Table 1

Figure 1. Idealized size distribution of traffic-related particulate matter (U.S. EPA 2004). Dp, particle diameter. The four polydisperse modes of traffic-related ambient particulate matter span approximately four orders of magnitude from < 1 nm to > 10 µm. Nucleation- and Aitken-mode particles are defined as UFPs (< approximately 100 nm). Source-dependent chemical composition is not well controlled and varies considerably. In contrast, NPs (1-100 nm) have well-controlled chemistry and are generally monodispersed.

Table 2

Figure 2. Surface molecules as a function of particle size. Surface molecules increase exponentially when particle size decreases < 100 nm, reflecting the importance of surface area for increased chemical and biologic activity of NSPs. The increased biologic activity can be positive and desirable (e.g., antioxidant activity, carrier capacity for therapeutics, penetration of cellular barriers), negative and undesirable (e.g., toxicity, induction of oxidative stress or of cellular dysfunction), or a mix of both. Figure courtesy of H. Fissan (personal communication).

Biologically based or naturally occurring molecules that are found inside organisms since the beginning of life can serve as model nanosized materials. For example, biogenic magnetite is a naturally occurring NSP that occurs in many species ranging from bacteria to protozoa to animals (Blakemore 1975; Kirschvink et al. 2001). Biogenic magnetite has even been found in brains of humans (Dunn et al. 1995; Kirschvink et al. 1992; Schultheiss-Grassi et al. 1999) and has been associated with neurodegenerative diseases (Dobson 2001; Hautot et al. 2003). A biologic model of coated nanomaterials can be found in ferritin, which is an approximately 12-nm-large iron storage protein that contains 5- to 7-nm-sized hydrous ferric oxide phosphate inside a protective protein shell (Donlin et al. 1998). Nanosized materials, including fullerenes, occur naturally from combustion processes such as forest fires and volcanoes.

Obvious differences between unintentional and intentional anthropogenic NSPs are the polydispersed and chemically complex nature (elemental, soluble, and volatile carbon compounds; soluble and poorly soluble inorganics; Cyrys et al. 2003; Hughes et al. 1998) of the former, in contrast to the monodisperse and precise chemically engineered characteristics and solid form of the latter, generated in gas or liquid phase [National Nanotechnology Initiative (NNI) 2004]. However, despite these differences, the same toxicologic principles are likely to apply for NPs, because not only size but also a number of other particle parameters determine their biologic activity. The multitude of interactions of these factors has yet to be assessed, and in this article we attempt to summarize our present knowledge.

NSPs are variably called ultrafine particles (UFPs) by toxicologists [U.S. Environmental Protection Agency (EPA) 2004], Aitken mode and nucleation mode particles by atmospheric scientists [Kulmala 2004; National Research Council (NRC) 1983], and engineered nanostructured materials by materials scientists (NNI 2004). Figure 1 depicts the range of sizes of airborne ambient particulate matter, including the nucleation-mode, Aitken-mode, accumulation-mode, and coarse-mode particles. Ambient particles < 0.1 µm, defined as UFPs in the toxicologic literature, consist of transient nuclei or Aitken nuclei (NRC 1983). More recently, even smaller particles in the nucleation mode with peak diameters around 4 nm have been observed (McMurry and Woo 2002). The distinction between NSPs generated by internal combustion engines and NPs becomes further clouded by the finding of nanotubes in diesel exhaust (Evelyn et al. 2003). The use of the term “nano” in this review reflects only particle size and not chemical composition. For the purposes of this review, we use the following terms: “Nanosized particle” (NSP) includes all engineered and ambient nanosized spherical particles < 100 nm. “Engineered nanoparticle” (NP) includes only spherical NSPs specifically engineered in the laboratory; other engineered nanosized structures will be labeled according to their shape, for example, nanotubes, nanofibers, nanowires, nanorings, and so on. “Ultrafine particle” (UFP) includes ambient and laboratory-generated NSPs that are not produced in a controlled, engineered way.

