Genotoxic Impact of Aluminum-containing Nanomaterials in Human Intestinal and Hepatic Cells.

Background: Exposure of consumers to aluminum-containing nanomaterials (Al NMs) through numerous products is an area of concern for public health agencies since human health risks are not completely elucidated. In addition, the available data on the genotoxicity of Al 2 O 3 and Al 0 NMs are inconclusive or rare. In order to provide further information, the present study investigated the in vitro genotoxic potential of Al 0 and Al 2 O 3 NMs in intestinal and liver cell models since these tissues represent organs which would be in direct contact or could experience potential accumulation following oral exposure. Methods: Differentiated human intestinal Caco-2 and hepatic HepaRG cells were exposed to Al 0 and Al 2 O 3 NMs (0.03 to 80 µg/cm2) and the results were compared with those obtained with the ionic form AlCl 3 . Several methods, including (cid:0) H2AX labelling, the alkaline comet assay and micronucleus (MN) assays were used. Oxidative stress and oxidative DNA damage were assessed using High Content Analysis (HCA) and the formamidopyrimidine DNA-glycosylase -modied comet assay respectively. Moreover, carcinogenic properties of Al NMs were investigated through the cell transforming assay (CTA) in Bhas 42 cells. Results: The three forms of Al did not induce chromosomal damage when tested in the MN assay. Furthermore, no cell transformation was observed in Bhas 42 cells. However, although no production of oxidative stress was detected in HCA assays, Al 2 O 3 NMs induced oxidative DNA damage in Caco-2 cells in the comet assay following a 24 h treatment. Considerable DNA damage was observed with Al 0 NMs in both cell lines in the comet assay, although this was likely due to interference with these NMs. Finally, no genotoxic effects were observed with AlCl 3 . Conclusion: The slight effects observed with Al NMs are therefore not likely to be related to ion release in the cell media. polydispersity in cell µg/ml. Three mean ± Al0, salt control MMS used as a positive control (25 µg/ml for Caco-2 cells and 30 µg/ml for HepaRG cells). Results are presented as means (±SEM) of the percentage of binucleated micronucleated cells (BNMN) scored from 1000 binucleated cells per slide. Two slides per concentration were scored per experiment. Viability was calculated by the replicative index (RI). Each concentration was tested in duplicate, n= 3. The percentages of BNMN cells were compared using the one-way Pearson chi-square test.***p<0.01.

contributions from products used in food (additives, contact materials) and in cosmetics, and concluded that adolescents were highly exposed [13].
Few studies on the genotoxicity of nanoscale forms of Al following oral ingestion have been performed, and most of the published literature has focused on Al 2 O 3 NMs only. DNA damage was reported in erythrocytes of rats after a single oral treatment with Al 2 O 3 NMs, although at high doses (≥1,000 mg/kg) [14,15]. Genotoxic effects were observed in bone marrow, but not in other organs, after a short-term treatement with lower doses of Al 2 O 3 NMs [16]. In vivo effects of Al 0 NMs following oral exposure are mostly lacking, although one study suggested cross-linking effects on DNA in the duodenum of rats [16]. Following oral exposure of rodents with ionic forms of Al, an increase in MN frequency was reported in bone marrow after a single oral administration [17] and in liver after a 30 day oral treatment [18].
Nevertheless, the induction of MN formation in liver was shown to decrease with an antioxidant treatment [18,19]. Consistent with these results, a slight oxidative DNA damage was observed in blood after a shortterm oral exposure [16].
The in vitro genotoxicity of Al 2 O 3 NMs has been assessed in several mammalian cell lines including human peripheral lymphocytes [20], primary human broblasts [21], hepatic HepG2 cells [22], and Chinese hamster ovary cells [23]. While some studies have not observed genotoxic effects of Al 2 O 3 NMs [20,24,25,26], others have reported a positive response [21,22] which may be associated with oxidative damage [22]. In contrast, no data on the in vitro genotoxicity of Al 0 NMs has been published so far, and only some cytotoxicity was detected in rat alveolar macrophages treated with Al 0 above 100 µg/ml [27].
For the salt AlCl 3, DNA damage has been reported in human peripheral blood lymphocytes, with positive results in micronucleus and chromosomal aberration tests, as well as in the comet assay [17,28,29,30].
According to an ECHA safety assessment [31], the data available on the genotoxicity of Al 2 O 3 NMs are inconclusive while few data on the genotoxicity of Al 0 NMs has been published so far. In addition to the direct contact of Al NMs present in food with the intestinal epithelium, Al accumulation in liver has been shown after oral exposure with Al 2 O 3 NMs [15,32,33].
Therefore, the aim of the current study was to evaluate the in vitro genotoxic potential of Al 0 and Al 2 O 3 NMs in two relevant human cell models of intestine and liver. Several endpoints of genotoxicity were investigated using the alkaline and Fpg-modi ed comet assays which detects DNA breakage including oxidative lesions, DNA double strand breaks were detected through phosphorylated histone H2AX (γH2AX), and the micronucleus assay which determines chromosome and genome damage. Furthermore, the capacity of aluminum-containing NMs to initiate or promote carcinogenesis was assessed by the Cell Transforming Assay (CTA) in Bhas-42 cells.
As these NMs can potentially dissolve in the dispersion solution or in media, the genotoxicity was compared to that of the metal salt AlCl 3 . Moreover, the interference of NMs, including with Al-NMs [34], has been demonstrated in numerous publications using a wide range of biological assays, and stresses the necessity to evaluate interference in order to assess the potential effect on the results [35,36,37,38]. In this study, various sources of interference have been taken into account within the different assays.

