1 Introduction

The human intestinal epithelium functions as a critical interface between the body and external environment, maintaining selective barrier properties while facilitating nutrient absorption and immune modulation. Understanding epithelial barrier function and regulation requires physiologically relevant in vitro models. While Caco-2 monocultures have been widely used due to their ability to differentiate into enterocyte-like cells with functional tight junctions and brush border microvilli (Basson et al., 1998; Sambuy et al., 2005), they lack the cellular heterogeneity of the intestinal epithelium (Pereira et al., 2016), particularly a functional mucus layer essential for barrier protection and immune regulation (Song et al., 2023).

Co-culture systems combining Caco-2 cells with mucus-producing HT29-MTX cells better replicate intestinal conditions. The HT29-MTX-E12 subclone exhibits enhanced goblet cell differentiation and mucus secretion (Lesuffleur et al., 1993; Behrens et al., 2001), providing a more physiologically relevant model for studying gut permeability, nanoparticle interactions, and responses to pro-oxidative or pro-inflammatory agents (Antunes et al., 2013; Martínez-Maqueda et al., 2015; Johansson and Hansson, 2016). However, achieving optimal differentiation in these co-culture systems remains challenging, requiring precise optimization of both biochemical stimuli—such as dexamethasone and butyrate (Liang et al., 2000; Willemsen et al., 2003)—and mechanical factors (Frohlich and Roblegg, 2012).

Traditional static culture conditions fail to replicate the physical forces—such as fluid shear stress and cyclic strain—present in the intestinal environment. These mechanical forces are essential for proper epithelial polarization and barrier function (Liu et al., 2022). While dynamic culture systems incorporating controlled mechanical stimulation show promise, their effects on differentiation timing and barrier development remain incompletely characterized. Additionally, established markers of epithelial differentiation like brush border enzyme expression and barrier integrity measurements may not fully capture the complexity of mechanical force-induced changes.

Dome formation—the development of fluid-filled, three-dimensional structures reflecting active ion transport beneath polarized monolayers—represents a promising but underutilized marker of epithelial differentiation (Lechner et al., 2011). These structures indicate functional barrier properties and polarized organization (Lever, 1985; Bohets et al., 2001), resembling the natural architecture of intestinal epithelium. Dome formation typically begins 5–8 days post-confluence, with structures increasing in size and density through fusion events (Hara et al., 1993). These regions exhibit specialized transport functions and increased brush border enzyme expression (Matsumoto et al., 1990; Ferraretto et al., 2007), making dome quantification valuable for assessing differentiation kinetics (Zweibaum et al., 2011). Advanced imaging technologies, particularly confocal microscopy, now enable detailed analysis of dome formation dynamics (Rotoli et al., 2002), potentially providing new insights into differentiation processes. While 3D models like organoids offer improved physiological relevance, their complexity and cost limit widespread application, making optimized 2D co-culture systems a practical alternative (Baptista et al., 2022).

In this study, we introduce an enhanced co-culture model incorporating orbital mechanical stimulation to accelerate differentiation. We establish dome formation as a quantitative marker of differentiation state, employing confocal microscopy and computational image analysis to characterize three-dimensional epithelial organization. Beyond testing the model’s potential, we also explore its application for studying epithelial cell interactions with food components by supplementing the model with in vitro semi-dynamic digested test food containing 1% w/v titanium dioxide (TiO2) nanoparticles. TiO2 is a common whitening agent employed as food additive (E171) (Weir et al., 2012; Peters et al., 2014; Rompelberg et al., 2016; Ropers et al., 2017). While TiO2 is currently approved by the Food and Drugs Administration (FDA) as safe for use in foods up to 1% by weight (FDA, 2024), a position recently reaffirmed by the Joint FAO/WHO Expert Committee on Food Additives (JECFA, 2023), research has shown that TiO2 food additive contain particles in the nanoscale range (<100 nm) that can accumulate in intestinal tissues. These nanoparticles have been associated with adverse effects including genotoxicity and oxidative stress in both in vitro studies (Koeneman et al., 2009; Gerloff et al., 2012; McCracken et al., 2013; Dorier et al., 2015; Cao et al., 2020), pre-clinical mouse models (Wang et al., 2007; Bettini et al., 2017), and humans (Heringa et al., 2018). Based on these concerns the European Food Safety Authority (EFSA) concluded that E171 could no longer be considered safe as a food additive (EFSA Panel on Food Additives and Flavourings et al., 2021; Boutillier et al., 2022).

Our objectives were to: (1) develop and test an enhanced differentiation protocol using orbital mechanical stimulation, (2) establish quantitative analysis methods for dome formation as a differentiation marker, and (3) demonstrate the model’s application by investigating epithelial responses to TiO₂ nanoparticles in the context of an in vitro digested food matrix. This optimized system bridges the gap between simple monocultures and complex 3D models, providing a robust platform for investigating intestinal barrier function, nutrient and drug absorption, while contributing to a growing field of research on nanoparticle interactions with the gut epithelium.

2 Materials and methods2.1 Materials

All materials used in this study were obtained from Sigma-Aldrich (St. Louis, MO, United States) or Merck (Darmstadt, Germany) unless otherwise specified.

2.2 Development of enhanced Caco-2/HT29-MTX-E12 co-culture2.2.1 Cell culture establishment and maintenance

Caco-2 cells and HT29-MTX-E12 cells (ECACC; Porton Down, UK) were maintained in complete Dulbecco’s Modified Eagle Medium (DMEM; Gibco, Waltham, MA, United States). The medium was supplemented with 10% fetal bovine serum (FBS), 1% non-essential amino acids (NEAA; Gibco, Waltham, MA, United States), 1% GlutaMAX™ (Gibco), and 1% penicillin-streptomycin (Gibco). Both cell lines were cultured at 37°C in a humidified atmosphere containing 5% CO2, with media changes performed every 48 h.

