Scholarly article on topic 'Characterization of the Vitrocell® 24/48 aerosol exposure system for its use in exposures to liquid aerosols'

Characterization of the Vitrocell® 24/48 aerosol exposure system for its use in exposures to liquid aerosols Academic research paper on "Environmental engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Toxicology in Vitro
OECD Field of science
Keywords
{"Aerosol exposure system" / Vitrocell / "Exposure characterization"}

Abstract of research paper on Environmental engineering, author of scientific article — Sandro Steiner, Shoaib Majeed, Gilles Kratzer, Grégory Vuillaume, Julia Hoeng, et al.

Abstract Background The Vitrocell® 24/48 is an advanced aerosol exposure system that has been widely used and characterized for exposure studies of cigarette smoke, but not for exposure to liquid aerosols with a low gas-vapor phase content such as the ones generated by electronic cigarettes. An experimental system characterization for this specific application was therefore performed. Methods Glycerol model aerosols of different particle size distributions, produced by a condensation monodisperse aerosol generator, were used for exposing small volumes of phosphate-buffered saline in the Vitrocell® 24/48. Disodium fluorescein, added as a tracer in the aerosol, allowed the exact aerosol mass deposition to be quantified fluorometrically. Results The aerosol mass delivery efficiency within the system showed variations in the range of ±25%. Aerosol dilution was not fully reflected in aerosol delivery, the achieved aerosol delivery should therefore be determined experimentally. Quartz crystal microbalances underestimated the deposition of liquid aerosols. Unequal delivery of particles of different sizes was detectable, although this effect is unlikely to be relevant under applied experimental conditions. Conclusions The Vitrocell® 24/48 aerosol exposure system can be used for exposures to liquid aerosols, such as those generated by electronic cigarettes. However, our results indicate that, compared with aerosol studies of cigarettes, a higher variability is to be expected.

Academic research paper on topic "Characterization of the Vitrocell® 24/48 aerosol exposure system for its use in exposures to liquid aerosols"

Accepted Manuscript

Characterization of the Vitrocell® 24/48 aerosol exposure system for its use in exposures to liquid aerosols

Sandro Steiner, Shoaib Majeed, Gilles Kratzer, Gregory Vuillaume, Julia Hoeng, Stefan Frentzel

PII: DOI:

Reference:

To appear in:

Received date: Revised date: Accepted date:

S0887-2333(17)30106-6 doi: 10.1016/j.tiv.2017.04.021 TIV 3985

Toxicology in Vitro

2 January 2017 19 April 2017 22 April 2017

Please cite this article as: Sandro Steiner, Shoaib Majeed, Gilles Kratzer, Gregory Vuillaume, Julia Hoeng, Stefan Frentzel, Characterization of the Vitrocell® 24/48 aerosol exposure system for its use in exposures to liquid aerosols. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Tiv(2017), doi: 10.1016/j.tiv.2017.04.021

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Characterization of the Vitrocell ® 24/48 aerosol exposure system for its use in exposures to liquid aerosols

Sandro Steiner PhD Shoaib Majeed MSc Gilles Kratzer MSc Gregory Vuillaume PhD Julia Hoeng PhD Stefan Frentzel PhD

Philip Morris International R&D, Philip Morris Products S.A. (part of Philip Morris International group of companies)

Quai Jeanrenaud 5

CH-2000 Neuchatel

Switzerland

Corresponding author

Sandro Steiner

Philip Morris Products S.A.

Quai Jeanrenaud 5 CH-2000 Neuchatel Switzerland

Tel: +41 (0)58 242 23 84 e-mail: Sandro.Steiner@pmi.com

Keywords

Aerosol Exposure system; Vitrocell; Exposure characterization Abbreviations

APS Aerodynamic Particle Sizer

CFP Cambridge Filter Pad

CMAG Condensation Monodisperse Aerosol Generator

CV Coefficient of Variation

DSF Disodium Fluorescein

GSD Geometric Standard Deviation

GVP Gas-Vapor Phase

PBS Phosphate Buffered Saline

QCM Quartz Crystal Microbalance

SD Standard Deviation

SMPS Scanning Mobility Particle Sizer

VC24/48 Vitrocell® 24/48 Aerosol Exposure System

WS Whole Smoke

n Delivery Efficiency

ncorr Dilution Corrected Delivery Efficiency

Highlights

Aerosol delivery in the Vitrocell 24/48 aerosol exposure system was characterized The system is suitable for liquid aerosols like the ones generated by e-cigarettes Empirical determination of aerosol delivery is recommended for each new aerosol type Quartz crystal microbalances are not suited for monitoring liquid aerosol deposition

Abstract

Background

The Vitrocell® 24/48 is an advanced aerosol exposure system that has been widely used and characterized for exposure studies of cigarette smoke, but not for exposure to liquid aerosols with a low gas-vapor phase content such as the ones generated by electronic cigarettes. An experimental system characterization for this specific application was therefore performed. Methods

Glycerol model aerosols of different particle size distributions, produced by a condensation monodisperse aerosol generator, were used for exposing small volumes of phosphate-buffered saline in the Vitrocell® 24/48. Disodium fluorescein, added as a tracer in the aerosol, allowed the exact aerosol mass deposition to be quantified fluorometrically. Results

The aerosol mass delivery efficiency within the system showed variations in the range of ±25%. Aerosol dilution was not fully reflected in aerosol delivery, the achieved aerosol delivery should therefore be determined experimentally. Quartz crystal microbalances underestimated the deposition of liquid aerosols. Unequal delivery of particles of different sizes was detectable, although this effect is unlikely to be relevant under applied experimental conditions. Conclusions

The Vitrocell® 24/48 aerosol exposure system can be used for exposures to liquid aerosols, such as those generated by electronic cigarettes. However, our results indicate that, compared with aerosol studies of cigarettes, a higher variability is to be expected.

Introduction

The in vitro toxicological testing of chemical compounds or mixtures of compounds commonly relies on dissolving the test substances in an adequate solvent. This solution can then be applied to the biological test system (e.g., a cell culture or a bacterial culture) at relevant concentrations. If aerosols are to be tested for their effects upon inhalation, this approach requires trapping the aerosols, usually on filters or in solvents, and eluting/processing the trapped material in adequate ways for the dissolution in a culture medium. However, this process has a limited feasibility as it may profoundly change the physical and chemical properties as well as the bioactivity of the test material [1-6]. In addition, the submersed state within a cell culture is not representative of the situation in the respiratory tract [6-8]. Toxicological testing of aerosols is therefore preferentially performed by exposure at the air-liquid interface, that is, the biological system is brought into direct contact with the aerosolized material to be tested.

Such toxicological testing requires the temperature and humidity of the test aerosols to be conditioned to meet specific requirements of the cell cultures, and to be diluted to achieve biologically relevant doses and dose responses [9, 10]. Therefore, aerosol exposure systems are commonly required. Various systems have been developed in the past few years [10, 11], all of which provide a high level of control over the exposure conditions, but also share common limitations. Owing to the complexity of the physical processes governing aerosol transport and deposition, aerosol delivery to the biological test system cannot be accurately predicted based on aerosol mass flows. In addition, aerosol parameters other than the relative humidity, temperature, and concentration may be changed; for instance, the particle number size distribution, the total particle mass, the partitioning of semi-volatile compounds between particulate matter and the gas-vapor phase (GVP), or even the chemical composition of the aerosol [2, 12-16]. In addition, a fully reproducible performance of the exposure systems cannot be implicitly assumed, as aerosol dilution may result in inaccurate dosing and a non-uniform aerosol delivery to replica cell culture inserts may occur [17].