Dispersion and characterization of NMs
Al 0 , Al 2 O 3 and ZnO NMs with a similar primary particle size were supplied from IoLiTec (Heilbronn, Germany). NM characteristics as provided by the supplier are presented in Table 1. AlCl 3 (hexahydrate) was purchased from Sigma Aldrich (Saint Louis, USA). NM dispersion was performed according to the NANOGENOTOX protocol [39], as described in [16].
The morphology and agglomeration of Al 0 and Al 2 O 3 NMs in the stock dispersion solution and in cell media were determined by transmission electron microscopy (TEM) (Figure S 1). For the characterization of NMs from stock solutions, TEM grids were prepared immediately after sonication and dilution (100 µg/mL) in the stock dispersion solution. For the characterization of NMs in cell culture media (DMEM +10% FBS and William's Medium +5% FBS), the samples were diluted with distilled water to 1.2 µg/mL prior to grid preparation. The TEM grids were prepared by deposition of a carbon-coated copper grid onto a drop of the stock solution for 20 s to allow adsorption of the NMs and were observed with an electron microscope (JEOL 1400 operated at 120 kV and coupled with a 2k-2k camera from Gatan (Orius 1000)).
The hydrodynamic diameter of Al 0 and Al 2 O 3 NMs were measured using a Malvern Zetasizer (Malvern Instruments, Malvern, UK) equipped with a 633-nm laser diode operating at an angle of 173°. To assess the stability of NM suspensions, following NM dispersion, samples were diluted to a nal concentration of 100 µg/ml in the stock dispersion solution or in cell media and measurements were performed at 0 and 24 h. The samples were equilibrated at 25 °C for 120 s prior to measurement. Ten repeated measurements for each sample were performed in 3 independent experiments. The mean hydrodynamic diameter Z ave was determined using cumulant analysis.

Cell culture and treatment
The human colorectal adenocarcinoma Caco-2 cell line was cultured (passages 25-38) until differentiation after 21 days as described in [40] including for cell seeding in various plate formats depending on the assay performed. Simarly, HepaRG cells (passages [13][14][15][16][17][18][19] were cultured and seeded for the various assays as previously described [40,41]. Differentiated Caco-2 and HepaRG cells were treated for 24 h with Al 0 and Al 2 O 3 NMs at concentrations ranging from 0.03 to 80 µg/cm 2 and with AlCl 3 as ionic salt control at 90 and 128 µg.mL -1 in DMEM + 10% FBS or William's medium + 5% FBS respectively. For some assays, ZnO NMs at concentrations from 1.5 to 6 µg/cm 2 were used as a positive NM control. Equivalence between volume concentration (µg/mL) and surface concentration (µg/cm 2 ) are shown in Table S 1B. Al content corresponding to the concentrations of Al-containing NMs and AlCl 3 that were used are summarized in Table S 1B.

Kinetics of nanoparticle sedimentation
The colloidal characterization of the suspended nanomaterials in the conditions of cellular uptake assay was achieved using the volumetric sedimentation method (VCM) as reported in DeLoid et al [42]. We rst measured the volume of the potentially agglomerated NM in DMEM and Williams media, at a NM concentration of 250 µg.mL -1 , using a speci c centrifugal tube and ruler device. From the measured pellet, the effective density ( eff ) is calculated using the following equation: Where m is the density of the medium in g.cm -3 , NP is the density of NP (2.7 g.cm -3 for Al and 3.95 for Al 2 O 3 ), M NP the total mass of NM in 1 mL of dispended volume and V the measured volume pellet. SF is a stacking factor and was set to 0.634, which generally is appropriate for random stacking. The loss of mass of NMs from ion release was estimated to be lower than 1% and was neglected in the density calculation. The viscosity of the cell culture media at 37°C was determined using a Nanoparticle Tracking Analysis device (Malvern Instrument) by measuring the apparent hydrodynamic radius of 400 nm standard particles in the media. Finally, the kinetics of sedimentation was calculated using the distorded grid (DG) model available from DeLoid et al [42]. The size of the NPS was taken from Table 2 (Z ave ).
Other model parameters are h=3.1mm (liquid column height), initial NM concentration : 0.250 mg.mL -1 , the dissolution and cell-NMs stickiness are neglected (parameters set to 0).

