 5.5, TiO2, and Al2O3 nanoparticles) or gum Arabic (0.25 wt%, nanotubes) or fetal bovine serum (SiC), at 4 °C, pulsed mode, 30 min for nanoparticles or nanotubes having the longest size (designated “long nanotubes”) and 8 h for the shortest nanotubes (designated “short nanotubes”; Glory et al. 2007). These suspensions were diluted in cell culture medium (DMEM medium supplemented with 50 IU/mL penicillin and 50 μg/mL streptomycin) as previously described (Simon-Deckers et al. 2008). Nanoparticles aggregation status was investigated after this dilution step in the cell culture medium. For titanium dioxide nanoparticles, results are presented elsewhere (Simon-Deckers et al. 2008). For carbon nanotubes, the presence of gum Arabic allows nanotubes to remain dispersed after dilution.

Cell culture

Cell lines were purchased from ATCC: A549 human lung carcinoma cells (CCL-185), NRK-52E rat kidney proximal cells (CRL-1571), LLC-PK1 pig kidney proximal cells (CL-101); or kindly provided by Dr. D. Cassio: WIF-B9 and Can-10 (Decaens et al. 2008). They were subcultured in DMEM containing 4.5 g/L glucose supplemented with 2 mM l-glutamine, penicillin/streptomycin (50 IU/mL and 50 μg/mL, respectively), and 10% (v/v) fetal bovine serum (A549, NRK-52E, LLC-PK1), or in a specific medium, described elsewhere (WIF-B9 and Can-10; Cassio et al. 2007). They were maintained at 37 °C in a 5% CO2/air incubator and passed at confluence.

Cytotoxicity assays

Cells were grown to sub-confluence in 96-well plates, exposed to 100 μL of 0.25–100 μg/mL nanoparticles suspensions, for 1–72 h (see figure captions for details). Nanomaterials cytotoxicity was assessed by using 3-(4,5-dimethylthiazol-z-yl)-2,5-diphenyl-tetrazotium bromide (MTT) assay (Mosmann 1983). Mitochondrial dehydrogenases of viable cells reduce MTT to water-insoluble blue formazan crystals which are then solubilised by dimethyl sulfoxide (DMSO); this assay thus indicates cell mitochondrial activity impairment. After exposure, 10 μL of a 5 mg/mL MTT solution was added to each well. After 1 h at 37 °C, medium was then replaced by 100 μL of DMSO and mixed thoroughly to dissolve the formazan crystals. In the case of nanoparticles toxicity assessment, methodological problems are noticed. For example, nanoparticles may interact with the substrate, thereby fixing free MTT and avoiding the crystallization of formazan, thus causing false negative results. Furthermore, due to their optical properties, some nanoparticles (particularly carbon-based nanoparticles) present either in the reaction mixture, or inside cells or absorbed on cells may directly influence the readout by increasing light absorption. Thus, the suitability of MTT for in vitro toxicity assessment of nanoparticles has come into question. In order to limit such potential interference of nanoparticles on the MTT assay, we carried out an additional methodological step in the MTT usual protocol. Nanoparticles were allowed to sediment during 1 h and 50 μL of each well were then transferred to another plate. Then, absorbance was measured at 570 nm. This procedure reduced the quantity of nanoparticles in suspension, which may interfere with MTT assay. Doing this, we noticed that MTT test was much more reliable for nanoparticle toxicity evaluation than in its usual form. Moreover, the absence of interference of nanoparticles with MTT was verified by mixing TiO2 nanoparticles with DMSO-dissolved formazan crystals. TiO2 nanoparticles did not induce any variation in absorbance (570 nm) measurement (not shown). Cell viability was then determined as a percentage of the negative control (untreated cells).

Cell membrane impairment was studied using lactate dehydrogenase (LDH) release assay (Sigma-Aldrich LDH assay kit; Decker and Lohmann-Matthes 1988). After nanoparticles exposure, 25 μL of cell supernatant were collected in a 96-well plate. Seventy-five microliters of LDH assay solution (prepared as described by the supplier) was added to each well. After 30 min of incubation, at room temperature, protected from light, 10 μL of HCl 1 N were added to wells and absorbance at 490 nm was read. Results were expressed as % of a positive control, cells exposed to 1 mM of CdCl2 for 24 or 48 h, condition in which 100% of intracellular LDH was released in cell supernatant.

