Tian Xia, 1, 2, Huan Meng, 1,2, Saji George, 1,2, Haiyuan Zhang, 1,2, Xiang Wang, 1,2, Zhaoxia Ji, 2, Jeffrey I. Zink, 2,3, Andre E. Nel, 1,2
1Division of NanoMedicine, Department of Medicine, University of California, 2California NanoSystems Institute, University of California, 3Department of Chemistry & Biochemistry, University of California
A key challenge for nanomaterial safety assessment is the ability to handle the large number of newly engineered nanomaterials (ENMs), including developing cost-effective methods that can be used for hazard screening.1 In order to develop an appropriate screening platform, it is necessary to assemble and synthesize nanomaterial libraries that can be used to screen for specific material compositions and properties that may lead to the generation of a biological hazard.1-3 Our opinion is that screening using rapid assessment platforms should initially take place at the biomolecular and cellular level to generate a comprehensive database of potentially hazardous events at the nanobio interface, and then, using these property-activity relationships, to prioritize animal studies that can validate the real-life significance of the in vitro observations.4 Both the National Toxicology Program as well as the National Research Council (NRC) in the US National Academy of Sciences (NAS) have recommended that toxicological testing in the 21st-century evolve from a predominantly descriptive science in animal models to a predictive scientific discipline premised on target-specific, mechanism-based biological screening.5,6 It was further recommended that biological testing should be based on robust scientific paradigms that can be used to concurrently screen multiple toxicants, rather than costly animal experiments that examine one toxicant at a time.4,7
We refer to the above approach as a predictive toxicological paradigm, which can be defined as assessing the in vivo toxic potential of a material or substance based on in vitro and in silico methods.4 There are four major requirements to keep in mind when establishing this paradigm. The first is to acquire or synthesize compositional and combinatorial ENM libraries that can be used for knowledge generation of the material properties that may lead to biological injury. The second requirement is to develop in vitro cellular screening assays that utilize mechanisms and pathways of injury. Third, is to develop high content or rapid throughput screening platforms that are capable of assessing the large number of material compositions and properties, and are based on the pathways of injury. Finally, the in vitro data should be used for in silico modeling to establish quantitative structure-activity relationships (QSARs) and to generate a hazard ranking that can be used to prioritize animal experiments.
The acquisition and characterization of standard reference nanomaterial (SRM) libraries forms the basic infrastructure requirement that is necessary to screen for toxicity and to elucidate the material properties that are most likely to result in biological injury.8 The selection of the library materials should take into consideration the commercial production volumes of the different nanomaterials, incorporating the major current ENM classes of materials (metals, metal oxides, silica, and carbon based nanomaterials). The choice of materials should also consider the exposure potential, route of exposure and delivery pathway. For example, free nanoparticles or powders are more likely to become airborne with the potential to generate pulmonary toxicity after inhalation. Thus, it is appropriate to investigate this scenario using lung cells (in vitro) and pulmonary exposure (in vivo) that ideally should be linked in terms of mechanisms of potential injury. Ideal ENM libraries should also include positive and negative control ENMs to provide a reference point for the evaluation of material toxicity.
Figure 1. Building of compositional and combinatorial ENM libraries. This process involves selection of well-characterized standard reference materials with similar size and surface area, and obtained either by in-house synthesis or from commercial sources. The libraries can be grouped into metals, metal oxides, carbon or silica-based nanomaterials. The materials in each library are subjected to high throughput screening in various screening systems such as mammalian cells, bacteria, yeast, and zebrafish. The identified toxic materials are used to develop combinatorial libraries that contain variations of the potential toxic properties of the ENM. The information obtained from the ENM screening can be subsequently used to build quantitative structure-activity relationship (QSAR) models. Selected ENMs will be used in in vivo assays to validate the results obtained from in vitro screening.
To establish the link from specific physicochemical properties of an ENM to its toxicity, it is necessary to establish combinatorial libraries, synthesized to vary or alter major physicochemical properties that may be involved in toxicity. Property variations may include nanoparticle size, surface area, shape, crystallinity, bandgap, porosity, solubility, charge, and surface functionalization (Figure 2). Since we identified the importance of dissolution in ZnO-induced toxicity, we hypothesized that modifications of this material could alter the dissolution rate and modify its toxicity.
Figure 2.Examples of combinatorial ENM libraries. Combinatorial libraries are built by synthesizing one of the compositional materials to vary one of their major physicochemical properties that may be involved in toxicity. Property variations may include nanoparticle size, shape, porosity, hydrophilicity/hydrophobicity, crystallinity, bandgap, photoactivation, solubility, charge, and surface area. A single property variation may also change other properties, and rigorous re-characterization is required.
One way to achieve this is to introduce another element, iron, into ZnO during the nanoparticle synthesis. Through careful synthesis we obtained an Fe-doped ZnO combinatorial library that included a series of nanoparticles with incremental percentages of Fe. Characterization of this nanoparticle library showed that increasing percentages of Fedoping decreased the ZnO dissolution rate in aqueous solutions without changing the crystal structure of the ZnO.14 We then tested the toxicity of these nanoparticles in vitro and found that the cytotoxicity decreased as the percentage of Fe doping increased, indicating that dissolution rate indeed plays an important role in cytotoxicity. Currently, we are performing in vivo tests using the Fe-doped ZnO library and preliminary data shows that the in vitro findings extend to multiple animal models including zebrafish, mouse and rat. These results show that building compositional ENM libraries can quickly identify potentially toxic nanomaterials and that modifying nanomaterial properties to build combinatorial ENM libraries can help link specific physicochemical properties to toxicological outcomes.