These limitations render the comprehensive interpretation of the biological responses to the exposure difficult and, therefore, the validity of an in vitro aerosol exposure experiment depends largely on the characterization of the exposure system. Hence, the general system performance (e.g., performance stability across repeated exposures and dosing accuracy) and the achievable dose delivery should be addressed. In addition, different aerosols may display strongly differing dynamics during transport, dilution and deposition. Therefore, aerosol exposure system characterization needs to address the abovementioned aspects in an aerosol specific manner. Such characterization will enable the suitability of a system for application to the aerosol of interest to be demonstrated. This characterization is relevant in the context of the increasingly performed in vitro toxicological assessment of electronic cigarettes (e-cigarettes). The aerosols generated by e-cigarettes and comparable products are low in GVP constituents and mainly consist of liquid particles, of which the bulk material is propylene glycol, glycerol and water [18]. The dynamic behavior of such aerosols is, in many aspects, not comparable with the aerosols generated by cigarettes [19-22]; nevertheless, e-cigarettes may be subjected to toxicological assessment studies using cigarette smoke as a reference aerosol in the same exposure system.

For in vitro exposures of organotypic cell cultures to aerosols generated by tobacco products at our facility, we use the Vitrocell® 24/48 aerosol exposure system (VC24/48, Vitrocell® Systems GmbH, Waldkirch, Germany). The VC24/48 and the smaller version of an identical working principle, the

VC24, are among the most sophisticated commercially available exposure systems. They allow simultaneous exposure to serially generated aerosol dilutions (seven dilutions and one negative control in the VC24/48) in six replica positions per dilution and are capable of measuring aerosol mass deposition online using quartz crystal microbalances (QCMs).

Although these features make the system suitable for controlled high-throughput applications, they also introduce the risk of generating system-related artifacts and biases in the aerosol delivery. These include aerosol losses on the internal system surfaces (resulting in lower aerosol delivery than expected), incomplete mixing of the aerosol with the dilution air (resulting in non-uniform or unstable delivery to replica positions and/or complete dilution rows), and non-representative internal aerosol sampling (resulting in a biased delivery of different particle sizes or the relative delivery of particles and GVP).

Although there is an experience-based confidence in the performance of Vitrocell® systems for exposures to combustion-generated aerosols [23-25], less is known about their feasibility for exposures to liquid aerosols. A system characterization for e-cigarette applications was therefore attempted in the present work. To avoid any bias originating from product variability and product condition (e.g., the battery state), we use a method that relies on labeled model aerosols specifically developed for simulating exposures to liquid aerosols with low GVP content [26]. The main focus was the dose delivery, dosing accuracy (the accurate translation of aerosol dilution to aerosol delivery), uniformity of aerosol delivery to replica positions and the stability of system performance across individual repetitions. In addition, the potential effects of the internal (anisoaxial and anisokinetic) aerosol sampling on the aerosol mass delivery to the exposure chambers, and the reliability of the built-in QCMs used for online measurement of aerosol mass delivery were assessed.

Materials and Methods

Experimental setup

Aerosol generation and characterization

The model aerosols were generated in a TSI 3475 condensation monodisperse aerosol generator (CMAG; TSI, Shoreview, MN, USA). A detailed description of the aerosol generation has been provided previously [26]. Briefly, dry aerosols consisting of salt crystals of roughly 100-nm diameter were generated and bubbled through a saturator containing the heated aerosol material. The obtained nuclei-vapor mixture was passively cooled down in a condensation chimney, which resulted in super saturation and condensation of the vapor onto the salt nuclei. By varying the temperature of the aerosol material in the saturator, varying the volume flow through the saturator, and removing a part of the salt nuclei by filtration, the amount of aerosol material per salt nucleus and the resulting particle size distribution could be modulated.

Using disodium fluorescein (DSF, Sigma-Aldrich, Munich, Germany) as the source of the salt nuclei and glycerol (Sigma-Aldrich) as the aerosol material, we generated fluorescently labeled water soluble aerosols with tunable mean particle sizes of low geometric standard deviation (GSD). Upon deposition within an exposure system, DSF could easily be retrieved and fluorometrically quantified. The applied CMAG settings as well as relevant parameters describing the generated aerosols are specified in Table 1.

Aerosol characterization

We characterized the key flow and aerosol parameters: i) the mean aerodynamic particle sizes (geometric mean) and their GSDs, ii) the aerosol mass flowrate, and iii) the DSF mass flowrate. Mean aerodynamic particle sizes and their GSDs were measured using a TSI 3321 Aerodynamic Particle Sizer (APS) installed downstream of the CMAG (the required aerosol dilution was achieved using a TSI 3302A Aerosol Diluter with a 100:1 capillary). The mean aerodynamic particle diameter of one of the

tested aerosols was below the lower size cutoff of the APS. The size and stability of the particles were measured using a scanning mobility particle sizer (SMPS; TSI); however, not during the exposures as it was done with APS, but in separate runs of aerosol generation. The SMPS reported a mean particle size of 0.16 ^m, but since mean particle sizes reported by different types of instruments cannot be directly compared with each other, this aerosol can be referred to as <0.5 ^m. Aerosol mass flowrates were measured by trapping the aerosol on weighted Cambridge filter pads (CFP; Borgwaldt-kc, Hamburg, Germany) for defined periods of time (1-5 min, depending on the aerosol size), followed by gravimetric determination of the trapped aerosol mass. DSF mass flowrates were determined fluorometrically after eluting the trapped aerosol in 10 mL of Dulbecco's phosphate-buffered saline (PBS; Sigma-Aldrich) for 30 min in the dark on a horizontal shaker. For increasing the fluorescence activity of DSF, the pH of the used PBS was in all cases adjusted to 9 [27].

VC24/48 exposures

A schematic representation of the VC 24/48 system is given in Supplemental Information Figure 1. More detailed information on the VC24/28 system is available on the manufacturer's website (www.Vitrocell.com). Briefly, the test aerosol passes through the VC24/48 via the dilution system, which is located on top of the exposure module. The exposure module provides 48 exposure chambers, which, owing to the layout of the dilution system, are grouped into eight rows, each providing six replica positions. Upstream of each row, the aerosol in the dilution system can be diluted with air, resulting in a total of seven dilutions that can be tested simultaneously, and an additional row for control exposures to air only. Downstream of each row, a QCM chamber allows monitoring the mass deposition in the according row online. During exposures, the aerosol passing the dilution system is sampled by negative pressure into the exposure trumpets, which project into the exposure chambers (and the QCM chambers) and generate a stagnation flow condition over the exposed cell culture inserts (and the QCM crystals). Assessing aerosol delivery to the exposure chambers

