The toxicity of AgNPs to eukaryotic cells, bacteria and multicellular organisms has been investigated in a number of studies, most of which overlook fundamental issues. For instance, not all studies indicated whether the nanoparticles were purified after synthesis or not, and many studies failed to describe the behavior of nanoparticles in the given biological media [26, 27]. The purpose of this study was to investigate the toxicity of a panel of highly purified and well-characterized AgNPs with a specific focus on size- and coating-dependent effects, and to explore the mechanisms of possible differences in toxicity.
In the present study we used exposure concentrations in the range of 5–50 μg/mL, primarily based on previous studies of Ag nanoparticles and eukaryotic cells . This may be related to a possible human exposure by using exposure data from a AgNPs manufacturing facility (maximum level 289 μg/m3), and by applying the same assumptions and calculations as in the study by Wang et al.. A concentration of 10 μg/mL would then approximately correspond to the total cellular deposition following 74 working weeks (8 h per day, 5 days per week). Thus, the doses used should be considered high but likely possible to be reached following years of exposure, or after acute accidental exposure.
The results showed a clear size dependent toxicity for the tested AgNPs since only the 10 nm AgNPs were cytotoxic for the BEAS-2B cells starting at doses of 20 μg/mL in the Alamar Blue assay. There was, however, no difference in toxicity between the 10 nm citrate and 10 nm PVP coated AgNPs, suggesting that the size rather than the capping agent was the property that triggered toxicity. Other studies have also reported higher toxicity for smaller compared to larger sized AgNPs. For example, Carlson et al. showed an increased ROS generation for 15 nm hydrocarbon coated AgNPs as compared to 55 nm, which also correlated with decreased cell viability in macrophages. Furthermore, Liu et al. found that 5 nm AgNPs were more toxic than 20 and 50 nm AgNPs in four cell lines (A549, HepG2, MCF-7, SGC-7901). Using the same kind of AgNPs as in the present study (10 nm PVP coated), George et al. reported approximately 35% cytotoxicity following exposure of fish gill cells to doses of 25 μg/mL; thus, a very similar extent of cytotoxicity as in the present study, and no cytotoxicity for the 40 nm. Recently also Wang et al. showed that 20 nm citrate and PVP coated AgNPs induced more cellular toxicity than larger particles (110 nm) and furthermore that the citrate coated 20 nm generated acute neutrophilic inflammation in the lungs of exposed mice to a much higher extent when compared to the larger ones.
In order to explore the genotoxicity of AgNPs in lung cells we used the alkaline version of the comet assay and γH2AX foci induction. In contrast to the size-dependent effect on cell viability, we found that all tested AgNPs induced DNA damage after 24 h as reported by the comet assay, but without γH2AX induction. There were, however, no signs of DNA damage at earlier time points (4 h) suggesting indirect genotoxic mechanisms that take more time to occur. The effect on cell viability and the DNA damage may potentially be explained by ROS generation . However, we could not provide any evidence of intracellular ROS production preceding (geno)toxicity, thus contradicting many other published in vitro studies . The comet assay is a highly sensitive method and widely applied in nanotoxicological studies , but it gives limited mechanistic insight. Thus, the more precise mechanism of genotoxicity warrants further investigation. One hypothetical explanation for the detected DNA damage could be the interaction of the particles with the DNA repair pathways. Such interactions have been previously reported for AgNPs e.g. reduction of the formamidopyrimidine DNA glycosylase activity  and down regulation of genes involved in DNA damage response/repair system (XRCC1 and 3, FEN1, RAD51C, RPA1) .
Next we investigated the mechanisms behind the observed size-dependent cytotoxicity by analysis of the cellular uptake and uptake mechanisms, intracellular localization, agglomeration and the released Ag fraction in cell medium. The TEM images showed that all AgNPs were mainly localized within membrane-bound structures. No AgNPs were detected in the cell nuclei, although nuclear presence has previously been reported for BEAS-2B cells (43–260 nm AgNPs) , U521 cells (6–20 nm AgNPs) , HaCaT cells (30 nm AgNPs)  and hMSC (AgNPs <50 nm) . Multi-lamellar structures consistent with autophagy were observed for the 10 nm sized AgNPs. Induction of autophagy has been reported for several engineered nanoparticles, including AgNPs and Ag nanowires [37, 39], and may represent a common cellular response to nanoparticles. In general, differences in the intracellular localization of the particles could not explain observed differences in toxicity. Moreover, when comparing the total cellular Ag content, determined by AAS, we could not detect a higher cellular dose of the most cytotoxic NPs, the 10 nm particles. Thus, the intracellular dose, that often is regarded to be of importance for toxicity, could not explain the higher toxicity of the 10 nm AgNPs. The total uptake was around 2–4 pg/cell for the coated and somewhat higher (approx. 10 pg/cell) for the uncoated AgNPs, in agreement with our previous studies of cellular uptake of AgNPs in BEAS-2B and A549 cells . Observed findings are also within the same range as reported in a recent study on HepG2 cells with 6.8 pg Ag/cell following exposure to 10 μg/mL AgNPs for 24 h . Interestingly, the same study further attempted to distinguish between AgNPs and Ag ions in the cells by using Triton-X 114-based cloud point extraction of the cell lysates. The authors concluded that approximately 10% of the total amount of Ag within the cells existed as Ag ions. Since this value was higher than the corresponding fraction of Ag ions prior to exposure (5.2%), they argued that transformation of AgNPs to Ag ions could have taken place intracellularly.
