An In Vitro Air-Liquid Interface Inhalation Platform for Petroleum Substances and Constituents

The goal is to optimize and show the validity of an in vitro method for inhalation testing of petroleum substances and its constituents at the air-liquid interface (ALI). The approach is demonstrated in a pilot study with ethylbenzene (EB), a mono-constituent petroleum substance using a human alveolar epithelial cell line model. This included the development and validation of a generation facility to obtain EB vapors and the optimization of an exposure system for a negative control (clean air, CA), positive control (nitrogen dioxide), and EB vapors. The optimal settings for the VITROCELL® 24/48 system were defined. Cytotoxicity, cell viability, inflammation, and oxidative stress were assessed in A549 after exposure to EB vapors. A concentration-dependent significant decrease in mean cell viability was observed after exposure, which was confirmed by a cytotoxicity test. The oxidative stress marker superoxide dismutase 2 was significantly increased, but no concentration-response was observed. A concentration-dependent significant increase in pro-inflammatory markers C-C motif chemokine ligand 2, interleukin (IL)6, and IL8 was observed for EB-exposed A549 cells compared to CA. The data demonstrated consistency between in vivo air concentrations at which adverse respiratory effects were observed and ALI-concentrations affecting cell viability, provided that the actual measured in vitro delivery efficiency of the compound were included. It can be concluded that extrapolating in vitro air concentrations (adjusted for delivery efficiency and absorption characteristics and applied for testing cell viability) to simulate in vivo air concentrations may be a promising method to screen for acute inhalation toxicity.


Introduction
There is an ever-increasing demand to implement human-relevant in vitro testing approaches. Inhalation is one of the major routes of exposure to xenobiotics and may serve as both the target tissue and portal of entry into the systemic circulation. The in vitro air-liquid interface (ALI) exposure method (i.e. cells cultured on a permeable insert with the basal surface of the cells in contact with liquid culture medium and the apical surface exposed to air) is promising, as this method: (i) does not require animals, except for animal components used for cell culture such as fetal bovine serum (FBS), (ii) is in total less costly and time consuming than an in vivo experiment; (iii) more reliably mimics human exposure compared to submerged cell culture or rodent exposures; and (iv) facilitates the evaluation of mechanistic effects of inhaled material on human lung cells and contributes to the development of adverse outcome pathways (Lacroix et al., 2018). The current study assesses the performance of an ALI in vitro system after exposure to petrochemical substances traditionally considered difficult to test in such systems. Two key hurdles must be addressed in order to achieve this goal. First, an experimental system must be designed to address the physicochemical properties that make these substances difficult to assess in cell-based systems, such as low aqueous solubility and high volatility. Secondly, substances derived from petroleum may contain many individual constituents with a range of physicochemical properties varying in relative proportion over time. Keeping in mind that individual constituents have a range of physicochemical properties, petrochemicals are often considered difficult to test due to the tendency to volatilize (e.g. gasoline constituents have a typical boiling point range of -20 C to 250 C (Nyer and Skladany, 1989)) and to partition into organic layers (e.g. octanol-water partition coefficients are greater than 6 (Reichenberg and Mayer, 2006)). Despite these issues, the chemicals must still be assessed per regulatory registration 2 Materials and methods

