Introduction

Nanotechnology refers to a wide range of technologies that measure, manipulate, or incorporate materials and/or features with at least one dimension between approximately 1 and 100 nm. At this size, materials may exhibit unique properties when compared with larger or bulk forms of the same material, and may have been engineered precisely for that reason. These unique properties can include altered color, magnetism or optical properties, increased strength, flexibility, or reactivity, and improved electrical conductivity or absorption. Such properties may arise owing to the smaller size, unique shape or arrangement, or the increased surface area of the material. As a result, engineered nanomaterials have become the focus of extensive research in diverse areas, including electronics, materials engineering, energy production and conservation, and biomedical applications. However, the same properties that make nanomaterials desirable in these various applications have the potential to alter the biological properties that impact the environment, health, and safety of these materials. Such properties could include altered environmental fate, hazard potential, or exposure potential when compared with the larger or bulk forms of the same material. Therefore, in addition to the extensive research on the commercial applications of engineered nanomaterials, there has been a focus on addressing their potential environmental and human health hazards and risks. This research may complement ongoing research to evaluate the hazards of naturally occurring small particles such as those from mining operations and anthropogenic road dust, but the discussion here is focused solely on engineered nanomaterials.

With the recognition that nanotechnology is undergoing rapid growth and research, questions about safe use of nanomaterials have been raised by consumer groups, regulators, and nongovernmental organizations. This situation has created a specific research interest in the area of toxicology that has been referred to as nanotoxicology. Until recently, few laboratories had experience in testing for the hazards of engineered nanomaterials. Others are now just beginning to assess what skills are necessary to test nanomaterials, whereas others have applied traditional toxicology test systems to the study of nanomaterials, often with little understanding of the complexities associated with the behavior of water-insoluble crystals in an aqueous biological environment. We have recognized that this last approach is not appropriate because it does not account for the physical properties of the particle that could alter how a study is conducted. Furthermore, insoluble nanoparticles can behave differently in different media, so what may appear to be a particle of one diameter in an aqueous environment of distilled water may be quite different in a biological medium of ionic buffers and proteins [1, 2] (Fig. 1). Given the knowledge that engineered nanomaterials may possess new properties when compared with the bulk form of the same material and that these properties may influence the hazard and risk potential of the material, proper characterization of these properties should be an integral part of the safety evaluation process. However, many of the early toxicology data on nanomaterials were generated without consideration of how nanomaterials behave in biological systems and with a lack of characterization of the particles, either outside or within the test system, which has prevented the nanotoxicology community from identifying and assigning basic parameters as most relevant for biological activity. Without the consideration of proper study design and end points, toxicology studies on nanomaterials will be prone to anecdotal findings which may differ among laboratories and slow progress. Indeed, international efforts are only now beginning to address these issues. That does not impact the present flood of studies being reported on the effects of nanomaterials on a biological system. In this article, we will highlight some of the reasons why good material characterization in nanotoxicology studies is important; why test systems for chemicals cannot simply be applied to nanoparticles without some understanding of how water-insoluble particles behave in these systems, including confounding issues due to vehicle effects, loss of nutrients, loss of assay dye, change in particle size and altered nanoscale phenomenon; and will provide some understanding of the current state of the science. It is hoped that this background will promote a dialogue between the physical chemist/material scientist and the toxicologist/biologist before they engage in a toxicology study for the health assessment of a nanomaterial.

Fig. 1
figure 1

Insoluble nanoparticles may not behave as individual particles when introduced into test media. Factors such as pH, ions, amino acids, and proteins can have a significant impact on the dispersion, aggregation, agglomeration, or the surface properties of nanomaterials. Adsorption of these factors onto the surface of the nanoparticles can also change the composition of the medium, thereby influencing downstream assay results. H+ hydronium ion, Tyr tyrosine amino acid, Alb albumin protein, Met methionine amino acid, PO 4 − phosphate ion, A single dispersed nanoparticles, B aggregated nanoparticles, C nanoparticles agglomerated into micron-sized particles, D nanoparticles with a corona