ThinCert™ cell culture inserts (24-well format, transparent insert membrane, 0.4-^m pore diameter; Greiner Bio-One, Kremsmunster, Austria) were placed into 42 positions of the exposure module of the VC24/48 (an air-only exposure was not included; the eighth row was left empty). PBS (100 ^L, pH 9) was added to each insert, providing an aqueous surface that served as a surrogate for cell cultures, and was exposed to fluorescently labeled model aerosols. The exposure duration and the settings of the VC24/48 were those routinely applied during exposures of organotypic cell cultures [28-31]: PBS samples were exposed for 28 min, and the aerosol volume flowrate entering the VC 24/48 was 0.41 L/min (corresponding to a 55 mL puff from a tested tobacco product smoked according to the Health Canada smoking regimen with a puff release duration of 8 s). The volume flowrate through the exposure trumpets was, for all experiments and all positions, set to 2 mL/min. During exposure experiments, the whole system was kept at a stable temperature of 37 °C. Before entering the dilution system, the dilution air was brought to 37 °C and 60±5% relative humidity. Aerosol characterization was performed before, in the middle of and after the exposures, as has previously been described in detail [26]. The used experimental setup is described in Supplemental Information Figure 2. Therefore, the exposures were interrupted after 14 min and, after particle sizing and aerosol trapping on CFPs, continued for another 14 min.

The test exposures were conducted in two distinct modes: i) a serial dilution mode for assessing the dosing accuracy: The aerosols were diluted serially with cumulate dilution air-flowrates of 0.1, 0.2, 0.5, 1.0, 1.5, 2.0 and 3.0 L/min, resulting in aerosol concentrations of 81%, 67%, 45%, 29%, 22%, 17%, and 12% compared with the CMAG output, respectively. ii) A single dilution mode for determining whether the VC24/48 affects aerosol delivery in unforeseen ways, for instance, owing to the travelling distance within the system and the potentially resulting aerosol losses: A dilution to 22%

aerosol (1.5 L/min dilution air-flowrate) was applied directly after the aerosol entered the VC24/48, that is, upstream of the first row.

In the serial dilution exposure mode, five aerosols with mean aerodynamic particle diameters of <0.5, 0.8, 1.1, 1.4 and 1.6 ^m were tested (Table 1), each in four independent experimental repetitions. In the single exposure mode, only the aerosol with a mean aerodynamic particle diameter of 0.8 ^m was used.

As the effect of aerosol size is a specific focus of this work, potential changes in the aerosol size distribution during the passage of the aerosol through the system were assessed by installing the APS alternately up- and downstream of the VC24/48. As reported previously [26], no relevant size changes were detected.

Assessing QCM accuracy

For testing the accuracy of the QCM mass deposition measurements, clean QCM crystals were placed into the QCM housing of row three and five and cell culture inserts containing 100 ^L PBS (pH 9) in the according exposure chambers in the exposure module. The PBS samples and the QCM crystals were exposed to the aerosol with a mean aerodynamic diameter of 0.8 ^m in the single dilution mode for 21 min, during which time the mass deposition on the QCMs was continuously monitored. A total of four repetitions were performed, and the aerosol was characterized before each repetition. After exposure, the PBS samples were retrieved from the cell culture inserts and QCM crystals were removed from the QCM chambers. The deposited aerosol was washed off the QCM crystal surfaces by incubation in 1 mL PBS (pH 9) for 10 min in the dark, and the DSF mass in in the pooled PBS samples was determined fluorometrically. The accuracy of the QCMs was assessed by comparing the deposited aerosol mass per area reported electronically by the QCMs with that determined fluorometrically.

Sample collection and fluorescence measurements

After exposure, PBS samples were retrieved from the culture inserts and, along with the samples obtained from CFPs and QCM crystals, stored at 4 °C in the dark until further processing. Fluorometric measurements were performed in a Fluostar Omega Microplate reader (BMG Labtech, Ortenberg, Germany) at an excitation wavelength of 485 nm and an emission wavelength of 520 nm, as described previously [26].

Data processing

APS measurements were used for assessing the stability within and between individual runs of aerosol generation and were not further processed.

The primary readout obtained from exposed PBS samples and exposed QCM crystals was the fluorescence intensity, which was converted to DSF concentrations (or masses) using DSF standard curves (and the sample volumes). DSF concentrations and masses, therefore, represent the most robust data generated and were directly used for assessing the uniformity of aerosol delivery to replica positions.

For comparing the mass deposition reported by QCMs with the fluorometrically determined mass deposition, the delivered DSF masses were converted to aerosol masses using the DSF contents of the aerosols that were obtained by dividing the DSF mass flowrates by the aerosol mass flowrates, according to Equation 1.

Aerosol mass delivery

Aerosol mass flow rate

=-—- x Delivered DSF mass

DSF mass flow rate

To eliminate variations in the DSF content of the generated aerosols, aerosol mass delivery efficiencies, eta, were calculated according to Equation 2.

A erosol mass delivery efficiency (n)

DSF mass delivered to position XY DSF mass flowrate through exposure system x time

For the comparison of the delivery efficiency at different aerosol concentrations, values for n were further converted to dilution-corrected aerosol delivery efficiencies, ncorr, by correcting for the applied aerosol dilution according to Equation 3.

Dilution-corrected aerosol mass delivery efficiency (neon)

DSF mass delivered to position XY Theoretical DSF mass flowrate at position XY x time

ncorr values are based on the assumption that, irrespective of the aerosol flow velocity in the dilution system, sampling efficiency at the trumpet inlet is equal; that is, particle size and dilution-dependent variations in ncorr would be indicative for an effect of anisokinetic and anisoaxial particle sampling.

Statistical analysis

The characteristics of the generated aerosols were summarized by their mean, standard deviation (SD) and coefficient of variation (CV). The SD was estimated by taking the square root of the sum of the between- and within-experimental repetition variabilities, while the CV was estimated by dividing the SD by its mean (and expressed as a percentage).

To analyze delivered DSF mass data, as collected during the serial dilution exposure mode (with 5 distinct aerosols of various nominal mean aerodynamic particle diameters), a linear mixed model was used which incorporated the 3 fixed factors that needed to be contrasted (aerosol settings, aerosol concentrations, and replica positions) and the 3 random factors arising from the experimental repetitions and their interactions with the aerosol setting alone, and with the aerosol setting and concentration combined (data were collected in a split-split-plot design, see [32]). Variance heterogeneity across aerosol concentration levels was accounted for in the model. Note that one individual value was missing owing to sample loss and was replaced by the average of the values collected during other experimental repetitions under the same aerosol setting, concentration, and replica position.

To analyze the delivered DSF mass data as collected during the single dilution exposure mode (one single aerosol setting at one given concentration), a linear mixed model was used which incorporated the 2 fixed factors that needed to be contrasted (rows and replica positions) and the 3 random factors arising from the experimental repetitions and their interactions with row and replica positions (data were collected in a split-block design, see [33]).

For QCM-related aerosol deposition data, a linear mixed model was used which, similar to the previous model, incorporated the 2 fixed factors of interest (rows 3 or 5 and possible ways to

estimate aerosol deposition: PBS in cell culture inserts, PBS used to wash QCM, or QCM readout) and the 3 random factors arising from the experimental repetitions and their interactions with row and aerosol deposition method.