Cell preparation for transmission electron microscopy (TEM)

Cells were exposed during 6 or 48 h to nanoparticles (50 μg/mL) or nanotubes (10 μg/mL). They were fixed with 2.5% glutaraldehyde, post-fixed with OsO4 and dehydrated in graded concentrations of ethanol (Strum et al. 1971) then embedded in Epon (nanoparticles) or Spurr (nanotubes). Ultra-thin sections were cut (80 nm), counterstained with lead citrate and uranyl acetate and observed with a CM 12 Philips electron microscope at 80 kV.

Intracellular reactive oxygen species (ROS) formation

The formation of ROS was measured via monitoring the increasing fluorescence of 2′,7′-dichlorodihydrofluorescein diacetate acetyl ester (H2DCF-DA, Invitrogen; Oyama et al. 1994). This dye is a stable cell-permeant indicator which becomes fluorescent when cleaved by intracellular esterases and oxidized by intracellular H2O2, ROO, ONOO. Fluorescence intensity is thus proportional to the amount of reactive oxygen species produced in cells. Cells were exposed to nanoparticles (see figure captions for details), washed twice with PBS, and incubated 30 min at 37 °C with 80 μM H2-DCF-DA. They were then harvested by scraping, centrifuged, and resuspended in PBS. The fluorescence intensity of cell suspensions was measured with excitation at 480 nm and emission at 530 nm (Molecular Devices Gemini X fluorescence spectrophotometer).

Genotoxicity assays

Comet assay was performed under alkaline conditions, according to the procedure of Singh et al. (1988). Microscope slides were precoated with 1% normal melting point agarose (NMA). Around 10,000 cells were mixed with 75 μL of 1% low melting point agarose (LMPA), pipetted over the first layer of agarose, and covered by a coverslip. The slides were placed on a flat tray and kept on ice for 5 min until agarose solidifies. The coverslips were removed and 75 μL of 0.5% LMPA were dispensed on the second layer of agarose, covered by a coverslip and allowed to solidify on ice. The slides were immersed in cold lysing solution (2.5 M NaCl, 100 mM EDTA, 10 mM Tris, 10% DMSO, 1% Triton X-100) for 1 h at 4 °C. DNA was then allowed to unwind for 30 min in alkaline electrophoresis solution (300 mM NaOH, 1 mM EDTA, pH >13). Electrophoresis was performed in a field of 0.7 V/cm and 300 mA current for 20 min. Slides were neutralized with 0.4 M Tris pH 7.5 and stained with 50 μL of 20 μg/mL ethidium bromide. Two slides per condition, and at least 50 comets per slide were analyzed under a fluorescence microscope (Olympus BX41) equipped with a 350–390 nm excitation and 456 nm emission filter at 20× magnification. Comet length and intensity were analyzed by using ImageJ software, comet tail moment was calculated as comet tail intensity/comet head intensity × comet tail length.

For γ-H2AX immunostaining, cells were fixed for 15 min with 3% paraformaldehyde at room temperature, and permeabilized for 10 min with 0.1% Triton X-100 prepared in PBS containing 2% bovine serum albumine. Classical immunostaining was then performed using anti-γ-H2AX as primary antibody (Interchim) and anti-mouse IgG-FITC secondary antibody (Invitrogen). Cell nuclei were stained by incubation for 20 min with Hoechst 33342 (5 μg/mL). Foci were numbered and expressed as number of foci per cell nucleus.

Statistical analysis

Results are expressed as mean ± standard deviation of at least three assays. Statistical evaluation was performed with STATISTICA software. One-way Kruskal–Wallis ANOVA was used, followed by Mann–Whitney U-test. p < 0.05 was considered to be statistically significant.

