GO used in our investigations showed the typical flake-like shape, as shown by SEM and TEM images (Supplementary Figure S1A,C). This shape is in line with previous observations reported by our group on the same material (Guarnieri et al., 2018; Di Cristo et al., 2020). The elemental analysis by EDS revealed a high content of oxygen atoms in the carbon-based structure of GO (Supplementary Figure S1B), confirming the graphene oxidation, as also shown by the previously reported thermogravimetric analysis of GO (Guarnieri et al., 2018; Di Cristo et al., 2020). Colloidal stability of GO suspensions in cell stimulation-like conditions was assessed using complete HHPM as dispersant medium at three representative GO concentrations (4, 20, 80 μg/ml). After a 24-h incubation (corresponding to an acute stimulation of upcyte^® hepatocytes), 4 μg/ml GO suspension presented two main peaks with a relative mean size of 6.0 ± 0.2 nm (peak #1) and 359.5 ± 1.3 nm (peak #2; Figure 1A), respectively. The first is identified by and assigned to the protein components supplemented in complete HHPM (Maiorano et al., 2010; Magrì et al., 2018) (Supplementary Figure S1D). Peak #2 corresponds, therefore, to the main GO population, with a relative intensity equal to 80.0%. The peak identity is further confirmed by the presence of a peak at a similar position shown by the GO sample dispersed in Milli-Q^® water at the same concentration even though a smaller relative mean size was evident (267.4 ± 15.8 nm; Figure 1A). At the same incubation time, as GO concentration increases (20 and 80 μg/ml), it is noticed a shift of peak #2 toward larger sizes in complete HHPM (505.6 ± 51.7 nm and 660.9 ± 55.5 nm, respectively; Figure 1A), along with a progressive rise of the relative peak intensity (85.4 and 98.3%, respectively). In Milli-Q^® water, the corresponding peaks were slightly stable, alongside the relative peak intensities. Peak position and relative intensity of all peaks identified in each tested condition are reported in Supplementary Table S1. Furthermore, consecutive DLS measurements (0–24 h) were acquired to follow the possible changes of the relative mean size of peak-#2 GO, likely due to suspension instability. Results indicate that, especially for the highest concentration tested, 80 μg/ml, the mean size rapidly increased in the first 6 h in complete HHPM, leaving a slight decrease at 24 h (Figure 1B). This effect is not observed when GO was incubated in Milli-Q^® water, where the mean size remained almost stable within the 24 h. Overall, these data, along with the occasional emergence of a peak #3 at the highest concentrations (20–80 μg/ml) characterized by relative mean sizes of 4.0–5.0 µm (see Supplementary Table S1), suggest that GO suspensions especially in complete HHPM are poly-dispersed and show a tendency to agglomerate/aggregate within the temporal range studied (0–24 h). The agglomeration/aggregation tendency of GO has been further evidenced by the presence of sediments at the highest concentrations (20–80 μg/ml), as reported in Figure 1C. The zeta potential analysis revealed a negative surface charge of GO in Milli-Q^® water, which goes from -35 mV to -41 mV as a function of the concentration applied (Figure 1D). After 24-h incubation in complete HHPM, the surface charge of GO increased to about −27.5 mV, as a consequence of the possible formation of a bio-macromolecular corona around GO (Maiorano et al., 2010). The particle-corona complexes (see technical details in Materials and Methods) show mean hydrodynamic diameters almost comparable to bare GO (Supplementary Figure S1E).