All statistical analyses presented in the manuscript were performed with the SAS system 9.2. Results

Aerosol generation and characterization

The measured aerosol parameters are summarized in Table 1. Note that the particle size distribution of the smallest aerosol (<0.5 ^m) was not measured during the exposure experiments, because its mean aerodynamic particle diameter was below the lower size cutoff of the APS (the size and stability of generation of this aerosol were measured using a scanning mobility particle sizer, not during the exposures but in separate runs of aerosol generation).

The aerosol sizes and size distributions were highly stable, with very small CVs for the mean aerodynamic particle diameters and for the GSDs, and was true within and across experimental repetitions. Aerosol and DSF mass flowrates (and consequentially their ratio, the DSF content of the aerosols) showed higher CVs (up to 32%).

VC24/48 characterization System performance stability

For assessing the stability of aerosol delivery, the delivery efficiencies n measured in the six positions of a given exposure row were averaged, yielding 140 values for serial dilution exposures and 28 values for single dilution exposures. These values were normalized to their averages across the four experimental repetitions, and the percentiles for the populations of the normalized values, giving a measure on the spread of the results, were calculated. The full range of variation (smallest to largest obtained value) spanned from 0.65 to 1.28 (serial dilution) and from 0.86 to 1.19 (single dilution). However, because the variations between repeated exposure runs cannot be fully attributed to the exposure system, but also contain a bias originating from sample processing and fluorescence measurement, the 10th and the 90th percentiles give a more reasonable measure for the stability of system performance across repeated exposures. In the data set obtained from serial dilution exposures, these percentiles captured values ranging from 0.87-1.16, 0.85-1.14, 0.74-1.20, 0.811.17 and 0.81-1.14 for the aerosols with mean aerodynamic diameters of <0.5, 0.8, 1.1, 1,4 and 1.6 ^m, respectively. The same analysis performed per dilution resulted in comparable values. Globally (across all tested aerosols and all dilutions), a range of 0.85-1.16 was covered, that is, 80% of all normalized values were within the range of the overall average ±15%.

The 10th and 90th percentile for single dilution exposures captured a comparable range of 0.88-1.15 (all exposure rows combined). The results are shown in Figure 1, with the data range of the average ±15 indicated as a visual aid.

Aerosol deposition uniformity across replica positions

Figures 2A and 3A show the delivered DSF mass for each position on the exposure plate, as measured in serial and single dilution exposures, normalized to the average per dilution row per repetition. Figures 2B and 3B show the same data after additionally averaging over the four repetitions. For our application of the VC24/48, aerosols with a mean particle diameter below 1 ^m are of the highest relevance; hence, aerosols with a mean aerodynamic particle diameter above 0.8 ^m were not displayed.

For deriving a measure for the uniformity of delivery to replica positions, dilution row-specific percentiles were calculated. The 10th and 90th percentiles are indicated in Figures 2 and 3; for

example, in serial dilution exposures to the aerosol with a mean aerodynamic diameter of <0.5 ^m at a concentration of 81% (first row), 80% of all values were captured by a range of 0.87-1.18. Note the occurrence of a few extreme outliers (mainly toward lower mass delivery). As stated above for the system's performance stability, the observed spread in the data cannot be entirely attributed to the system, but may also include variations from sample processing and fluorescence measurement. We therefore consider the range defined by the 10th and 90th percentiles as a reasonable estimate on the fold-aerosol mass delivery to an individual position in the VC24/48 as compared with the average delivery to all positions of the according dilution row in one individual repetition.

The uniformity of delivery to replica positions in serial dilution exposures was not equally distributed between different aerosol dilutions. For instance, for the 0.8-^m diameter aerosol, the 10th and 90th percentiles at an aerosol concentration of 81% ranged from 0.5 to 1.36, whereas the respective range for the <0.5 ^m diameter aerosol was 0.87 to 1.18.

Single dilution exposures resulted in higher delivery uniformity than serial dilution exposures. The 10th and 90th percentiles span a range of 0.8 at the lowest and 1.2 at the highest (Figure 3A). Moreover, no extreme outliers were detected.

Global system performance (performance stability and delivery uniformity combined) A measure for the overall performance of the VC24/48, accounting for inter-repetition variation and uniformity of delivery to replica positions, was derived using the aerosol-specific ncorr values for all dilutions and positions. A selection of calculated percentiles is listed in Table 2. Again, assuming that the sample collection and fluorescence measurement added additional variation, the 10th and 90th percentiles provide a reasonable measure for the variation introduced by the system. When repeated exposures are performed in the VC24/48, the aerosol delivery to a position of choice in a repetition of choice can reasonably be expected to lie within the range of the average delivery to all positions exposed to the same aerosol concentration ±25%.

Aerosol dilution versus aerosol delivery

Serial dilution resulted in a clearly discernible dose response (Figure 1) and the statistical model for the serial dilution experiment led to a p-value <0.0001 for the effect of aerosol dilution. However, aerosol concentration and aerosol delivery did not generally change by the same factor, as shown by the not fully linear shape of the graphs in Figure 1 and by the normalized representation of the dosing in Figure 4. The most evident and consistent inaccuracies in the relative dosing were observed when the aerosol was diluted from 81% to 67% and from 45% to 29%. When diluting the aerosol with a mean aerodynamic diameter of 0.8 ^m from 81 to 67%, the 0.84-fold change in the aerosol concentration resulted in a 0.99-fold change in n. When diluting from 45% to 29%, the 0.64-fold change in aerosol concentration resulted in a 0.81-fold change in n.

Single dilution exposures resulted in more accurate dosing in all rows, except for row 7, where a statistically significant overdosing was detected (a weaker, but significant dosing inaccuracy between rows 3 and 4 was also measured). Compared with row 6, a 1.28-fold higher delivery efficiency was measured in row 7. In contrast, at the dilution step from 17% to 12% aerosol (between the same dilution rows), serial dilution resulted in relative under-dosing in row 7 (i.e., only a 0.83-fold delivery was expected based on the dilution). This result indicates that the inaccurate dosing observed in serial dilution experiments may be a result of the applied dilution scheme and not of the intrinsic system properties.

Effect of the mean particle diameter on aerosol delivery

Dilution-corrected delivery efficiencies, ncorr, for each aerosol and aerosol dilution, averaged over the four experimental repetitions, are shown in Figure 5. The data show a trend toward a lower delivery efficiency with increasing particle size (this was confirmed by the statistical model with a p-value >0.0001).

QCM accuracy

Figure 6 shows the aerosol mass deposition per area on QCMs as directly reported by the QCMs and as determined by DSF quantification in the PBS used for washing the QCMs, as well as the area-specific aerosol mass delivery to PBS samples in the exposure chambers.

The aerosol mass delivery per area to cell culture inserts and the mass deposition per area on QCM crystals, as measured by DSF quantification, are in good agreement. Nevertheless, the delivery to QCMs was consistently lower (85-95%) than the delivery to PBS samples, the difference being close to statistical significance (p « 0.1).

The difference between the mass deposition per area on QCM crystals, as determined fluorometrically, and the mass deposition per area, reported by the QCMs, was large and statistically highly significant (p<0.0001). The QCMs detected only 10-30% of the total deposited mass, if the delivery to cell culture inserts was considered to be 100%.