Results and discussion

Physico-chemical characterization

The size and shape of each tested nanoparticle were analyzed from TEM images (Fig. 1).
Fig. 1

TEM observation of nanoparticles. Nanoparticles were deposited on TEM grids and directly observed. a TiO2-CEA, b TiO2-Degussa, c TiO2-Sigma, d Al2O3, e SiC, f CNT, g TiO2-Sigma-R

TiO2 nanoparticles diameters ranged from 12 to 140 nm, their shape was either spherical or elongated. As shown in Fig. 1, nanoparticles were slightly agglomerated and/or aggregated when dispersed in water. Still it has to be noted that the TEM preparation protocol enhances agglomeration since air-drying of a nanoparticle suspension leads to the formation of clusters. Figure 1 may thus overestimate nanoparticles agglomeration state. We consider from these images that nanoparticles are loosely agglomerated. This hypothesis was confirmed by photon correlation spectroscopy analysis, which characterized nanoparticles clusters as agglomerates having diameters in the range of 20–40 nm (Simon-Deckers et al. 2008). Physico-chemical characteristics of the nanoparticles we used are summarized in Table 1. From XRD analysis, crystalline phases of TiO2 nanoparticles were determined: TiO2-CEA were 95% anatase, TiO2-Degussa were nearly 90% anatase, TiO2-Sigma were 100% anatase, and TiO2-Sigma-R were 100% rutile (Simon-Deckers et al. 2008). Metal oxide nanoparticles purity was evaluated by energy dispersive spectroscopy (EDS) analysis after TEM observation. Only TiO2-Sigma-R nanoparticles contained impurities, composed of silicium. Depending on the duration of sonication, the length of CNT was controlled: it was reduced from the 400 μm initial length to 0.1–20 μm when sonicated during 30 mn (long CNT Fe) or to 0.1–5 μm when sonicated during 8 h (short CNT Fe). A fraction of the “long CNT Fe” batch was annealed, these purified CNT contained 0.08 wt% Fe (long CNT) while non-purified CNT contained 4.24 wt% Fe (Long CNT Fe). We thus had a panel of three CNT samples: Long CNT, Long CNT Fe, and Short CNT Fe.
Table 1

Nanoparticles characteristics


Supplier, ref.


Size TEM (nm)

SSA (m²/g)

Size BET (nm)


ζ (DMEM) (mV)

Crystalline phase


Degussa, Aeroxide® AluC


11 ± 3

83 ± 1







12 ± 3

92 ± 3




95% anatase


Degussa, Aeroxide®TiO2 P25


24 ± 7

46 ± 1




89% anatase


Sigma, T8141


142 ± 36

10 ± 1




100% anatase


Sigma, 637262


L: 68 ± 17

D: 9 ± 3

118 ± 2



100% rutile




17 ± 3

125 ± 5




Long CNT



L: 0.1–20

D: 8–177

42 ± 2



Short CNT



L: 0.1–5

D: 7–180

42 ± 2



Size was determined by TEM observation

SSA Specific surface area, measured according to Brunauer Emmett and Teller protocol, PZC point of zero charge, obtained by zeta potential measurements, L length, D diameter, n.a. non applicable

When prepared in cell culture medium in order to expose cells, nanoparticles agglomerated as 200–400 nm clusters except for TiO2-Sigma particles which agglomerated to >3 μm clusters (Simon-Deckers et al. 2008). This agglomeration can be explained by the fact that their point of zero charge (PZC) is close to cell culture medium pH, and by the salinity of this medium. Only CNT could be maintained in a correct dispersion state, thanks to the amendment of CNT initial aqueous suspension with gum Arabic.


Cytotoxicity of nanoparticles was first evaluated in NRK-52E renal cells (Fig. 2a–d). Cell mortality was not observed for exposure periods shorter than 24 h. After 48 h of exposure, cell mortality rapidly reached a plateau for concentrations higher than 20 or 40 μg/mL until concentrations as high as 200 μg/mL (data not shown). When assessed with MTT test, cell mortality induced by nanoparticles (Fig. 2a) was more important than cell mortality induced by CNT (Fig. 2b), with a maximal cell death rate of 35% (TiO2-Degussa and Al2O3) versus 23% (Long CNT Fe) when exposed to 50 μg/mL of these nanoparticles. When results were compared two by two at the exposure concentration of 40 μg/mL, this difference was statistically significant (p = 0.002) whatever metal oxide nanoparticle size and chemical composition and whatever CNT length and purity.
Fig. 2

