ENN A1 and B1 In Vitro Toxicological Effects on 2D and 3D Organ-on-Chip HepaRG Liver Cells

TL;DR

This study examines the toxic effects of ENN A1 and B1 on liver cells in 2D and 3D models, finding that 3D spheroids offer better insights into their toxicity and distinct cell death pathways.

Contribution

The study introduces a 3D HepaRG spheroid model to better understand the differential toxicity and genotoxicity of ENN A1 and B1 compared to 2D models.

Findings

3D spheroids showed greater sensitivity to ENN A1 and B1 toxicity compared to 2D cultures.

ENN A1 induced significant DNA damage at low concentrations without cytotoxicity, suggesting genotoxic potential.

ENN A1 and B1 triggered distinct cell death pathways, with A1 primarily causing apoptosis and B1 inducing necrosis and limited autophagy.

Abstract

Enniatins (ENNs) are emerging Fusarium mycotoxins detected in food and feed. Despite their widespread occurrence, their toxicity remains poorly understood; thus, advanced in vitro systems that can mimic human physiology are of interest. We evaluated the cytotoxic and genotoxic effects of ENN A1 and ENN B1 exposure on differentiated (DIFF) and undifferentiated (UND) HepaRG liver cells cultured as 2D monolayers and 3D spheroids. Cytotoxicity, assessed by ATP-based luminescence, revealed a time-dependent decrease in inhibitory concentration 50 (IC50) values between 24 h and 48 h across all models. In DIFF HepaRG cells, ENN A1 IC50 values in 3D spheroids decreased from 14.4–18.2 µM at 24 h to 2.2–3.0 µM at 48 h, reaching values comparable to those measured in 2D DIFF cells at 48 h (2.2–2.6 µM), while no IC50 could be determined in 2D at 24 h. For ENN B1, a pronounced time-dependent toxicity was observed, with IC50 values in 3D DIFF spheroids decreasing from 4.1–6.6 µM at 24 h to 1.3–1.6 µM at 48 h, remaining lower than those measured in 2D DIFF cells at 48 h (2.4–3.0 µM). ENN A1 primarily induced apoptotic responses, whereas both ENN A1 and B1 were associated with necrotic responses, and ENN B1 induced a transient and limited autophagic signal, suggesting a minor role for autophagy. To further characterize cellular responses to ENN exposure, spheroids cultured in microfluidic chips were sectioned, and proliferation (Ki67), DNA damage (γH2AX), and apoptosis (cleaved caspase-3) was assessed. Immunostaining revealed no proliferative response, whereas significant DNA damage was detected, particularly in DIFF spheroids. At low, sub-cytotoxic concentrations (~5 µM, 24 h), ENN A1 induced significant DNA damage, as shown by increased γH2AX levels, while cytotoxic effects were only observed at higher concentrations (IC50 ~ 18 µM, 24 h), supporting a potential genotoxic effect independent of cytotoxicity. Despite the structural similarities between ENN A1 and ENN B1, our results highlighted distinct cell death pathways between the two analogues. Both ENNs were detected throughout spheroids without evidence of peripheral restriction, although a homogeneous functional test could not be conclusively demonstrated. Overall, the 3D HepaRG spheroid model proved to be a more physiologically relevant system, offering differential sensitivity, as well as enhanced mechanistic insight, compared to 2D cultures.

Figures

Full text

1. Introduction

Mycotoxins are secondary metabolites produced by many fungal species that frequently contaminate food and feed products worldwide. Downstream factors such as inadequate harvesting practices, improper storage, and non-optimal conditions during transportation, processing, and retail can promote fungal growth and increase the risk of mycotoxin contamination [1]. Given the ubiquitous presence of fungi, mycotoxins have become a growing concern for health organizations, as their occurrence in food products is difficult to avoid and currently poses a risk to consumers [2]. Recent advancements in analytical detection methods have facilitated the identification of previously unrecognized mycotoxins, leading to the introduction of the “emerging mycotoxins” concept in 2008 [3]. This term designates fungal secondary metabolites that are widespread, yet remain excluded from routine surveillance programs, regulatory policies, and established maximum residue thresholds [4,5]. Among them, enniatins (ENNs), a group of mycotoxins mainly produced by Fusarium species, have raised increasing concern due to their widespread occurrence and potential effects on human health [6]. ENNs are cyclohexadepsipeptides with ionophoric properties, which thus can disrupt cellular homeostasis by forming cation-selective pores in cell membranes. In foodstuffs, four major ENNs (A, A1, B and B1) have been commonly detected, especially in cereals and cereal-based food products [6].

Recent DSM food product surveys conducted between 2022 and 2025 indicated relatively stable average concentrations of ENNs around 50 µg/kg, with high prevalence rates for ENN B (52–66%) and B1 (56–63%), which have consistently ranked among the five most frequently detected mycotoxins since 2023. In parallel, ENN A1 shows a continuous increase in detection frequency, while ENN A remains relatively stable, further highlighting the growing relevance of ENNs in a food safety context [7,8,9,10].

In this context, both in vivo and in vitro methods have been applied to evaluate the toxicological effects of ENNs, although current regulatory initiatives promote the reduction in animal models. Indeed, regulatory initiatives, such as the FDA Modernization Act 2.0, Canada Bill S-5, and European Medicines Agency, promote the reduction in animal use in safety assessment, thereby fostering the development of human-relevant and high-throughput in vitro models [11,12]. These models aim to better reproduce tissue architecture and function, including cell–cell interactions, than conventional 2D cultures while meeting regulatory and ethical requirements. Although in vivo studies provide an integrated assessment of toxicological effects, they are increasingly complemented by innovative in vitro approaches designed to better investigate human physiology and toxicological responses while reducing reliance on animal experimentation.