Table 2 shows the tremendous differences in particle number concentrations and particle surface areas for particles of the four ambient modes, assuming an airborne concentration of 10 µg/m³ of unit density particles of each size. The extraordinarily high number concentrations of NSPs per given mass will likely be of toxicologic significance when these particles interact with cells and subcellular components. Likewise, their increased surface area per unit mass can be toxicologically important if other characteristics such as surface chemistry and bulk chemistry are the same. Although the mass of UFPs in ambient air is very low, approaching only 0.5-2 µg/m³ at background levels (Hughes et al. 1998), it can increase several-fold during high pollution episodes or on highways (Brand et al. 1991; Shi et al. 2001; Zhu et al. 2002).

Physicochemical characteristics as determinants of biologic activity. The small size and corresponding large specific surface area of solid NSPs (Table 2) confer specific properties to them, for example, making them desirable as catalysts for chemical reactions. The importance of surface area becomes evident when considering that surface atoms or molecules play a dominant role in determining bulk properties (Amato 1989); the ratio of surface to total atoms or molecules increases exponentially with decreasing particle size (Figure 2). Increased surface reactivity predicts that NSPs exhibit greater biologic activity per given mass compared with larger particles, should they be taken up into living organisms and provided they are solid rather than solute particles. This increased biologic activity can be either positive and desirable (e.g., antioxidant activity, carrier capacity for therapeutics, penetration of cellular barriers for drug delivery) or negative and undesirable (e.g., toxicity, induction of oxidative stress or of cellular dysfunction), or a mix of both. Not only may adverse effects be induced, but interactions of NSPs with cells and subcellular structures and their biokinetics are likely to be very different from those of larger-sized particles. For example, more than 60 years ago virologists described the translocation of 30- to 50-nm-sized virus particles along axons and dendrites of neurons and across epithelia (Howe and Bodian 1940), whereas first reports about increased inflammatory activity and epithelial translocation of man-made 20- and 30-nm solid particles appeared only more recently (Ferin et al. 1990; Oberdörster et al. 1990).

The characteristic biokinetic behaviors of NPs are attractive qualities for promising applications in medicine as diagnostic and therapeutic devices and as tools to investigate and understand molecular processes and structures in living cells (Akerman et al. 2002; Foley et al. 2002; Kreuter 2001; Li et al. 2003). For example, targeted drug delivery to tissues that are difficult to reach [e.g., central nervous system (CNS)], NPs for the fight against cancer, intravascular nanosensor and nanorobotic devices, and diagnostic and imaging procedures are presently under development. The discipline of nanomedicine--defined as medical application of nanotechnology and related research--has arisen to design, test, and optimize these applications so that they can eventually be used routinely by physicians (Freitas 1999).

However, in apparent stark contrast to the many efforts aimed at exploiting desirable properties of NPs for improving human health are the limited attempts to evaluate potential undesirable effects of NPs when administered intentionally for medicinal purposes, or after unintentional exposure during manufacture or processing for industrial applications. The same properties that make NPs so attractive for development in nanomedicine and for specific industrial processes could also prove deleterious when NPs interact with cells. Thus, evaluating the safety of NPs should be of highest priority given their expected worldwide distribution for industrial applications and the likelihood of human exposure, directly or through release into the environment (air, water, soil). Nanotoxicology--an emerging discipline that can be defined as “science of engineered nanodevices and nanostructures that deals with their effects in living organisms”--is gaining increased attention. Nanotoxicology research not only will provide data for safety evaluation of engineered nanostructures and devices but also will help to advance the field of nanomedicine by providing information about their undesirable properties and means to avoid them.