For co-culture establishment, both cell lines were independently cultured until reaching 80% confluence, after which they were harvested using 0.25% trypsin-EDTA solution. Cell viability and counts were determined using the Trypan Blue exclusion assay (Bio-Rad Laboratories, Hercules, CA, United States) on a TC20 Automated Cell Counter (Bio-Rad Laboratories). Subsequently, the cells were seeded in a physiologically relevant ratio of 9:1 (Caco-2:HT29-MTX-E12), representing the approximate proportion of enterocytes to goblet cells found in the human small intestinal epithelium (Welcome, 2018; Paone and Cani, 2020). The cell mixture was seeded onto Transwell® permeable supports (pore size: 0.4 μm, surface area: 1.12 cm2; Corning, New York, NY, United States) at a density of 1 × 10⁵ cells/cm2 in 12-well cell culture multiwell plates (Corning). Co-cultures were maintained in complete DMEM with media changes every 48 h throughout the experimental period.

2.2.2 Static and dynamic culture conditions for cell differentiation

Following seeding, co-cultures were initially maintained under static conditions until reaching confluence (approximately 7 days, designated as T0). Post-confluence, cultures were divided into two experimental groups:

(i) Static condition: co-cultures were maintained in a standard cell culture incubator without mechanical stimulation.

(ii) Dynamic condition: co-cultures were placed on a Celltron orbital shaker (Infors HT, Basel, Switzerland) set to 55 rpm, housed within a cell culture incubator. The shaker’s orbital diameter of 25 mm generated an estimated fluid shear stress of 0.17 Pa (1.7 dynes/cm2) at the cell surface [calculated according to Dardik et al. (2005)]. This value falls within the physiological range, representing moderate-high fluid shear stress, as intestinal cells experience shear forces of 1–5 dynes/cm2 during digestion, attenuated to <1 dyne/cm2 by the microvilli barrier (Guo et al., 2000).

Both conditions were maintained at 37°C with 5% CO2 throughout the 21-day differentiation period. Media was changed every 48 h, with care taken to maintain identical handling procedures between static and dynamic conditions except for the orbital motion.

2.3 Characterization of the co-culture model

The development and functional differentiation of the co-culture model were assessed through multiple complementary approaches, including barrier function, molecular markers, morphological analysis, and imaging techniques.

2.3.1 Functional assessment

The co-culture model’s functional properties were evaluated by measuring barrier integrity, permeability, gene expression of differentiation markers, and mucin production.

2.3.1.1 Transepithelial electrical resistance (TEER) measurement

The integrity of the barrier in the Caco-2/HT29-MTX-E12 co-cultures was monitored by measuring transepithelial electrical resistance (TEER) every other day during the 21-day differentiation period. TEER measurements were performed under sterile conditions using a Millicell ERS-2 Voltohmmeter (Millipore, Burlington, MA, United States). The culture medium was replaced before each TEER measurement, and the culture was equilibrated for 30 min at 37°C and 5% CO2. Prior to each use, the electrode was sterilized with 70% ethanol and air-dried for 10 min. To maintain optimal cell conditions during measurement, the culture plate was placed on a temperature-controlled heating block (StableTemp Dry Block Heater, Cole-Parmer; Vernon Hills, IL, United States) set to 37°C. TEER values were corrected by subtracting the resistance of a cell-free Transwell insert, and final TEER values were normalized to the surface area of the insert. For each condition, three independent co-cultures were established, with each time point representing the average value across replicates.

2.3.1.2 Paracellular permeability assessment

Paracellular permeability was assessed by measuring the transepithelial transport of phenol red (phenolsulfonphthalein), which was selected as a low molecular weight marker (376 Da) allowing detection of changes in tight junction integrity and barrier formation (Smetanová et al., 2011). Following protocols adapted from Ferruzza et al. (2003) and Jiang et al. (2013), the co-cultures were rinsed with pre-warmed Receiving Buffer (RB), composed of Hank’s Balanced Salt Solution (HBSS; Gibco) supplemented with 11 mM glucose and 25 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) pH 7.4 (Lonza, Basel, Switzerland). The apical (AP) compartment received 0.5 mL of Donor Buffer (DB: HBSS containing 1 mM phenol red, 11 mM glucose, and 25 mM HEPES, pH 7.4), while 1.5 mL of RB was added to the basolateral (BL) compartment. Samples (100 µL) were collected from the BL compartment at 20-minute intervals over 3 h, with equivalent volumes of fresh RB added to maintain constant volume. The phenol red concentration in BL samples was quantified spectrophotometrically at 479 nm. Transepithelial flux was expressed as the apparent permeability coefficient (Papp), calculated using the equation from Jiang et al. (2013):

where K represents the steady-state rate of change in phenol red concentration over time in the BL compartment (s⁻1), Vr is the volume of the receiver chamber (1.5 mL), and A is the surface area of the membrane (1.12 cm2). The Area Under the Curve (AUC) was calculated for each differentiation time point to quantify and represent the co-culture permeability over time.

