Background: Animal models remain at that time a reference tool to predict potential pulmonary adverse effects of nanomaterials in humans. However, in a context of reduction of the number of animals used in experimentation, there is a need for reliable alternatives. In vitro models using lung cells represent relevant alternatives to assess potential nanomaterial acute toxicity by inhalation, particularly since advanced in vitro methods and models have been developed. Nevertheless, the ability of in vitro experiments to replace animal experimentation for predicting potential acute pulmonary toxicity in human still needs to be carefully assessed. The aim of the study was to evaluate the differences existing between the in vivo and the in vitro approaches for the prediction of nanomaterial toxicity and to find advanced methods to enhance in vitro predictivity. For this purpose, rats or pneumocytes in co-culture with macrophages were exposed to the same poorly soluble and poorly toxic TiO and CeO nanomaterials, by the respiratory route in vivo or using more or less advanced methodologies in 2 2 vitro. After 24 h of exposure, biological responses were assessed focusing on pro-inflammatory effects and quantitative comparisons were performed between the in vivo and in vitro methods, using compatible dose metrics. Results: For each dose metric used (mass/alveolar surface or mass/macrophage), we observed that the most realistic in vitro exposure method, the air-liquid interface method, was the most predictive of in vivo effects regarding biological activation levels. We also noted less differences between in vivo and in vitro results when doses were normalized by the number of macrophages rather than by the alveolar surface. Lastly, although we observed similarities in the nanomaterial ranking using in vivo and in vitro approaches, the quality of the data-set was insufficient to provide clear ranking comparisons. Conclusions: We showed that advanced methods could be used to enhance in vitro experiments ability to predict potential acute pulmonary toxicity in vivo. Moreover, we showed that the timing of the dose delivery could be controlled to enhance the predictivity. Further studies should be necessary to assess if air-liquid interface provide more reliable ranking of nanomaterials than submerged methods. Keywords: Poorly soluble nanomaterials, Acute exposure, Pulmonary toxicity, Alternative toxicity testing, Air-liquid interface, In vivo - in vitro comparison * Correspondence: email@example.com Institut National de l’Environnement Industriel et des Risques (INERIS), (DRC/ VIVA/TOXI), Parc Technologique ALATA - BP 2, F-60550 Verneuil-en-Halatte, France Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 2 of 20 Background exposed in vivo by oropharyngeal aspiration and lung Inhalation is an important exposure route for many me- slices or alveolar macrophages exposed in vitro to sus- tallic and poorly soluble nanomaterials (NMs) , in- pensions of TiO and CeO NMs. For some NMs, they 2 2 cluding TiO or CeO , which are among the most showed pro-inflammatory effects at similar doses in vivo 2 2 commonly used in nanotechnologies . To assess the and in vitro when the doses were expressed in mass of pulmonary toxicity of these NMs after acute exposure, NM per surface unit, both in vivo and in vitro. Teeguar- in vivo assays using animal models remain the most reli- den et al.  compared the pulmonary toxicity between able approach to predict potential adverse effects in mice exposed in vivo by inhalation and alveolar epithe- humans , because of similar levels of complexity. lial cells or macrophages exposed in vitro in submerged Nevertheless, considering the high number of NMs used conditions to iron oxide NMs. They showed inflamma- and their physico-chemical diversity, it seems difficult, tory effects at lower doses in vivo compared to in vitro for ethical and financial reasons, to rely on animal ex- when the doses were expressed in μg/cm and better perimentation only. It is therefore necessary to find reli- similarity when the doses were expressed in mass of NM able methods that can be used as alternatives to in vivo per number of macrophages. Donaldson et al.  models in this context. showed good correlations between pro-inflamatory re- In vitro studies using lung cells represent an inexpen- sponses in vivo in rats (neutrophil influx) and in vitro sive and easy-to-use alternative to assess pulmonary (IL-8 expression) in A549 cells when the doses expressed acute toxicity after exposure to NMs . Usually in in μg/cm where normalized by NM surface areas. vitro, the cells are exposed in submerged conditions to Nevertheless, only few in vitro experiments performed suspensions of NMs for 24 h. However, these simplistic in submerged conditions were compared to in vivo ex- experimental conditions do not accurately mimic the in- periments and it remains unclear whether better predic- teractions between particles and lungs in the human tion could be obtained by using more advanced in vitro body . This may lead to different biological responses methods, like ALI exposures. between in vivo and in vitro approaches. Recently, many In this context, the aim of our study was to assess the progresses have been made to simulate in vitro the ability of several in vitro methods, more or less ad- cell–particle interactions occurring in the lungs in vivo. vanced, to predict the adverse effects observed in vivo Importantly, advanced cellular models including after exposure to poorly toxic and poorly soluble metal- co-cultures or 3D-cultures  and physiological expos- lic NMs. The perspective is to promote reliable alterna- ure methods, including systems allowing exposure of tive methodologies to predict the potential inhalation cells at the air-liquid interface (ALI) to aerosols of NMs toxicity of NMs in humans. For this purpose, in vivo and , have been developed. These new methodologies in vitro experiments were performed using the same could help to predict more reliably the pulmonary ef- TiO and CeO NMs. In vivo, rats were exposed to the 2 2 fects observed in vivo. NMs by intratracheal instillation and then sacrificed Comparisons of NMs toxicity between in vivo and in after 24 h of exposure. In vitro, alveolar epithelial cells vitro approaches were performed in several studies to in co-culture with macrophages were exposed for 24 h assess if similar toxicity patterns could be found. at the ALI to aerosols or in submerged conditions to Qualitative vivo-vitro comparisons were performed. In suspensions of NMs. Moreover, different deposition kin- their study, Sayes et al.  compared cytotoxic and in- etics were tested. The results of the in vitro study were flammatory responses, between rats exposed in vivo by published previously by our team . In this paper we intratracheal instillation and alveolar epithelial cells and showed toxic effects at lower doses when cells were ex- macrophages exposed in vitro in submerged conditions posed at the ALI to aerosols of NMs compared to expos- to silicium and ZnO NMs. The authors didn’t observe ure to suspensions. We also showed the relevance of correlations between the in vitro and in vivo results. timing consideration for the dose delivery when asses- Nevertheless, Rushton et al.  highlighted that better sing poorly soluble NM toxicity in vitro. Both in vivo vivo-vitro correlations could be obtained when the toxi- and in vitro, cytotoxic, inflammatory and oxidative stress cological responses were normalized by the NM surface responses were assessed after 24 h of exposure and areas. In this work, the authors normalized the data pub- qualitative and quantitative comparisons were per- lished by Sayes et al. by the surface area of the NMs and formed. To perform in vivo - in vitro comparisons, com- showed that the NMs used could be ranked similarly in mon dose metrics were selected between in vivo and in vivo and in vitro in function of their toxicity. Recently, it vitro methods and normalizations were performed. has also been shown that advanced comparisons could be performed by using similar dose metrics between in Results vivo and in vitro approaches. For example, Kim et al. The ability of several in vitro methods (ALI and sub-  performed quantitative comparisons between mice merged) to predict potential adverse effects in vivo in Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 3 of 20 lungs, after exposure to poorly toxic and poorly soluble of the dose delivery. Nevertheless, as shown in our pre- metallic NMs, was assessed in this study. vious paper , NMs concentration in suspensions For this purpose, we performed in vivo and in vitro ex- were adjusted to obtain similar deposited doses periments using the same TiO (NMs 105, 101 and 100) (Additional file 1: Table S1). First, cells were exposed to and CeO (NM212) NMs. The physico-chemical charac- suspensions in inserts and the NM deposition was main- teristics of the four NMs were characterized in exposure tained for 3 h. After 3 h of exposure, the deposition was media (Table 1). Furthermore, the number size distribu- stopped and the cells were kept during the remaining tions and densities of NMs in suspensions (for exposure 21 h in submerged condition in the incubator. Secondly, of cells in submerged conditions in vitro or rats by intra- cells were exposed in plates to suspensions of NMs for tracheal instillation in vivo) and in aerosols (for exposure 24 h. In that situation, the NM deposition was main- of cells at the ALI) were assessed. Surprisingly, similar tained for the whole exposure time, meaning that the results were observed between NM suspended in water final deposited dose was reached within 24 h. After 24 h and in culture medium . Number size distributions of exposure, inflammation, cytotoxicity and oxidative and densities determined in exposure media were then stress were assessed. Lowest Observed Adverse Effects used to calculate the mean surface area of NM agglom- Levels (LOAELs) and critical effect dose intervals were erates in suspensions and in aerosols (for ALI exposures then determined, using first significant effects measured only). Based on our previous electron microscopy obser- or benchmark dose response modeling, respectively. vations , agglomerates were considered spherical for Focusing on these LOAELs and critical effect dose inter- the calculation. The mean surface area calculated in ex- vals, quantitative and qualitative comparisons were per- posure media was then used for vivo-vitro comparisons. formed between in vivo and in vitro results. With dose The major innovation of this study was to compare intervals, contrary to with LOAELs, dose-response NM toxicities between in vivo and in vitro approaches, curves were taken into account and uncertainty was in- using several more or less advanced in vitro methods cluded in the data. Comparisons were performed with and testing different timings of the dose delivery in vitro. LOAELs and dose intervals to assess if similar conclu- The Fig. 1, which was adapted from our previous pub- sions could be made using the two criteria of effect dose. lished paper  to take into account our new in vivo For these comparisons, normalizations were performed experiments, is presented here and proposes an overview to have common dose metrics between in vivo and in of the study design. For the study, we focused on the vitro approaches. NM surface areas were also considered doses deposited into the lungs (in vivo) or on cells (in for ranking comparisons. vitro) because we postulated that metallic and poorly soluble NMs exert their toxicity by direct contact with Pro-inflammatory responses in vivo and in vitro the cells. In vivo and in vitro, the biological responses were