An in vitro coculture system for the detection of sensitization following aerosol exposure 1

The aim of the study was to develop an in vitro model that mimics the alveolar-capillary barrier and that allows assessment of the respiratory sensitizing potential of respiratory sensitizers. The 3D in vitro model cultured at the air liquid interface consists of alveolar type II epithelial cells (A549), endothelial cells (EA.hy926), macrophage-like cells (PMA-differentiated THP-1) and dendritic-like cells (non-differentiated THP-1). This alveolar model was exposed apically to nebulized chemical respiratory sensitizers (Phthalic Anhydride (PA) and TriMellitic Anhydride (TMA)) or irritants (Methyl Salicylate (MeSa) and Acrolein (Acr)) at concentrations inducing at maximum 25% of cytotoxicity. The exposure to respiratory sensitizers induced dendritic cells activation and a specific cytokine release pattern, while the irritants did not. In addition, the cell surface marker OX40L was determined for dendritic like cells activation to identify high molecular weight allergens. With this in vitro model we can postulate a set of promising markers based on the studied compounds that allow the discrimination of chemical respiratory sensitizers from irritants.


Introduction
Due to their peculiar toxicity and systemic effects, respiratory sensitizers are receiving increasing attention within toxicity evaluation studies and risk assessment.In certain cases, as under article 57 of the EU REACH regulation, respiratory sensitizers can be considered as substances of very high concerns (SVHC) following the principle of equivalent level of concern to carcinogens, mutagens, substances toxic for reproduction, PBT substances (persistent, bio-accumulative and toxic) and vPvB (very persistent and very bio-accumulative) substances (EC, 2006).The inclusion of a substance on the ECHA's list for SVHC does not translate in the automatic ban for the substance, but it implies that additional documentations and authorizations are required prior of the placing on the market, use or import of the substance, with high expenditure for the company and high social impact.Respiratory sensitization to chemical (or Low Molecular Weight; LMW) compounds mostly occurs at the workplace, leading to the development of occupational allergies such as allergic asthma, rhinitis and conjunctivitis (Kimber et al., 2010).Asthma alone represents the most prevalent occupational disease of the lungs in developed countries and it is worth noting that occupational exposures to dusts, gases, fumes, vapours and chemicals are responsible of 16.3% of asthma in adults (Torén and Blanc, 2009).Around a hundred of chemicals were described to act as respiratory sensitizers (Bloemen et al., 2009), among which different chemical classes such as acid anhydrids, diisocyanates and chloroplatinate salts were identified.Protein allergens (or High Molecular Weight; HMW compounds), which are also a potential cause of occupational asthma, are the primary cause for the development of respiratory allergy in the general population.Indeed, the exposure to certain proteins occurring in environmental sources such as pollen, animal dander and house dust mites (HDM) are common causes of asthma.Depending on the geographical location, 50 to 85% of people with asthma are allergic to certain HDM proteins (Gregory and Lloyd, 2011).Despite the health effects and social and economic impact, the early identification of the compounds with the potential to act as respiratory sensitizers is still difficult.This is due to the lack of complete understanding of the systemic mechanisms involved in the development of respiratory sensitization and in the absence of fit for purpose validated or even widely accepted in vivo models or in vitro assay to identify respiratory sensitizers.Currently, the respiratory sensitizing potential of compounds can only be assessed using the rodent LLNA (Local Lymph Node Assay) and is thus completely dependent on in vivo methods.Significant differences in the pulmonary and immune systems between rodents and humans make the extrapolation to humans really difficult.Even if it is generally accepted that LMW respiratory sensitizers are positive in the LLNA as both skin and respiratory sensitizations share similar mechanisms during induction phase, it is not possible to distinguish respiratory from skin sensitizers using this assay.Also, the route of exposure is still under debate for the investigation of respiratory sensitizers using the LLNA.In addition to the LLNA more work is needed to classify a compound as a respiratory sensitizer such as the evaluation of the cytokine fingerprint and the IgE serum levels.Thus it is still not known whether the LLNA could predict the potency of chemicals to sensitize the respiratory tract (Anderson et al., 2011).Regarding the ethical and economic issues raised by the use of in vivo methods, the development of in vitro alternatives to study respiratory sensitization has been intensified.
Dendritic Cells (DCs) represent a key player of the respiratory sensitization process.For this reason, an in vitro model for identification of respiratory sensitizers should definitely be based and rotate around DCs.However, events triggered at the alveolar epithelium and the microenvironment created by the other cell types present at the alveolar region play an important role for the development of respiratory sensitization.For instance, the alveolar epithelium provides the major source of danger signals that directs the immune response and it acts as the first line of defence against airborne pathogens and xenobiotics.With a surface area up to 100 m 2 , alveoli represent the largest exposure area of the lung.Besides the alveolar epithelial barrier, the endothelium of the capillary also acts as structural barrier in the alveoli.In the lower airways, a network of DC is directly located above and beneath the basement membrane.From this "launch base" DCs project dendrites to sample and uptake foreign material (Lambrecht and Hammad, 2003;van Rijt and Lambrecht, 2005;Holt, 2012;van Helden and Lambrecht, 2013).Upon uptake, DCs can undergo a process of activation and maturation during which they express costimulatory molecules and release cytokines and chemokines (Fig. 1A).Macrophages present in the alveolar interstitium and on the alveolar surface, not only contribute to the first line defence against inhaled materials but may also differentially affect DC function and the induction of adaptive immunity.Interstitial macrophages can prevent DC activation and maturation upon allergen encounter through an IL-10 dependant manner (Sibille and Reynolds, 1990;Bedoret et al., 2009;Toussaint et al., 2013;Lauzon-Joset et al., 2014).Differently, alveolar macrophages, which are mostly involved in the removing of foreign material from the alveolar space, can promote an early inflammatory response.Thus, macrophages represent a heterogeneous population that is highly versatile in its responses to foreign materials, which helps maintaining the immune homeostasis in the airways (Chung and Adcock, 2014).In this work we investigated the in vitro effects of respiratory sensitizers (LMW: Phthalic Anhydride (PA) and TriMellitic Anhydride (TMA); HMW: Bet v1 purified protein and HDM extract) and irritants (Methyl Salicylate (MeSa) and Acrolein (Acr)) at the lung-blood barrier using a novel 3D in vitro coculture model developed by our group on the basis of the lung in vitro model for respiratory irritation developed by Klein et al. (Klein et al., 2013, 2017;Fizeșan et al., 2018) (Fig. 1B).The model for respiratory irritation combines alveolar type II epithelial cell line A549 grown at the Air-Liquid-Interface (ALI) on a porous membrane of a Transwell™ insert, monocyte cell line THP-1 cells differentiated into macrophage like cells, endothelial cells EA.hy926 and human mast cell line HMC-1.This model was originally developed to assess the toxic effects of particles at the alveolar barrier.The possibility to expose the system at the ALI represents an advantage as it mimics the in vivo situation.However, the original model from Klein et al. (2013) is not suitable for of respiratory sensitization, since it lacks DCs, which, as previously described, are pivotal for the respiratory sensitization process.Due to this limitation, further modifications, such as the addition of DCs in the basolateral compartment, simplifications, such as the removal of mast cells from the model, and optimisations, such as the increase of the insert pores' size to allow migration of cells, were necessary in order to generate a suitable in vitro model for the identification of respiratory sensitizers (Fig. 1C).
It has been suggested that respiratory sensitizers may elicit a response in in vitro assay for skin sensitization (Basketter et al., 2017).Even if most of chemical respiratory sensitizers tested were positive in assays built for skin sensitization, data are still missing and several chemical respiratory sensitizers are not evaluated correctly using skin models.Furthermore, this approach does not allow the differentiation of skin vs. respiratory sensitizers.Different DC-like cell lines were previously described to assess the sensitizing potential of chemicals, among which THP-1 cells are used for the human cell line activation test (h-CLAT), which represents the first in vitro model validated in 2017 by OECD to replace in vivo testing for skin sensitizers (Ashikaga et al., 2006;OECD, 2017).For this reason, THP-1 cells are used as a model for DCs in our in vitro model for respiratory sensitization.In this study, we intended to investigate the upregulation of cell surface markers (i.e.CD86 and CD54, also called B7-2 and Intracellular Adhesion Molecule-1 (ICAM-1), respectively) on THP-1, as described for the h-CLAT model, together with other surface markers, soluble mediators (e.g.cytokines and chemokines) and gene expression.
The newly developed lung in vitro model was designed to allow the prediction of the respiratory sensitization potential of inhaled compounds and we also wanted to identify additional markers that increase the accuracy and sensitivity of the system.During the characterization of the model, we investigated the effects of the microenvironment generated by the presence of endothelial, epithelial and macrophages on dendritic cells.Indeed, alveolar epithelial cells, together with macrophages, constitute the first line of defence of the alveoli from exogenous material.The role of the alveolar epithelium, in close physical association with DCs, is not only restricted to its barrier role, but it also represents a central element for the process of sensitization by regulating the function of DCs through the release of cytokines which promotes Th2 responses.The airway epithelium is the major source of cytokine and chemokine production in the lung and the resulting microenvironment strongly influence DCs maturation and the outcome of the immune response (van Rijt and Lambrecht, 2005;Hammad and Lambrecht, 2011).The presence of epithelial cells (ECs) up-regulate, as compared to the THP-1 cultured alone, the expression of a wide variety of cytokines, such as CCL20 or Granulocyte-macrophage colony-stimulating factor (GM-CSF), which recruit and activate DCs to drive TH2-mediated immune response (Schuijs et al., 2013;Lambrecht and Hammad, 2015).
Promising models have been developed in the last years to study respiratory sensitization to chemicals (Huang et al., 2013;Dik et al., 2015;Forreryd et al., 2015;Hermanns et al., 2015;Mizoguchi et al., 2017) but they are targeting only one of the key steps undertaken in the mechanism of respiratory sensitization.It is widely assumed that only one in vitro assay would not be sufficient to determine by itself the respiratory sensitizing potential of compounds.For instance for the assessment of the potential of a chemical to trigger skin sensitization, the use of an Integrated Test Strategy (ITS) by using validated models responding at least to two of the key events of the AOP built for skin sensitization proved to accurately determine the sensitizing potential of chemicals (Clouet et al., 2017;Otsubo et al., 2017;Strickland et al., 2017;Ohtake et al., 2018).The concept of adverse outcome pathways (AOP) became these last years a considerable tool of interest by outlining key events leading to an adverse heath outcome, which then provides a basis for focusing future research and developing alternative methods for hazard characterisation.An AOP for respiratory sensitization was recently proposed based on key steps present in the AOP for skin sensitization but with specific differences (Kimber et al., 2014;Sullivan et al., 2017).This AOP includes (1) a molecular initiating event with the protein-binding reactions, followed by cellular key events corresponding to the (2) epithelial response with the release of danger signals and cytokines, (3) the DC activation which preferentially promotes a TH2 pro-environment and (4) finally the TH2 cell proliferation (Sullivan et al., 2017).By measuring cytokines released at the alveolar barrier and by evaluating the activation of DCs, our model is already able to answer 2 of the 4 key events of the AOP.Several parameters could be added to complete the set of markers to address better the different key events involved in respiratory sensitization.