Full size image

Carrying chemical testing strategies to the world of nanoparticles

Dose preparation and delivery

With the emergence of nanotechnology into the public arena came questions about how to evaluate the hazards and if the battery of test designs used for chemicals (i.e., molecules) had to be completely revised to fully assess the hazards of nanomaterials [3, 4]. Up to that point, information on nanomaterials had come from two primary sources: from academic laboratories with focused but limited study designs for biological responses to nanomaterials, and from industrial laboratories using test methods associated with testing of chemicals. The early results from these two sources suggested that some unique effects were observed following exposure to nanomaterials, which led some to questions about whether nanoparticles require a unique testing strategy. That debate continues, but most have recognized that other than adjusting the exposures and expression of dose to take into account the characteristics of nanomaterials, the toxicology hazard end points to be monitored and measured are the same as for soluble chemicals [5, 6].

Certainly, it is true that certain aspects of existing testing strategies for chemicals do not always apply to nanoparticles for several obvious reasons. First, nanoparticles consist of more than one molecule and dissolution of nanoparticles in aqueous media is expected to be limited. Although nanomaterials are chemicals, their crystalline structure, water insolubility, and slow dissolution means that molecules do not as easily reach the biological systems as they do for soluble chemicals. Furthermore, in some organs such as the respiratory tract, persistence (defined here as the lack of elimination from the organ by biological processes) impacts the biological response. Slow dissolution or increased persistence affect interaction with the test system: for pulmonary effects, the slower the dissolution rate, the potentially more persistent the particle and the more likely the fibrogenic or even carcinogenic response [7], which is why dissolution rates have been measured for the hazard assessment of man-made fibrous substances. With systemic toxicity, the slower the dissolution rate, the less likely the particle will generate a toxic response [8] because individual molecules are released slowly. For example, Xia et al. [8] studied the effects of TiO2 and ZnO on the cytotoxicity to different lines of pulmonary cells (macrophages and bronchial epithelium). Their results show that although TiO2 was incorporated into the cell, it did not elicit a cytotoxic response (or reactive oxygen species formation), whereas cytotoxicity was observed with ZnO. Furthermore, the cytotoxic response was directly related to zinc ions released from the ZnO into the medium and cell. Similar results have been shown with cadmium release from CdSe quantum dots [9]. Other researchers have shown that the length of carbon nanotubes is an important determinant of their potential to induce inflammation and the formation of granulomas [10]. These data suggest that it is not only the dose and composition of the material that can influence toxicity, but that the shape and dimensions also play a role. Concepts such as these are not new to the field of toxicology, but are established concepts in the field of particle and fiber toxicology, where the three D’s—dose, durability, and dimension—are well known to have a significant influence on the hazard potential of particles and fibers [11].

Therefore, a testing strategy for inhalation exposure to a soluble chemical might easily be adopted for a nanomaterial by leveraging knowledge from the field of particle and fiber toxicology. However, another difficult issue to address involves the generation and measurement of the aerosol or test atmosphere. Generation of an aerosol of nanoparticles was initially thought to be a major technical problem. Generators used for bulk dusts are clearly inappropriate because they do not produce individual nanoparticles; rather they generate agglomerates of these materials. Some have had success with dispersing the nanoparticle in an aqueous suspension and nebulizing it [12]. This does produce a nanoparticle, but it also produces larger agglomerates. For carbon nanotubes, brush-feed and acoustic generators have been used with good success [12, 13]

The issue of particle agglomeration is an important issue for inhalation studies because the size of the inhaled particulate, whether an individual particle or an agglomerate, will determine where the particle will be deposited in the respiratory tract, which will be a key determinant of its potential for hazard [14]. For example, larger particles may be trapped in the upper respiratory tract, whereas small particles can penetrate deep into the more sensitive regions of the lung. On the other hand, if the purpose of the investigation is to evaluate the potential hazards in the workplace, where exposures are primarily to agglomerated particles, then a wide distribution of particle size may be appropriate. Therefore, the ability to properly generate a nanoparticulate exposure atmosphere as well as consideration of the likely exposure form of the particulate (individual or agglomerated) will be important considerations for drawing conclusions about the hazard and risk potential for nanomaterials.