Nanoparticles cytotoxicity. NRK-52E cells were exposed 48 h to nanoparticles (a, c) or carbon nanotubes (b, d), and cell mortality was assessed by MTT assay (a, b) or LDH assay (c, d). For MTT tests, results are expressed as a percentage difference compared to a negative control group. Whereas for LDH tests, results are expressed as a percentage difference compared to a positive control group

According to MTT assay results, nanoparticles physico-chemical characteristics affected their impact on cell viability (Fig. 2a). A statistically lower cytotoxicity was observed for the largest nanoparticles (TiO2-Sigma, 140 nm), as compared to the thinnest ones (TiO2-CEA, 12 nm, p = 0.004 at 80 μg/mL and TiO2-Degussa, 25 nm, p = 0.04 at 80 μg/mL). Nanoparticles cytotoxicity thus depended on nanoparticles size.

The cell death rate induced by Al2O3 nanoparticles (11 nm) was not statistically different from mortality induced by TiO2-CEA (12 nm) and TiO2-Degussa (24 nm). SiC nanoparticles (17 nm) induced less cell mortality than metal oxide nanoparticles. Nanoparticles cytotoxicity thus also depended on nanoparticle chemical composition, metal oxide nanoparticles being more toxic for NRK-52E cells than SiC nanoparticles.

The toxicological effect of TiO2 nanoparticles crystalline phase was also evaluated. TiO2-Sigma-R (100% rutile, 68 × 9 nm) nanoparticles induced a cell death rate comparable to the one obtained after exposure to TiO2-Sigma (100% anatase, 142 ± 36 nm) nanoparticles (data not shown). Considering the effect of the size of nanoparticles on cytotoxicity (the largest nanoparticles inducing the lowest cell death), we would have expected TiO2-Sigma-R nanoparticles to induce more cell death than TiO2-Sigma nanoparticles, which is not the case. This lesser important cytotoxic effect can be explained by the crystalline phase of TiO2 nanoparticles, rutile nanoparticles being less reactive than anatase nanoparticles. Moreover, TiO2-Sigma-R nanoparticles were also in elongated shape. We then cannot definitely conclude on the impact of crystalline phase on toxicity, since both nanoparticles shape and crystalline phase varied.

Among CNT, the long CNT containing Fe impurities induced more cell death than the short CNT containing Fe impurities (p = 0.004 at 50 μg/mL) and more cell death than the long, purified CNT (p = 0.002 at 50 μg/mL). These results suggest that, in the range of tested CNT lengths, CNT cytotoxicity depended on length and purity (Fig. 2b), the shortest and the purest CNT of our panel being the less toxic. It is thus important to precisely characterize CNT, in particular in terms of metallic impurities, in order to be able to evaluate the relative part of the observed toxicological effect that can be attributed to the presence of these impurities. Regarding metallic impurities in carbon nanotubes, some studies recently came out, dealing with this problem (Ge et al. 2008). The authors underlined the need to establish standard methods for the quantitative analysis of metal impurities. Such recommendations should thus be taken into account in experimental design.

Lactate dehydrogenase assay gave complementary results (Fig. 2c, d): observations concerning the influence of physico-chemical characteristics on cytotoxicity were similar, but cytotoxicity of metal oxide nanoparticles was significantly lower than when assessed with MTT test (Fig. 2c). Conversely, CNT cytotoxicity was higher than when assessed with MTT test (Fig. 2d). LDH assay gives information on membrane impairment and MTT assay gives information on cell mitochondrial enzymes activity. These results suggest that CNT impair cell membranes more than cell metabolic activity, which can be caused by a mechanical action of CNT on these membranes. Inversely, metal oxide nanoparticles may enter the cell without damaging membranes, and cause a metabolic response because of their accumulation in cell organites.