Cell lines identity
The correct identity of each cell line (A549, THP-1 and Ea.hy926) used in the present work was confirmed via Human STR profiling cell authentication service provided by American Type Culture Collection (ATCC, Manassas, VA, USA).Briefly, about 40 µL of cell suspension (containing at least 1 x 10 6 cells) were spotted on the Sample Collection Card provided by ATCC.The sample collection cards were allowed to dry and then shipped back to ATCC for the genotyping of the cell lines.

Cell lines
The human cell lines A549 (Lieber et al., 1976), THP-1 (Tsuchiya et al., 1980) and EA.hy926 (Butterfield et al., 1988) were obtained from ATCC.The cells were cultured using different media (Table 1).Adherent cell lines were trypsinized twice a week and medium was changed every other day.Cells were maintained in a humidified atmosphere with 5% CO2 at 37 °C and tested regularly for mycoplasma contamination.

Coculture workflow
On day 0, THP-1 cells differentiation into macrophage-like cells was started by seeding 4 × 10 5 cells/mL in complete medium with addition of phorbol-12-myristate-13-acetate (PMA; 20 ng/mL; Deisenhofen, Germany) and incubation over night at 37 °C and 5% CO2.The differentiation of PMA treated cells was enhanced on day 1 by removing the PMAcontaining media and replacing it by fresh THP-1 complete medium for a further 5 days.On day 2, EA.hy926 and A549 cells were seeded on Millipore cell culture inserts (surface area of 4.5 cm 2 ; 5 μm pore size; high pore density PET membranes for 6-well plates; Millipore, Molsheim, France).EA.hy926 cells were seeded on inverted Transwell™ inserts (2.4 x 10 4 cells/cm 2 ).Upon attachment on the basolateral side of the Transwell™ insert, the plate with the transwell inserts was turned back before the A549 cells were seeded inside the transwell (6 x 10 4 cells/cm 2 ).EA.hy926 and A549 cells were grown for three days at 37 °C and 5% CO2 in a humidified incubator.On day 5, medium was replaced by complete coculture medium containing 10% of FBS.On day 6, differentiated THP-1 (2.4 x 10 4 cells/cm 2 ) cells were added into the inserts on the top of the confluent layer of A549.Upon attachment, typically after 4 hours, medium from the upper compartment was removed and the coculture was cultivated overnight at the ALI until exposure (Fig. 2).

THP-1 cell migration assay
The ability of THP-1 cells to migrate from one side to the other of a Transwell™ (e.g. from the basolateral compartment to the apical compartment) was tested using 1, 3, 5 and 8 µm pore size Transwell™ inserts (Millipore).1,5 x 10 6 cells/mL were seeded in the apical side of each insert.0; 50 and 100 ng/mL of chemoattractant (MCP-1 -Monocyte Chemoattractant Protein-1) (Sigma) were spiked in THP-1 medium at the basolateral side of the inserts.After 2h, cells that migrated through the membrane were counted microscopically.Trypan blue was used for the determination of dead cells (dilution 1:2 v/v in trypan blue).The percentage of cell migration was calculated as number of cells that migrated / total number of cells x 100.
Exposure of the monocultures THP-1 cells were cultured in THP-1 medium (Table 1).THP-1 cells (1x10 6 cells/mL) were exposed to different concentrations of chemicals and to their vehicle controls (Phthalic Anhydride (PA), TriMellitic Anhydride (TMA) and Methyl Salicylate (MeSa) in DMSO and Acrolein (Acr) in water, respectively).Concentration of vehicle control did not exceed 0.2%.

Assessment of cell viability
24h after exposure, the cell viability of the cocultures was assessed using 400 µM of Alamar Blue diluted in culture medium.After 1h incubation at 37 °C in the dark, medium from the apical and basolateral compartments was collected and the fluorescence measured at 530 nm excitation and 590 nm emission using a fluorescence microplate reader (Spark 20M, Tecan, Mechelen, Belgium).Filters used were 530/25 (excitation) and 590/20 (emission) respectively.
Cytotoxicity on THP-1 cells was assessed after 24h exposure to chemicals using 1µM Sytox Blue (Thermofischer).Sytox blue positive cells were measured using cytofluorimetric analysis on a SLR Fortessa system (BD Biosciences, Heidelberg, Germany) and analysed using the FlowJo software (V10) (Ashland, Oregon -USA).

Aerosol exposure
The Vitrocell® Cloud-6 system (Vitrocell®, Waldkirch, Germany) was used for the exposure of the coculture in vitro model to test compounds and vehicle controls, which were considered as negative controls.Compounds were diluted in 50% (v/v) sterile water in PBS (1X).PA, TMA and MeSa were dissolved in DMSO prior of the dilution in water/PBS solution.The cocultures were exposed to different concentrations of the compounds in their vehicle controls (PA, TMA and MeSa in DMSO and Acr, Lipopolysaccharide (LPS), Bet v1 and HDM in water).Briefly, the system makes use of a vibrating membrane to generate the aerosol from liquids and suspensions in a closed chamber.The device comprises a cloud aerosol chamber on the top of which the cloud of aerosol is generated and allowed to settle on the cultivation module, usually within a period of 15 minutes, allowing the simultaneous exposure of up to six cell culture inserts.The system is designed in order to ensure for generated aerosol cloud to evenly distribute inside the chamber.The nebulized cloud allows a dose-controlled and spatially uniform aerosol deposition on the top of the inserts.In order to keep the cells viable, the module is heated at a steady temperature of 37 °C.
Purified natural Bet v1 (Indoor biotechnologies, Cardiff, United Kingdom) from birch pollen was nebulized on the inserts at a concentration of 0,02 µg/cm 2 .
Surface marker expression was measured using LSRFortessa cell analyser and analysed with the FlowJo software V10.Sytox Blue-positive cells (dead cells) were excluded from the analysis.Relative geometric mean fluorescence intensities (rMFI) were expressed in % and calculated as follow: Cytokine release 24 hours after exposure, aliquots from the medium in the basolateral compartment were collected and GM-CSF, CCL20, IL-6, IL-10, IL-7, MCP-1, RANTES, sCD40L, IL-33, IL-25 and IL-1α were quantified.Quantification was performed on a Luminex 200™ (Luminex Corporation, Oosterhout, the Netherlands) using the MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel -Immunology Multiplex Assay and the MILLIPLEX MAP Human TH17 Magnetic Bead Panel -Immunology Multiplex Assay (Millipore) following manufacturer's instruction.

Gene expression
24 hours after exposure, total RNA was isolated from THP-1 cells using the RNeasy Mini Kit (Qiagen, Leusden, The Netherlands) according to the manufacturer's protocol.Nanodrop ND1000 spectrophotometer (Thermo Scientific, Villebonsur-Yvette, France) was used to measure the purity and the concentration of extracted RNA.The Agilent 2100 Bioanalyzer electrophoretic system (Agilent Technologies, Diegem, Belgium) was used to verify the integrity of extracted RNA.All samples displayed high purity and RNA integrity values (RINs) above 8, making them suitable for qRT-PCR.cDNAs were prepared from 1 μg of RNA using the following reagents: Protoscript II reverse transcriptase and murine RNase inhibitor (New England Biolabs, Ipswich, MA, USA), dNTPs (Promega) and random primers (Invitrogen, Carlsbad, NM, USA) following the manufacturer's instructions.qRT-PCR was then performed on a ViiA 7 Real-Time PCR System (Thermo Scientific, Villebon-sur-Yvette, France) using the Takyon low ROX SYBR MasterMix dTTP Blue Kit (Eurogentec, Liège, Belgium) on 10 μL of qRT-PCR mix with the following concentrations: 1× MasterMix, 100 nM of each forward and reverse primers, and 0.4 ng/μL of cDNA.The thermal cycling conditions were as follows: denaturation for 5 minutes at 95 °C, followed by 45 cycles of denaturation for 15 seconds at 95 °C and annealing and extension for 1 minute at 60 °C.A final dissociation step (melting curve) was used to determine the primer specificity.No-Template and genomic DNA controls were added in each plate to exclude a possible contamination from the used reagents and the presence of genomic DNA.All PCR reactions were carried out in triplicates.Gene expression analysis was calculated based on the ΔΔCT method using the Biogazelle qbase PLUS software 2.5 (Gent, Belgium).Four reference genes (B2M, HPRT1, YWHAZ, SDHA) were selected in agreement with our previous study (Klein et al., 2017) and most stable candidates between experimental samples were determined using GeNorm in the Biogazelle qBase PLUS software and used for the analysis of the qRT-PCR results.The list of primers for the reference genes and genes of interest is summarized in Table 2.
Tab. 2: List of primers sequences used for qRT-PCR experiments on RNA isolated from THP-1 cells in the basolateral side of the coculture Unless further mentioned, qRT-PCR primer were designed for the experiments of this study Gene Forward primer (5′➔3′) Reverse primer (5′➔3′) TGGGAACAAGAGGGCATCTG CCACCACTGCATCAAATTCATG1 : (Klein et al., 2017).

Statistical analysis
All results are presented as mean ± s.e.m.Data were analyzed with an unpaired Student's t-test, or by one-way ANOVA using Graphpad Prism software (GraphPad Software Inc., San Diego, CA, USA).The macro Regtox on the Excel software (Microsoft, USA) was used to determine the CV75.Differences between groups were considered statistically significant when P<0.05.Table S1 1 summarizes comprehensive information such as the n values, statistical tests, P values, F values and degrees of freedom used to do the statistical analyses.Linear Discriminant Analysis was done using PAST3 software.

Development of a functional in vitro system of the alveolar barrier to assess respiratory sensitization
Different DC like cell lines were previously described as potential candidates to assess the sensitizing potential of chemicals (Galvão dos Santos et al., 2009).Among these there is the THP-1 cell line, which is used for the human cell line activation test (h-CLAT) that has been validated in 2017 by OECD to replace in vivo testing for skin sensitizers (Ashikaga et al., 2006;EURL ECVAM, 2015;OECD, 2017).To make the 3D in vitro model representative of the alveolar anatomy and functionality it is necessary that the DCs seeded at the basolateral side of the Transwell™ insert are able to migrate to the apical side, in order to capture the antigens and start the sensitization process.A chemotaxis assay was performed to determine the suitable pore size for the membrane of the Transwell™ inserts, which allows the migration of THP-1 DC like cells through the membrane (Fig. S1A 1 ).The percentage of migration of THP-1 cells through 1, 3, 5 and 8 µm pore size inserts was evaluated after 2h exposure to 0, 50 and 100 ng/mL of MCP-1.No migration to very low migration was observed using the 1 and 3 µm pore size membranes for all tested concentrations of MCP-1 (Fig. S1B 1 ).On the contrary, 8 µm pore size inserts allowed the migration of THP-1 cells even in the absence of chemoattractant.In addition, for the 8 µm pore size inserts, no significant concentration response curve was observed after 2h incubation to the different MCP-1 concentrations.
In the case of the 5 µm pore size inserts the "basal" migration rate (in the absence of chemoattractant) was about 2 % migration without MCP-1 and a dose response curve was clearly observable upon exposure to different concentration of MCP-1, with a percentage of migration of about 12 % when the highest MCP-1 concentration was used.Based on these observations, the 5 µm pore sizes insert was selected to build an in vitro system for the assessment of respiratory sensitization.
Since the choice of 5 µm pore size inserts represent a major modification to the original model of Klein et al. (2013), which is based on the 1 µm pore size inserts, it was deemed necessary to verify optimal cell viability and epithelial surface properties (presence of lung surfactant).No influence on cell viability (Fig. S1C 1 ) or on the lung surfactant (Fig. S1E 1 ) were observed with the new conditions as compared to the originals.
Further modifications were necessary to optimise the new lung in vitro model, such as the cell densities and the volume of the cell culture medium in the basolateral compartment.Compared to the original model for alveolar toxicity (Klein et al., 2013(Klein et al., , 2017)), cell densities were adapted and reduced to avoid the presence of detached dead endothelial cells and cellular debris in the basal compartment.The volume of cell culture medium in the basal compartment was reduced from 2 to 1 mL to avoid leakage of medium to the apical compartment, due to capillarity because of the higher pore size of the Transwells™ (5 µm versus 1 µm).Leakage of medium from basolateral to apical compartment would hinder the surfactant production and reverse the ALI status to submerged conditions.The new volume did not affect the viability of the system (Fig. S1D 1 ).Finally, to allow a stable differentiation of THP-1 cells into macrophage like cells, differentiation of THP-1 cells with PMA was followed by a 5-days resting period during which a greater degree of differentiation of THP-1 cells was observed, reflected by a better adherence of THP-1 macrophages-like cells to the epithelial layer.Five days resting allows macrophages to express surface markers associated with macrophage differentiation (such as the upregulation of CD11b, CD44 and CD49e and the downregulation of CD14 and TLR2) in a pattern similar to the in vivo situation (Daigneault et al., 2010;Mittar et al., 2011).The macrophage resting period allowed the reduction of the basal level of inflammation of the system (Marescotti et al., 2019).
To mimic at best the in vivo situation and to allow the system to be versatile and suitable for chemicals, the coculture system was further developed for exposure at the ALI (Lacroix et al., 2018).In our study the Vitrocell® cloud system was used with a modified basement module in which the inserts are placed during aerosol exposure to allow the deposition of the aerosol only through the apical side of the insert (Lenz et al., 2009).To ensure that the exposure would occur exclusively into the apical side of the inserts, a custom-made sealing device (Vitrocell®, Waldkirch, Germany) was used to seal the empty space between the inserts and the exposure module (Fig. S2A-D 1 ).Thereby direct exposure of the lower compartment is completely avoided.Direct exposure of the basal compartment to the aerosol was thereby completely excluded and there was no possibility for direct exposure of DCs located beneath the membrane.After exposure to water and DMSO vehicle controls, the complete coculture system did not show any statistical difference in viability after 24h and 48h incubation as compared to unexposed inserts kept into the incubator (Fig. S2E 1 ).

Influence of the microenvironment on THP-1 cells
To assess the impact of the multicellular environment on THP-1 cells, we compared cell surface markers expression of CD54 and CD86 on THP-1 cells after exposure to chemicals in monoculture and in coculture.Different fold increase levels for both markers, depending on the culturing condition and on the microenvironment were observed.The basal levels of CD54 and CD86 on THP-1 cells was determined in monoculture and coculture in the absence of external stimuli.The Mean of Fluorescence Intensity (MFI) of both markers was measured on THP-1 cells in both culture conditions after exposure to water.Results indicate an increase of MFI levels of both CD86 and CD54 basal levels after coculturing in the alveolar model (Fig. 3A-B) as compared to the measured basal level of the monoculture.Basal level expression of CD54 and of CD86 was about 28 times and 6 times higher, respectively, in cocultures than in THP-1 cells cultivated alone.
LPS-induced cell surface marker expression was also measured on THP-1 cells in monoculture and in coculture after 24h exposure.Despite not being a respiratory sensitizers itself, LPS is known to exacerbate the allergic response and respiratory inflammation (Kumari et al., 2015).In this study, LPS was used as a positive control to verify the response of THP-1 cells.LPS exposure did not affect cellular viability in monoculture and coculture (Fig. S3A 1 ).Exposure of THP-1 cells in both culture conditions to LPS induced overexpression of CD86 and CD54 as compared to the negative control (water).No statistical differences in the relative expressions of CD86 as compared to the negative control were observed with CD86 being expressed at 165% and 139% for the monoculture and the coculture, respectively.On the contrary, the expression of CD54 was upregulated to 2044% and 240% as compared to the vehicle control for the monoculture and the coculture, respectively (Fig. 3A-C).In the coculture, THP-1 cells were exposed indirectly to the cloud of LPS which could explain the apparent reduced relative expression of CD54 as compared to the one measured in monoculture where LPS was diluted directly in the medium.In this case, LPS had first to cross the alveolar barrier in order to activate THP-1 cells.In addition, even if THP-1 cells density was the same for monoculture and coculture, it must be considered that three additional cell types are present in the coculture, thus reducing the relative number of LPS molecules available for each THP-1 cell in the coculture.The basal expression of CD54 is 4 times higher in coculture as compared to monoculture (28 times and 6 times higher than the negative control for the coculture and monoculture, respectively.The relative expression is from the basal expression, thus explaining why LPS induced THP-1 cells in coculture presented a lower induction compared to the monoculture.Finally, the microenvironment produced at the alveolar barrier in the coculture system could also explain the difference in relative cell surface marker expression. In order to better understand the effect of the microenvironment on THP-1 cells, the basal cytokine release was measured in supernatants after 24h in both culture conditions.Coculture microenvironment contributed to DC recruitment and activation through the release of CCL20, GM-CSF, MCP-1, RANTES and IL-6.Significant increased levels of these mediators were found in coculture compared to monoculture: RANTES (3 times), GM-CSF (70 times), CCL20 (122 times) and MCP-1 (344 times) (Fig. 3D).In the coculture, 53 pg of IL-6/mL were released in the medium while its concentration in monoculture was below the detection level limit (Fig. S3B 1 ).IL-6, which participates in a broad spectrum of biological activities, is also involved in the regulation of cellular adhesion molecules such as CD54.The coculture microenvironment also provides pro-TH2 cytokines such as the release of IL-1α (2,8 pg/mL in coculture but below the detection limit in monoculture) and IL-7 cytokines (3 times higher in coculture) (Fig. 3D).Finally, the microenvironment could also have an immunosuppressive effect which protect cells from an exaggerated inflammatory response with the production of IL-10 (15 pg/mL released in coculture, but below the detection limit in monoculture).
In addition, cytokines release and gene expression in THP-1 cells were measured in the coculture after the exposure to LPS, which is generally used as a positive control for DC activation (Miyazawa et al., 2007).LPS is considered as an adjuvant of allergy and was shown to exacerbate hypersensitivity response and asthma (Liu, 2002;Kumari et al., 2015).In our system, LPS stimulation leads to the overexpression of IL-6, CCL20, RANTES, IL-1a and GM-CSF (Fig. S3B 1 ).IL-10, which has important immune regulatory function in vivo, is 10 times more released after LPS stimulation as compared to the monocultures.IL-10 production is known to affect DC function by downregulating surface expression of class II MHC molecules (Akdis et al., 2011) that could be linked to the down regulation of CIITA, and HLA-DMA gene expression in our model (Fig. S3C 1 ).

Assessing the respiratory sensitizing potential of chemicals: THP-1 in monoculture compared to the coculture with the alveolar barrier.
The first objective of our study was to compare the response of common established markers for sensitization on THP-1 cells (i.e.CD54 and CD86) when exposed directly in submerged monoculture conditions to respiratory sensitizers and respiratory irritants (following the h-CLAT protocol) and after exposure within the coculture system at the ALI.Experiments were designed in order to evaluate the possible benefit of such system for the assessment of respiratory sensitization and in order to justify the use of advanced and more complex coculture system, keeping in mind that in vitro models should be "as complex as necessary and as simple as possible" (Pridgeon et al., 2018).
It has been found in the h-CLAT that many skin sensitizing chemicals need a certain cytotoxicity for inducing sufficient activation of THP-1 cells.Both for monoculture and for coculture, a working concentration of chemicals (respiratory irritants and sensitizers) leading to 75% of cell viability (CV75) was established as prescribed by the h-CLAT protocol.These concentrations lead to the best results for cell surface markers expression in the h-CLAT model (Sakaguchi et al., 2009).Non-toxic and poorly soluble compounds were used at their limit of solubility and toxic compounds were used at the concentration leading to the CV75 (Fig. S4 1 ).The same cell density (1.0 x 10 6 cells/mL) of THP-1 cells was used in both the monoculture and the coculture systems.However, it must be considered that the coculture consists of three additional cell types instead of only one as it is in the monoculture.Thus, the relative concentration of the test compounds and the resulting calculated dose (micrograms of compound per million cells in coculture) is lower for the coculture as compared to the monoculture.Contrary to the monoculture, where cytotoxicity was assessed using Sytox blue (a nucleic acid stain) for its good reproducibility and rapid measurement, the Alamar Blue assay was used to evaluate the viability of the cocultures, since it allows the assessment of the cell viability in the whole Transwell™ insert.Dose responses curves were obtained after exposure of cocultures to Acr, PA and TMA (Fig. S4A 1 ).No cytotoxicity was observed exposing cells to MeSa at the concentration corresponding to the maximum solubility of the compound (56 µg/cm 2 ).For the monoculture of THP-1 cells, it was possible to obtain dose response curves only for Acr, which was used for the following experiments at the concentration corresponding to the CV75.For the other chemicals the concentrations corresponding to the maximum of solubility were used, since it was not possible to reach the CV75 (Fig. S4B 1 ).The determined CV75 in monoculture and coculture are reported in figure S3C (Fig. S4C 1 ).
In monoculture of THP-1 cells, (Figure 4B) the exposure to the respiratory sensitizers PA and TMA did not influence the expression of CD54 which was about 105% for PA and 96% for TMA.Similar results were observed for the irritant MeSa where an expression of 103% was measured for CD54.Acr exposure leads to a CD54 upregulation up to 165 % that is still below the threshold of 200% set for the h-CLAT to consider a compound as a sensitizer (Fig. 4B1).Similar results were obtained for the CD86 expression in the THP-1 cell monocultures where exposure to MeSa leads to the highest expression of CD86 with 131% relative expression measured (Fig. 4B2).
Most importantly, the use of the coculture system (Fig. 4A) allowed a clear differentiation between respiratory sensitizers and respiratory irritants based on the expression of CD54.Indeed, the exposure to the sensitizers PA and TMA lead to a significant up-regulation of the expression of CD54 on the surface of THP-1 cells of up to 186% and 204% as compared to vehicle controls, respectively, while no increase was observed for the respiratory irritants MeSa and Acr (Fig. 4A1).For what concerns the cell surface marker CD86, TMA showed an upregulation of up to 156% whereas PA exposure only lead to a 122% upregulation.Exposure of the coculture model to the respiratory irritant MeSa induced a slight increase of CD86 expression of 110% in THP-1 cells as compared to the vehicle control, while the respiratory irritant Acr induced a down-regulation (82% as compared to the vehicle control) (Fig. 4A2).Thus, the cellular context given by the alveolar barrier seems to have a valuable role for the in vitro detection of chemical respiratory sensitizers.The microenvironment at the epithelial barrier may indeed condition the immune response by sending danger signals that could recruit and activate DCs or, in specific cases, prevent the activation of DCs (van Rijt and Lambrecht, 2005).

Fig. 4: THP-1 cells present a different expression of costimulatory molecules depending on cultures conditions
Activation of THP-1 cells exposed to chemical respiratory sensitizers and irritants through exposure at the air liquid interface in the coculture (A) and in monoculture under submerged conditions (B).Relative mean of fluorescence intensity (rMFI, %) measured after 24h to chemical sensitizers and irritants on THP-1 cells of CD54 in coculture (A1) and in monoculture (B1) and of CD86 in tetraculture (A2) and in monoculture (B2).Red dotted lines represent the threshold levels used in the human-cell line activation (h-CLAT) test.Different letters illustrate significant differences between different treatments (Mean +/-SEM).

Cell surface markers
In addition to CD54 and CD86, other markers involved in the respiratory sensitization process were evaluated for their potential use for the in vitro prediction of the respiratory sensitization potential of airborne chemical substances.Naïve T cells differentiation into Th2 cells strongly depends on co-stimulatory molecules expressed by DCs.One of the most critical co-stimulatory molecules is the receptor OX40 expressed by T-cells and its ligand OX40L.At protein levels, slight increase in the expression of OX40L was observed in THP-1 cells after 24h exposure to the chemical respiratory sensitizer TMA but no increase was observed after exposure to the chemical respiratory sensitizer PA.In contrast, exposure to MeSa did not result in any significant effect in THP-1 cells, while the exposure to acrolein down-regulated the release of OX40L to 62%, as compared to the negative control.These results show that, despite being up-regulated by certain respiratory sensitizers, but not by others, and being down-regulated by certain respiratory irritants, but not by others, OX40L does not represent a good marker for the prediction of respiratory sensitization (Fig. S5A 1 ).The receptor for TSLP (TSLPr), a cytokine strongly involved in the OX40L upregulation on DCs, showed an increased expression on THP-1 cells after exposure to chemical sensitizers.The exposure to PA and TMA upregulated the TSLPr expression on THP-1 cells up to 162% and 151% respectively while no statistically significant differences in the expression of this marker were measured after exposure to the chemical irritants Acr and MeSa with 86% and 105% expression, respectively (Fig. 5A).TSLPr combines with IL7 receptor α (IL7Rα) to constitute the high-affinity-binding complex for TSLP, which is able to trigger signalling via the phosphorylation of signal transducer and activator of transcription 5 (P-STAT5), while TSLPr alone has low affinity for its ligand (He and Geha, 2010).IL7-Rα expression is upregulated after PA exposure and downregulated after Acr exposure, allowing the discrimination between a sensitizer and an irritant.However, no differences compared to their controls were observed after TMA and MeSa exposure (Fig. S5B 1 ).PA has been classified in the LLNA by topical exposure as a strong sensitizer while TMA was identified as a moderate sensitizer.The potency of these chemical sensitizers can explain why PA is able to upregulate the expression of IL7Rα on THP-1 cells while TMA does not.

Chemokines
DC activation leading to the up-regulation of specific cell markers including cell surface markers, depends on danger signals that are present in the microenvironment.In combination with other biological endpoints, specific cytokines patterns could enable the identification of respiratory sensitizers.Therefore, the impact of chemical respiratory sensitizers and chemical respiratory irritants on the cytokine pattern was assessed by measuring cytokines release in the medium in the basolateral compartment using a cytokine multiplex array on a Luminex system (Fig. 5B).Although the pattern of the released cytokines showed differences for the two chemical respiratory sensitizers, noteworthy differences of the cytokines release compared with the controls were observed (Fig. 5B).CCL20 secretion, which triggers DC recruitment to the alveolar barrier was increased up to 2.5, in the medium from the basolateral compartment, after PA exposure and 2.6 times after TMA exposure in but was not statistically significant different from the chemical respiratory irritant MeSa, which resulted in a 1.6 fold increase.In contrast, released CCL20 after Acr exposure was strongly decreased down to 10% of the vehicle control.
Released GM-CSF after 24h was 2,4 to 3,4 times higher in PA and TMA exposed cocultures as compared to their vehicle control.On the contrary, released GM-CSF decreased to 0,2 times after exposure to Acr while the exposure to MeSa had no influence on it.The epithelial-derived GM-CSF is described as an early critical signal in respiratory sensitization (Sheih et al., 2016), which also contributes to the development of asthma (Schuijs et al., 2013).On the other hand, IL-1a cytokine, which acts on epithelial cells in an autocrine signalling manner to trigger GM-CSF production, was not different from the vehicle controls after exposure to respiratory sensitizers and irritants.IL-10, which is secreted by macrophages to prevent DC activation, allows the distinction between chemical respiratory sensitizers and chemical respiratory irritants, being more secreted by sensitizers and decreased or unchanged by irritants.RANTES is described by several authors as a cytokine of interest for the discrimination between respiratory sensitizers and irritants (Huang et al., 2013;Hermanns et al., 2015).However, in our model RANTES does not clearly contribute to the distinction between irritants and sensitizers.MCP-1 and IL-7 were also described as possible discriminant markers for the identification of respiratory sensitizers (Huang et al., 2013).MCP-1 release was only different after the exposure to Acr, for which 10 times less MCP-1 was release as compared to the control, while its amount was not different for all other tested compounds.Other cytokines measured (such as IL-6, sCD40L, IL-25 and IL-33) did not contribute to the distinction between controls, irritants and sensitizers or were even below the detection level (data not shown).

Gene expression
An assay based on a genomic biomarker signature, the so-called Genomic Allergen Rapid Detection (GARDair), identified hundreds of genes which are regulated in the MUTZ-3 cell line after exposure to respiratory sensitizers (Forreryd et al., 2015).Based on these findings, a set of genes that are strongly involved in immune response pathways was chosen to be tested in THP-1 cells from our coculture.IL1RL1 (also called ST2), which encodes for IL-33 receptor allowed the classification between sensitizers and non-sensitizers with a log2 fold increase expression higher than 1 following exposure to PA and TMA (Table in Fig. 5C) On the contrary, CIITA expression, which encodes for the Major Histocompatibility Complex class II (MHCII) showed a decrease expression at the gene level after exposure to PA and TMA.By looking to the expression of the set of genes, MeSa is the only compound, which increased expression of certain genes.Indeed, log2 fold increase of the gene expression was higher than 1 for MyD88, G-CSF-R and HLA-DMA and it resulted 10 times higher than its vehicle control for the expression of HLA-DRA.

Discriminant analysis
A Linear Discriminant Analysis (LDA) was built in order to investigate which markers influenced the most the classification of the compounds as respiratory sensitizers or irritants (Fig. 5D).The LDA allows the distinction between the three groups: vehicle controls, chemical irritants and chemical respiratory sensitizers.Cell surface markers, cytokines and gene expression were included in the LDA.Among all tested markers for the respiratory sensitizers group, GM-CSF, CCL20, IL-10 and IL1R1-1 were the markers that most strongly influence the distribution of the groups along the axes.RANTES, CD54 and TSLPr allowed the discrimination between the three groups only to a lesser extent.Irritant identification was mostly influenced by the gene expression of HLA-DRA and CD80, and G-CSF-R and HLA-DMA at a lesser extent.

What about protein allergens?
Contrary to chemical sensitizers, which are mostly involved in occupational allergies, respiratory allergies affect all layers of the general population.To verify if the in vitro model proposed in this work is able to predict potential respiratory sensitization induced by environmental proteins (e.g.pollen), the coculture was exposed to different protein allergens such as House Dust Mite (HDM) extract (containing 20 µg/mL of Der p1) and to Bet v1, the major protein allergen from Birch Pollen (BP).Both Der p1 and Bet v1 represent common causes of asthma and seasonal allergy (Gregory and Lloyd, 2011;Weber, 2014).
Both exposures to Bet v1 and HDM extract did not influence the viability of the system (Fig. S6A 1 ).All markers used in our model for the assessment of respiratory sensitization potential of chemicals were investigated after exposure to both proteins for 24h.While the exposure to chemical respiratory sensitizers did not influence the OX40L expression on THP-1 cells, the exposure to protein allergen Bet v1 and to HDM extract lead to the upregulation of OX40L to 158 and 150% respectively (Fig. 6A).MCP-1 release in the medium was 2.3 times increased after exposure to Bet v1 while the exposure to HDM extract did not influence MCP-1 release compared to the control.MCP-1 is thus not the most suitable to identify protein allergen in the system.The same observation could be drawn for IL-10, MCP-1, CCL20 and IL-6 cytokines, where a higher response is observed after exposure to Bet v1 while HDM extract exposure did not differ from the control (Fig S6B 1 ).Also for gene expression, Bet v1 exposure is different from the vehicle control for IL1R1, MAP2K1, G-CSF-R and CIITA, while the exposure to HDM extract had no impact on gene expression (Fig. S6C 1 ).Differently from chemical respiratory sensitizers, IL1R1 gene expression tended to decrease after exposure to protein allergens.Exposure to chemicals, either sensitizers or irritants, did not influence MAP2K1 gene expression.Expression of MAP2K1 tended to decrease after exposure to protein allergens.Finally, Bet v1 exposure lead to the upregulation of CIITA and G-CSF-R gene expression.LDA performed on protein sensitizers does not allow discriminating them from the vehicle controls (Fig. 6B).The only discriminant markers influencing the distribution of the samples along the axis are the cell surface marker OX40L and the MCP-1 cytokine.

Discussion and conclusion
One suggested way to assess the respiratory sensitizing potential of chemicals began with the use of models developed for skin sensitization adapted if needed for respiratory purpose (Basketter et al., 2017).Indeed, most chemical respiratory allergens were positive in assay for skin sensitization.Tests performed with the DPRA (Direct Peptide Reactivity Assay) present a rich database for chemical respiratory sensitizers while less data is available for the h-CLAT where in many instances no information is available.False negative results are also observed, for instance the respiratory sensitizers PA and hexamethylene diisocyanate (HDI) (Bloemen et al., 2009) are positive in the in vivo LLNA assay and classified accordingly as strong and extreme sensitizers but are negative in the h-CLAT (Dearman et al., 2013;Basketter et al., 2017;Ohtake et al., 2018).In monoculture of THP-1 cells, all respiratory sensitizers and irritants that we tested induced upregulation of CD86 which is below the threshold of 150% set for the h-CLAT to consider a compound as a sensitizer.The results of these experiments are different from what was observed using the h-CLAT, which classified TMA as a sensitizer based on a CD86 expression higher than 150% compared to the negative control.Similar inconsistencies in the h-CLAT results for the expression of CD86 have been reported before by others (Parise et al., 2015), who reported CD86 values above the 150% threshold for only 8 out of 23 chemical skin sensitizers tested.The majority of chemicals that have the potential to induce sensitization of the respiratory tract could induce a positive prediction in assay for skin sensitization, it is thus not possible to discriminate skin from respiratory sensitizers.Especially for regulatory needs, the distinction between both kinds of sensitizers is crucial as the classification differs.Indeed, respiratory sensitizers are considered as SVHC while skin sensitizers are not.In order to overcome these issues, the model presented in this work was specifically designed to predict respiratory sensitization.Based on the current knowledge on the relevant cell interactions involved in sensitization we have developed a coculture model of the alveolar barrier which include alveolar type II epithelial cells, endothelial cells, macrophages and dendritic cells.This model mimics the alveolo-capillary barrier and allows the communication of all cell types.Apical exposure through the ALI to the compounds elicits the basolateral release of cytokines which participated, along with chemicals that could have crossed the alveolar barrier, to the activation of dendritic like cells.Our hypothesis is that the microenvironment did not only induce DC activation, but it also directed the immune response towards a pro-TH2 environment through the release of IL-1α and IL-7 cytokines.IL-1α controls in an autocrine manner the release of cytokines such as IL-33, which is involved in TH2 immunity (Willart et al., 2012).IL-7, which is involved in T-cells proliferation, maturation and survival (Yeon et al., 2017).Finally, the microenvironment could also have an immunosuppressive effect which protect cells from an exaggerated inflammatory response with the production of IL-10.We hypothesize that such cytokine release is required for DCs activation in the context of lung sensitization.
A first set of compounds comprising two respiratory sensitizers (PA and TMA), two irritants (Acr and MeSa) and two protein sensitizers (HDM and Bet v1) allowed the identification of discriminating markers either on the surface of DCs or released in the medium of the basolateral compartment such as cytokines and biomolecular events triggered in DCs.Our in vitro coculture model is able to discriminate chemical respiratory sensitizers from chemical irritants through the TSLPr and CD54 markers expression measured on the surface DCs after 24h.CD54, involved in cellular adhesion, was described in previous studies for the assessment of the skin sensitizing potential of chemicals (Galvão dos Santos et al., 2009).The TSLP cytokine was described in the development of allergic diseases, mostly during sensitization and allergic inflammation to HMW allergens (Reefer et al., 2010).An increased epithelial expression of TSLP has been shown in patients suffering from asthma (Bosnjak et al., 2011).However, its receptor TSLPr was never measured so far for the identification of respiratory sensitizers to the best of our knowledge.We found that TSLPr expression increased on DCs after 24h exposure to chemical respiratory sensitizers in our model, while no increase was measured after exposure to irritants.TSLP instructs DCs to mature and drive TH2 polarization by upregulating the expression of co-stimulatory markers such as OX40L.This signalling pathway may explain why, in our model, DCs do not express TSLPr after exposure to both protein sensitizers although the expression of OX40L on DCs is upregulated for both of them.Even if we did not detect an increase of OX40L on the cell surface of THP-1 under the chosen experimental conditions, the OX40L gene expression was demonstrated to be useful for the identification of respiratory chemical sensitizers while helping the discrimination of chemical respiratory sensitizers from both irritants and chemical skin sensitizers (Mizoguchi et al., 2017).To conclude, we found suitable markers in our model based on a representative set of learning compounds but only few substances were tested.Additional chemical respiratory sensitizers need to be tested in order to validate the model.A further improvement would be the testing of chemicals inducing exclusively skin sensitization such as dinitrochlorbenzene (DNCB) in order to establish suitable markers that could distinguish respiratory from skin sensitizers, which would represent an added value of our model over the monoculture of only THP-1 cells.
After this proof-of-concept using one representative concentration per chemical (yielding comparable cytotoxicity if possible), next important step could be the measurement of concentration-dependent responses in order to be able to better discriminate the responses to different chemicals and to obtain information relevant for potency assessment.An additional improvement for the further optimisation of our system would also be to measure the different endpoints at different time points.For instance, no difference between exposed cells and the vehicle control was observed for CD86 and CD80 at the mRNA expression level in THP-1 cells after 24h.At the difference, an increase of these genes was measured after 9h exposure to chemicals in monocytes in a coculture system (Mizoguchi et al., 2017).There are differences between both models like the way of exposure (submerged conditions vs at the ALI in our model) and the cells which are used (bronchiolar BEAS-2B cell line and fresh human monocytes from peripheral blood).In addition, the in chemico method DPRA, validated in 2015 by OECD to assess the sensitizing potential of chemicals, was normally used at 24h with only one concentration of the test compounds (Lalko et al., 2012;OECD, 2015).Measuring the peptide reactivity in a kinetic DPRA, meaning with different concentrations and at different time points (30 minutes, 2h and 24h), resulted in a better accuracy for the assessment of sensitization potential.A possible determination of the potency of the sensitizers, allowing the classification of the chemicals into the 1A and 1B classes, which was only possible in vivo (Wareing et al., 2017).In our model the concentration CV75 was used as a starting point of our experiment as it is the case for the h-CLAT.While the CV75 could be useful for the determination of cell surface markers expression, the use of sub toxic concentrations may be more appropriate for the gene expression and cytokines measurements.For instance, sub toxic concentrations of different respiratory sensitizers and irritants were applied in submerged conditions for 24h to an in vitro lung capillary barrier comprising bronchiolar club cells NCIH-441 and microvascular endothelial cells ISO-HAS-1 to determine a pattern of cytokines released (Hermanns et al., 2015).Here, authors identify RANTES as a cytokine able to discriminate respiratory sensitizers from irritants while, in our model, only TMA induced an increase of RANTES release which was different to the irritants.On the contrary, upregulation of the release of GM-CSF, MIP-3a (CCL20) and IL-10 proved to be suitable markers for the discrimination between chemical sensitizers and irritants in our system.Again, even if our model is different from this one, information given by this additional concentration would be interesting and could help for the harmonisation of the concentration of chemicals applied to our model, as for now TMA, PA and Acr are used at the CV75 while MeSa, which is not soluble enough, did not have effect on cell viability at the tested concentration.Having comparable level of cytotoxicity may help in the determination of concentrations to test in our model.The difference between the levels of cellular viability taken for both irritants could explain why we observed so important differences between the results of cytokine expression.MIP-3α (CCL20) and MCP-1 released in the basolateral compartment were highly down-regulated after exposure to Acr while no difference was observed after MeSa exposure.The same conclusions can be drawn for the gene expression for which MeSa exposure induced the upregulation of G-CSF, HLA-DMA and HLA-DRA genes while Acr did not.Forreryd and co-workers used concentrations leading to 90% of cell viability (CV90) for the GARD assay for respiratory sensitizers to establish their genomic signature on the MUTZ-3 dendritic like cell line (Forreryd et al., 2015).Particular attention needs also to be raised regarding the solubility of the compounds.Indeed, certain sensitizers are either water insoluble such as diisocyanates or are directly hydrolysed after being in contact with water as it is the case for acid anhydrides which could have led to false negative results in the h-CLAT.The low solubility of some compounds also represents a limiting factor in our model.The use of compounds such as HDI remains problematic as it polymerises in contact with water making the nebulization impossible.Extensive work needs to be done in order to be able to expose our system to all kinds of chemical compounds at the ALI.
HMW allergens such as pollen grains were believed to be unable to reach the lower airways and thus to induce an immunologic response there.However, a small percentage of large-sized particles may penetrate into the lower airways (Michel et al., 1977).Lower airways and in particular alveoli are of interest for respiratory sensitization as DC are strongly active with the most robust air surveillance in the lung observed at this site (Thornton et al., 2012).The presence of alveolar type II epithelial cells, as we have in our model, is important for HMW sensitization assessment as it has been demonstrated that lung surfactant proteins SP-A and SP-D from alveolar type II cells attenuate the allergic inflammation by directly binding to the protein allergens (Stewart et al., 2014).Therefore, the use of our model is justified also for the testing of HMW allergens.Most of the markers identified for chemical respiratory sensitizers were not upregulated by the exposure of HMW sensitizers.Indeed, only OX40L expression on THP-1 cells was up-regulated after the exposure to HMW sensitizers while this marker was down-regulated after the exposure to irritant.This marker could be used for the identification of HMW respiratory sensitizers in our model.Mechanism of action of HMW sensitizers differ from the one of chemical respiratory sensitizers.Many known HMW allergens have a protease activity, such as those present in HDM or pollens which are able to disrupt the barrier function of the epithelial layer by cleaving tight junctions, facilitating the passage of allergen across the epithelial surface and hence activating protease-dependent and protease-independent signalling pathways (Kauffman et al., 2006;Hammad et al., 2009).It is still unclear how important is the protease activity of allergens as protease activity is not a general feature of all protein allergens (Papazian et al., 2015).Allergens and cytokines released at the epithelial barrier then trigger activation of DCs and sensitization (Lambrecht and Hammad, 2015).In our model, MCP-1, IL-10, CCL20 and IL-6 cytokines released were increased in the basolateral compartment compared to the vehicle control only after the exposure to Bet v1 but not to HDM.Considering the different forms of the allergen, differences in the responses were expected for the purified protein Bet v1 and the extracts of HDM.Indeed, cytokine released after exposure to natural purified Der p1, which is one of the major proteins of HDM, and to the whole HDM extract showed different dose response curves by A549 cells (Kauffman et al., 2006).In this experiment, IL-6 cytokine released was about two times higher after exposure to the natural purified Der 1 protein compared to the HDM extract at the same concentration.Noteworthy, the release of IL-6 and IL-8 cytokines after HDM extract exposure declined at highest exposures, which was already observed using different HMW extracts (Kauffman et al., 2000(Kauffman et al., , 2006)).Using lowest concentrations of HDM in our model could have showed differences in the cytokines release.
Sensitization and subsequent allergic reaction could occur after repeated exposure to compounds.However, it is not possible to do multiple exposures in our model due to its limited shelf life.Extensive work has to be done in order to extend the shelf life of the model.
We focused our research on the alveolar part.The same approach applied to the different parts of the respiratory tract such as the nose, the trachea and the bronchi would be needed to assess the respiratory sensitizing potential of chemicals through the complete respiratory system, but our system could represent at least one of the links of the chain.
In conclusion, the learning set of compounds showed that TSLPr and CD54 cell surface markers expression, IL-10, GM-CSF and CCL20 cytokines release and IL1-R1 gene expression at 24h may be used to differentiate chemical respiratory sensitizers from irritants in our model and OX40L cell surface marker expression could be used to identify HMW respiratory sensitizers.