In vivo, rats were exposed to NMs by intratracheal in- assessed after 24 h of exposure to three TiO (NMs 105, stillation and sacrificed after 24 h of exposure. In vitro, 101, 100) and one CeO (NM212) NMs. alveolar epithelial cells in co-culture with macrophages In vivo, pro-inflammatory effects (neutrophil influx were exposed for 24 h at the ALI to aerosols or in sub- and levels of the pro-inflammatory mediators IL-1β, merged conditions to the NMs. At the ALI, the cells IL-6, KC-GRO and TNF-α) were assessed in bronchoal- were exposed to aerosols of NMs for 3 h, meaning that veolar lavage fluids (BALF) of rats exposed by intratra- the final deposited dose was reached within 3 h. The cheal instillation (IT) to the NMs (around 4.5 mL cells were then kept in the incubator for the remaining recovered for each sample). We observed significant ef- 21 h at the ALI with the NMs deposited on their surface. fects with TiO NMs 105 and 101 and CeO NM212, 2 2 In submerged conditions, we used two different timings but not with TiO NM100. Significant pro-inflammatory Table 1 Physico-chemical properties of TiO (NMs 105, 101, 100) and CeO (NM212) nanomaterials in exposure media 2 2 Critallinity Coating Primary Primary Primarysurface Mean size in Mean density in Mean surface particle density area, BET exposure exposure media area in exposure 3 2 3 2 size (nm) (g/cm ) (m /g) media (nm) (g/cm ) media (m /g) Susp Aero Susp Aero Susp Aero NM105 80% anatase / No 21 4.2 46.1 318 240 1.4 0.7 13.5 37.7 20% rutile NM101 Anatase Hydrophobic 8 3.9 316 567 80 1.6 0.9 6.7 83.3 NM100 Anatase No 100 3.9 10 286 320 1.8 0.6 11.7 31.3 NM212 Cubic cerionite No 29 7.2 27 233 200 2.1 1.1 12.5 27.3 Susp Suspension, Aero Aerosol Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 4 of 20 Fig. 1 Experimental conditions used for the in vivo/in vitro comparisons (adapted from ). In vitro and in vivo experiments were performed using the same TiO (NM105, NM101, NM100) and CeO (NM212) NMs. In vitro, alveolar epithelial cells in co-culture with macrophages were 2 2 exposed for 24 h at the air-liquid interface (ALI) to aerosols or in submerged conditions to suspensions of NMs. Different deposition kinetics were tested. At the ALI the NM deposition via aerosol was maintained for 3 h. The cells were then kept at the incubator for the remaining 21 h (3 h + 21 h). In submerged conditions, two deposition kinetics were used. In inserts, the deposition was maintained for 3 h. After 3 h, NM suspensions were replaced by fresh medium and the cells were then kept a the incubator for the remaining 21 h (3 h + 21 h) with the NMs deposited on their surface. In plates, classic exposure conditions were used and NM depositions were maintained for 24 h. In vivo, rats were exposed by intratracheal instillation with NM suspension and the NM were deposited almost instantly into the lungs. After 24 h of exposure, the biological activity was assessed, focusing more particularly on pro-inflammatory markers, including cytokine secretions and neutrophil influx (in vivo only) effects were noted at the maximum dose tested: 400 μg/ significant increases in neutrophils or cytokines were 2 6 lungs, corresponding to around 0.1 μg/cm or 20 μg/10 noted for TiO NM100. Moreover, for all the NMs macrophages after normalization by the alveolar surface tested, we did not observe any significant changes in (4000 cm ) or the number of alveolar macrophages macrophages or total cell numbers. (25 million), respectively (Fig. 2). After exposure to Based on the significant responses detected in vivo, TiO NMs 105 and 101, this was characterized by a sig- lowest observed adverse effects levels (LOAELs) were nificant neutrophil influx in BALF supernatants, associ- determined for pro-inflammatory effects with NMs 105, ated with increased concentrations of TNF-α for the 101 and 212, but not with NM100. These LOAELs were NMs 105 and 101 and KC-GRO for NM105 only. We used in the present study to compare in vivo and in vitro also noted significant increases in IL-1β,IL-6 and results. We also used benchmark dose-response model- TNF-α secretion with NM212, although no significant ing to determine critical effect doses for a 20% increase neutrophil influx was detected. The absence of signifi- of cytokine/chemokine response with an interval of cant neutrophil influx with NM212 may have been due doses corresponding to a 90% confidence, to compare in to a high variability in the control sample. No vivo and in vitro results. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 5 of 20 Fig. 2 Cytology and cytokines/chemokine levels in bronchoalveolar lavage fluids 24 h after instillation with the NMs. Rats were instilled after hyperventilation with suspensions of TiO (NM105, NM101, NM100) and CeO (NM212). After sacrifice, bronchoalveolar lavages were performed 2 2 using PBS. The bronchoalveolar lavage fluids were recovered and centrifuged to separate cells from supernatant. For cytology analysis, the cells were resuspended in RMPI medium and then seeded on slides at 300000 cells/spots using a cytospin and then fixated and coloured in May-Grunwald Giemsa. The percentage of different cell types in BALF was determined using optical microscopy. For cytokine/chemokine analysis, supernatants were dosed using ELISA multiplex to determine IL-1β, IL-6, KC-GRO and TNF-α levels. Data represent the mean ± SD of six animals. Kruskal-Wallis test followed by Dunn’s post-hoc test were performed to compare treated groups to controls (*p < 0.05; **p < 0.01; ***p < 0.001) In vitro, pro-inflammatory responses were assessed apical and basolateral compartments of the inserts after 24 h of exposure by evaluating the levels of (containing 1 and 2 mL of culture medium, respect- pro-inflammatory mediators IL-1β,IL-6, IL-8 and ively). In submerged conditions in plates, which repre- TNF-α in cell supernatants. After ALI exposure to sents the classic exposure conditions usually used in aerosols of NM in inserts, cytokine levels were only vitro, the NM deposition was maintained for the 24 h measured in the basolateral compartment (containing of exposure and cytokine levels were exclusively mea- 2 mL of culture medium) as the cells were maintained sured on the apical side of the cells (containing 0.5 mL at the ALI for the 3 h of exposure to the aerosols and of culture medium) due to the absence of a basolateral for the remaining 21 h into the incubator with the NMs compartment. deposited on their surface. After exposure in sub- Briefly, as demonstrated in our previous in vitro study merged conditions to suspensions in inserts, using the , we observed significant pro-inflammatory re- similar dose rate timing of the dose delivery than at the sponses at the ALI with all tested NMs. We also ob- ALI (3 h), cytokine levels were assessed both in the served effects in submerged conditions in inserts and in Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 6 of 20 plates, but mainly with NMs 105 and 101. A compilation exposure method using these two parameters because of the pro-inflammatory results published in our previ- too little significant cytotoxicity and oxidative stress ef- ous study  is available in the Additional files of the fects were detected in vitro. However, it could not be ex- present paper (Additional file 1: Figure S1). According cluded that less cytotoxicity and oxidative stress effects to the first significant pro-inflammatory responses de- were observed compared to pro-inflammatory effects, tected, LOAELs were determined for each assay per- both in vivo and in vitro, because of a lack of sensitivity formed (Additional file 1: Table S2). For all NMs, the of the assays performed. LOAELs were determined at lower doses at the ALI compared to submerged conditions in inserts and also at Vivo-vitro comparisons using the inflammation results lower doses when the final dose was deposited within As described previously, inflammation was the most sen- 3 h rather than within 24 h. In the present study, bench- sitive marker of biological responses at 24 h in our mark dose-response modeling was also used with the in study, both in vivo and in vitro. For this reason, we fo- vitro data to determine an interval of dose for a 20% in- cused on the pro-inflammatory responses to perform crease of cytokine/chemokine response with a 90% con- vivo-vitro comparisons. To perform quantitative com- fidence, to compare in vivo and in vitro results. parisons, the LOAELs determined in vivo and in vitro for the first significant pro-inflammatory responses ob- Cytotoxicity and oxidative stress effects in vivo and in vitro served were first used. Dose-response comparisons were Cytotoxicity and oxidative stress responses were also then performed using dose intervals determined by assessed, both in vivo and in vitro. In vivo, LDH levels benchmark modeling. For dose intervals calculation, we were evaluated in BALF supernatants and Reactive determined a critical effect dose corresponding to a 20% Oxygen Species (ROS) levels were measured in BALF increase of pro-inflammatory mediator levels compared cells. Although significant pro-inflammatory responses to non-exposed controls and the Benchmark Dose were noted, we did not observe any significant cytotoxic Lower confidence limit (BMDL) and the Benchmark or oxidative stress effects after 24 h of exposure to the Dose Upper confidence limit (BMDU) of the interval for NMs (Table 2 and Additional file 1: Figure S2). In vitro, a 90% confidence. This was performed for each cytotoxicity was assessed by using the alamar blue test pro-inflammatory mediator, each exposure method and and by measuring LDH levels in cell supernatants. ROS each NM used. Examples of benchmark dose-response levels were measured in cells as marker of oxidative modeling for the calculation of critical effect doses and stress. We observed few significant cytotoxicity and oxi- dose intervals are shown in the Additional file 1: Figure dative responses at the ALI (and only with the NMs 105 S3. For each NM and each exposure method, we then and 101). LOAELs were also determined for cytotoxicity calculated the median value of the BMDL and the me- and oxidative stress in submerged conditions and more dian value of the BMDU for the four pro-inflammatory particularly in inserts (Table 2). As described in our pre- mediators (IL-1β, IL-6, IL-8/KC-GRO, TNF-α), to vious article , it was not possible to perform clear determine a median dose interval for general quantitative comparisons between the different in vitro pro-inflammatory response, as shown in the Table 3.We Table 2 LOAELs (μg/cm ) for cytotoxicity and oxidative stress effects determined after 24 h of exposure LOAELs indicated represent significant cytotoxicity > 5% Significant effects allowing the determination of a LOAEL No significant adverse effects observed b 2 Doses tested at the ALI: 0.1, 1, 3 μg/cm c 2 Doses tested in submerged conditions in inserts: 1, 3, 10 μg/cm d 2 Doses tested in submerged conditions in plates: 1, 3, 10, 20 μg/cm e 2 Doses tested in vivo: 0.001, 0.01, 0.1 μg/cm Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 7 of 20 calculated the median dose interval of the four cytokines of toxicity for poorly soluble NMs [15–17]. To express the because similar results could be observed in our study metric in mass per alveolar surface unit, masses deposited when pooling the results of the four cytokines and when into the lungs or on the cells were divided by the total al- comparing dose intervals of each cytokine one by one. veolar surface in vivo (4000 cm )[18, 19] or by the surface 2 2 We believe that comparisons performed in our study of the cell layer in vitro (4,67 cm in inserts and 2 cm in were easier to follow when a general pro-inflammatory