Lung, liver, and kidney cell lines were then used to identify differences of sensitivity to TiO2 nanoparticles and CNT (Table 2).
Table 2

Sensitivity to TiO2 nanoparticles and CNT of lung, liver, and kidney cells

Cell line

TiO2-CEA (%)

Long CNT (%)



















Cell mortality induced in liver, kidney, and lung cell models exposed to 100 μg/mL of nanoparticles for 48 h. Mortality is expressed as percentage of negative control, i.e., unexposed cells

In our conditions, liver and lung cells were more sensitive to carbon nanotubes than renal cells. This higher sensitivity of lung cells to carbon nanotubes can be explained by their ability to secrete lung surfactant. This substance, covering cell surface, could trap CNT at the vicinity of cells and then increase the duration of contact between cells and CNT, thus enhancing cell membrane alterations. Moreover, liver cells are known to secrete most of the plasma proteins, which can also trap carbon nanotubes at the vicinity of cell membranes (Yamashita et al. 2007). Conversely, renal cells were more sensitive to TiO2 nanoparticles than liver and lung cells. This can be explained by the role of proximal cells, specialized in uptaking, and recycling molecules from kidney tubules. These cells may thus have a higher uptake activity than lung and liver cells, permitting to nanoparticles to be more dynamically internalized in cells, and consequently to find rapidly a way to intracellular organites. This observation also suggests that the uptake pathway and mechanisms underlying nanoparticles and carbon nanotubes cytotoxicity may be different. However, it must be noticed that the present results were obtained via an in vitro set of experiments. It would be inconsistent to directly extrapolate potential consequences of in vivo exposure scenarios from this dataset. Moreover, it has to be noted that the nanoparticles concentrations required to observe this impact are very important, and would not represent the reality of an environmental contamination. However, if nanoparticles are locally accumulated in an organ and not eliminated, such concentrations would be locally reached, and such cell response would potentially be observed.

Intracellular accumulation

Intracellular accumulation of nanoparticles was observed by TEM. All the nanoparticles we tested were uptaken by cells. Figure 3 represents the results obtained after exposure of NRK-52E kidney cells (Fig. 3a) or A549 lung cells (Fig. 3b, c) to TiO2-CEA nanoparticles or CNT. TiO2 nanoparticles uptake was observed in the three tested cell types. Whatever the considered cell type, nanoparticles were localized in the cytoplasm, either in vesicles or isolated. It was striking to note that even if TiO2-CEA nanoparticles were agglomerated into 200–400 nm clusters in exposure medium, inside cells these clusters seemed to be smaller. As already described, clusters could have been dissociated at the vicinity of cells or after uptake (Soto et al. 2007). Intracellular CNT accumulation was also observed, but only the shortest ones were accumulated (Simon-Deckers et al. 2008).
Fig. 3

Nanoparticles intracellular accumulation. Nanoparticles internalization in NRK-52E (a), and A549 (b, c) after exposure to 50 μg/mL of TiO2-CEA nanoparticles (ab) for 24 h (a) or 48 h (b) or 10 μg/mL of Long CNT Fe for 48 h (c). Nanoparticles localizations are indicated by arrows. C: Cytoplasmic compartment, N cell nucleus, E extracellular domain

Nanoparticles intracellular accumulation mechanisms have not been rationally studied and are yet not known. Several endocytosis pathways exist, leading to the internalization of cargoes differing in their size and composition. The route of uptake of 50–200 nm particles is presumed to be clathrin-mediated endocytosis, while 200–1,000 nm particles would preferentially taken up by caveolae-mediated endocytosis (Rejman et al. 2004), or by macropinocytosis. If macropinocytosis has been observed constitutively in several cell types, it is primarily a response to cell stimulation by growth factors (Jones 2007). The size of intracellular vesicles containing nanoparticles which were observed in this TEM study (more than 200 nm) would rather let us think that nanoparticles entered cells via macropinocytosis or caveolae-mediated endocytosis, as hypothesised in other studies (Stearns et al. 2001; Singh et al. 2007). However, the presence of isolated nanoparticles in cell cytoplasms suggest that it was not the only mechanism involved in their uptake, and that they could find their way to cell cytoplasms through multiple pathways in the fluid phase and via non-specific adsorptive endocytosis, or by direct diffusion through plasma membrane like some viruses do (Marsh and Pelchen-Matthews 2000). Since these isolated nanoparticles were mostly present in cells after 4 h of exposure, and more rarely after 24 h of exposure, we rather think that multiple uptake mechanisms co-exist and result in the entrapment of nanoparticles in large cytoplasmic vesicles. This entrapment would be a mean for cells to counteract the intracytoplasmic presence of nanoparticles, this sequestration phenomenon allowing cells to avoid interaction with vital organelles or cytoplasmic processes. We did not observe any evidence of exocytosis of these large vesicles or of isolated nanoparticles.