A549 cell model and culture conditions
A new vial of the human alveolar epithelial type 2-like A549 cell line was obtained from American Type Culture Collection (ATCC number: CCL-185, 80 passages, Manassas, USA), which was originally derived from a lung carcinomatous tissue from a 58-year-old Caucasian male. A549 cells were grown in T-75 culture flasks and routinely maintained in Minimal Essential Medium (MEM) 1x with GlutaMAX™-1 (Brand Gibco, ThermoFisher Scientific, Waltham, USA) supplemented with 10% non-heat inactivated FBS superior (Merck, Darmstadt, Germany) at 37 °C under 5 % carbon dioxide (CO2). Before reaching 70-80 % confluence, cells were subcultured using (0.05 %) Trypsin-EDTA solution (Brand Gibco, ThermoFisher Scientific). Medium was refreshed every 2 days and cells were subcultured every 3 (9x10 5 cells in 20 ml cell culture medium (CCM)) or 4 days (4.5x10 5 cells in 20 ml CCM). Cells from the work cell bank were passaged at least 2 times before use in experiments and no more than 20 times in total. A549 cells were negative for mycoplasma.
In vitro ALI exposure A549 cells were seeded at a density of 50,000 cells/insert (~151,000 cells/cm²) on ThinCert™ polystyrene membrane inserts, pore size 0.4 µm, surface area 0.33 cm² (24-well format) (Greiner Bio-One, Kremsmünster, Austria, Catalogue number 662641). Inserts were placed in a sterile 24-well plate, and CCM was added to both sides, 600 µl basolateral and 100 µl apical. Plates were incubated for ± 72 hours (h) at 37 °C, 5 % CO2 in a humidified incubator. Immediately before exposure, CCM was completely removed from the apical side and the inserts were transferred into the VITROCELL ® 24/48 device (VITROCELL ® Systems Gmbh, Waldkirch, Germany). Before positioning the inserts in the base module, each row was separately filled with CCM (20 ml) allowing cells to be nourished from the bottom while being exposed to vapor on the top side. The base bottom module with inserts and CCM was positioned on a temperature-controlled heating plate set at 37 °C. The base top, inlet, and exhaust module including the main distribution line were heated using a water circulation system set at 37 °C. The cells were exposed to the atmosphere from the main distribution line by extraction at each of the 48 positions in the base top module using a vacuum pump. Each insert was exposed to a flow of 1.5 milliliter per minute (ml/m) for 4 h. The distance between the trumpet and insert was 2 mm. Before cells were exposed to EB vapors, the ALI system was thoroughly tested for uniformity to clean air (CA) and nitrogen dioxide (NO2) exposure. At least 42 positions with A549 inserts in the plate were exposed to either CA (negative control) or NO2 (in-house used positive control, about 12 ppm) with a flow of 500 ml/m in the main distribution line. Experiments were repeated three times using cells of a different passage.
Incubator control (IC) cells, consisted of 6 cell culture inserts without apical medium, were kept in a humidified 37 °C incubator with 5 % CO2 for 24 h and served as control for CA exposure. A549 cells were exposed to 3 concentrations of EB vapors i.e. 30000 mg/m³ (800 ml/m), 40000 mg/m³ (600 ml/m), and 50000 mg/m³ (490 ml/m) (in rows 6 and 5, 4 and 3, 2 and 1, respectively 6 positions per row), CA (row 8, 6 positions), and NO2 (20 ppm or 41.1 mg/m 3 , row 7, 6 positions). After exposure, the inserts were placed in a new sterile 24-well plate (different plates for control versus exposed inserts to avoid carry over) with 600 µl CCM on the basolateral side and allowed a recovery period of 1 or 20 h in a humidified 37 °C incubator with 5 % CO2. For each biologically independent run, 12 replicate cell cultures spread over 2 rows of each EB concentration were treated in parallel in the VITROCELL ® 24/48 system, of which 4 replicates per condition were incubated post-exposure for 1 h and analyzed for gene expression by real-time quantitative reverse transcription polymerase chain reaction (real-time qRT-PCR), and another 4 replicates were incubated post-exposure for 20 h for further assessment of cell viability/cytotoxicity, and protein secretion. In addition, 4 replicates were used for optimization of HS-GC-MS analysis (which is later replaced by dose determination in stainless steel inserts). For NO2 and CA only 1 row was exposed for each compound, and 3 inserts were analyzed for real-time qRT-PCR and 3 for cell viability/cytotoxicity, and protein secretion. Five biologically independent runs, using different cell passages, were performed. In Figure 1, a schematic drawing of the experimental process and design is shown. Generation and characterization of EB Ethylbenzene (CAS 100-41-4, Sigma-Aldrich) was volatized using the following generation set-up (Fig. 2). A closed stainlesssteel vessel (100 ml volume) with two fittings was put on an enclosed microbalance (Mettler-Toledo AB104S, VWR, Radnor, Pennsylvania, USA). EB was injected with a pipet in the vessel and weighed continuously during the experiment to measure consumption. A pressure gauge was used to control the nitrogen pressure (50-4000 mbar) through the capillary into the vessel using one of the fittings. The second fitting was connected to the second capillary (outlet). By controlling the pressure, capillary size and length, the sample was precisely dosed in a tank with a heated base plate set at 195 °C. Temperature of the base plate was controlled with a Variac temperature controller and monitored with a thermocouple connected to a thermometer (Fluke 52 II, Distrelec,'s-Hertogenbosch, The Netherlands). The capillary dosage unit was based on the method described by Goelen et al. (1992) and further improved for the EB case.
When the substance dripped on the bottom plate of the tank (150-155 mg/minute (min)) it vaporized and was mixed with a controlled (Mass Flow Controller, MFC, Brooks Instrument, Hatfield, Pennsylvania, USA) humidified air flow of 3 liter per minute (l/m). The humidification was obtained by using two Nafion (Perma Pure, Lakewood, USA) humidifiers in parallel. A relative humidity (RH) of more than 80 % is necessary to keep the A549 cells alive. Subsequently, the humidified vapor flow was guided through a double-walled glass tube positioned in a thermostatic circulating water bath (Lauda Alpha A6, VWR) set at 42 °C.
The exhaust of the glass tube was connected to a glass T-piece. One side was connected to a 2-way valve to be able to build up enough pressure in the flow direction of the VITROCELL® 24/48 ALI platform (VITROCELL ® Systems Gmbh). The exhaust of the needle valve (overflow) was guided into the fume cupboard. The other side of the T-piece was connected through a heated sampling line (45 °C) to a low-pressure MFC (Brooks Instrument) which was positioned inside the climatic chamber (37 °C). This MFC was set at the desired flow for the first row of the ALI platform (484-492 ml/m). Before cell exposure, temperature and RH of the flow were measured (Temperature and Humidity probe 635-2135, Testo NAV/SA, Ternat, Belgium) using a T-piece. Generated concentrations were determined by combining microbalance consumption and used airflows on the one hand and a Flame Ionization Detector (FID) analyzer (J.U.M. Engineering 3-300A, Karlsfeld, Germany) on the other hand.

