Curcumin (Sigma cat no. C1386), dichlorvos (Sigma cat no. 45441), and mannitol (Sigma cat no. M9647) were purchased from Sigma Aldrich (Amsterdam, Netherlands), and dimethyl sulfoxide (DMSO) (Arcos cat no. 167852500) was purchased from Acros Organics (Geel, Belgium). Dulbecco’s Modified Eagle Medium with Ham’s Nutrient Mixture F-12 (1:1) (DMEM/F12) without phenol red (Gibco cat no. 31331-028), Trypsin 0.5% EDTA (10x) (Gibco cat no. 15400-054), nonessential amino acids (NEAA) (Gibco cat no. 11140-035), and phosphate-buffered saline (Gibco cat no. 20012019) were from Gibco (Carlsbad, CA), geneticin (G418) (Duchefa cat no. G0175001) from Duchefa (Haarlem, Netherlands), and penicillin/streptomycin, pH 7.4 (P/S) (Invitrogen cat no. 15070063) from Invitrogen (Breda, Netherlands). Fetal bovine serum (FBS) (Gibco cat no. 10270-106) and dextran-coated charcoal-stripped fetal calf serum (DCC-FCS) (Gibco cat no. 12676029) were both purchased from Gibco.

Fe₃O₄-PEG-PLGA nanomaterials were provided by Colorobbia Holding S.p.A (Firenze, Italy) in the framework of the BIORIMA research project and synthesized as described (D'Elios et al., 2018). Both Ag and TiO₂ nanomaterials (designated NM300K and NM101, respectively) were from the nanomaterial repository of the Joint Research Center of the European Commission (Ispra, Italy). Ag nanomaterials were provided as a suspension. The NANOGENOTOX protocol was used for dispersion of TiO₂ (Farcal et al., 2015).

The Limulus Amoebocyte Lysate (LAL) assay was applied to detect bacterial endotoxin contamination as described earlier (Kroll et al., 2013; Eder et al., 2022). The Limulus Amoebocyte Lysate PYROTELL^®–T assay was purchased from Associates of Cape Cod, Inc. (East Falmouth, MA) and used according to the manufacturer’s instructions. Data analysis was performed using PYROS^® Software (Associates of Cape Cod, Inc.).

U2OS-NRF2 cells were kindly provided by Bio Detection Systems (Amsterdam, Netherlands). The human osteoblastic osteosarcoma U2OS-NRF2 cells (van der Linden et al., 2014) express two oligos containing four different EPRE sequences: 1 × consensus EPRE (TCA​CAG​TGA​CTA​AGC​AAA​AT), 1 × hNQO1 EPRE (TCACAGTGAC TCAGCA-GAAT), 1 × hGCLM EPRE (AGA​CAA​TGA​CTA​AGC​AGA​AA) and 1 × hGCLC EPRE(TCACAGTCAGTAAGTGATGG). The two oligos were ligated into a promoter-less luciferase reporter-construct pLuc. Because the U2OS cells express the NRF2 pathway endogenously, a selection construct (pSG5-neo) was used. The cells were cultured in DMEM/F12 supplemented with 10% FCS and penicillin/streptomycin (final concentrations 10 U/ml and 10 μg/ml, respectively) (designated as growth medium). Once per week, 200 μg/ml G418 was added to the culture medium to maintain selection pressure. Cells were maintained at 37°C in a humidified atmosphere with 5% CO₂.

Cytotoxicity of nanobiomaterials was evaluated by the Alamar blue (resazurin) assay as described (Keshavan et al., 2021). The cell viability experiments were performed by one of the participating laboratories, prior the “round robin” pre-validation experiments. The cells were trypsinized, counted, and resuspended in cell culture medium without phenol red and supplemented with 5% dextran-coated charcoal-stripped FCS (DCC–FCS), to a final concentration of 10⁴cells/well (100 µl). Cells were seeded in 96-well plates and exposed to test materials or were maintained in DCC–FCS alone (negative control). The assay reagent (Thermo Scientific, Sweden) (10% [v/v] solution of AlamarBlue^® reagent) was added to each well to monitor the cellular metabolic function. The samples were analyzed using a spectrophotometer (Tecan Infinite^® F200).