To understand the possible detrimental impact of GO on upcyte^® hepatocytes, we first evaluated its impact on cell viability and cell membrane integrity. Confluent cell cultures were treated for 24 h with increasing GO concentrations, as reported in Figure 2A. The half-maximal inhibitory concentration (IC₅₀) was calculated testing a wide concentration range of GO (4–320 μg/ml). As evidenced by the reduction of the metabolic activity, GO determined a dose-dependent decrease of cell viability, with an IC₅₀ equal to 102.2 μg/ml (Figure 2A). Note that the highest GO concentration tested, corresponding to 320 μg/ml, caused a cell viability reduction comparable to the lethal effect exerted by the positive control, which was 0.03% Triton X-100 (Supplementary Figure S2A). As revealed by the cytotoxicity assay, treated cells also showed a significant release in the culture medium of the cytosolic lactate dehydrogenase (LDH), indicating cell membrane damage (Figure 2B). In particular, at 80 μg/ml GO, a concentration value lower than the IC₅₀, it is possible observing a significant increase of cytotoxicity to about 11.5% (p = 0.0015) with respect to the control, whereas at lethal doses (320 μg/ml), the damage weakly increased only up to 17% (p < 0.0001). In addition, with the increase of GO concentration, the decrease of cell viability alongside the rise in released LDH showed a significant negative linear correlation (p < 0.0001 and Pearson’s r = 0.9521; Supplementary Figure S2B). As a function of the GO concentration, the progressive deposition of GO on the top surface of treated cells was observed. In fact, the number of GO deposits and the percentage area covered by GO dose-dependently increased (Supplementary Figures S2C–E). Considering three representative GO concentrations (4, 20, and 80 μg/ml), the mean lateral size relative to GO deposits by SEM images analysis was about 2.0 μm, with no differences among the GO concentrations (Supplementary Figures S2F,G). Nevertheless, the presence of large deposits does not likely prevent the uptake of the nano-sized GO present in suspension. The GO deposition is indicative of a partial colloidal instability of GO suspensions (as also evidenced by DLS analysis), so mechanical cell damage due to particle deposition cannot be excluded.

To assess apoptosis, oxidative stress, or inflammatory response potentially triggered by GO, upcyte^® hepatocytes were treated with increasing sub-IC₅₀ doses (2–80 μg/ml). Regarding apoptosis induction, treated cells were screened to follow the extent of phosphatidylserine (PtdSer) exposure over time as early indicator of apoptosis (Figure 3A), finding no significant variations at any of the tested concentrations for the early GO treatment times (3 and 6 h). After 24 h of treatment, however, PtdSer exposure significantly increased about 1.3-fold with respect to the control upon the highest concentrations (20 μg/ml GO, p = 0.0352; 40 μg/ml GO, p = 0.0043; and 80 μg/ml GO, p = 0.0429). It is important to underline that GO showed just mild induction of PtdSer exposure and a different kinetic compared to the positive control (0.5 μM Staurosporine). Indeed, Staurosporine induced an earlier significant increase of PtdSer exposure, reaching a maximum peak at 6 h (p < 0.0001). PtdSer level reached the basal value again after 24 h of Staurosporine treatment in correspondence to the significant increase of necrotic cells (p < 0.0001, Supplementary Figure S3). For evidencing a further sign of cellular apoptosis, the intracellular amount of cleaved-PARP in GO-treated cells was measured. We found no alteration of cleaved-PARP levels with respect to the control at any of the tested concentrations, further confirming the mild apoptotic effect induced by GO (Figure 3B). Since oxidative stress is involved in many mechanisms of cytotoxicity (such as apoptosis, DNA damage, lipid peroxidation), we investigated the possible role of GO as an oxidative stress inducer (Di Cristo et al., 2018). We found that GO did not modulate the gene expression of two main cellular antioxidants, HO-1 and SOD1, at any of the tested concentrations (Figure 3C). By contrast, the positive control (400 μM H₂O₂) induced a significant upregulation of HO-1 and SOD1 gene expression (p = 0.0094 and p = 0.0269, respectively). Upon GO treatment, western blot analysis of HO-1 and SOD1 confirmed the absence of significant variations of the corresponding proteins with respect to the control (Figure 3D). As far as an inflammatory response is concerned, GO did not induce in our cell model any modulation of the gene expression of TNFα, IL-1β, IL-6, and IL-8, considered mediators of inflammation (Figure 3E). On the other hand, upcyte^® hepatocytes showed a significant upregulation of all those genes (p = 0.0008 for TNFα; p = 0.0482 for IL-1β; p < 0.0001 for IL-6; p = 0.0001 for IL-8), when exposed to exogenous pro-inflammatory stimuli (see technical details in Materials and Methods). These data indicate that the cell model employed has a poor ability for inflammatory response; however, based on only this evidence, we cannot conclude that GO does not induce inflammation in vivo.