Ion release from NMs
Following the dispersion of Al 0 and Al 2 O 3 NMs, suspensions were diluted in stock solution (ultra pure water + 0.05 % BSA) or cell culture media (DMEM +10% FBS and William's Medium +5% FBS) at concentrations of 25, 50 and 100 µg/mL. After 24 h, ion release from NMs was determined by ultracentrifugation at 16,000 g for 1 h at 4°C (Hettich Zentrifuge Mikro 220R). The supernatants were processed through acidic hydrolysis (69% HNO 3 , 180°C for 20 min in an MLS-ETHOS Microwave system) before detection of Al species with a quadrupole Inductively Coupled Plasma Mass Spectrometry (ICP-MS) (iCAP Q, Thermo Fisher Scienti c GmbH, Dreieich, Germany) equipped with a PFA ST Nebulizer, a quartz cyclonic spray chamber and a 2.5 mm quartz injector (Thermo Fisher Scienti c). The gas ows were set to 14 L/min, and 0.65 L/min for the cool gas (Ar) and the auxiliary gas (Ar) respectively. The ow rate of the sample was 0.39 mL/min. Results are given as percentage of the initial Al amount.

Uptake observations by TEM
Following a 24 h treatment, cells were xed by glutaraldehyde (2.5%) and embedded in DMP30-epon before cutting ultra-thin sections (90 nm) for TEM observation as described in [40].

Cellular imaging and High Content Analysis (HCA)
After 24 h treatment with Al NMs and AlCl 3 , plates were processed for HCA with an ArrayScan VTI HCS Reader (Thermo Scienti c, Waltham, USA) as described in [40]. Cell numbers were determined from DAPI staining, active caspase-3 was quanti ed in the total cell compartment and H2AX in cell nuclei.
Oxidative stress was measured using CellROX Deep Red Reagent (Fisher Scienti c, Illkirch, France). Brie y, cells were pre-incubated for 1 h with 5 M CellROX in serum-free media and washed twice with PBS before treatment with NMs and AlCl 3 . After 24 h and twice washing with PBS, cells were incubated with 3 M Hoescht 33342 for 20 min at 37°C. Cells were then washed twice with PBS and were scanned and analyzed using the Compartimental Analysis module of the Bioapplication software. For each well, images from 7 elds (20 × magni cation) were analyzed for quanti cation of uorescence at 647 nm.

Comet assay
After a 5 h ( Figure S 3) or 24 h treatment with Al NMs and AlCl 3 , the comet assay was performed as described in [40,43]. The individual tail intensity of at least 50 cells per slide were analyzed using the Comet Assay IV software (Perceptive Instruments, Haverhill, UK). Cells were considered as hedgehogs when DNA damage was too high to score. At least three independent experiments were performed. Methyl methanesulfonate (MMS) was used as positive control.
The level of oxidized bases was determined with the modi ed comet assay using the bacterial DNA repair enzyme Fpg through the formation of single-strand breaks (SSB) induced by the excision of oxidized purines [44,45]. Some additional steps to the protocol described above were performed such as incubation with enzyme buffer (0.1 M KCl, 0.2 mM EDTA, 40 mM HEPES, 0.2 mg/ml BSA) after lysis.
Two slides, one incubated with enzyme buffer (control slide) and the other with 9 U/slide Fpg at 37°C for 30 min, were then processed as described previously.

Particle interaction with DNA during the comet assay
The interaction of NMs with DNA migration during the comet assay was evaluated as described previously [35,40]. Brie y, dilutions of Al 0 or Al 2 O 3 NMs in 0.5% low-melting point agarose (LMP) were prepared at nal concentrations of 28 and 128 µg/mL (corresponding to 9 and 40 µg/cm 2 conditions). After trypsinization and centrifugation (2 min, 136 g), untreated Caco-2 and HepaRG cells were resuspended in the LMP/NM mixture, loaded on pre-coated slides and processed in the alkaline comet assay as previously described, in the presence or absence of Fpg. A negative control consisting of untreated cells in LMP-agarose in the absence of NMs was performed in order to compare the results.