plates), respectively. To express the metric in mass per response was used instead of comparing the dose inter- macrophage number, the mass of NM in lungs (in vivo) or vals of each cytokine one by one. With dose intervals, the deposited mass per cm (in vitro) were divided by the contrary to with LOAELs, dose-response curves were total number of alveolar macrophages in vivo (around 25 taken into account and uncertainty was included in the million) [18, 20] or in vitro (60,000 or 25,000/cm in in- data. Comparisons were performed with LOAELs and serts or in plates, respectively) . The doses expressed dose intervals to assess if similar conclusions could be in mass/macrophages were also normalized by the surface made using the two criteria of dose. For the compari- area for each NM. This normalization was performed be- sons, LOAELs and dose intervals were expressed using cause it was shown that the surface area was the most ef- different dose metrics which were compatible between fective dose metric to explain acute NM toxicity in the in vivo and in vitro methods. lung [15–17]. For that, doses expressed in mass/macro- phages were multiplied by NM surface areas, calculated using NM primary characteristics in powders (BET Selection of relevant dose metrics method) or mean sizes and densities in exposure media. Common dose metrics that could be used with all expos- Based on our previous observations, NMs were assumed ure methods (submerged, ALI, instillation or inhalation) to be spherical for surface area calculations in exposure were selected, as shown in Fig. 3, to compare NM toxicity media . Nevertheless, it was not possible to ensure between in vivo and in vitro approaches. To generate that NM agglomerates were strictly spherical and relative common dose metrics, we normalized the deposited uncertainties remain regarding the mean surface area cal- masses by the alveolar surface or by the macrophage num- culated in exposure media. ber . These two normalizations were performed as Doses could also be expressed in number of NMs per they take into account the direct contact between the surface area or per cell. Nevertheless, these metrics were NMs and the tissues, that was shown to be the main cause not chosen due to the difficulty to characterize NM size Table 3 Dose intervals (in μg/cm ) determined for each NM and each methodology Cytokines NM105 NM101 NM100 NM212 Dose interval Dose interval Dose interval Dose interval BMDL BMDU Median BMDL BMDU Median BMDL BMDU Median BMDL BMDU Median In vitro, IL-1β 0.10 11.90 4.63–11.20 0.53 2.85 2.82–6.36 10.68 32.59 10.96–26.19 0.06 10.26 0.430–14.168 suspension IL-6 4.50 11.11 4.86 10.47 ND ND ND ND (24 h) IL-8 ND ND ND ND ND ND ND ND TNF-α 4.76 11.20 2.82 6.36 11.25 19.80 0.80 18.08 In vitro, IL-1β 0.43 2.25 0.26–1.51 0.60 6.61 0.80–6.19 3.47 53.23 3.47–55.64 3.47 54.81 3.37–52.50 suspension IL-6 0.20 1.36 1.28 8.17 ND ND 2.30 50.25 (3 h + 21 h) IL-8 0.11 1.11 0.84 3.73 3.35 58.05 3.42 9.64 TNF-α 0.31 1.66 0.76 5.78 3.83 52.87 3.31 54.75 ALI IL-1β 0.051 0.80 0.061–0.82 0.061 0.74 0.099–0.80 0.006 0.91 0.045–0.90 0.63 2.61 0.88–2.71 (3 h + 21 h) IL-6 0.037 0.81 0.089 0.77 0.012 0.88 0.88 2.71 IL-8 0.078 0.83 1.13 11.87 0.078 0.90 ND ND TNF-α 0.070 0.82 0.11 0.84 0.24 0.97 ND ND In vivo IL-1β 0.0011 0.067 0.0007–0.075 ND ND 0.0022–0.084 ND ND ND ND ND 0.0000–0.074 IL-6 0.00069 0.13 0.0044 0.086 ND ND 0.000 0.019 IL-8 0.00019 0.084 ND ND ND ND 0.0033 0.082 TNF-α 0.00075 0.0091 0.000 0.082 ND ND 0.000 0.074 BMDL Benchmark Dose Lower confidence limit, BMDU Benchmark Dose Upper confidence limit. Median: Median BMDL and BMDU values calculated by pooling the four cytokines, to have a dose-interval for a general pro-inflammatory response Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 8 of 20 Fig. 3 Compatibility of the different dose metrics between in vivo and in vitro approaches. In order to compare in vitro and in vivo conditions it is important to use common dose metrics. The doses are often expressed as concentrations, including mass/volume of liquid in vitro in submerged conditions and mass/volume of air in vivo in inhalation studies. However, these metrics cannot be used within the different in vivo (inhalation or instillation) and in vitro (ALI or submerged) methodologies. Moreover, using concentrations in mass/volume does not take into account the real contact between the NMs and the cells or tissues. Thus it does not seem appropriate to use such dose metrics for in vivo-in vitro comparisons; more particularly for poorly soluble NMs as their toxicity is attributable to their surface reactivity. In vivo, the total mass of NMs administered per lungs, animal or mass is often used as dose metric. This dose metric takes into account the deposition in the overall organ, but cannot be used in vitro. Nevertheless, common dose metrics can be used by normalizing the mass deposited on cells in vitro or into the lungs in vivo by the surface of the tissues or by the number of cells. Doses expressed in mass can also be normalized NM surface areas, that has been shown to be the most effective dose metric for acute NM toxicity in the lung distributions in the lungs. Moreover, the number metric conditions in inserts (3 h + 21 h) and in submerged was not shown to be more relevant than the mass metric conditions in plates (24 h), compared to in vivo, when assessing NM toxicity . respectively. Comparisons in mass/alveolar surface Comparisons in mass/macrophages Doses of NMs were first expressed in μg/cm , after Doses were also normalized by the total number of mac- normalization of the deposited doses by the total alveo- rophages and expressed in μg/10 macrophages to com- lar surface in vivo (4000 cm ) or by the surface of the pare in vivo and in vitro LOAELs and dose intervals cell layer in vitro. All the LOAELs and the dose intervals (Table 5, Fig. 4b and Additional file 1: Table S3). For that determined for pro-inflammatory effects were expressed purpose, in vivo doses expressed in μg were normalized using this dose metric (Tables 3 and 4, Fig. 4a) and by the number of alveolar macrophages. In vitro, depos- vivo-vitro comparisons were performed. Generally, for ited doses expressed in μg/cm were normalized by the each NM, we noted pro-inflammatory effects at lower total number of alveolar macrophages-like cells per cm . doses in vivo compared to in vitro. We also observed We noticed that the LOAELs and the dose intervals de- that the LOAELs and the dose intervals determined in termined in vitro were closer to those observed in vivo vitro after exposure at the ALI were closer to those in when the doses were normalized by the number of mac- vivo than those determined in vitro in submerged condi- rophages rather than by the alveolar surface (Table 5 tions. Moreover, we noted that the LOAELs determined and Fig. 4). When looking at the LOAELs, the in vitro were closer to those in vivo when the final dose pro-inflammatory responses were observed at similar was achieved in vitro within 3 h rather than within 24 h. doses in vivo and in vitro at the ALI, whereas a differ- When comparing the LOAELs for each NM, differences ence of at least a factor of 10 was observed when the of a factor of 10, 30 and 100 were noted for exposure at LOAELs were expressed in μg/cm . Differences of the ALI to aerosols (3 h + 21 h), exposure in submerged around a factor of 3 and 20 were observed between the Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 9 of 20 Table 4 LOAELs (in μg/cm for 24 h of exposure) determined for pro-inflammatory effects Significant effects allowing the determination of a LOAEL No significant adverse effects observed a 2 Doses tested at the ALI: 0.1, 1, 3 μg/cm b 2 Doses tested in submerged conditions in inserts: 1, 3, 10 μg/cm c 2 Doses tested in submerged conditions in plates: 1, 3, 10, 20 μg/cm d 2 Doses tested in vivo: 0.001, 0.01, 0.1 μg/cm in vivo experiments and the in vitro experiments per- Ranking using the surface area as dose metric formed in submerged conditions in inserts (3 h + 21 h) Doses in mass/macrophages were normalized by the sur- and in plates (24 h), respectively. face area of each NM, to assess how the surface reactiv- ity influenced the biological responses in vivo and in 2 6 vitro. The dose intervals were expressed in cm /10 Ranking of the NMs according to the methodology used macrophages (Additional file 1: Tables S4 and S5) and a For each methodology used, a ranking of the four NMs ranking was provided for the four poorly toxic and used was provided according to the inflammation results poorly soluble NMs used in our study. Doses in mass/al- and the dose intervals that had been determined. Rank- veolar surface could also be normalized by the surface ing comparisons were performed using dose intervals area, generating the same ranking of the NMs as using only, because better screening could be performed be- the mass/macrophages dose metric (Fig. 5). tween NMs by using this criterion of effect compared to First, the dose intervals were normalized by the calcu- the use of LOAELs. Comparisons were performed to as- lated surface area using the NM primary sizes and dens- sess whether the four poorly toxic and poorly soluble ities (BET method) (Additional file 1:Table S4).This NMs could be ranked similarly, based on the different normalization had an influence on the ranking of the methodologies tested. The dose intervals were also nor- NMs. The NM101 was ranked with a lower toxicity than malized by NM primary surface areas and agglomerate expected, both in vivo and in vitro (Fig. 5). In vivo, we ob- surface areas to understand the differences in toxicity served dose intervals at lower doses for NM105 and existing between the NMs. A toxicity ranking of the NM212 than for NM101, however, it was not possible to NMs according to the different methodologies and dose include NM100 in this ranking, as no significant effects metrics used is presented in the Fig. 5. were observed, probably because significantly lower doses 2 6 (in cm /10 macrophages) were tested compared to the three other NMs, due to a lower surface area. In vitro, 2 6 Ranking using mass as dose metric when dose intervals were expressed in cm /10 macro- 2 6 In mass (μg/cm or μg/10 macrophages) some differ- phages, using surface area calculated according to primary ences were observed between in vivo and in vitro condi- sizes, we also observed effects at lower doses for the tions (Fig. 5). In vivo, NMs 105, 101 and 212 were NM100, the NM105 and the NM212 than for the NM101. observed to be clearly more toxic than NM100, as we Secondly, the dose intervals were normalized by the did not observe any significant effects with the NM100. surface area calculated according to NM agglomerate In vitro, we noticed pro-inflammatory responses for mean sizes and densities in exposure media (suspensions NMs 105 and 101, at lower doses than for NM212, at or aerosols) (Additional file 1: Table S5). Interestingly, the ALI and in submerged conditions. NM100, similarly less changes in the ranking of the NMs were observed as NM105 and NM101, seemed to elicit more when performing this normalization, for all the method- pro-inflammatory responses at the ALI than NM212, ologies used (in vitro ALI, in vitro in submerged and in but this was not observed in submerged conditions. vivo) (Fig. 5). In vivo, no clear discrepancy could be Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 10 of 20 Fig. 4 Dose intervals calculated for a 20% increase in inflammation markers in function of methodologies used. Comparisons of dose intervals were performed between the in vivo and in vitro methods used. The comparisons were performed using two dose metrics: the mass/alveolar surface (a) or the mass/macrophages (b). In