Estimation of in vitro test concentrations for EB
The methodology that was used to calculate in vitro air concentrations from concentrations of in vivo studies is demonstrated in Table 1. For the initial calculations, the in vivo absorption was considered to be equal to the in vitro absorption and a flow per insert of 3 ml/m; the latter was changed later during study design optimisation to 1.5 ml/m (Tab. S1 1 ). The in vitro air concentrations were calculated with the following formula (eq.1): = * * ℎ * * * Cvitro = in vitro air concentration (mg/m³) Cvivo = in vivo exposure concentration (mg/m³) tvivo = in vivo exposure time (h) Vinh = volume inhaled air per h (m³/h) SAinsert = surface area insert (cm²) SAlung = lung surface area (cm²) Finsert = flow per insert (m³/min) tvitro = in vitro exposure time (min) To illustrate the reasoning behind the calculations, the 4 h-LC50 rat (i.e. the concentration of the chemical that will kill 50 % of the test animals) is chosen as an example. Starting from the in vivo dose of 17360 mg/m 3 , the in vivo mass per cm 2 lung is calculated. Hereto, the in vivo exposure time, volume inhaled air per h, and lung surface area are taken into account. This in vivo mass per cm 2 lung is set equal to the mass per cm 2 insert; from here the in vitro concentration in air is calculated, taking into account in vitro exposure time and flow per insert (m³/min). In vivo 17360 mg/m 3 corresponds to in vitro 170 mg/m 3 , both for 4 h exposure and assuming in vivo absorption to be the same as in vitro absorption. The dose-range resulting from 10 in vivo studies considered (Tab. 1) was 0.02-355 mg/m³ for 4 h exposure. The initially calculated ALI in vitro dose derived thereof was 1-625 mg/m³ for 4 h exposure to EB, with each time a factor of 5 difference between two consecutive Volunteers: no effect on pulmonary function (Moscato et al., 1987) Indicative occupational exposure limit (OEL), short (EC, 2000) Volunteers: irritation of throat/nose and feelings of "chest congestion" (Yant et al., 1930) Volunteers: no nasal irritation (van Thriel et al., 2003) Volunteers: severe irritation of throat and nose (immediately) (Yant et al., 1930) RD50*, mouse, breathing rate reduction (de Ceaurriz et al., 1981) RD50, mouse, breathing rate reduction (Nielsen and Alarie 1982) LC50** , rat (Smyth et al., 1962) LC100, rat (Smyth et al., 1962) LC100, rat (Ivanov, 1962) In vivo  (Paur H.R. et al., 2011), mouse (Knust J. et al., 2009), rat (Stone K.C. et al., 1992); *RD50: Concentration which elicits a respiratory rate decrease of 50%; **LC50: Lethal concentration that will kill 50 % of the test animals ALTEX preprint published April 20, 2021 doi:10.14573/altex.2010211 6 concentrations. The lowest concentration (1 mg/m³) is chosen because it is based on in vitro concentrations derived from acute human studies showing throat irritation (e.g. van Thriel et al., 2003), the highest concentration (625 mg/m³) is a concentration that is above in vitro concentrations derived from acute lethal concentrations in animal studies (e.g. Smyth et al., 1962).
Due to lack of biological effects (data not shown) in initial ALI experiments (up to 625 mg/m³) and during process optimization (at 7000 mg/m 3 ), and based on experimental results obtained for delivery efficiency of EB (see results section 'Absorption of EB'), these initial in vitro concentrations derived through calculation from in vivo data had to be modified and high EB concentrations of 30000-50000 mg/m 3 were finally tested.

Chemical analysis of delivered dose
The actual delivered dose is a result of the delivery efficiency of the substance from an aerosol to the liquid lining of the cell layer (linked to its physicochemical properties) and the deposition efficiency of the exposure system.
Analysis of EB-exposed cells showed that necessary cell handling prior to headspace-gas chromatography-mass spectrometry (HS-GC-MS) analysis caused high evaporation rates which was not representative for the actual exposure conditions. Dose determination was performed using stainless steel inserts with 125 µl CCM to minimize sample handling time and thus evaporation. Inserts were placed in the VITROCELL 24/48 module under the same conditions as the final EB experiments with cells. The mean delivered dose inside the stainless steel inserts is a proxy for the dose deposited on the cell surface and absorbed in the cells (~intracellular dose).
The samples were diluted with mineral water (Spa Reine) and doped with the isotope-labeled compound D10-EB in a sealed vial. The HS sampler heats the vial at 70 °C for 30 min. During this period the EB transitions from the sample matrix into the vapor phase above. A fixed volume of the HS vapor is extracted from the vial and injected into a capillary column for GC separation. A mass spectrometer is used to detect and quantify the EB (Thermo HS-GC-MS). The MS single-ion monitoring mode of operation was used to enhance the detector sensitivity and selectivity. The internal standard method is used for the quantitative determination of EB. The quantification is based on the integrated peak areas of the most characteristic ions for EB and D10-EB.
Cell viability/cytotoxicity determination To assess cell viability or cytotoxicity, several assays are available for application in an ALI set-up. We performed the MTT (measurement of mitochondrial activity) and lactate dehydrogenase (LDH) assay (measurement of membrane integrity).
The conversion of MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) tetrazolium salt into its reduced formazan form was assessed. A MTT stock was prepared in Dulbecco's Phosphate Buffered saline (DPBS) at a concentration of 5 mg/ml. The MTT substrate is prepared in CCM and added to cells in culture at a final concentration of 1 mg/ml, and incubated for 2-3 h at 37 °C and 5 % CO2. The formazan product of the MTT tetrazolium accumulates as an insoluble precipitate inside cells as well as being deposited near the cell surface and in the CCM. The formazan must be solubilized prior to recording absorbance readings by e.g. isopropanol (2 h incubation, shaking at room temperature). The quantity of formazan (presumably directly proportional to the number of viable cells) is measured by recording changes using a multi-mode microplate reader in absorbance mode (570 nm; Clariostar, BMG Labtech, Offenburg, Germany). Results were expressed as percentages of CA-exposed negative control cells.
Cytotoxicity was assessed using the LDH detection kit, CytoTox-one Homogeneous Membrane integrity assay (Cytotoxicity Assay, Promega, Madison, USA) according to the manufacturer's instructions. Briefly, 20 h post-exposure, 200 μL of medium was added to the apical side from the inserts and incubated at 37 °C and 5 % CO2. After 20 min, medium from the apical and basolateral compartments were pooled and 100 µl of medium was incubated for 10 min with 100 μL LDH substrate mix. The reaction was stopped by the subsequent addition of 50 μL stop solution.
As high control, cells were exposed to the lysis solution of the CytoTox-one kit. Complete CCM incubated with the quantification reagents was used as background control. Fluorescence measurements were done using a multi-mode microplate reader in fluorescence mode (ex: 530-15 nm, em: 600-20 nm; Clariostar, BMG Labtech, Isogen Life Science, De Meern, The Netherlands).
Both assays were conducted in technical duplicates for five biological independent runs with each run existing of 4 replicate inserts. Changes in cell viability (MTT) and cytotoxicity (LDH) were analysed relative to CA and were assessed by mixed models while considering experiment ID (biological replicate) as random factor. Data were analysed using R 2 (R version 3.5.1 (2018-07-02) --"Feather Spray") and specific packages for mixed model analyses "lme4" (Bates et al., 2015) and "lmerTest" (Kuznetsova et al., 2017). Linear mixed models were fit by REML (restricted maximum likelihood) using function lmer with default parameters. P-value of fixed effects smaller than 0.05 was used as cut-off for statistical significance. For gene expression and protein expression, fixed effects are presented in this manuscript together with their 95% confidence interval (CI) in bar plots. 95 % CI means that there is 95 % chance that range contains the true mean or in other words we can interpret that if we repeat the study 100 times then we get the same values in 95 % of cases. Significance of result from CI can be assessed in the following way. If 95% CI captures the value of no effect (e.g. 0 for log2 fold change in case of gene expression) this represents a statistically non significant result (at significance level 0.05). If 95% CI does not include the value of no effect, then this represents statistically significant result.

Real-time qRT-PCR of stress response and pro-inflammatory markers
At 1 h post-exposure, total RNA was isolated from 2 pooled replicate inserts of A549 cells using the RNeasy Plus Micro Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. Extracted RNA was stored at -80 °C until further processing. Purity and concentration of extracted RNA were measured using the Nanodrop ND1000 spectrophotometer (NanoDrop Technologies, Wilmington, USA).
Then, complementary DNA (cDNA) was prepared from 500 ng RNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics, Mannheim, Germany) following manufacturer's instructions. qRT-PCR was then performed on a Lightcycler ® 480 RT-PCR System (Roche, Basel, Switzerland) using the following concentrations: 5 µl 2x Lightcycler ® 480 Probes Master (Roche Diagnostics), 0.5 µl 20x PrimeTime ® Assay (Integrated DNA Technologies, IDT, Iowa, USA), 2 µl RNAse-free water, and 2.5 µl cDNA (10 ng). The PrimeTime ® Assays are shown in Table 2. A non-template control (RNAsefree water, Probes Master, and PrimeTime ® Assay) was taken into account as negative control for each analyzed gene to exclude possible contamination from the used reagents. The thermal cycling conditions were as follows: pre-incubation for 10 min at 95 °C, followed by 45 cycles of denaturation 10 seconds (sec) at 95 °C; annealing for 30 sec at 62 °C; extension for 1 sec at 72 °C, followed by cooling for 10 sec at 40 °C. All PCR reactions were carried out in duplicate using 384-well plates.
To identify crossing point (Cp) values of the PCR reactions, the 'Second Derivative Maximum' method was used, which is included in the LightCycler ® 480 software. This method identifies the point where the reaction's fluorescence reaches the maximum of the second derivative of the amplification curve, which corresponds to the point where the acceleration of the fluorescence signal is at its maximum. The obtained Cp values were processed using qbase + (ΔΔCT method) to calculate relative gene expression of treated samples compared to (negative) control samples. The amplification efficiency was set to default '2'.
One of the key features in qbase + is that it includes a multiple reference gene normalization strategy to remove nonbiological variation, which is based on the GeNorm algorithm. Therefore, 3 reference genes were included in the analysis, i.e. actin beta (ACTB), glyceraldehyde 3-phosphate dehydrogenase (GAPDH), hypoxanthine Phosphoribosyltransferase 1 (HPRT1) (Casadei et al., 2011).
Fold changes (FC) relative to CA were calculated and logarithmically transformed (log2 scale) prior to statistical analysis. Significant changes relative to CA were assessed by mixed models while considering experiment ID (biological replicate) as random factor. Data were analysed using R and specific packages for mixed model analyses "lme4" (Bates et al., 2015) and "lmerTest" (Kuznetsova et al., 2017). P-value smaller than 0.05 was used as cut-off for statistical significance, and abs(mean(log2 FC)) above log2(1.5) was used to focus on biologically relevant expression changes.

Measurement of pro-inflammatory markers
After 20 h post-exposure, 200 µl of medium from the apical (washing for 20 min at 37 °C, 5 % CO2) and basolateral compartment were pooled and stored at -80 °C until use. Halt Protease inhibitor cocktail (ThermoFisher Scientific), 100 x diluted was added before freezing. As positive control, 20 µg/ml lipopolysaccharide (LPS, stock 1 mg/ml in PBS) was added to the apical medium of an insert and cells were incubated submerged in a CO2 incubator for 24 h. The samples (2 technical repeats/sample) were assayed using human IL-6 and IL-8 uncoated enzyme-linked immunosorbent assay (ELISA) assay (ThermoFisher Scientific). Briefly, 96-well plates were coated with capture antibody overnight at 4 °C. After 3 washings, wells were blocked with ELISA/ELISPOT diluent and incubated at room temperature for 1 h. A 2-fold serial dilution was made of the IL-6 or IL-8 standard. Standard dilutions and samples (100 µl; IL-8 samples were diluted 10 times) were added to the wells of the coated plate and these were incubated at room temperature for 2 h. After 4 washes, 100 µl detection antibody was added followed by an incubation of 1 h at room temperature. After 4 washes 100 µl diluted Avidin-HRP was added followed by an incubation of 30 min at room temperature. After 5 washes 100 µl 1X 3,3',5,5'-Tetramethylbenzidine solution was added followed by an incubation of 15 min at room temperature. Finally, 100 µl stop solution was added and the absorbance was measured at 450 nm (Clariostar, BMG Labtech). R software was used for analysis. The response curve between cytokine concentration and absorbance was analyzed using the four-parameter log-logistic function (package "drc", LL.4 method). Based on the output statistics of the model, cytokine concentrations of the samples were calculated and expressed in pg/ml.

ALTEX preprint published April 20, 2021 doi:10.14573/altex.2010211
8 Cytokine concentrations were adjusted for cell viability (MTT) and cytotoxicity (LDH) prior to analysis. Significant changes in cytokine concentration were analyzed relative to CA using ANOVA.

Quantitative in vitro-to-in vivo extrapolation (QIVIVE) calculations
Air concentrations used in ALI testing can be back-calculated to in vivo air concentrations. The calculation method applied here is the reverse of the method (eq. 1) used to calculate in vitro air concentrations from in vivo exposure concentrations as explained in former paragraph 'Estimation of in vitro test concentrations for EB'.

Results
Having performed preliminary tests and several follow-up optimizations, a simultaneous 4 h ALI cell culture exposure to CA, NO2, and EB vapors, followed by incubation under ALI conditions (20 h for cell viability/cytotoxicity and cytokine secretion, 1 h for gene expression) has been established.

Generated exposure concentrations
The measured NO2 and EB exposure concentrations for the 5 independent cell culture exposure runs can be found in Table 3. The data show a reproducible generation method with a coefficient of variation <5% for all tested concentrations. Relative humidity and temperature of the flow directly before cell exposure was 88.6 ± 4.0 % RH and 37.4 ± 0.1 °C.

Chemical analysis of delivered dose, calculation of delivery efficiency, and new in vitro dose-range
Delivered doses were determined in three other experimental runs on the same day by exposing stainless steel inserts containing 125 µl CCM to EB. After exposure, CCM was chemically analysed using HS-GC-MS. The generated EB concentration and deposited EB dose can be found in Table 4. The average delivered dose for about 30000, 40000, and 50000 mg/m³ exposure in stainless steel inserts was 9.10, 14.3, and 22.7 µg/4 h (Tab. 4). The average delivery efficiency calculated thereof is 0.10 % (0.08-0.12 %) as shown in Table  5, which has been used for the new dose-range calculation below. That the delivery efficiency would be low could be expected, given that the measured delivery efficiency of the structurally related methylbenzene is also low (0.036%) (Steiner et al., 2018).
The in vitro dose-range calculated previously from experimental data was 1-625 mg/m³ for 4 h exposure and assumed the same absorption in vivo as in vitro (see Materials

and Methods section 'Estimation of in vitro test concentrations for EB').
This range can be recalculated from the same acute studies to a new 4 h dose-range that considers the difference between in vivo absorption and in vitro absorption (delivery efficiency by proxy, calculated in current VITROCELL® 24/48 set-up), and the modified flow from 3 to 1.5 ml/m (Tab. S1 1 ). The reported in vivo absorption values for inhalation are 64 % for humans (Chin et al., 1980) and 44 % for rats (Bardodej and Bardodejova, 1970). The average measured in vitro delivery efficiency is 0.1 % (Tab. 5).

Tab. 5: Delivery efficiency (%) of EB measured in stainless steel inserts
Average ( Based on the preliminary dose-range (1-625 mg/m³), the in vivo absorption, and in vitro delivery efficiency, the new lower concentration limit of the dose-range becomes 640 mg/m³ (1*64 %/0.1 %); the new upper limit, becomes 275000 mg/m³ (625*44 %/0.1 %). The lower limit takes into account human absorption (ABS) (64%) as the lower in vivo concentrations are for human data (Tab. S1 1 ); the upper limit takes into account rat studies (hence 44% ABS). As a result, the preliminary dose-ALTEX preprint published April 20, 2021 doi:10.14573/altex.2010211 range of 1-625 mg/m³ changes to the current dose-range of 640 -275000 mg/m³ during the 4 h exposure window. An overview of the preliminary and current dose-ranges are presented in Table 6.
Four tested concentrations were within the current dose-range: 7000 mg/m 3 which resulted in no biological effects (data not shown), and the concentrations 30000, 40000, and 50000 mg/m 3 which caused concentration-related biological responses as shown in this study.

1-625
In vivo studies in Table  1 3 preliminary 640-275000 Preliminary dose range, in vivo ABS and in vitro delivery efficiency ('ABS')

Endpoint measurements
Quality control charts for cell viability (MTT) of negative (CA) and positive (NO2) controls were set up to check uniformity in the VITROCELL 24/48® device. The lower and upper limits are 85 % and 115 % cell viability for CA and 40 % and 69 % cell viability for NO2. The mean cell viability and Stdev (MTT) for each of the five CA and NO2 runs fitted within the limits (data not shown), so the experiments were valid for further endpoint measurement analysis. Different biological endpoints that are relevant for acute in vitro ALI screening of EB were assessed, including cell viability (MTT), cytotoxicity (LDH release), oxidative stress (HMOX1 and SOD2 expression), and pro-inflammatory response (CCL2, IL6, and IL8 expression and secretion). CA exposures were used as negative control. ALI cultures were exposed to NO2 (about 20 ppm in air) as a gaseous positive control for cell viability (MTT).

Cell viability/cytotoxicity determination
The mean cell viability (MTT) for A549 cells was 94 % after exposure to CA versus IC (P = 0.10) (Fig. 3). Exposure of A549 cells to NO2 (about 20 ppm) showed a significantly decreased mean cell viability of 67 % (P = 1.33E-13) compared to CA as expected. ALI exposure of A549 cells to EB induced a concentration-dependent decreased mean cell viability of 86 %, 77 %, and 47 % for exposure to about 30000, 40000, and 50000 mg/m 3 , respectively, as compared to CA. The results were statistically significant for the lower to higher tested concentrations, respectively P = 3.59E-4, P = 9.64E-9, and P = 4.19E-26.
The mean cytotoxicity (LDH release) for A549 cells was 4 % after exposure to CA versus IC (P = 0.09) (Fig. 4). Exposure of A549 cells to NO2 (about 20 ppm) showed a significantly increased mean cytotoxicity of 9 % (P = 2.55E-3), as compared to CA. A possible reason for a lower effect on cells compared to MTT might be that we measured LDH release too late (after 20 h post-incubation), since LDH has a half-life of approximately 9 h in CCM and NO2 might induce cell death faster than EB. Again, EB induced a concentration-dependent increased mean cytotoxicity which was significant for all concentrations 30000 (9 %, P = 7.74E-4), 40000 (20 %, P = 8.34E-12), and 50000 (36 %, P = 1.27E-24) mg/m³ as compared to CA.

Fig. 3: Change in cell viability (as % compared to clean air (CA)) of A549 cells after 4 hours (h) exposure to ethylbenzene (EB, nominal concentration 30000, 40000, and 50000 mg/m3), nitrogen dioxide (NO2, nominal concentration 20 ppm), and incubator controls (IC) based on 5 independent biological experiments
Box and whisker plots visualizing the range of the individual data points per condition. The upper whisker extends from the hinge to the largest value no further than 1.5 * IQR from the hinge (where IQR is the inter-quartile range, or distance between the first and third quartiles). The lower whisker extends from the hinge to the smallest value at most 1.5 * IQR of the hinge. Data beyond the end of the whiskers are called "outlying" points and are plotted with an "x"; other individual data points are overlayed and plotted with filled dots. The mean is indicated by a blue square. Different colors represent different biological experiments with their technical replicates.

ppm) and incubator controls (IC) based on 5 independent biological experiments
Box and whisker plots visualizing the range of the individual data points per condition. The upper whisker extends from the hinge to the largest value no further than 1.5 * IQR from the hinge (where IQR is the inter-quartile range, or distance between the first and third quartiles). The lower whisker extends from the hinge to the smallest value at most 1.5 * IQR of the hinge. Data beyond the end of the whiskers are called "outlying" points and are plotted with an "x"; other individual data points are overlayed and plotted with filled dots. The mean is indicated by a blue square. Different colors represent different biological experiments with their technical replicates.

Biomarker analysis
CA exposure significantly induced IL6 (log2 FC = 2.9, P = 3.21E-5) and IL8 (log2 FC = 2.5, P = 8.79E-6) gene expression as compared to IC (Fig. 5A,B), whereas at the level of cell viability/cytotoxicity, no significant difference between CA and IC was shown. No increase was observed for CCL2 expression in CA-exposed cells (log2 FC = 0.5, P = 0.25) compared to IC (Fig. 5C). NO2, which was used as positive control for cell viability, did not induce CCL2, IL6, and IL8 expression as compared to CA. For IL-8 secretion (protein level), normalized for MTT cell viability, the opposite was shown (Fig. 5D). A significant release of IL-8 was observed for exposure of A549 cells to NO2 (log2 FC = 0.94, P = 0.01), whereas CA exposure was comparable to IC. IL-6 secretion was below detection limit (data not shown), while CCL-2 secretion was not measured.
The oxidative stress marker SOD2 was significantly increased for all EB exposure concentrations (log2 FC of ~ 0.8), but no concentration-response was observed (Fig. 6A). NO2 showed a modest statistically significant increase of HMOX1 expression compared to CA (log2 FC 0.8, P = 9.04E-3). For EB exposure, there was a concentration-dependent decreased expression which was significant for all concentrations compared to CA, but the same decrease was observed for IC (log2 FC -3.7, P = 5.95E-16) and, for that reason, not a relevant marker to be studied in the context of EB exposure (Fig. 6B).