Cytokinesis-block micronucleus assay (CBMN)
The CBMN assay was performed as described in [40] according to the guideline n°487 of the Organization for Economic Co-operation and Development (OECD) [46]. After staining of the slides with acridine orange (100 μg/mL), at least 1000 binucleated cells per slide were scored. Three independent experiments were carried outand each concentration was tested in duplicate. The replication index (RI) was calculated using the formula recommended by OCDE guideline n°487. MMS and ZnO NM were used as positive controls.

Bhas 42 Cell Transformation Assay (CTA)
Originally established from the v-Ha-ras-transfected BALB/c 3T3 cells by Sasaki et al [47], Bhas 42 cells used in this study (passage 23) were obtained from Harlan Laboratories (Rossdorf, Germany). Both the CTA and concurrent cell growth assays were performed in their 6-well format and in accordance with a guidance document produced by the OECD [48], with some modi cations. The protocol, including both an initiation and a promotion assay, was previously described by Fontana et al [49].
In the initiation assay, 24 h after seeding (420 cells/cm 2 ) (Day 1), the cells were treated with Al NMs and AlCl 3 for 72 h (Day 4). Then, the cells were cultivated in fresh medium until Day 21, with medium changes on Day 7, Day 10 and Day 14. MCA (1 µg/mL) was used as positive control.
In the promotion assay, the cells were seeded (1,500 cells/cm 2 ) and cultured for 4 days without changing the media. On Day 4, 7, and 10, the culture medium was replaced with fresh media containing Al NMs or AlCl 3 . The treatment continued until Day 14. The cells were then cultured in fresh medium in the absence of NMs until Day 21. TPA (0.05 µg/mL) was used as positive control.
In both assays, the cells were xed with ethanol on Day 21 and stained with a 5 % Giemsa solution. The morphological criteria recommended by OECD were followed for the evaluation of transformed foci. The mean of the number of transformed foci was calculated from six replicate wells.
Cell growth assays in both the initiation and promotion conditions were performed on Day 7 using three replicate wells for each condition. The cells were xed in 4% formaldehyde and stained with 1 µg/mL DAPI. The number of cell in wells was determined by automated microscopy with an Arrayscan VTi using the Target Activation module of the BioApplication software. The relative cell growth (%) was calculated as follows: (number of cells in treated cultures / number of cell in control cultures) x100.

Statistical analysis
The statistical signi cance of HCA results was tested using one-way Analysis of variance (ANOVA) followed by Dunnett's post-hoc tests with GraphPad Prism 5.
For the comet assay, the one-way Analysis of variance (ANOVA) was used followed by Dunnett's post-hoc test.
For the micronucleus assay, the percentages of micronucleated cells in treated and solvent control cultures were compared using the one-way Pearson chi-square test.
For the CTA, data were statistically analysed by multiple comparison using the one-sided Dunnett's test (p<0.05, upper-sided). The signi cance of the positive controls (MCA and TPA) was evaluated relative to the control (p < 0.05) by the one-sided Student's t-test.

Nanomaterial characterization
Information concerning the physico-chemical characterization, including the morphology, primary size, surface speci c area (SSA), purity and density of the Al 0 , Al 2 O 3 and ZnO NMs used in this study are provided in Table 1. However, in contrast to the information provided by the suppliers, the particle morphology of Al 2 O 3 NMs in the stock dispersion solution cannot be considered as being spherical, but rather have a rod-like shape when observed by TEM ( Figure S1). Although Al 0 particles exhibit a spherical shape, numerous elongated protrusions are also observed ( Figure S1). Therefore, values of "average particle size" (Table 1) should then be considered with caution. Due to the drying step for preparation of TEM grids, the crystallization of different components of the culture media did not allow a proper characterization of the morphology of Al NMs in cell culture media (data not shown).
Particle hydrodynamic diameter and stability in the stock dispersion solution, as well as in cell media, were assessed by DLS immediately (0 h), as well as after 24 h ( Table 2). The hydrodynamic diameters of Al 0 , Al 2 O 3 and ZnO NMs in the dispersion stock solution were 254 ± 4 nm for Al 0 , 168 ± 3 nm for Al 2 O 3 and 233 ± 11 nm for ZnO immediately following dispersion and were stable over time. The stability over time of these NMs was also observed in cell media (Table 2)