Several in vitro studies have reported the toxicological effects of ENNs, primarily using intestinal cell models. In colorectal adenocarcinoma Caco-2 cells, ENNs induced pronounced dose-dependent cytotoxicity, particularly ENN A and ENN A1, with IC_50_ values ranging from 2 to 10 µM after 48 h of exposure, depending on the assay employed [13,14,15]. However, fewer studies have addressed hepatic toxicity, despite the liver being the central organ for xenobiotic metabolism and a known target of ENNs [16,17]. In hepatic cell lines such as HepG2, ENNs exhibited cytotoxic effects in the low micromolar range (1–5 µM), depending on the endpoint measured [18,19].

Mechanistic investigations further suggest that ENNs, particularly ENN B1, disrupt mitochondrial membrane integrity, induce oxidative stress, and impair cellular energy metabolism, indicating a key role of mitochondrial dysfunction in ENN toxicity [20]. ENN A1 and B1 are two of the most extensively studied analogues, consistently exhibiting cytotoxic effects in human cell models at low micromolar concentrations [13,18,19], as well as being shown to have ionophoric properties that disrupt intracellular ion homeostasis and mitochondrial function. Several in vitro studies have reported mitochondrial membrane depolarization, ATP depletion, increased oxidative stress, and impaired cellular energy metabolism following exposure to these enniatins. Together, these findings support the notion that mitochondrial dysfunction plays a central role in ENN toxicity and may contribute to the activation of distinct cellular responses depending on the enniatin and the cellular context [13,15,18,19,21,22].

However, even though in vitro studies consistently demonstrate cytotoxicity of ENNs in a low micromolar range, important knowledge gaps remain regarding their overall toxicity. Only a limited number of studies have focused on the understanding of cell death mechanisms involved, such as apoptosis, necrosis or autophagy. Moreover, the genotoxic potential of ENNs remains unclear, as existing in vitro studies provide limited and sometimes contradictory evidence of DNA damage and chromosomal alterations.

In this study, we aimed to investigate the cellular responses involved in HepaRG liver cell toxicity following exposure to ENN A1 or B1, selected based on their known toxicological effects and prevalence in foodstuffs [9,23]. HepaRG cells are an immortalized human hepatic cell line widely used in in vitro toxicology due to their metabolic competence. We specifically focused on apoptosis, necrosis and autophagy. We employed both 2D cell monolayers and advanced 3D spheroid cultures. For the first time, ENN toxicity was evaluated using an organ-on-chip 3D liver cell culture system to better reflect the hepatic tissue architecture and microenvironment, combined with high-resolution imaging, thus improving the physiological relevance of this model compared with traditional 2D in vitro systems. We also carried out spheroid sectioning to determine to what extent ENN toxic effects impacted the peripheral cell layers or penetrated across the entire spheroid.

2. Results

2.1. Impact of ENN A1 and B1 Exposure on Cell Viability

The cell viability was assessed using luminescence-based ATP quantification assays (CellTiter-Glo 2D and CellTiter-Glo 3D, Promega, Madison, WI, USA) on HepaRG cell in both 2D and 3D models and in differentiated (DIFF) and undifferentiated (UND) states. The dose–response curves showed reliable R^2^ values, confirming the quality of the sigmoidal regression. Lower R^2^ values were mainly observed when abrupt changes in viability occurred between two consecutive concentrations (e.g., for DIFF 3D HepaRG model at 24 h, at 15 µM, viability ~45%, and at 20 µM, ~1%), particularly for ENN A1, which makes sigmoidal fitting more challenging (Table 1).

Across all experimental conditions, ENN B1 consistently exhibited higher cytotoxic potency than ENN A1, shown by lower IC_50_ values. This difference was particularly evident in DIFF 3D spheroids, for which the IC_50_ values at 24 h were approximately threefold lower for ENN B1 (~5.3 µM) than for ENN A1 (~16.2 µM). A similar relationship was observed after 48 h exposure, with IC_50_ values of ~1.4 µM for ENN B1 and ~2.5 µM for ENN A1. These differences between ENNs were statistically significant (e.g., ENN A1_3D_DIFF_24h vs. ENN B1_3D_DIFF_24h, adjusted p < 0.001).

For both enniatins, the IC_50_ values generally decreased between 24 h and 48 h, indicating increased cytotoxicity with prolonged exposure. In DIFF 3D spheroids, ENN A1 IC_50_ values decreased from 14.39–18.15 µM at 24 h to 2.22–2.97 µM at 48 h, while ENN B1 IC_50_ values decreased from 4.09–6.59 µM to 1.30–1.57 µM over the same time period. A similar time-dependent decrease was observed in UND 2D monolayer cells and 3D spheroids. This time-dependent increase in cytotoxicity was statistically significant between 24 h and 48 h within the same model and differentiation state (e.g., ENN A1_3D_DIFF_24h vs. ENN A1_3D_DIFF_48h, adjusted p < 0.001).

When comparing 2D and 3D cultures, lower IC_50_ values were observed in 3D spheroids compared with 2D monolayers. After 24 h of exposure, the IC_50_ values for ENN A1 and B1 in DIFF 3D HepaRG spheroids ranged from 14.4 to 18.1 µM and 4.1 to 6.6 µM, respectively, whereas in differentiated 2D cultures, the IC_50_ values could not be determined at this time point and remained above 30 µM for both compounds. Based on the IC_50_ ranges, ENN B1 exhibited an approximately three- to five-fold lower IC_50_ in the 3D model compared with 2D cultures, corresponding to a log_10_ fold change of ~0.5–0.7, indicating differences associated with culture dimensionality. Accordingly, differentiated HepaRG cells displayed significantly different responses between 2D and 3D models at 24 h (e.g., ENN A1_2D_DIFF_24h vs. ENN A1_3D_DIFF_24h, adjusted p < 0.001).

