Due to the well-known shortcomings of submerged cell exposure systems for pharmaco-toxicological studies on pulmonary epithelial cells (such as uncertain effective biological dose for nanoparticles and unrealistic exposure scenario due to missing air-liquid interface and absence of mucus or lining fluid on respiratory epithelial cells), there have been several approaches to develop exposure systems for cells at the air-liquid interface. In the CULTEX system, particles are deposited from a continuous aerosol flow by diffusional and gravitational deposition onto cells . Bitterle and coworkers refined this approach by establishing a stagnation point flow over the cell layer, which enhanced the deposition efficiency of 200 nm submicron sized particles from about 0.7%  to 2% . Deposition efficiencies of 15-30% were obtained for charged 50-600 nm particles under the influence of an alternating electrostatic field  or in the EPDExS system utilizing unipolar electrostatic deposition downstream of a differential mobility particle sizer . None of these types of devices was designed for nebulized liquid substances (micron-sized droplets) and none of them (except for one Bitterle type system using a QCM for dose measurement ) allows for real-time, direct measurement of the amount of particles deposited onto the cells, but rather infer the cell deposited dose indirectly from the product of the measured particle concentration in the air and the empirically determined deposition efficiency.
Previously described methods for deposition of micron-sized particles or droplets onto cells have relied on inertial impaction as the deposition mechanism. The use of impactors or impingers [15, 17, 18, 35] requires high flow rates for effective particle deposition, which may impair cell viability  and the deposition efficiency is also strongly dependent on particle size . While the high flow speed (30 cm/s in the device by ) may simulate in vivo conditions in the respiratory bronchioles after a sharp inspiration , it exceeds typical flow speeds under normal breathing conditions and is completely unrealistic for the alveolar regime, where flow velocities are so small that deposition due to impaction can be disregarded . A second type of cell exposure system for liquids utilizes high velocity sprays to deposit the liquid directly onto the cells via inertial impaction [19, 20]. High-speed spray-deposition does not occur under in vivo conditions (except in the mouth after using a medical spray nebulizer), and both systems may induce cellular stress due to high speed droplet collisions. Again, none of the systems above provides information on the cell deposited dose.
The ALICE utilizes a completely different, more "gentle" technique of particle generation and deposition and it provides direct dose measurements. As discussed above, micron-sized droplets (MMD = 4.5-5.4 μm) are generated at a very high concentration (80-130 g/m3) by a vibrating membrane nebulizer and are uniformly deposited via cloud and single particle settling onto cells at the air-liquid interface. It is noteworthy that the cloud settling speed of 13-17 cm/s is too small for impaction to occur, as evidenced by the absence of enhanced droplet deposition near the center of the chamber, where the cloud speed is largest (see foils 5 and 8 in Figure 5). While impaction is a relevant deposition process in the upper respiratory tract, gravimetric sedimentation simulates the in vivo conditions of supermicron particle deposition in the alveolar region. The gentle form of particle deposition (compared to impaction) minimizes mechanical strain and subsequent stress for the exposed biological material, as was confirmed by the unaffected viability of human alveolar epithelial A549 cells after ALICE exposure (Figure 8). Furthermore, none of the ZnO nanoparticle doses applied here had any significant effect on cell viability, which is a pre-requisite for reliable measurements of particle-induced oxidative stress response, as performed in the current study.
In contrast to jet nebulizers or ultrasonic generators, vibrating membrane nebulizers do not exert high shear forces and do not heat the nebulized liquid during the generation process. Hence, it does not jeopardize the biological activity of potentially delicate therapeutic agents, such as biopharmaceuticals [23, 38]. Other positive aspects of the eFlow technology are the low residual volume in the reservoir (5%-10%), the large and stable mass/volume output (80-130 mg/m3 for 5 liter/min flow rate) and the low electrostatic charging of the droplets. Furthermore, the near micron-sized membrane pores may prohibit supermicron particles and nanoparticle agglomerates to exit the nebulizer. This could explain the measured reduced deposition efficiencies (0.3-0.4) for ZnO nanoparticle suspensions, which were not freshly prepared prior to nebulization. The prolonged residence time of the ZnO nanoparticles in suspension allowed for enhanced agglomeration (as confirmed by DLS measurements) leading to supermicron-sized agglomerates, which were too large to pass through the pores of the nebulizer membrane. Due to the increasing interest in the pharmacological and toxicological effects of nanoparticles, the prevention of large agglomerates of nanoparticles from getting delivered to the cells may be an attractive feature of the ALICE, which comes at the expense of reduced deposition efficiency (less than 0.57 ± 0.07).
