Alignment of in vitro and in vivo pulmonary inflammation models using crystalline quartz silica

Abstract

Background

Systematic in vitro–in vivo comparisons are increasingly used to assess the relevance and predictivity of in vitro lung models for inhalation toxicology and regulatory risk assessment. Here, we compared inflammatory endpoints across established in vitro and in vivo pulmonary models following exposure to crystalline quartz silica particles (DQ₁₂). To better align exposure timelines, in vitro responses assessed at 24 h were extended to 7 days, matching the post-exposure period recommended in OECD inhalation testing guidelines for animal testing. To test its potential and limitations, we utilized a harmonized in vitro co-culture model consisting of the human bronchial cell line Calu-3 and human monocyte-derived macrophages, which were exposed to DQ₁₂ particles.

Results

No increased cytotoxicity or impairment of barrier integrity, as assessed by transepithelial electrical resistance (TEER) and tight junction protein 1 (TJP1) gene expression, was observed 7 days after exposure in vitro, in contrast to clear tissue damage detected in vivo. However, we observed increased release of interleukin (IL)-6 and IL-8, measured at both protein and gene levels. Gene expressions of IL-1β, IL-6, and IL-8 showed positive correlations between the in vitro and in vivo models.

Conclusions

By extending exposure duration and aligning time points, this study identified inflammatory biomarkers that correlate between an in vitro lung model and in vivo data. These findings demonstrate the value of refined in vitro models for assessing particle-induced lung inflammation and support their relevance for hazard assessment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12989-026-00665-w.

Background

The urgency for human-relevant in vitro lung models in inhalation toxicology is driven by the need for accurate hazard and risk assessment while reducing reliance on animal testing. Airborne particulate matter, including pollutants, chemical dust, and metal particles [1, 2], is a major contributor to chronic pulmonary diseases, such as silicosis, fibrosis, and even lung cancer [3, 4]. Among these are crystalline silica particles (DQ₁₂). They have been widely used as well-characterized reference quartz material in inhalation toxicology. Moreover, silica-based particles are frequently used as model materials because silica, as a particle class, spans a wide range of sizes, surface chemistries, and reactivities, allowing systematic investigation of how specific physicochemical properties influence biological responses [5]. Traditional in vivo models, particularly rodents, have long been the standard for evaluating the safety of inhaled substances. The Organization for Economic Co-operation and Development (OECD) has established several test guidelines (TGs) for inhalation toxicity studies, each regarding different exposure durations to assess various toxicological endpoints. For instance, TG 403 [6] addresses acute inhalation toxicity with single exposures, TG 412 [7] focuses on subacute (28-day) studies, and TG 413 to sub-chronic (90-day) toxicity assessments. These guidelines typically recommend using male and female rodents, often rats, to evaluate potential sex-related differences in response. The number of animals per group is generally set at a minimum of 10 per sex to ensure statistical robustness. In TG 413 [8], animals are exposed to the test substance for 6 h per day, 5 days a week, over 90 days, to observe potential adverse effects from prolonged exposure. However, key interspecies differences in lung structure and function, such as airway branching complexity, epithelial cell composition, and breathing patterns, make rodent data only partially predictive for human outcomes [9]. Moreover, ethical concerns, high costs, and the limited translatability of animal data to humans have accelerated the move toward in vitro alternatives using human cell cultures.

In this context, New Approach Methodologies (NAMs), such as organotypic airway models and computational dosimetry, have emerged as key tools for providing human-relevant data and supporting regulatory decision-making in inhalation toxicology [10]. Regulatory acceptance of in vitro lung models depends on four essential qualities: physiological and biological relevance, reliability, predictive capacity, and relevant endpoints [11]. While such models are rapidly being developed and refined, remaining challenges include ensuring these criteria are met and building confidence in their predictive capacity [10, 12]. Several systematic comparisons have shown that advanced air–liquid interface (ALI) and three-dimensional airway models reproduce in vivo outcomes more accurately than simple submerged cultures, particularly for endpoints such as cytokine release, transepithelial electrical resistance (TEER), oxidative stress, and transcriptomics [13–16]. These studies emphasize the importance of model complexity, dose normalization, and endpoint selection for strengthening in vitro–in vivo concordance. However, such direct comparisons are not always required, as regulatory case studies already demonstrate that NAMs can inform decision-making without new animal testing [17, 18].

Inadequate exposure regimes and dose metrics are common limitations of in vitro models for observing significant changes when investigating the risk of inhaled toxicants [19]. This underscores the importance of studies that better align in vitro and in vivo experimental designs by selecting appropriate exposure scenarios, dose considerations, endpoints, and biomarkers to strengthen the predictive power of in vitro lung models. Several recent regulatory case studies demonstrate how these limitations can be overcome through NAMs and in vivo-in vitro extrapolation (IVIVE)-based approaches. For example, the U.S. EPA applied an organotypic 3D airway model, combined with benchmark dose modeling, to refine inhalation risk assessment for chlorothalonil, thereby reducing reliance on a 90-day rat study [20]. This approach was formalized in the OECD Integrated Approaches to Testing and Assessment (IATA) Case Study No. 367 [21], providing a structured framework for integrating in vitro data and IVIVE into decision-making. Similarly, the Scientific Committee on Consumer Safety (SCCS) used the MucilAir™ airway model to evaluate the inhalation safety of acetylated vetiver oil in sprayable cosmetics [22]. Although these examples involve vapors or aerosols rather than particles, they demonstrate how NAMs, supported by IVIVE, can inform regulatory endpoints and reduce animal testing. Importantly, such approaches also help to identify the endpoints that are most relevant for decision-making. These are often those that are anchored in Adverse Outcome Pathways (AOPs) and linked to early key events. Building on these developments, our study applies similar principles to particulate exposures, using human-relevant in vitro models and AOP anchoring to enhance the translational value of in vitro findings for regulatory contexts.

Achieving reliable IVIVE requires a systematic approach, including (i) selecting a relevant Adverse Outcome Pathway (AOP) to guide endpoint identification [23], (ii) utilizing a robust, reliable, and predictive in vitro model [24], (iii) aligning dose setting and -metrics with real-world exposures, (iv) identifying biomarkers that facilitate meaningful comparisons, and (v) adopting new correlation methodologies [25]. This framework is essential to advance the regulatory acceptance of in vitro models and to improve their use in risk assessment.

This study focused on the AOP173 [26], which outlines the progression from initial exposure caused by respiratory inflammation to lung fibrosis. It consists of six key events (KEs) following the molecular initiating event (MIE 1495), and each KE is well described in the literature [27]. Briefly, the interaction between inhaled substances and the lung cell membrane components, i.e., the MIE, is followed by a cascade of biological responses. The danger signals, or alarmins, are released prior to the secretion of pro-inflammatory and pro-fibrotic mediators (KE 1496). These mediators recruit pro-inflammatory leukocytes into the lung (KE 1497). Persistent inflammation and resulting tissue damage cause a loss of alveolar-capillary membrane integrity (KE 1498) and activate T helper type 2 (Th2) signaling (KE 1499). This response drives fibroblast proliferation and differentiation into myofibroblasts (KE 1500), ultimately leading to excessive collagen deposition and other extracellular matrix components (KE 68). The resulting structural changes, such as thickened alveolar walls and reduced lung function, lead to pulmonary fibrosis (AO 1458). Central to the pathway described is the persistent or non-resolving inflammation, which drives tissue injury and ultimately leads to fibrosis in the presence of continuous stimuli. Therefore, we decided to focus on the early KEs in this pathway regarding inflammation, including increased pro-inflammatory markers (KE 1496), recruitment of inflammatory cells (KE 1497), and loss of membrane integrity (KE 1498). Since DQ₁₂ particles are known to induce inflammation-driven fibrosis in humans [28] and rats, these KEs can be measured across both in vitro and in vivo systems to establish meaningful correlations. Our in vitro model employs the Calu-3 cell line and is a well-established model for the human bronchial epithelium that has been widely used for in vitro studies [12, 29, 30], particularly in inhalation toxicology. It shows great potential to reduce and eventually replace animal testing by providing a reliable alternative for mechanistic and toxicity studies. Incorporating monocyte-derived macrophages (MDMs) as free immune cells and important key players in innate and adaptive immunity makes this model well-suited for evaluating inflammatory endpoints. In a previous interlaboratory comparison study, the Calu-3 – macrophage co-culture model demonstrated consistent results in 5 out of 7 laboratories upon bacterial lipopolysaccharide (LPS) exposure. Similarly, exposures to DQ₁₂ and titanium dioxide (TiO₂) particles at low concentrations, which were selected to reflect in vivo conditions, did not significantly affect cell viability or barrier integrity in any of the involved laboratories [31]. Although the Calu-3-MDMs co-culture model demonstrated transferability and reproducibility, the absence of statistically significant pro-inflammatory cytokine release under the tested conditions suggests that higher deposited doses may be necessary to enhance predictive value, improve comparability with in vivo responses, and identify key endpoints for future integration into quantitative IVIVE and regulatory applications. However, such increases must be carefully balanced to maintain relevance to realistic human exposures. This study aims to further optimize the Calu-3-MDMs model by aligning the timelines and endpoints of in vitro and in vivo experiments, with a focus on inflammation as a critical early event in AOP173. Using DQ₁₂ particles as a well-characterized reference quartz material, we investigated key pro-inflammatory cytokines, immune cell recruitment, and epithelial barrier integrity in both systems. Moreover, we designed this study to address the limitations of short-term in vitro exposure by adapting the timeline to more closely match the in vivo experiments. As the 90-day exposure specified in TG413 is not compatible with in vitro macrophage cultures, shorter exposure time points were selected and applied consistently in both in vivo and in vitro experiments to enable direct alignment between the two models, which is the primary objective of this study. We incorporated endpoints in our in vivo experiments, such as gene expression, to further align the models, focusing on the early stages of AOPs related to inflammation and cytokine production. The outcome of this alignment is to enhance the model’s predictive power, providing a more accurate evaluation of inflammatory responses and improving the relevance of in vitro findings by comparing them to in vivo conditions.

