Introduction

Inhaled drug delivery is an effective treatment paradigm for respiratory diseases, such as asthma and chronic obstructive pulmonary disease (COPD), due to direct access to the intended target organ and the potential to minimize systemic exposure and consequent side effects (Ruge et al. 2013). The development of innovative inhaled medicines is often impacted by toxicity in the respiratory tract in non-clinical species, leading to attrition or dose limitations in clinic, typically after the selection of novel drug candidates (David et al. 2014). As part of the industry shift from observational to predictive toxicology, a proactive discovery safety strategy aims to identify and mitigate these safety risks early during drug discovery, and thereby enable the design and selection of drug candidates with the right safety profile (Hornberg et al. 2014a; Hornberg and Mow 2014; Morgan et al. 2018). This relies heavily on in vitro methods that are predictive for in vivo toxicity, clinical safety and compatible with the design-make-test-analyze (DMTA) cycle during lead optimization campaigns (Hornberg et al. 2014b; Johansson et al. 2019). A plethora of high-content screening (HCS) methodologies have been developed and deployed to successfully identify for example cardio-, hepato-, neuro-, and nephrotoxicity (Li and ** the modelled ELF concentrations onto the dose response curve. Interestingly, we found that by map** the modelled ELF concentrations onto the occludin dose–response curve, we could show that pathological changes in vivo were only observed in areas of the respiratory tract that were exposed to drug concentrations (14 µM) similar to that of the in vitro IC50 (27 µM) (Fig. 5c). Equally, no pathological changes were observed in areas of the respiratory tract that were exposed to concentrations (3.2 & 0.02 µM Cmax) that did not elicit an effect in the occludin assay. This suggests that a combination of PBPK modelling and this novel in vitro assay can inform translational and quantitative risk assessment for inhaled small molecules.

Discussion

Pharmaceutical R&D productivity requires a proactive discovery safety strategy to identify and mitigate safety risks early to bring forward candidate drugs with the Right Safety profile (Hornberg and Mow 2014; Morgan et al. 2018). To inform decision-making during drug discovery, robust in vitro assays are required that are compatible with integration in the DMTA cycle namely being cost-effective, have sufficient throughput, and with readouts that are predictive for (non)clinical safety (Blass 2021; Hornberg et al. 2014b; Johansson et al. 2019; McKim 2010; Sanders et al. 2017; Szymanski et al. 2012). Currently, a range of organ-specific assays exist, which use a single cell type rather than attempting to fully recapitulate the complexity of the organ they represent, several of which are applicable to early drug screening, for example those aimed towards the cardiovascular system, liver and kidney (Gustafsson et al. 2014; Persson et al. 2013; Pognan et al. 2023; Pointon et al. 2017; Sjogren et al. 2018). Options to screen inhaled drugs for lung toxicity however have lagged behind. Some bronchial cell-based assays have shown some promise when grown in ALI conditions, such as our predictive TEER-based ALI assay for the detection of lung toxicants (Balogh Sivars et al. 2018) and an imaging-based assay correlating junctional protein staining with TEER measurements as a primary readout of lung barrier integrity (Pell et al. 2021). Recently published breathing lung-on-chip models have also shown significant promise as an in vitro tool to study inhalation toxicity (Sengupta et al. 2022). However, these assays come with significant limitations in terms of throughput, cost, and their ability to stratify compounds within a chemical series. Relevant comparative assay details are listed in Table 2. A549 cells have been widely used as a cell model to study toxicological responses, such as cytotoxicity, oxidative stress and inflammatory responses (Barosova et al. 2021). However, TEER values from these cells when elevated to ALI conditions are far lower than alternative alveolar cell lines such as the lentivirus immortalized hAELVi cells. Indeed hAELVi cells have also been shown to possess a more type I phenotype than the type II of the A549 cells and thus representing a larger proportion of the alveolar population (Kuehn et al. 2016). In an effort to develop an assay which can be rapidly deployed for screening within a DMTA cycle, we focused on measuring a single endpoint from a widely studied cell line which, whilst not recapitulating the full complexity of the lung alveolar space allowed for robust measurement of a physiologically relevant parameter such as junctional protein morphology. Here, we showed that quantification of occludin staining in A549 cells in a simple 2D format can detect lung toxicants/irritants with 90% sensitivity and 100% specificity after 24 h compound administration. This is likely driven by the central role played by junctional proteins in maintaining an effective epithelial barrier in the lung. Occludin is a classic tight junction protein which functions to link cells to adjacent cells and to the internal cytoskeletal structures (Kojima et al. 2013). It has been shown to have a clear link to barrier maintenance, ATP production and gene regulation (Castro et al. 2018), whilst studies with occludin knock-out mice have also shown chronic inflammation amongst other pathologies, indicating not only a wide role for occludin within cellular homeostasis but also a pivotal link to inflammation and pathological processes (Saitou et al. 2000; Sugita and Kabashima 2020).