The differentiation status also influenced the cytotoxic sensitivity. In both 2D and 3D systems, DIFF HepaRG cells tended to exhibit lower IC_50_ values than undifferentiated cells, particularly at 48 h. This effect of differentiation was statistically significant under identical experimental conditions (e.g., ENN A1_2D_IND_24h vs. ENN A1_2D_DIFF_24h, adjusted p < 0.001). For instance, at 48 h in 3D spheroids, ENN A1 IC_50_ values ranged from 2.22 to 2.97 µM in differentiated cells compared with 3.19–3.80 µM in undifferentiated cells, while the ENN B1 IC_50_ values were 1.30–1.57 µM in differentiated spheroids and below 2.14 µM in undifferentiated spheroids.

Overall, two-way ANOVA confirmed the significant effects of the culture model (2D vs. 3D), mycotoxins exposure and differentiation status (DIFF vs. UND). Detailed ANOVA results are provided in the Supplementary Materials. These results clearly showed that the in vitro 3D DIFF HepaRG spheroid model, corresponding to the most complex studied model, was of interest to predict hepatotoxicity. This 3D model was therefore selected for further investigation. The diffusion of mycotoxins within spheroids was evaluated after sectioning and staining. This model was the most relevant model of liver architecture.

2.2. Dynamic Effects on 2D and 3D HepaRG Cell Models After Acute ENN A1 and B1 Exposure

We then investigated the dynamic and long-term HepaRG 2D monolayer cell response following ENN A1 or B1 acute exposure with five different doses (1, 5, 15, 30, 48 (ENN A1) and 50 (ENN B1) µM). Cell proliferation was monitored using an Incucyte S3 Live-Cell Analysis System (Sartorius, Göttingen, Germany), and results showed that cytotoxic effects persisted up to four days post-exposure, with a progressive decline in cell confluence (Figure 1A). Notably, treated cells did not proliferate again, indicating irreversible cytotoxic damage after acute exposure. A dose- and time-dependent increase in compromised cell membranes was observed as shown by a change in green fluorescence intensity when compared to the control condition (Figure 1B). A progressive loss of adherent cells, with a reduction exceeding 30% cell confluence at the highest tested concentrations after four days exposure was also observed (Figure 1C). In general, a marked increase in compromised cell membranes (observed by green fluorescence intensity as mentioned) and a decrease in cell confluence was observed at 15 µM for both ENN A1 and B1. However, this effect was more pronounced after ENN B1 exposure, suggesting an overall stronger cytotoxic effect when compared to ENN A1. As shown in Figure 1A, some dead cells remained attached to the well surface, which is likely due to the absence of wash steps during the kinetic assay. We intentionally carried out the experiment this way to preserve time-lapse measurements and to better observe actual cell behavior. It can be assumed that intermediate washing would have removed these non-viable cells and potentially amplified the observed cytotoxic effects.

2.3. Proliferative Analysis in Sectioned 3D HepaRG Spheroids After ENN A1 and B1 Exposure

HepaRG spheroids, using both DIFF and UND cell states, were then cultured in an organ-on-chip system [24] to evaluate the cell proliferation after acute ENN A1 or B1 exposure. Spheroids were exposed for 24 h to either ENN A1 using three different doses (0, 15 and 30 µM) or ENN B1 using four different doses (0, 10, 30 and 50 µM). After treatment, spheroids were fixed before being sectioned by a cryostat. We then used immunostaining with a primary antibody against the Ki67 protein involved in cellular division to detect proliferating cells (Figure 2A,B). Firstly, as exposure concentrations increased (30 µM for ENN A1 and 50 µM for ENN B1), the integrity of the spheroid peripherals was compromised, with evident cell loss during recuperation of HepaRG spheroids. Then, nuclei were clearly visualized by DAPI staining, confirming nuclear labeling within spheroid sections. No differences were observed in the Ki67 signal intensity regardless of the ENNs tested or exposure concentrations (two-way ANOVA followed by Tukey’s post hoc test, p > 0.05) (Figure 2C,D). These results thus suggest that ENN exposure did not increase or reduce cell proliferation compared to untreated controls. It is hypothesized that the effects observed after 24 h of acute exposure may evolve over longer periods. This is supported by Figure 1, which demonstrates a time-dependent impact and the absence of re-proliferation. The three-dimensional organization of spheroids, including cell–cell communication and structural architecture, together with toxin-related parameters such as molecular weight and physicochemical properties, may influence toxin diffusion and cellular stress responses.

2.4. Investigation of Cell Death Mechanisms After ENN A1 and B1 Exposure

We focused on the cell death mechanisms potentially involved in the observed cytotoxicity and absence of cell proliferation after HepaRG spheroid exposure to ENN.

Due to the complex nature of regulated cell death pathways, we assessed apoptosis using several apoptotic markers. We combined complementary assays to target the different stages of apoptosis and evaluate whether the apoptotic processes were engaged following acute ENN exposure in HepaRG spheroids.

We also performed a series of dynamic assays using the spheroid model both in DIFF and UND cell states. Early apoptosis after ENN A1 or B1 exposure was evaluated after 48 h, using five different concentrations (5, 10, 15, 30, and 50 µM). As shown in Figure 3, an increase in early apoptosis was observed, especially during the first 6 h of exposure. However, no statistical differences were observed after 48 h when compared to the normalized control (Kruskal–Wallis test, p > 0.05). Interestingly, at 18 h exposure, a decrease below control levels was observed for all the tested ENN A1 and B1 concentrations, and this shift appeared even earlier, around 12 h, for the highest tested concentration (i.e., 50 µM) (Figure 3A–D). After exposure of both DIFF and UND cell spheroids to ENN A1, similar results were observed, with the effects generally more pronounced at 50 µM (Figure 3A,B), but after exposure of the DIFF model to ENN B1, no sign of an early apoptotic response was observed when compared to the control (Figure 3C,D). In contrast, using the UND cell model, ENN B1 showed a clearer dose-dependent trend (Figure 3B,D). These results are consistent with the determined cytotoxicities (Table 1), highlighting distinct responses depending on the cell differentiation state.