Methods

Chemicals and reagents

All the chemical reagents were purchased from Sigma-Aldrich (Switzerland), while all cell culture reagents were purchased from Gibco and Thermo Fisher Scientific (Switzerland) unless stated otherwise. At every step during cell culture and exposure, cells were kept in an incubator under controlled conditions (humidified atmosphere, 37 °C, 5% CO₂). Cell numbers were determined using an EVE™ bench-top automated cell counter (Witec AG, Switzerland) with the trypan blue exclusion method (0.4% trypan blue solution in phosphate buffer saline (PBS) (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 10010-015).

Crystalline quartz silica particles (DQ₁₂)

Crystalline quartz silica particles (DQ₁₂) were received from the Institute of Occupational Medicine (IOM), UK. They are produced by The Health and Safety Executive (HSE) (Harpur Hill, Buxton, Derbyshire, UK). Composition is 100% silicon dioxide (SiO₂), α-quartz well-characterized reference quartz material (A9950) with 89.3% crystallinity. More information can be found in the material safety data sheet [32].

Particle suspension preparation

DQ₁₂ particles were dispersed according to the Nanogenotox protocol [33] for toxicity testing in a volume range between 4 mL and 10 mL in endotoxin-free water (Thermo Fisher Scientific, Cat ref. 1897399), but without bovine serum albumin (BSA). Briefly, 180 mg of the material powder was weighed into glass vials (20 mL, N24, flat-bottom; MACHEREY-NAGEL, Düren, Germany; Cat. No. 702021), pre-wetted with 100 µL ethanol, and mixed with 8 mL of endotoxin-free water to a stock concentration of 22.5 mg/mL. The solution was sonicated using a Branson SFX-550 sonifier model equipped with a model 102-C converter (Emerson, Baar, Switzerland) to reach a total power of 7056 J (7.35 W for 21 min). During the sonication, the glass vial was kept on ice to prevent excess heating of samples during sonication.

Final exposure concentrations were prepared by serial dilution of the stock in cell culture medium to achieve doses ranging from 10 to 300 µg/cm² on the 0.9 cm² inserts. For example, a 10 µg/cm² dose corresponds to 300 µg/mL in 30 µL applied per insert. Higher doses were prepared proportionally to reach 50, 100, 200, and 300 µg/cm². All suspensions were always prepared on the day of cell exposure and vortexed right before exposure (Vortex-Genie 2, Scientific Industries, USA).

Exposure concentrations were selected to balance the detection of measurable biological responses while avoiding unrealistic or overload conditions. The lower concentration (10 µg/cm²) was chosen based on PATROLS guidance for lung dosing of engineered nanomaterials [34], while higher concentrations were included to explore dose–response relationships. Concentrations above 100 µg/cm² were considered increasingly unrealistic; therefore, 10 and 100 µg/cm² represent biologically plausible exposures.

TEM imaging of DQ₁₂ particles

To image the DQ₁₂ particles, transmission electron microscopy (TEM) using FEI Technai Spirit (FEI, Hillsboro, OR, USA) operating at 120 kV was used. Ten microliters of the sample was drop-casted onto carbon film on copper 300 square mesh (Electron Microscopy Sciences, Pennsylvania, PA, USA) and dried at RT before visualizing the particles with a 2048 × 2048-pixel wide angle Veleta CCD camera (Olympus, Toyko, Japan). Image processing, including scale bar inclusion, contrast adjustments, and particle analysis, was conducted using Fiji (ImageJ version 1.54) software. The size was measured automatically with the analyze particles option after thresholding and watershed. Only correctly watershed particles were considered for the size evaluation from 4 TEM images, and sixty-two particles were analyzed. Frequency distribution results were plotted using the GraphPad Prism 10.2.3 software (San Diego, CA, USA).

Due to challenges in automated segmentation and pronounced particle heterogeneity, 62 particles could be confidently measured. Although larger sample sizes are sometimes recommended, statistical checks (convergence of the mean, median, and trimmed mean; acceptable confidence intervals; and consistent bootstrap results) demonstrate that this sample size provides stable, representative estimates of the particle population. Gaussian Mixture Modeling supports a single unimodal distribution, indicating that the observed variability is intrinsic.

SEM imaging of DQ₁₂ particles

DQ₁₂ particle suspension was prepared in MiliQ water and imaged using a scanning electron microscope (SEM, TESCAN Mira 3 LM field emission, Kohoutovice, Czech Republic). Briefly, 10 µL of diluted stock particles were dried on silicon wafer slides affixed to aluminum SEM stubs (Agar Scientific, Stansted, UK) with copper conductive tape double coated (Ted Pella, Redding, California, US) and sputter coated with a 2 nm thick layer of gold using a 208 h sputter coater (Cressington Scientific Instruments, Watford, UK). Imaging was performed at an accelerating voltage of 7 kV, a working distance of 9.45 mm, and using the secondary electron detector (SE). The field of view was 4.83 μm at a magnification of 114,000×. Scale bar inclusion was done in Fiji (ImageJ version 1.54) software.

DLS measurement

The hydrodynamic diameter was determined by dynamic light scattering (DLS) using an Anton Paar Litesizer 500 particle analyzer (Anton Paar, Graz, Austria) operating with a 658 nm laser and a scattering angle of 175°. Thirty measurements of 10 s were performed at 37 °C. Stock suspensions were diluted with Milli-Q water (pH = 6.0) or Calu-3 cell culture medium to a final concentration of 90 µg/mL for measurement. This corresponds to a dilution factor of 3–100× relative to the concentrations used in cell exposure experiments, ensuring that particle concentrations were suitable for reliable DLS measurements without multiple scattering or detector saturation.

Cell culture

The human bronchial epithelial cells (Calu-3) were purchased from the American Type Culture Collection (ATCC, Rockville, USA). Calu-3 cells were cultured in MEM +GlutaMAX (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 41090-028) supplemented with heat inactivated 10% fetal bovine serum FBS (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. A5256701), 1% Pen Strep (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 15140-122), 2% Amphotericin B (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 15290-026) and 1% Non-Essential Amino Acids (NEAA) solution (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 11140-035), which will be called complete cell culture media in the following text (CCM). Heat inactivation was done by incubating the pre-thawed FBS bottle for 30 min at 56 °C in a water bath with occasional shaking. The cells were seeded at a density of 90,000 cells/cm² Falcon^® 12-well PET permeable supports (0.4 μm pore size, 0.9 cm² growth area; Corning, Corning, NY, USA; Cat. No. 353180) placed in a companion 12-well flat-bottom plate with a low-evaporation lid (Corning Life Sciences, Corning, NY, USA, Cat. No. 353043). The cells were incubated at 37 °C in a 100% humidified atmosphere containing 5% CO₂. The medium was renewed every 2–3 days. Human MDMs were differentiated from monocytes isolated from buffy coats provided by the Swiss Transfusion Center (Bern, Switzerland), as described previously [35, 36]. In brief, isolated blood monocytes were cultured in 6-well tissue culture plates (Corning, FALCON^®, United States) for six days at a density of 10⁶ cells/mL in 3 mL Roswell Park Memorial Institute (RPMI)−1640 (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 42401-018) cell culture medium supplemented with 10% (v/v) heat inactivated FBS (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. A5256701), 1% L-glutamine (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 25030-024), and 1% Pen Strep (Gibco™, Thermo Fisher Scientific, Paisley, UK, Cat. No. 15140-122), referred to as the complete cell culture medium (cRPMI). For MDM differentiation, 10 ng/mL of macrophage colony-stimulating factor (M-CSF; Miltenyi 130–096–485) was added to cRPMI for 6 days.