Oxidative response and genotoxicity

Since TiO2 anatase nanoparticles are known to be highly reactive and able to promote oxido-reduction reactions, and since the Fe contained in carbon nanotubes would possibly initiate Fenton or Fenton-like oxido-reduction reactions, we then measured reactive oxygen generation inside cells after exposure to these nanoparticles.

An elevation of free radicals concentration was measured in NRK-52E cells exposed for 24 h to nanoparticles (Fig. 4). This elevation was statistically significant except for low concentrations of CNT containing Fe impurities, and SiC nanoparticles at all the tested concentrations. TiO2, Al2O3 nanoparticles, and CNT thus generate an oxidative stress in cells.
Fig. 4

ROS production induced by nanoparticles in NRK-52E cells. Cells were exposed for 24 h to 1–200 μg/mL of nanoparticles. ROS production was evaluated by H2DCF-DA assay. Fluorescence was normalized by total protein content, and expressed as percentage of the negative control, i.e., non exposed cells

Contrary to results obtained in cytotoxicity assays (MTT and LDH assays), cellular ROS production consecutive to nanoparticles exposure did not depend on the size of these materials. Indeed, ROS generation observed after exposure to TiO2-Sigma (140 nm) nanoparticles tended to be higher than ROS generation induced by TiO2-CEA (12 nm), TiO2-Degussa (24 nm), and Al2O3 (11 nm) nanoparticles exposure, but this difference was not statistically significant. This absence of size-dependence in ROS production could be interpreted as a proof that cell death is not directly due to intracellular oxidative stress. Indeed, same levels of intracellular ROS can be linked either with high or low cytotoxic effects. We thus can conclude that oxidative stress is not the only mechanism of nanoparticles toxic action that finally leads to cell death.

ROS production was dependent on nanoparticle chemical composition, since SiC nanoparticles (17 nm) did not induce a statistically significant elevation of intracellular ROS generation whereas 12–25 nm metal oxide nanoparticles did. This chemical dependence can let us draw the hypothesis of a chemical source of ROS generation, with potentially redox reactions.

Carbon nanotubes induced the same level of ROS production whatever their length and purity. If we consider the fact that CNT are less accumulated than metal oxide nanoparticles, we can suggest that intracellular ROS production induced by CNT exposure refers to more complex mechanisms than simple redox reactions. This can explain the fact that iron impurities are not involved in ROS overproduction.

Toxicological studies have already been carried out to assess the ability of nanoparticles to induce cellular overproduction of reactive species in biological samples. Some of them succeeded in demonstrating such kind of toxicological effects after nanoparticles exposure. Indeed, Wang et al. (2008b) showed that oxidative stress occurred in whole brain of mice intranasally instilled with TiO2 nanoparticles every other day during a 30-day exposure period experiment. It means that nanoparticles were translocated to the animal’s brain where they caused oxidative damage. The authors notably reported abnormal lipid peroxidation, protein oxidation, and increased activity of catalase (an antioxidant enzyme). The present experiment proves that if nanoparticles are translocated through epithelial barriers and reach kidneys, they may cause oxidative damage to proximal tubule cells.