We then dynamically monitored caspase 3/7 activity, as a marker of the late-stage irreversible step in apoptosis in HepaRG DIFF and UND spheroids after 18, 24, and 48 h exposure to five different ENN A1 or ENN B1 concentrations (5, 10, 15, 30, and 50 µM). As shown in Figure 4, ENN A1 significantly increased the caspase activity at 18 h in DIFF spheroids at the highest tested concentration (i.e., 50 µM, p-value < 0.001) (Figure 4A). Interestingly, at 24 h and for the lower concentrations (namely, 5 and 15 µM, p-value < 0.023 and 0.014, respectively), a significant increase in caspase 3/7 activity was also observed, which was not measured at 48 h, likely due to increased cell death. Moreover, the fact that significant effects were only observed at 5 and 15 µM but not at 10 µM may be attributed to variability among the individual spheroids due to their physiological state that may influence measurements and increase the standard deviation. On UND HepaRG spheroids, significant caspase 3/7 activity was observed at 5 µM after 24 h exposure, but no effects were measured at 18 and 48 h (Figure 4B).

In contrast, exposure to ENN B1 (Figure 4C,D) did not result in any statistically significant changes in caspase 3/7 activity across all the studied time points and concentrations for either DIFF or UND cell spheroids (two-way ANOVA with Tukey’s post-hoc test, p-value > 0.05).

Overall, the significant increase in caspase 3/7 activity induced by ENN A1 but not by ENN B1 aligns with our earlier results using an early apoptosis assay.

Following these findings, HepaRG spheroid sections were immunostained using cleaved caspase-3 (CC3), a marker associated with early-stage apoptosis. To do so, spheroids in DIFF and UND states were exposed to ENN A1 at 0, 5, 15, and 30 µM and ENN B1 at 0, 10, 30, and 50 µM for 24h (Figure 5). Overall, an increase in late-stage apoptotic execution was observed, especially in DIFF spheroids after ENN A1 exposure (Figure 5C,D). Consistent with the results obtained for early apoptosis and caspase 3/7 activity, ENN A1 induced a more visible late-stage apoptotic response than ENN B1. Moreover, a significant difference in caspase-3 cleavage (CC3 signal), was detected between DIFF and UND spheroids exposed to 15 µM ENN A1, reinforcing the fact that cellular differentiation state influences the response to mycotoxin exposure, with the DIFF cell state being generally more sensitive to ENN exposure.

While a slight increase in apoptotic markers (early apoptosis, caspase-3/7 activity, and CC3 staining) was observed following exposure of spheroids cultured in the organ-on-chip system to ENN A1, no significant apoptotic signal was detected for ENN B1. We, therefore, explored whether an additional cell death mechanism could be involved following acute exposure of differentiated (DIFF) and undifferentiated (UND) spheroids to ENN A1 or ENN B1. The membrane integrity was assessed using a fluorescent DNA-intercalating dye, which emits a signal upon loss of plasma membrane integrity. This signal reflects compromised membrane integrity and may result from late-stage apoptosis (secondary necrosis) or from primary necrotic processes induced by severe cellular stress.

ENN A1 induced a clear and sustained loss of membrane integrity over 48 h in DIFF spheroids (Figure 6A), whereas no marked membrane disruption was observed in UND spheroids (Figure 6B). In contrast, ENN B1 exposure resulted in pronounced membrane permeability, particularly in UND spheroids, with a strong and statistically significant signal detected at the highest tested concentration (50 µM; Kruskal–Wallis test, p < 0.05) (Figure 6C,D). Overall, except for ENN A1 in UND spheroids, ENN exposure induced cellular stress associated with rapid membrane permeabilization, which may reflect necrotic processes and/or late-stage apoptotic events.

Given that both apoptotic activation and loss of membrane integrity were observed within 24 h following exposure to ENN A1 or ENN B1, we next evaluated whether autophagy was also involved in the cellular response. As shown in Supplementary Figure S2, a significant increase in autophagic activity was transiently observed following exposure to 30 µM ENN B1, suggesting induction of a stress response. However, this signal rapidly declined, likely due to excessive toxicity and cell death at this concentration. In contrast, ENN A1 did not induce significant changes in autophagic fluorescence under any of the tested conditions. These results indicate that distinct stress-associated cellular responses are engaged depending on the enniatin considered, with ENN B1 transiently activating autophagy, although the overall autophagic activity remained limited.

2.5. DNA Damage

Exposure to ENN A1 or B1 appears to lead to multiple cell death pathways raising questions concerning DNA damage which may also be a major consequence. To assess whether ENNs induced DNA damage, HepaRG spheroids were cultured in the organ-on-chip model, in both DIFF and UND conditions, then exposed for 24 h to different concentrations of ENN A1 or B1. Following treatment, spheroids were sectioned and analyzed by immunostaining using the γH2AX marker, a phosphorylated form of the histone indicating a double-strand break. As shown in Figure 7A,B, the DNA double-strand break formed was clearly visible, as shown by green subnuclear foci. In general, for both ENN A1 and B1, higher DNA damage ratios were observed for the DIFF cell state spheroids than for the UND ones. The DNA damage was most intense for 30 µM ENN B1 exposure (Figure 7C,D). To confirm these results, the DNA double-strand break was also quantified, and a marked increase in DNA damage was clearly observed for the 30 µM ENN B1 condition, with a signal approximately seven times higher than the control, while the signal was two times higher for 10 µM and four times for 50 µM exposure (Figure 7D).

For ENN A1, the quantified signal, regardless of concentration, was about four times higher than the control. As the IC_50_ values for HepaRG 3D DIFF models ranged from 14.39 to 18.15 µM for ENN A1, and DNA damage appeared at around 5 µM, ENN A1 was thus shown to induce a genotoxic effect (Figure 7C).