For evaluating the GO impact on the cytochrome P450 system of upcyte^® hepatocytes, cells were treated for 24 h with increasing sub-IC₅₀ GO doses (2–80 μg/ml). We measured the gene expression and the corresponding metabolic activity of CYP3A4 and CYP2C9, which are the most representative enzymes of that system. In parallel, the responsiveness of cytochromes P450 was assessed, making use of well-known drugs. For this aim, cells were daily incubated up to 72 h (Bjornsson et al., 2003; Fahmi et al., 2008) with 50 μM Rifampicin or 100 μM Ciprofloxacin, which act as inducer or inhibitor of both CYP3A4 and CYP2C9, respectively. Results show that, upon treatment with Rifampicin, CYP3A4 gene expression was about 41-fold upregulated with respect to untreated cells (p = 0.0052; Figure 4A, upper panel), whereas, when cells were treated with Ciprofloxacin, 0.6-fold downregulated in comparison to the control level (p = 0.0138; Figure 4B, upper panel). Similarly, CYP2C9 gene expression was significantly regulated upon the same treatments, with a Rifampicin-mediated upregulation of about 4.0-fold (p = 0.0047; Figure 4D, upper panel) and a Ciprofloxacin-mediated downregulation to about 0.7-fold with respect to untreated cells (p = 0.0444; Figure 4E, upper panel). Therefore, we evaluated the metabolic activity of CYP3A4 and CYP2C9 using as substrates two fluorescent compounds, 7-benzyloxy-4-trifluoromethylcoumarin (BFC) and 7-methoxy-4-trifluoromethylcoumarin (MFC), respectively. Results indicate that, in line with the gene expression data, the activity of CYP3A4 was statistically increased by about 290% in Rifampicin-treated cells with respect to the control (p = 0.0247; Figure 4A, lower panel), whereas it appeared to be reduced to about 64% in the case of Ciprofloxacin-treated cells (p = 0.0006; Figure 4B, lower panel). Comparably, CYP2C9 activity was modulated by the same treatments consistently with gene expression regulation (Figures 4D,E, lower panels). Afterward, we evaluated the effect of sub-IC₅₀ concentrations of GO on the cytochrome P450 system. We observed a dose-dependent downregulation of CYP3A4 gene expression (starting from 4 μg/ml, p = 0.0015), which was equal to about 0.37-fold with respect to the control level at the highest concentration (80 μg/ml GO, p < 0.0001; Figure 4C, upper panel). Consistent with the gene expression, CYP3A4 metabolic activity showed a significant and progressive dose-dependent reduction, reaching a level of about 18.6% with respect to the control, at a concentration value of 80 μg/ml GO (p = 0.0002; Figure 4C, lower panel). In parallel, CYP2C9 gene expression significantly decreased starting from 20 μg/ml GO (p = 0.0012), down to about 1.9% with respect to the control level with 80 μg/ml GO (p < 0.0001; Figure 4F, upper panel). CYP2C9 metabolic activity was also significantly reduced, reaching about 10.9% of the basal level with the highest GO concentration (80 μg/ml, p = 0.0006; Figure 4F, lower panel). For both CYP3A4 and CYP2C9, GO-mediated gene expression downregulation had a significant positive linear correlation with the metabolic activity impairment (p = 0.0119 and Pearson’s r = 0.9095 for CYP3A4 and p = 0.0015 and Pearson’s r = 0.9682 for CYP2C9; Supplementary Figure S4A, B). Our results indicate a strong inhibition activity of GO on CYP3A4 and CYP2C9 gene expression and relative metabolic activities. Such an inhibitory response appears to be dose-dependent, and the negative effect of 4 μg/ml GO could be approximately compared to that induced by 100 μM Ciprofloxacin (∼33 μg/ml). The inhibition of the cytochrome P450 system by sub-lethal doses of GO was also confirmed on CYP2B6 and CYP1A2 (Supplementary Figures S4E, H). For CYP2B6, gene expression decreased starting from 20 μg/ml GO (p = 0.0069), while GO affected CYP1A2 gene expression at lower doses (2 μg/ml; p = 0.0052). Upon 80 μg/ml GO treatment, CYP2B6 and CYP1A2 reached a gene expression equal to about 4.3 and 11.3% of the control level (p < 0.0001 for both), respectively. In contrast to the effects exerted on CYP3A4 and CYP2C9 gene expression, Rifampicin and Ciprofloxacin had no effects on CYP2B6 and CYP1A2, revealing their specific mechanism of action only on CYP3A4 and CYP2C9 in upcyte^® hepatocytes (Supplementary Figures S4C, D and Supplementary Figures S4F, G). Conversely, GO seems to act on a wide range of targets within the cytochrome P450 system.