The potential induction of NRF2 mediated gene expression by nanobiomaterials was tested by measuring the induction of luciferase activity in the NRF2-U2OS cells. Protocols are available upon request. In brief, the cells were trypsinized, counted, and resuspended in cell culture medium without phenol red and supplemented with 5% dextran-coated charcoal-stripped FCS (DCC–FCS) at a final concentration of 10⁴cells/well (100 µl) in a 96-well plate without using the most outer wells. The plates were incubated for 24 h in a humidified atmosphere at 37°C under 5% CO₂. Following this pre-incubation one reference plate was exposed containing 9 serial dilutions in the range of 1 × 10⁻⁴ M to 1 × 10⁻⁸ M (log10 dilution steps) of the reference compound curcumin, as well as a positive control dichlorvos (1 × 10⁻⁵–7 × 10⁻⁷ M) and a negative control mannitol (1 × 10⁻³–1 × 10⁻⁵ M). Dichlorvos was included as a positive control as it is known to induce a response in this assay, while the negative control (i.e., mannitol) should not. Curcumin was chosen as reference compound, as it usually results in a dose-effect response in the current assay. It is good practice to select different chemicals as reference chemical and positive control. The cells were exposed to reference compounds by adding the compounds from a 200 x concentrated stock solution in DMSO to exposure medium (5% DCC-FCS in DMEM/F12 without phenol red). Following exposure to the test materials, cells were further processed for the luciferase induction assay. Cells were rinsed using PBS followed by lysis through 30 µl low salt buffer (Tris, 25 mM, DTT 2.0 mM, CDTA 2.0 mM), and a subsequent freezing step at −80°C ensured complete cell lysis. Luciferase was measured using a flash mix protocol (BDS, Amsterdam, Netherlands). The flash mix or illuminate mix contained 20 mM tricine, 1.07 mM (MgCO₃)4 Mg(OH)2.5 H₂O, 2.67 mM MgSO₄ x 7 H₂O, EDTA 0.1 mM, DTT 1.5 mM, D-Luciferine 539 mM, ATP 5.49 mM. The measurements were performed in the different laboratories using a luminometer with two injectors, one to initiate the reaction (through the addition of the Luciferin present in the illuminate mix) and one for stopping the enzymatic reaction with NaOH. The reaction was thus stopped by adding 100 µL of 0.2 M NaOH. A threshold of induction factor of 1.5 was set for the NRF2 mediated gene expression, as described before (van der Linden et al., 2014).

Fe₃O₄-PEG-PLGA nanomaterials were selected as a representative and novel nanobiomaterial for the present study. These nanomaterials are envisioned both for therapeutic and diagnostic applications. The “benchmark” TiO₂ nanomaterials were included as an inert (non-cytotoxic) nanomaterial and Ag nanomaterials were included as a nanomaterial that most likely would elicit NRF2 mediated gene expression (based on the available literature, see above), though cytotoxicity at higher concentrations of the latter nanomaterials could not be excluded. Additional positive and negative chemical controls (dichlorvos and mannitol) were included for the assay based on the manufacturer’s recommendations.

The participating laboratories were trained (online) on the execution of the NRF2 reporter gene assay, quality control measures, and data analysis (for a schematic of the workflow, refer to Figure 1). The following laboratories/institutions participated in the pre-validation study: Karolinska Institutet, Wageningen University, University of Torino, Université Grenoble-Alpes, Edinburgh Napier University, University of Rome Tor Vergata, Université Paris Cité, and Tokyo University of Science. However, one of these laboratories only tested Fe₃O₄-PEG-PLGA and not the other “benchmark” nanomaterials and the results are therefore shown separately. Protocols were extensively discussed and agreed upon during online meetings and tutorials. Chemicals and cell culture reagents were procured from the same source, and the NRF2-U2OS cell line was distributed to all the laboratories. The plate layout for the reporter gene assay was decided. Each experiment thus included one reference plate and three experimental plates. The three upper rows (B-C-D) of the reference plate as well as each experimental plate contained a full concentration range of the reference compound curcumin dissolved in DMSO. The lower part (rows E-F-G) contained the positive and negative control (reference plate) or one of the three nanobiomaterials under investigation (Fe₃O₄-PEG-PLGA, TiO₂, Ag). The participating laboratories also harmonized the exposure conditions. Hence, Fe₃O₄-PEG-PLGA and Ag nanomaterials were diluted from a stock of 3000 μg/ml in dispersant provided with the particles at a concentration range of 0.21 µg/ml–3000 μg/ml followed by a second 30 x dilution step in exposure medium to an exposure range of 0.001–100 μg/ml. For TiO₂, freshly prepared suspensions were made using the NANOGENOTOX dispersion protocol (Farcal et al., 2015). The reporter cells were exposed for 24 h in a humidified atmosphere at 37°C under 5% CO₂.

Data were exported to Excel (Microsoft) for further processing. Cytotoxicity was expressed as % viability towards the unexposed cells. For the NRF2 reporter gene experiments, the results were presented as the Induction Factor (IF), which is the measured relative light unit (RLU) value divided by the mean RLU value of the solvent control. When the induction factor of curcumin was over 8, the NRF2-U2OS reporter gene assay was regarded to be effective. Samples presenting 1.5 fold or higher induction were considered as inducers of NRF2 mediated-gene expression (van der Linden et al., 2014). Graphs were prepared in Prism 9.0 (GraphPad Software, Inc.) by analysing data using non-linear curve fitting (agonist versus response). To evaluate the variability of results and reproducibility of the assay, both the intra-laboratory and inter laboratory standard deviations were calculated across all NRF2 reporter gene assay results and plotted in a heatmap. Statistical analysis was performed using GraphPad Prism version 8.3.0.