For ENN B1, cytotoxicity was observed at 24 h, with IC_50_ values ranging from 4.09 to 6.59 µM. Under the experimental conditions used in this study, DNA damage was assessed only at concentrations ≥10 µM, which are above the IC_50_ values determined at 24 h; so, further studies need to be conducted at very low concentrations below the IC_50_ at 24 h. Across all the markers analyzed, staining was distributed throughout the spheroid structure rather than being restricted to peripheral cell layers, indicating that the ENN effects were not limited to the outer cell layers.

To validate the obtained results, the same treatment was carried out using a standard 2D HepaRG monolayer (either DIFF or UND). After 24 h of exposure, cells were fixed and stained with DAPI, γH2AX and a CellMask to better visualize the cell membranes (Figure 8A).

The results showed better detection of DNA damage (γH2AX foci) using the 2D model, likely due to the microscopy method used. However, a strong reduction in cell number was clearly observed with increasing ENN A1 or B1 concentrations (Figure 8A,B). For both ENN A1 and B1, the cell numbers were reduced at intermediate concentrations (15 and 30 µM, respectively) using the UND cell state, while at the higher concentrations (30 µM for ENN A1 and 50 µM for ENN B1), only limited cell death was observed, as cells were attached to the surface (Figure 8A).

To better quantify the cell death in the 2D monolayer, nuclei were counted following DAPI staining, as a reduction in the number of measurable nuclei directly reflects cell loss (Figure 8B). A significant reduction in the total measured nuclei, from 5 µM for ENN A1 in DIFF cells (group BC vs. control AB) and from 15 µM ENN A1 in UND cells (group AC vs. control A), was observed. At the highest concentration (30 µM), complete loss of cells was observed, regardless of the cell state. For ENN B1, the toxic effects were more pronounced, with a significant decrease in the number of nuclei already visible at 10 µM, and a complete loss of cells was observed at 50 µM for both DIFF and UND cells. In summary, while some peripheral cells may be lost during spheroid treatment, the 3D model clearly preserved a spherical and compact cell architecture during exposure experiments. This architecture also allowed us to visualize the internal damage, especially at higher ENN concentrations, meaning that the mycotoxins did not only impact the surface cells, but they also penetrated the spheroids. In contrast, using the 2D monolayer model, the damaged cells tended to detach from the surface; thus, some results could not be fully understood, and this clearly shows that, in the latter conditions, there could be an overestimation of the cellular stress observed after exposure to mycotoxins.

3. Discussion

Concerning ENNs’ cytotoxicity, the IC_50_ values obtained in the present study confirm the cytotoxic effects of ENNs, with clear time- and dose-dependent responses between 24 h and 48 h of exposure. In DIFF HepaRG cells, a time-dependent decrease in IC_50_ values was observed, reaching 2.22–2.97 µM for ENN A1 and 1.30–1.57 µM for ENN B1 after 48 h, corresponding to approximately seven- and fourfold reductions, respectively, and indicating the higher cytotoxicity of ENN B1. In contrast, Coulet et al. (2024) [23], using the same 3D HepaRG model, reported higher IC_50_ values, with ENN A1 appearing more cytotoxic than ENN B1 (~3 µM for ENN A1 and ~8 µM for ENN B1 at 48 h). When compared with studies performed in HepG2 cells, our results fall within the lower range of IC_50_ values reported in the literature, although substantial variability exists depending on the exposure duration and the cytotoxicity assay employed. For instance, ENN A1 IC_50_ values after 24 h exposure ranged from 11.6 ± 5.7 µM using the MTT assay [13] to 2.6–4.2 µM with the BrdU assay [18], while higher values were obtained using Alamar Blue [18]. Similar variability was observed for ENN B1, with IC_50_ values ranging from 24.3 ± 6.3 µM (MTT, 24 h) [11] to 2.8–3.5 µM (BrdU) [18], highlighting the strong influence of the cytotoxicity assay on IC_50_ determination.

In the present study, an ATP-based luminescence assay was employed to evaluate the cytotoxicity, which may explain the slight difference IC_50_ values, as it directly quantifies intracellular ATP levels and reflects global metabolic disturbances. Importantly, ATP quantification is particularly suited to 3D spheroid models, as MTT- and resazurin-based assays are limited by diffusion constraints. The same assay was therefore applied to both the 2D and 3D HepaRG models to directly compare the obtained data. Moreover, the cell differentiation status (DIFF versus UND) and cell model (2D versus 3D) significantly influenced the cytotoxic responses. DIFF HepaRG cells represent a more physiologically relevant liver model than UND HepaRG or HepG2 cells, as they display hepatocyte-like morphology and retain key liver-specific functions, including phase I and II enzymes, transporters, and functional bile canaliculi [25,26], whereas HepG2 cells show limited CYP450 expression [27]. In 3D spheroids, these properties are further reinforced by spatial organization, cell–cell interactions, and diffusion gradients, which collectively modulate compound penetration, metabolism, and stress responses. Consistent with previous observations [28], DIFF 3D spheroids generally exhibited a differential sensitivity compared with UND spheroids, supporting the relevance of differentiated 3D HepaRG models for the hepatotoxicity test.

Beyond cytotoxicity, we also investigated the mitogenic effects and cell-death mechanisms. No increase in Ki67-positive cells was detected following acute exposure to ENN A1 or ENN B1, confirming the absence of mitogenic activity, in agreement with previous studies [29,30]. ENN A1 induced moderate apoptotic responses, particularly in DIFF 3D spheroids, as evidenced by early apoptosis markers, caspase-3/7 activation, and CC3 immunostaining. In contrast, ENN B1 did not induce detectable apoptosis. Both enniatins were associated with the rapid loss of membrane integrity, reflecting severe cellular stress. As discussed in the Results Section, this membrane permeabilization may arise from late-stage apoptosis or primary necrotic processes. Autophagy was only transiently induced following ENN B1 exposure and remained limited overall, suggesting a minor contribution under the tested conditions.