vitro and in vivo experiments were performed using the same TiO (NM105, NM101, NM100) and CeO (NM212) NMs. In vitro, alveolar epithelial cells in co-culture with macrophages were exposed for 24 h at the air-liquid interface (ALI) to aerosols or in submerged conditions to suspensions of NMs. Different deposition kinetics were tested. At the ALI the NM deposition via aerosol was maintained for 3 h. The cells were then kept at the incubator for the remaining 21 h (3 h + 21 h). In submerged conditions, two deposition kinetics were used. In inserts, the deposition was maintained for 3 h. After 3 h, NM suspensions were replaced by fresh medium and the cells were then kept a the incubator for the remaining 21 h (3 h + 21 h) with the NMs deposited on their surface. In plates, classic exposure conditions were used and NM depositions were maintained for 24 h. In vivo, rats were exposed by intratracheal instillation with NM suspensions and the NMs were deposited almost instantly into the lungs. After 24 h of exposure, the biological activity was assessed, focusing more particularly on pro-inflammatory mediators. For each exposure method and for each NM, benchmark dose-response modeling was used to estimate the critical dose related to a 20% increase of pro-inflammatory mediator level and the lowest (BMDL) and the highest (BMDU) dose of the interval corresponding to confidence interval of 90%. A median dose intervals was then calculated by pooling the dose intervals of the four cytokine to have a general pro-inflammatory response made between NMs 105, 101 and 212; all three were ob- indicates that toxicity may be due to NM agglomerates served to be more toxic than NM100. In vitro at the rather than isolated NMs. ALI, NMs 105, 101 and 100 were observed toxic at lower dose than NM 212, although this was clearly more pronounced for NM105 and NM100. In submerged con- Discussion ditions, similarly as when the dose intervals were The aim of this study was to assess the ability of several expressed in mass/macrophages, NMs 105 and 101 were more or less advanced in vitro methods, to predict the observed to be more toxic than NMs 100 and 212. This pulmonary adverse effects observed in vivo after acute better correlation in the ranking between doses exposure to poorly toxic and poorly soluble metallic expressed in mass and doses expressed in surface area, NMs. The perspective is to promote reliable alternative when normalizing the dose intervals by mean agglomer- methodologies to animal testing for the prediction of ates surface area rather than by primary surface areas, pulmonary toxicity of NMs in humans. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 11 of 20 Table 5 LOAELs (μg/10 macrophages for 24 h exposure) determined for pro-inflammatory-effects Significant effects allowing the determination of a LOAEL No significant adverse effects observed a 6 Doses tested at the ALI: 1.67, 16.7, 50 μg/10 macrophages b 6 Doses tested in submerged conditions in inserts: 16.7, 50, 167 μg/10 macrophages c 6 Doses tested in submerged conditions in plates: 40, 120, 400, 800 μg/10 macrophages d 6 Doses tested in vivo: 0.16, 1.6, 16 μg/10 macrophages The selection of relevant in vivo and in vitro models in our study, it remains unclear whether the cells and the to predict potential biological responses in humans was NMs were covered by surfactant. thus very important. In vivo, the rat was selected be- Different exposure methods were assessed. In vivo, cause it is the recommended species to assess inhalation rats were exposed for 24 h by intratracheal instillation of toxicity in humans . In vitro, human rather than rat NM suspensions, after hyperventilation. In vitro, cells cells were chosen because they were more likely to were exposed using more or less advanced methods. model responses of cells in the human body. Moreover, Co-cultures were exposed at the ALI in inserts to simulate because the principal pulmonary target for inhaled NMs more closely the interactions between NMs and alveolar remains the alveoli , we focused on this part of the cells occurring in vivo, and to avoid contamination with lungs. At the alveolar surface in vivo, macrophages are culture medium. The ALI exposure system used in our in close contact with epithelial cells (i.e type and I and II study  was selected for its ability to deposit sufficient pneumocytes), at a ratio of approximately one macro- amounts of NMs on cells to observe biological adverse ef- phage to ten pneumocytes [18, 20]. The main role of the fects [31, 32]. In parallel we also used a more classical macrophage is to engulf particles to eliminate them from exposure method and cells were exposed to NM suspen- the alveolar space . Type I pneumocytes serve as a sions in submerged conditions to assess the general im- thin gas-permeable epithelial barrier . Type II pneu- pact of the culture medium surrounding poorly soluble mocytes have a role in defense of the alveoli thanks to NMs on the cell biological response. their physiological abilities . In the alveoli, macro- Both in vivo and in vitro, the toxicity was assessed phages and epithelial cells are at the ALI and are cov- after 24 h of exposure to the same TiO and CeO NMs. 2 2 ered by a thin layer of surfactant secreted at the apical For that, we focused on the deposited doses on cells or side by the type II pneumocytes. into the lungs, because metallic and poorly soluble NMs To mimic the cell organization at the alveolar surface exert their toxicity mainly by direct contact with the and the potential interactions between the cells and the cells . Moreover, we tested different timings of the NMs, a co-culture using two cell types was selected in dose delivery in vitro, to assess if that factor could influ- vitro. The A549 alveolar epithelial cell line was selected ence the cell response. In vivo, the final doses of NMs for its ability to form a cell layer and to secrete surfactant were deposited by instillation almost instantly into the [23, 26]. The THP-1 monocyte cell line was chosen for its lungs. At the ALI in vitro, NMs were deposited on the capacity to differentiate into macrophage-like cells with cells using a very low aerosol flow rate of 5 mL/min to Phorbol Myristate Acetate (PMA) . This model was prevent cell damages due to the air flux. To deposit a selected for its increased sensitivity compared to dose sufficient to observe biological effects, cells were mono-cultures of alveolar epithelial cells at the ALI  exposed for 3 h to aerosols. After exposure, the cells and in submerged conditions [28–30]. We also postulated were then kept in the incubator for the remaining 24 h that with this co-culture model, the deposited NMs could with NMs deposited on their surface. In submerged become covered with surfactant before they interact with conditions, it was not possible to deposit the NMs in- the cells, as observed in the alveoli in vivo . Neverthe- stantly on the cells either, as the deposition kinetics less, because the presence of surfactant was not evaluated depended mostly on their sedimentation rate. In Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 12 of 20 Fig. 5 Dose intervals of NMs for inflammation according to methodologies and dose metrics used. Dose intervals calculated for general acute pro-inflammatory response were used to compare the ranking of each NM in function of each exposure method used. Comparisons were also performed according to the four dose metrics used in our study (a: mass/alveolar surface), (b: mass/macrophages), (c: dose in mass/macrophages normalized by primary surface area), (d: dose in mass/macrophages normalized by agglomerate surface area). In vitro, alveolar epithelial cells in co-culture with macrophages were exposed for 24 h at the air-liquid interface (ALI) or in submerged conditions to suspensions of NMs. In vivo, rats were exposed by intratracheal instillation of NM suspensions. After 24 h of exposure, the biological activity was assessed, focusing on pro-inflammatory mediators. For each exposure method, each NM and each cytokine, benchmark dose-response modeling was used to estimate the critical dose related to a 20% increase of pro-inflammatory mediator level and the lowest (BMDL) and the highest (BMDU) dose of the interval corresponding to a confidence interval of 90%. A median dose intervals was then calculated by pooling the dose intervals of the four cytokine to have a general pro-inflammatory response inserts, we used thedurationof3hfor thedosede- 21 h with NMs deposited on their surface. In sub- livery, in order to provide comparisons as accurate as merged conditions in plates, we used classical expos- possible between ALI and submerged exposure. As ure conditions and the NM deposition on the cells for the ALI, the cells were kept for the remaining was maintained for 24 h. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 13 of 20 Similar endpoints were selected to compare in vivo NOAELs or LOAELs cannot be quantified and can be and in vitro toxicity. The biological responses were considerable as it sticks only to the dose tested. Thus assessed after 24 h of exposure to the NMs, by perform- NOAELs and LOAELS determined depend strongly on ing cytotoxicity, oxidative stress and inflammation as- the study design [37–39]. On the contrary, dose intervals says, to determine the absolute toxicity of each NM. determined by benchmark modeling take the potency of Both in vivo and in vitro, we observed that inflammation NMs to induce an effect into account and take uncertainty was the most sensitive parameter for detection of bio- in the data into account. The question is not whether an logical responses at 24 h. After NM exposure, we de- effect is induced or not, but at what dose an effect of tected significantly more pro-inflammatory effects than interest is induced. These dose intervals provide more ac- cytotoxicity and oxidative stress responses, and generally curate comparisons between in vivo and in vitro data. at lower doses. We were not surprised about the absence [37–39]. of clear pulmonary cytotoxic effects in our study as Although there were clear advantages in using dose in- poorly soluble TiO and CeO NMs were shown to be tervals, some uncertainties remain in our study regarding 2 2 not very cytotoxic at 24 h both in vivo and in vitro . their determination due to limited log dose data intervals Regarding oxidative stress production, as ROS are and because the pro-inflammatory effects were observed known to interact quickly with molecules present in the mostly at the highest doses tested. This highlights the im- cells, better detection could have been achieved by per- portance of providing experimental data that should allow forming several measurements during the 24 h of expos- to model a reasonable response slope by either providing ure. Moreover, in our protocol, cells were incubated a clear dose-response pattern of toxicity (that implies the with DCFDA probe after exposure and not before expos- use of toxic compounds) or more refined tested doses in ure which may have reduced assay sensitivity . case of low toxicity. The more experimental doses are Nevertheless, in absence of cytotoxicity at 24 h, which is tested, the more accurate is the analysis and this is true the case in our study, the authors did not show a clear also for NOAEL/LOAEL assessment. increase of ROS measurement sensitivity when the Comparisons were performed with LOAELs and dose probe was added before NM exposure . intervals to assess if similar conclusions could be made For these reasons, comparisons were performed using the two criteria of effect dose. As the LOAELs and mostly with inflammation markers as readout for NM dose intervals were associated with dose metrics, the