2.3.1.3 Gene expression analysis of intestinal differentiation markers

RNA was isolated from co-cultures at four time points (T0, T1, T2, and T3; days 0, 7, 14, and 21, respectively). Cells were washed twice with 1 mL warm Dulbecco’s Phosphate-Buffered Saline (DPBS), scraped in 1 mL cold DPBS, and collected by centrifugation at 250 × g for 5 min at 4°C. Cell pellets were lysed in 300 µL TRI Reagent (Zymo Research; Orange, CA, United States) and stored at −80°C until processing. Total RNA was extracted using the Direct-zol RNA MiniPrep kit (Zymo Research) following manufacturer’s protocol with modifications. Cell lysis was enhanced using a Bioruptor sonicator (Diagenode; Denville, NJ, United States) for three cycles (30 s ON/30 s OFF) at 4°C with high power settings. Lysates were centrifuged at 845 × g for 5 min at room temperature. RNA quality and quantity were assessed using a NanoDrop ND-2000 spectrophotometer (Thermo Fisher Scientific; Wilmington, DE, United States), with acceptable quality indicated by 260/280 ratios of approximately 2.0–2.1 and 260/230 ratios between 2.0 and 2.2.

cDNA synthesis was performed using 2 µg total RNA with the High-Capacity RNA-to-cDNA Kit (Applied Biosystems; Foster City, CA, United States). Quantitative PCR (qPCR) was conducted using a CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories) with TaqMan Fast Advanced Master Mix (Applied Biosystems). The thermal cycler conditions were as follows: uracil-N glycosylase incubation at 50°C for 2 min and polymerase activation at 95°C for 20 s, followed by 40 cycles of denaturation at 95°C for 3 s and annealing/extension at 60°C for 30 s. The target genes were selected to assess key intestinal functions, including those encoding tight junction proteins [CDH1 (cadherin-1) and TJP1 (zonula occludens-1)], intestinal brush border enzymes [ALPI (intestinal alkaline phosphatase), DPP4 (dipeptidyl peptidase-4), SI (sucrase-isomaltase)], and mucin glycoproteins [MUC2 (mucin-2) and MUC5AC (mucin-5AC)]. All TaqMan probe sets used in this study were purchased from Applied Biosystems (Supplementary Table S1). Among four candidate reference genes (ACTB, actin beta; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; PPIA, peptidylprolyl isomerase A; RPLP0, ribosomal protein lateral stalk subunit P0), GAPDH and RPLP0 genes were identified as the most stable reference genes, with M values <0.5, and used to normalize the expression of the target genes using the ΔCt method. Target gene expression was normalized to the geometric mean of these two reference genes (Vandesompele et al., 2002). Samples were analyzed in duplicate, with replicates showing cycle threshold (Ct) differences greater than 0.25 being reanalyzed. Data were processed using Bio-Rad CFX Maestro 2.3 Version 5.3.

2.3.1.4 Mucin quantification

Intracellular mucins were quantified using a modified periodic acid-Schiff (PAS) assay based on the methods of Mantle and Allen (1978), with modifications from Yamabayashi (1987) and Miner-Williams et al. (2009). The PAS method was selected for its specificity in quantifying glycoprotein content through reaction with 1,2-glycol groups (Dharmani et al., 2009; Harrop et al., 2012), allowing reliable detection even at low concentrations (Kilcoyne et al., 2011). Briefly, total soluble proteins were extracted from cell cultures using RIPA buffer, following manufacturer’s instructions. The samples were centrifuged at 2,000 × g for 10 min, and the supernatant was diluted 1:5 in DPBS. A standard curve was prepared using porcine stomach mucin (10–1,500 μg/mL). Both samples and standards underwent oxidation with periodic acid at 37°C for 2 h, followed by staining with Schiff’s reagent for 30 min at room temperature. Absorbance measurements were performed at 550 nm using an Infinite M200 microplate reader (Tecan; Männeford, Switzerland). Mucin concentrations were determined based on the standard curve.

2.3.2 Morphological assessment

To visualize the morphology of the co-cultures, the presence of the mucus layer, and formation of domes, Caco-2/HT29-MTX-E12 cells grown on the Transwell permeable supports under static and dynamic conditions were examined at post-confluence by confocal microscopy at days 0, 7, 14, and 21 (T0, T1, T2, T3, respectively).

The co-cultures were incubated with fluorescent wheat germ agglutinin (WGA Oregon Green® 488 Conjugate; Ex 496 nm, Em 524 nm) (Thermo Fisher Scientific) at a concentration of 5 μg/mL for 20 min in cell culture conditions (37°C and 5% CO2). WGA specifically binds sialic acid and N-acetylglucosaminyl residues, which are predominant in mucins (Kilcoyne et al., 2011), thus used for mucus layer visualization. Then, co-cultures were fixed with 4% formaldehyde (in DPBS) for 15 min at 37°C and then washed multiple times with DPBS. To stain nuclei, fixed cells were incubated with TO-PRO-3 iodide (Ex 642 nm, Em 661 nm) (Thermo Fisher Scientific) at a 1:2000 dilution in DPBS for 10 min at room temperature. After staining, the co-cultures were washed multiple times with DPBS. For microscopy preparation, the Transwell membranes were carefully excised from their plastic inserts using a scalpel by inserting the blade from the underside of the membrane and cutting around the perimeter, following an adaptation of protocol by Hiebl et al. (2020). The membranes were then gently lifted using fine forceps, gripping only the circular edges to preserve the cell layer integrity. The excised membranes were mounted on glass slides using anti-fade medium [2.5% 1,4-diazabicyclo [2.2.2] octane (DABCO), 50 mM Tris, pH 8, and 90% glycerol], taking special care to avoid air bubbles and maintain proper orientation for subsequent microscopic analysis.

2.3.2.1 Confocal microscopy analysis

Images were acquired with a Nikon A1R + HD25 confocal laser scanning microscope with a 60 × oil immersion objective (NA 1.4). To capture a large area at high resolution, we employed the Large Image function of the Nikon software (Nis-Elements AR 5.20). This function created composite images by automatically assembling 6 × 6 individual image fields into larger mosaic images. The tile scanning was performed across multiple z-planes to generate comprehensive 3D mosaics of adjacent image stacks. Z-stack images were acquired (Schindelin et al., 2012), with three areas of ∼1 mm2 imaged from the central region of the Transwell membrane for each time point sample (T0, T1, T2, T3) and for each of the two treatments (static and dynamic condition). Images were acquired at 2 µm intervals, starting from the Transwell membrane on which the cells were grown to the top of the cell multi-layer, and numbered in ascending order. The appearance of the mucus layer that covers the cultures was indeed taken as an indication of cell top reaching. In this way, each image corresponded along z to the image with the same number in other samples (e.g., image #4 of all the samples had the same height, that is at ∼8 µm of height, from the membrane surface). The confocal microscopy imaging allowed for the analyses of the overall morphology of the co-cultures, including cell distribution and dome formation. The continuity of the mucus layer was evaluated by examining the WGA staining in the z-stack images.