Table 2 Assay comparison to ALI-based TEER measurements
Full size table

To develop an imaging-based assay measuring localization of a tight junction protein in a simple 2D culture, growth conditions of the chosen cell line were found to be of utmost importance. Specifically, in the development of this assay, it was found that the morphological characteristics of the bronchial cell lines 16HBE and Calu3 were incompatible with the requirements of the imaging algorithm. In order for robust measurement of the occludin staining pattern, it was necessary to use a whole cell stain to allow for determination of cell edges independent of the junctional marker to narrow the region of measurement to those edges. In the case of Calu3, these cells displayed a highly overlap** growth pattern, tending to form multilayered colonies rather than a uniform monolayer of cells. 16HBE, on the other hand, grew in a highly consistent monolayer fashion but displayed an undulating pattern of cell-to-cell contact rather than a liner line. These inherent conditions posed a challenge when it came to determining the degree of overlap of the junctional marker with the defined cell edge. Alternative seeding densities and culture time points were explored (data not shown). However, A549 cells were consistently found to provide the most robust basis to enable screening assay development.

In order to apply in vitro toxicology assays to quantitative risk assessment, it is essential to incorporate drug exposure information and establish how in vitro concentrations relate to clinical exposure (Bell et al. 2018). While there are challenges to scale compound concentrations and elicited effects in in vitro assays to plasma drug exposure and adverse events in the clinic, it has been possible to use in vitro assay data for quantitative risk assessment of some organ toxicities (Albrecht et al. 2019; Archer et al. 2018; Hengstler et al. 2020; Kappenberg et al. 2020; O'Brien et al. 2006; Persson et al. 2013; Sjogren et al. 2018). Quantitative risk assessment has been more challenging for inhaled drugs because, unlike for orally administered drugs, the systemic plasma exposure at a given inhaled dose does not accurately reflect the relevant exposure in the lung (Frohlich 2019; Kassinos et al. 2021). To address this gap, we applied an existing PBPK modelling approach, which mathematically describes respiratory tract physiology and the fate of deposited drug dose, and accurately predicts PK, target site exposure, and interestingly, spatial heterogeneity in target site concentrations within the lung (Boger et al. 2016; Boger and Friden 2019). This approach allowed us to directly correlate actual compound concentrations from the in vitro assay with calculated compound exposure in the ELF in a toxicity study and demonstrate that the pathological changes consistent with airway irritation were restricted to areas of the respiratory tract that reached concentrations that aligned with the IC50 of the imaging assay. While recent studies have shown the applicability of modelling to in vitro to in vivo extrapolation (IVIVE) for toxic vapours from agrichemicals (Moreau et al. 2022) and environmental contaminants such as endocrine disrupters (**e et al. 2023), to our knowledge, this is the first-time in vitro toxicity data have been quantitatively linked to exposure levels in the lung for compound risk assessment in drug development.

The propensity of a compound to cause to toxicity may depend on primary target pharmacology, off-target secondary pharmacology and chemical reactivity, which originate from its chemical structure and physicochemical properties, and can manifest in many different downstream effects in the cell, including direct damage to organelles or DNA, membrane permeability, impaired cellular signalling or metabolism. In general, assessing cell health parameters has proven an effective trade-off between, on the one hand, ensuring high sensitivity (compared to more generic cell survival or cytotoxicity assessment) and, on the other hand, providing robust, accessible screening options with throughput (compared to more specific but therefore less-sensitive molecular endpoints). Here, the validation compound set encompassed diverse physicochemical properties, therapeutic targets, off-target pharmacology, and a range of pathologies across multiple pre-clinical tox species, including epithelial erosion, atrophy, oedema, hyperplasia and inflammation. In summary, we present a novel imaging-based assay that can be applied to identify and mitigate the risk for lung toxicity of inhaled compounds during early drug discovery, and for quantitative risk assessment in drug development. This will enable acceleration of the identification of inhaled drug candidates with the right safety profile without the need for animal studies.