QIVIVE calculations
Three in vitro concentrations were selected for the QIVIVE calculations and were based on the in vitro results for cell viability: the highest tested concentration with no observed effect (NOEC, 7000 mg/m³, experimental data not shown), and both the lowest tested concentration (LOEC, 86% cell viability, 30000 mg/m³), and the highest tested concentration (47% cell viability, 50000 mg/m³) with observed effect, as shown in Figure 3. Results of the reverse calculations are presented in Supplementary Table 2 (values in italic). As no experimental value for absorption of EB in the mouse is available, the QIVIVE calculations were performed twice for the mouse: once with the value for human absorption (64%) and once with the value for absorption in rat (44%). The QIVIVE calculations were performed for different scenarios, described by the in vitro concentration (7000, 30000 or 50000 mg/m³), the species (human, rat, or mouse), and exposure time (15, 30, or 240 minutes); only the scenarios that are relevant for the discussion are shown in Supplementary Table 2.

Biological responses
ALI exposure of A549 cells to EB resulted in a significant concentration-dependent decrease in mean cell viability. To compare these biological changes with previous work, only a few papers on in vitro studies with EB or as BTEX mixture for respiratory toxicity were found (Liu et al., 2013(Liu et al., , 2014(Liu et al., , 2015. Short-term exposure of A549 cells in a hanging drop system, shown to be reliable for volatile compounds with high sensitivity (Liu et al., 2015), allowed the calculation of mass balances and derivation of an EC50 value for EB from the internal cellular concentration (11 mmol/kg), which was then related to the nominal air concentration of 9980 ppm (~ 43313 mg/m 3 ) for a 1 h exposure (Liu et al., 2014). This value is in the same order of magnitude as in our study. This points to the importance of a more reliable ALI exposure set-up for those compounds which occur in the gaseous phase upon exposure, and which are easily lost in a classical cell culture set-up (e.g. submerged cells, open system). Based on a literature review of gasoline and related compounds (Bisig et al., 2015(Bisig et al., , 2016(Bisig et al., , 2018Kunzi L. et al., 2015), markers for inflammation (e.g. tumor necrosis factor (TNF)α, IL6, IL8, CCL2) and oxidative stress (e.g. HMOX1, SOD2) should primarily be considered for these compounds. In this pilot study using EB, a concentration-dependent increase of proinflammatory markers CCL2, IL6, and IL8 was observed for EB-exposed A549 cells compared to CA, which was significant for all concentrations. Exposure of A549 cells induced the release of IL-8 in a dose-dependent manner, which was significant compared to CA. The oxidative stress marker SOD2 was significantly increased for all EB exposure concentrations, but no concentration-response was observed.

Delivery efficiency
This ALI EB vapor study calculated an average delivery efficiency of 0.10 %. The delivered dose of aerosol components to cells after in vitro ALI exposure depends predominantly on the physicochemical properties of the substances as was demonstrated by Steiner et al. (2018). The delivery efficiency of EB was not measured by Steiner and coworkers, but those of two structurally related substances (benzene and methylbenzene) were. For comparison, relevant physicochemical properties and (low) delivery efficiencies for these compounds are listed in Table 7.  The initially calculated administered in vitro dose-range of 1-625 mg/m³ for 4 h exposure to EB vapors was based on human and animal data (Tab. 1). In our calculation of the air concentrations, we assumed the same absorption in vitro as in vivo. The ALI experiments including this dose-range showed that exposure to different EB concentrations (up to 7000 mg/m³), induced no significant reduction in cell viability (data not shown). One of the reasons for initially showing less effect than expected could be that the in vitro absorption is considerably lower than the in vivo absorption, which was quantified as EB delivery efficiency in the course of this pilot study (Tab. 5). The link between 4 independent acute in vivo studies (Tab. S1 1 ) and the current dose-range for ALI exposure studies (Tab. 6), and the in vitro biological results is discussed below. In a first study, human volunteers who were exposed for 4 h to 425 mg EB/m³ experienced no nasal irritation (van Thriel et al., 2003); this no-effect dose corresponds with a 4 h in vitro dose of 846 mg/m³/4 h, which is just above the lower limit of the current dose-range (640 mg/m³/4 h) and below the highest tested in vitro concentration (7000 mg/m³/4 h) with no reduction in cell viability in the current study. In a second study with volunteers, subjects suffered severe and immediate nose and throat irritation after a 6 min exposure to 5000 ppm (21700 mg/m³) of EB (Yant et al., 1930). The corresponding recalculated in vitro dose of 1080 mg/m³/4 h is about 1.7 times above the lower limit of the current dose-range (640 mg/m³/4 h) and below the lowest in vitro tested dose with effect on cell viability (30000 mg/m³/4 h). At a dose of 7000 mg/m³/4 h that is about 7 times higher than the recalculated in vitro effect concentration of 1080 mg/m³/4 h, no reduction in cell viability was seen in the current study. Cell death seems to require a higher dose than irritation. Mechanistically, it is difficult to justify a comparison between in vitro cell death and severe irritation; it may be more justifiable to compare irritation with stress markers. In this study, the oxidative stress marker SOD2 was significantly increased for all EB exposure concentrations, but no concentration-dependent response was observed (Fig. 6A). In a third study, the RD50 (17620 mg/m³) for breathing reduction of mice exposed for 30 min to EB (Nielsen and Alarie, 1982) corresponds with a recalculated in vitro dose of 160000 mg/m³/4 h (calculated with 64 % absorption), which is above half of the upper limit of the current dose-range (275000 mg/m³/4 h). The 4 h LC50 rat (17360 mg/m³) of a fourth study (Smyth et al., 1962) corresponds with a recalculated in vitro dose of 150000 mg/m³/4 h which is above half of the upper limit of the current dose-range, and is 3 times above the highest tested dose in the current project (50000 mg/m³/4 h).
The doses tested with ALI in the current case showing effects on cell viability (30000 to 50000 mg/m³/4 h) are in the lower part of the current dose-range (640 -275000 mg/m³/4 h).
The 4 h in vitro test concentrations of 30000 and 50000 mg/m³ correspond with an insert mass load of 32.1 and 53.6 µg/cm² respectively (Tab. S2 1 ) and a decreased mean cell viability of 86 % and 47 %, respectively. These findings are in line with the results of Gohlsch et al. (2019) who demonstrated that cytotoxicity in A549 cells tested up to 100 µg/cm 2 could be a reliable in vitro indicator for in vivo toxicity.
It can be concluded that the preliminary dose-range was justified as starting point for setting the in vitro dosimetry but, as demonstrated in this pilot study, differences between in vivo absorption rate and in vitro deposition and delivery efficiency should be considered before setting the final dose-range for ALI experiments. 13 the acute DNEL for local effect (irritation of the respiratory tract) for workers (293 mg/m³) is 12 times lower. The indicative EU-OEL (occupational exposure limit) is 8 times lower (442 mg/m³, 8 h time weighted average (TWA) for working lifetime exposure) (EC, 2000). Volunteers exposed for 4 h to 425 mg/m³ experienced no nasal irritation (Tab. S1 1 ). Furthermore, the reverse calculation of an in vitro 4 h concentration of 7000 mg/m³ gives for short term exposure a human 15 min exposure concentration of 56300 mg/m³. For comparison, the indicative short EU-OEL is about 60 times lower (884 mg/m³ for 15 min (EC, 2000)), and in volunteers exposed to 239 mg/m³ for 15 min, no effect on pulmonary function was observed (Moscato et al., 1987, Tab. 1).

QIVIVE
At an in vitro 4 h concentration of 50000 mg/m³, the average cell viability was reduced to about 50 %. Reverse calculation of this in vitro concentration to an in vivo concentration resulted firstly in a rat 4 h exposure concentration of 5800 mg/m³. For comparison, the 4 h LC50 rat is 17000 mg/m³ (Smyth et al., 1962), and shows that in vitro cell death may be at least a qualitative measure for in vivo lethality caused by EB, which is in agreement with Gohlsch et al. (2019) who demonstrated that cytotoxicity in A549 cells could be a reliable in vitro indicator for in vivo respiratory toxicity. However, the in vivo lethality of EB could also be a result of systemic effects due to transport of the contaminant via the blood stream from the lungs to other organs, a process that is not simulated in the ALI exposure with lung cells; or it may be that the in vitro absorption is underestimated. Secondly, the IVIVE calculations of 50000 mg/m³/h resulted in a mouse 30 min exposure concentration of 7990 mg/m³ (with 44% absorption mouse = absorption rat). For comparison, the 30 min RD50 mouse for breathing rate reduction is two times higher (17200 mg/m³ (Nielsen and Alarie Y., 1982)). Thirdly, the IVIVE calculations of 50000 mg/m³/h resulted in a mouse 30 min exposure concentration of 5490 mg/m³ (with 64% absorption mouse = absorption human). For comparison, the 30 min RD50 mouse for breathing rate reduction is three times higher (17200 mg/m³). This may be an indication that in vitro cell death not being a measure for non-lethal effects (in this case breathing rate reduction in mice). To our knowledge, there are no data on lethality for mice available for comparison with the in vitro concentration causing cell death for the chemical tested in this study.
The QIVIVE calculation from in vitro air concentrations applied for testing cell viability to in vivo air concentrations appeared to be a promising method for screening for respiratory toxicity, as was shown with the LC50-rat, and also by Gohlsch et al. (2019). Regarding humans, no in vitro cell death may indicate that non-lethal effects (pulmonary function and nasal irritation) in humans will not occur. Since the mode of action is not reported, it may also be that cough/irritation in humans may occur from stimulation of nerves. A549 may not be the optimal model for examining this effect. For mice, the QIVIVE concentration causing 50 % breathing reduction exceeded the QIVIVE 50 % cell death concentration, showing that for breathing rate reduction, in vitro cell death does not seem to be the right measure.

Conclusion
This pilot study exposed a frequently used in vitro model (A549 cells) at the ALI to assess inhalation toxicity of the single compound EB. A generation facility to obtain EB vapors was successfully developed. Experimental conditions using the VITROCELL® 24/48 exposure system were optimized to achieve a (low) delivery efficiency that resulted in dose-dependent biological changes. The data demonstrate consistency in effect levels when comparing cell viability in the ALI experiments with known in vivo non-lethal effects in humans. It can be concluded that QIVIVE from in vitro air concentrations applied for testing cell viability to in vivo air concentrations may be a promising method for screening for acute inhalation toxicity. This pilot study with EB as a test compound demonstrated the approach of an ALI set-up, complemented with QVIVE calculations to predict human in vivo inhalation toxicity, which should be further validated with other respiratory toxicants.