After having observed the effects of GO on phase-I drug metabolism enzymes, we investigated the impact of sub-lethal GO exposures on phase-II and phase-III drug metabolism/transport enzymes in upcyte^® hepatocytes. In particular, the gene expression of Glutathione S-Transferase (GST) and ATP Binding Cassette Subfamily G Member 2 (ABCG2) was evaluated. Results show that GST and ABCG2 gene expression was not modulated by GO at any of the tested concentrations (Figures 5A,B). For confirming that GO effects observed on phase-I CYPs were specific and not related to a global de-differentiation of upcyte^® hepatocytes, we also evaluated in control and GO-treated cells the expression of Glucose Transporter 2 (GLUT2), which is a solute carrier transporter catalysing the passive transport of glucose across the cell membrane in functional hepatocytes (Kim et al., 1998; Thorens 2015). FACS analysis revealed a stable expression of GLUT2 over time in the control cells, confirming the phenotypic stability of the cell model employed, and only a moderate or absent modulation by GO, depending on the dose (Supplementary Figure S5A). Regarding the xenobiotic-sensing receptors, PXR and CAR, which are transcription regulators of a large part of phase-I, -II, and -III executioners (Evans and Mangelsdorf 2014), we observed that PXR was significantly downregulated only by 80 μg/ml GO (p = 0.0224), while CAR gene expression was not modulated at the experimental doses applied in this study (Supplementary Figures S5B, C).

We also analysed the impact of sub-lethal GO doses on Albumin gene expression and translation. In the case of gene expression, a dose-dependent downregulation was observed with a resulting reduction to about 5.2% with respect to the control transcript level by 80 μg/ml GO (p < 0.0001; Figure 6A). A similar trend in the Albumin transcript translation was found, with a reduction of the intracellular protein amount equal to about 12.5% of the control by the highest GO dose (p < 0.0001; Figure 6B). Considering two other hepatic proteins, GO had a dose-dependent negative impact on the gene expression of Transferrin and Transthyretin with a downregulation equal to about 16.5 and 16.4% of the control, respectively, by 80 μg/ml GO (p = 0.0051 and p = 0.0023, respectively; Supplementary Figure S6A, B).

Having in mind that inflammation and its mediators in vivo determine the inhibition of cytochromes P450 via a multitude of complex pre- and post-transcriptional mechanisms (de Jong et al., 2020; Stanke-Labesque et al., 2020; Lenoir et al., 2021), and considering that at the experimental conditions applied (i.e., sub-lethal doses) GO did not stimulate an inflammatory response in our cell model (see above), we assessed if GO may display a pro-inflammatory cytokine-like behaviour, provoking the impairment of the cytochrome P450 system and, at the same time, the secretion of a “positive” acute-phase protein. In this regard, it has been widely reported that IL-6 acts as the chief regulator of APPs in human hepatocytes during inflammation (Castell et al., 1989), as well as a suppressor of CYP3A4 (Jover et al., 2002). Hence, IL-6 was applied herein as a positive control.

To address the point and verify this hypothesis, upcyte^® hepatocytes were treated with a very low, sub-lethal dose of GO (4 μg/ml) in order to minimize as much as possible other toxic effects (such as cell membrane damage and early apoptosis), and the gene expression of CYP3A4 and CRP (Gabay and Kushner 1999) was monitored over time.

We found an early significant downregulation of CYP3A4 after 3 h of treatment (p = 0.0165; Figure 7A). The transcript level of CYP3A4 remained comparably reduced at 8 h (p = 0.0256) and decreased further at 24 h (p = 0.0086). In parallel, the gene expression of CRP was upregulated starting from 3 h (p = 0.0012) and reached a maximal peak at 8 h (p = 0.0018; Figure 7B). At the end of GO treatment (24 h), the CRP transcript level returned to the basal level. At the peak of CRP upregulation, we found a significantly increased secretion of CRP (p = 0.0303; Figure 7C). Similarly, after 8 h of treatment, 20 ng/ml rhIL-6 serving as positive control provoked the downregulation of CYP3A4 (p = 0.0006) and the upregulation of CRP (p = 0.0057; Supplementary Figure S7A). Accordingly, the secretion of CRP was increased by rhIL-6 at 8 h (p = 0.0101; Supplementary Figure S7B).