In addition to apoptosis and necrosis, ferroptosis may represent an alternative mechanism contributing to ENN-induced cell death. ENN B has been shown to activate ferroptosis signaling and increase intracellular iron in Atlantic salmon hepatocytes [31], and ENNs have been reported to induce ROS generation and lipid peroxidation in Caco-2 cells [15]. These observations support the further investigation of ferroptosis in HepaRG models.

Regarding genotoxicity, our results showed increased γH2AX staining in DIFF HepaRG spheroids, while UND cells were less affected. ENN A1 induced a fourfold increase in the γH2AX positive area at all tested concentrations (5–30 µM), including at 5 µM, a dose below the IC_50_, supporting a genotoxic effect. ENN B1 induced a marked increase in γH2AX staining only at 30 µM, a concentration associated with cytotoxicity. These findings are consistent with previous reports describing ENN DNA damage, including studies by De Felice et al. (2023) and Behr et al. (2025) [20,32]. Genotoxic effects have been reported in an immortalized human embryonic kidney cell line (HEK293T) exposed to ENNs at 25 µM [33], in Caco-2 cells at concentrations ranging from 1.5 to 3 µM without associated cytotoxicity [15], in patient-derived lymphocytes at 0.14 µM [29], and in UND 2D HepaRG model exposed to 1.56 µM of ENN B, but with confirmed cytotoxicity [34]. In contrast, no increase in DNA damage was observed in V79 Chinese hamster lung fibroblasts following 24 h exposure to ENN B at concentrations ranging from 0.1 to 100 µM [35]. Together, these data highlight the marked variability of ENN-induced genotoxic responses depending on the enniatin considered, exposure concentration, and cellular context and underline the need for improved characterization of ENN genotoxic profiles.

Notably, despite their close structural similarity, ENN A1 and ENN B1 exhibited divergent effects, emphasizing the need to consider individual enniatins to fully understand their toxicological effects. As most ENN toxicity data rely on 2D models that do not fully reflect tissue architecture, the 3D systems used in this research provide a relevant alternative. Using a liver organ-on-chip model, ENN associated signals were detected throughout spheroid sections rather than being limited to peripheral layers, indicating penetration within the 3D structure. This is a highly novel result proving that ENNs’ toxicity is not only on peripheral cells, as well as the relevance of more complex 3D spheroid models. Together, these findings highlight the overall value of advanced 3D in vitro approaches for refining the assessment of ENNs’ toxicity.

4. Conclusions

This study demonstrates that ENN A1 and B1 induce cytotoxic effects in HepaRG cells at low micromolar concentrations, with IC_50_ values ranging between approximately 1 and 4 µM and without evidence of mitogenic activity. The mechanisms of cell death were enniatin-dependent: ENN A1 primarily activated apoptotic responses associated with the subsequent loss of membrane integrity, whereas ENN B1 predominantly induced membrane permeabilization consistent with necrotic processes. Under the experimental conditions used, autophagy played only a limited role, while the contribution of other regulated cell-death pathways, such as ferroptosis, warrants further investigation. Importantly, ENN A1 induced significant DNA damage at non-cytotoxic concentrations as early as 24 h of exposure (5 µM), supporting a potential genotoxic effect. In contrast, ENN B1 did not exhibit the same pattern under the conditions tested, despite close structural similarity, highlighting that small chemical differences between enniatins may result in distinct biological responses. These findings emphasize the need to further investigate individual enniatins as well as their combined effects, as co-exposure scenarios may lead to different toxicological outcomes. Using a 3D liver organ-on-chip model combined with spheroid sectioning and high-resolution imaging, we showed that ENN-induced signals were detected throughout spheroid sections rather than being restricted to peripheral cell layers, indicating penetration in the spheroid for the very first time. However, while this observation supports ENN accessibility within the 3D structure, it does not demonstrate a homogeneous functional impact across the spheroid. The present study focused on acute exposure at relatively high concentrations, which may not fully reflect chronic dietary exposure scenarios. In addition, potential interactions between enniatins and other mycotoxins were not addressed and should be considered in future studies. In conclusion, this approach represents a highly valuable step toward more physiologically relevant in vitro liver models and provides a foundation for future studies incorporating chronic exposure, co-exposure, and dynamic flow conditions.

5. Materials and Methods

5.1. Chemicals and Reagents

ENN A1 and B1 were purchased from Sigma-Aldrich (Merck, Darmstadt, Germany) with a stated purity of ≥98%, as verified by the supplier using HPLC analysis, according to the certificate of analysis. William’s E medium, fetal bovine serum (FBS), insulin–transferrin–selenium (ITS-G), hydrocortisone, and phosphate-buffered saline (PBS) were obtained from Gibco (Thermo FisherScientific, Saint-Laurent, QC, Canada). Penicillin–streptomycin solution and trypsin–EDTA were purchased from Wisent Inc (Saint-Jean-Baptiste, QC, Canada).

5.2. Cell Culture

HepaRG cells, a human hepatoma cell line, were obtained from Biopredic International (Saint-Grégoire, France). Cells were cultured in William’s E medium supplemented with 10% fetal bovine serum, 1% penicillin–streptomycin (10,000 U/mL), 0.1% hydrocortisone (50 µM in PBS 1×), and 0.1% insulin–transferrin–selenium (ITS-G, 100). Cells were maintained at 37 °C in a humidified atmosphere containing 5% CO_2_. The culture medium was changed every two to three days, and cells were maintained at approximately 80% confluency before subculturing. For passaging, cells were washed with PBS and detached using 0.025% trypsin–EDTA for 2–3 min at 37 °C. Once detached, cells were resuspended in supplemented medium to stop enzymatic activity. Cell suspensions were either diluted for continued culture or centrifuged at 1000 rpm for 5 min, and cell pellets were resuspended at the appropriate density for seeding in microfluidic devices.