Human exposure to nanosized materials. In addition to natural and anthropogenic sources of UFPs in the ambient air, certain workplace conditions also generate NSPs that can reach much higher exposure concentrations, up to several hundred micrograms per cubic meter, than is typically found at ambient levels. In contrast to airborne UFP exposures occurring via inhalation at the workplace and from ambient air, not much is known about levels of exposure via different routes for NPs, whether it is by direct human exposure or indirectly through contamination of the environment. For example, are there or will there be significant exposures to airborne singlet engineered carbon nanotubes or C₆₀ fullerenes? First measurements at a model workplace gave only very low concentrations, < 50 µg/m³, and these were most likely in the form of aggregates (Maynard et al. 2004). However, even very low concentrations of nanosized materials in the air represent very high particle number concentrations, as is well known from measurements of ambient UFPs (Hughes et al. 1998). For example, a low concentration of 10 µg/m³ of unit density 20-nm particles translates into > 1 10⁶ particles/cm³ (Table 2). Inhalation may be the major route of exposure for NPs, yet ingestion and dermal exposures also need to be considered during manufacture, use, and disposal of engineered nanomaterials, and specific biomedical applications for diagnostic and therapeutic purposes will require intravenous, subcutaneous, or intramuscular administration (Table l). It can be assumed, however, that the toxicology of NPs can build on the experience and data already present in the toxicology literature of ambient UFPs. [Additional details provided in Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf).]

Manufactured nanomaterials in the environment. Manufactured nanomaterials are likely to enter the environment for several reasons. Some are and others will be produced by the ton, and some of any material produced in such mass quantities is likely to reach the environment from manufacturing effluent or from spillage during shipping and handling. They are being used in personal-care products such as cosmetics and sunscreens and can therefore enter the environment on a continual basis from washing off of consumer products (Daughton and Ternes 1999). They are being used in electronics, tires, fuel cells, and many other products, and it is unknown whether some of these materials may leak out or be worn off over the period of use. They are also being used in disposable materials such as filters and electronics and may therefore reach the environment through landfills and other methods of disposal.

Scientists have also found ways of using nanomaterials in remediation. Although many of these are still in testing stages (Chitose et al. 2003; Jaques and Kim 2000; Joo et al. 2004; Nagaveni et al. 2004; Nghiem et al. 2004; Tungittiplakorn et al. 2004), dozens of sites have already been injected with various nanomaterials, including nano-iron (Mach 2004). Testing to determine the safety of these NPs used for remediation to environmentally relevant species has not yet been done. Although most people are concerned with effects on large wildlife, the basis of many food chains depends on the benthic and soil flora and fauna, which could be dramatically affected by such NP injections. In addition, as noted by Lecoanet et al. (2004), nanosized materials may not migrate through soils at rapid enough rates to be valuable in remediation. Future laboratory and field trials will help clarify the line between remediation and contamination [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

In recent years, interest in potential effects of exposure to airborne UFPs has increased considerably, and studies have shown that they can contribute to adverse health effects in the respiratory tract as well as in extrapulmonary organs. Results on direct effects of ambient and model UFPs have been reported from epidemiologic studies and controlled clinical studies in humans, inhalation/instillation studies in rodents, or in vitro cell culture systems. For example, several epidemiologic studies have found associations of ambient UFPs with adverse respiratory and cardiovascular effects resulting in morbidity and mortality in susceptible parts of the population (Pekkanen et al. 1997; Penttinen et al. 2001; Peters et al. 1997a, 1997b; von Klot et al. 2002; Wichmann et al. 2002), whereas other epidemiologic studies have not seen such associations (Pekkanen et al. 1997; Tiittanen et al. 1999). Controlled clinical studies evaluated deposition and effects of laboratory-generated UFPs. High deposition efficiencies in the total respiratory tract of healthy subjects were found, and deposition was even greater in subjects with asthma or chronic obstructive pulmonary disease. In addition, effects on the cardiovascular system, including blood markers of coagulation and systemic inflammation and pulmonary diffusion capacity, were observed after controlled exposures to carbonaceous UFPs (Anderson et al. 1990; Brown et al. 2002; Chalupa et al. 2004; Henneberger et al., 2005; Jaques and Kim 2000; Pekkanen et al. 2002; Pietropaoli et al. 2004; Wichmann et al. 2000).