Second-generation upcyte^® hepatocytes (donor 653-03) and Hepatocyte High Performance Medium (complete HHPM: basal medium supplemented with supplement A and L-glutamine) were purchased from upcyte^® technologies GmbH (Hamburg, DE). Cell culture flasks and 96- and 24-well plates were from Corning Incorporated (Corning, NY, United States). Collagen Type I solution from rat tail, Rifampicin, Ciprofloxacin, high glucose phenol red-free Dulbecco’s Modified Eagle’s Medium (DMEM), Dulbecco’s Phosphate Buffered Saline (DPBS), resazurin sodium salt, 7-benzyloxy-4-trifluoromethylcoumarin, 7-methoxy-4-trifluoromethylcoumarin, β-glucuronidase/arylsulfatase, Lipopolysaccharide (LPS), recombinant human Interferon gamma (rhIFNγ), Triton X-100 and Staurosporine (Stau) were from Sigma Aldrich (Merck KGaA, Darmstadt, DE). Tecan Spark^® multimode microplate reader was from Tecan (Männedorf, CH). Recombinant human Interleukin 6 (rhIL-6) and Human CRP Quantikine^® ELISA kit were from R&D Systems (Bio-Techne, Minneapolis, MN, United States). CytoTox96^® Non-Radioactive Cytotoxicity Assay and RealTime-Glo™ Annexin V Apoptosis and Necrosis Assay were purchased from Promega Corporation (Madison, WI, United States). Prism software was from GraphPad Software (San Diego, CA, United States). Zetasizer Nano-ZS was from Malvern Instruments (Worcestershire, United Kingdom). Pierce™ BCA Protein Assay Kit, TBS Buffer and primary antibody anti-cleaved-PARP were obtained from Thermo Fisher Scientific (Waltham, MA, United States). TRIzol™ Reagent, SuperScript™ VILO™ cDNA Synthesis Kit, qPCR primers and NuPAGE™ 4–12% Bis-Tris gel were obtained from Invitrogen (Thermo Fisher Scientific, Waltham, MA, United States). NanoDrop One^C was from Thermo Scientific™ (Thermo Fisher Scientific, Waltham, MA, United States). iTaq™ Universal SYBR^® Green Supermix, Clarity™ Western ECL Substrate and ChemiDoc™ MP Imaging System were from Bio-Rad (Hercules, CA, United States). Applied Biosystems ViiA 7 Real-Time PCR System was from Life Technologies (Thermo Fisher Scientific, Waltham, MA, United States). Amersham™ Protran™ 0.2 μm NC nitrocellulose blotting membrane was from GE Healthcare (Buckinghamshire, United Kingdom). Primary antibodies anti-HO-1, anti-SOD1, anti-Albumin and anti-β-Actin, and secondary anti-rabbit and anti-mouse HRP-linked antibodies were purchased from Cell Signaling Technology (Danvers, MA, United States). The illustrations of the cell components are from smart.servier.com (LES LABORATOIRES SERVIER, SAS, Suresnes, FR).

Graphene oxide (GO) was kindly provided by Dr E. Vazquez (Universidad de Castilla-La Mancha, Spain). Characterization of the nanomaterial dispersed in Milli-Q^® water was previously provided by our group (Guarnieri et al., 2018; Di Cristo et al., 2020). GO colloidal stability was analysed as a function of concentration (4, 20, and 80 μg/ml) and incubation time (0, 0.5, 1.5, 2, 3, 6, and 24 h at 37°C) when dispersed in Milli-Q^® water or complete HHPM. GO size distribution profiles were determined via dynamic light scattering (DLS) analysis using a Zetasizer Nano-ZS at 37°C, even though such technique is not entirely accurate for the study of non-spherical particles. Each suspension was prepared directly into a cap-closed disposable cuvette for DLS measurements and incubated at 37°C with no further transfers in order to replicate the same conditions of biological experiments. Three consecutive measurements, with 7 runs per measurement, were performed for each GO suspension. The attenuator and the optimal measurement position were set automatically. GO-free complete HHPM was used as background control. For aqueous suspensions, GO surface charge was measured via Zeta Potential (ZP) analysis using a Zetasizer Nano-ZS at 25°C. Five consecutive measurements, with an automatically set number of runs per measurement, were taken for each GO suspension. To study particle-corona complexes, at the end of 24-h incubation in complete HHPM, GO suspensions were centrifuged at 15 000 g for 15 min at 4°C, and corresponding pellets were washed three times, adding a volume of Milli-Q^® water equal to the initial volume (Maiorano et al., 2010; Magrì et al., 2018). Particle-corona complexes were characterized via DLS and Zeta Potential, as described above. GO was endotoxin-free, as previously reported by Di Cristo et al. (Di Cristo et al., 2020), in accordance with US Food and Drug Administration guidelines (https://www.fda.gov/regulatory-information/search-fda-guidance-documents/guidance-industry-pyrogen-and-endotoxins-testing-questions-and-answers).