Characterization of particles in the air, as it is in any medium, is also important. Using more than one method of measurement has value, even for parameters such as particle number, because instruments vary in their sensitivity [12]. As with any inhalation toxicity study, the mass, median size, and size distribution are important parameters. Particle number can also be a useful measurand especially because mass is so insensitive for small particles. Surface area and other physical characteristics may be more difficult to measure during exposure as continuous measurements. The International Standards Organization (ISO) has proposals for what parameters are appropriate for measuring nanoparticles, and these are being considered by the scientific experts engaged in ISO TC229.

For some, the issues of aerosol generation are difficult to overcome and the linking of body burden to exposure is a challenge, so alternative methods have been used to deliver the nanomaterial to the animal. These approaches often involve direct delivery of the material to the lungs (e.g., intratracheal instillation, intranasal aspiration, or oropharyngeal aspiration). Although these methods deliver known lung burdens, they artificially bypass the normal processes for distribution in the respiratory tract. Additionally, the substance is delivered in a bolus, which can produce artifactual effects [15]. Such results emphasize the challenge of dispersing the particle in a medium that is compatible with the respiratory tract while maintaining the particle characteristics that are most appropriate for assessing its toxicity [16]. Once the particles has been administered to the lung, assessing the effects on exposure to the respiratory tract can include a variety of techniques, the most commonly used being bronchoalveolar lavage and histological examination; no nano-specific end points need to be used or have been identified [17].

For assessment of systemic toxicity, simply adding the nanomaterials to the test system is not appropriate, particularly if the delivery vehicle alters the agglomeration state of the particle. To many toxicologists, this problem is easily solved by mechanical means, such as sonication, prior to adding the test material to the test system. Sonication of test material solutions has been a standard practice for insoluble chemicals and is thought to work reasonably well—at least as best as can be detected by the naked eye. However, the approach may not be appropriate for nanomaterials, especially when conclusions will be drawn on the size-specific toxic effects. Glory et al. [18] showed that sonication of carbon nanotubes breaks the original structure into much smaller units; in other words, the carbon nanotubes that were initially purchased and characterized were very different after sonication. Therefore, many of the studies already published that utilized this standard technique to mechanically deagglomerate substances to primary particles may have reported results on something completely different unless the particle was characterized after sonication. In another study, sonication of TiO2 in dimethyl sulfoxide was found to result in the formation of unanticipated breakdown products of dimethyl sulfoxide which influenced the test results, once again highlighting that care needs to be taken when preparing nanomaterial testing solutions [19]. The key aspect that researchers are trying to address with each of these approaches is the hazard potential of the nanomaterial; however, unless we understand what that end size really is, how can we relate it to the biological response?

Considerations for the correct dose metric for nanomaterials

A central tenet of toxicology is the concept that it is the dose that makes the poison. In this regard, toxicologists have traditionally assessed dose on a mass basis (e.g., milligrams per kilogram body weight). However, a number of studies have suggested that this may not be the most appropriate dose metric for the assessment of particulates, including nanomaterials. For example, researchers have shown that ultrafine (20-nm) TiO2 induced a greater acute pulmonary inflammatory response in mice and rats than did fine (250-nm) TiO2 when given at the same mass dose [14]. However, when the doses of the particulates were expressed in terms of their surface area, a similar inflammatory response was observed at equal surface area doses. Similar effects were observed in an independent laboratory that also examined TiO2 [20] and another laboratory that examined six differently sized carbon particles [21]. Observations such as these have led researchers to conclude that the surface area of a particulate sample may be the most appropriate dose metric for hazard assessment of particulate materials, including nanomaterials. However, an independent examination of these data has suggested that surface area may not be the most appropriate dose metric; rather it is particle number which most accurately predicts the pulmonary inflammatory potential [22]. Still other studies using TiO2 and quartz particulates have shown that nanoscale materials may not be more inflammatory than larger particles and that surface reactivity (as measured by electron spin resonance or hemolytic potential), not surface area or particle number, is a more accurate predictor of the pulmonary inflammation outcome [23, 24].

Despite the lack of resolution on the appropriate dose metric [25–27], these data highlight the importance of characterizing nanomaterial parameters such as particle size, surface area, surface reactivity, and particle number, in addition to mass, such that retrospective analyses of multiple dose metrics are possible. Such analyses may provide important insights into the dose metric that most accurately predicts the biological response, which is likely to vary depending on the nanomaterial type, the synthesis method, and surface coating, to name a few factors. These data may ultimately shed light on the potential toxic mechanisms of action for nanomaterials, which in turn will allow for a more complete and accurate assessment of their true risk.