Studies in animals using laboratory-generated model UFPs or ambient UFPs showed that UFPs consistently induced mild yet significant pulmonary inflammatory responses as well as effects in extrapulmonary organs. Animal inhalation studies included the use of different susceptibility models in rodents, with analysis of lung lavage parameters and lung histopathology, effects on the blood coagulation cascade, and translocation studies to extrapulmonary tissues (Elder et al. 2000, 2002, 2004; Ferin et al. 1991; Ferin and Oberdörster 1992; Kreyling et al. 2002; Li et al. 1999; Nemmar et al. 1999, 2002a, 2002b, 2003; Oberdörster et al. 1992a, 1995, 2000, 2002, 2004; Semmler et al. 2004; Zhou et al. 2003).

Figure 3. Hypothetical cellular interaction of NSPs (adapted from Donaldson and Tran 2002). EGFR, epidermal growth factor receptor. Inflammation and oxidative stress can be mediated by several primary pathways: a) the particle surface causes oxidative stress resulting in increased intracellular calcium and gene activation; b) transition metals released from particles result in oxidative stress, increased intracellular calcium, and gene activation; c) cell surface receptors are activated by transition metals released from particles, resulting in subsequent gene activation; or d) intracellular distribution of NSPs to mitochondria generates oxidative stress.

Figure 4. Percentage of neutrophils in lung lavage of rats (A,B) and mice (C,D) as indicators of inflammation 24 hr after intratracheal instillation of different mass doses of 20-nm and 250-nm TiO2 particles in rats and mice. (A,C) The steeper dose response of nanosized TiO2 is obvious when the dose is expressed as mass. (B,D) The same dose response relationship as in (A,C) but with dose expressed as particle surface area; this indicates that particle surface area seems to be a more appropriate dosemetric for comparing effects of different-sized particles, provided they are of the same chemical structure (anatase TiO2 in this case). Data show mean ± SD.

In vitro studies using different cell systems showed varying degrees of proinflammatory- and oxidative-stress-related cellular responses after dosing with laboratory-generated or filter-collected ambient UFPs (Brown et al. 2000, 2001; Li et al. 2003). Collectively, the in vitro results have identified oxidative-stress-related changes of gene expression and cell signaling pathways as underlying mechanisms of UFP effects, as well as a role of transition metals and certain organic compounds on combustion-generated UFPs (Figure 3). These can alter cell signaling pathways, including Ca²⁺ signaling and cytokine signaling (e.g., interleukin-8) (Donaldson et al. 2002; Donaldson and Stone 2003). Effects were on a mass basis greater for model UFPs than for fine particles, whereas effects for ambient UFP cellular responses were sometimes greater and sometimes less than those of fine and coarse particles. The interpretation of the in vitro studies is oftentimes difficult because particles of different chemical compositions were used, target cells were different, and duration, end points, and generally high dose levels also differed. Results from high doses in particular should be viewed with caution if they are orders of magnitude higher than predicted from relevant ambient exposures (see “Exposure dose-response considerations,” below). [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/
7339/supplemental.pdf).]

Laboratory rodent studies. With respect to potential health effects of NSPs, two examples should serve to illustrate a) that NSPs have a higher inflammatory potential per given mass than do larger particles, provided they are chemically the same, and b) that NSPs generated under certain occupational conditions can elicit severe acute lung injury.