Second-generation human upcyte^® hepatocytes from a female Caucasian donor (653-03) were cultured following the manufacturer’s indications. Upcyte^® hepatocytes were seeded at the concentration of 10 000 cells/cm² into cell culture flasks coated with 0.1 ml/cm² of 50 μg/ml collagen-type I in 20 mM acetic acid, and they were cultured in complete HHPM, in incubation in a humidified atmosphere at 37°C, with 5% CO₂. No antibiotics were added to the culture medium to do not alter cytochromes P450 activity. The culture medium was changed 3 times per week. Cells were expanded for 1 or 2 passages before being treated, as described below.

The cell viability upon GO treatment was evaluated by colorimetric resazurin reduction test. Upcyte^® hepatocytes were cultured into collagen-coated, flat-bottom 96-well plates (cell growth area equal to 0.3 cm², approximately) and, at the confluence, they were treated with different GO concentrations (final volume equal to 75 μL per well) for 24 h. A 24-h treatment with the complete medium supplemented with 0.03% Triton X-100 was used as a positive control of cell viability reduction. At the end of GO treatment, stimulation media were replaced with 100 μL per well of serum-free phenol red-free high glucose DMEM, supplemented with 44 μM resazurin sodium salt, after having extensively washed with DPBS (O'Brien et al., 2000). After 1 h of incubation at 37°C in a humidified atmosphere with 5% CO₂ in the dark, the resazurin solution was transferred into a clean 96-well plate, and fluorescence was measured at 535 nm by Tecan Spark^® reader. For each culture condition, three independent experiments were performed, each one with a technical triplicate. In each experiment, a couple of cell-free, collagen-coated wells was incubated with the stimulation medium per culture condition (containing GO at all the concentration tested) and used as blank value to be subtracted during the data analysis and to exclude any GO induced optical interferences. The reported results are expressed as percentage values (means ± SD) compared to the control condition (set as 100%). The half maximal inhibitory concentration (IC₅₀) of GO was calculated using the log (inhibitor) vs. normalized response curves model on Prism software.

For evaluating the cell membrane damage upon GO treatment, the colorimetric CytoTox96^® Non-Radioactive Cytotoxicity Assay was used. Upcyte^® hepatocytes were cultured into collagen-coated, flat-bottom 96-well plates (cell growth area equal to 0.3 cm², approximately). At the confluence, cells were treated with different GO concentrations (final volume equal to 75 μL per well) for 24 h. As a positive control, confluent cells were treated with the complete medium supplemented with 0.03% Triton X-100 for 24 h. At the end of stimulation, conditioned media were collected, centrifuged at 15 000 g for 15 min at 4°C, and analysed following the manufacturer’s instructions. For each culture condition, three independent experiments were performed, each one with a technical triplicate. In each experiment, a couple of cell-free, collagen-coated wells was incubated with the stimulation medium per culture condition (containing GO at all the concentration tested) and used as blank value to be subtracted during the data analysis and to exclude any GO induced optical interferences. The results are expressed as a percentage net increase of the absorbance (means ± SD) compared to the positive control (set as 100%) after having subtracted the basal value of the control condition (set as 0%).

For assessing the induction of apoptosis and necrosis, the RealTime-Glo™ Annexin V Apoptosis and Necrosis Assay was used, according to the manufacturer’s instructions. Upcyte^® hepatocytes were cultured into collagen-coated, flat-bottom 96-well plates (cell growth area equal to 0.3 cm², approximately). At the confluence, cells were treated for 24 h with the complete medium supplemented with GO at different concentrations, in the presence of the detection reagent (final volume equal to 75 μL per well). As a positive control, confluent cells were treated for 24 h with 0.5 μM Staurosporine (Stau), diluted in the same incubation medium. After 3, 6, and 24 h of incubation, luminescence and fluorescence (at 530 nm) were measured by Tecan Spark^® reader for the induction of apoptosis and necrosis, respectively. For each culture condition, three independent experiments were performed, each one with a technical triplicate. In each experiment, a couple of cell-free, collagen-coated wells was incubated with the stimulation medium per culture condition (containing GO at all the concentration tested) and used as blank value to be subtracted during the data analysis and to exclude any GO induced optical interferences. Apoptosis data are expressed as an n-fold increase over the control conditions (means ± SD) for each incubation time. Necrosis results are expressed as net increase of fluorescence for each condition (means ± SD), after having subtracted the control cells’ basal value obtained after 3 h of incubation (set as 0 a.u.).