5.3. Cell Viability

The cell viability was assessed using the CellTiter-Glo Luminescent Cell Viability Assay (Promega France, Charbonnières-les-Bains, France). For 2D cultures, the CellTiter-Glo 2.0 kit was used, whereas the CellTiter-Glo 3D kit was applied for spheroids. After equilibration of plates and reagents to room temperature, an equal volume of reagent was added directly to the wells (1:1 with the culture medium). Plates were shaken on an orbital shaker to ensure efficient cell lysis and incubated at room temperature (20 °C) to stabilize the luminescent signal (10 min for 2D; 30 min for 3D). The luminescence was measured after 24 h and 48 h using a PerkinElmer Nivo multimode plate reader, 700 nm IR blocker by the top (PerkinElmer, Waltham, MA, USA), and background signals (medium + reagent without cells) were subtracted. Data were normalized to controls and expressed as percentage viability.

5.4. Design and Preparation of Microfluidic Systems and µ-Slide Well Ibidi Treatment (IbidiTreat) for the Organ-on-Chip Model

5.4.1. Chip Manufacturing

The materials and the protocol used were designed and conceived by [24] “https://doi.org/10.3390/cancers13164046 (accessed on 25 July 2025)”. Briefly, the two layers of the chip were cast in a mix of polymethylsiloxane (PDMS) Dow SYLGARD 184 Silicon Elastomer Clear with a 1:10 ratio of Dow SYLGARD 184 curing agent (Ellsworth Adhesive, Germantown, WI, USA) and then placed in a desiccator for 20 min to eliminate air bubbles before being cured at 80 °C for 45 min in the oven. The layers were demolded and assembled manually after a 30 s exposure to atmospheric plasma using an Enercon plasma gun (Enercon Industries Corporation, Menomonee Falls, WI, USA).

5.4.2. Microfluidic Model Preparation for Cell Culture

After assembling the two layers and adding inlets and outlets (91145A138, McMaster-Carr, Elmhurst, IL, USA) to facilitate pipetting, the chips were autoclaved. The channels were then washed once with isopropanol, followed by three washes with Dulbecco’s Phosphate Buffered Saline (D-PBS) and three times with a non-ionic triblock copolymer of poly(ethylene glycol) and poly(propylene glycol) (PEG-PPG-PEG, Pluronic F-108) (Sigma-Aldrich Canada Co., Oakville, ON, Canada) to prevent the attachment of biological material. After 2 to 3 h of incubation at 37 °C, 5% CO_2_, and 100% humidity, the channels were rinsed three times with D-PBS and three times with the appropriate culture medium.

5.4.3. Spheroid Formation in the Microfluidic Model

A suspension of 8 × 10^6^ cells/mL was prepared for seeding. Then, 200 µL of cell suspension was pipetted three times into the inlet of the channels and then three times into the outlet to homogenize the dispersion of the cells. The appropriate medium was then changed every 24 h until spheroids formed (approximately 4 days after seeding). Prior to ENN exposure, seeding conditions with cells differentiated (no-proliferative cells) were optimized to ensure a reproducible spheroid formation. The spheroid size and morphology were monitored microscopically, confirming consistent diameters before treatment.

5.4.4. Preparation of µ-Slide Well ibidiTreat

µ-Slide Well ibiTreat plates (Ibidi, cat. no. 80826) were used. HepaRG cells in both differentiated and undifferentiated states were seeded at a concentration of 50,000 cells/mL in 300 µL of culture medium per well. The plates were then incubated for 48 h at 37 °C with 5% of CO_2_ and 100% relative humidity, to form a cell monolayer. Each well was seeded with a final volume of 300 µL containing ENN A1 at either 5 µM, 15 µM, or 30 µM and ENN B1 at either 10 µM, 30 µM, or 50 µM. Plates were incubated for 48 h at 37 °C with 5% of CO_2_ and 100% relative humidity.

5.5. Immunofluorescence Analyses on 2D HepaRG Monolayers and Spheroids-on-Chip

After 24 h of ENNs exposure, both monolayer and spheroids were washed three times with PBS, using 300 µL per wash for monolayers and 200 µL for spheroids, directly in wells or channels. Cells were fixed with 10% formalin (Fisher Scientific Company, Toronto, ON, Canada) for 45 min and then washed approximately 5 to 6 times with PBS. For the 3D microfluidic model, the top layer of the chip was removed, and the spheroids were collected and placed in the optimal cutting temperature compound OCT (Leica, Buffalo Grove, IL, USA). Samples were left at 20 °C for 48 h to sediment allowing the spheroids to settle on the same plane. Then, they were frozen on dry ice and stored at −80 °C. The blocks were sent to the Molecular Pathology core facility of the Research Centre of the University of Montreal Hospital Centre (CRCHUM), in Montreal, QC, Canada, for the preparation of 8 µm thick sections using a Leica cryostat (Leica, Buffalo Grove, IL, USA).