Discussion

The repeated use of the VC24/48 for exposure studies of combustion-generated aerosols and studies on system characterization have led to an empirically-based confidence in the performance of the system [23-25] in this application. However, no such data-based evidence is available for its use in conjunction with liquid aerosols with low GVP content. Owing to the increasing interest in these aerosols, for example, in the context of the assessment of e-cigarettes or medical inhalation devices, a proactive system characterization in this regard represents a major asset for future research. The dosing accuracy, the stability of system performance across repeated exposures, and the uniformity of aerosol delivery to replica cell cultures are the key parameters defining the performance of an in vitro aerosol exposure system. For the latter two parameters, acceptance criteria as "in a defined number of repeated exposures, for X% of the exposed positions, the aerosol delivery should not deviate more than Y% from the average delivery to all positions exposed to the same concentration of the same aerosol" can be defined. We therefore report observed ranges of variation by providing a lower and an upper percentile defined by the values of X. The range of values captured by the two percentiles defines the spread of the data Y.

The differences in aerosol delivery to individual positions observed in the current work cannot solely be attributed to the VC24/48, but also include a bias potentially introduced during sample collection, sample processing and fluorescence measurements. As a consequence, an X-value of 100% and the according values for Y do not give a reasonable measure for the performance of the system, but rather under-estimate system performance. Exact determination of the variation introduced by sample collection, processing and analysis is experimentally challenging, but we estimate it to increase the overall spread by roughly 10%. Consequentially, we only consider the range covered by 80% of the measured values - located symmetrically around their average - to be descriptive for the true system related variation. Based on this approximation, it can reasonably be expected that, when repeated exposures are performed in the VC24/48, the aerosol delivery to a position of choice in a repetition of choice will lie within the range of the average delivery to all positions exposed to the same aerosol concentration ±25%. This still includes system-related outliers, that is, the range covering the large majority of achieved aerosol deliveries is narrower, as for instance indicated in Figure 2 with regard to delivery uniformity.

Furthermore, under the applied settings, acceptable values for X and Y will depend on parameters such as the biological test system used, the biological endpoints to be addressed, and the bioactivity of the test aerosol. Therefore, there is not a single value to define system performance. Taking this into account, Table 2 provides an estimation of the feasibility of the VC24/48 for an application of interest by giving an overview on the system performance for X values ranging from 40% (30th and 70th percentiles) to 100%.

Besides giving a measure of system performance, we also aimed at identifying properties and operation modes that affect system performance adversely that should be further investigated, adapted, or avoided.

In this context, with regard to the performance stability of the system, it is important to note that even though inter-repetition variations were observed, the global pattern of aerosol delivery/dosing (the relative delivery to individual dilution rows) was commonly comparable between repeated exposures (see Figure 1). This may indicate that the variation was the result of the overall state of the system, which, as a comparison of the stability observed under serial and single dilution settings indicates, was affected slightly stronger when highly concentrated aerosol was passing through parts of the dilution system. We therefore consider it likely that the stability of performance of the system was reduced by the buildup of deposits on the internal surfaces that affected the overall aerosol delivery. This is supported by the reduced dosing accuracy observed at high aerosol concentrations in serial dilution exposures and is in line with the literature, as, for example, Adamson and co-workers [17] showed that cleaning/maintenance has a relevant impact on the performance of Vitrocell® systems.

We consequentially conclude that, to achieve optimal system stability, exposures using high aerosol concentrations (in the range of >30%) should be avoided and that a high aerosol mass delivery should preferentially be achieved by prolonged exposure using low aerosol concentrations. The fact that the 22% aerosol was delivered more uniformly in the single than in the serial dilution mode (compare Figures 2 and 3) indicates that the level to which uniform delivery to replica positions is achieved is not a function of aerosol concentration or the aerosol flow velocity at the trumpet inlet. We assume that the process of dilution, as determined by the relationship between the dilution airflow, aerosol flow velocity and the aerosol concentration upstream of the dilution point, are of larger relevance in this context. Owing to the low sampling flowrate into the exposure trumpets, the aerosol is only sampled in close proximity to the walls of the dilution system (see Supplemental Information Figure 3). It is likely that the aerosol in this boundary layer is strongly affected by inefficient mixing, even if a laminar homogeneous flow in the center of the tubes of the dilution system is established. Although yet to be tested, for instance by computational fluid dynamics simulations, we hypothesize that incomplete mixing and/or the generation of turbulent flow patterns downstream of a dilution point affects the homogeneity with which the aerosol is sampled into the exposure trumpets of the according dilution row. Consequently, by optimizing the dilution scheme, the delivery uniformity might be considerably improved, which will be addressed in future work.

With regard to dosing accuracy, our results show that the inaccuracy in the relative dosing cannot be the result of the travelling distance/travelling time of the aerosol and the potentially occurring aerosol losses inside the system. The 29% aerosol was, for instance, consistently over-dosed relative to the 45% aerosol. In addition, the dosing accuracy of the 22% aerosol in the single dilution mode was generally higher than that in the serial dilution mode (relative to the 29% aerosol), that is, the absence of dilution upstream of an exposure row might have improved the dosing accuracy in this row. As for the delivery uniformity across replica positions, we therefore hypothesize that the dilution process itself and the resulting aerosol dynamics downstream of the dilution point has a

relevant impact on aerosol delivery and that dosing accuracy may be improved by identifying optimal dilution schemes.

The delivery of smoke generated from cigarettes to replica positions in Vitrocell® systems has been reported to be of higher homogeneity in previous studies [17, 23], most notably in one of them (Majeed et al. 2014) under identical system settings. In contrast to these studies, in the present work, large deviations occurred mainly toward lower aerosol delivery, whereas the upper limits were comparably well defined. In addition, Majeed and co-workers [23] reported a higher reproducibility at the highest smoke concentration tested. Other findings included that for volatile smoke constituents and chemical activity (that cannot be unequivocally attributed to particulate matter or the GVP), accurate dosing was achieved across the whole concentration range, whereas for particle mass deposition, this was only the case at low and intermediate smoke concentrations. Evidently, the physicochemical differences between glycerol aerosols, as used in this work, and complex aerosols, such as cigarette smoke, translate into globally different delivery patterns in the VC24/48. Although this is in part certainly the result of volatile compounds and particulate matter being delivered differently, differences in particle dynamics taking place during aerosol transport, dilution, sampling at the exposure trumpets and deposition in the exposure chambers may play important roles. Specifically, processes at the trumpet inlets were assessed in more detail in the present work by investigating the effect of particle size on particle delivery. Under the settings applied in this work, the aerosol reaches flow velocities of 0.3-2.0 m/s in the dilution system, whereas the velocity in the exposure trumpets is held stable at 0.005 m/s. This has two potential consequences: i) with increasing aerosol dilution, higher particle losses in the dilution system were expected, due to inertial impact in the U-turns that connect the dilution rows. Ii) Since sampling of an aerosol into the exposure trumpet requires particles to change the direction of movement by 90°, increasing the aerosol dilution was expected to decrease the efficiency of particle sampling at the trumpet inlets. Importantly, as both effects are the result of particle inertia, they were expected to be more pronounced for large particles than for small particles. Taken together, the two effects were expected to result in a decreased dilution corrected aerosol delivery efficiency ncorr with increasing aerosol dilution and with increasing particle size.