Dispersion of nanomaterial in the test system—what is the size seen by the target?

The recurring issue that exists for the hazard assessment of nanomaterials is that unless we understand what that end size really is, we cannot draw definitive conclusions on the size-dependent properties that may be driving the biological response. Murdock et al. [1] showed that particle size changed dramatically in biological media compared with distilled water. Using a variety of particles, including elemental silver, copper, and aluminum, metal oxides of titanium, silica, and aluminum, and carbon nanotubes, these researchers reported that primary particles of size less than 100 nm agglomerated to particles nearly 3–6 times greater depending on the ionic strength of the solvent. Interestingly, the addition of serum acted as a dispersing agent and reduced the agglomerates to a size comparable to that measured in distilled water. Other researchers have shown similar effects when dispersing nanomaterials in water versus phosphate-buffed saline, with the results indicating particle size distributions that varied by an order of magnitude (100 nm versus 2,000 nm) [28]. The importance of size in the biological medium has not been fully recognized by many investigators, but data from Murdock et al. [1] and Ahamed et al. [29] clearly show differences in biological response in cell-based in vitro test systems. Murdock et al. [1] showed that the cytotoxicity of silver and copper particles increased if the particles were suspended in a medium with serum compared with a medium without serum. Ahamed et al. [29] reported similar results for coated and uncoated silver particles, demonstrating that agglomerated uncoated silver appeared to be excluded from organelles such as the nucleus, whereas coated silver, which was not agglomerated, was widely distributed. Guo et al. [2] found that amino acids in the nutrient medium adsorbed onto carbon nanotubes. Such results point to the need to evaluate the particle that comes in contact with the target organ or cell. However, we cannot assume that simply dispersing the particle will automatically result in increased toxicity because Murdock et al. [1] also reported no differences in toxicity for aluminum, aluminum oxide, or titanium dioxide despite differing particle sizes.

Much of the research and discussion on the medium-dependent size effects of the nanomaterial have been focused on their influence for in vitro studies; however, these same principles also likely to dictate in vivo responses. The complicating factor is that the in vivo test system is more complex than the in vitro one as the ionic and protein content is self-regulated and changes in different organ, tissue, and cellular compartments. Just as serum can disperse agglomerated nanoparticles in vitro, serum protein should be able to do the same in vivo. Albumin coating enhances cellular uptake of gold nanorods in vitro [30] so why should this not occur in vivo? There are currently no experimental data to confirm protein-coated state or influence of this on uptake in vivo.

Certainly, the size of the nanoparticle has an impact on absorption in vivo. Jani et al. [31,32] demonstrated that smaller size improved absorption; absorption of 50-nm polystyrene beads was more efficient compared with that of 500-nm polystyrene beads [31,32]. Furthermore, Florence et al. [33] have shown that 59-nm uncoated polystyrene beads were absorbed from the intestinal tract but surfactant-coated beads were not absorbed; and carboxylate-coated polystyrene beads were not absorbed, but nonionic beads were absorbed. These data suggest that the coating could have significant effects on particle absorption by altering the surface environment; however, confirmation of the actual mechanism involved will require additional experimental data combined with good particle characterization.

The use of in vitro approaches for the assessment of nanomaterial hazard

In an effort to reduce, refine, and replace the use of animal testing in toxicology, in vitro cell-based assay systems have become important tools in the field of toxicology. A number of the considerations mentioned above with respect to understanding the particle shape, size, and agglomeration in the test system are especially important for in vitro systems because these properties are determined in culture medium. The formulation of the culture medium with respect to serum concentration, pH, and other factors can vary from laboratory to laboratory and therefore can influence the response in the cellular system, as described above. However, there are a number of additional considerations for the use of in vitro systems that further highlight the need for careful conduct when examining the hazard potential of nanomaterials. For example, researchers have shown that classic dye-based in vitro cytotoxicity assays such as 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide and neutral red produce invalid results with some nanomaterials owing to interactions between the dye and the nanomaterial, such as adsorption of the indicator dye by the nanomaterial [34]. In other instances, the assessment of cellular responses can be difficult owing to direct physical interference of the insoluble particles with the ability to properly visualize the cells [34]. Other researchers have shown that the results of in vitro studies with nanoparticles need to be carefully interpreted as diverse particles have been shown to cause similar oxidative stress responses in vitro but to exhibit dramatically different pathological effects in vivo [35]. Therefore, scientists have highlighted the need for better information on in vitro dosimetry and toxicokinetics as a vital component to increase the utility of these in vitro assays for hazard characterization [35].