The first example involves studies with ultrafine and fine titanium dioxide (TiO₂) particles, which showed that ultrafine anatase TiO₂ (20 nm), when instilled intratracheally into rats and mice, induced a much greater pulmonary-inflammatory neutrophil response (determined by lung lavage 24 hr after dosing) than did fine anatase TiO₂ (250 nm) when both types of particles were instilled at the same mass dose (Figure 4A,C). However, when the instilled dose was expressed as particle surface area, it became obvious that the neutrophil response in the lung for both ultrafine and fine TiO₂ fitted the same dose-response curve (Figure 4B,D), suggesting that particle surface area for particles of different sizes but of the same chemistry, such as TiO₂, is a better dosemetric than is particle mass or particle number (Oberdörster G 2000). Moreover, normalizing the particle surface dose to lung weight shows excellent agreement of the inflammatory response in both rats and mice [Figure S-2 in Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/
supplemental.pdf)]. The better fit of dose-response relationships by expressing the dose as surface area rather than mass when describing toxicologic effects of inhaled solid particles of different sizes has been pointed out repeatedly, especially when particles of different size ranges--nano to fine--were studied (Brown et al. 2001; Donaldson et al. 1998, 2002; Driscoll 1996; Oberdörster and Yu 1990; Oberdörster et al. 1992a; Tran et al. 1998, 2000) [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

Particle chemistry, and specifically surface chemistry, plays a decisive role in addition to particle size, as shown in the second example: exposure of rats to polytetrafluoroethylene (PTFE) fume. PTFE fume (generated by heating PTFE) has long been known to be of high acute toxicity to birds and mammals, including humans (Cavagna et al. 1961; Coleman et al. 1968; Griffith et al. 1973; Nuttall et al. 1964; Waritz and Kwon 1968). Analysis of these fumes revealed the nanosized nature of the particles generated by heating PTFE to about 480°C, with a count median diameter (CMD) of 18 nm. They were highly toxic to rats, causing severe acute lung injury with high mortality within 4 hr after a 15-min inhalation exposure to 50 µg/m³ (Oberdörster et al. 1995). This short exposure resulted in an estimated deposited dose in the alveolar regions of only approximately 60 ng. In humans, acute lung injury, known as polymer fume fever, can result from exposures to PTFE fumes (Auclair et al. 1983; Goldstein et al. 1987; Lee et al. 1997; Williams et al. 1974; Woo et al. 2001). Additional rat studies showed that the gas phase alone was not acutely toxic and that aging of the PTFE fume particles for 3 min increased their particle size to > 100 nm because of accumulation, which resulted in a loss of toxicity (Johnston et al. 2000). However, it is most likely that changes in particle surface chemistry during the aging period contributed to this loss of toxicity, and that this is not just an effect of the accumulated larger particle size [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

These examples seem to represent the extremes of NSPs in terms of pulmonary toxicity, one (TiO₂) being rather benign yet still inducing significantly greater inflammatory effects on a mass basis than fine particles of the same chemical makeup, the other (PTFE fumes) inducing very high acute toxicity, possibly related to reactive groups on the large surface per unit mass.