Interlaboratory standard deviation of the assay results of all participating laboratories was calculated in accordance with ISO standards 5725-1 and 5725-2 for accuracy (trueness and precision) of measurement methods and results.

The Fe₃O₄-PEG-PLGA nanomaterials obtained from Colorobbia and the corresponding dispersant were evaluated for sterility (endotoxin content). Both were found to contain endotoxin levels below the US FDA-mandated level for medical devices (data not shown). The Fe₃O₄-PEG-PLGA nanomaterials were visualized by TEM. TEM phase contrast images (Figure 2A) and HAADF-STEM images (Figure 2B) indicated regular morphology with a mean particle diameter of 12 ± 4 nm. The higher magnification HREM phase contrast images (Figure 2C) disclosed a cubic crystal structure consistent with the magnetite lattice, and polycrystalline pattern rings collected by SAED (Figure 2D) were indexed as crystalline magnetite, identified as the unique phase composition. The benchmark materials were fully characterised, see Comero et al. (2011) for Ag, and Rasmussen et al. (2014) for the TiO₂ nanomaterials.

SAXS analysis showed that there was little or no dissolution of the Ag nanomaterials following incubation at 37°C for 18 days in culture media. The average size of these nanomaterials did not change during incubation (15 ± 0.2 nm and 15 ± 0.2 nm at t = 0 and t = 18 days, respectively). The nanobiomaterials were also analysed with respect to hydrodynamic diameter and zeta potential in the relevant cell culture medium. Previously the dissolution of the Fe₃O₄-PEG-PLGA nanomaterials in cell culture media was shown to be less than 0.5% within 24 h (data not shown). Together the data indicated that all the test materials were stable following incubation in cell culture medium for 24 h at the exposure conditions for the NRF2 reporter gene assay (Supplementary Figures S1A,B).

For a correct interpretation of the results from the reporter gene assay, the potential of the test materials to reduce cell viability should be assessed. To this end, the Alamar blue assay was used. Fe₃O₄-PEG-PLGA nanomaterials were non-cytotoxic towards U2OS cells and only a slight decrease in cell viability (metabolic capacity) was evidenced at the highest tested concentration of 100 μg/ml (Figure 3A). Similarly, TiO₂ nanomaterials were non-cytotoxic at low concentrations but a markedly decreased viability at the highest concentration of 100 μg/ml (Figure 3B) was noted. In contrast, for Ag nanomaterials, a dose-dependent loss of cell viability was observed (Figure 3C). The potential cytotoxic effects of the reference compounds curcumin, dichlorvos (positive control) and mannitol (negative control) were also evaluated (Figure 4). Neither dichlorvos or mannitol affected cell viability of the U2OS cells, while curcumin at a concentration of 500 nM and higher reduced U2OS cell viability in a dose-dependent manner (Figure 4).

Next, the induction of the NRF2 pathway was assessed. NRF2-U2OS cells were exposed to increasing concentrations of the reference compound (curcumin), and to the positive and negative controls (Figure 5). Both the reference compound and the positive control (dichlorvos) induced NRF2 mediated gene expression while exposure to mannitol did not (Figure 5).

Eight laboratories participated in the “round robin” pre-validation study, of which seven used Fe₃O₄-PEG-PLGA, Ag, and TiO₂ nanomaterials (Figure 6), whereas one partner only used Fe₃O₄-PEG-PLGA nanomaterials (Supplementary Figure S2). The results consistently showed that TiO₂ did not induce NRF2-mediated gene expression. However, exposure to Ag nanomaterials induced NRF2-mediated gene expression in a dose-dependent manner in all experiments (Figures 6A–G). Some differences could be observed in the magnitude of responses (i.e., induction factor) of similar concentration in the different laboratories. It was consistently found that the NRF2 mediated gene expression declined at the highest concentrations which likely is due to the cytotoxicity following exposure to the Ag nanomaterials at the higher concentrations. Finally, following exposure to Fe₃O₄-PEG-PLGA minimal induction of NRF2-mediated gene expression was observed (Figures 6A–G). Hence, while three of the participating laboratories reported no induction, the results from 5 other laboratories showed a minor induction at 30 or 100 μg/ml, while some reported a lower induction factor for the 100 μg/ml samples compared to 30 μg/ml. Finally the inter and intra-laboratory standard deviations of the assay results were calculated.

The inter-laboratory standard deviation ranged from 0.044 to 1.221 with a mean of 0.28 (Figure 7). The mean intra-laboratory standard deviation was 0.16 (Supplementary Figure S3).

To verify the lack of interference of the test materials with the measurement of luciferase activity, the U2OS-NRF2 cells were fixed at the end of exposure by adding 50 µl of paraformaldehyde at 4% in PBS for 30 min at room temperature just before cell lysis to perform the luciferase induction assay as described above. No interference was observed (data not shown).