Ion release in stock solution and cell culture media
Ion release from Al NMs was investigated using ICP-MS (Table 3) and results are presented as percentages with respect to the initial concentration of aluminum. A decrease in the percentage of ion release from Al 0 NMs was observed with increasing NMs concentrations in both the stock dispersion solution (1.30 % at 25 µg/mL and 0.48 % at 100 µg/mL) and in media (3.88 % to 0.95 % in DMEM, and 2.42 % to 0.68 % in William's for 25 and 100 µg/mL respectively). Nevertheless, ion release from Al 0 NMs was slightly higher in media when compared to the dispersion stock solution. A concentration-dependent decrease in ion release was also observed for ZnO NMs (Table 3). Ion release was also higher in media compared to dispersion stock solution.
In contrast to Al 0 and ZnO NMs, for Al 2 O 3 NMs, the percentage of ion release with respect to the initial concentration was very low, relatively stable and independent of the NM concentration, although ion release was slightly higher in cell media.
The level of ions from AlCl 3 solutions was stable and independent of the concentration in the stock solution, but decreased with increasing concentration in media. This decrease of ion concentration with higher concentrations is likely due to the precipitate formed by AlCl 3 in cell media.

Uptake of Al NMs in Caco-2 and HepaRG cells
The uptake and the intracellular distribution of Al 0 and Al 2 O 3 NMs following a 24 h treatment in Caco-2 and HepaRG cells were investigated by TEM (Figure 2 and 3). In both cell lines, the majority of Al 0 NMs were found as dense agglomerates of various sizes in the cytoplasm embedded in electron lucent or dense vesicles which are likely endosomes and lysosomes (Figure 2 and 3 NMs was also seen, this occured less frequently than for Al 0 NMs (Figure 2

Cytotoxicity
Viability and apoptosis in Caco-2 and HepaRG cells following a 24 h treatment with Al NMs were investigated by cell counts (Figure 4 A)

Comet assay
The potential for Al 0 and Al 2 O 3 NMs to induce DNA damage in Caco-2 and HepaRG cells was investigated with the alkaline comet assay after a 24 h treatment (Figure 7 A and B). A modi ed comet assay with the Fpg enzyme was also performed to detect oxidative DNA damage (Figure 7 C and D).
In Caco-2 cells, a signi cant increase in tail DNA was observed with Al 0 NMs from 28 to 80 µg/cm 2 in the alkaline comet assay (Figure 7 A). In contrast, neither Al 2 O 3 and ZnO NMs, or the ionic salt control AlCl 3 induced any signi cant increase in tail DNA. In the Fpg-modi ed comet assay, a signi cant increase in tail DNA was observed in cells treated with Al 2 O 3 NMs at 3, 9 and 80 µg/cm 2 (Figure 7 C).

In HepaRG cells, tail DNA signi cantly increased in a dose-dependent manner in cells treated with Al 0
NMs, including a very considerable effect starting at 28 µg/cm 2 . In contrast, no effect was observed for cells treated with Al 2 O 3 and ZnO NMs, or the ionic control AlCl 3 (Figure 7 B). Similarly, an increase in tail DNA in the Fpg-modi ed comet assay was observed for Al 0 NMs at all concentrations tested with a very signi cant effect observed at concentrations above 9 µg/cm 2 . No signi cant changes in tail DNA were detected in HepaRG cells treated with Al 2 O 3 and ZnO NMs or AlCl 3 (Figure 7 D).
DNA damage in Caco-2 and HepaRG was also investigated by the alkaline comet assay after a 5 h treatment ( Figure S 3). In Caco-2 cells, a concentration-dependent increase in tail DNA was observed in cells treated with Al 0 NMs from 28 to 80 µg/cm 2

. No effect was detected in cells treated with Al 2 O 3 and
ZnO NMs, or the ionic salt control AlCl 3 . In HepaRG cells, a concentration-dependent increase was also observed with Al 0 NMs from 9 to 80 µg/cm 2 . In the Fpg-modi ed comet assay, an increase in tail DNA was detected in Caco-2 cells treated with Al 0 NMs from 9 to 80 µg/cm 2 and with Al 2 O 3 NMs at 3, 9 and 28 µg/cm 2 . Interestingly, in HepaRG cells treated for 5 h with Al NMs, results from the Fpg-modi ed comet assay showed that at every concentration tested, only hedgehogs were observed for all NMs (data not shown).

Interaction of NMs with DNA during the comet assay
The interference of NMs with the comet assay was assessed ( Figure 8) according to the protocol described by Bessa et al [35]. Compared to the untreated control, a concentration-dependent increase in % tail DNA was observed when Al 0 NMs are added at nal concentrations of 9 and 40 µg/cm 2 . A similar effect was also observed when Fpg was included in the assay. Compared to the negative control, no difference was detected for Al 2 O 3 NMs in the absence of Fpg, while a slight increase was observed with Fpg.

Micronucleus assay
In order to evaluate chromosome damage, the cytokinesis-block micronucleus assay was performed in Caco-2 and HepaRG cells treated for 24 h (Table 4)

Discussion
Exposure of the general population to NMs present in consumer products, including food, has increased dramatically within the last decade, and a thorough evaluation of the potential adverse effects resulting from exposure to NMs following ingestion is necessary. Among the toxic effects of Al-containing NMs that have been shown in several studies, genetic damage is of particular concern [15,21,22,23]. Both intestine and liver are considered key organs for investigating genotoxic effect of nanomaterials found in food since they represent the main organ of contact and the main organ of accumulation, respectively.
Nevertheless, in our recent in vivo study investigating the genotoxicity of Al NMs, only a very limited genotoxic response was observed. In fact, only a cross-linking effect was suggested in the rat duodenum with Al 0 NMs [16]. As the in vivo treatment duration was rather short (3 administrations over 2 days), and that it cannot be excluded that the level of NMs in the organs would be low, we chose to complete our study by investigating the in vitro genotoxicity of Al NMs in human intestinal Caco-2 and hepatic HepaRG cells using complementary tests.
Despite the uptake and presence of Al NMs in Caco-2 and HepaRG cells, no cytotoxicity or apoptotic response was observed following treatment with Al 2 O 3 NMs. Our results are in agreement with data from various publications that have reported little or no cytotoxicity in various cell lines [21,27,34,50,51], including in Caco-2 cells [52,53] and HepG2 cells [22].
No induction of chromosomal damage was observed in the micronucleus assay in either Caco-2 or HepaRG cells exposed to Al 2 O 3 NMs. Moreover, we did not observe a transforming activity in the CBA assay, supporting the absence of mutagenic potential for Al 2 O 3 NMs. Our results are consistent with two recent studies that reported a negative response in the chromosomal aberration and the micronucleus assays in human lymphocytes treated with Al 2 O 3 NMs with a smaller size (3 to 4 nm) than the one used in this study (20 nm), and for a longer incubation time (72 h) [20,54]. In contrast, other studies have reported an increase in micronucleus formation following a 24 h treatment with Al 2 O 3 NMs in other cell lines, including CHO cells [23], human broblasts [21] and RAW264 murine macrophages [24]. Interestingly, Al 2 O 3 NMs were shown to inhibit the replication e ciency of high-delity DNA polymerase [55]. Nevertheless, such inhibition did not affect the mutation rate at the single nucleotide level of replication products compared to controls [55]. Further investigation demonstrated that Al 2 O 3 NMs did not induce a clastogenic effect but rather chromosome loss and polyploidy, although these effects were observed only at one concentration [21]. An aneugenic effect of Al NMs was not observed in our study (data not shown). The discrepancy may be explained by the fact that our tests were performed in nonproliferating cells.
Similarly, numerical chromosomal damage (aneuploidy and polyploidy) and abnormal metaphases were reported in the bone marrow of rats 48 hours after a single oral dose of Al 2 O 3 NMs while no effect was observed with bulk Al 2 O 3 [14,15]. In addition, induction of micronuclei in erythrocytes was also observed.
However, this genotoxic effect on erythrocytes was concomitant with a cytotoxic effect, while no toxicity was observed in our study [16], or in the study of Zhang et al [56]. In contrast, other results obtained from in vivo studies are in agreement with the lack of chromosomal damage observed in vitro in our study following treatment with Al 2 O 3 NMs. In fact, with the same Al 2 O 3 NMs used in this study, we did not observe an induction of micronuclei in either bone marrow or in the colon of rats after a short-term oral treatment [16]. Similarly, no induction of micronuclei in the bone marrow of mice was detected following intraperitoneal injections, irrespective of the size of the Al 2 O 3 particles [56].
The absence of genotoxic activity of Al 2 O 3 NMs in Caco-2 and HepaRG cells was further con rmed in the H2AX assay as well as the comet assay. We did not observe any increase in H2AX levels in either cell line, which is in agreement with results from a study by Tsaousi et al [21] in primary human broblasts.
Additionally, Al 2 O 3 NMs did not induce DNA damage in the alkaline comet assay in Caco-2 and HepaRG cells following a 24 h treatment. Although some studies have reported negative results in the comet assay in human lymphocytes and in human embryonic kidney cells [20,26], others have demonstrated time-and/or concentration-dependent genotoxic effects in Chinese hamster lung broblasts [56], in RAW264 murine macrophages [24] and in human liver HepG2 cells [22] treated with Al 2 O 3 NMs.
Nevertheless, the increase of DNA fragmentation in these latter studies was probably linked to cell death detected by Trypan blue exclusion [56] or by apoptotic markers [22,24].
In vivo, after a short-term treatment using the same Al 2 O 3 NMs, we only observed an increase in DNA damage in the comet assay in bone marrow, while no effect was observed in intestine, colon, kidney, spleen or blood [16]. Balasubramanyam et al [14] showed a time-and concentration-dependent increase in DNA damage in blood with the comet assay with both bulk and nano Al 2 O 3 forms after a single gavage but the effect decreased at 48 h before disappearing at 72 h. DNA breakage associated with necrosis and apoptosis was observed in liver and kidney of rats after a repeated oral treatment for 75 days with 70 mg/kg bw Al 2 O 3 NMs [57]. Therefore, it seems that both the in vitro and in vivo results with Al 2 O 3 NMs support the conclusion that DNA breaks detected by the comet assay were mostly related to cell death rather than to a clear genotoxicity.
Nevertheless, we have shown that Al 2 O 3 NMs induced oxidative DNA damage in Caco-2 cells following a 24 h treatment, despite no signi cant ROS induction. Furthermore, a concentration-dependent trend towards oxidative damage was observed at 5 h. This could suggest the rapid formation of oxidative DNA damage which is further repaired, as previously demonstrated [58,59]. Evidence from in vitro experiments in a variety of different cell lines suggests that treatment with Al 2 O 3 NMs can induce oxidative stress [20,56,60,61] including in Caco2 cells [53]. Interestingly, Alari et al [22] reported positive results in the comet assay in HepG2 cells which was accompanied by oxidative damage and cell death. In the present study, no oxidative DNA damage or oxidative stress was observed in HepaRG cells. Differentiated HepaRG cells represent a model which is more similar to human hepatocytes when compared to HepG2 cells, and could therefore be less sensitive to oxidative damage resulting from Al 2 O 3 NMs. Similarly, we did not detect oxidative DNA damage in liver, or in other organs of rats after oral exposure [16]. In contrast, an increase in oxidative stress was observed in several tissues including liver after acute and repeated oral exposure of rats with Al 2 O 3 NMs [33].
Similar to the results obtained for Al 2 O 3 NMs, no cytotoxicity or apoptotic response was observed following treatment with Al 0 NMs, despite their presence in the cytoplasm of Caco-2 and HepaRG cells. In contrast to our results in differentiated Caco-2 and HepaRG cells, Al 0 NMs were found to induce a decrease in viability in rat alveolar macrophages and in BRL3A rat liver cells following 24 h exposure at concentrations similar to those used in our study [27,62]. This discrepancy could be explained by a difference in relative cell density for a similar concentration of Al 0 NMs tested with a lower NM:cell ratio in differentiated Caco2 and HepaRG cells compared to the two other proliferating cell systems.
Despite only a slight increase in H2AX levels observed only in HepaRG cells and only at the highest concentration tested, a dose-dependent increase in tail DNA was observed in both Caco-2 and HepaRG cell lines treated with Al 0 NMs using the alkaline comet assay after both 5 h and 24 h treatments.
Nevertheless, this result required further investigation due to possible interference of NMs with the alkaline comet assay that has been widely documented in the literature [35,36,63,64]. Indeed, NMs present in the cytoplasm of cells following uptake can interact with DNA following the lysis step of the comet assay, and could therefore induce additional breaks or inhibit DNA migration. In addition, a dissolution due to the conditions of the comet assay could result in reaction of aluminum ions with DNA, especially the phosphate backbone, as reported in some studies [65,66]. Such reactions may then induce DNA damage revealed during the comet assay as suggested by Zhang et al [67]. Our results clearly demonstrate that, unlike Al 2 O 3 NMs, Al 0 NMs can induce DNA damage when in contact with DNA and interfere signi cantly with the comet assay. Consequently, the positive results in cells treated with Al 0 NMs obtained in this study should therefore be treated with caution. In vivo, using the same Al 0 NMs as the present study, no genotoxic response was observed in several key tissues, with the exception in rat duodenum where a cross-linking effect was suggested [16].
The carcinogenic potential of Al NMs was investigated using the cell transformation assay with Bhas 42 cells. Neither Al 0 nor Al 2 O 3 NMs induced cell transformation, although a decrease in the number of transformed foci was observed. This decrease, observed at concentrations inducing a weak inhibition of cell proliferation at Day 7, is likely explained by a more pronounced inhibition of cell growth after 21 days of culture due to the three repeated treatments during the promotion assay. This phenomena was also observed with amorphous silica NMs [49] as well as with other non-carcinogenic chemicals such as Lascorbic acid and caffeine [68].