Nanoparticles were rarely observed in cell nuclei; however, DNA damage can result from oxidative reactions caused in cell cytoplasm by some toxicants, which find echoes in cell nuclei. Since exposure to nanoparticles caused intracytoplasmic ROS production, their possible induction of damage to DNA was assessed by two complementary techniques: the alkaline comet assay and the immunostaining of γ-H2AX foci (Fig. 5).
Fig. 5

DNA damage induced by nanoparticles in NRK-52E cells. After exposure to 20-200 μg/mL of nanoparticles for 24 h, alkaline comet assay (a) and γ-H2AX foci immunostaining (b) methods were applied to cells. Results are compared to negative control (Ctl−), i.e., unexposed cells, or positive control (Ctl+), i.e., cells exposed to 50 μM of etoposide for 24 h

The alkaline comet assay is an indicator of DNA single and double strand breaks, and alkali-labile sites. It was performed on cells exposed to the most cytotoxic nanoparticles (Fig. 5a), i.e., one type of small and spherical TiO2 nanoparticles (TiO2-CEA was chosen) and CNT. For short CNT and SiC, no significant effect on DNA integrity was observed (not shown).

DNA damage caused by exposure to long CNT was of low intensity; however, a statistically significant response was observed in the comet assay after exposure to high (100 μg/mL) concentrations. DNA damage induced by TiO2-CEA nanoparticles was much more important, increased with exposure concentration and was statistically significant for exposure concentrations equal or higher than 100 μg/mL. Since ROS level after cell exposure to 50, 100, and 200 μg/mL of TiO2-CEA are quite similar, oxidative stress may not be the unique cause of DNA damage. One hypothesis would be that DNA repair processes might also be disturbed when nanoparticle uptake increases.

Among DNA damage produced by nanoparticles exposure, double strand breaks were searched for, by monitoring γ-H2AX foci expression. H2AX histone is phosphorylated to the γ-H2AX form through constitutive processes such as oxidative metabolism or stalled replication forks, or in response to agents that introduce DNA double strand breaks (Park et al. 2003). Exposure to TiO2, Al2O3 nanoparticles or CNT did not significantly increase the amount of γ-H2AX foci in NRK-52E cell nuclei (Fig. 5b). Same results were obtained with SiC nanoparticles (data not shown).

These data suggest that nanoparticles are genotoxic, causing the formation of single strand breaks or alkali-labile sites, but not double strand breaks. An in-deep study of the nature of these DNA damage would now permit to identify the mechanisms leading to nanoparticles genotoxicity.


In this study preliminary results of nanoparticles and carbon nanotubes cytotoxicity, genotoxicity and oxidative effects on NRK-52E cells are presented Furthermore, cytotoxicity results obtained on different cell types (liver, kidney, and lung cell lines) are compared.

Results show that TiO2, Al2O3, SiC nanoparticles, and multi-walled carbon nanotubes induced a significant cell death. Among nanoparticles, the smallest, anatase, and spherical nanoparticles induced the highest cytotoxic effects. Cytotoxic effects of CNT did not depend on their length and purity. These toxicological effects were confirmed by the use of a second cytotoxicity assay test, which also permitted to highlight two routes of cell death induction: membrane damage, which was the major way of CNT cell injury, and mitochondrial activity impairment which was the mode of action of metal oxide and SiC nanoparticles. Cell mortality induced by nanoparticles and CNT was cell line-dependent. All these nanoparticles, except the SiC nanoparticle batch used here, induced the generation of reactive oxygen species in cells. Since we could not draw a direct parallel between ROS level and cytotoxicity occurrence, this oxidative stress may not be the only mechanism explaining cytotoxicity. Moreover, DNA damage was revealed consecutive to nanoparticles exposure. This damage was also probably due to ROS intracellular generation, but again no direct correlation between the level of ROS and the intensity of DNA damage could be done. For this reason more complex mechanisms need to be pointed in order to have a clear view on nanoparticles mode of action after intracellular accumulation.


This work was supported by ADEME (French Environment and Energy Management Agency), by the Région Ile-de-France in the framework of C’nano IdF. C’Nano-IdF is the nanoscience competence center of Paris Region, supported by CNRS, CEA, MESR and Région Ile-de-France. It was also funded by the French National Research Agency (ANR) and the AFSSET (the French Agency for Environmental and Occupational Health Safety).

Copyright information

© Springer Science+Business Media B.V. 2009