key toxicity. Nevertheless, it has to be noted that the inflam- point for the comparisons was to select similar metrics for matory effects were observed at high doses (at least both in vivo and in vitro methodologies. For that, we fo- 10-fold higher) compared to realistic human exposure cused on the deposited masses because it take into ac- scenarios . The markers of inflammation used in our count the direct contact between the NMs and the tissues, study were selected according to their relevance in that was shown to be the main cause of toxicity for poorly representing the pro-inflammatory acute response in the soluble NMs [15–17]. Doses in mass were first normalized lungs after exposure to particles [35, 36]. Similar to the total alveolar surface in vivo or to the surface of the pro-inflammatory mediators were chosen in vitro and in cell layer in vitro. This normalization was based on the as- vivo, however those significantly secreted in vitro were sumption that the alveolar epithelium may be the main not necessarily predominant in vivo. On the basis of this target after acute exposure to NMs [14, 22]. finding we assumed that better comparisons could be Expressing results in mass/alveolar surface, we observed provided in our study by considering the global inflam- responses at doses around ten times lower in vivo 2 2 matory response, more particularly the secretion of (LOAELs at 0.1 μg/cm and BMDU around 0.08 μg/cm ) pro-inflammatory mediators that can be measured both than in vitro at the ALI (LOAELs at 1 μg/cm ,BMDU in vivo and in vitro. from 0.8 to 2.7 μg/cm ). The differences were slightly To perform quantitative comparisons, LOAELs were more pronounced in submerged conditions when the dose first determined according to the significant was delivered in 3 h (at least 20-fold compared to in vivo) 2 2 pro-inflammatory responses observed. Secondly, we (LOAELs at 3 μg/cm , BMDU from 1.5 to 55 μg/cm ), used benchmark dose-response modeling to determine and much more important, with a factor of around 100, dose intervals related to increase in pro-inflammatory when the deposition of the NMs was continuous during mediator levels. There are multiple advantages of the 24 h of in vitro exposure (LOAELs at 10 μg/cm , using benchmark dose modeling instead of a No BMDU from 6.4 to 26 μg/cm ). Interestingly, when com- Observed Adverse Level (NOAEL) or LOAEL approach in paring in vivo and in vitro biological activation levels using hazard assessment. Using NOAEL/LOAEL approach, the mass/alveolar surface metric, similar differences were toxic effects are reduced to a yes/no question and are typ- observed with LOAELs and dose intervals determined ically determined based on the presence or absence of using benchmark dose-response modeling. This indicates statistical significance. Moreover, the uncertainty in the that similar conclusions could be made using these two Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 14 of 20 criteria of effect, when comparing biological activation submerged exposure in plates for 24 h were less obvious levels in vivo and in vitro. when looking at the interval of doses instead of the In vivo and in vitro results expressed in mass/alveolar LOAELs. Finally, pro-inflammatory effects were still ob- surface were also compared in three noteworthy studies served at higher doses in vitro in submerged conditions [10, 11, 40]. In their study, Kim et al.  observed simi- compared to at the ALI or in vivo. Although serum was lar pro-inflammatory profiles when expressing the doses added in our in vitro experiments in submerged condi- in μg/cm , after exposing mice by oropharyngeal aspir- tions to keep the cells in their best physiological condi- ation and macrophages or lung slices to suspensions of tions, we hypothesized that the presence of serum may TiO , CeO and SiO NMs. Nevertheless, the real have reduced potential NM toxicity. Indeed, it has been 2 2 2 masses of NMs deposited in vitro and ex vivo were not shown that NMs were less toxic in vitro in presence of assessed and no clear quantitative comparisons were serum in suspensions compared to in absence of serum performed between the in vivo and the in vitro ap- [41, 42]. Taking that into account, better correlations at proaches, which renders the interpretation of the results 24 h may be provided in absence of serum, however this from their study difficult. Jing et al.  compared the point still has to be demonstrated. responses after acute exposure to Cu NPs, in mice lungs Although differences exist regarding the cellular and ani- and in alveolar epithelial cells at the ALI. They observed mal models and the duration of exposure between the similar responses (chemokine and LDH release) but ap- Teeguarden study and ours, this seems to indicate that fo- parently at lower doses in vitro compared to in vivo, cusing on the cell number might better explain the general when expressing the dose in ng/cm . Nevertheless, the acute pro-inflammatory response elicited by NMs in the al- responses were assessed at 2 h or 4 h in vitro and at veoli, both in vivo and in vitro. However, in our study we 24 h or 40 h in vivo, which brings uncertainties towards did not discriminate between the responses from the alveo- their comparative results. Teeguarden et al.  exposed lar epithelial cells and the macrophages. Moreover, it mice in vivo by inhalation or lung epithelial cells in vitro should be noted that although a ratio of one macrophage with suspensions of FeO NMs for 4 h with similar tim- for ten pneumocytes was used to mimic in vitro the ratio ings of the dose delivery and assessed the mass and re- present in vivo in rat or human lungs , the number of 2 2 2 gional deposition of the NMs. When focusing on the macrophages/cm (60,000/cm in inserts or 25,000/cm in tracheobronchal part of the lung, they observed plates) was higher in vitro compared to in vivo (around pro-inflammatory responses with doses about 10 to 6000 macrophages/cm ) because the A549 cell surface in 100-time lower in vivo than in vitro. However, they did vitro was much lower than the alveolar epithelial cell sur- not study the response of the alveolar part of the lung face in vivo. This could explain why in vivo and in vitro re- with this metric. Instead, they normalized the doses in sults were matching better when the doses were expressed mass by the number of alveolar macrophages in vivo in mass/macrophages rather by in mass/alveolar surface. and in vitro. Indeed, monocultures of murine macro- Ranking comparisons were also performed between phages were also exposed to suspensions in their study the four NMs tested. For ranking comparisons, we fo- . When focusing on the alveolar macrophages, they cused on dose intervals only. Indeed, LOAELs are de- showed that pro-inflammatory responses were triggered pending a lot on the experimental design as they are at similar doses in vitro and in vivo. strictly determined according to the doses tested. In the Interestingly, we also observed significant effects at case of few doses are tested and when NMs are observed much closer doses in vivo (LOAELs at 16 μg/10 macro- to be toxic only at the highest doses tested, like in our phages, BMDU around 12 μg/10 macrophages) and in study, LOAELs do not allow to make clear differences vitro using this dose metric, and more particularly at the between the NMs. However, with dose intervals deter- ALI (LOAELs at 16.7 μg/10 macrophages, BMDU from mined using benchmark dose-response modeling, more 13 to 45 μg/10 macrophages) and when the dose was accurate effect doses were determined for each NM and deposited in submerged conditions on the cells within each exposure method. Thanks to this criterion of dose, 3 h (LOAELs at 50 μg/10 macrophages, BMDU from a better screening of the four NMs has been performed. 25 to 900 μg/10 macrophage), rather than in 24 h The comparisons were performed using the mass metric (LOAELs at 400 μg/10 macrophages, BMDU from 250 but also with the surface area metric, since it was shown to 1000 μg/10 macrophage). As shown by using the that the surface area was the most effective dose metric mass/alveolar surface metric, similar differences were to explain acute NM toxicity in the lung [16, 17, 43]. observed using LOAELs and dose intervals, when com- Expressing the doses in mass (mass/alveolar surface or paring in vivo and in vitro biological activation levels mass/macrophages), similarity in the rankings were ob- using the mass/macrophage metric. Nevertheless, for the served between in vivo and in vitro conditions for the NMs 100 and 212 the differences of toxicity observed three TiO NMs. Both in vivo and in vitro, NMs 105 between submerged exposure in inserts for 3 h and and 101 appeared more toxic than NM100, except at the Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 15 of 20 ALI were similar toxicity was observed for the three and realistic in vitro methodologies allows to predict TiO NMs. However, it was not the case for the CeO more closely the biological responses observed in vivo 2 2 NM212. NM212 was observed as toxic as NMs 105 and and thus might give a better estimation of the potential 101 and more toxic than NM100 in vivo, whereas it was absolute toxicity in humans. observed to be less toxic than the TiO NMs 105 and Nevertheless, further improvements still need to be 101 in vitro. Moreover, we noticed that the ranking of made to draw clear conclusions. In our study, the ani- the NMs could change according to the dose metric mals were exposed by suspension instillation and not by used. Generally, when using the NM mass as dose inhalation of aerosols containing NMs. The instillation metric, the NM101 appeared as toxic as the NM105 and method remains less physiologic than the inhalation more toxic than the NMs 212 and 100. Nevertheless, al- route, especially because the dose is instantly deposited though we observed similarities in nanomaterial rank- into the lungs using a bolus. This could induce a greater ings between in vivo and in vitro approaches, biological response compared to inhalation, where the benchmark dose intervals were too large to make clear final dose is generally deposited within 4 h . More- ranking comparisons, due to the insufficient quality of over, although the instillation method allows to deposit the data-set. This underlines the importance of provid- NMs more deeply into the lungs , there was a lack ing good quality data to perform reliable comparisons. of accurate dosimetry in our study as the Multiple-Path Because of the insufficient quality of the data-set, it re- Particle Dosimetry Model (MPPD)  could not be mains thus undetermined if ALI exposure methods used. Thus, the regional deposition and more particu- could provide better predictivity than submerged larly the real dose distributed to the alveoli was not ac- methods regarding the ranking of the NMs. curately evaluated. When doses in mass were normalized by NM primary Furthermore, some limitations remain regarding the as- surface areas, the NM101, that has the highest surface sessment of dose delivery in vitro, more particularly in activity, was observed to be less toxic than expected and submerged conditions as the deposited dose on cells was clearly appeared less toxic than the other NMs. Indeed, estimated using the ISDD model and not directly mea- based on the surface reactivity theory which implies that sured. Nevertheless, the relative uncertainty was probably higher NM surface areas induce higher potential toxic- low as good similarities were observed between the esti- ities , similar responses were to be expected from mated and measured deposited doses of poorly soluble these three NMs when normalizing the dose by surface NMs at 24 h . At 3 h, the