2.3.2.2 Image processing and analysis

Post-acquisition image processing and dome formation analysis were conducted using ImageJ Fiji (v1.54f). Confocal image stacks were imported using the Bio-Formats plugin in hyperstack format, which enabled accurate handling of metadata and preserved multidimensional data integrity. For three-dimensional visualization, surface topography was reconstructed using the “3D Surface Plot” plugin, with parameters optimized for dome structure analysis. These included a grid size of 128 to achieve optimal spatial resolution, “Filled Gradient” appearance settings to enhance surface continuity visualization, and a Z-scale factor of 0.1 to maintain proportional spatial representation.

For detailed dome formation analysis, each section from the stacked images was exported individually as a composite file containing merged channels. To ensure data quality, sections containing glass slide grid artifacts were excluded from the analysis. Areas with weak green signal representing cellular presence were selectively enhanced in Adobe Photoshop 2024 (Adobe Inc., San Jose, CA, United States) to improve visualization clarity while maintaining signal fidelity.

2.3.2.3 Quantitative and qualitative assessment of domes

Dome formations in 3D reconstructed images were detected and quantified using Python with the OpenCV library, along with additional image processing packages (scikit-image, SciPy, and Matplotlib). Each image was analyzed from four different angle views to enhance detection accuracy. The images were first normalized to a 0–1 intensity range, followed by conversion to grayscale and adaptive segmentation to isolate dome-like structures. Each detected region was further characterized by measuring its area and height, though these detailed measurements were not included in the final analysis (data not shown).

Each composite image was analyzed using Python’s Open-Source Computer Vision Library (OpenCV). Specifically, after binarizing each image, contiguous cell-covered objects were identified and characterized by two parameters: area extension (measured in pixels) and eccentricity (ranging from 0 to 1). The eccentricity parameter reflects the shape’s elongation, calculated as the ratio of the distance between the foci of an ellipse (fit to each contour) to its major axis length, with values closer to 0 indicating circular shapes and values near 1 indicating highly elongated shapes.

2.4 Assessment of co-culture model with food components

To evaluate the enhanced co-culture model’s utility for food safety assessment, we first evaluated its biocompatibility with digested skimmed milk powder (SMP) (SaporePuro; Gioia Group, Torino, Italy), as test food matrix, followed by investigating cellular responses to SMP supplemented with TiO₂ nanoparticles.

2.4.1 Digestion and supplementation of SMP with TiO₂ nanoparticles2.4.1.1 Semi-dynamic in vitro digestion of SMP

A semi-dynamic in vitro digestion protocol was performed on 10% w/v SMP solution (10 g SMP dissolved in 100 mL distilled water) following the standardized INFOGEST method (Mulet-Cabero et al., 2020) with bio-compatibility modifications. The process began with an oral phase where 30 mL of SMP solution was combined with 10 mL simulated salivary fluid (SSF) without amylase (since the food matrix contained no starch; Supplementary Table S2) and maintained at 37°C with constant stirring for 2 min. For the gastric phase, the oral bolus was mixed with simulated gastric fluid (SGF) containing porcine pepsin (2,000 U/mL, Sigma-Aldrich). To simulate physiological gastric emptying, an automated titrator (Metrohm, Herisau, Switzerland) gradually reduced the pH to 3.0 over 60 min, with emptying events occurring every 20 min. The intestinal phase commenced with the addition of simulated intestinal fluid (SIF) containing porcine pancreatin (100 U/mL, Sigma-Aldrich), maintaining pH 7.0 for 120 min at 37°C under continuous stirring in a climate-controlled hood (Model 810; ASAL s.r.l., Cernusco sul Naviglio, Italy) equipped with an orbital shaker (Model 709; ASAL s.r.l.). Bile was omitted from the intestinal phase since previous studies showed they are a major source of cytotoxicity in cell experiments (Vieira et al., 2022). Additionally, the SMP used contained relatively low lipid content (Supplementary Table S2), reducing its need in the digestion process (Atallah et al., 2020). Three independent digestions were performed. After pooling the samples, they were immediately frozen in liquid nitrogen to halt enzymatic activity (Kondrashina et al., 2024). After rapid thawing, the samples were processed to obtain the bioaccessible fraction. First, to obtain a clear supernatant, samples were centrifuged (10,000 × g, 30 min, 4°C) and filtered through a 0.45 μm polyethersulfone (PES) membrane (Millipore). Then, the filtrate was ultrafiltered using a 3 kDa Vivaspin® 20 centrifugal filter (Sartorius AG, Göttingen, Germany) to remove larger proteins and enzymes (Giromini et al., 2019), and finally sterilized through a 0.22 μm PES membrane (Millipore) (Segeritz and Vallier, 2017).

2.4.1.2 TiO₂ solubilization in digested SMP

TiO₂ effect on epithelial cells was investigated by incorporating anatase TiO2 nanoparticles (−325 mesh, catalog #248576) into the digested SMP bioaccessible fraction (Faria et al., 2020). Following a modified Nanogenotox dispersion protocol (Jensen et al., 2011), TiO₂ was pre-wetted with 0.05% ethanol, centrifuged (3,000 × g, 1 min), and the pellet was resuspended in digested SMP to achieve 0.25% TiO₂ concentration. This concentration accounts for the 1:8 dilution factor inherent to the digestion protocol while maintaining equivalence to the FDA-permitted maximum of 1% w/v TiO₂ in the original food matrix (FDA, 2024). The suspension underwent 32 sonication cycles (30 s each) followed by 0.22 μm PES membrane filtration to ensure complete dispersion and sterility.

2.4.1.3 Titanium quantification

Titanium content in digesta samples was determined using Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES, Spectro Arcos-Ametek, Kleve, Germany). The instrument was equipped with an axial torch and high salinity kit. Analysis was performed in duplicate with 12-second measurement runs, each preceded by a 60-second stabilization period. Quantification was achieved using a calibration curve constructed with certified aqueous titanium standards, with a limit of detection (LOD) of 0.33 ppb.

2.4.2 Impact of TiO₂-supplemented SMP digesta on co-culture model

The potential cytotoxicity and oxidative stress effects of TiO₂-supplemented digesta were evaluated through multiple complementary assays examining cell viability, membrane integrity, and redox status markers in the co-culture model. Results in the treated cells were compared to control cells receiving only serum- and phenol red-free DMEM and otherwise handled identically to treated cells.

2.4.2.1 Cell viability assessment

The bioaccessible fraction of digested SMP (with and without TiO₂) was prepared for cell exposure by dilution in serum- and phenol red-free DMEM. Non-spiked digesta was diluted at 1:3, 1:10, and 1:20 (v/v) ratios, while TiO₂-spiked digesta was diluted at 1:3 and 1:10 based on results reported in section 3.3. Cell viability was assessed using PrestoBlue Cell Viability Reagent (PB, Thermo Fisher Scientific, 1:10 dilution). Differentiated co-cultures (21 days, dynamic conditions) were exposed apically to 0.5 mL of diluted digesta for 3 h at 37°C, approximating physiological intestinal transit time (Hardy et al., 1989). Control cells received serum-free DMEM with 1:10 PB. Fluorescence measurements (excitation: 560 nm, emission: 590 nm) were performed using an Infinite F200 microplate reader (Tecan), with blank correction (PB in serum-free DMEM). Cell viability was expressed as the percentage of blank-corrected fluorescence of SMP digesta-exposed cells relative to blank-corrected fluorescence of control cells.

2.4.2.2 Membrane integrity assessment

Cell membrane integrity was evaluated by measuring lactate dehydrogenase (LDH) release from the cytoplasm into the culture medium (phenol red-free DMEM) using the CyQUANT LDH Cytotoxicity Assay Kit (Invitrogen). Absorbance was measured at 490 nm with background correction at 680 nm (Infinite M200, Tecan). LDH release was expressed as a percentage of maximum LDH activity (obtained from lysed control cells).

2.4.2.3 Evaluation of oxidative status and antioxidant defense markers

Multiple parameters were assessed to characterize cellular oxidative status. For lipid peroxidation assessment, thiobarbituric acid reactive substances (TBARS) levels were quantified in culture media (phenol red-free DMEM) following Potter et al. (2011). Media samples (500 µL) were combined with 400 µL of 15% trichloroacetic acid (TCA) and 800 µL of thiobarbituric acid (TBA) solution [0.67% TBA, 0.01% butylated hydroxytoluene (BHT)]. After heating (95°C, 20 min) and cooling, the organic phase was extracted with 3 mL butanol. The upper phase (200 µL) was analyzed for fluorescence, with TBARS expressed as malondialdehyde (MDA) equivalents using a standard curve (0.007–4 nmol/mL).

For glutathione status determination and ROS measurements, treated and control cells were lysed with RIPA buffer following manufacturer’s instructions to extract total soluble proteins, which were quantified using Quick Start™ Bradford Protein Assay (Bio-Rad Laboratories). For GSH and GSSG were measured using a modified Ellman’s method (Vuolo et al., 2022). Samples were deproteinated (3 volumes 5% TCA, 12,000 × g, 5 min, 4°C) and analyzed in a 96-well format. For GSH measurement, 50 µL sample or standard (0–500 nmol/mL) was combined with 50 µL Tris/EDTA buffer (1 mM Tris, 2 mM EDTA, pH 8.2) and 20 µL 5,5′-dithiobis (2-nitrobenzoic acid) (DTNB, 10 mmol/L). After dark incubation (15 min, room temperature), absorbance was measured at 412 nm. For GSSG measurement, deproteinated samples were first reduced with dithiothreitol (DTT, 10 mM, 15 min, room temperature).

For reactive oxygen species (ROS) production measurements, cell lysates (1:10 dilution) were combined with 100 µM 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) in DPBS according to Kim and Xue (2020). After dark incubation (37°C, 30 min), fluorescence was measured (excitation: 485 nm, emission: 530 nm) and normalized to total soluble protein content.

2.5 Statistical analysis

All statistical analyses were performed using GraphPad Prism version 10.3 (GraphPad Software, Inc., San Diego, CA, United States). Data are presented as mean ± standard deviation (SD). Statistical significance was set at p < 0.05. TEER measurements, paracellular permeability, and gene expression data were analyzed using one-way ANOVA followed by Tukey’s post hoc test. Mucin quantification and dome count data were evaluated using two-way ANOVA followed by Šidák’s multiple comparison test. Through Python’s OpenCV library, dome structure comparisons (coverage, eccentricity, and contiguous objects) between static and dynamic conditions were analyzed by non-parametric Mann-Whitney U test at each time point. Cell viability, LDH release, redox status markers, and ROS measurements were analyzed using one-way ANOVA followed by Tukey’s post hoc test.

3 Results3.1 Dynamic culture conditions promote mucin production without affecting epithelial differentiation

Barrier integrity development, monitored through TEER measurements, showed progressive enhancement under both static and dynamic conditions throughout the 21-day culture period. TEER values exhibited consistent increases, reaching approximately 1,100 Ω × cm2 by day 19 and stabilizing around 1,000 Ω × cm2 by day 21 (Figure 1A). This pattern demonstrates successful establishment of the epithelial barrier through tight junctions by day 19, regardless of mechanical stimulation. In contrast to TEER measurements, paracellular permeability analysis revealed distinct temporal dynamics. While permeability remained comparable between static and dynamic conditions during early culture periods (T0–T2), both conditions showed a significant increase in permeability at T3 (p < 0.05 vs. T0, T1, and T2; Figure 1B).

Figure 1. Development of barrier properties in Caco-2/HT29-MTX-E12 co-cultures under static and dynamic conditions during differentiation. (A) Transepithelial Electrical Resistance (TEER) measurements. Values were recorded at ten time points: 0 (T0), 3, 5, 7 (T1), 10, 12, 14 (T2), 17, 19, and 21 (T3) days post-confluence and expressed as Ω × cm2. Lines represent third-order polynomial fits to the experimental values. Data points represent mean ± SD (n = 23–36 from three independent experiments). Statistical significance was assessed using one-way ANOVA followed by Tukey’s post hoc test, with no significant differences between conditions. (B) Paracellular permeability assessed by phenol red transport, expressed as area under the curve (AUC) of transepithelial flux. Measurements were taken at four time points: immediately after reaching confluence (T0), and at 7 (T1), 14 (T2), and 21 (T3) days post-confluence. Data shown as mean ± SD (n = 4 from two independent experiments). Statistical significance was assessed using one-way ANOVA followed by Tukey’s post hoc test. No significant differences were observed between static and dynamic conditions at each time point. Both conditions showed significantly increased permeability at T3 compared to earlier time points (T0, T1, and T2; p < 0.05).

Assessment of key intestinal differentiation markers revealed distinct temporal expression profiles between barrier-associated and secretory genes. Expression analysis of genes related to epithelial integrity and enterocyte function (CDH1, TJP1) (Figures 2A, B) and brush border enzymes (ALPI, DPP4, SI) (Figures 2C–E) demonstrated similar upward trends between static and dynamic conditions. The comparable expression levels of these markers establish that mechanical stimulation did not significantly alter the differentiation program of intestinal epithelial cells. However, mechanical stimulation markedly influenced goblet cell-specific genes. By day 21, cells cultured under dynamic conditions exhibited significant upregulation of both mucin genes, MUC2 and MUC5AC, compared to static cultures (Figures 2F, G).

Figure 2. Expression profiles of key intestinal differentiation markers in Caco-2/HT29-MTX-E12 co-cultures under static and dynamic conditions. (A–B) Expression of genes encoding tight junction proteins: cadherin-1 (CDH1) and tight junction protein-1 (TJP1). (C–E) Expression of genes encoding brush border enzymes: intestinal alkaline phosphatase (ALPI), dipeptidyl peptidase-4 (DPP4), and sucrase-isomaltase (SI). (F–G) Expression of genes encoding mucins: mucin-2 (MUC2) and mucin-5AC (MUC5AC). Gene expression was analyzed at four time points: immediately after confluence (T0), and at 7 (T1), 14 (T2), and 21 (T3) days post-confluence. Data are expressed as relative expression values normalized to the geometric mean of GAPDH and RPLP0 reference genes. Values represent mean ± SD from three biological replicates. Statistical significance between static and dynamic conditions at each time point was assessed using one-way ANOVA followed by Tukey’s post hoc test (*p < 0.05, ***p < 0.001).

To confirm these transcriptional changes at the protein level, we quantified in cellulo mucin production throughout the differentiation period. While cytosolic mucin levels were initially comparable between conditions at day 7 (T1), dynamic cultures showed significantly higher concentrations by day 14 (T2), displaying more than two-fold increase compared to static conditions (Figure 3). Interestingly, this difference was not maintained at T3, where mucin levels were similar between conditions. This enhanced intracellular mucin content at T2 aligns with the gene expression data, though with a temporal offset.

Figure 3. Quantification of intracellular mucins in Caco-2/HT29-MTX-E12 co-cultures under static and dynamic conditions. Intracellular mucin content, expressed as μg/mL, was measured using periodic acid-Schiff (PAS) assay at four time points: immediately after confluence (T0), and at 7 (T1), 14 (T2), and 21 (T3) days post-confluence. Values represent mean ± SD from four biological replicates across two independent experiments (n = 4). Statistical significance was assessed using two-way ANOVA followed by Šidák’s multiple comparison test, with significant differences between static and dynamic conditions at T2 (*p < 0.05).

3.2 Dynamic culture conditions enhance three-dimensional epithelial organization

Confocal microscopy analysis characterized the three-dimensional organization of the epithelial layer through sequential optical sections and 3D surface plot reconstructions (Figure 4). Images were acquired at 2 µm intervals from the Transwell membrane to the top of the cell multilayer under both static (Figures 4A–C) and dynamic (Figures 4E–G) conditions. WGA staining (green) labeled the mucus layer, while TO-PRO-3 (blue) marked cell nuclei, enabling visualization of the spatial distribution of cellular and mucus components, as previously described (García-Rodríguez et al., 2018). The WGA-stained mucus layer outlined the outer culture layer, facilitating accurate three-dimensional reconstruction of the epithelial architecture, as shown in the surface plots (Figures 4D, H).

Figure 4. Evaluation of the epithelial organization of Caco-2/HT29-MTX-E12 co-culture using confocal microscopy at T2 as representative time point. (A–C) Representative optical sections along the z-axis under static conditions, acquired at increasing heights (6, 12, 18 μm) from the Transwell membrane. (D) 3D surface reconstruction of the complete z-stack from static condition. (E–G) Representative optical sections along the z-axis under dynamic conditions, acquired at increasing heights (6, 12, 18 μm) from the Transwell membrane. (H) 3D surface reconstruction of the complete z-stack from dynamic condition. Nuclei were stained with TO-PRO-3 (blue) and mucus layer was visualized using WGA (green). Images were acquired at day 14 (T2) post-confluence. Scale bar = 100 μm.

Building on this imaging approach, detailed analysis revealed distinct patterns in the three-dimensional organization of the epithelial layer between static and dynamic conditions. Initial 3D surface plot reconstructions showed similar dome formation patterns across both conditions from T0 to T3 (Supplementary Figure S1), with no significant differences in dome counts (Figure 5). This led us to conduct a more detailed analysis using sequential optical sections (z-stacks) to examine three key parameters: cell coverage density, multicellular organization (through contiguous cellular structures), and geometric arrangement (via eccentricity measurements).

Figure 5. Quantification of dome structures in Caco-2/HT29-MTX-E12 co-cultures under static and dynamic conditions. Dome counts, expressed as average number of domes per angle view, were performed at four time points: immediately after confluence (T0), and at 7 (T1), 14 (T2), and 21 (T3) days post-confluence using 3D surface reconstructions from confocal z-stacks. Values represent mean ± SD from three technical replicates. Statistical significance was assessed using two-way ANOVA followed by Šidák’s multiple comparison test, with no significant differences between static and dynamic conditions across all time points.

The z-stack analysis revealed progressive differences in epithelial organization between conditions (Figure 6). At T1, both conditions achieved similar initial monolayer formation, with 75%–90% surface coverage through the first 14 µm from the insert. However, by T2, distinct organizational patterns emerged. Dynamic cultures maintained approximately 80% cell coverage up to 14 µm height, with gradual decrease in higher sections, while static cultures showed steeper coverage reduction, reaching only around 50% at comparable heights. These differences became more pronounced at T3, where static cultures displayed a steeper decline in coverage in higher sections, falling below 20%.

Figure 6. Quantitative analysis of epithelial coverage across optical sections in Caco-2/HT29-MTX-E12 co-cultures under static and dynamic conditions. Cell coverage was analyzed across sequential optical sections (2 μm intervals, sections 0–27 corresponding to heights 0–54 μm) at three time points: 7 (T1), 14 (T2), and 21 (T3) days post-confluence. Individual measurements from three technical replicates are shown as thin lines (red for static, blue for dynamic conditions). Dashed lines with markers represent full averages (circles for static, squares for dynamic), while partial averages (due to different cell scattering in optical sections) are shown with reduced opacity. Statistical significance between static and dynamic conditions was assessed using Mann-Whitney U test (*p < 0.05, **p < 0.01 for multiple comparisons), with significant differences (p < 0.01 for overall test across sections) observed at T2.

The formation of three-dimensional epithelial structures also showed temporal evolution (Supplementary Figure S2). At T1, both conditions displayed minimal organization with only 2–3 contiguous structures present in the lower cell layers. By T2, dynamic conditions supported a more organized and cohesive formation of cellular structures in the intermediate layers, with approximately 10 objects. In contrast, static cultures showed significantly higher counts but with greater variability across sections. At T3, dynamic cultures established a stable epithelial architecture with 15–20 organized structures across the upper layers, whereas static cultures, although achieving similar counts, exhibited less consistent organization in the upper regions (Supplementary Figure S3A).

Geometric analysis through eccentricity measurements revealed distinct temporal patterns (Supplementary Figure S3B). Initial structures showed low eccentricity values (0.2–0.4) in both conditions. At T2, eccentricity values in the lower sections under dynamic conditions were lower than those observed in static cultures, indicating a more rounded cell structure at the base. This difference diminished in the upper sections, where both static and dynamic conditions converged around higher eccentricity values (0.7–0.8). By T3, eccentricity profiles across sections for both conditions became more similar, though static cultures exhibited more variable values throughout the culture period.

3.3 Digested SMP maintains cell viability across multiple exposure ratios

Since the dynamic growth model demonstrated comparable or superior results to the static model, subsequent experimental supplementation with digested SMP with TiO₂ were conducted exclusively under dynamic conditions. To optimize the exposure protocol, it was first necessary to determine the optimal exposure ratio for digested SMP that would maintain cell viability while maximizing exposure to the digesta. The results shown in Supplementary Figure S4 illustrate the relative cell viability of the co-culture exposed to digested SMP at three dilution ratios: 1:3, 1:10, and 1:20. In comparison to the control cells, no significant reduction in cell viability was observed at any dilution, with viability remaining above 90% across all conditions. Based on these results, the 1:3 and 1:10 dilutions were selected for subsequent experiments supplementation of the co-culture dynamic model with digested SMP with TiO₂.

3.4 TiO₂-supplemented digesta induces concentration-dependent oxidative stress

ICP-OES analysis of the SMP digesta revealed a Ti concentration of 1.25 μg/mL. This concentration matched the expected theoretical value for 1% w/v TiO2-supplemented SMP, accounting for the eight-fold dilution that occurs during the semi-dynamic in vitro digestion process. To investigate the biological impact of TiO2 at this concentration, the digested samples were tested on the enhanced co-culture model at selected dilution ratios (1:3 and 1:10), evaluating cell viability, membrane integrity, and oxidative status biomarkers.

The results show that cell viability, reported in Supplementary Figure S5A, decreased significantly at the 1:3 dilution compared to the control, while no significant reduction was observed at the 1:10 dilution. Interestingly, LDH assay (Supplementary Figure S5B), which measures cell membrane damage, showed no significant differences between the control and treated cells at either concentration, establishing that the reduced viability at 1:3 dilution was not due to direct membrane damage.

To further characterize the mechanism of viability loss, we evaluated the oxidative status of the supplemented cell co-cultures. TBARS levels in the media, an indicator of lipid peroxidation, were significantly elevated at the 1:3 dilution (Supplementary Figure S5C), indicating increased oxidative damage, while the 1:10 dilution showed no significant change compared to the control. The 1:3 dilution also showed higher, though not statistically significant, ROS accumulation (Supplementary Figure S5D).

The assessment of antioxidant defenses revealed that glutathione (GSH) levels (Supplementary Figure S5E) decreased significantly at both concentrations, indicating a depletion of cellular antioxidant capacity. Moreover, the ratio of reduced (GSH) to oxidized (GSSG) glutathione significantly declined at the 1:3 dilution (Supplementary Figure S5F), further supporting the induction of oxidative stress under these conditions.

4 Discussion4.1 Development of enhanced Caco-2/HT29-MTX-E12 co-culture

Our investigation revealed distinct patterns of epithelial development between static and dynamic culture conditions, with mechanical stimulation selectively enhancing specific aspects of cellular differentiation and organization. The most interesting differences we observed concerned dome formation at T2, which occurs through active transepithelial fluid transport and fluid entrapment between cells and the substrate (Lever, 1985). These structures, recognized as one of the hallmarks of advanced epithelial differentiation (Rotoli et al., 2002), were particularly prominent under dynamic conditions and maintained strong apical-basal polarity essential for nutrient transport and barrier function (Zweibaum et al., 2011).

When quantifying dome formation using 3D surface plots, we observed no significant differences between conditions, with dome numbers stabilizing around T3. This outcome contrasted with earlier observations by Hara et al. (1993), who reported a gradual increase in dome numbers plateauing around day 15. This discrepancy likely stems from methodological differences - our use of confocal microscopy for 3D analysis versus the bright-field imaging employed in earlier studies (Matsumoto et al., 1990; Hauck and Stanners, 1991; Herold et al., 1994), which could affect dome detection and quantification. Furthermore, the field lacks a standardized definition of what constitutes a “dome,” contributing to variability in reporting and cross-study comparisons.

To address these methodological limitations, we developed a comprehensive analysis approach using confocal z-stacks that integrated cell coverage, eccentricity, and continuity parameters. This method revealed that dynamic conditions enhanced differentiation within a shorter timeframe, with distinct patterns in vertical organization. The mechanical forces specifically promoted cellular polarization processes, resulting in superior structural integrity and organization. Dynamic cultures progressed from early developmental stages reminiscent of initial epithelial formation (Fantini et al., 1986) to structures characteristic of differentiated intestinal epithelial monolayers (Ferraretto et al., 2018), maintaining stable organization with contiguous cell clusters.

Previous research established that domes typically appear around day 6 post-confluency, peak by day 15, and undergo fusion events that expand to cover larger portions of the monolayer (Pinto et al., 1983; Scaglione-Sewell et al., 1998). Our observations extended these findings, showing that mechanical forces not only accelerated dome formation but also enhanced their structural integration, promoting the fusion of smaller, circular domes into larger, more cohesive formations compared to static conditions.

The barrier function development, assessed through TEER measurements, showed an interesting pattern with no significant differences between static and dynamic conditions. This finding contrasts with previous studies where fluid shear stress upregulated tight junction proteins in both Caco-2 cells (Delon et al., 2019) and lung epithelial cells exposed to stretch-induced forces (Cavanaugh et al., 2001). The discrepancy between our co-culture results and previous monoculture studies demonstrates that the presence of mucus-producing HT29-MTX-E12 cells may modulate mechano-transduction pathways affecting tight junction assembly. The interaction between enterocytes and goblet cells likely introduces additional complexity to the mechanical regulation of barrier formation, potentially through paracrine signaling or altered mechanical force distribution across the heterogeneous epithelial layer.

An unexpected discrepancy emerged between barrier integrity measurements: while TEER values remained stable, indicating consistent ionic barrier function, paracellular permeability increased at T3 in both conditions, revealing a decline in size-selective barrier properties by the end of the time-course. This pattern mirrors in vivo observations of age-related increases in epithelial permeability, where tight junction function deteriorates over time (Salazar et al., 2023). The divergence between stable TEER values and increased permeability provides evidence of a distinct regulation of different transport pathways. While TEER primarily reflects paracellular resistance through tight junctions (Pongkorpsakol et al., 2021), permeability changes may reflect alterations in specific transport mechanisms or selective barrier properties (Horowitz et al., 2023). This differential response could result from two distinct mechanisms: the development of transcellular transport pathways and the selective modulation of specific tight junction proteins. The latter can regulate size-dependent permeability without affecting overall electrical resistance, as previously demonstrated for tight junction dynamics (Tervonen et al., 2019).

The mechanical stimulation’s biological impact extended to cellular differentiation pathways, particularly affecting mucin production by goblet cells. While epithelial differentiation markers showed comparable expression levels between conditions, mucin-related genes displayed significant upregulation at T3 under dynamic conditions. This selective enhancement aligns with previous studies demonstrating positive effects of shear stress on goblet cell function (Reuter and Oelschlaeger, 2018; Lindner et al., 2021; Xu et al., 2021). The temporal dynamics showed distinct patterns between protein and gene expression levels. Increased intracellular mucin content was detected at T2, while peak mucin gene expression occurred later at T3. These observations reflect the complex dynamics of mucin synthesis, processing and secretion, where mucins undergo elaborate biosynthetic pathways including initial synthesis, dimerization, and extensive glycosylation (Asker et al., 1998; Thornton et al., 2008). The early accumulation of mucin content represents a regulatory response to mechanical forces, where initial mucin storage occurs independently of transcrip