Summary

In summary, we should recognize that standard study designs for airborne particles should be applicable to nanoparticles, and end points currently used to assess the biological responses to other particles can be used for nanoparticles. In addition, proper attention should be given to particle characterization to facilitate a comprehensive assessment of the appropriate dose metric. For systemic toxicity, characterizing the particle in the dosing solution is important, as is understanding how the physiological environment in the body may alter the particle size and surface. This also applies to in vitro studies, which should characterize the particle that comes into contact with the test system, while also considering any assay interferences the particles may introduce.

Status of characterization in studies

Why is it an issue?

Uncertainty makes for anxiety. To many people, the uncertainty about the effects of nanomaterials on humans and the environment are a reason to limit commercialization of these materials. Some of the uncertainty comes from the anecdotal evidence of hazard and risk from the literature. Until overarching strategies were developed to assess the hazards of nanomaterials using defined protocols, investigators took liberal approaches to studying toxicity. That, coupled with the lack of appropriate characterization, has led to some information that appears conflicting, which makes our search for rules governing structure–activity relationships difficult and adds to the uncertainty. For example, Renwick et al. [36] described a study in two different sizes of either carbon black or TiO2 were administered directly into the lungs of rats and the inflammatory responses were measured. Their results showed that whereas the bulk material may be “nontoxic,” the nano-sized material was “toxic.” These data supported the concept of “smaller size equals greater hazard.” However, Barlow et al. [37] reported only minor differences in the response of L-2 type II cells in culture exposed to nano-sized carbon black or TiO2 compared with larger sized particles, in apparent conflict with Renwick et al. [36]. However, Barlow et al. [37] did not describe the characterization of their materials, and there was no characterization of the materials in the culture medium. How the particles behaved in the medium could have influenced the results. Furthermore, without characterization, important parameters that may be contributing to the discordant results cannot be ascertained, which does not allow for the scientific community to build a greater understanding of the mechanisms involved with nanomaterial toxicity.

When characterization is conducted well, it becomes apparent how the biological phenomena that we observe are related to particle characteristics. For example, Stoeger et al. [21] researched a wide range of ultrafine carbon particles and suggested that there is a threshold for particle surface area at an instilled dose of 20 cm2, below which there is no pulmonary inflammatory response in mice. In that study, the lungs of mice were instilled with different amounts of carbon black from a variety of sources and the inflammatory response was evaluated using bronchoalveolar lavage. The pulmonary inflammatory response, relative to the amount of material dosed, varied greatly, but only when the response was correlated with the Brunauer–Emmett–Teller surface area did all the substances align along a dose–response curve. The data indicate that although size may be an important factor, it alone does not convey a greater potential for hazard, and is only one of the many components of nanomaterials that may influence their interaction with biological systems.

Clearly, many aspects for understanding nanomaterial hazard may be related to how dose is defined, as was described above. Brunner et al. [38] showed that characterization was essential to understand the dose metric and the dose response in simple cytotoxicity assays, whereas others who did not characterize the nanomaterial in the medium mistook the effect of the agglomerated particle to represent the primary particle. However, the issue of dose metric is only one part of the equation. How the material is manufactured, including the subtleties of coating, is also important. For example, TiO2 is a well-studied nanomaterial and separate studies have been conducted with materials from Degussa (Evonik), DuPont, and BASF. Although these are all the same substance, they are not all the same particle if one considers surface coating. It is quite clear that such differences may influence the downstream toxicological responses and therefore should be characterized as part of the hazard assessment for these materials. Failure to characterize these parameters could once again lead to conflicting results among different laboratories, thereby slowing progress. Support for the influence of surface coating is provided by research by Sayes and Warheit [39], who reported different levels of cellular damage as measured by LDH release using alumina-coated and alumina–silica-coated TiO2.