Calu-3 and MDMs co-culture model preparation

The cell culture model is well-established and used in a recent interlaboratory comparison study [31]. More details on the Calu-3 model can be found in [36–38]. In brief, Calu-3 cells were cultured under submerged condition for 7 days. After 7 days, the apical medium was removed, and the cells were cultured at the ALI for an additional 7 days. On day 14, after differentiation into MDM, cells were scraped off the wells, counted, and added to the Calu-3 monolayer to obtain co-culture. As not all MDM attach to the Calu-3 cells, 39,000 MDM were added in a volume of 0.2 mL to achieve a final concentration of 25,000 MDM/cm² [29]. After 4 h, the apical medium was removed, and the co-culture was kept at 37 °C overnight. Monocytes were derived from different donors for each biological repetition.

Pseudo-ALI exposure

Exposure was done according to the already developed PATROLS Guidance Document 3016 [38] without washing the apical side of the insert and adjusting the volume to the surface of the 12-well insert. In brief, at least 4 h after adding the MDMs, the cell culture medium was carefully removed from the apical side of the co-culture. Using the warm CCM, dilutions from 0.3 to 9 mg/mL of DQ₁₂ were made from the stock solution. Before and after dilution, the particle solution was shortly vortexed to ensure complete mixing. A volume of 30 µL of the solution was then added to the apical side of the co-culture insert with a surface area of 0.9 cm². Using this information for the volume and surface area of the inserts, a rough calculation was made, and the concentrations were reported as 10–300 µg/cm², respectively. The plate was returned to the incubator at 37 °C and 5% CO₂ for the desired exposure period of 24 h, 3 days, or 7 days. During exposures, only the basal medium was refreshed every 2 days, while the apical compartment was left undisturbed and the apically applied DQ₁₂ remained in situ throughout the exposure period.

Endotoxin detection

Both cell culture media and particle suspension media aliquots were regularly analyzed for endotoxin contamination using PierceTM Chromogenic Endotoxin Quant Kit (Thermo Scientific ref. A39552) using the manufacturers’ user guide.

Mycoplasma detection

A Mycoplasma test was performed regularly to detect potential mycoplasma infections in cell culture. Two mL of cell supernatant from a cell culture flask, which has been in contact with cells for at least 24 h (better 48 h), was collected for the mycoplasma detection. MycoAlert™ Assay was performed according to the manufacturers’ user guide Lonza protocol (Catalog #: LT07-318).

Measurements of transepithelial electrical resistance (TEER)

At 24 h after exposure, 500 µL cell culture medium was added to the apical side of the cells, and the epithelial barrier integrity was evaluated by TEER measurements using a Millicell^® ERS-2 (Electrical Resistance System, Millipore, Switzerland) equipped with an STX01 electrode (World Precision Instruments, Switzerland). All TEER measurements were recorded in the presence of a 500 µL cell culture medium at 37 °C in the apical compartment of the inserts and a 1500 µL medium in the basolateral compartment. Measurements were performed inside a biosafety cabinet at room temperature, with cultures being outside the incubator for 25 min. Before each TEER recording, the electrodes were sterilized in 70% ethanol, rinsed with pre-warmed sterile 1x HBSS (Gibco™, Thermo Fisher Scientific, Cat. No. 14025-092) and briefly equilibrated in pre-warmed cell culture medium. This procedure was done before measuring TEER in each insert. Blank insert (without cells) was measured to determine offset values, which were subtracted from sample readings. Measurements were performed in three technical replicates (each insert was measured 3 times with the probe positioned each time at different areas of the insert), and a minimum of 3 biological replicates.

Cytotoxicity

The release of LDH into the supernatant was assessed using an LDH cytotoxicity detection kit (Roche Applied Science, Mannheim, Germany, Ref. 11644793001), according to the manufacturer’s protocol. Each sample in vitro was tested in a single experiment on three to four independent occasions, while in vivo, it was evaluated with at least six repetitions (n = 6) and eight repetitions (n = 8) for the control group (dH₂O instillation). LDH activity was quantified photometrically by measuring at 490 nm (reference wavelength at 630 nm). For positive controls, cell cultures were exposed apically to 2% Triton X-100 in PBS for 24 h, and the negative control was untreated tissue. Samples from both apical and basal sides were tested.

Inflammatory cytokines and chemokines

At 24 h, 3 days, or 7 days after exposure, 500 µL cell culture medium was added to the apical side of the inserts. After 30 min, the apical and basolateral supernatants were collected separately for cytokine and chemokine analysis. Release of inflammatory cytokines interleukin IL-6, IL-8, and IL-1β were measured in the supernatants using ELISA kits (Biotechne, R&D systems, Minneapolis, USA, IL-6 Cat. No DY206, IL-8 Cat. No. DY208, and IL-1β Cat. No. DY201). Samples were collected from three independent experiments. Cytokine analysis was performed according to the manufacturer’s protocol. For ELISA, absorbance was measured using a spectrophotometer at 450 nm and a reference wavelength of 570 nm.

Gene analysis

After collecting the supernatant for the ELISA assay, the cells were carefully washed 3 times with PBS, and the total RNA was isolated from the co-culture. Cell lysis was performed directly on the insert, using 100 µL of BL + TG buffer (Promega Madison, WI, USA, Ref. Z6011), and total RNA was extracted using ReliaPrep™ RNA Cell Miniprep System (Promega, Madison, WI, USA, Ref. Z6011) following the technical manual from the manufacturer. The quantity and quality of RNA were assessed by Thermo Scientific™ NanoDrop™ 2000 Spectrophotometer. The 260/280 and 260/230 absorbance ratios were evaluated, and only RNA samples with ratios of ~ 2.0 (typically 1.9–2.2) were taken forward for cDNA synthesis and RT-qPCR analysis.

Real-time qPCR for inflammatory cytokines and chemokines

The reverse transcriptase reaction was performed with the Omniscript RT system (Qiagen, Hilden, Germany, Cat. No. 205113), OligodT (Microsynth, Balgach, Switzerland, Cat. No. 35 and 36), and RNasin Plus RNase Inhibitor (Promega, Madison, WI, USA, Switzerland, Cat. No. NZ2611), described previously [39]. In brief, the synthesis of complementary DNA (cDNA) was performed by using 6.5 µL of isolated RNA (250 ng), 1 µL oligo-dT primer (10 µM), 0.25 µL RNase inhibitor, 1 µL dNTP Mix (5 mM), 0.25 µL Omniscript reverse transcriptase (1 Unit), and 1 µL buffer RT.

The real-time PCR was performed on the 7500 fast real-time PCR system (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA). Reactions were carried out in a final volume of 10 µL by mixing 2 µL 5-fold diluted cDNA with 5 µL SYBR-green master mix (Fast SYBR Green master mix, Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA, Cat. No. 4385617), 2 µL nuclease-free water (Promega, Madison, WI, USA, Cat. No. P119A), and 2 µL primer mix (91 nM).

Thermal cycling conditions consisted of an initial denaturation at 95 °C for 20 s, followed by 40 amplification cycles of denaturation at 95 °C for 3 s and annealing/extension at 60 °C for 30 s. A melt curve analysis was performed at the end of each run, consisting of a preconditioning step at 95 °C for 15 s followed by incubation at 60 °C for 1 min and a continuous temperature ramp from 60 °C to 95 °C with fluorescence acquisition to verify amplification specificity.

Relative expression levels were calculated as fold change compared to untreated tissues with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein zeta (YWHAZ) as internal standard genes. These reference genes were selected based on empirical assessment of expression stability, as their Ct values showed minimal variation (CV < 3%) across all experimental conditions and treatments. Primers were purchased from Thermo Fisher Scientific (Zug, Switzerland). Details about the primers are included in the (Table S1).