Engineered nanomaterials can have very different shapes, for example, spheres, fibers, tubes, rings, and planes. Toxicologic studies of spherical and fibrous particles have well established that natural (e.g., asbestos) and man-made (e.g., biopersistent vitreous) fibers are associated with increased risks of pulmonary fibrosis and cancer after prolonged exposures [Greim et al. 2001; International Agency for Research on Cancer (IARC) 2002]. Critical parameters are the three Ds: dose, dimension, and durability of the fibers. Fibers are defined as elongated structures with a diameter-to-length ratio (aspect ratio) of 1: 3 or greater and with a length of > 5 µm and diameter ≤ 3 µm [World Health Organization (WHO) 1985]. Carbon nanotubes have aspect ratios of up to ≥ 100, and length can exceed 5 µm with diameters ranging from 0.7 to 1.5 nm for single-walled nanotubes, and 2 to 50 nm for multiwalled nanotubes. Results from three studies using intratracheal dosing of carbon nanotubes in rodents indicate significant acute inflammatory pulmonary effects that either subsided in rats (Warheit et al. 2004) or were more persistent in mice (Lam et al. 2004; Shvedova et al. 2004b). Administered doses were very high, ranging from 1 to 5 mg/kg in rats; in mice doses ranged from 3.3 to 16.6 mg/kg (Lam et al. 2004) or somewhat lower, from 0.3 to 1.3 mg/kg (Shvedova et al. 2004a). Granuloma formation as a normal foreign body response of the lung to high doses of a persistent particulate material was a consistent finding in these studies. Metal impurities (e.g., iron) from the nanotube generation process may also have contributed to the observed effects. Although these in vivo first studies revealed high acute effects, including mortality, this was explained by the large doses of the instilled highly aggregated nanotubes that caused death by obstructing the airways and should not be considered a nanotube effect per se (Warheit et al. 2004). Invitro studies with carbon nanotubes also reported significant effects. Dosing keratinocytes and bronchial epithelial cells invitro with single-walled carbon nanotubes (SWNTs) resulted in oxidative stress, as evidenced by the formation of free radicals, accumulation of peroxidative products, and depletion of cell antioxidants (Shvedova et al. 2004a, 2004b). Multiwalled carbon nanotubes (MWNTs) showed proinflammatory effects and were internalized in keratinocytes (Monteiro-Riviere et al. 2005). Again, the relatively high doses applied in these studies need to be considered when discussing the relevancy of these findings for invivo exposures. A most recent study in macrophages comparing SWNTs and MWNTs with C₆₀ fullerenes found a cytotoxicity ranking on a mass basis in the order SWNT > MWNT > C₆₀. Profound cytotoxicity (mitochondrial function, cell morphology, phagocytic function) was seen for SWNTs, even at a low concentration of 0.38 µg/cm². The possible contribution of metal impurities of the nanotubes still needs to be assessed. Therefore, whether the generally recognized principles of fiber toxicology apply to these nanofiber structures needs still to be determined (Huczko et al. 2001).

Future studies should be designed to investigate both effects and also the fate of nanotubes after deposition in the respiratory tract, preferentially by inhalation using well-dispersed (singlet) airborne nanotubes. In order to design the studies using appropriate dosing, it is necessary to assess the likelihood and degree of human exposure. It is of utmost importance to characterize human exposures in terms of the physicochemical nature, the aggregation state, and concentration (number, mass, surface area) of engineered nanomaterials and perform animal and invitro studies accordingly. If using direct instillation into the lower respiratory tract, a large range of doses, which include expected realistic exposures of appropriately prepared samples, needs to be considered [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

Ecotoxicologic studies. Studies have been carried out to date on only a few species that have been accepted by regulatory agencies as models for defining ecotoxicologic effects. Tests with uncoated, water-soluble, colloidal fullerenes (nC₆₀) show that the 48-hr LC₅₀ (median lethal concentration) in Daphnia magna is 800 ppb (Oberdörster E 2004b), using standard U.S. EPA protocols (U.S. EPA 1994). In largemouth bass (Micropterus salmoides), although no mortality was seen, lipid peroxidation in the brain and glutathione depletion in the gill were observed after exposure to 0.5 ppm nC₆₀ for 48 hr (Oberdörster E 2004a). There are several hypotheses as to how lipid damage may have occurred in the brain, including direct redox activity by fullerenes reaching the brain via circulation or axonal translocation (see also “Disposition of NSPs in the respiratory tract,” below) and dissolving into the lipid-rich brain tissue; oxyradical production by microglia; or reactive fullerene metabolites may be produced by cytochrome P450 metabolism. Initial follow-up studies using suppressive subtractive hybridization of pooled control fish versus pooled 0.5-ppm-exposed fish liver mRNA were also performed. Proteins related to immune responses and tissue repair were up-regulated, and several proteins related to homeostatic control and immune control were down-regulated. A cytochrome P450 (CYP2K4) involved in lipid metabolism was up-regulated [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)]. In addition to these biochemical and molecular-level changes in fish, bactericidal properties of fullerenes have also been reported and are being explored as potential new antimicrobial agents (Yamakoshi et al. 2003). Engineered nanomaterials used as antimicrobials may shift microbial communities if they are released into the environment via effluents. As we know from anthropogenic endocrine-disrupting compounds, interference of signaling between nitrogen-fixing bacteria and their plant hosts could be extremely harmful both ecologically and economically in terms of crop production (Fox et al. 2001).