Much of the knowledge about ENM cellular toxicity has been generated using fairly straightforward cellular viability assays such as the lactate dehydrogenase (LDH) and the colorimetric MTT/MTS assays or propidium iodide (PI) staining. The major drawback is that these assays are often not informative of a specific toxicological pathway because multiple stimuli can result in the same assay outcome, establishing little connectivity between the biological outcome and specific ENM properties. Moreover, cellular viability assays also do not reflect sublethal toxicity effects. For these reasons we advocate developing mechanism-based in vitro assays because this is conceptually the easiest way to link in vitro toxicity screening with pathological effects in vivo. Currently, there are approximately ten major mechanistic pathways of toxicity that have been linked to ENMs (Table 1). These include injury paradigms such as the generation of reactive oxygen species and oxidative stress, frustrated phagocytosis (e.g., in mesothelial surfaces), changes in protein structure and function (e.g., loss of enzymatic activity), protein unfolding response, immune response activation (e.g., through exposure of cryptic epitopes or immunostimulatory effects), fibrogenesis and tissue remodeling, blood clotting, vascular injury, neurotoxicity (e.g., oxidative stress, protein fibrillation), and genotoxicity.
It is important to note that additional mechanisms of toxicity are possible given the wide range of novel ENM physicochemical properties. For the purposes of this article, we will focus on the generation of oxidative stress.
Particle-induced oxidative stress invokes three tiers of cellular responses including cellular antioxidant defense, activation of pro-inflammatory signaling pathways leading to the production of cytokines/chemokines, and mitochondria-mediated cell death.3,9,14, However, performing the entire panel of tests necessary to study the three tiers of oxidative stress requires at least 2-3 weeks of labor-intensive effort. A rapid throughput screening approach could offer several advantages over conventional assays. First, this approach speeds up the pace of knowledge generation that is possible with compositional and combinatorial ENM libraries. High throughput screening (HTS) provides a rapid readout because of the standardization of the procedure, automation (e.g., cell seeding, liquid handling, imaging, image analysis), and miniaturization (requiring smaller amounts of reagents and lowering the cost per assay). Not only is HTS capable of screening large libraries, but it can also accommodate multiple cell lines, time points and doses of exposure within the same experiment. Coupled with good bioinformatics and decision making tools, this approach can significantly improve the reliability of toxicological screening as well as establishment of property-activity relationships. To develop rapid throughput platforms based on mechanisms of toxicity, it is advantageous to combine different steps or nodal points in the injury pathway. Such multi-parametric screening efforts enhance the utility of the procedure, cover lethal and sublethal cellular responses and improve the predictive value of the assay. As an example of such an assay, we recently developed a multi-parametric screening procedure that incorporates several of the cellular oxidative stress responses involved in the advanced tier of oxidative stress (Figure 3).14
Figure 3. Schematic illustrating the relationships of cellular responses induced by metal oxide nanomaterials. We established multi-parametric HTS assays based on nanoparticleinduced Reactive Oxygen Species (ROS) production and oxidative stress. Nanomaterials induce ROS production as a direct consequence of specific material properties or as a consequence of triggering cellular injury responses leading to generation of the oxidant radicals. ROS production could trigger a range of oxidative stress effects. The induction of cellular toxicity at the highest level of oxidative stress involves a number of interrelated cellular responses, including intracellular Ca2+ release and perturbation of the mitochondrial membrane potential (MMP) that prestages cell death and the accompanying changes in cell membrane integrity and nuclear propidium iodide (PI) uptake. The parameters chosen in multi-parametric HTS assays are highlighted in yellow.
In vivo screening is time consuming and expensive. A complete set of toxicological assays for a single chemical, including assessment of carcinogenicity, chronic, reproduction and developmental effects could involve hundreds of animals and costs in the range of $1–3 million per test. As a result, less than 2% of industrial chemicals have undergone toxicity testing in rodents. We believe that by using a predictive toxicology approach it is possible to avoid a similar conundrum in nano-safety testing. Using the mechanism-based in vitro HTS screening, we should be able to identify the major mechanisms of toxicity and perform hazard ranking, which can then be used to prioritize in vivo testing. This approach also allows us to obtain data on the dose and kinetics related to the nanomaterial physicochemical properties, and to quantify biological response and exposure outcomes. The in vivo results are important to validate the in vitro screening as “predictive,” thereby allowing the in vitro platform to be used as the primary screening procedure. In vitro HTS data will also be used to establish in vitro nano-QSARs by in silico modeling that will use statistics, mathematics and machine learning to perform hazard ranking as a prelude to future risk prediction studies.
We propose the implementation of predictive toxicology for toxicity screening of nanomaterials. This exercise is based on the establishment of compositional and combinatorial libraries, the development of mechanism-based in vitro toxicity screening assays, the development of multi-parametric high throughput screening assays, the building of computerized QSAR models, and the prioritization of in vivo assays to validate the predictability of in vitro assays. We think this is an appropriate approach to building a knowledge base that can meet the challenges of an expanding nanotechnology enterprise.
This work was funded by the U.S. Public Health Service Grants, RO1 CA133697, RO1 ES016746 and RC2 ES018766. Support was also provided by the National Science Foundation and the Environmental Protection Agency under Cooperative Agreement Number EF 0830117. The work was also supported by the NSF USDOD HDTRA 1-08-1-0041 grants. Any opinions, findings, conclusions or recommendations expressed herein are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or the Environmental Protection Agency.