For the gene expression analysis of GO-treated cells, upcyte^® hepatocytes were cultured into collagen-coated, flat-bottom 24-well plates (cell growth area equal to 2.0 cm², approximately) and, at the confluence, they were treated with different GO concentrations (final volume equal to 500 μL per well) for 24 h. Cells treated with 400 μM H₂O₂ overnight were used as positive control for the oxidative stress. Regarding the inflammatory response, a 24-h treatment with 1 μg/ml LPS and 50 ng/ml rhIFNγ was used for the upregulation of TNFα, while 100 μg/ml LPS overnight was used for the upregulation of IL-1β, IL-6, and IL-8. As positive and negative controls of CYP3A4/2C9 induction, confluent cells were treated for 72 h with 50 μM Rifampicin or 100 μM Ciprofloxacin, respectively, changing the stimulation medium every day after an extensive wash with DPBS (Bjornsson et al., 2003; Fahmi et al., 2008). For studying the gene expression of CYP3A4 and CRP over time, upcyte^® hepatocytes at P3 were cultured into collagen-coated 24-well plates (65 000 cells/cm²) for 1 day and, then, treated with 4 μg/ml GO (500 μL/well) for 3, 8 and 24 h. As positive control, cells were treated with 20 ng/ml rhIL-6 (specific activity equal to about 1.1 × 10⁵ IU/μg) for 8 h. At the end of any stimulation, cells were extensively washed with DPBS and incubated with TRIzol™ Reagent (500 µL per well) at −80°C for at least one night. For each sample, total RNA was isolated according to Chomczynski’s method (Chomczynski 1993). RNA yield was determined using NanoDrop One^C, and RNA purity was examined considering A260/A280 and A260/A230 ratios. Total RNA (2 μg/sample in 20 µL of total volume reaction) was reverse-transcribed to first-strand cDNA by SuperScript™ VILO™ cDNA Synthesis Kit, following manufacturer’s instructions. Transcript levels of target genes were measured by quantitative Real-Time PCR (qPCR) using iTaq™ Universal SYBR^® Green Supermix on Applied Biosystems ViiA 7 Real-Time PCR System. Primer sequences, annealing temperatures, corresponding amplicon sizes, and qPCR efficiencies are reported in Supplementary Table S2. The qPCR efficiency of each primer pair was obtained using the dilution model and calculated according to the equation: E = 10^(−1/slope), where “slope” is the linear regression slope of a standard curve (Pfaffl and Dorak 2007). For each primer pair, melting curve analysis was carried out to verify the production of a single amplicon and, consequently, primer specificity. The gene expression of GAPDH was used as endogenous control (reference gene). For assessing the reference gene stability across the samples, the n-fold changes in GAPDH transcript level were calculated as E^−ΔCq, where Cq refers to the quantification cycles, for all samples versus the control condition and ordinary one-way ANOVA was run on those transcript fold changes, obtaining no significant differences (p = 0.6161; n = 3) (Sundaram et al., 2019). The target transcript levels were calculated using Pfaffl’s model for relative quantification (Pfaffl 2001). For each sample, the reaction was performed in technical triplicate. The results are expressed as the average of at least three independent experiments, with relative SD values.

For evaluating GO effects on apoptosis and oxidative stress, the protein content from GO-treated upcyte^® hepatocytes was analysed by western blot. In particular, proteins were isolated starting from the same homogenates used for RNA isolation as described by Chomczynski (Chomczynski 1993). The protein content was quantified by Pierce™ BCA Protein Assay Kit, following the manufacturer’s instructions. Electrophoresis was performed loading 20 μg of total protein per sample on NuPAGE™ 4–12% Bis-Tris gel under reducing conditions. Proteins were transferred to an Amersham™ Protran™ 0.2 μm NC nitrocellulose blotting membrane. The blot was blocked with 5% non-fat milk in TBS Buffer, supplemented with 0.1% Tween 20 (T-TBS), for at least 1 h at room temperature and then probed with primary antibodies raised against cleaved-PARP (1:1,000), HO-1 (1:1,000), SOD1 (1:500), Albumin (1:1,000) or β-Actin (1:1,000), overnight in a cold room. β-Actin was considered as the internal control. The blot was extensively washed with T-TBS and incubated with secondary anti-rabbit (1:15,000) or anti-mouse (1:10,000) HRP-linked IgG antibodies for 1 h at room temperature. The blot was then washed with T-TBS and incubated with Clarity™ Western ECL Substrate, following the manufacturer’s instructions. The blot image was acquired by ChemiDoc™ MP Imaging System. Each considered marker was analysed by western blot in three independent experiments. The densitometric analysis was performed by quantifying band densities by Fiji software (http://fiji.sc (Schindelin et al., 2012)). The reported results are expressed as n-fold changes (means ± SD) over the control condition (set as 1.0).