uncertainty may have been area. This has been shown in vivo  and in vitro . higher and could have led to an underestimation of the Because this was not the case, we hypothesized that the deposited dose . This may have contributed to in- hydrophobic surface coating that surrounds the NM101 creased differences between the ALI and the submerged but not NM105, NM100 and NM212 may have contrib- exposures in terms of biological activation levels. uted to reduce the toxic potential of NM101. This was Another reason why it is difficult to conclude clearly not surprising as it was shown in several studies that that the use of advanced and realistic in vitro method- NM acute toxicity was more dependent on coating than ologies might give a better estimation of the potential on core properties [46, 47]. absolute toxicity in humans is that some uncertainties However, when the doses in mass/macrophages were exist regarding the dose metrics selected. To compare normalized by surface areas calculated using mean ag- the in vivo to the in vitro approach, we normalized the glomerate sizes and densities, we did not observe this dose in vivo in mass by the total alveolar surface. We de- clear change in the NM101 ranking. Indeed, similarities cided to use the value of 4000 cm [18, 19], which in the rankings were observed between doses expressed seemed suitable for 7 weeks old male rats. Nevertheless, in mass/macrophages and in cm /macrophages when this may represent an overestimation as alveolar surfaces surface areas were calculated using mean agglomerate of around 2000 , and 3400 cm  have also been sizes and densities. This indicates that focusing on mean calculated for 6 weeks and 60 days old rats, respectively. surface areas in exposure media rather than on primary To normalize the dose by the number of macrophages, surface areas may better explain the biological responses we assumed that around 25 million of macrophages observed with poorly soluble NMs. Nevertheless, further were in the alveoli in vivo and we used the number of investigations are necessary to confirm this allegation. counted macrophages in vitro. Although we based our- Comparing several in vitro methods to the in vivo ap- selves on two publications [18, 54] to determine the proach, that was considered as the reference method in number of alveolar macrophages in vivo, it remains un- our study to estimate the potential toxicity of NMs in clear whether all of them were in contact and contrib- humans, allowed us to evaluate the predictive ability of uted to the biological response elicited by the NMs, different in vitro system in absolute terms. Finally, ac- more particularly considering that only around 8 million cording to our results, it seems that the use of advanced of macrophages were retrieved in the BALF in vivo. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 16 of 20 Nevertheless, we decided to use the value of 25 million critical effect dose intervals could be used instead of of macrophages instead of the measured value of 8 mil- LOAELs to provide more accurate comparisons between lion because we observed that the number of macro- the NMs. Regarding toxicity rankings of NMs, relative phages retrieved from the BALF was depending a lot on similarities were shown between in vivo and in vitro the experimenter and because the protocol used in our methodologies. Nevertheless, we could not conclude study was not implemented to retrieve all the alveolar clearly about each in vitro methodology ability to predict macrophages. the NM rankings observed in vivo because the quality of Some uncertainties also remain because our experi- the data-set was insufficient to determine accurate dose mental data-set did not allow to provide a clear intervals. Interestingly, we also observed when normaliz- dose-response pattern of toxicity. That may had an im- ing the doses by NM surface areas, that the toxic effects pact on the accuracy of our comparisons. For example were probably more attributable to agglomerates, rather in vivo, there was a difference of a factor of ten between than to isolate NMs. each dose tested; this might prevent us to determine ac- In conclusion, we showed that advanced methods curate LOAELs and dose intervals. This is particularly could be used to enhance the in vitro experiments ability true because the pro-inflammatory effects were observed to predict potential acute pulmonary toxicity in vivo. at the highest dose tested. Although using intermediary Moreover, we highlighted that careful consideration of doses might have enabled to determine more precisely some key methodological points in vitro could contrib- LOAELs and critical dose intervals, this has no impact ute to improve in vitro methods predictivity, including on our general conclusions regarding comparison of bio- control of the timing of the dose delivery. Although logical activation levels between the different exposure these conclusions are inferred from our experimental methods used in our study: regardless the criterion of data-set and should be further confirmed with other comparison used, the in vivo methodology remains the nanomaterials, including more toxic NMs, this study most sensitive one in our study, to predict potential ad- brings new perspectives regarding the usage and devel- verse effects after acute exposure to poorly soluble NMs. opment of advanced in vitro methods. Regarding NM rankings, we observed that it was difficult to use LOAELs to rank NMs in function of each expos- Methods ure methodology used and that determining dose inter- Nanomaterials vals using benchmark dose-response modeling was very Four poorly toxic and poorly soluble NMs were used in important for this purpose. However, because the the study. The TiO NM100 and NM101 and the CeO 2 2 data-set quality used in our study was not optimal, the NM212 were obtained from the Joint research center dose intervals determined were too large to provide clear (JRC). The TiO NM105 was obtained from Evonik In- and reliable comparisons of NM rankings between each dustries (AEROXIDE® TiO P25). Data indicated in our methodology used. To perform in vivo - in vitro com- study regarding primary sizes and specific surface areas parisons we thus recommend to test more doses and to (BET) were provided by the manufacturer (Table 1). reduce the interval between each doses, in order to de- TiO and CeO primary physico-chemical properties 2 2 termine more accurate dose intervals. were also well characterized by the JRC. [55, 56]. The endotoxin levels of the NMs were tested by partners of Conclusion the European project NANoREG. They were below the Quantitative comparisons were performed between in limit of detection (data not shown). vivo and in vitro acute pro-inflammatory responses using compatible dose metrics. Biological activation In vivo study levels were compared and we showed better in vivo- in Animals vitro correlations when doses were expressed in mass/ Pathogens free 7 weeks old male rats (WISTAR RjHan:WI, macrophages rather than in mass/alveolar surface. Using JANVIER LABS, France; 250 g), were housed in polycar- the determined LOAELs and critical effect dose inter- bonate cages, in a temperature and humidity controlled vals, we assessed the ability of each in vitro method used room, and had free access to food and water ad libitum. in our study to predict the biological responses in vivo. All the in vivo experiments were approved by the “Comité We showed that the most realistic in vitro exposure Régional d’Ethique en Matière d’Expérimentation Animale method: the ALI method, was the most predictive in de Picardie” (CREMEAP) (C2EA – 96). terms of absolute toxicity, whatever the dose metric used. In vitro, we also showed better vivo-vitro correla- Preparation of NM suspensions tions while using timings of dose delivery of 3 h rather Similarly as for the in vitro study, suspensions of TiO than 24 h. For each exposure method, we ranked NMs (NM105, NM101, NM100) and CeO (NM212) at in function of their toxicity and we highlighted that 10 mg/mL in Mili-Q water were prepared and then Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 17 of 20 sonicated at amplitude 100 during 2 min (1 min on, Collected BALFs were centrifuged at 350 g for 10 min 1 min off, 1 min on) using a cuphorn sonicator (QSO- and 4 °C, to separate the cells from the supernatant. The NICA, Q700). Suspensions in Milli-Q water at 5; 0.5 and supernatants recovered from the first lavage (around 0.05 mg/mL were prepared to expose rats to 500; 50 and 4.5 mL for each sample) were aliquoted in eppendorf 5 μg/animal, respectively. tubes and stored at − 80 °C until analysis. Characterization of NM suspensions Cell counting After centrifugation, the cell pellets were For each NM, DLS measurements were performed resuspended in 5 mL of RPMI medium (Gibco, 61870), (Malvern, Zetasizer Nano S) on NM suspensions to meas- 20 μL of cell suspension were mixed with 20 μL of pro- ure the hydrodynamic diameter and to assess the size dis- pidium iodide containing accridine solution (Nexcelom, tribution of the particles in suspensions. DLS results on CS2-0106) and the cells were counted using a cell coun- water suspensions used to instill animals are presented in ter equipped with a fluorometer (Nexcelom, Cellometer® the Additional files section (Additional file 1:FigureS4). Auto 2000), to differentiate the dead cells and the eryth- Regarding in vitro experiments, DLS measurements were rocytes from the pulmonary cells. performed after sonication in stock suspensions (2.56 mg/ mL in milli-Q water) and just after dilution in 0.4 mg/mL BALF cytology After counting, the cells were diluted in suspensions in culture medium. These in vitro results RPMI, seeded on slides at 300000 cells/spots using a were presented in our previous article . Surprisingly, cytospin (300 g, 5 min) (Shandon, cytospin2) and similar results were observed between NM suspended in then fixated and coloured in May-Grunwald Giemsa water and in culture medium. (MGG). Briefly, the slides were fixated in MG pure for3minfollowedby 2mininMGdiluted at 50% in Intratracheal instillation Mili-Q water, rinsed 2 times with Mili-Q water for Rats were anesthetized (0.5 mg/kg ketamine hydro- 20 min and then coloured in Giemsa. The percentage chloride, 0.1 mg/kg atropine and 1 mg/kg xylazine), of the different cell types (macrophages, neutrophils, endotracheal intubation was performed using a canula eosinophil) in BALF was then determined using op- and animals were connected to a respirator (Harvard tical microscopy. Apparatus, ventilator model 683) for 30 s to create a hyperventilation. Rats were disconnected from the ap- Intracellular ROS levels (DCF assay) After counting, paratus, 100 μL of NMs suspension in water or vehicle the BALF cells were seeded at 1 × 10 cells/mL in 24 (Mili-Q water) was added in the cannula and suspen- well plates (Falcon, 353047) (in RPMI medium supple- sions were directly aerosolized into rat lungs by physio- mented with 10% of FCS: 0,5 mL/well), and were then logical aspiration. It was chosen to disperse NMs in incubated for 18 h at 37 °C and with 5% of CO to let Milli-Q water and not in physiological saline buffer to the cells (mostly macrophages) to adhere on the plate. enhance NM stability in suspension. This choice was The cells were then rinsed with PBS and incubated for made since it was shown that intratracheal instillation 35 min with 10 μM of 5-(and-6)-chloromethyl-2′,7′-di- of distilled water in rats, like physiological saline, did chlorodihydrofluorescein diacetate (CM-H DCFDA) not induce significant inflammatory responses at 24 h . probe (Life technologies, C6827) in PBS (0.5 mL/well). After 30 min of incubation, the probe was removed in Dosing and biodistribution analysis some control wells, 1 mM of H O in PBS was added 2 2 After instillation, rats were sacrificed 3 h after instillation and the cells were incubated for 5 min, to serve as posi- to evaluate the lung burden (n = 2). Mass of NM was mea- tive control. After incubation, the cells were washed sured in collected lungs by inductively coupled plasma with PBS and incubated for 5 min in 90% of Dimethyl mass spectrometry (ICP-MS) analysis. Briefly, a procedure Sulfoxide (DMSO) (Sigma-Aldrich, D2438) in PBS consisting of incubation with a mixture of nitric acid (0.5 mL/well). The cells were then scraped using (HNO3) and hydrofluoric acid (HF), and heating was ap- scrapers (TPP, 99002), the well contents were retrieved plied to digest lungs and TiO nanomaterials in order to in tubes (Eppendorf, 3810X) and the tubes were centri- determine the total Ti content by ICP-MS . fuged at 10000 g and 4 °C for 5 min, to eliminate the dead cells and to remove the remnants particles. The Assessment of biological activity tube contents were transferred in 96 well black plate Animals (n = 6) were sacrificed 24 h after instillation and (150 μL/well) (Greiner Bio-one, 655076) and the fluor- bronchoalveolar lavages were performed with PBS. A escence of the samples was read (excitation: 488 nm, first bronchoalveolar lavage was performed using 5 mL emission: 530 nm) using a spectrophotometer (TECAN, of PBS for biochemical analysis. Two other lavages were infinite 2000). The value of each sample was expressed in then performed with 10 mL of PBS to collect more cells. percentage of intracellular ROS compared to the control. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 18 of 20 Pro-inflammatory release in BALF supernatants IL-8, TNF-α levels in culture medium (ELISA)), and Il-1β, IL-6, IL-8 and TNF-α releases were measured in oxidative stress assays (DCF assay). collected supernatants using a commercial available ELISA multiplex kit (Mesoscale discovery, Proinflamma- Deposited dose assessment tory Panel 2, N05059A-1) and a multiplex reader In vivo, the mass of each NM instilled into the lungs (Mesoscale discovery, Sector Imager 24000) according to was measured by ICP-MS. The nominal doses (5; 50; supplier recommendations. 500 μg/animal) were corrected to 4; 40 and 400 μg/ani- mal according to dosage results (Additional file 1: Figure LDH release and protein levels in BALF supernatants S5). According to the lung alveolar surface (4000 cm ) LDH release were quantified in BALF using a commer- or the number of alveolar macrophages (25 million), this cially available kit (Promega, CytoTox-ONE Homoge- corresponds to theoretical deposited doses in lungs of 2 6 neous Membrane Integrity assay). Proteins levels were around 0.1; 0.01; 0.001 μg/cm or 16; 1.6; 0.16 μg/10 measured in BALF using a Bradford assay (Biorad, pro- macrophages, respectively. tein assay kit). In vitro, the real mass deposited on the cells was ei- ther assessed by ICP-MS dosage (for ALI exposures) or estimated (in submerged conditions) using the in vitro Statistical analysis All data were expressed as mean ± sedimentation diffusion and dosimetry model (ISDD) standard deviation (SD) (n = 6). Statistical analyses were , after measuring the hydrodynamic diameter by dy- performed using Graphpad Prism 5.0 (GraphPad Soft- namic light scattering and the effective density of the ware Inc., San Diego, CA). Results were analyzed by a agglomerates following the Volumetric Centrifugation non-parametric Kruskal-Wallis test followed by Dunn’s Method (VCM) . The detailed material and post-hoc test to compare the different treated groups to methods used in vitro and all the deposition data are the non-exposed control. available in the following paper . Deposited masses on cells in vitro are also presented in the Additional In vitro study files section of the present manuscript (Additional file All materials and methods used in the in vitro study are 1: Tables S1 and S6). The final measured or calculated fully detailed in the following article . Briefly, alveolar doses tested were around 0,1; 1; 3 μg/cm at the ALI epithelial cells (A549) in co-culture with macrophages (for 3 h of maintained deposition + 21 h without depos- (THP-1) were exposed either at the ALI to aerosols or in ition in the incubator), 1; 3; 10 μg/cm in submerged in submerged condition to suspensions of TiO and CeO 2 2 inserts (for 3 h of maintained deposition + 21 h without NMs. A ratio of ten A549 for one THP-1 was used to deposition in the incubator) and 1, 3, 10, 20 μg/cm in mimic the ratio existing in vivo in the lungs. submerged in plates (24 h of maintained deposition). Different timings of the dose delivery were used in vitro. At the ALI, cells were exposed to aerosol of NMs using a Vitrocell® system. In this system, the NM depos- Determination of critical dose intervals using benchmark ition was maintained for 3 h, meaning that the final dose-response modeling dose was reached within 3 h. The cells were then kept All the in vivo and in vitro data were analyzed using the in the incubator for the remaining 21 h at the ALI with benchmark dose-response modeling software PROAST the NMs deposited on their surface. In submerged con- (RIVM, Bilthoven, The Netherlands). The PROAST soft- ditions, two different dose rates were used. Cells were ware selects the optimal data fitting model from an ex- exposed to suspensions of NMs in inserts using similar ponential family of models. Briefly, for each cytokine timing of thedosedeliveryasatthe ALI. TheNM de- and each exposure method used (in vivo, ALI (3 h + position was maintained for 3 h. After 3 h of exposure, 21 h), submerged in inserts (3 h + 21 h), submerged in the deposition was stopped by replacing NM suspen- plates (24 h)), we determined the critical effect dose cor- sions by fresh medium and cells were then kept during responding to a 20% increase of pro-inflammatory medi- the remaining 21 h in submerged condition in the incu- ator levels compared to non-exposed controls and the bator. Cells were also exposed in plates to suspensions benchmark dose lower confidence limit (BMDL) and the of NMs for 24 h, to represent the exposure conditions benchmark dose upper confidence limit (BMDU) of the usually used in vitro. In that situation, the NM depos- interval for a 90% confidence. For each exposure method ition was maintained for the whole exposure time, used and each NM, we then calculated the median value meaning that the final dose was reached within 24 h. of the BMDL and the median value of the BMDU of the After 24 h of exposure, the biological activity of the four pro-inflammatory mediators (IL-1β, IL-6, IL-8/ cells was assessed for all methodologies using cytotox- KC-GRO, TNF-α), to determine a median dose interval icity (Alamar blue, LDH), inflammation (IL-1b, IL-6, for general pro-inflammatory response. We decided to Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 19 of 20 calculate the median dose interval of the four cytokines Ethics approval All the in vivo experiments were approved by the “Comité Régional d’Ethique because similar results could be observed when pooling en Matière d’Expérimentation Animale de Picardie” (CREMEAP) (C2EA – 96). the results of the four cytokines and when comparing dose intervals of each cytokine one by one. We believe Competing interests that our comparisons were easier to interpret in our The authors declare that they have no competing interests. study when using a general pro-inflammatory response. We choose a critical effect of 20% based on the magni- Publisher’sNote tude of effect in several notable studies [16, 60, 61]. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Author details Additional files Institut National de l’Environnement Industriel et des Risques (INERIS), (DRC/ VIVA/TOXI), Parc Technologique ALATA - BP 2, F-60550 Verneuil-en-Halatte, Additional file 1: Figure S1. Levels of pro-inflammatory mediators in France. Université de Technologie de Compiègne (UTC), Laboratoire cell supernatants in vitro (adapted from ). Figure S2. Levels of BioMécanique et BioIngénierie (BMBI), UMR CNRS 7338, 60205 Compiègne, proteins, LDH (cytotoxicity) and intracellular ROS (oxidative stress) in BALF. France. Department of Biomedical Engineering, Tufts University, Medford, Figure S3. Examples of critical effect doses (CED) and dose intervals MA, USA. (CEDL/BMDL and CEDU/BMDU) determined using benchmark dose response modeling. Figure S4. Size distribution of the NMs in the Received: 7 November 2017 Accepted: 9 May 2018 suspensions used to expose rats. Figure S5. Initial lung burden in vivo assessed by ICP-MS 3 h after instillation (n =2). Table S1. Doses deposited in vitro in submerged conditions in function of nominal concentrations in References suspensions (First published in ). Table S2. LOAELs (in μg/cm ) 1. Bakand S, Hayes A, Dechsakulthorn F. Nanoparticles: a review of particle determined in vitro with the pro-inflammatory effects for each exposure toxicology following inhalation exposure. Inhal Toxicol. 2012;24(2):125–35. method used (First published in ). Table S3. Dose intervals (in μg/10 2. Piccinno F, Gottschalk F, Seeger S, Nowack B. Industrial production macrophages) determined for each NM and each methodology. Table S4. 2 6 quantities and uses of ten engineered nanomaterials in Europe and the Dose intervals normalized by primary surface areas (in cm /10 - world. J Nanopart Res. 2012;14(9):1109. macrophages) for each NM and methodology. Table S5. Dose intervals 2 6 3. Oomen AG, Bos PM, Fernandes TF, Hund-Rinke K, Boraschi D, Byrne HJ, et normalized by agglomerate surface areas (in cm /10 macrophages) for al. Concern-driven integrated approaches to nanomaterial testing and each NM and methodology. Table S6. Characterization of mass deposited assessment–report of the NanoSafety Cluster Working Group 10. in vitro on cells after 3 h exposure at the ALI to aerosols of NMs (Adapted Nanotoxicology. 2014;8(3):334–48. from ). (DOCX 831 kb) 4. Nel A, Xia T, Meng H, Wang X, Lin S, Ji Z, et al. Nanomaterial toxicity testing in the 21st century: use of a predictive toxicological approach and high- throughput screening. Acc Chem Res. 2013;46(3):607–21. Abbreviations 5. Joris F, Manshian BB, Peynshaert K, De Smedt SC, Braeckmans K, Soenen SJ. ALI: Air-Liquid Interface; BALF: BronchoAlveolar Lavage Fluids; BET: Brunauer– Assessing nanoparticle toxicity in cell-based assays: influence of cell culture Emmett–Teller; BMDL: Benchmark Dose Lower confidence limit; parameters and optimized models for bridging the in vitro-in vivo gap. BMDU: Benchmark Dose Upper confidence limit; DLS: Dynamic Light Chem Soc Rev. 2013;42(21):8339–59. Scattering; DMSO: Dimethyl Sulfoxide; ICP-MS: Inductively Coupled Plasma - 6. BeruBe K, Aufderheide M, Breheny D, Clothier R, Combes R, Duffin R, et al. In Mass Spectrometry; ISDD: In vitro Sedimentation Diffusion and Dosimetry; vitro models of inhalation toxicity and disease. The report of a FRAME JRC: Joint Research center; LDH: Lactate DeHydrogenase; LOAEL: Lowest workshop. ATLA. 2009;37(1):89–141. Observed Adverse Effect Level; MGG: May-Grunwald Giemsa; 7. Paur HR, Cassee FR, Teeguarden J, Fissan H, Diabate S, Aufderheide M, et al. NM(s): NanoMaterial(s); NOAEL: No Observed Adverse Effect Level; In-vitro cell exposure studies for the assessment of nanoparticle toxicity in PMA: Phorbol Myristate Acetate; ROS: Reactive Oxygen Species; SD: Standard the lung—a dialog between aerosol science and biology. J Aerosol Sci. Deviation; VCM: Volumetric Centrifugation Method 2011;42(10):668–92. 