Macrophage DiD live staining

To visualize MDMs on top of Calu-3 cells throughout the fluorescence microscopy experiment and to determine their number, cells were stained with Vybrant DiD Cell Labeling Solution (Molecular Probes, Cat. No. V22887). Imaging was conducted using an inverted confocal laser scanning microscope (CLSM; Leica Stellaris 5, Heerbrugg, Switzerland) equipped with Power HyD S detectors, a Plan FLUOTAR 10x/0.32 Dry objective, a Plan-Apochromat 20x/0.75 Dry objective, and a Plan-Apochromat 63x/1.4 Oil CS2 objective (Leica, Heerbrugg, Switzerland). The system was operated using LAS X software version 4.6.1. Stack images of the cells were acquired sequentially at 10x magnification, providing a field of view of 1163.64 μm × 1163.64 μm. The resulting tiles were merged using the LAS X software. Fluorescence imaging was performed using a laser excitation wavelength of 647 nm for Alexa Fluor 647.

Before adding MDMs to Calu-3, monolayer cells were suspended at a density of 1 × 10⁶/mL in RPMI, and 1.5 µL of DiD labeling solution per mL of cell suspension was added. Cell suspension was gently mixed and incubated for 10 min at 37 °C. After incubation, the suspension was centrifuged at 1500 rpm for 5 min at room temperature using an Eppendorf Centrifuge 5702 R (Eppendorf, Germany). The supernatant was removed, and cells were gently resuspended in a warm (37 °C) medium. The washing step was repeated two more times. Cells were then seeded at the previously mentioned density on top of the Calu-3 epithelium.

For macrophage counting, the ImageJ software (version 1.54) was used. Counting was performed using an automatic function on three randomly chosen regions of interest where Z-stacks were acquired. Before counting, MAX-Z projection and eroding to points were performed (Figure S1). The average number of macrophages counted was calculated from the processed Z-stacks and then divided by the size of the acquired image area, 1163.64 × 1163.64 μm. The final number was then converted to cell/cm².

Tight junctions, F-actin and nuclei staining

Staining was performed to visualize tight junctions and the monolayer structure of epithelial cells. Cells grown on cell culture inserts were washed 3 times with phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (Thermo Fisher Scientific Inc., the Netherlands, Cat. No. 158121) for 20 min. After fixation with 4% paraformaldehyde, the membrane was carefully cut out from the plastic insert frame, and all subsequent staining steps were carried out on the excised membrane. Then, they were permeabilized with 0.1% Triton X-100 (Thermo Fisher Scientific Inc., the Netherlands, Cat. No. T8787) for 10 min. After permeabilization cells were incubated with the ZO-1 rabbit polyclonal antibody (Thermo Fisher Scientific, Switzerland, Cat. No. 617300) 1:100 in 0.1% bovine serum albumin (BSA) (Sigma Aldrich, Cat. No. A7030) in PBS) for 2 h, followed by another 2 h incubation with a mixture of secondary antibodies in 0.1% BSA in PBS: goat anti-rabbit immunoglobulin G (IgG) DyLight 488 conjugated (Agrisera, Sweden, 1:100 dilution), rhodamine-phalloidin (Thermo Fisher Scientific, Switzerland, 1:100 dilution, Cat No A22287) and 1 µg/mL 4′ 6-diamidino-2-phenylindole (DAPI) (Roche, Manheim, Germany, Cat No. 10236276001). All the staining steps were performed in the dark at room temperature. After staining, the cells were washed with PBS and mounted in Kaiser’s glycerol gelatine, phenol-free (Merck KGaA, Germany, Cat. No. 108635) on microscopy slides.

Imaging was conducted using an inverted confocal laser scanning microscope (CLSM; Leica Stellaris 5, Heerbrugg, Switzerland) equipped with Power HyD S detectors, a Plan FLUOTAR 10x/0.32 Dry objective, a Plan-Apochromat 20x/0.75 Dry objective, and a Plan-Apochromat 63x/1.4 Oil CS2 objective (Leica, Heerbrugg, Switzerland). The system was operated using LAS X software version 4.6.1. Stack images of the cells were acquired sequentially at 10x magnification, providing a field of view of 1163.64 μm × 1163.64 μm. The resulting tiles were merged using the LAS X software. Fluorescence imaging was performed using three laser excitation wavelengths: 405 nm for DAPI, 488 nm for Alexa Fluor 488, and 647 nm for Alexa Fluor 647.

Rats

Eight-week-old male Fisher 344 rats (Janvier Labs, Le Genest Saint Isle, Saint Berthevin, France) were used for the study. Rats were housed in a specific pathogen-free facility at the Department of BioMedical Research, University of Bern (Bern, Switzerland) in individual cages with food and water ad libitum in controlled temperature and light conditions.

Study design

The rats were randomized into three groups: a control group receiving only distilled H₂O by oropharyngeal aspiration and two groups receiving 8 mg/kg of DQ₁₂ (approximately 2 mg/rat) particles by oropharyngeal aspiration. In the acute group, all endpoints were assessed 24 h after exposure to particles, and in a sub-acute group, all endpoints were evaluated 3 and 7 days post-particle administration. Each group consisted of six to eight rats.

Anesthesia

Isoflurane was used to anesthetize rats before oropharyngeal aspiration. Briefly, one rat at a time was placed in an anesthetic chamber filled with 4% isoflurane until no pinching reflex was observed. After the procedure, rats were checked for weight and physical condition daily to observe and monitor any decline in their health.

Particle administration to the lungs

After anesthesia, rats were placed on a vertical holder, and the mouth was opened with a tweezer to apply a drop of particle suspension (in 200 µL of dH₂O). Oropharyngeal aspiration was performed in two repetitions to avoid complications with the animal’s breathing during aspiration of the suspension. Briefly, the tongue was held with a tweezer, and the particle suspension was placed at the back of the tongue. The nose was closed until the rats inhaled the liquid directly. Then, the operation was repeated. Rats were given time to recover before being placed back in their cage.

Calculated deposited dose of DQ₁₂ by oropharyngeal aspiration

To compare the in vitro and in vivo deposited doses of DQ₁₂ following exposure, we estimated the lung surface area based on values reported in the literature. The lung surface area of an adult Fisher 344 rat is ranging from 4,000 to 10,000 cm² [40, 41]. Using these values, we calculated the lung tissue’s deposited dose per cm² of lung surface area, ranging from 0.2 to 0.5 µg/cm². This estimation was used to approximate biologically effective doses in vivo and to facilitate comparison with the in vitro model. The mass-per-kilogram values reported later in the manuscript reflect the nominal dose prepared for animal exposure, normalized to rat body weight.

Euthanasia

Rats received an intra-peritoneal injection of 1000 µL Esconarkon (300 mg/mL, Pentobarbitalum natricum, Streuli Pharma AG) before collecting organs.

Assessment

Rats were sacrificed 24 h, 3 days, and 7 days post oropharyngeal aspiration of the particles. The rats were euthanized; blood was collected by puncturing the heart with a 24G syringe before dissection. The lung, heart, and trachea were exposed by bilateral thoracotomy. Before collecting bronchoalveolar lavage fluid (BALF), the left main bronchus was closed with surgical forceps to ensure BALF retrieval only from the right lungs. A cut was applied in the upper third part of the trachea in which a 24G cannula was inserted, and 2 mL of ice-cold PBS was instilled into the lung and aspirated back again for further analysis (this step was repeated three times with fresh PBS for a total of 6 mL of collected BALF). The forceps were changed on the left main bronchus to cut off the right lobes of the lungs for further analysis, and they were placed in different tubes on ice. The heart was perfused through the right ventricle with a minimum of 10 mL of ice-cold 4% PFA to fix the following organs via perfusion before collection: left lung, lung draining lymph nodes (LDLN), trachea, spleen, and one lobe of the liver for histopathology assessment. Organs were fixed in 4% paraformaldehyde (PFA) overnight at 4 °C prior to dehydration. Briefly, tissues were perfused and immersed in ice-cold 4% PFA overnight, then rinsed in PBS for 10 min the following day. Dehydration was performed on ice with 50% ethanol for 30 min, followed by 70% ethanol for 30 min. Samples were then transferred to room temperature and incubated in 90% ethanol for 30 min, 100% ethanol (2 × 30 min), and 100% xylene (2 × 30 min).

For paraffin embedding, tissues were incubated in paraffin I (100%) at 60 °C for 3 h, followed by paraffin II (100%) at 60 °C for 1 h. Organs were embedded individually in paraffin cassettes, and once solidified, paraffin blocks were stored at 4 °C until sectioning.

Bronchoalveolar lavage fluid analysis (cytospin, differential count)

After collecting BALF, the samples were kept on ice until centrifugation (10 min at 200 g). The cell pellet was resuspended in 1 mL of ice-cold PBS, and cells were counted manually. Then, 60,000 cells in 150 µL were collected and spun down for 5 min at 800 rpm onto glass slides (CytospinTM cytocentrifuge; Thermofisher Scientific). The remaining cells were frozen at −80 °C for RNA extraction. BALF was collected by slowly instilling 5 mL of ice-cold PBS into the lungs via the trachea, followed by gentle retrieval of the lavage fluid. This procedure was repeated three times, yielding three separate BALF aliquots. From the first aliquot, 200 µL was kept for the LDH assay, and 200 µL was kept for the BCA assay (total protein) before freezing the remaining volume. After cytospin, the cells were stained with the Diff Quick staining kit (RAL Diagnostics, France) and mounted before evaluation by a trained diagnostic biomedical scientist, which included the differential counting of inflammatory cells.

Hematoxylin and eosin staining

Hematoxylin and eosin (H&E) staining was conducted as part of the standard procedure on formalin-fixed and paraffin-embedded sections. For consistency purposes, we chose to work on the left lobe only as the right lobes were used for BALF retrieval.

Real-time quantitative PCR (in vivo samples)

Total RNA was extracted from lung tissue using the NucleospinR RNA kit (Macherey Nagel, Germany). Following the manufacturer’s instructions, reverse transcription was performed with the Omniscript RT kit (Qiagen, Germany). Quantitative RT-PCR was conducted using the FastStart Universal SYBR Green Master (Rox) (Merck, Germany) with the Quantstudio 6 Real-Time PCR machine (ThermoFisher Scientific). Relative mRNA expression was calculated using the ΔΔCt method, and fold change was determined by FC = 2^-ΔΔCt. Data were normalized to the housekeeping gene TATA Binding Protein (TBP). The use of a single reference gene was justified by its validated expression stability across all experimental conditions, time points, and treatment groups, with minimal Ct variation, supporting its suitability for normalization in this dataset. Primers used are listed in the supplementary figure (Table S2).

Enzyme-like immunosorbent assay

Rat BALF protein content was screened for IL-1β (Invitrogen, 88–6010 A), CXCL5 (Invitrogen; ERCXCL5), and IL-6 (Sigma Aldrich, RAB0311) with an ELISA kit. The assays were performed following the manufacturer’s instructions.

Statistics

For the in vivo analysis, statistics were conducted using GraphPad Prism 8.4.0 software (GraphPad Software Inc., La Jolla, California, USA). A one-way ANOVA followed by Dunnett’s multiple comparisons test, with a single pooled variance, was applied. Data with P-values of p < 0.01(*), p < 0.001(**), p < 0.0001(***), p < 0.00001(****) were considered statistically significant. All endpoints were evaluated with at least six repetitions (n = 6) for 24 h, 3- and 7-days post-particle administration, and eight repetitions (n = 8) for the control group (dH₂O instillation). For the in vitro experiments, GraphPad Prism 10.2.3 software (San Diego, CA, USA) was used for statistical analyses. Absolute values were used for the statistical analyses of TEER and cytokine measurements. Normalized values were used for the statistical analyses of LDH release. For the in vitro experiments, normalization was performed to the positive control (2% Triton X-100 in PBS, defined as 100% cytotoxicity) based on absorbance measurements. We performed 3 to 5 independent replicate experiments using 3 or 2 inserts per condition. Replicates of the experiments were used to calculate the standard deviation (SD) and mean values. Statistical tests were performed as indicated in the Figure legends.

In vitro – in vivo correlation of gene expression data

Values for gene expression fold change normalized to respective untreated control and housekeeping genes were used for analysis. We used the Spearman rank correlation coefficient (r_s) to determine whether there is a monotonic relationship between the results obtained by three methods/two models (in vivo, in vitro at low dose, and in vitro at high dose). The numbers of repetitions for each method were not equal; thus, we did not use conventional central tendency statistics such as mean or median. Instead, the correlation coefficients were computed on 2-tuples. Accordingly, we performed pairwise combinatorial selections as follows: for a given timepoint t1, if method/model 1 (e.g., in vivo) had n1(t1) independent repetitions, and method/model 2 (e.g., in vitro low dose) had n2(t1) independent repetitions, the total number of statistically equivalent pairs between these two sets of data is n1(t1) x n2(t1). Therefore, in case of three timepoints (t1,t2,t3), the total number of pairs of is n1(t1) x n2(t1) x n1(t2) x n2(t2) x n1(t3) x n2(t3). Therefore, correlation coefficients were computed on each possible pair corresponding to the three time points, and we reported the mean value. (Table S3)

Results

3D lung cell model characterization

Calu-3 cells reached confluency on day 8 (Fig. 1A), forming a tight monolayer (Fig. 1B), after which they were transferred to ALI by removing the cell culture medium from the apical side. At ALI, they maintained high TEER values in the range 400–800 Ωcm² (Fig. 1C) throughout the whole co-culture preparation period for 16 days. On day 15, MDMs were added. The number of MDMs was assessed to be 16,863 ± 3570 cells/cm² by staining the cells with the DiD fluorescent dye and imaging the cells without fixation (Fig. 1D).

Fig. 1.

Characterization of the Calu-3 and MDMs co-culture model. A Brightfield images with a scale bar of 100 μm show Calu-3 cells after seeding on top of the insert in a submerged condition until day 8. After 8 days, they were transferred to ALI. MDMs were added on day 15, as indicated with a yellow arrow, and exposure was performed on day 16. B Calu-3 monolayer nuclei DAPI staining (cyan) and F-actin (magenta). C TEER values from cell seeding until the exposure; data are presented as mean ± standard deviation, N = 4. D MDMs (yellow) live imaging with Vibrant DiD staining

DQ₁₂ characterization

Scanning electron microscopy (SEM) analysis showed that DQ₁₂ particles have a broad particle size distribution from 100 to 1000 nm and sharp, spiked edges as their distinctive topographical features (Fig. 2A). To precisely determine the size of DQ₁₂ particles and deposition, the 10 mg/cm² suspension was prepared in MiliQ water. The suspension was drop cast on a TEM grid and left to dry for 3 days. TEM imaging showed homogenous deposition on the grids (Fig. 2B). From image analysis, the calculated size of the particles was 557 ± 283 nm (mean ± standard deviation) (Fig. 2C), which aligned with DLS measurements and with what we observed on SEM images. To test the potential aggregation of the particles in cell culture media before exposure to the cells, the 100x dilution from stock solution was prepared. Particles showed no apparent aggregation upon preparing the suspension measured by DLS (Figure S2).

Fig. 2.

Characterization of DQ₁₂ particles. A SEM image of DQ₁₂ topographic features. B TEM image of DQ₁₂ deposition on TEM grids. C The size of DQ₁₂ particles was determined using TEM image analysis

Acute exposure to DQ₁₂ - Impact on cytotoxicity, membrane integrity, and pro-inflammatory response in vitro

After 24 h post-exposure of the co-culture model with different concentrations of DQ₁₂ ranging from low, 10 µg/cm² to high, 300 µg/cm², there was no observed cytotoxicity by measuring LDH release from the apical side (Fig. 3A). Constant TEER values indicated no changes in membrane integrity (Fig. 3B), which was also confirmed by ZO-1 staining, where we observed the presence of tight junctions (Figure S3). When looking at inflammatory markers, only a significant increase in the secretion of IL-6 was observed for the doses of 200 µg/cm² and higher (p < 0.05) (Fig. 3C) but not for IL-1β and IL-8 (Figure S4). Moreover, no changes in IL-1β, IL-6, or IL-8 gene expression from the co-culture model were observed (Fig. 3D).

Fig. 3.

Effects on in vitro co-culture model upon 24 h post-exposure to DQ₁₂. A Cytotoxicity measured as LDH release from the apical side of the insert 24 h after exposure, normalized to 0.2% Triton X-100 control. Data are presented as mean ± standard deviation (N = 4); B Membrane integrity evaluated by TEER measurements, data are presented as mean ± standard deviation (N = 4); C IL-6 release measured by ELISA assay, positive control was LPS treated cells. Data are presented as mean ± standard deviation (N = 3), results marked as with ** were considered statistically significant (p < 0.01) according to ordinary one-way ANOVA with Dunnett’s multiple comparison test; D qPCR data of IL-1β, IL-6, and IL-8 gene expression profile as markers of pro-inflammatory response upon exposure to DQ₁₂, dose range from 10–300 µg/cm². Data for gene expression are presented as fold-change to the untreated control. The positive control was LPS-treated cells. Data are presented as mean ± standard deviation (N = 3)

In the initial 24-hour in vitro experiments, no broad inflammatory response was observed, as inflammatory gene expression remained unchanged and cytokine secretion was largely unaffected, with the exception of IL-6 release at the two highest DQ₁₂ doses, suggesting that short-term exposure may not be sufficient to elicit measurable effects across multiple inflammatory endpoint, even with significantly high doses of DQ₁₂. To investigate this further, we conducted in vivo experiments and adapted the experimental design in vitro for improved alignment with the in vivo readouts. This included extending the exposure duration to 3 and 7 days and adjusting the endpoint measurements in both the in vitro and in vivo models. Additionally, we conducted in vivo experiments to verify in vitro findings and simultaneously confirm the responsiveness of DQ₁₂ using the same batch of material to eliminate potential discrepancies in results obtained from both models that could occur because of the differences in the starting material (Fig. 4).

Fig. 4.

Experimental design for in vivo- in vitro comparison. Exposures were performed in both models, starting with the same batch of DQ₁₂ material. Cytotoxicity and inflammatory endpoints were measured after days 1, 3 and 7

Prolonged exposure to DQ₁₂ - Impact on cytotoxicity and lung tissue morphology in vitro and in vivo

No cytotoxicity was observed at 3 and 7 days post-exposure, with concentrations of 10 µg/cm² and 100 µg/cm² of DQ₁₂ in vitro (Fig. 5A). On the other hand, in vivo experiments revealed significant cytotoxicity on day 3 (p < 0.05) (Fig. 5B). Changes in membrane integrity compared to non-treated cells when measuring TEER (Fig. 5C), tight junctions gene expression (Figure S5), and ZO-1 staining (Figure S3) showed a decreasing trend but not statistically significant. These findings highlight no morphological change or disruption of the integrity of the epithelium upon exposure to DQ₁₂ in vitro. On the other hand, histological analysis revealed cellular infiltration in rat lung tissue 24 h after DQ₁₂ exposure, which persisted and intensified over the subsequent three days, culminating in patchy areas of infiltrated cells. By day 7 post-exposure, the extent of cellular infiltration was notably reduced; however, the lung architecture exhibited altered morphology, including smaller alveoli and evidence of tissue destruction (Fig. 5D).

Fig. 5.

Effects on in vitro co-culture model upon 7 days post-exposure to DQ₁₂. A Cytotoxicity measured as LDH release after 7 days from the apical side of the insert, normalized to 0.2% Triton X-100 control, data are presented as mean ± standard deviation (N = 3); B Cytotoxicity was assessed in vivo by measuring LDH release in BALF. Rats were administered 8 mg/kg of DQ₁₂ via oropharyngeal aspiration in two doses of 200 µl of dH₂O, and BALF was collected at 24 h, 3 days, and 7 days post-administration. Control rats received 400 µl of dH₂O via oropharyngeal aspiration.; C Membrane integrity evaluated by TEER measurements. D H&E staining was performed in rat lung tissue following DQ₁₂ administration, following the same treatment protocol described in (B). Data are presented as mean ± standard deviation (N = 4 for in vitro, N = 6–8 for in vivo experiments). It should be noted that the in vitro doses (10–100 µg/cm²) are not intended to be dosimetrically equivalent to the calculated in vivo deposited dose (0.2–0.5 µg/cm²). The comparison in Fig. 5 is qualitative and endpoint-based rather than quantitative dosimetry

Prolonged exposure to DQ₁₂ - Impact on pro-inflammatory response in vitro and in vivo

Enhanced secretion of cytokine IL-8 in vitro, which is responsible for attracting immune cells, specifically neutrophils, to the site of inflammation in humans, was observed on days 3 and 7 following exposure (p < 0.05) (Fig. 6A). A significant increase in the secretion of IL-6 in vitro was also observed for the dose of 100 µg/cm² for both days 3 and 7 (p < 0.05) (Figure S6). Differential counts from cells collected in BALF of the rats revealed an enhanced influx of total immune cells on days 3 and 7 (Fig. 6B). This was due to increased numbers of lymphocytes, monocytes, eosinophils, and neutrophils (Figure S7), which was also visualized and confirmed by Diff Quick Staining (Fig. 6C). To correlate gene expression in vitro with in vivo data obtained from rats, a Spearman’s correlation coefficient was calculated between the two data sets for both low and high doses of DQ₁₂ (Table S3). By simultaneously measuring identical inflammatory markers across both models, we observed that the in vivo results mirrored those of the extended in vitro experiments. Specifically, the same inflammatory markers exhibited significant upregulation, confirming the delayed inflammatory response (Fig. 6D). Expression of IL-1β was significantly upregulated on day 7 in both models for both doses, with a moderate positive correlation for the low 10 µg/cm² dose (r_s=0.50) (Table S3). Expression of IL-6 reached significant upregulation on day 3 and then dropped to baseline expression on day 7 in both models. However, despite the same trend, the correlation was very weak when the statistical analysis was performed (r_s=0.04 and r_s=0.11) (Table S3). The expression of IL-8 cytokine was constantly enhanced on all time points analyzed, peaking on day 3 in the in vitro model exposed with the 100 µg/cm² dose and between day 3 and day 7 with the 10 µg/cm² dose with the moderate positive correlation for the low 10 µg/cm² dose (r_s=0.50) (Table S3). Since rodents do not produce IL-8, the expression of IL-8 receptor CXCR2, present in both humans and rodents [42] was measured in vivo. CXCR2 receptor showed upregulation on day 3, which persisted on day 7 (Fig. 6D). Even though rodents do not have the IL-8 gene, they produce multiple IL-8 homologs that bind to the CXCR2 receptor and have the same function as IL-8 [43]. Gene expression of the two IL-8 homologs, CXCL2 and CXCL5-6 [43], was analyzed. They showed a peak on day 3, which, in contrast to CXCR2, dropped on day 7 (Figure S8). Therefore, for the best comparison with IL-8 in vitro, CXCR2 was chosen in vivo.

Fig. 6.

In vitro-in vivo comparisons following exposure to DQ₁₂. A Increased IL-8 secretion in the in vitro model on days 3 and 7 following exposure. B Total number of cells counted in BALF at different time points post-DQ₁₂ administration in rats. Rats were administered 8 mg/kg of DQ₁₂ via oropharyngeal aspiration in two doses of 200 µl of dH₂O, and BALF was collected 24 h, 3 days, and 7 days post-administration. Control rats received 400 µl of dH₂O via oropharyngeal aspiration. C Diff Quick Staining of BALF cells following the exposure protocol as described in (B). D Interleukin 1β, interleukin-6, and interleukin-8 gene expression profile in vivo and in vitro on days 1,3 and 7 upon exposure to DQ₁₂ at 10 µg/cm² (low) or 100 µg/cm² (high). Rats were exposed as described in (B), receiving a calculated dose of 0.2-0.2.5ug/cm². Data are presented relative to the negative control (line); untreated cells served as a baseline; the positive control was LPS-treated cells. Data are presented as mean ± standard deviation (N = 5 for in vitro and N = 6–8 for in vivo)

The consistency between the in vitro and in vivo data underscores the importance of prolonged post-exposure time and multiple sampling points in observing the dynamics of the inflammatory response. Furthermore, the endpoints identified in our in vitro model will serve as valid indicators of inflammation in vivo, improving in vitro relevance following inflammatory response to inhaled particles such as DQ₁₂.

Discussion

Understanding the health risks associated with inhalation exposure to hazardous substances is essential in toxicology, as it directly relates to the impact on human respiratory health. However, traditional and regulatory-approved in vivo methods for assessing local respiratory toxicity raise ethical concerns, are costly, time-consuming, and might fail to predict human outcomes because of inter-species differences. In recent years, governments and regulatory bodies have actively supported the development and uptake of alternative methods to animal testing. For example, the UK government has published a comprehensive strategy to phase out regulatory animal testing and accelerate the adoption of non‑animal methods such as advanced in vitro systems and organ‑on‑a‑chip technologies, aiming to end specific animal tests by 2026–2030 and support human relevant alternatives where effective replacements exist [44]. In the EU, animal testing for cosmetic products and ingredients has been banned for many years, with a full marketing ban in force since 2013, which has driven investment in non‑animal approaches to safety assessment [45]. Developing robust in vitro lung models has therefore become not only a humane, but also a scientifically advanced and regulatory-relevant alternative to traditional animal tests [46].

Establishing confidence in NAMs for regulatory use requires more than technical feasibility; it depends on a structured evaluation of the context of use, biological relevance, technical performance, data integrity, transparency, and independent review, as outlined by van der Zalm et al. (2022) [10] and ICCVAM (2024) [47]. Our study addresses several of these elements by employing a human bronchial co-culture model, anchoring endpoints in an AOP, and aligning exposure.

Our study aims to contribute to this paradigm shift by developing an in vitro model that helps build confidence in non-animal methods for inhalation toxicology. To achieve this, we employed an in vivo-in vitro comparative approach to translate findings between in vitro experiments conducted in the Calu-3-MDMs co-culture model and the in vivo rat model recommended by OECD TG413 for inhalation toxicology. Moreover, our approach aligns with four key actions recently proposed to enhance the adoption of in vitro methods and meet the criteria required to replace animal models in regulatory settings [48]. In particular, (i) providing transparent and detailed methodological descriptions consistent with the principles of GIVIMP and utilizing established SOPs; (ii) selecting endpoints corresponding to key molecular and cellular events described in inhalation-related AOPs; (iii) employing an air–liquid interface exposure system to capture physiologically relevant responses, while acknowledging that deposited doses require further refinement for quantitative extrapolation; and (iv) using a human-relevant co-culture combining Calu-3 bronchial epithelial cells with MDMs from different donors. The inclusion of donor-derived MDMs introduces inter-donor variability in immune responsiveness, which may contribute to the observed experimental differences but also reflects real-world biological diversity, thereby enhancing the model’s human relevance.

Cell model and dose

The Calu-3-MDMs co-culture model is well-established, and an ALI-exposed co-culture lung model, tested across different laboratories [31] with a detailed SOP published during the PATROLS project [49], ensuring transferability and reproducibility to support regulatory evaluation. This model shows the epithelial monolayer integrity under prolonged ALI culture; therefore, it is suitable for long-term ALI exposure to exposed aerosols [29]. Although more physiologically relevant ALI systems derived from human donor biopsies have been recommended [48], our study supports the established robustness of the Calu-3/MDM co-culture model, previously evaluated in inter-laboratory studies [31], demonstrating intra-laboratory consistency in barrier integrity and inflammatory trends. Despite donor-related variability, the observed responses align with in vivo inflammatory patterns, underscoring the model’s biological relevance and predictive potential. Additionally, the model presented here can be easily transferred to other laboratories that lack access to human biopsy materials.

A critical challenge in IVIVE is the discrepancy between in vivo and in vitro dose metrics. In vivo exposures typically involve low, chronic doses, while in vitro systems often require higher doses to elicit measurable effects. In the previously published research, the Calu-3-MDM co-culture model did not show an inflammatory response after 24 h exposure, with doses extrapolated from human exposure when exposed to DQ₁₂ at the ALI via the VITROCELL^® Cloud system [31]. Also, other studies reported that DQ₁₂ does not upregulate pro-inflammatory cytokines in ALI-exposed A549 – THP1 co-cultures at concentrations up to 31 mg/cm² [50]. The present study utilizes a dose setting employing higher doses of DQ₁₂ to explore the potential of the Calu-3-MDMs co-culture model in capturing inflammation, as our focus was on optimizing endpoints and alignment with the rat model rather than employing the most realistic dosage. The pseudo-ALI method allowed us to deposit higher concentrations of particles compared to aerosol exposure systems [51] and is more comparable to the oropharyngeal aspiration we used in vivo. Implementing aerosolized inhalation exposures in both systems would have introduced substantial technical complexity and uncertainty, thereby reducing the robustness of endpoint comparability. Pseudo-ALI and oropharyngeal aspiration were therefore chosen as pragmatic, dose-controlled exposure approaches that reduce technical variability and facilitate cross-system comparison. The principle of pseudo-ALI [38] exposure is to deposit a small volume of the particle suspension directly on top of the co-culture model, ensuring a precise and high concentration of particles deposited. This simulates oropharyngeal aspiration, where depositing a suspension of particles into the trachea leads to more control over the deposited dose and more uniform deposition in the alveolar region compared to intra-tracheal instillation [52]. Both methods are more easily reproducible, as they do not require high-level, complex, and expensive equipment, and therefore show potential for standardization when achieving consistent experimental outcomes. The challenge of low in vivo relevant doses failing to elicit robust effects in vitro using various materials, such as SiO_2, is a limitation commonly reported in the literature [53]. One explanation for this is the reactivity of DQ₁₂ and how the particles are produced, which influences their biological effect, and the surface reactivity of freshly fractured silica particles, which influences their toxicity [54, 55]. Besides the surface of the particles, the chemical composition can also influence toxicity; e.g., a real-life sample with Al and Fe in traces causes a lower response than standard quartz [56]. Therefore, in-depth characterization of the material to be analyzed is essential before conducting experiments. This can also significantly contribute to the development of Quantitative Structure-Activity Relationship (QSAR) models in the future, which will aid in predicting biological effects based on the material’s structure, serving as an additional tool for toxicity assessment [57]. Another important reason for the importance of in vitro dose setting becomes obvious when considering the differences between in vivo situations and the substantially less complex in vitro models. These differences affect the individual exposure of target cells in the in vitro and in vivo test system, leading to entirely different toxicokinetic and toxicodynamic, and consequently leading to different cellular key events progressing toward adverse outcomes when comparing a complex organ with different cell types to a much simpler in vitro model [58]. Approaches to minimize or overcome these differences are improved with the validation of more complex (or more fitting) in vitro models and accurate in vivo to in vitro dose extrapolations. In our study, even with significantly higher doses of up to 300 µg/cm² DQ₁₂ over 24 h, we observed an absence of response at both the gene expression and cytokine secretion levels, apart from IL-6 secretion, in our in vitro lung model. Despite the well-documented inflammatory potential of DQ₁₂, our initial 24-hour experiments showed no significant inflammatory response. Therefore, we explored whether the absence of a measurable short-term inflammatory response could indicate a later or progressive inflammatory response. We designed our in vitro experiments to mirror the in vivo exposure conditions while measuring identical endpoints across both models. In addition, the same DQ₁₂ batch was used in vitro and in vivo. This approach enabled tracking the same inflammatory markers and cellular responses over a longer period in a single study, which was not reported previously. Extending the post-exposure duration revealed a progressive inflammatory response, mirroring in vivo observations. Potential mechanisms underlying the timing of the inflammatory response observed in the in vitro model could include the time required for particle internalization by macrophages and the particles’ surface-based reactivity [55], which together determine the specific toxicokinetic and toxicodynamic profiles. Therefore, to improve the in vitro study design, optimizing both dose levels and post-exposure durations was needed to capture the dynamic inflammatory response [59].

AOP and biomarkers

The application of IVIVE models to predict the toxicity of certain material depends on identifying the AOP-relevant in vitro bioassays; in our study, we selected AOP173, which links early inflammatory responses in the respiratory system to downstream adverse outcomes of fibrosis, which makes it relevant for DQ₁₂ exposure [23]. This choice guided our study design, focusing on measuring key inflammatory events such as cytokine release (IL-6, IL-1β, and IL-8), recruitment of inflammatory cells, and markers of epithelial barrier integrity. These endpoints align with the AOP’s critical molecular and cellular events, ensuring that our findings are directly relevant to the mechanistic understanding of particle-induced lung toxicity. Subsequently, we identified several key biomarkers for the three earliest key events of AOP173 that reflect early inflammatory processes. For the KE1496, we measured gene expression of IL-8, IL-6, and IL-1β from rat lung tissue and in vitro cell culture. IL-6 and IL-1β were particularly valuable due to their conserved expression across species and their roles in early inflammatory processes, with IL-6 acting as a driver of fibrosis pathophysiology [60] and IL-1β promoting inflammation and fibroblast activation [61], making them essential biomarkers for linking early inflammation to downstream fibrotic outcomes in the AOP framework. IL-8, while not expressed in rodents, and its receptors, CXCR1 and CXCR2, have a significant role in various pulmonary diseases and neutrophil activation [62, 63]. Therefore, this cytokine could be used to evaluate both events: KE1496 increased pro-inflammatory markers and KE1497 recruitment of inflammatory cells. We tested several rat homologs of IL-8, such as CXCL1, CXCL2, and CXCL5/6, as well as the human and rodent receptors for IL-8, CXCR1, and CXCR2, to see which of these markers in rats correlated best with IL-8 expression in vitro [64]. Moreover, there was significant secretion of IL-8 in the supernatant compared to increased differential cell counts in BALF, particularly a significant increase of neutrophils on the third day. For the KE1498, loss of membrane integrity, we included epithelial integrity marker TJP1 gene expression since it can be measured in both systems besides TEER measurement, which is routinely done in vitro, and histology analysis, which is normally done in vivo to assess membrane disruption. While in vivo, the tissue damage was evident, in vitro, we observed a decreased trend that was not statistically significant. This might be explained by the fact that the co-culture model only features two cell types: epithelial cells and macrophages, which do not fully reflect the in vivo complexity, where, for example, activated neutrophils can significantly contribute to tissue damage.

Several studies compared in vitro and in vivo pulmonary responses for inhalation toxicity testing, yet reliable IVIVE remains challenging. Loret et al. demonstrated that ALI exposure enhances in vitro predictivity; however, their study was limited to acute (24-hour) responses and did not assess longer-term inflammatory processes [13]. Similarly, Sauer et al. classified 19 mostly water-soluble chemicals based on acute toxicity predictions using the EpiAirway^TM model, focusing primarily on cytotoxicity and TEER measurements without incorporating detailed inflammatory marker analyses [65]. Building on this, Jackson et al. conducted a pre-validation study using the EpiAirway^TM model with a broader set of 59 chemicals, the largest test set to date, which included more water-insoluble compounds [66]. However, their study primarily aimed at identifying hazardous chemicals and respiratory tract corrosives, whereas our research focuses on particle toxicity. One study examined the effects of nano-sized titanium dioxide on the EpiAirway^TM tissue model using the same 3-hour exposure profile as the aforementioned chemical studies, but it did not include in vivo results for comparison [67]. More broadly, exposure durations in organotypic lung models cultured at the air–liquid interface span a wide temporal range, from acute exposures of hours to sub-chronic designs extending over several weeks [68], depending on the biological question and model stability. Short-term exposures (typically ≤ 24–48 h) are frequently employed to characterize acute stress responses and early pathway perturbations, whereas longer exposure periods are used to investigate cumulative inflammatory effects, tissue remodeling, or disease-relevant mechanisms. However, extended culturing time is limited by model-specific constraints on tissue integrity. In this context, our study applies a 7-day post-exposure time, representing an intermediate duration that allows assessment of delayed and sustained inflammatory responses. This approach is particularly relevant for particle-induced effects, which may involve modes of action and temporal dynamics distinct from those of soluble chemicals. Unlike studies that primarily assess cytotoxicity by MTT assay or histopathological evaluations, our approach examines the dynamic progression of inflammation over time with unique one-to-one comparisons to the in vivo results. Studies that use simple toxicity assays may overlook the importance of the multifactorial nature of inhalation toxicity, which requires a deeper understanding of its underlying mechanisms. This underscores the importance of developing pathway-based predictive approaches, such as AOPs, to more effectively assess inhalation toxicity [69, 70]. While AOPs have been theoretically proposed and specific in vitro models and assays have been outlined to evaluate nanomaterial-induced fibrosis [71], their practical implementation remains limited. In contrast, this study demonstrates that AOPs are not merely conceptual tools but can be integrated with experimental data, highlighting their practical use. Our application of AOP173 represents an innovative step in establishing a more robust framework for predicting long-term health risks. Moreover, this approach leaves space for using multiple pre-validated models in the future to predict other relevant key events beyond the initial inflammation and complement the whole AOP.

While our study demonstrates a promising comparison between in vitro and in vivo inflammatory responses by establishing a successful correlation of highly relevant biomarkers, a more rigorous and standardized correlation analysis framework is needed to advance regulatory acceptance of in vitro methodologies. For example, integrating omics data could enhance the complexity and comparability of biomarker measurements across systems. Standardized reporting systems like the OECD’s Omics Reporting Framework (OORF) [72] are some of the initiatives to standardize omics reporting and incorporate these advanced methodologies into toxicological studies.

Beyond cytokine measurements, future studies should explore additional endpoints, such as oxidative stress markers and fibrosis mediators [73], which could further test the predictive power of in vitro models used for IVIVE. Moreover, although comparable results for inflammatory response were observed between the in vitro model using pseudo-ALI exposure and the rat model that uses oropharyngeal aspiration as a method of exposure, both approaches typically employ particle doses higher than those expected under realistic environmental exposure conditions [74]. The use of more realistic exposure scenarios should be explored and considered to further improve the translational relevance of in vitro models to humans, such as aerosol delivery systems that better mimic environmental exposures. In vivo tools, such as the Ventilator-Assisted Aerosol Delivery (VAAD) machine [75], developed by Scireq for precise dose delivery directly to the trachea of rodents, could serve as a benchmark for comparison with corresponding in vitro aerosol exposure systems. In this study, generating our own in vivo data in addition to using data from the existing literature was necessary to ensure accurate and complete verification and control of the protocols and materials used. Relying solely on results from published studies remains problematic, given the incomplete reporting of methods and materials in the existing literature. Authors of in vitro and in vivo studies alike should therefore strictly adhere to the general reporting guidelines proposed by platforms such as SciRAP [76, 77].

Finally, extending the approach to a broader range of particle types, including standard reference materials and real-life samples, as well as incorporating endpoints linked to chronic outcomes such as fibrosis, would contribute to the validation of our model as a reliable alternative for assessing particle-induced effects.

Conclusion

This study demonstrates that aligning in vitro and in vivo experimental designs enables identification of inflammatory endpoints that show concordance across systems, thereby supporting quantitative IVIVE. Using DQ₁₂ as a model material, we show that an in vitro exposure strategy incorporating multiple post-exposure points captures biologically meaningful inflammatory responses that are predictive of in vivo outcomes. These findings support the utility of advanced in vitro lung models for improving the translational relevance of non-animal approaches in inhalation toxicology.

Abbreviations

ALI

Air-Liquid Interface

AOP

Adverse Outcome Pathway

AOP173

Adverse Outcome Pathway 173 (specific pathway focusing on lung fibrosis progression)

BALF

Bronchoalveolar Lavage Fluid

BSA

Bovine Serum Albumin

CCM

Complete Cell Culture Medium

cRPMI

Complete Roswell Park Memorial Institute Medium

CXCR

C-X-C Chemokine Receptor

DAPI − 4',6

Diamidino-2-phenylindole

DLS

Dynamic Light Scattering

DQ₁₂

Crystalline Quartz Silica Particles

ELISA

Enzyme-Linked Immunosorbent Assay

FBS

Fetal Bovine Serum

GAPDH

Glyceraldehyde-3-Phosphate Dehydrogenase

HBSS

Hanks’ Balanced Salt Solution

IgG

Immunoglobulin G

IL

Interleukin

IVIVE

In Vivo-In Vitro Extrapolation

KE

Key Event

LDH

Lactate Dehydrogenase

MDMs

Monocyte-Derived Macrophages

MEM

Minimum Essential Medium

MIE

Molecular Initiating Event

MDM

Monocyte-Derived Macrophages

M-CSF

Macrophage Colony-Stimulating Factor

NEAA

Non-Essential Amino Acids

OECD

Organization for Economic Co-operation and Development

OORF

Omics Reporting Framework

PATROLS

Physiologically Anchored Tools for Realistic Nanomaterial Hazard Assessment

PBS

Phosphate-Buffered Saline

qRT-PCR

Quantitative Reverse Transcription Polymerase Chain Reaction

QSAR

Quantitative Structure-Activity Relationship

RPMI

Roswell Park Memorial Institute Medium

SD

Standard Deviation

SEM

Scanning Electron Microscopy

SOP

Standard Operating Procedure

TEER

Transepithelial Electrical Resistance

TEM

Transmission Electron Microscopy

TG

Test Guideline

TJP1

Tight Junction Protein 1

VAAD

Ventilator-Assisted Aerosol Delivery

ZO-1

Zonula Occludens-1

Author contributions

IL: Writing-Original Draft Preparation, Experimental Design, Data Analysis, Acquisition, Investigation, Interpretation and Visualization. FB and BRR: Conceptualization, Methodology, Validation, Investigation, Resources, Reviewing- Original draft preparation, Supervision, Project administration. Writing-Review and Editing: IL, SM, FB, BRR, APF SM and MD: Experimental Design (*in vivo*), Data Analysis, Acquisition, Interpretation, and Reviewing. SB: Advanced Statistical Analysis. All authors read and approved the final manuscript.

Funding

This research is supported by the Swiss National Science Foundation NRP 79 program (Nr. 407940_206331/1) and the Adolphe Merkle Foundation.

Data availability

The datasets generated during the current study are available in the Zenodo repository [DOI: 10.5281/zenodo.14916634].

Declarations

Ethics approval and consent to participate 

Experiments involving primary monocyte isolation from human blood were approved by the committee of the Federal Office for Public Health Switzerland (reference number: 611-1, Meldung A110635/2) for the Adolphe Merkle Institute. Animal experiments were approved by The Swiss Federal Veterinary Office of the Cantonal Ethical Committee for Animal Experiments (Amt für Landwirtschaft und Natur des Kantons Bern) under permission number BE 90/2022. All rats received humane care in compliance with the “Principles of Laboratory Animal Care” formulated by the National Society for Medical Research and the “Guide for the Care and Use of Laboratory Animals” prepared by the Institute of Laboratory Animal Research. All experiments were performed following the European Convention of Animal Care standards.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.