Whereas a trend toward lower ncorr values was indeed observed for larger particles, a unidirectional dependency on the aerosol dilution could not be detected, i.e. a positive correlation between increased particle size and decreased absolute values of ncorr was not observable. Anisoaxial/anisokinetic sampling can therefore be ruled out as a cause for the lower delivery efficiency of larger particles, or, at least, is not the only relevant mechanism involved. We consider the parabolic velocity profile of the laminar flow in the tubing of the dilution system in combination with the limited volume accessible to the sampling into the exposure trumpets as responsible for this, as it largely diminishes the effective difference between the flow velocity in the dilution system and the trumpet. (As shown in the Supplemental Information, computational fluid dynamics confirm the absence of a relevant difference between the flow velocity along the tubing of the dilution system and the flow into the exposure trumpet in close proximity to the system walls). Whether mixing effects as assumed to affect delivery uniformity and performance stability might be involved is to be addressed in follow-up studies; for instance, larger particles may be under-represented in close proximity to the walls of the dilution system—for example, owing to diffusional limitations— and accordingly under-sampled into the exposure trumpets.

Taken together, the presented data underline the need for aerosol-specific exposure system characterization, especially when studies comparing different aerosols are to be performed. We consequentially recommend that for any aerosol of interest, the aerosol dose delivery is determined under the system settings to be applied during the exposure experiments, before the actual exposures are conducted. Whereas for aerosols consisting of dry powders, QCMs provide the

method of choice for this, we demonstrate that for liquid aerosols, QCMs are not feasible. QCM mass measurements rely on the resonant frequency of the quartz crystals, which changes with the mass adhering to the crystal. The proportionality between a change in the resonant frequency and the deposited mass causing the change is only given for rigid deposits. In contrast, the mass of liquid deposits, depending on the thickness of the deposited layer and the viscosity of the material, may be underestimated because of viscous energy dissipation into the liquid layer [34, 35]. We indeed could show that this under-estimation of the deposited mass occurs and can clearly state that for liquid aerosols consisting of a material with a viscosity comparable to or lower than that of glycerol, QCMs cannot be expected to report accurate mass deposition and alternative methods for determining aerosol delivery have to be applied.

Summary and conclusion

In a detailed characterization of the delivery of liquid aerosols in the VC24/48, the performance stability of the system as well as the delivery uniformity to replica positions were assessed. When repeated exposures are performed in the VC24/48, the aerosol delivery to a position of choice in a repetition of choice can reasonably be expected to lie within the range of the average delivery to all positions exposed to the same aerosol concentration ±25%. As a result of the applied dilution scheme as well as the used aerosol concentration showing a relevant impact on system performance, the identification of optimal system operation conditions allows reducing the variability in aerosol delivery.

The aerosol dosing achieved by aerosol dilution is precise, but only partly accurate. Aerosol mass delivery can therefore not be inferred from the aerosol mass fed into the system and the applied dilution. Differences in the dilution corrected delivery efficiencies ^corr for different particle sizes are detectable, but they are independent of aerosol dilution. The size distribution of the fraction of a test aerosol ultimately reaching the exposure chamber is therefore the same at all applied aerosol dilutions - although not identical to the size distribution of the native test aerosol fed into the Vitrocell system. If different aerosols of highly different mean particle sizes are tested, the delivery efficiency will be lower for the aerosol of larger mean particle size, which should be compensated by adjusting the aerosol, dilution accordingly.

In conclusion, the VC24/48 is suitable for the controlled exposures to liquid test aerosols. However, compared with exposure studies using cigarette smoke, a higher variability between the deliveries to individual exposure chambers has to be taken into account. Therefore, we recommend empirically determining the optimal system settings that will result in the desired dose delivery and minimal variation. As the differential delivery patterns of highly different aerosols must be assumed to not be a VC24/48 specific phenomenon, this recommendation can be expanded to all aerosol exposure systems. Importantly, although intended to be used for this purpose, QCMs are not a feasible tool for this, as they considerably underestimate the mass deposition of liquid aerosols.

Acknowledgments

We thank Edanz Group Ltd. for their editorial assistance during the development of this manuscript and Samuel Kleinhans for assisting with the statistical analyses. The authors are grateful to Arkadiusz Kuczaj for providing Supplemental Information concerning aerosol flow in the system.

Competing interests

Authors are employees of Philip Morris Products S.A., which is the sole source of funding and sponsor of this project.

References

1. Murdock, R.C., et al., Characterization of nanomaterial dispersion in solution prior to in vitro exposure using dynamic light scattering technique. Toxicological sciences, 2008. 101(2): p. 239-253.

2. Sahu, S., et al., Particle size distribution of mainstream and exhaled cigarette smoke and predictive deposition in human respiratory tract. Aerosol Air Qual Res, 2013. 13: p. 324-32.

3. Hirsch, V., et al., Surface charge of polymer coated SPIONs influences the serum protein adsorption, colloidal stability and subsequent cell interaction in vitro. Nanoscale, 2013. 5(9): p. 3723-3732.

4. Panas, A., et al., Silica nanoparticles are less toxic to human lung cells when deposited at the air-liquid interface compared to conventional submerged exposure. Beilstein journal of nanotechnology, 2014. 5(1): p. 1590-1602.

5. Laurent, S., et al., Crucial ignored parameters on nanotoxicology: the importance of toxicity assay modifications and "cell vision". PloS one, 2012. 7(1): p. e29997.

6. Lenz, A.-G., et al., Inflammatory and oxidative stress responses of an alveolar epithelial cell line to airborne zinc oxide nanoparticles at the air-liquid interface: a comparison with conventional, submerged cell-culture conditions. BioMed research international, 2013. 2013.

7. Paur, H.R., et al., In-vitro cell exposure studies for the assessment of nanoparticle toxicity in the lung-A dialog between aerosol science and biology. Journal of Aerosol Science, 2011. 42(10): p. 668-692.

8. BéruBé, K., et al., Human primary bronchial lung cell constructs: The new respiratory models. Toxicology, 2010. 278(3): p. 311-318.

9. Grass, R.N., et al., Exposure of aerosols and nanoparticle dispersions to in vitro cell cultures: a review on the dose relevance of size, mass, surface and concentration. Journal of Aerosol Science, 2010. 41(12): p. 1123-1142.

10. Paur, H.R., et al., In Vitro exposure systems and bioassays for the assessment of toxicity of nanoparticles to the human lung. Journal Fur Verbraucherschutz Und LebensmittelsicherheitJournal of Consumer Protection and Food Safety, 2008. 3(3): p. 319-329.

11. Thorne, D. and J. Adamson, A review of in vitro cigarette smoke exposure systems. Experimental and Toxicologic Pathology, 2013. 65(7): p. 1183-1193.

12. Alonso, M., et al., Aerosol particle size growth by simultaneous coagulation and condensation with diffusion losses in laminar flow tubes. Journal of aerosol science, 1999. 30(9): p. 11911199.

13. Chang, P.-T., L.K. Peters, and Y. Ueno, Particle size distribution of mainstream cigarette smoke undergoing dilution. Aerosol science and technology, 1985. 4(2): p. 191-207.

14. Fujitani, Y., et al., Particle deposition efficiency at air-liquid interface of a cell exposure chamber. Journal of Aerosol Science, 2015. 81: p. 90-99.