8. Sayes CM, Reed KL, Warheit DB. Assessing toxicity of fine and nanoparticles: Acknowledgements comparing in vitro measurements to in vivo pulmonary toxicity profiles. We thank Christelle Gamez, Kelly Blazy, Franck Robidel, Anthony Lecomte Toxicol Sci. 2007;97(1):163–80. and Emanuel Peyret for their technical assistance regarding animal experimentations. 9. Rushton EK, Jiang J, Leonard SS, Eberly S, Castranova V, Biswas P, et al. We thank Camille Rey for the critical revision and the language corrections. Concept of assessing nanoparticle hazards considering nanoparticle dosemetric and chemical/biological response metrics. J Toxicol Environ Health A. 2010;73(5):445–61. Funding 10. Kim YH, Boykin E, Stevens T, Lavrich K, Gilmour MI. Comparative lung This work was supported by the French ministry of environment (“Programme toxicity of engineered nanomaterials utilizing in vitro, ex vivo and in vivo 190”) and by the EU-FP7 (NANoREG project). approaches. J Nanobiotechnol. 2014;12:47. 11. Teeguarden JG, Mikheev VB, Minard KR, Forsythe WC, Wang W, Sharma G, Availability of data and materials et al. Comparative iron oxide nanoparticle cellular dosimetry and response The datasets supporting the conclusions of this article are included within in mice by the inhalation and liquid cell culture exposure routes. Part Fibre the article (and its additional files) and in the following paper  for the in Toxicol. 2014;11:46. vivo and in vitro experimentations, respectively. 12. Donaldson K, Schinwald A, Murphy F, Cho WS, Duffin R, Tran L, et al. The biologically effective dose in inhalation nanotoxicology. Acc Chem Res. Authors’ contributions 2013;46(3):723–32. The manuscript was written through contributions of all authors. All authors 13. Loret T, Peyret E, Dubreuil M, Aguerre-Chariol O, Bressot C, le Bihan O, et al. gave approval to the final version of the manuscript. TL designed and Air-liquid interface exposure to aerosols of poorly soluble nanomaterials performed the experiments, analyzed and interpreted the data, wrote and induces different biological activation levels compared to exposure to revised the manuscript. FR was involved in the design, the performance of in suspensions. Part Fibre Toxicol. 2016;13(1):58. vivo experiments. BT, AB, CE were involved in the writing and the revisions. GL 14. U.S. EPA. Air quality criteria for particulate matter (Final Report, 2004). U.S. designed the study and was involved in the performance of the in vivo Environmental Protection Agency, Washington, DC, EPA 600/P-99/002aF-bF. experiments, the writing and the revisions. 2004. https://cfpub.epa.gov/si/si_public_record_report.cfm?dirEntryId=87903. Loret et al. Particle and Fibre Toxicology (2018) 15:25 Page 20 of 20 15. Schmid O, Stoeger T. Surface area is the biologically most effective dose metric 40. Jing X, Park JH, Peters TM, Thorne PS. Toxicity of copper oxide nanoparticles for acute nanoparticle toxicity in the lung. J Aerosol Sci. 2016;99:133–43. in lung epithelial cells exposed at the air-liquid interface compared with in 16. Braakhuis HM, Cassee FR, Fokkens PHB, de la Fonteyne LJJ, Oomen AG, vivo assessment. Toxicol in Vitro. 2015;29(3):502–11. Krystek P, et al. Identification of the appropriate dose metric for pulmonary 41. Panas A, Marquardt C, Nalcaci O, Bockhorn H, Baumann W, Paur HR, et al. inflammation of silver nanoparticles in an inhalation toxicity study. Screening of different metal oxide nanoparticles reveals selective toxicity Nanotoxicology. 2016;10(1):63–73. and inflammatory potential of silica nanoparticles in lung epithelial cells 17. Duffin R, Tran L, Brown D, Stone V, Donaldson K. Proinflammogenic effects and macrophages. Nanotoxicology. 2013;7(3):259–73. of low-toxicity and metal nanoparticles in vivo and in vitro: highlighting the 42. Wiemann M, Vennemann A, Sauer UG, Wiench K, Ma-Hock L, Landsiedel R. role of particle surface area and surface reactivity. Inhal Toxicol. 2007;19(10): An in vitro alveolar macrophage assay for predicting the short-term 849–56. inhalation toxicity of nanomaterials. J Nanobiotechnol. 2016;14:16. 43. Stoeger T, Schmid O, Takenaka S, Schulz H. Inflammatory response to TiO2 18. Stone KC, Mercer RR, Gehr P, Stockstill B, Crapo JD. Allometric relationships and carbonaceous particles scales best with BET surface area. Environ of cell numbers and size in the mammalian lung. Am J Respir Cell Mol Biol. Health Perspect. 2007;115(6):A290-A1. 1992;6(2):235–43. 44. Oberdorster G, Oberdorster E, Oberdorster J. Nanotoxicology: an emerging 19. Ohashi T, Pinkerton K, Ikegami M, Jobe AH. Changes in alveolar surface-area, discipline evolving from studies of ultrafine particles. Environ Health surfactant protein-A, and saturated phosphatidylcholine with postnatal rat Perspect. 2005;113(7):823–39. lung growth. Pediatr Res. 1994;35(6):685–9. 45. Braakhuis HM, Giannakou C, Peijnenburg WJ, Vermeulen J, van Loveren H, 20. Crapo JD, Barry BE, Gehr P, Bachofen M, Weibel ER. Cell number and cell Park MV. Simple in vitro models can predict pulmonary toxicity of silver characteristics of the normal human lung. Am Rev Respir Dis. 1982;126(2):332–7. nanoparticles. Nanotoxicology. 2016;10(6):770–9. 21. OECD. Guidance document on acute inhalation toxicity testing. ENV/JM/ 46. Fubini B, Ghiazza M, Fenoglio I. Physico-chemical features of engineered MONO(2009)28. Series on testing and assesment: number 39. 2009. nanoparticles relevant to their toxicity. Nanotoxicology. 2010;4(4):347–63. 22. Geiser M, Kreyling WG. Deposition and biokinetics of inhaled nanoparticles. 47. Demokritou P, Gass S, Pyrgiotakis G, Cohen JM, Goldsmith W, McKinney W, Part Fibre Toxicol. 2010;7:2. et al. An in vivo and in vitro toxicological characterisation of realistic 23. Lehnert BE. Pulmonary and thoracic macrophage subpopulations and nanoscale CeO2 inhalation exposures. Nanotoxicology. 2013;7(8):1338–50. clearance of particles from the lung. Environ Health Perspect. 1992;97: 48. Baisch BL, Corson NM, Wade-Mercer P, Gelein R, Kennell AJ, Oberdorster G, 17–46. et al. Equivalent titanium dioxide nanoparticle deposition by intratracheal 24. Mann-Jong M, Shih L, Wu R. Pulmonary epithelium: cell types and functions. instillation and whole body inhalation: the effect of dose rate on acute In: Proud M, editor. The pulmonary epithelium in health and disease. West respiratory tract inflammation. Part Fibre Toxicol. 2014;11:5. Sussex: Wiley; 2008. 49. Wong BA. Inhalation exposure systems: design, methods and operation. 25. Fehrenbach H. Alveolar epithelial type II cell: defender of the alveolus Toxicol Pathol. 2007;35(1):3–14. revisited. Respir Res. 2001;2(1):33–52. 50. Anjilvel S, Asgharian B. A multiple-path model of particle deposition in the 26. Blank F, Rothen-Rutishauser BM, Schurch S, Gehr P. An optimized in vitro rat lung. Fundam Appl Toxicol. 1995;28(1):41–50. model of the respiratory tract wall to study particle cell interactions. J 51. Cohen JM, Teeguarden JG, Demokritou P. An integrated approach for the in Aerosol Med. 2006;19(3):392–405. vitro dosimetry of engineered nanomaterials. Part Fibre Toxicol. 2014;11:20. 27. Park EK, Jung HS, Yang HI, Yoo MC, Kim C, Kim KS. Optimized THP-1 52. Pinkerton KE, Barry BE, Oneil JJ, Raub JA, Pratt PC, Crapo JD. Morphologic differentiation is required for the detection of responses to weak stimuli. changes in the lung during the lifespan of Fischer 344 rats. Am J Anat. Inflamm Res. 2007;56(1):45–50. 1982;164(2):155–74. 28. Wottrich R, Diabate S, Krug HF. Biological effects of ultrafine model particles 53. Massaro GD, Mortola JP, Massaro D. Sexual dimorphism in the architecture of in human macrophages and epithelial cells in mono- and co-culture. Int J the lungs gas-exchange region. Proc Natl Acad Sci U S A. 1995;92(4):1105–7. Hyg Environ Health. 2004;207(4):353–61. 54. Miller FJ. Dosimetry of particles in laboratory animals and humans in 29. Napierska D, Thomassen LC, Vanaudenaerde B, Luyts K, Lison D, Martens JA, relationship to issues surrounding lung overload and human health risk et al. Cytokine production by co-cultures exposed to monodisperse assessment: a critical review. Inhal Toxicol. 2000;12(1):19–57. amorphous silica nanoparticles: the role of size and surface area. Toxicol 55. Singh C, Friedrichs S, Ceccone G, Gibson N, Jensen K, Levin M, et al. Cerium Lett. 2012;211(2):98–104. dioxide, NM-211, NM-212, NM-213. Characterisation and test item 30. Rothen-Rutishauser B, Blank F, Muhlfeld C, Gehr P. In vitro models of the preparation, JRC repository: NM-series of representative manufactured human epithelial airway barrier to study the toxic potential of particulate nanomaterials Ispra, Italy: European Commission Joint Research Centre matter. Expert Opin Drug Metab Toxicol. 2008;4(8):1075–89. Institute for health and consumer protection. 2014. 31. Kim JS, Peters TM, O’Shaughnessy PT, Adamcakova-Dodd A, Thorne PS. 56. Rasmussen K, Mast J, De Temmerman P-J, Verleysen E, Waegeneers N, Van Validation of an in vitro exposure system for toxicity assessment of air- Steen F, et al. Titanium dioxide, NM-100, NM-101, NM-102, NM-103, NM-104, delivered nanomaterials. Toxicol in Vitro. 2013;27(1):164–73. NM-105: characterisation and physico-chemical properties. JRC science and 32. Xie YM, Williams NG, Tolic A, Chrisler WB, Teeguarden JG, Maddux BLS, et al. policy reports. 2014. Aerosolized ZnO nanoparticles induce toxicity in alveolar type II epithelial 57. Krystek P, Tentschert J, Nia Y, Trouiller B, Noel L, Goetz ME, et al. Method cells at the air-liquid interface. Toxicol Sci. 2012;125(2):450–61. development and inter-laboratory comparison about the determination of 33. Landsiedel R, Sauer UG, Ma-Hock L, Schnekenburger J, Wiemann M. titanium from titanium dioxide nanoparticles in tissues by inductively coupled Pulmonary toxicity of nanomaterials: a critical comparison of published in plasma mass spectrometry. Anal Bioanal Chem. 2014;406(16):3853–61. vitro assays and in vivo inhalation or instillation studies. Nanomedicine 58. Hinderliter PM, Minard KR, Orr G, Chrisler WB, Thrall BD, Pounds JG, et al. (Lond). 2014;9(16):2557–85. ISDD: a computational model of particle sedimentation, diffusion and target 34. Hoet PH, Nemery B, Napierska D. Intracellular oxidative stress caused by cell dosimetry for in vitro toxicity studies. Part Fibre Toxicol. 2010;7(1):36. nanoparticles: what do we measure with the dichlorofluorescein assay? 59. DeLoid G, Cohen JM, Darrah T, Derk R, Rojanasakul L, Pyrgiotakis G, et al. Nano Today. 2013;8(3):223–7. Estimating the effective density of engineered nanomaterials for in vitro 35. Driscoll KE, Carter JM, Hassenbein DG, Howard B. Cytokines and particle-induced dosimetry. Nat Commun. 2014;5:3514. inflammatory cell recruitment. Environ Health Perspect. 1997;105:1159–64. 60. Gosens I, Mathijssen LE, Bokkers BG, Muijser H, Cassee FR. Comparative 36. Oberdorster G, Finkelstein JN, Johnston C, Gelein R, Cox C, Baggs R, et al. hazard identification of nano- and micro-sized cerium oxide particles based Acute pulmonary effects of ultrafine particles in rats and mice. Res Rep on 28-day inhalation studies in rats. Nanotoxicology. 2014;8(6):643–53. Health Eff Inst. 2000;96:5–74; disc 5–86 61. Stoeger T, Reinhard C, Takenaka S, Schroeppel A, Karg E, Ritter B, et al. 37. Organization WH. Guidance document on evaluating and expressing Instillation of six different ultrafine carbon particles indicates a surface area uncertainty in hazard characterization. 2018. threshold dose for acute lung inflammation in mice. Environ Health 38. Hardy A, Benford D, Halldorsson T, Jeger MJ, Knutsen KH, More S, et al. Perspect. 2006;114(3):328–33. Update: use of the benchmark dose approach in risk assessment. EFSA J. 2017;15(1):4658. 39. Slob W. Dose-response modeling of continuous endpoints. Toxicol Sci. 2002;66(2):298–312.
Particle and Fibre Toxicology – Springer Journals
Published: Jun 4, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud