Comparative study of the transcriptomes of Caco-2 cells cultured under dynamic vs. static conditions following exposure to titanium dioxide and zinc oxide nanomaterials

Abstract

Due to the widespread application of food-relevant inorganic nanomaterials, the gastrointestinal tract is potentially exposed to these materials. Gut-on-chip in vitro systems are proposed for the investigation of compound toxicity as they better recapitulate the in vivo human intestinal environment than static models, due to the added shear stresses associated with the flow of the medium. We aimed to compare cellular responses of intestinal epithelial Caco-2 cells at the gene expression level upon TiO₂ (E171) and ZnO (NM110) nanomaterial exposure when cultured under dynamic and conventionally applied static conditions. Whole-genome transcriptome analyses upon exposure of the cells to TiO₂ and ZnO nanomaterials revealed differentially expressed genes and related biological processes that were culture condition specific. The total number of differentially expressed genes (p < 0.01) and affected pathways (p < 0.05 and FDR < 0.25) after nanomaterial exposure was higher under dynamic culture conditions than under static conditions for both nanomaterials. The observed increase in nanomaterial-induced responses in the gut-on-chip model indicates that shear stress might be a major factor in cell susceptibility. This is the first report on the application of a gut-on-chip system in which gene expression responses upon TiO₂ and ZnO nanomaterial exposure are evaluated and compared to a static system. It extends current knowledge on nanomaterial toxicity assessment and the influence of a dynamic environment on cellular responses. Application of the gut-on-chip system resulted in higher sensitivity of the cells and might thus be an attractive system for use in the toxicological hazard characterization of nanomaterials.

Keywords:

• Gut-on-chip
• nanotoxicology
• titanium dioxide
• zinc oxide
• transcriptomics

Introduction

Various metal-oxide nanomaterials (NMs) (e.g. TiO₂, SiO₂, ZnO, and MgO) have been used in the food processing industry either as food additives or incorporated in food packaging materials attempting to optimize product properties and to increase their shelf life (Weir et al. 2012; Dekkers et al. 2011; Espitia, Otoni, and Soares 2016; He et al. 2016). The rapid increase in production and use of engineered NMs demands a thorough investigation of their potential toxicological effects (Ranjan et al. 2019; Bahadar et al. 2016). As oral exposure to these food-associated NMs can be expected, it is particularly important to study the possibility of harmful effects on the human gastrointestinal tract. Human intestinal epithelial cell line-based models, specifically Caco-2 cells, are frequently used to study the potential effects of NMs on the gastrointestinal epithelium (Bajak et al. 2015; Mao et al. 2016; Chen et al. 2016).

With the emergence of microfluidic technology various organ-on-chip platforms have been launched (Bhise et al. 2014; Kimura, Sakai, and Fujii 2018). More specifically, gut-on-chip devices have been introduced that allow to culture epithelial cells under continuous perfusion resulting in physiological shear stress (Kim et al. 2012). This novel technology is used in an attempt to better recapitulate the physiological microenvironment and the functionality of the human intestine compared to static culturing methods (Kim et al. 2012; Kim and Ingber 2013). Gut-on-chip models with varying cellular make up have been used to study the cellular responses to, and translocation of drugs and (environmental) chemicals and revealed the potential for these dynamic models as an alternative in vitro models to replace animal studies (Beaurivage et al. 2019; Kulthong et al. 2020; Santbergen et al. 2020; Shin and Kim 2018; Kulthong et al. 2018). Recent studies, including those from our group, indicate differences in the basal gene expression and cellular functionality of Caco-2 cells between dynamic and static culture conditions (Kulthong et al. 2021; Chi et al. 2015; Sakharov et al. 2019; Kim et al. 2016).

TiO₂ and ZnO NMs are the most produced NMs worldwide (Vance et al. 2015). TiO₂ powders are allowed to be used as a food additive (E171) and have been shown to contain TiO₂ particles with varying sizes, including a fraction of particles smaller than 100 nm, so-called nanoparticles (NPs) (Rompelberg et al. 2016; Weir et al. 2012; Yang et al. 2014; Peters et al. 2014; Younes et al. 2019). The fraction of particles smaller than 100 nm has been described to lie between 10 and 49%, based on electron microscopy measurement, and recently also percentages above 50% have been reported (Geiss et al. 2020). ZnO powders are considered a ‘GRAS’ (generally recognized as safe) substance and are allowed to be used in biomedical applications by the FDA (Rasmussen et al. 2010). Recent studies on gene expression in epithelial intestinal cells (Caco-2 or Caco-2 co-cultures with HT29 cells) indicated that exposure to TiO₂ NMs (relevant for E171) affected the expression of genes involved in oxidative stress, inflammation, and DNA repair (Proquin et al. 2019; Dorier et al. 2017). Exposure to ZnO NMs affected the expression of genes involved in inflammation, and metal responses (Moos et al. 2011; Moreno-Olivas, Tako, and Mahler 2019). Thus, extensive reports are available on the effects of TiO₂ and ZnO NMs on gene expression in Caco-2 cells cultured under conventional static conditions, but no studies so far focused on cellular responses following exposure to these NMs under dynamic culture conditions. For this study, we have selected both TiO₂ and ZnO NMs because of their different material properties. TiO₂ is recognized as a low soluble inorganic material, therefore, possible toxicological impacts are likely mainly due to cell-particle contact (Warheit and Brown 2019; Singh et al. 2007). ZnO is a more soluble material of which the degree of dissolution depends on particle size, crystal form and the biochemical composition of the solvent the material is dispersed into (Mudunkotuwa et al. 2012; Singh et al. 2011). Thus, possible biological/toxicological effects are likely (partly) due to zinc ions dissolved from ZnO NMs (Cao et al. 2015).

The aim of the current comparative study was to investigate the effects of TiO₂ and ZnO NM exposure on gene expression in Caco-2 cells when cultured under dynamic (gut-on-chip) or static (Transwell) conditions. For this, we used industrial representative NMs, namely a food-grade E171 TiO₂ material, and representative material from the EU nanomaterials repository (NM110) for ZnO. Briefly, Caco-2 cells were grown for 21 days in Transwells, according to a standard protocol (Hubatsch, Ragnarsson, and Artursson 2007), and in our gut-on-chip device (Figure 1), as previously described (Kulthong et al. 2020). The cells were subsequently treated with non-toxic doses of TiO₂ and ZnO NMs. Gene expression data were obtained using a microarray platform. A bioinformatics approach, consisting of linear models and intensity-based moderated t-statistics, was used for the identification of differentially expressed genes and gene set enrichment analysis (GSEA) was applied for the identification of affected biological pathways.

Figure 1. Schematic representation of the gut-on-chip system.

Materials and methods

Chemicals and reagents

Bovine serum albumin (BSA), Dulbecco’s Modified Eagle Medium (DMEM), penicillin-streptomycin, Hank’s balanced salt solution (HBSS) were obtained from Sigma-Aldrich (Zwijndrecht, Netherlands). Phosphate Buffered Saline (PBS), heat-inactivated fetal bovine serum (FBS) and MEM-non-essential amino acids were purchased from Fisher Scientific (Landsmeer, Netherlands). TiO₂ NM E171 was provided by a commercial supplier and was characterized and used in earlier studies (Helsper et al. 2016; Peters et al. 2014). NM110 was obtained from the Joined Research Center (JRC nanomaterials repository, ISPRA, Italy).

Design of the gut-on-chip system

The microfluidic gut-on-chip system has been developed and described previously (Kulthong et al. 2020), a schematic picture of the design is given in Figure 1. In brief, the chip consists of three 15 × 45 mm (width × length) re-sealable glass slides that result in two flow chambers [i.e. an upper apical (AP) and lower basolateral (BL) chamber] upon assembly (Micronit, Enschede, Netherlands). Both the upper and lower glass slides were spaced from the middle layer membrane by a 0.25 mm thick silicone gasket. The flow chambers were separated by a glass slide containing a polyester (PET) porous cell culture membrane (0.4 µm pore size and a surface area of ∼1.6 cm²) on which the cells are grown. The volume of the AP chamber is 75 mm³ with a chamber height of 0.25 mm (membrane to top layer) and the BL chamber is 110 mm³ with a chamber height of 0.65 mm (bottom layer to membrane), resulting in a total volume of 185 mm³ (µL) of the device. The chip was placed in a chip holder with a quick locking mechanism, constructed for sealing the glass slides of the chip together and for connection of external capillaries to the chip via specific ferrules to ensure tight connections and a leak-free system.

Cell culture

A Caco-2 cell line (HTB-37), derived from a human colorectal adenocarcinoma, was obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). The cells were grown (at passage number 29–45) in a complete culture medium, consisting of DMEM supplemented with 10% FBS, 1% penicillin-streptomycin, and 1% MEM non-essential amino acids, further referred to as DMEM⁺. Conventional DMEM contains sodium bicarbonate at a concentration that is optimized for interaction with the standard enriched (i.e. 5%) CO₂ conditions in a cell culture incubator. However, the gut-on-chip system is a closed system that does not allow for medium interaction with the CO₂ in the incubator. Therefore, 10 mM sodium bicarbonate was added to the DMEM⁺ in the gut-on-chip system to optimize the pH buffering capacity. For the Transwell experiments, the cells were seeded at a density of 75 000 cells per cm² on 12-well Transwell polyester inserts (0.4 µm pore size, 1.12 cm² surface area; Corning, Amsterdam, Netherlands) and cultured in DMEM⁺ for 21 days. The medium was changed every two to three days. For the gut-on-chip experiments, the cells were also seeded at a density of 75 000 cells per cm² and the cells were allowed to attach to the membrane on the middle layer of the chip for 24 h. After attachment, the membrane middle layer of the chip was inserted in the microfluidic chip, and the cells were exposed to a continuous flow of 100 µL/h DMEM⁺ until day 21 of culturing. By doing so, the shear stress in the apical compartment was ∼0.002 dyne/cm² at the membrane surface, where the cells are grown. Previously, proliferation and differentiation of the cells, and barrier integrity and permeability have been evaluated in both models. The results indicated fully differentiated cells and a barrier with proper integrity and functionality after 21 days of culture in both models (Kulthong et al. 2020).

A constant flow was introduced in the chip using a microsyringe pump (NE-4000, New Era Pump Systems, Inc.) equipped with two polypropylene syringes (30 mL, Luer-locktm, Becton, Dickinson and company), with each syringe connected to either the AP or the BL compartment using Fluorinated Ethylene Propylene (FEP) tubing (0.50 mm inner diameter, with a length of 25 and 15 cm for the inlet and outlet, respectively). Before the start of each experiment, all tubing and chips were sterilized using an autoclave and rinsed with 70% ethanol. Tubing and chips were prefilled with medium to eliminate air bubbles in the system. The entire system was placed in an incubator at 37 °C to maintain appropriate cell culture conditions.

Size distribution and sample preparation of NMs

NMs were prepared and dispersed according to the Nanogenotox dispersion protocol (Jensen et al. 2011). Briefly, an accurately weighed sample of 15.4 mg of finely powdered TiO₂ and ZnO was added to 30 μL of 96% ethanol to pre-wet the NMs followed by the addition of 570 μL 0.5 mg/mL BSA solution (final NM concentration 2.6 mg/mL). The mixture was shaken and put on ice. The suspension was then sonicated using a 400 W probe sonicator set at 10% amplitude for 16 min (Branson Ultrasonics SonifierTM S-450). After that, the stock solution was diluted in DMEM⁺ for dynamic light scattering (DLS) measurement. The concentrations were optimized based on the scattering intensity of the materials, which was higher for the TiO₂ NMs than for the ZnO NMs. The concentrations used for size measurement were 10 and 100 µg/mL for TiO₂ and ZnO NMs, respectively. The hydrodynamic diameters of the NMs were determined using a custom-built DLS setup with a fixed 90 degrees detection angle. The instrument consists of an ALV/SO SIPD Single-Photon Detector with ALV Static and Dynamic Enhancer Fiber optics, an ALV7002-USB correlator, and a Cobolt Samba-300 DPSS laser with a wavelength of 532 nm. Samples (n = 3) were analyzed with each measurement consisting of 10 technical replicate measurements of 30 s each. The results are expressed as the average hydrodynamic diameter (nm) ± standard deviation (SD) that was calculated using AfterALV (Dullware, USA) software.

In vitro sedimentation, diffusion, and dosimetry (ISDD) modeling of NMs

The deposited fraction of the exposure doses of the particles was calculated using the In vitro Sedimentation, Diffusion and Dosimetry (ISDD) model developed by the Pacific Northwest National Laboratory (Hinderliter et al. 2010). The following parameters were used as input in the ISDD model: the hydrodynamic diameters of the NMs in DMEM⁺ as measured by DLS (Table 1), medium column height (44.6 and 0.25 mm for the Transwell and gut-on-chip, respectively), temperature (310 °K), media density (1 g/ml), and media viscosity (0.0009 N s/m²) (Abdelkhaliq et al. 2018).

Table 1. Physicochemical characteristics of the selected NMs.

Material code	Core material	Product type	Crystal structure	Size (TEM; nm)	Hydrodynamic diameter (nm) ± SD of NMs in DMEM^+after 1 h	PDI ± SD
E171	TiO2	Powder	Rutile/anatase	60–300^a	262 ± 21	0.27 ± 0.11
NM110	ZnO	Powder	Zincite	114 ± 97^b	229 ± 13	0.17 ± 0.04

^aHelsper et al. (2016), ^bSingh et al. (2011) average equivalent circular diameter of well dispersed primary particles and their aggregates/agglomerates.

Viability assay

Cytotoxicity of NMs was assessed using a WST-1 assay, which is a mitochondrial activity-based cell viability assay. Caco-2 cells (50 000 cells/cm²) were seeded in 96-well plates. After 24 h, the medium was discarded and was subsequently replaced with various concentrations of TiO₂ (0, 10, 25, 50, 100, or 150 µg/mL) and ZnO (0, 0.4, 2, 10, 50, or 100 µg/mL) NMs dispersed according to the protocol mentioned above and diluted in DMEM⁺ for 24 h. Concentration ranges were based on data from the literature, which indicated that TiO₂ NM is less cytotoxic than ZnO NM, hence the higher concentration ranges (Singh et al. 2011; Proquin et al. 2017). Triton X-100 (0.25%, v/v) was used as a positive control. At the end of the treatment period, 10 µL WST-1 solution was added and further incubated for 1 h. The plate was centrifuged at 200 g for 10 min and 70 µL supernatant was collected. The absorbance of the supernatant was measured at 440 and 630 nm using a microplate reader (Synergy HT, BioTek, VT, USA), and the background absorbance of the NMs was subtracted. The percentage of cell viability was calculated by dividing the absorbance of the treatment group with the absorbance of the negative control (cells without NMs exposure) and multiplying the value by 100.

NM exposure to Caco-2 cells in the gene expression study

At day 21 post-seeding, Caco-2 cells in the gut-on-chip and Transwell were exposed to equal deposited NM doses based on the ISDD modeling. For the gut-on-chip, NM suspensions of 100 µg/mL TiO₂ (n = 4) and 80 µg/mL ZnO (n = 4) were prepared in DMEM⁺ [DMEM⁺ was used as a control (n = 4)] and perfused via the upper channel with a flow rate of 100 µL/h. DMEM⁺ without NMs was pumped through the basolateral channel with an equal flow rate. After about 42 min the upper chamber was fully filled with the NM solution and the flow was stopped for 6 h to exclude the possible influence of laminar flow in the gut-on-chip system resulting in different exposure doses as reported in our previous study (Kulthong et al. 2020). This will enable the best comparison of the cellular response by cells grown in the two systems. A 6 h incubation period was selected based on a whole-genome gene expression study with Caco-2 cells exposed to TiO₂ NM (E171) for time periods ranging between 2 to 24 h (Proquin et al. 2017). While differences in gene expression at the various incubation times were described, similar pathways were reported to be affected. We, therefore, selected a timepoint in the time range they described.

In the Transwells, the cell culture medium on the AP side was replaced with 500 uL NM suspension of 50 or 10 µg/mL for TiO₂ (n = 4) or ZnO (n = 4), respectively, and the medium of the basolateral side was refreshed with DMEM⁺ without NMs followed by 6 h incubation.

RNA isolation

Caco-2 cells were treated with NMs in the gut-on-chip or Transwell for 6 h. After the 6 h incubation, the apical chamber of the chip was perfused with 100 µL DMEM⁺ to wash the cells. After that, 100 µL RLT lysis buffer was perfused through the chip and incubated for 2 min followed by another 100 µL RLT lysis buffer. The entire volume of RLT solution was collected as cell lysate and total RNA was extracted using the Qiagen RNAeasy Micro kit according to the manufacturer’s instructions. The RNA amount and quality were determined using a Nanodrop system (ND-1000 Thermoscientific Wilmington, Delaware, USA).

To the cells cultured on Transwell membranes 350 µL of RLT lysis buffer were added after washing the cells with DMEM⁺, cell lysates were then collected and processed using the same protocol as above.

Affymetrix microarray processing and analysis

The isolated RNA was subjected to genome‐wide expression profiling. In brief, total RNA was labeled using the Whole-Transcript Sense Target Assay (Affymetrix, Santa Clara, CA, USA) and hybridized on Human Gene 2.1 ST arrays (Affymetrix). The quality control and data analysis pipeline have been described in detail previously (Lin et al. 2011). Normalized expression estimates of probe sets were computed by the robust multiarray analysis (RMA) algorithm (Bolstad et al. 2003; Dai et al. 2005) as implemented in the Bioconductor library affyPLM. Probe sets were redefined using current genome definitions available from the NCBI database, which resulted in the profiling of 29 597 unique genes (custom CDF version 23) (Ritchie et al. 2015). Differentially expressed probe sets (genes) were identified by using linear models (library limma) and an intensity-based moderated t-statistic (Abatangelo et al. 2009; Sartor et al. 2006). Probe sets that satisfied the criterion of p < 0.01 were considered to be significantly regulated. Microarray data have been submitted to the Gene Expression Omnibus (accession number: GSE158620). All groups were n = 4, with the exception of one outlier in the TiO₂ NM treatment group in the Transwells, which was excluded based on a multi-dimensional scaling plot resulting in n = 3 for this group.

Biological interpretation of array data

Changes in gene expression were related to biologically meaningful changes using gene set enrichment analysis (GSEA). It is well-accepted that GSEA has multiple advantages over analyses performed on the level of individual genes (Allison et al. 2006; Subramanian et al. 2005; Kanehisa et al. 2017). GSEA evaluates gene expression on the level of gene sets that are based on prior biological knowledge, GSEA is unbiased, because no gene selection step (fold change and/or p‐value cutoff) is used; a GSEA score is computed based on all genes in the gene set, which boosts the S/N ratio and allows to detect affected biological processes that are due to only subtle changes in expression of individual genes. Gene sets were retrieved from the expert‐curated KEGG database (Kanehisa et al. 2017) (BRITE Functional Hierarchy level 1). Moreover, only gene sets comprising more than 15 and fewer than 500 genes were taken into account. For each comparison, genes were ranked on their t‐value that was calculated by the moderated t‐test. Statistical significance of GSEA results was determined using 1000 permutations.

Results

Physicochemical characterization of the NMs

Both TiO₂ and ZnO NMs were obtained in powdered form and were dispersed according to the Nanogenotox protocol (Jensen et al. 2011). The resulting suspension was subjected to DLS analysis to determine the hydrodynamic size of the nanomaterials. The average hydrodynamic sizes of the materials after 1 h of incubation in DMEM⁺ are given in Table 1. The polydispersity index (PDI), indicating whether the size distribution of the particles in the suspension is homo- or heterogenous, of both suspensions, was <0.7 showing a homogenous size distribution (Stetefeld, McKenna, and Patel 2016). The hydrodynamic sizes of the materials in DMEM⁺ were stable for 24 h (data not shown), which is the longest exposure time used in experiments.

Selection of non-cytotoxic concentrations of NMs

Cytotoxicity experiments, using a WST-1 assay, were performed to select non-toxic concentrations of NMs to be applied in the subsequent gene expression study. Proliferating (1-day old) Caco-2 cells were used to model the worst-case scenario for cytotoxicity as they are generally considered more sensitive to the toxicity of NMs than differentiated (21-day old) cells (Gerloff et al. 2013; Thompson et al. 2012; Ude et al. 2017). The cells were exposed to TiO₂ and ZnO NM concentrations up to 150 and 100 µg/mL, respectively, for 24 h. As shown in Figure 2, no cytotoxicity (>80% viability) was observed for TiO₂ NMs for any of the tested concentrations, whereas cytotoxic effects of ZnO NMs were observed at concentrations of 50 µg/mL and higher. No cell viability data of Caco-2 cells exposed to ZnO (NM110) NM is described in the literature, but cell viability data of A549 cells using a WST-1 assay showed results similar to our data (i.e. 85% viability at 25 ug/ml and 45% viability at 50 ug/ml; Singh et al. 2011). As the design of the 96-well plates (used for the cytotoxicity experiments) and the gut-on-chips and Transwells (used for the gene expression experiments) are different, the exposure is also given as the deposited mass of the NMs for which the calculations are described below.

Figure 2. Cell viability of 1-day old Caco-2 cells exposed for 24 h to increasing concentrations (also given as deposited mass) of (A) TiO₂, or (B) ZnO NM, given as a percentage (±SEM) of the negative control (n = 3); *indicates a significant difference compared with the negative control (one-way ANOVA, Dunnett, p < 0.05).

Prediction of particle sedimentation using the ISDD model

In our experiments, we have used three different culturing devices for cell culture (i.e. 96-well plates for the cell viability experiments and Transwells and gut-on-chips for the gene expression experiments). The three different culturing devices have different dimensions (i.e. the height and width of the cell culture compartment) and volumes. To be able to compare the gene expression profiles, we exposed the cells in the Transwells and gut-on-chips to equal NM deposited mass doses (see Supplementary Table 1 for final deposited mass doses used). To achieve this, nominal exposure concentrations of TiO₂ and ZnO NMs were calculated for each culture device using the In vitro Sedimentation, Diffusion and Dosimetry (ISDD) model (Hinderliter et al. 2010). The calculations showed that to attain equal deposited NM doses in both devices, higher nominal NM concentrations were needed for the chip (i.e. 2-fold higher for TiO₂ and 8-fold higher for ZnO NMs) than for the Transwell exposure. The final selected deposited mass dose of ∼2.0 µg/cm² ZnO NM (which corresponds to an NM concentration of ∼6.5 µg/mL in the cell viability assay) was well below the dose that indicated cytotoxicity in the 96-well plate cytotoxicity measurement. Therefore, this dose even accommodates a potential higher sensitivity of cells cultured under shear stress. The selected deposited mass dose of ∼2.5 µg/cm² TiO₂ NM (corresponding to an NM concentration of ∼12.5 µg/mL in the cell viability assay) is a rather low dose when considering the cell viability results. Due to the smaller volume in the gut-on-chip system, a higher nominal NM concentration was used to reach a deposited mass dose equal to that in the Transwell system. Because of this higher concentration the maximum concentration limit at which suspension stability could be maintained was reached at a deposited mass dose of ∼2.5 µg/cm2 TiO₂ NM.

Gene expression in Caco-2 cells exposed to NMs cultured under dynamic and static conditions

NM exposure induced differentially expressed genes in Caco-2 cells were identified in both cell culture models by comparison of the NM exposure groups (i.e. TiO₂ and ZnO) to their respective control groups. In total, the expression of 29 597 genes was evaluated. The total number of differentially expressed genes (p < 0.01) after NM exposure was higher in the gut-on-chip devices than in the Transwells for both the TiO₂ and ZnO NMs (Figure 3).

Figure 3. Differential expression of genes in 21-day old Caco-2 cells grown in a gut-on-chip and in a Transwell exposed to NMs for 6 h. A Volcano plot showing the t-test statistics (p-value) plotted against the fold change of genes in response to TiO₂ NM exposure in cells cultured in (A) Transwells (B) gut-on-chips and ZnO exposure in cells cultured in (C) Transwells (D) gut-on-chips; the dotted line indicates a p-value of 0.01.

When applying a selection based on an FDR <0.25 there were no differentially expressed genes in cells cultured in Transwells exposed to either TiO₂ or ZnO NMs. Therefore, the applied criteria for differentially expressed genes used for further analysis were p < 0.01 and Log2 FC ≥1.2 (Supplementary Table 2). After TiO₂ NM exposure, a low number of 8 (for cells grown in Transwells) and 41 (for cells grown in gut-on-chips) differentially expressed genes (p < 0.01; Log2 FC ≥ 1.2) was identified (Figure 4). There was no overlap of differentially expressed genes between the two models induced by the NM exposure (Figure 4(A)). Of the eight differentially expressed genes in the Transwell after TiO₂ NM exposure, seven genes had no known biological related function and one gene (which was downregulated) was involved in DNA binding processes. Among the 41 differentially expressed genes in the gut-on-chip after exposure to TiO₂ NM, nine genes had a known biological related function (three upregulated genes and six downregulated genes). The biological functions of the upregulated genes (i.e. SLC10A5, SLC38A4, and KRT34) were related to sodium/bile acid transport and cellular structure, while the downregulated genes (i.e. EBLN2, MT1G, TAS2R20, and EID3) were associated with RNA-binding, metal binding, DNA-repair and taste receptors. The other 32 genes were coding for siRNA, small nucleolus, non-coding RNA, or genes without known functions (Supplementary Table 3).

Figure 4. Venn diagrams showing the differentially expressed genes (p < 0.01 and Log2 FC ≥1.2) in Caco-2 cells cultured in the gut-on-chip and in the Transwell after exposure to (A) TiO₂ and (B) ZnO NMs for 6 h; ↑ and ↓ represents up- and downregulation, respectively.

Exposure of the Caco-2 cells to ZnO NMs resulted in 5 (in the Transwell) and 21 (in the gut-on-chip) differentially expressed genes (p < 0.01; Log2 FC ≥1.2). Like for the TiO₂ NM exposure, there was no overlap in differentially expressed genes between the different culture conditions after ZnO NM exposure (Figure 4(B)). All of the differentially expressed genes in the Transwell had an unknown biological related function (e.g. microRNA, uncharacterized genes), whereas in the gut-on-chip 13 (all upregulated) out of 21 differentially expressed genes had a known biological related function. Six of the upregulated genes (i.e. MT1B, MT1A, MT1M, MT1E, MT1G, and MT1H) are coding for metallothionines. The other seven upregulated genes, namely; HSPA6, SLC30A2, OR11H6, EGR2, GADD45B, RGS16, and USP17L5, code for genes involved in stress responses, zinc transposers, sensory transduction, transcription regulation, DNA-damage responses, signal transduction, and protein metabolism, respectively. The eight genes that had no biological relevant function were mostly downregulated (Supplementary Table 4).

Gene set enrichment analysis on gene expression data in Caco-2 cells exposed to NMs under dynamic and static culture conditions

GSEA was performed to elucidate the responses of Caco-2 cells to NM exposure when cultured under dynamic conditions compared to cells cultured under static conditions. The studied pathways were derived from the KEGG database. This database is structured into KEGG categories that are subdivided into category subgroups. Each category subgroup contains various pathways, each represented by a gene set. The seven KEGG categories are: ‘metabolism’, ‘genetic information processing’, ‘environmental information processing’, ‘cellular processes’ and ‘organismal systems’, ‘human disease’ and ‘drug development’ (BRITE Functional Hierarchy level 1).

An overview of enriched pathways in the cells cultured in the Transwells and gut-on-chips after exposure to TiO₂ NMs is presented in Figure 5. The total number of gene sets based on the KEGG database that was analyzed was 312. Of these 312 gene sets, five and 38 pathways were downregulated in the cells grown in the Transwell and gut-on-chip, respectively (p < 0.05 and FDR < 0.25). There were no significantly upregulated pathways in Caco-2 cells exposed to TiO₂ NM under both culture conditions when compared to their unexposed controls. In addition, none of the affected downregulated pathways overlapped between the two culture models (Figure 5(A)).

Figure 5. Overview of downregulated pathways in Caco-2 cells cultured in the Transwells and gut-on-chip devices after exposure to TiO₂ NMs for 6 h. (A) Venn diagram showing downregulated pathways under each culture condition. Pie chart showing the distribution (in percentage) of the downregulated pathways over the KEGG categories in (B) the Transwell and (C) the gut-on-chip.

Following TiO₂ NM exposure of the cells grown in the Transwells, the most prominently downregulated pathway of the five pathways represented the ‘allograft rejection’ pathway (normalized enrichment score, NES = −1.800) which falls under the KEGG category ‘human disease’ and the KEGG category subgroup ‘immune disease’. In the gut-on-chip 38 pathways were downregulated of which the most prominently downregulated pathway was the ‘homologous recombination’ pathway (NES = −2.050), which falls under the KEGG category ‘genetic information processing’ and the KEGG category subgroup ‘replication and repair’ (p < 0.05 and FDR < 0.25) (Supplementary Tables 5, 6). The distribution of the downregulated pathways of each treatment condition over the KEGG categories is presented in Figures 5(B,C). Pathways under the KEGG categories ‘human disease’ and ‘metabolism’ were most affected under the Transwell and gut-on-chip culturing conditions, respectively. The top five downregulated pathways, selected based on association with epithelial cell functions, in both culture models exposed to TiO₂ NM are presented in Table 2.

Table 2. Top five downregulated pathways upon TiO₂ exposure to cells cultured under dynamic and static conditions.

KEGG pathway name	KEGG category	KEGG category subgroup	Size	NES	p-Value	FDR
Downregulated pathways in Transwells
Intestinal immune network for IgA production	Organismal systems	Immune system	44	−1.691	0.006	0.112
Downregulated pathways in gut-on-chips
Homologous recombination	Genetic information processing	Replication and repair	41	−2.050	0.000	0.008
Carbon metabolism	Metabolism	Global and overview maps	114	−2.040	0.000	0.004
Pyruvate metabolism	Metabolism	Carbohydrate metabolism	39	−1.965	0.000	0.009
Propanoate metabolism	Metabolism	Carbohydrate metabolism	33	−1.894	0.000	0.017
DNA replication	Genetic information processing	Replication and repair	36	−1.880	0.001	0.017

Upon ZnO NM exposure, 50 pathways were affected in the Transwell (35 up- and 15 downregulated) and 122 pathways were affected in the gut-on-chip (88 up- and 34 downregulated) (Supplementary Tables 7–10) (Figures 6(A,B)). The most affected upregulated pathways in the Transwell and gut-on-chip models were included in the ‘human diseases’ KEGG category and the most affected downregulated pathways in the ‘metabolism’ KEGG category (Figures 6(C–F)).

Figure 6. Overview of up- and downregulated pathways in Caco-2 cells cultured in the Transwells and gut-on-chip devices after exposure to ZnO NMs for 6 h. (A) Venn diagrams showing the (A) upregulated and (B) downregulated pathways. (C–F) Pie charts showing the distribution (in percentage) of the affected pathways over the KEGG categories. Upregulated pathways in (C) the Transwell and in (D) the gut-on-chip and downregulated pathways in (E) the Transwell and in (F) the gut-on-chip.

In cells grown in Transwells, the most prominently upregulated pathway represented the ‘mRNA surveillance pathway’ (NES = 2.062) under the KEGG category ‘genetic information processing’ and KEGG category subgroup ‘translation’. The most prominently downregulated gene set was associated with ‘fructose and mannose metabolism’ under the KEGG category ‘metabolism’ and KEGG category subgroup ‘carbohydrate metabolism’ (NES = −1.965). In the cells grown in the gut-on-chip, the most prominently upregulated gene set was associated with ‘mineral absorption’ (NES = 2.439) under the KEGG category ‘digestive system’ and KEGG category subgroup ‘organismal systems’. The most prominently downregulated pathway was the ‘pyruvate metabolism’ under the ‘carbohydrate metabolism’ KEGG category and KEGG category subgroup ‘metabolism’ with an NES of −2.327. The top five of most up- and downregulated gens sets, selected based on association with epithelial cell functions, in both culture models exposed to ZnO NM are presented in Table 3.

Table 3. Top five enriched pathways upon ZnO exposure to cells cultured under dynamic and static conditions.

KEGG pathway name	KEGG category	KEGG category subgroup	Size	NES	p-Value	FDR
Upregulated pathways in Transwells
mRNA surveillance pathway	Genetic information processing	Translation	87	2.062	0.000	0.009
Ribosome biogenesis in eukaryotes	Genetic information processing	Translation	76	2.059	0.000	0.005
RNA transport	Genetic information processing	Translation	146	1.943	0.000	0.013
Mineral absorption	Organismal systems	Digestive system	49	1.862	0.000	0.028
Spliceosome	Genetic information processing	Transcription	128	1.807	0.000	0.040
Upregulated pathways in gut-on-chips
Mineral absorption	Organismal systems	Digestive system	49	2.438	0.000	0.000
Protein processing in endoplasmic reticulum	Genetic information processing	Folding, sorting and degradation	160	2.375	0.000	0.000
Apoptosis	Cellular processes	Cell growth and death	133	2.077	0.000	0.001
SNARE interactions in vesicular transport	Genetic information processing	Folding, sorting and degradation	33	2.059	0.000	0.001
Apoptosis-multiple species	Cellular processes	Cell growth and death	31	2.035	0.000	0.001
Downregulated pathways in Transwells
Fructose and mannose metabolism	Metabolism	Carbohydrate metabolism	32	−1.965	0.000	0.021
Folate biosynthesis	Metabolism	Metabolism of cofactors and vitamins	26	−1.700	0.008	0.085
Galactose metabolism	Metabolism	Carbohydrate metabolism	31	−1.628	0.014	0.137
Staphylococcus aureus infection	Human diseases	Infectious disease: bacterial	57	−1.604	0.009	0.127
Citrate cycle (TCA cycle)	Metabolism	Carbohydrate metabolism	29	−1.522	0.032	0.157
Downregulated pathways in gut-on-chip
Pyruvate metabolism	Metabolism	Carbohydrate metabolism	39	−2.327	0.000	0.000
DNA replication	Genetic information processing	Replication and repair	36	−2.149	0.000	0.001
Mismatch repair	Genetic information processing	Replication and repair	23	−2.084	0.000	0.001
Ribosome	Genetic information processing	Translation	113	−2.016	0.000	0.002
Homologous recombination	Genetic information processing	Replication and repair	41	−2.001	0.001	0.002

The exposure of Caco-2 cells to ZnO NM resulted in 27 enriched pathways that were shared between both culture conditions. Of these pathways 18 were up- and 6 were downregulated similarly (p < 0.05 and FDR < 0.25) (Figures 6(A,B); Supplementary Table 11). Among the upregulated pathways were pathways associated with epithelial cell function (e.g. ‘autophagy’, ‘colorectal cancer’, ‘TNF signaling pathway’) (Table 4). The downregulated pathways associated with epithelial cell function included ‘propanoate metabolism’, ‘oxidative phosphorylation’, ‘fat digestion and absorption’. The other three pathways that were shared between both culture conditions had opposite responses. The enriched pathways ‘cell cycle’, ‘RNA degradation’ and ‘lysine degradation’ were upregulated in Caco-2 cells grown in the Transwell but downregulated in the gut-on-chip.

Table 4. Overlapping pathways upon ZnO exposure to cells cultured under dynamic and static conditions.

KEGG pathway name	KEGG category	KEGG category subgroup	Size	NES*	p-Value	FDR
Overlapping upregulated pathways
Autophagy-other	Cellular processes	Transport and catabolism	31	1.639/1.532	0.014/0.028	0.117/0.068
Viral carcinogenesis	Human diseases	Cancer: overview	162	1.634/1.592	0.000/0.000	0.146/0.000
Protein processing in endoplasmic reticulum	Genetic information processing	Folding, sorting and degradation	160	1.563/2.375	0.000/0.000	0.146/0.000
Colorectal cancer	Human diseases	Cancer: specific types	86	1.562/1.495	0.006/0.013	0.136/0.078
Hippo signaling pathway-multiple species	Environmental information processing	Signal transduction	28	1.519/1.681	0.035/0.009	0.141/0.031
Ferroptosis	Cellular processes	Cell growth and death	39	1.496/1.663	0.029/0.008	0.148/0.035
Pathogenic Escherichia coli infection	Human diseases	Infectious disease: bacterial	53	1.486/1.405	0.027/0.048	0.154/0/111
Mitophagy-animal	Cellular processes	Transport and catabolism	64	1.462/2.138	0.029/0.000	0.178/0.001
Autophagy-animal	Cellular processes	Transport and catabolism	127	1.457/1.811	0.008/0.000	0.177/0.014
Epithelial cell signaling in helicobacter pylori infection	Human diseases	Infectious disease: bacterial	70	1.456/1.696	0.023/0.002	0.173/0.030
TNF signaling pathway	Environmental information processing	Signal transduction	110	1.405/1.703	0.024/0.001	0.209/0.031
FoxO signaling pathway	Environmental information processing	Signal transduction	129	1.402/1.507	0.018/0.005	0.200/0.072
Overlapping downregulated pathways
Propanoate metabolism	Metabolism	Carbohydrate metabolism	33	−1.736/−1.952	0.004/0.000	0.152/0.004
Glycolysis/gluconeogenesis	Metabolism	Carbohydrate metabolism	68	−1.708/−1.691	0.002/0.002	0.132/0.027
Valine, leucine, and isoleucine degradation	Metabolism	Amino acid metabolism	47	−1.568/−1.915	0.013/0.000	0.155/0.005
Oxidative phosphorylation	Metabolism	Energy metabolism	116	−1.543/−1.684	0.004/0.000	0.172/0.027
Carbon metabolism	Metabolism	Global and overview maps	114	−1.531/−1.570	0.004/0.003	0.173/0.054
Fat digestion and absorption	Organismal systems	Digestive system	41	−1.461/−1.445	0.037/0/040	0.223/0.114

*Transwell/gut-on-chip.

Discussion

The aim of this study was to assess the changes in gene expression and associated biological pathways in Caco-2 intestinal epithelial cells cultured dynamically in a gut-on-chip or statically in a Transwell following exposure to TiO₂ (E171) and ZnO (NM110) NMs. In vivo, intestinal cells are continuously exposed to shear stress, which has been shown to affect gene expression (Kulthong et al. 2021), function and morphology (Delon et al. 2019; Kim et al. 2017) of Caco-2 cells in vitro. Therefore, a dynamic model better mimics the in vivo microphysiological environment. While inconsistent data is available on the in vivo intestinal shear stress, the estimated shear stress in our model is relatively low (compared to other dynamic intestinal in vitro models). Gene expression studies like the present study aim to provide a better understanding of the influence of flow on cell behavior, but more knowledge is needed on the relation between microphysiological conditions in the gut and cell homeostasis in vivo and in vitro. Here we show that, at the level of gene expression, cells cultured in the gut-on-chip respond stronger to exposure to TiO₂ and ZnO NMs compared with cells cultured in a Transwell. These stronger responses were seen both at the individual gene level and in terms of affected pathways. Different responses of cells cultured under the dynamic and static conditions to compounds or toxicants have been reported previously in other cell types, like endothelial and proximal tubular epithelial cells (Feng et al. 2019; Sakolish, Philip, and Mahler 2019).

We exposed Caco-2 cells to two types of NMs with different physicochemical properties. TiO₂ NMs are generally regarded as stable in suspension and do not dissociate ions in cell culture medium (Warheit and Brown 2019; Singh et al. 2007). ZnO NMs are known to dissolve readily under physiological conditions, the dissolution is influenced by the size and surface properties of the NM and by the biochemical conditions of the dispersion solvent (Mudunkotuwa et al. 2012; Singh et al. 2011). The effects observed following exposure to the ZnO NM (NM110) used in this study can thus likely be attributed to a combination of the ZnO particulate material and Zn ions, whereas the TiO₂ NM (E171) exposure can likely only be attributed to a particulate effect. Therefore, it is not surprising that at the individual gene level, the majority of the exposure-induced differentially expressed genes were different upon exposure to TiO₂ or ZnO NMs. In this study, we ensured that the NM exposure conditions between the static and dynamic systems were comparable. Previously, we have seen that flow highly influences compound-cell interactions (Kulthong et al. 2020), which is expected to be even more pronounced when using particluate exposure, in which sedimentation is a critical issue. Sedimentation can also not be predicted by the ISDD modeling in a dynamic system. Therefore, the flow in the gut-on-chip system was stopped during the NM exposure to the cells. However, as the results of this study show, shear stresses, when applied for a prolonged time previous to NM exposure, still have a profound effect on gene expression by the exposed cells.

Exposure of Caco-2 cells to TiO₂ NM under both culturing conditions induced a very limited number of significantly differentially expressed individual genes with known biological function (mainly downregulated). Interestingly, significantly downregulated genes were the DNA-binding gene (ZNF117) in the cells grown in Transwell and the DNA-repair (EID3) gene in the cells grown in the gut-on-chip device. Both genes are involved in DNA-damage/genotoxicity pathways (Cornu et al. 2008; Wang et al. 2018) and have previously been reported to be affected in Caco-2 cells exposed to E171 (TiO₂ NM) (Dorier et al. 2017; Proquin et al. 2017). At the biological pathway level, five and 38 gene sets (in the Transwell and gut-on-chip, respectively) were significantly downregulated. The affected pathways were different for the two culturing conditions. For Caco-2 cells grown in Transwells, the pathway associated with the immune response ‘intestinal immune network for IgA production’ was most prominently downregulated. Pathways that were affected in Caco-2 cells grown in the gut-on-chip were associated with DNA damage. The pathway, ‘homologous recombination’ was the most prominently downregulated pathway. These results corroborate previous observations of effects of different types of TiO₂ NMs (including E171) on gene expression in epithelial cells that showed involvement of pathways related to oxidative stress, DNA repair, and immune system impairment (Gerloff et al. 2012; Dorier et al. 2017; Proquin et al. 2017; Proquin et al. 2019).

Exposure to ZnO NM induced much stronger responses in Caco-2 cells cultured in the gut-on-chip compared to cells grown in the Transwell model at the individual gene expression level. Twenty-one individual genes were differentially regulated in Caco-2 cells grown in the gut-on-chip. Of the 16 upregulated genes, the highest upregulated genes were metallothionein genes (i.e. MT1B, MT1M, MT1H). High upregulation of metallothionein genes is a common response after exposure to metal NMs and especially metallic ions (van der Zande et al. 2016; Sahu 2016), but metallothionein genes have also been shown to be induced by shear stress alone, as observed in our previous study (Kulthong et al. 2021). In cells grown in Transwells only five genes were differentially expressed, all without a relation to a known biological function. Also, on a pathway level, stronger responses were observed to ZnO NM exposure in cells cultured in the gut-on-chip than in the Transwell.

In the cells grown in gut-on-chip a total of 122 pathways were enriched (i.e. 88 pathways upregulated and 34 pathways downregulated), while in the static system 50 pathways were enriched (i.e. 35 pathways upregulated and 15 pathways downregulated). Seventy pathways were specifically upregulated in cells grown in the gut-on-chip device, which included pathways associated with apoptosis, MAPK-, Jak-STAT-, p53-, and NF-Kappa B- and NOD-like receptor signaling. These pathways have important regulatory roles in a wide variety of cellular processes including cell proliferation, differentiation, apoptosis, stress responses, and immune responses in mammalian cells (Chen et al. 2009; Harrison 2012; Ho and Woodgett 2010; Mitchell, Vargas, and Hoffmann 2016).

Comparison of differentially regulated pathways due to dynamic culture conditions alone, as observed in our previous study (Kulthong et al. 2021), with those that were due to NM exposure under dynamic conditions in the present study showed that several of the affected pathways in the present study were also affected in the previous study by shear stress only (Kulthong et al. 2021). Comparison of the data from the previous study vs. data from TiO₂ NM exposed cells cultured under dynamic conditions in the present study showed 15 pathways that were downregulated in both studies (thus due to shear stress only and due to TiO₂ NM exposure under shear stress). Ten of the 15 pathways were related to metabolism. Comparison of data from ZnO NM exposure under shear stress conditions with shear stress only conditions showed 14 downregulated pathways that were downregulated under both conditions, of which nine were related to metabolism. Finally, a comparison of upregulated pathways due to ZnO NM exposure under shear stress conditions with those due to shear stress only conditions showed 24 pathways that were upregulated under both conditions. Most of these 24 pathways were related to cellular stress and immune responses. This suggests a partial contribution of shear stress-related responses. However, the total number of upregulated pathways after ZnO NM exposure under shear stress conditions was higher (i.e. 88) than the total number of upregulated pathways due to shear stress only (i.e. 52; Kulthong et al. 2021). Also, several of the pathways that were upregulated due to ZnO exposure under shear stress conditions, but not due to shear stress only conditions, were related to cytotoxicity or cellular stress (e.g. ‘toll-like receptor signaling pathway’, ‘natural killer cell-mediated cytotoxicity’ and ‘leukocyte transendothelial migration‘) indicating an additional effect of the NM exposure on the cells.

When considering the effects (on a pathway level) of the ZnO NM on the cells that were shared between static and dynamic culture conditions, it stands out that many of the shared downregulated pathways relate to cellular metabolism (like ‘propanoate metabolism’ under the ‘metabolism’ KEGG category). ZnO NM exposure in both culture conditions also regulated additional non-shared downregulated pathways involved in metabolism, like the ‘fructose and mannose metabolism’ pathway in cells grown under static and the ‘pyruvate and butanoate metabolism’ pathway in cells grown under dynamic conditions. This might indicate that ZnO NM exposure negatively modulated cellular metabolism, interfering with the cellular energy levels in Caco-2 cells. This corroborates the observed decrease of glucose metabolism in lung epithelial cells exposed to ZnO nanoparticles reported in the study of Lai et al. (2015). The mechanism or degree of modulation of the cellular metabolism, however, also appears to depend on the culture conditions as mentioned earlier, indicating a possible additional or sensitizing effect of shear stresses on cellular metabolism (Kulthong et al. 2021). Evaluation of the 24 overlapping gene sets that were modulated in the Caco-2 cells cultured under both culturing conditions showed a strong correlation of ZnO exposure with the autophagy pathway. Autophagy has been proposed as a mechanism involved in nanomaterial toxicity (Stern, Adiseshaiah, and Crist 2012) and has been identified as a major modulator of ZnO NM induced cellular toxicity (Song et al. 2019; Roy et al. 2014). While no other shared pathways associated with autophagy were observed, the Caco-2 cells cultured under dynamic conditions showed several downregulated pathways associated with DNA repair and replication (under the ‘replication and repair’ KEGG category subgroup) including DNA replication, mismatch repair, homologous recombination (in the top 5 of downregulated pathways). Effects on DNA repair and replication are consistent with the Nanogenotox report where ZnO (NM110) has been considered as a potential positive control for genotoxicity [study in Caco-2 cells (Norppa et al. 2013)].

This study provides insights into the underlying mechanisms of the effects of NM exposure using a whole-genome gene expression analysis approach in two different in vitro cell culture systems. A study comparing transcriptomics and proteomics data reports a good correlation between the two, but discrepancies can exist, meaning that not necessarily all genes that are regulated will result in protein translation (Ning, Fermin, and Nesvizhskii 2012). It would therefore be interesting to study the findings of the gene expression analysis presented in this study using a proteomics approach in a future study. Furthermore, since the results indicate differences in responses to NMs, it would be interesting to study the uptake of NM in the cells, to evaluate whether increased uptake, due to a changed functionality of the cells, in the cells could explain the observed differences.

The gut-on-chip reported in our study could be further developed depending on the research question, for instance by incorporating M (microfold) cells, which are known to play an important role in the uptake of NMs (Schimpel et al. 2014; Walczak et al. 2015), by combining the intestinal barrier with immune cells [e.g. THP-1, PMBCs, or RAW 264.7 cells (Kämpfer et al. 2020; Ghasemi et al. 2020; de León-Rodríguez et al. 2019)] to add immunocompetence to the model that could be used to study immunomodulatory effects of NMs, or by incorporating intestinal organoids to further improve the biological relevance and/or to study interindividual differences (Naumovska et al. 2020; Kasendra et al. 2018). Finally, it would be interesting to include the intestinal microbiota into the model to study the interactions between the microbiome and intestinal wall (Jalili-Firoozinezhad et al. 2019).

Conclusions

The obtained results showed that there was no direct overlap in either differential expression of individual genes or of affected pathways in Caco-2 cells cultured under the two different conditions after exposure to TiO₂ NM. However, in both cases, pathways associated with DNA damage were affected. Following exposure to ZnO NM, there was no direct overlap in differentially expressed genes at the individual gene level between both culture conditions, but there were several shared affected pathways. Pathway analysis indicated that, in general, the responses upon ZnO NM exposure were stronger than those upon TiO₂ NM exposure. Both materials caused a stronger transcriptional impact at both the individual gene expression and pathway level in cells cultured under dynamic conditions in a gut-on-chip than under static conditions in a Transwell. Part of these responses can be attributed to shear stress-related effects alone as similarities in gene expression profiles due to the shear stress and NM exposure were noted, but an increased number of affected pathways related to stress responses points toward additional effects of NM exposure. This suggests that shear stress renders the Caco-2 cells more sensitive and might imply an underestimation of NM-induced effects in intestinal epithelial cells cultured under static conditions in vitro. Nonetheless, to the authors’ knowledge, there is no study revealing the gene expression profile of human epithelial cells in vivo upon TiO₂ and ZnO NM exposure, therefore it is quite challenging to clearly indicate which of the two culturing conditions of the Caco-2 cells resembles the human in vivo situation best. This manuscript is the first effort to evaluate the responses of (food-associated) NMs in Caco-2 cells cultured under different conditions (i.e. dynamic and static). The data in this study suggest that the dynamically cultured cells are more sensitive, and thus might be an attractive model to be used for toxicological hazard characterization.

Author contributions

K.K. performed the experiments and together with M.Z. and H.B. wrote the main manuscript text. L.D. partially performed data analysis. G.H. with the support of K.K. and I.M.E. performed the biostatistical analysis.

Acknowledgments

The authors sincerely acknowledge the support and critical comments of Prof. Dr. Ir. IMCM (Ivonne) Rietjens.

Disclosure statement

All authors declare that they have no competing interests.

Data availability statement

All data generated or analyzed during this are included in this published article.

Additional information

Funding

K.K. is supported by a Royal Thai Government Scholarship. This work was in part supported by the Dutch Ministry of Agriculture, Nature and Food Quality (project KB-23-002-022).