Validation of an air–liquid interface toxicological set-up using Cu, Pd, and Ag well-characterized nanostructured aggregates and spheres

The deposition efficiency (E _SEM ) was defined as the fraction of the particles entering the NACIVT that were deposited on the surface of the silica wafers placed in the cell inserts. Silica wafers were placed in the inserts and treated as if they had been seeded with cells for an in vitro experiment, with 400 µl liquid in the well below the insert and no apical media in the inserts. As for the cell cultures, the deposition voltage was 2 kV. The total number of particles entering the NACIVT was determined by the SMPS. The aerosol particles used for the determination of E _SEM were sintered Ag with a count median diameter (CMD) of 53.2 nm, and a geometric standard deviation (GSTD) of 1.55.

The SEM images were analyzed using imageJ software (Rasband 1997–2015). By using the Smooth, Threshold, and Make Binary functions, the deposited particles were separated from the wafer background. The particles were then automatically registered using the Count Particles function. In the process, noise consisting of clusters of 1–3 pixels was also registered in significant amounts for several images. In order to avoid this, a lower size limit was set below which no particles were considered.

Wells 6, 18, and 24 of the NACIVT were used to determine E _SEM . For well 6, a total of five sub-areas were investigated. For each of the sub-areas, three SEM images were analyzed. For wells 18 and 24, three SEM images were analyzed.

The aerosol particles generated were characterized using a DMA coupled in series with an aerosol particle mass analyzer (APM, Kanomax model 3600) and CPC; the system is referred to as DMA-APM. The DMA selects a particle mobility size that in turn is characterized with respect to mass by the APM. By stepping the voltage in the APM, the average mass of the individual particles of the selected distribution can be estimated (Ehara et al. 1996). The specific set-up used in the study is described in more detail by Rissler et al. (2013). The set-up was calibrated using polystyrene latex spheres of a known density (1.05 g/cm³) as described by McMurry et al. (2002).

For spherical particles, the mass and diameter are related according to m∝d _p ³ . For aggregates formed by diffusion-limited cluster aggregation, the mass-mobility relationship is instead typically described by a power law (Park et al. 2003; Rissler et al. 2013):

The particle mass, m _agg , is a function of the mobility diameter, d _me , the K-factor, k, and the mass-mobility exponent, D _mm . The power law fitted to the experimental data is expressed in terms of SI units. The characterization of size-dependent mass was performed as a separate experiment, not during the actual cell exposures, but found to be stable between experiments. The power law relation allows for the transformation of the size number distribution to mass size distribution.

Samples of Cu and Pd aggregate particles were collected to determine the Sauter primary particle diameter. The Sauter diameters of the primary particles were calculated based on an analysis using imageJ (Rasband 1997–2015). The Sauter diameter is used to calculate the specific surface area, SSA _TEM , of the metal aggregates. The method, and the Sauter diameter, is described in detail by Bau et al. (2010) and Svensson et al. (2015).

The collection of particle samples took place in the NACIVT, alongside actual exposure of cell cultures, onto lacey carbon-coated copper grids glued onto semiconductor silicon substrates. The samples were then visualized using transmission electron microscopy (TEM, model 3000F 300 kV, JEOL).

The particle number dose, N _dose , was calculated as the product of the average total particle number concentration entering the NACIVT, c _N , the total aerosol volume passed over the insert, v, the deposition efficiency E _SEM divided by the insert membrane surface area, SA _insert :

Assuming a confluent layer of cells on the insert membrane N _dose represents the deposited particle number per cm² of cells (#/\({\text{cm}}_{\text{cell}}^{ 2}\)). For an experiment of 1 h exposure, v was 1500 cm³ for one insert of 6.5 mm in membrane diameter (0.33 cm²).

The mass dose, m _dose , was calculated using the total aerosol particle mass concentration, c _m , as follows:

For Cu and Pd aggregates, the c _m was calculated as the sum of the mass size distribution. The mass size distribution was calculated as follows:

The sum of the mass size distribution was calculated as

For Ag aggregates, the total mass concentration, c _m , was determined using TEOM measurements, c _m-TEOM .

The surface area distributions of the Cu and Pd aggregates were obtained by multiplying Eq. 4 by the specific surface area SSA_TEM, determined by methods described earlier:

The total surface area concentration, c _SA , was calculated as the sum of the surface area distribution, similar to Eq. 5. The surface area dose, SA _dose , was calculated as

For sintered Cu and Pd aerosol particles, the m _dose and SA _dose were calculated assuming that the d _me was the true physical diameter. Bulk densities of 8.96 and 12.02 g/cm³ for Cu and Pd, respectively, were used for the calculations.