Fig. 1 :
Fig. 1: The alveolar barrier and the in vitro system.(A) Schematic representation in vivo of the alveoli.The alveoli represent an extensive surface area.The close contact of alveolar epithelial cells and endothelial cells from the capillary allows for gases exchange.Surfactant, produced by type II epithelial cells, lower the surface tension and participate to alveolar host defense mechanism by interacting with immune cells such as alveolar macrophages for the binding and the removal of foreign materials.Dendritic cells (DCs) are located above and beneath the basement membrane.Alveolar macrophages have an active role in immune response and they also mediate DCs functions.(Adapted and modified from Lambrecht and Hammad, 2003; Whitsett and Alenghat, 2015), (B) Coculture system previously developed by Klein et al. 2013 to study the inflammatory effect of NPs at the alveolar barrier.Cells were seeded on a porous membrane of a Transwell™ insert and cultured at the air liquid interface (ALI) for aerosol exposure.(C) Modification of the coculture system allowing the assessment of the respiratory sensitizing potential (adapted and modified from Chary et al. 2017, WO2018/122219 A1).

Fig. 2 :
Fig. 2: Workflow for the seeding and exposure of the coculture in vitro model

Fig
Fig. 3: THP-1 cells response are conditioned and influenced by the coculture micro environment (A) Expression of CD54 and CD86 costimulatory molecules on THP-1 cells after lipopolysaccharide (LPS) stimulation or exposed to the vehicle control (i.e.water) in submerged conditions in monoculture or exposed through the air-liquid interface in the coculture (Mean +/-SEM).(B) Basal expression of CD54 and CD86 costimulatory molecules (Mean +/-SEM).(C) Induced expression of CD54 and CD86 after 24h exposure to LPS, (D) Basal cytokine (GM-CSF, MCP-1, MIP-3α/CCL20, RANTES and IL-7) levels released in the medium in both cultures (Scatter plot and mean).

Fig. 5 :
Fig. 5: GM-CSF, IL-10 and IL1-R1 might be used to identify respiratory sensitizers in the coculture system:) (A) TSLPr expressed on THP-1 cells in the coculture after 24h exposure to chemical sensitizers and irritants at the ALI.Different letters illustrate significant differences between different treatments (Mean +/-SEM).(B) Differential cytokines release in medium in the basolateral compartment of the coculture after 24h exposure to chemical sensitizers and irritants.Results were normalized to the level of expression of the vehicle control (Dark blue: inducedlight blue: repressed more than 2 times).(C) Differential gene expression of relevant markers for sensitization in THP-1 cells after 24h exposure to chemical sensitizers and irritants.Results were normalized to the level of expression of the vehicle control and expressed on the log2 scale (Dark blue: inducedlight blue: repressed more than 2 times (or log2 of -1).(D) Linear Discriminant Analysis (LDA) (Past3 software) to point out which markers (CD86, CD54, IL7ra, TSLPr and OX40L cell surface markers, CIITA, MyD88, HLA-DMA, HLA-DRA, CD80, IL1R1-1 and MAP2K1 genes and MCP-1, IL7, IL10, MIP3a, RANTES and GM-CSF cytokines) influenced the distribution of samples within the different groups (irritants, sensitizers or controls).Convex hulls gather samples from the same group together and deep blue lines indicates markers influencing the distribution.Red dotted line corresponds to the part of the LDA which is enlarged.

Fig. 6 :
Fig. 6: OX40L could be used to identify protein respiratory sensitizers in our model (A) OX40L expression on THP-1 cells after 24h exposure in the coculture to Bet v1 protein and HDM extract containing Der p1.Different letters illustrate significant differences between different treatments (Mean +/-SEM).(B) Linear Discriminant Analysis (LDA) (Past3 software) to point out which markers (CD86, CD54, IL7ra, TSLPr and OX40L cell surface markers, CIITA, MyD88, HLA-DMA, HLA-DRA, CD80, IL1R1-1 and MAP2K1 genes and MCP-1, IL7, IL10, MIP3a, RANTES and GM-CSF cytokines) influenced the distribution of samples within the different groups (irritants, sensitizers or controls).Convex hulls gather samples from the same group together and deep blue lines indicates markers influencing the distribution.