Another example of potential conflicting data is from Barnes et al. [40], who evaluated the genotoxicity of well-characterized SiO2 in the 3T3-L1 fibroblast cell line and found no DNA damage on the basis of the results of the Comet assay. Barnes et al. [40] discussed their results being in contrast to the reported DNA damaging effects observed in MCF-7 cells, a breast cancer cell line. They attributed the difference in response to differences between the cell lines that were used; however, without proper characterization one cannot exclude differences in purity, size, shape, or other particle parameters. Such a perspective is supported by a recent review by Singh et al. [41], who listed the results of a number of genotoxicity and cytotoxicity studies for nanomaterials. Their assessment indicated that frequently results of studies on the same material conflict with one another, and that the conflicting results are associated with a lack of characterization, which is necessary for a proper interpretation and comparison of the results.

It is clear that there are many characteristics of nanomaterials that have the potential to influence their toxicity. Without proper material characterization conflicting results from independent toxicology studies, which are conducted on what is thought to be the same material, are likely to emerge. These issues further highlight the need for adequate material characterization as part of any assessment of nanomaterial hazard.

The path forward: the need for consistent characterization

With such uncertainty about the safety of nanomaterials, regulators have expressed the need to develop “rules of thumb” for structure–activity relationships (or more appropriately, property–activity relationships) to predict hazards without requiring excessive animal testing. Although everyone involved in the field of nanotoxicology recognizes this need, rules for structure–activity relationships can only be established once the scientific community can effectively compare the physical–chemical characteristics against the biological responses across a range of nanomaterials. One fact is certain: size does matter, but it is not the only factor that may be influencing the biological responses to nanomaterials. What is required is a comprehensive characterization of multiple parameters so that we can develop a better understanding of the properties that are important determinants of hazard potential.

Owing to the inherent need for material characterization and the apparent lack of characterization in many research publications, several publications as well as national and international symposia and work groups convened in the past 2–3 years have promoted good characterization, and have listed what parameters (some call them measurands) must be determined and reported by investigators conducting toxicology studies on nanomaterials [4,42,43]. Since the key characteristics that may influence the toxicity of nanomaterials are not yet fully understood, the characterization of test materials must initially be broad in scope, which has led to intimidating long lists of parameters that many investigators have found overwhelming. As a result, recent proposals have become more limited in their recommendations to help gain support and compliance with a minimal or base set of parameters that are thought to be important. One such effort is the MINChar initiative (http://www.characterizationmatters.org), which represents a grassroots effort aimed at improving the quality of nanotoxicology studies [44]. The initiative advocates for a minimum base set of characterization parameters. This simplified approach with basic questions researchers should be asking and the corresponding parameters that should be addressed at each step of the evaluation are listed in Table 1. It is important to note that this is a recommended minimum set of parameters, and is not intended to replace more robust guidelines from governments and organizations such as the ISO and the Organization for Economic Cooperation and Development (OECD). Furthermore, highlighting only a key set of parameters without specifying prescriptive approaches was intentional and done as a means to foster a culture of effective characterization and allow for flexibility and concordance with more authoritative guidelines. Inclusion of these characterization end points as a part of any toxicological evaluation of nanomaterials will facilitate a more accurate interpretation of the results in addition to enabling researchers to reproduce data or explain discordant findings. More importantly, the consistent interpretation of the toxicology responses relative to the characteristics of the nanomaterials will help to develop a greater understanding of the exact properties of these materials that result in toxicity or the absence of toxicity. This will help to facilitate the consistent, responsible, and sustainable development of nanotechnology, including the development of effective regulatory policies that ensure the safety of humans and the environment while still effectively enabling the promises of nanomaterials and nanotechnology to be realized.

Table 1 Recommended minimum physical and chemical parameters for characterizing nanomaterials in toxicology studies
Full size table

With the need for characterization highlighted and a minimal list of parameters presented, the remaining need is for the application of these principles. This will require an effective collaboration between physical chemists/material scientists and toxicologists/biologists before they engage in a toxicology study for the assessment of a nanomaterial. We hope that this article, and those that follow in this special issue, will help to increase awareness and promote these collaborations.