Aggregates of Cu and Pd were, as visualized by TEM, fractal structures composed of smaller primary particles partly fused together (Fig. 6a–d). The sintered Pd aggregates were compacted to near spherical shapes, appearing crystalline. Sintered Cu aggregates appear spherical and no apparent crystal structure can be observed (Fig. 6e–h). The TEM images used for primary particle analysis were taken from the lowest exposure level of Cu and Pd aggregates according to methodology described in Section “Primary particle size and specific surface area—SSA.”

The numbers of primary particles measured for Cu and Pd aggregates were 886 and 705, respectively. The cumulative frequency distributions are shown in Fig. 7, revealing similar distributions for Cu and Pd.

The mean, µ, and standard deviations, σ, of the distributions were determined for the number, surface area, and mass-weighted distributions of the primary particles. The Sauter primary particle diameter, d _va , and specific surface area (SSA_TEM) were calculated using the methodology described by Svensson et al. (2015) (Table 2).

The SSA_TEM was used for the conversion from mass to surface area distribution, as described in Section “Calculation of dose.”

The number size distributions of the aggregates and spheres were measured online during the experiments using SMPS. Number size distributions of Cu₁, Pd₁, and Ag₁ are presented in Fig. 8a. The dilution was also successful and did not systematically affect the count median diameter (CMD) or geometric standard deviation (GSTD), see Fig. 8b and Table 3. Both particle number concentration, measured by SMPS, and mass concentration, measured by TEOM, indicate that the generated aerosols of metals aggregates or spheres were stable for the duration of the experiments.

From the number size distributions, the total particle number concentration, CMD, and GSTD were calculated (Table 3). In order for efficient referencing, the exposure levels were indexed numerically.

Due to short-circuiting of the unipolar charger in the NACIVT, the c _N and c _m-TEOM were not well correlated with the number, N _dose , and mass dose, m _dose , for Pd₁ and Pd₂. During the period for which the charger had short-circuited, an assumption of 20 % charged particles (instead of 100 %) was used based on the Boltzmann charge distribution. This means that rather than a deposition efficiency, E _SEM , of 36 %, the resulting average deposition efficiencies were 14 % and 26 % for Pd₁ and Pd₂, respectively. The difference between Pd₁ and Pd₂ with respect to deposition efficiency is explained by a difference in when the short circuit occurred. No TEOM measurement could be obtained for Cu₃ due to a sudden pressure change in the experimental system. The dose for Cu₃ is also lower than can be expected based on c _N and c _m-TEOM ; this is due to a lower particle flow as a result of the pressure change.

Due to the relatively high particle concentrations in the SDG-generated aerosols, the CPC functioned largely in its photometric mode. This mode registers the total light scattering of the aerosol and calculates a particle concentration. The normal CPC operating mode registers individual light pulses from individual particles. Instead of using a CPC, an upgrade of the experimental set-up could utilize an electrometer, which measures the current in the aerosol. This would make the set-up more robust regarding high-aerosol concentrations. Alternatively, a second dilution system could be introduced, allowing dilution before both the SMPS and TEOM.

The mass-mobility relations for Cu and Pd aggregates were determined in the size range of 40–360 nm (Fig. 9).

The k-factor and mass-mobility exponent, D _mm , were determined for the aerosol particles by fitting Eq. 1 to the mass-mobility data of the aggregates. The parameters were used to transform the number size distribution of the aggregates into mass size distribution and subsequently the delivered mass dose, see Section “Calculation of dose.” For the sintered particles, spherical particles were assumed (i.e., that the mobility diameter was equal to the geometric diameter).

The deposition efficiency, using polydisperse sintered Ag aerosol particles, was determined in the three wells of the NACIVT: wells 6, 18, and 24. For well 6, three SEM images from each of five separate regions on the wafer were analyzed (Fig. 10a). SEM images from wells 18 and 24, three from each well, were chosen from the central region of the wafers.

Results from well 6 show that there is variability in the deposition pattern over the wafer surface. Wells 18 and 24 lie in the same range of efficiency as those of well 6 (Table 4).