15. Ishikawa, S., Y. Nagata, and T. Suzuki, Analysis of Cigarette Smoke Deposition Within an In Vitro Exposure System for Simulating Exposure in the Human Respiratory Tract. Beiträge zur Tabakforschung/Contributions to Tobacco Research, 2016. 27(1): p. 20-29.

16. Guha, A., Transport and deposition of particles in turbulent and laminar flow. 2008.

17. Adamson, J., et al., An inter-machine comparison of tobacco smoke particle deposition in vitro from six independent smoke exposure systems. Toxicology in Vitro, 2014. 28(7): p. 1320-1328.

18. Hajek, P., et al., Electronic cigarettes: review of use, content, safety, effects on smokers and potential for harm and benefit. Addiction, 2014. 109(11): p. 1801-1810.

19. Pichelstorfer, L., et al., Simulation of aerosol dynamics and deposition of combustible and electronic cigarette aerosols in the human respiratory tract. Journal of Aerosol Science, 2016.

20. Kleinstreuer, C. and Y. Feng, Lung deposition analyses of inhaled toxic aerosols in conventional and less harmful cigarette smoke: a review. International journal of environmental research and public health, 2013. 10(9): p. 4454-4485.

21. Feng, Y., et al., Computational transport, phase change and deposition analysis of inhaled multicomponent droplet-vapor mixtures in an idealized human upper lung model. Journal of Aerosol Science, 2016. 96: p. 96-123.

22. Zhang, Z., C. Kleinstreuer, and Y. Feng, Vapor deposition during cigarette smoke inhalation in a subject-specific human airway model. Journal of Aerosol Science, 2012. 53: p. 40-60.

23. Majeed, S., et al., Characterization of the Vitrocell® 24/48 in vitro aerosol exposure system using mainstream cigarette smoke. Chemistry Central Journal, 2014. 8(1): p. 62.

24. Talikka, M., et al., The response of human nasal and bronchial organotypic tissue cultures to repeated whole cigarette smoke exposure. International journal of toxicology, 2014: p. 1091581814551647.

25. Schlage, W.K., et al., In vitro systems toxicology approach to investigate the effects of repeated cigarette smoke exposure on human buccal and gingival organotypic epithelial tissue cultures. Toxicology mechanisms and methods, 2014. 24(7): p. 470-487.

26. Steiner, S., et al., A new fluorescence-based method for characterizing in vitro aerosol exposure systems. Toxicology in Vitro, 2017. 38: p. 150-158.

27. Kesavan, J. and R.W. Doherty, Use of Fluorescein in Aerosol Studies. 2000, DTIC Document.

28. Mathis, C., et al., A Systems Biology Approach Reveals the Dose-and Time-Dependent Effect of Primary Human Airway Epithelium Tissue Culture After Exposure to Cigarette Smoke In Vitro. Bioinformatics and Biology insights, 2015. 9: p. 19.

29. Iskandar, A.R., et al., Impact assessment of cigarette smoke exposure on organotypic bronchial epithelial tissue cultures: a comparison of mono-culture and co-culture model containing fibroblasts. Toxicological Sciences, 2015: p. kfv122.

30. Iskandar, A., et al., 3-D nasal cultures: Systems toxicological assessment of a candidate modified-risk tobacco product. ALTEX, 2016.

31. Zanetti, F., et al., Systems toxicology assessment of the biological impact of a candidate Modified Risk Tobacco Product on human organotypic oral epithelial cultures. Chemical Research in Toxicology, 2016. 29(8): p. 1252-1269.

32. Montgomery, D.C., Design and analysis of experiments. 2008: John Wiley & Sons.

33. Littell, R.C., et al., SAS for mixed models. 2006: SAS institute.

34. Zhuang, H., et al., Frequency response of a quartz crystal microbalance loaded by liquid drops. Langmuir, 2007. 23(13): p. 7392-7397.

35. Voinova, M., M. Jonson, and B. Kasemo, 'Missing mass' effect in biosensor's QCM applications. Biosensors and Bioelectronics, 2002. 17(10): p. 835-841.

Figure legends

Figure 1: Inter-repetition variation and aerosol dosing, based on disodium fluorescein delivery efficiencies, n, per aerosol and exposure mode. Individual repetitions and the average over four repetitions are displayed. Error bars represent standard deviations. The gray shading indicates the range of average ±15%.

Figure 2: Delivery uniformity to replica positions in serial dilution exposures. A) Disodium fluorescein (DSF) mass deliveries per position, for each repetition normalized to the averages per dilution row. Only the results for the 0.5 ^m and 0.8 ^m diameter aerosols are shown. The dashed lines and the numbers in the plots indicate the 10th and 90th percentiles of the collection of 24 data points per dilution row. B) The same experiments averaged over four repetitions. Error bars represent standard deviations, the dashed lines indicate the 10th and 90th percentiles of the collection of 42 data points.

Figure 3: Delivery uniformity to replica positions in single dilution exposures. A) Disodium fluorescein (DSF) mass deliveries per position, for each repetition normalized to the averages per dilution row. The dashed lines and the numbers in the plots indicate the 10th and 90th percentiles of the collection of 24 data points per dilution row. B) The same experiment averaged over four repetitions. Error bars represent standard deviations, the dashed lines indicate the 10th and 90th percentiles of the collection of 42 data points.

Figure 4: Relative dosing in serial and single dilution exposures. Aerosol delivery efficiencies, n, per dilution row (average over six positions) were normalized to the applied aerosol concentration and averaged over the four performed repetitions. For better readability, all obtained values were further normalized to those obtained for row 1. Horizontal connections between two subsequent rows indicate a full correlation between aerosol dilution and aerosol delivery at the according dilution step. A value of 1 indicates ideal dosing relative to row 1. Values above 1 indicate over-dosing, and values below 1 indicate under-dosing relative to row 1. Error bars represent standard deviations.

Figure 5: Dilution corrected delivery efficiencies, ncorr, measured in serial dilution exposures for the five tested aerosols, averaged over four repetitions and six positions per dilution row. Error bars represent standard deviations.

Figure 6: Comparison of the aerosol mass deposition reported by the QCMs and that determined fluorometrically. Exposures were conducted during 21 minutes in single dilution mode (22% aerosol), with the 0.8-^m diameter aerosol. Samples obtained from two dilution rows (two QCMs, 12 cell culture inserts) were pooled.

Figure 1

Serial dilution, <0.5 |im

Serial dilution, 0.8 um

67 45 29 22 17 12

Aerosol concentration (%]

67 45 29 22 17 12

Aerosol concentration (%)

Serial dilution, 1.111m

Serial dilution, 1.4 um

0.0025 0.002 0.0015 0.001 0.0005 0

67 45 29 22 17 12

Aerosol concentration {%)

67 45 29 22 17 12

Aerosol concentration (%)

0.0025 0.002 0.0015 0.001 0.0005 0

Serial dilution, 1.6 (im

Single dilution (22% aerosol), 0.8 (im

67 45 29 22 17 12

Aerosol concentration (%]

0.0004 0.0003 0.0002 0.0001 0

Row 1 Row 2 Row 3 Row 4 Row 5 Row 6 Row 7

Fig. 1

Figure 2

Serial dilution, <0.5 um Serial dilution, 0.8 tun

* Rep I a Rep II + Rep III o Rep IV * Rep I n Rep II ♦ Rep III • Rep IV

----- ------ "»"""ill o "o lAr> p„ ft ^ d"0" x,uo " o« * A ~M

V. ..His ik'H p°s -v" il 1s!:-- .1«! m

• „„ a» . ™ «--?... .......• 6Ü »", "* ' * . * ♦ .