The QCM is a highly sensitive, fast-response instrument for real-time determination of the deposited substance mass on the cells. The QCM has a response time of ~1 s, a large dynamic range (0.018-1800 μg/cm2; extendable to larger mass with non-linear correction factors) and high accuracy (here: <7.3% agreement with gravimetry). Its drawback for droplet measurements is that the QCM is sensitive to viscoelastic effects, which requires drying of the deposited droplets for accurate determination of the deposited nanoparticle/solute mass. However, the current study has shown that the QCM can be used as a real-time indicator for droplet deposition as long as the deposited liquid layer induces a frequency shift (QCM signal) of less than 715 Hz, the asymptotic value of an ''infinitely'' thick water layer. Finally, it must be stipulated that for stabilized nanoparticle suspensions, the (dry) nanoparticle mass can only be obtained from the QCM data, if the mass ratio of nanoparticles and stabilizing agent is known.
The repeatability of 12% in mass dose delivery is similar to the value of about 10% reported for the spray exposure unit RHINOCON (; error bar of the mean deposition in their figure 2). The entire area of the ALICE chamber can be used for cell exposure experiments, since the spatial variability of the ALICE is small (better than 1.6%). For comparison, the spatial variability of the RHINOCON system was 8% , which is probably due to the less uniform spray produced by commercially available spray units. In spite of excellent uniformity in the ALICE, agglomeration of nanoparticles, both prior and after nebulization of unstable nanoparticle suspensions, cannot be ruled out. Hence, for unstable nanoparticle suspensions the authors recommend minimization of potential agglomeration by reducing the processing time (less time for agglomeration), choosing low nanoparticle concentrations (less collision probability) and providing visual confirmation of spatial uniformity following nebulization (TEM measurements).
Although the deposition efficiency in the exposure chamber of the ALICE is relatively large (0.57 ± 0.07), the cell-specific deposition efficiency is currently limited to 0.072 (for two 6-well plates) due to the poor fractional cell coverage of standard transwell plates. For comparison with previously described cell exposure systems, it is important to note that the deposition efficiency of 0.072 represents the overall deposition efficiency, which is defined as the ratio of cell deposited and total mass of the substance filled into the nebulizer. Previous studies typically reported the internal deposition efficiency, the fractional deposition of the substance entering the exposure system, which does not account for reduced overall deposition efficiency due to, for example, residual substance in the nebulizer (particle generator) or substance loss in the conductive tubing upstream of the exposure chamber. The overall deposition efficiency is the more relevant parameter for materials, which are expensive or in limited supply (e.g. modern drugs), since it provides the basis for estimating the true costs of exposure experiments.
Unfortunately, previous exposure systems have either not been characterized in terms of deposition efficiency (no deposition efficiencies are given for any of the cell exposure systems for liquid substances described in the literature [18–20]) or the reported deposition efficiencies refer to the internal deposition efficiency, that is it discards residual material in the particle generator or losses upstream of the exposure system. For purely diffusion-based cell exposure systems, the internal deposition efficiencies were limited to 0.02 (relative to the dose entering the exposure system). Internal deposition efficiencies near unity (100%) were reported for electrostatic deposition of charged particles. Stevens and coworkers  deposited size-selected charged particles by passing the aerosol through a bipolar charger and subsequently through a differential mobility analyzer (DMA). Since the charging efficiency of a bipolar charger is limited to about 0.30 , the overall deposition efficiency is limited to about 0.3. For the bipolar electrostatic deposition system by Savi and coworkers , deposition efficiencies of 0.15-0.30 were reported. However, all of these values are upper limits for the overall deposition efficiency, since they do not account for additional losses due to, for example, residual mass in the particle generator (which can be large depending on the particle generator), losses in the transport lines and waste material during turn-on/turn-off phase of the exposure system. Hence, the overall deposition efficiency of 0.072 for the ALICE is better than, or within the range of, typically reported upper limits of overall deposition efficiencies for air-liquid cell exposure systems.