For primary cell cultures, this study's observation agrees with that of air–liquid interface exposure of Cu aggregates generated by spark discharge by Jing et al. (2015). The Cu aggregates by Jing et al. (2015) have a primary particle size of 5 nm and number mean diameter of 9.2 nm. Both studies observe a decrease in cellular viability for primary cell cultures with an increase in dose, expressed in terms of mass. For cytokine expression of SAEC because of Cu aerosol particle exposure, a significant dose response for TNF-α could be asserted (p < 0.05) by this study. However, it is clear from the tier 2 analysis that a dose of 0.4–10 µg/\({\text{cm}}_{\text{cell}}^{ 2}\) Cu aerosol particles produces a statistically significant (p < 0.05) increase in Il-8 expression. This is comparable to that of Kim et al. (2013), where a dose of 1 µg/cm² Cu, generated by nebulizing a suspension of 12 nm Cu, is associated with a fourfold to fivefold increase in Il-8 expression as compared to the control.

For cancer cell line A549, the tier 2 analysis in our study showed that exposure to Cu aerosol particles significantly reduced cellular viability. However, no significant dose response for A549 exposed to Cu aerosol particles could be asserted. This is in contrast to observations by both Jing et al. (2015) and Aufderheide et al. (2013). They reported a dose-dependent decrease in viability for A549 cell cultures. The doses of Cu achieved by Aufderheide et al. (2013), Cu aerosol with aerodynamic diameter of 0.9 µm, are higher by approximately a factor of 10, while showing a response regarding viability that is similar to the one found in our study. In Jing et al. (2015), the trend is the opposite; here, doses are approximately a factor of 10 lower than this study, but still associated with a similar decrease in viability to ours. In Elihn et al. (2013), a dose of 4.1 µg/cm² Cu aerosol particles was associated with a viability of 44 %, related to the control for A549 cells. Lower doses did not induce a drop in viability. The Cu aerosol particles, by Elihn et al. (2013), were generated from dry powder with a GMD of 180 nm, primary particles approximately 100 nm.

Herzog et al. (2013) produced cellular doses of Ag aerosol of 0.03 and 0.28 µg/cm², generated using a nebulizer from a suspension of <100 nm Ag nanoparticles. Triple cell co-cultures were used for the study, consisting of A549 with other epithelial-airway relevant cell types added. No significant effects on cytotoxicity, expression of Il-8, or genetic marker of Il-8 and TNF-α were observed for either dose levels of Ag on the cell cultures. Our study also found that Ag in similar doses to those of Herzog et al. (2013) does not induce any significant effects on A549 cell culture viability. However, we found that the Ag aerosol, in the tier 2 analysis, induced a significant effect regarding SAEC viability, as well as for Il-6, MCP, and TNF-α expression. In Herzog et al. (2014), a dose of 3.0 µg/cm² nebulized Ag, 116 nm in diameter measured by light scattering technique, with a subsequent 24 h post exposure period was associated with what is described as a moderate response in TNF-α in A549 cell co-cultures. In Holder and Marr (2013), cultures of A549 cells were exposed to either polydisperse or monodisperse sizes of Ag aerosols generated by an atomizer; the monodisperse fractions were selected using a DMA. The delivered dose of the polydisperse Ag aerosol was 0.7 µg/cm² and did not have any significant effects on viability, cytotoxicity, or Il-8 expression. The size-selected fractions of Ag aerosol were 50, 75, and 100 nm. For these exposures, the dose was expressed in terms of number, mass, and surface area. Although lower doses for the size-selected aerosols were in the range of 5–26 ng/cm², a statistically significant decrease in viability was observed. Translated into a number-weighted dose, for the size-selected fractions, it was in the range of 4.7–7.6 10⁶ #/cm², which is approximately a factor of 100 below doses in this study.

With regard to the Ag exposures in this study, the lack of primary particle analysis in the calculation of surface area dose was not possible, according to methodology described by Svensson et al. (2015).