For evaluating the enzymatic activity of CYP3A4 and CYP2C9 in GO-treated cells, the conversion of 7-benzyloxy-4-trifluoromethylcoumarin (BFC) or 7-methoxy-4-trifluoromethylcoumarin (MFC) to 7-hydroxy-4-trifluoromethylcoumarin (HFC) was monitored as described by Donato et al. (Donato et al., 2004), with some modifications. Upcyte^® hepatocytes were seeded into collagen-coated, flat-bottom 96-well plates (cell growth area equal to 0.3 cm², approximately) and, at the confluence, they were treated with different GO concentrations for 24 h (final volume equal to 75 μL per well). As positive and negative controls, confluent cells were treated for 72 h with complete medium supplemented with 50 μM Rifampicin or 100 μM Ciprofloxacin, respectively, changing the stimulation medium every day after an extensive wash with DPBS (Bjornsson et al., 2003; Fahmi et al., 2008). At the end of treatments, cells were washed with DPBS, and they were incubated with 100 μL per well of 100 μM BFC or 150 μM MFC in the incubation medium (1 mM Na₂HPO₄, 137 mM NaCl, 5 mM KCl, 0.5 mM MgCl₂, 2 mM CaCl₂, 10 mM glucose, 10 mM Hepes; pH 7.4 buffered solution) for 2 h at 37°C, in a humidified atmosphere with 5% CO₂. After the incubation with BFC or MFC, the supernatant was collected, diluted 1:1 (v/v) with β-glucuronidase/arylsulfatase (150 Fishman units/mL and 1,200 Roy units/mL, respectively) and incubated for 2 h at 37°C. At the end of this step, the reaction mixture was diluted 1:1 (v/v) with the quenching solution (0.25 M Tris in 60% acetonitrile). Finally, the fluorescent HFC metabolite formation was quantified at the wavelength of 410 nm (excitation) and 510 nm (emission) by Tecan Spark^® reader. Each culture condition was assayed in technical triplicate in three independent experiments. In each experiment, a couple of cell-free, collagen-coated wells was incubated with the stimulation medium per culture condition (containing GO at all the concentration tested) and used as blank value to be subtracted during the data analysis and to exclude any GO induced optical interferences. The reported results are expressed as percentage values (means ± SD) over the control condition (set as 100%).

For determining the secretion of CRP into the cell supernatants, the Human CRP Quantikine^® ELISA kit was used. At the end of the 8-h incubation, the stimulation media derived from the corresponding qPCR experiments were saved, centrifuged at 15,000 g for 15 min at 4°C and analysed according to manufacturer’s instructions. The cell supernatants derived from cells treated with 20 ng/ml rhIL-6 for 8 h served as positive control of CRP secretion. For each culture condition, four independent experiments were performed. Results are expressed as n-fold changes (means ± SD) over the control condition (set as 1.0).

Statistical analysis was run on Prism software. Ordinary one-way ANOVA was performed for cell viability assay, cytotoxicity assay, apoptosis assay, qPCR analysis, western blot analysis, and cytochrome P450 activity assay. If ANOVA detected statistically significant differences within the data set, Dunnett’s multiple comparisons test (cytotoxicity assay, apoptosis assay, qPCR analysis, and cytochrome P450 activity assay) or Tukey’s multiple comparisons test (cell viability assay) were used to calculate significant differences. Two-way ANOVA was used for necrosis assay, with Sidak’s multiple comparisons test. Unpaired t-Test was used to calculate significant differences between the control condition and the drug-treated cells within data sets obtained by qPCR analysis or cytochrome P450 activity assay, and between the control condition and the GO- or rhIL-6-treated cells in CRP quantitative determinations. GO IC₅₀ was calculated using the log (inhibitor) vs. normalized response curves model. Correlation between cell viability and cytotoxicity data and between relative gene expression and metabolic activity data of CYP3A4 and CYP2C9 was calculated using a linear regression. All tests were run setting a confidence interval of 95%. When p < 0.05, differences were considered statistically significant. All data are presented as means ± standard deviations (SD) of at least three independent experiments.