For both models the selected samples were incubated with a blocking buffer (PBS 1X, 3% BSA, IgG-free, protease-free, 0.5% Triton 100 10X) for 1 h at room temperature. The samples were either incubated with mouse anti-γH2AX antibodies (targeting the phosphorylated histone H2AX at Ser139) (1:250) (05-636, Sigma Aldrich, Louis, MO, USA), mouse anti-Ki67 antibodies (targeting protein Ki-67, expressed during active phases of the cell cycle) (1:600) (94995, CellSignaling, Withby, ON, Canada), or rabbit anti-CC3 antibodies (targeting the activated form of caspase-3) (1:600) (96615, ThermoFisher Scientific, Waltham, MA, USA) overnight at 4 °C. After one wash with PBS, the wells or slides were incubated in a secondary antibody buffer (PBS 1X, 3% BSA) with AlexaFluor-647 and AlexaFluor-488 secondary antibodies (both at 1:650) (A21244 and A11029, respectively, Invitrogen, ThermoFisher Scientific, Waltham, MA, USA) for 1 h at room temperature. The slides or well were stained with DAPI (1:5000), prepared from a stock solution of 5 mg/mL (D3571, Invitrogen, ThermoFisher Scientific, Waltham, MA, USA), to label the nuclei. Finally, for the 2D model, CellMask Plasma Membrane Stain (C10046, Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) was used to label the plasma membranes. The stain was applied at a 1:1000 dilution and incubated for 15 min at 37 °C, followed by PBS 1X washes.

Slides for the 3D model were then mounted with ProLong Gold Antifade Mountant (P36934, Invitrogen, Invitrogen, Carlsbad, CA, USA). 2D well samples were visualized at the Microscopy Platform of the University of Western Brittany (UBO) using a Confocal ZEISS LSM780 microscope, whereas spheroid images were acquired at the Molecular Pathology Core Facility at CRCHUM using the Olympus VS110 microscope and processed with OlyVIA software (Version 2.9.1, Olympus, Tokyo, Japan). Image analysis was conducted using the ImageJ software, version 1.54f, Java 1.8.0-322 (Fiji/ImageJ, National Institutes of Health, Bethesda, MD, USA).

5.6. Death Mechanism Determination

5.6.1. Caspase 3–7 Measurement

HepaRG cell spheroids were generated in both differentiated and undifferentiated states. Cells were seeded in ultra-low attachment 96-well round-bottom plates at a density of 50,000 cell/mL, with 100 µL of cell suspension per well. Spheroids were maintained in culture for five days at 37 °C, 5% CO_2_, and 100% humidity before treatment. Both ENN A1 and B1 were tested at concentrations of 5, 10, 15, 30, and 50 µM. Apoptosis was measured at 18, 24 and 48 h post treatment by adding 100 µL of Caspase-Glo 3/7 3D Assay reagent (Promega France, Charbonnières-les-Bains, France) directly to each well, following the manufacturer’s recommendations (Supplementary Table S1). The plates were incubated at ambient temperature (20 °C) for 35 min to stabilize the signal. Luminescence was measured using a PerkinElmer Nivo multimode plate reader, 700 nm IR blocker by the top (PerkinElmer, Waltham, MA, USA).

5.6.2. Quantification of Apoptotic Cells and Necrotic Cells

Spheroids were generated using the HepaRG cell line in both differentiated and undifferentiated states, following the same culture conditions and treatment concentrations as described previously for caspase tests (Supplementary Table S1). The RealTime-Glo Annexin V Apoptosis and Necrosis Assay (Promega France, Charbonnières-les-Bains, France) was prepared according to the manufacturer’s instructions. Briefly, the Annexin V NanoBiT Substrate, CaCl_2_, and Necrosis Detection Reagent were thawed at room temperature, while the Annexin V-SmBiT and Annexin V-LgBiT were kept on ice. The 2X Detection Reagent was prepared by sequentially adding the components to prewarmed culture medium, following the manufacturer’s protocol. The mixed solution was added to each well at a 1:1 ratio. Plates were incubated at 37 °C with 5% CO_2_ and 100% relative humidity. Luminescence and fluorescence signals were recorded using a PerkinElmer Nivo multimode plate reader (PerkinElmer, Waltham, MA, USA) every 3 h up to 12 h post-exposure and then every 6 h until 48 h.

5.6.3. Autophagy Measurement

Autophagy activity was assessed using the Autophagy Assay Kit (MAK138, Sigma Aldrich, Louis, MO, USA), which uses a blue fluorescent probe that specifically stains autophagic vacuoles. Cells were seeded in 96-well black plates (Corning Inc., Corning, NY, USA) at 50,000 cell/mL and allowed to adhere overnight. Kinetic analysis was performed to monitor the autophagic activity over time. Following treatment with ENN A1 or ENN B1 at 5, 15 or 30 µM, they were analyzed at multiple time points: every 1.5 h up to 6 h, 9 h, 12 h, 24 h and 48 h (Supplementary Table S1). Then, the medium was removed, and the staining solution was prepared, according to the manufacturer’s instructions, by diluting the Autophagy Detection Reagent in the Assay Buffer. Cells were incubated with the staining solution for 45 min at 37 °C, 100% relative humidity, 5% CO_2_, and protected from light. After incubation, cells were washed three times with water Assay Buffer to remove excess dye. Fluorescence was measured using a microplate reader VICTOR Nivo (PerkinElmer, Waltham, MA, USA) with 355/40 nm (excitation) and 530/30 nm (emission).

5.7. Statistical Analyses

Statistical analyses were performed using GraphPad Prism (version 10.5.0). The IC_50_ values were calculated by non-linear regression using a four-parameter logistic (4PL) model. The normality of the data was assessed using either the Shapiro–Wilk or Kolmogorov–Smirnov test. For normally distributed data, parametric tests were applied, with two-way ANOVA followed by Tukey’s post hoc test. For non-normally distributed data, the non-parametric Kruskal–Wallis test was used. All experiments were performed using at least three independent biological replicates, with a minimum of two technical replicates per condition. For 2D imaging experiments, analyses were conducted on three independent experiments, with nine images acquired per replicate. Image analysis was performed using custom scripts developed in ImageJ software (version 1.54f, Java 1.8.0-322) to quantify the fluorescence intensity and cell number. Data are presented as the mean ± SEM of n = 3 independent experiments, unless otherwise stated. The results were generally expressed after normalization to control conditions, and the statistical significance was defined as p < 0.05 (), p < 0.01 (), and p < 0.001 ().