Ion release from NMs in cell culture media, or in intracellular compartments can contribute to cytotoxic effects in vitro. The soluble fraction of Al 0 and Al 2 O 3 NMs measured by ICP-MS demonstrated a very low solubility of Al 0 and Al 2 O 3 NMs in both cell media . However, ion release may occur after cell uptake in speci c compartments with low pH such as lysosomes [69] as suggested for Al 2 O 3 NMs [24]. In such a scenario, secondary effects affecting mitochondria and resulting in the generation of ROS cannot be excluded. In the case of Al 0 , the formation of a passivating oxide layer may in uence its dissolution behavior [70]. Consequently, effects could be induced by ionic Al released from the NMs rather than effects related to the particulate form [1]. As a strong oxygen acceptor, the Al ion tends to bind to citrate, phosphate, and catecholamine, generating oxygen radicals [1,71]. In addition, Al ions can also bind to negatively charged phospholipids, which are easily attacked by reactive oxygen species such as O 2 ·, H 2 O 2 , and OH· [72,73] as well as DNA [66].
No genotoxic effects were observed in differentiated Caco-2 or HepaRG cells treated with AlCl 3 at concentrations up to 128 µg/mL Al content corresponding to 1.16 mg/mL AlCl 3 . At the concentrations of AlCl 3 tested, no effects were observed in the different assays following 5 or 24 h treatments. Indeed, negative results were obtained for promotion and initiation, as well as for genotoxic and oxidative stress responses. Our results are consistent with Villarini et al [74] who observed no genotoxicity in response to Al ions in neuroblastoma cells with the comet assay, as well as no cytotoxicity or oxidative stress.
However, other studies have reported genotoxicity of AlCl 3 in human lymphocytes [17,29]. Interestingly, the authors of this study observed the highest level of micronuclei during the G1-phase of the cell cycle. The differentiated HepaRG and Caco2 cells used in our study are not proliferating, and therefore could explain the discrepancy between the studies. In vitro, chromosomal damage observed in blood cells at AlCl 3 concentrations below 25 µg/mL, was associated with apoptosis [28,29,30]. Moreover it was shown that Al ions can induce oxidative DNA damage irrespective of the cell cycle phase [29]. Indeed, the role of Al ions in mediating genotoxic effects may be more complex, as it has been suggested that Al ions may inhibit several DNA repair proteins with zinc nger domains [29,75].
In our study, as the soluble fraction of AlCl 3 was always higher than that for Al 0 and Al 2 O 3 NMs, the effects observed for Al 0 and Al 2 O 3 NMs are not likely to be related to ion release in cell media. Although ECHA emphasized that the difference in the toxicological pro le between soluble aluminum compounds and insoluble aluminum oxide may be explained by lower bioavailability of insoluble test compounds, it was recently shown that the content of Al in blood of rats treated orally was higher with Al 2 O 3 NMs than with AlCl 3 [76]. Moreover, the persistence of NMs in organs long after intial exposures has been described, and the accumulation of Al NMs in organs following repeated exposure could poteniate adverse effects in tissues in the long term. Further studies are clearly needed to investigate the fate of accumulated NMs in tissue, including possible effects due to ion release, as well as toxic effects related to particle accumulation.

Conclusion
In summary, despite the uptake and presence of Al NMs in the cytoplasm of differentiated Caco-2 and HepaRG cells, we have shown that Al 2 O 3 NMs do not induce apoptosis, oxidative stress, or cytotoxic effects following a 24 h treatment. In addition, Al 2 O 3 NMs were negative in the micronucleus assay, and in initiation and promotion in the CTA. Nevertheless, oxidative DNA damage was observed in Caco-2 cells.
The assays performed with Al 0 NMs and AlCl 3 were also negative except a slight increase of H2AX levels only in HepaRG cells, and only at the highest concentration tested. Considerable DNA damage was observed with Al 0 NMs in both Caco-2 and HepaRG cells in the comet assay, although this was likely associated with the signi cant interference with these NMs, and these results must be taken with caution.
As ion release from Al NMs was shown to be very limited in cell media, the effects are rather due to the particulate form or to ion release inside the cells. Further investigation is needed to clarify the extent of intracellular ion release from NMs, its contribution to cytotoxic effects compared to the direct impact of the presence of intracellular particles.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SuppdatPFT.ppt