After nebulization of 1 mL of liquid, a 14 μm liquid layer is formed on the cells in the ALICE. Since the layer thickness is small compared to the diameter of the transwell inserts for the cells (>6.4 mm for standard 6-, 12-, and 24-well plates), nearly 100% of the nanoparticles (or solute molecules) interact with the cells due to diffusional motion and diffusional losses to lateral walls are small. Furthermore, in vivo epithelial cells are typically covered by a thin liquid layer. Hence, generation of a thin liquid film on the ALICE cells prevents evaporation of liquid from the cells and resembles physiological conditions, especially since the thickness of the deposited film can be regulated by epithelial cells via water transport to the basal side .
Although the nebulizer used in the current study can spray up to 5 mL of liquid per filling, the authors recommend using 1 mL for the following reasons: The QCM can be used as real-time indicator for 1 mL, or less, of sprayed liquid (frequency shift <715 Hz), the deposited liquid layer is sufficiently thin (14 μm) for efficient nanoparticle-cell interaction and the corresponding exposure time is short (5 min, which increases by about 2 min per additionally sprayed mL of liquid). Using less liquid reduces the liquid layer thickness and exposure time, but it also enhances the residual liquid fraction remaining in the nebulizer (0.05-0.1 mL; 10-20% for 0.5 mL) and hence the cell deposition efficiency.
The suitable concentration range of the nanoparticle suspension used in the ALICE is determined by the detection limit of the QCM. For nebulization of 1 mL of liquid, the lower detection limit of the QCM (0.018 μg/cm2) corresponds to a solute/nanoparticle concentration of 12.4 ppm (mass) in water. Assuming a maximum solute/nanoparticle concentration of 10%, a dose of 160 μg/cm2 per ALICE run can be supplied to the cells. The lowest concentration applied in the ALICE as yet, was 40 ppm of 15 nm gold nanoparticles, which resulted in a dose of 0.061 μg/cm2 (Figure 7). This dose level and a 10-fold higher dose were used by Brandenberger and coworkers who applied the ALICE to study cellular uptake and toxicological effects of a triple cell co-culture model simulating the alveolar lung epithelium due to gold nanoparticle exposure at the air-liquid interface .
Although air-liquid interface exposures have become more widely used in recent years, there are very few quantitative comparisons between air-liquid interface and submerged (conventional) dose-response curves after nanoparticle exposure. Since the ALICE provides direct accurate dose measurements, the data set provided here can be used for such a comparison. For both exposure conditions, A549 cells showed no significant response in IL-8 and HO-1 mRNA expression after exposure to less than 1.0 μg/cm2 ZnO nanoparticles (Figure 9). In contrast, significant differences between exposure methods were observed for larger concentrations. For the highest investigated dose (submerged: 5.0 μg/cm2; ALICE: 8.5 μg/cm2), the ratio of ALICE and submerged response was 0.26 and 5.7 for IL-8 and HO-1, respectively. This indicates a substantially mitigated and enhanced response for the ALICE relative to submerged conditions for the pro-inflammatory (IL-8) and oxidative stress marker (HO-1), respectively. The underlying reasons for these differences are currently unknown.
To put the dose levels obtained during in vitro exposures (ALICE: 0.3 - 8.5 μg/cm2; submerged: 0.75 - 5 μg/cm2) into perspective, it is instructive to consider that the current recommended Occupational Safety and Health Administration (OSHA) standard for ZnO fumes is 5 mg of ZnO per cubic meter of air (5 mg/m3) averaged over an eight hour work shift. Assuming an accumulated breathing volume of 3 m3 in 8 h, a lung surface area of 140 m2, an alveolar deposition efficiency of 10-50% (depending on particle size) and a 70% long-term clearance from the alveolar regime  the OSHA standard corresponds to an average (long-term) daily alveolar surface dose of 0.32-1.6 ng/cm2. Hence, the maximum lifetime dose accumulated by a worker is 3.6-18 μg/cm2 (5 workdays per week for 50 weeks per year over 45 years). The upper limits used for the current in vitro experiments are within this lifetime range and the lowest submerged dose of 0.75 μg/cm2 still corresponds to several years of exposure at the maximum allowed dose level. Since no significant in vitro cellular response was observed after challenging the A549 cells with 0.75 μg/cm2 (for 3 h post-incubation time), the current data suggest that ZnO nanoparticles do not pose a significant health risk, if the OSHA exposure limits are obeyed. However, it is unclear whether this result also holds for a chronic exposure scenario, that is continuous delivery of the mean daily dose (0.32-1.6 ng/cm2) over a lifetime.