A comparison of air–liquid interface dose levels and of liquid suspension is difficult. The theoretical tools described by Teeguarden et al. (2007) can translate a dose expressed in terms of units/cm³ (volume) to that of units/cm² (surface). A possible first-order translation of the doses in this study in order to compare them with other studies on toxicology suspension would be to simply add a length dimension to the dose, as was done by Karlsson et al. (2009). The dose of Cu₁ would be 4.52 µg/cm³ (µg/cm³) rather than µg/cm². A physical interpretation could be that all particles in 1 cm³ solution sediment onto a surface of 1 cm². The results from Karlsson et al. (2009) show that doses of 40 and 20 µg/cm² Cu nanoparticles, 200 nm in solution measured by light scattering, are associated with a drop in viable A549 cells to <10 %, compared to control, for both doses. In Lanone et al. (2009), a reduction in viability of A549 and THP-1 as a result of Cu particle exposure was observed in a dose-dependent manner in the range of 1–100 µg/cm³. The dose of Cu used in Karlsson et al. (2009) was higher than any achieved in this study and also associated with larger viability effects. Given the translation of dose previously discussed, Lanone et al.'s (2009) results are comparable to ours, where a significant response was found for both A549 and SAEC in the grouped analysis for viability. For Ag exposure in solution, Foldbjerg et al. (2011) found adverse and dose-dependent decreases in A549 viability as a result of Ag exposure, ~150 nm in diameter measured by light scattering technique, in the dose range of 5–15 µg/cm³. Beer et al. (2012) showed that a dose of 0.5–2 µg/cm³ Ag, 17–70 nm by TEM measurements, was associated with a dose-dependent decrease in A549 viability; our study found no significant effects on viability for A549 in either the grouped or differentiated analysis. Beer et al. also found that the effects of silver ions on the toxicity of Ag and Cu particles could not be neglected. Some research efforts have been devoted to the toxicology of Pd in suspension. Petrarca et al. (2014) showed that a dose of 10–100 µg/cm³ was associated with a dose-dependent drop in viability, down to <10 % for the highest dose. The effect is also dependent on the post exposure incubation time, and the ionic Pd component is important.

Ideally, in the field of air–liquid interface research, the question of solubility would pose less of a problem in comparison to solution toxicology. This is because the particles would encounter the cell cultures and their mucosa in the same fashion as that of an in vivo situation. This is the case for the SAEC cultures of this study. However, as described in the methodology section for the A549, a small liquid volume was added on the apical side of the insert during exposure. Without this apical medium, it is possible that a more clear response in A549 viability could be observed in this study. It should also be noted that real lung cells are in motion, stretching, and shrinking, as we breathe. This can lead to a higher degree of particle movement through the mucosa to the cell surface compared to the static inserts used in vitro.

Different air–liquid interface exposure systems will have different deposition efficiencies, depending on their design and functionality. The NACIVT utilizes an electrical field to ensure a high-deposition efficiency. This study used a 2 kV alternating field achieving a 36 % deposition efficiency, which is similar to that found in Jeannet et al. (2014) for the same system, with a constant field strength and particle size range. The electrostatic aerosol in vitro exposure system (EAVES) reports a deposition efficiency for 200 and 500 nm PSL spheres of 35 and 47 %, respectively (de Bruijne et al. 2009). Since deposition efficiency is expected to drop with the mobility diameter of the particles, the reported efficiencies for the EAVES system should be interpreted as higher than that of the NACIVT. This is to some degree in contrast to the NACIVT, where efficiency is reported to drop as size increases: for 500 nm, the deposition efficiency is below 15 % (Jeannet et al. 2014). The higher efficiency of the EAVES is most likely explained by a combination of design and higher deposition field strength. Another electrostatic device is the electrostatic particulate dosage and exposure system (EPDExS). This system is characterized with regard to deposition efficiency for several particle sizes and deposition voltages, 50–530 nm and 0–10 kV, respectively. For 2 kV deposition voltage in the EPDExS, which was used in this study in the NACIVT, 100 and 200 nm particles had a deposition efficiency of >95 and 55–60 %, respectively. The EPDExS, together with the EAVES, have a higher deposition efficiency for particles over 100–200 nm compared to the NACIVT. Again, this may be due to a difference in design in the exposure chambers: Both the EAVES and the EPDExS place the cell cultures closer to the aerosol inlet than the NACIVT. This is certain to affect the measured deposition efficiency as losses in the exposure chambers are dependent on both geometry and particle residence time before arriving at the site of intended deposition.

There are also air–liquid exposure devices that use mechanisms other than electrostatic deposition. The air–liquid interface cell exposure system (ALICE) that uses a cloud droplet formation and sedimentation to ensure deposition of particles has reported deposition efficiencies from 7 to 47–50 % (Herzog et al. 2013; Lenz et al. 2009). The Vitrocell^® system has been used in several air–liquid interface toxicological studies. The most common version of the system relies on sedimentation and diffusion for particle deposition. In Xie et al. (2012), a Vitrocell^® module was built into an in-house system where cells were exposed to nebulized ZnO. No clear deposition efficiency was reported in the study, making comparisons with the literature difficult. In Kooter et al. (2013), a deposition efficiency for 50 nm particles, based on modeling rather than experiments, was reported as ~2 % for the Vitrocell^®. The CULTEX RFS reports a size-dependent deposition efficiency over 70 % for particles in the size range from 100 nm to 10 µm. For aerosol particles below 100 nm, the efficiency decreases in a size-dependent manner to approximately 30 % (Aufderheide et al. 2013).