In the present study we carefully addressed time-dependent changes in agglomeration, an aspect often completely overlooked in studies within the field of nanotoxicology. There are several factors that should be taken into account when evaluating the agglomeration of the different particles. For example, it is well known that the intensity as measured using light scattering techniques increases with particles size in a non-linear manner, and at the same time sedimentation reduces the intensity thus making the interpretation non-trivial. Clearly, however, there was an evident difference in stability between the citrate and PVP coated 10 nm particles. This could be explained by the more rapid displacement of the electrostatically weakly bound citrate with medium components (e.g. aminoacids), triggered by the high ionic strength of the medium, when compared to the non-charged larger PVP polymer capping agent. A more rapid breakdown of the stabilizing coating will evidently affect the stability of the particles. The lower stability of the citrate coating also resulted in higher Ag release compared with the PVP coated Ag NPs in cell medium after 4 h. However, observed differences in agglomeration did not translate to differences in Ag release or toxicity after 24 h. This is perfectly in line with the recent study by Wang et al. showing higher Ag release in BEGM media from 20 nm citrate coated Ag nanoparticles when compared to PVP coated particles at 6 h, followed by a very similar release at 24 h. Also, in accordance with our results, they report higher Ag release and toxicity from the smaller (20 nm) compared to the larger (110 nm) Ag nanoparticles. In all, the primary particle size seems to be more important than the size of the agglomerates for Ag release  and, according to the present study, for toxicity as well.
Proteins in the cell medium are known to be important for the stabilization of citrate coated AgNPs via the formation of a protein corona . Therefore the low protein content of our working medium (serum free) could partially explain the agglomeration of the citrate coated particles upon dispersion. Ultimately the protein corona may play a role in the cellular uptake. Monteiro-Riviere et al. recently showed that pre-incubation of citrate coated Ag nanoparticles with different proteins (albumin, IgG, transferrin) reduced the cellular uptake for both 20 nm and 110 nm particles. Yet, the similar behavior of the different sized nanoparticles used in this study together with the low protein content in the working cell medium, suggest that the protein corona is unlikely to explain the observed differences in toxicity. Differences in nanoparticle agglomeration affect sedimentation and may ultimately result in changes in the exposure doses and uptake rates . However, the uptake of the 10 nm citrate and 10 nm PVP coated AgNPs was similar (2.9 pg/cell and 2.1 pg/cell) and in the same range as the 75 nm citrate coated AgNPs (3.2 pg/cell). Next we explored the uptake mechanisms for the 10 and 75 nm citrate coated AgNPs and found that both particles were internalized by active mechanisms as shown by the negligible uptake at 4°C. A combination of different active pathways was involved for both particles as previously shown for AgNPs  as well as other nanomaterials e.g. quantum dots . Thus, while we acknowledge the importance of agglomeration for particle stability, and the fact that this, as well as the protein corona can affect cellular uptake, metal release and toxicity, it appears not to play a major role in the toxicity observed for the 10 nm citrate and 10 nm PVP coated particles.
The main difference between the AgNPs in our study was the released amount of Ag in cell medium, which was significantly higher for the 10 nm AgNPs. One explanation for this is obviously the increased surface area and increased particle number (196 fold increase for 10 nm vs. 40 nm particles and 1140 fold increase for 10 nm vs. 75 nm particles) for the same mass/volume dose. This is in line with previous reports showing that the release of Ag is directly related to the total surface of the particles as well as the composition of the experimental media . Ag release has previously been reported to increase with smaller particle size in a non-linear manner , thus explaining the much higher release from the 10 nm particles when compared to the other sizes. To further explore the role of the released Ag, we also investigated the toxicity of the “released fraction” (i.e. the supernatant of centrifuged particles after incubation in cell medium for 24 h thus likely containing various Ag-complexes). However, no effect on cell viability was observed after incubating BEAS-2B cells with this fraction and we therefore concluded that the extracellular release and presence of ionic species in cell medium could not account for the observed differences in toxicity. We thus posit that the size dependent toxicity relates to the intracellular release of Ag ions. When we attempted to mimic one intracellular compartment, the lysosome, by using artificial lysosomal fluid (ALF), very little release was observed (Additional file 6: Figure S6). This is explained by the severe agglomeration that takes place in this solution due to the very high ionic strength  since low pH (in low ionic strength solutions) is known to cause higher Ag release . In addition, ALF does not contain any proteins that can serve to stabilize the particles and we conclude that mimicking various intracellular compartments is challenging. Previous studies have shown that Ag ions interfere with cellular functions by interacting with the thiol and amino groups of biomolecules , thus providing an explanation for the toxicity. Ag release has also been reported to govern the toxicity of AgNPs towards bacteria, where the particles act as a vehicle for Ag delivery . In the same study the antibacterial effect was hindered under anaerobic conditions (that prevent oxidation and Ag release) . Moreover, AgNPs with higher Ag release were shown to be more toxic in Caenorhabditis elegans. In all, this suggests that AgNPs may change the transport rate of Ag ions into cells and organisms and that subsequently released Ag ions exert the detrimental effects.