E I o I

Aerosol concentration and sample position

123456 123456 123456 123456 123456 123456 123456 81% 67% 45% 29% 22% 17% 12%

Aerosol concentration and sample position

5erial dilution, <0.5nm

Serial dilution, 0.8 fim

ft ! s

£ I Q I

uHt -- 4- [u T

№... i№ ; 0.85 J! J1 i ML;d|

•I a.

■S s

i'liLfi^J-

123456 123456 123456 123456 123456 123456 123456 81% 67% 45% 29% 22% 17% 12%

Aerosol concentration and sample position

123456 123456 123456 123456 123456 123456 123456 81% 67% 45% 29% 22% 17% 12%

Aerosol concentration and sample position

Figure 3

Single dilution (22%) O.S |im

A Rep I □ Rep II o Rep III o Rep IV

£ 1.5

£ i iq

a 1.18 . _ 1.13 - - y

1.18 ♦

S | 'L" 1.03 0 * A g-a-o----- .....-

- * i □

2 S .........°°9 "0.90 og9 091 A. .........

t oj o.84 u'sy 0.84 2

irt ™

Q E 0.5

□ 0.78

123456 123456 123456 123456 123456 123456 123456 Row 1 Row 2 Row 3 Row 4 Row 5 Row 6 Row 7

Sample position

Single dilution (22%), 0.8 um Averages over repetitions

i/> o re -

£"Q <u

l/l ra Q E

"iTiiT

0.91 1

123456 123456 123456 123456 123456 123456 123456 Row 1 Row 2 Row 3 Row 4 Row 5 Row 6 Row 7

Sample position

Figure 4

-s a 1.2

in □ >- QJ

ro — O fü

E bo z

—A— Serial dilution

■<>•• Single dilution

.........5" i .......... Y........^ t

Row 1 Row 2 Row 3 Row 4 Row 5 Row 6 Row 7 81% 67% 45% 29% 22% 17% 12%

Row number and aerosol concentration (serial dilution)

Figure 5

0.81 0.67 0.45 0.29 0.22 0.17 12% Aerosol concentration

■ <0.5 tun

■ 0.8 tun

□ 1.1 iim

□ 1.4 tun

□ 1.6 tun

Figure 6

"S O £L

to CUO

35 30 25 20 15 10 5

Rep 1 Rep 2 Rep 3 Rep 4

□ Cell culture inserts (PBS, fluorometry)

□ QCM wash (PBS, fluorometry)

■ QCM readout

Table 1: Aerosol generation and characterization (CMAG: condensation monodisperse aerosol generator, DSF: disodium fluorescein, GSD:

geometric standard deviation)._

Nominal

mean aerodyna

mic particle diameter

<0.5 um

0.8 urn

1.1 um

1.4 um

1.6 um

CMAG settings

Total flow (rotam eter scale value3) Saturat or flow (rotam eter scale valuea) Screen flow (rotam eter scale valuea) Saturat or

temper ature (°C) Reheat er

temper ature (°C) Aerosol characteriz ation (n=12) Aerosol mass flow rate (mg/mi n) DSF mass flow rate (^g/mi n)

Mass ratio DSF/Ae rosol Mean aerody namic particle diamet er (^m)

a exact volume closed, a value

Aver ± sd (C

age V)

0.02 0.0 (3

8 ± 09 2)

9.8 ± 1.8

0.38 ±

0.1 (3 3 4)

0.16 0.0 (1. b ± 02 1)

1.78 0.0 (0.

b ± 1 5)

1.5 - 2

Aver age

0.00 5

0.83 ±

1.29 ±

0.1 (9. 82 6)

10.2 ± 0.9

0.0 01

(9. 3)

0.0 (3. 2 0)

0.0 (1. 2 6)

3.75 - 4.25

Aver age

4.28 3

0.00 25

1.13 ±

1.36 ±

0.2 (5. 14 0)

10.6 ± 2.1

(3. 0)

0.0 (1.

6.75 - 7.25

Aver age

1.41 ±

9.6 ± 1.4

0.00 1

1.41 ±

1.39 ±

(4. 6)

0.0 (1. 2 3)

0.0 (1. 2 6)

7.75 - 8

Aver age

9.30 3

3.5 - 4.5

0.29 (3. 4 2)

6.5 ± 0.3

0.00 07

0.00 (5. 004 4)

1.62 ± 0.03

1.36 ± 0.02

flow rates for the indicated scale values are specified by TSI. A value of 1 refers to the according valve being completely of 8 corresponds to roughly 3.6 L/minute

bthe size distribution of the aerosol with a mean aerodynamic particle diameter below 0.5 ^m was measured with SMPS. Measurements during the exposures were not performed, the values listed here orginate from separate runs of aerosol generation.

Table 2: Values for the dilution corrected DSF mass delivery efficiency ncorr to individual replica positions were normalized to the average over all exposed replica positions. A selection of lower and upper percentiles describes the spread of the obtained data around the average values (which due to the normalization is equal to 1)

Mean 70th- 30th 75th- 25th 80th- 20th 85th- 15th 90th- 10th 95th- 5th 100% coverag e

exposur e mode aerodynami c particle percentile s (40% coverage (50% coverage (60% coverage (70% coverage (80% coverage (90% coverage

diameter ) ) ) ) ) )

<0.5 ^m upper lower 1.12 0.93 1.14 0.90 1.17 0.87 1.22 0.84 1.25 0.75 1.30 0.62 1.64 0.11

0.8 ^m upper 1.09 1.11 1.12 1.15 1.19 1.25 1.97

lower 0.94 0.90 0.89 0.86 0.82 0.65 0.14

c o 1.1 ^m upper lower 1.14 0.95 1.16 0.92 1.18 0.90 1.20 0.82 1.23 0.70 1.34 0.37 1.57 0.12

d ÜS 'H 1.4 ^m upper 1.13 1.16 1.19 1.22 1.33 1.40 1.57

e to lower 0.88 0.85 0.82 0.77 0.70 0.50 0.32

1.6 ^m upper 1.09 1.12 1.16 1.20 1.29 1.41 2.71

lower 0.92 0.90 0.86 0.78 0.60 0.52 0.39

0.5 - 1.6 ^m upper lower 1.11 0.93 1.14 0.90 1.17 0.86 1.19 0.82 1.24 0.74 1.34 0.51 2.71 0.11

Single dilution 0.8 ^m upper lower 1.07 0.90 1.10 0.88 1.13 0.86 1.17 0.83 1.23 0.81 1.30 0.74 1.81 0.62

Highlights

• Aerosol delivery in the Vitrocell 24/48 aerosol exposure system was characterized

• The system is suitable for liquid aerosols like the ones generated by e-cigarettes

• Empirical determination of aerosol delivery is recommended for each new aerosol type

• Quartz crystal microbalances are not suited for monitoring liquid aerosol deposition