Figure 5. Routes of exposure, uptake, distribution, and degradation of NSPs in the environment. Solid lines indicate routes that have been demonstrated in the laboratory or field or that are currently in use (remediation). Magenta lettering indicates possible degradation routes, and blue lettering indicates possible sinks and sources of NSPs.

Figure 6. NPs have been shown to release oxyradicals [pictured here is the mechanism of C60 as determined by Yamakoshi et al. (2003)], which can interact with the antioxidant defense system. Abbreviations: GPx, glutathione peroxidase; GSH, reduced glutathione; GSSG, oxidized glutathione; ISC, intersystem crossing; R, any organic molecule; SOD, superoxide dismutase. In addition to fullerenes, metals such as cadmium, iron, or nickel quantum dots, or iron from SWNT manufacturing, could also act in Fenton-type reactions. Phase II biotransformation, ascorbic acid, vitamin E, beta carotene, and other interactions are not shown.

Aqueous fullerenes and coated SWNTs are stable in salt solutions (Cheng et al. 2004; Warheit et al. 2004), cell culture media (Lu et al. 2004; Sayes et al. 2004), reconstituted hard water, and MilliQ water (Dieckmann et al. 2003; Oberdörster E 2004a, 2004b). NSPs will tend to sorb onto sediment and soil particles and be immobilized because of their high surface area:mass ratio (Lecoanet and Wiesner 2004). Biologic transport would occur from ingested sediments, and one would expect movement of nanomaterials through the food chain (Figure 5).

To make engineered nanomaterials more biocompatible, both surface coatings and covalent surface modifications have been incorporated. Some studies have shown that both the surface coating and the covalent modifications can be weathered either by exposure to the oxygen in air or by ultraviolet (UV) irradiation for 1-4 hr (Derfus et al. 2004; Rancan et al. 2002). Therefore, although coatings and surface modifications may be critically important in drug-delivery devices, the likelihood of weathering under environmental conditions makes it important to study toxicity under UV and air exposure conditions. Even coatings used in drug delivery of NPs may not be biopersistent or could be metabolized to expose the core NP material [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/
supplemental.pdf)].

Reactive oxygen species mechanisms of NSP toxicity. Both in vivo and in vitro, NSPs of various chemistries have been shown to create reactive oxygen species (ROS). ROS production has been found in NPs as diverse as C₆₀ fullerenes, SWNTs, quantum dots, and UFPs, especially under concomitant exposure to light, UV, or transition metals (Brown et al. 2000, 2001; Derfus et al. 2004; Joo et al. 2004; Li et al. 2003; Nagaveni et al. 2004; Oberdörster E 2004a; Rancan et al. 2002; Sayes et al. 2004; Shvedova et al. 2004a, 2004b; Wilson et al. 2002; Yamakoshi et al. 2003). It has been demonstrated that NSPs of various sizes and various chemical compositions preferentially mobilize to mitochondria (de Lorenzo 1970; Foley et al. 2002; Gopinath et al. 1978; Li et al. 2003; Rodoslav et al. 2003). Because mitochondria are redox active organelles, there is a likelihood of altering ROS production and thereby overloading or interfering with antioxidant defenses (Figure 3).

Figure 6 diagrams some of the antioxidant defense systems that occur in animals, and possible areas where NSPs may create oxyradicals. The C₆₀ fullerene is shown as a model NP producing superoxide, as has been shown by Yamakoshi et al. (2003). The exact mechanism by which each of these diverse NPs cause ROS is not yet fully understood, but suggested mechanisms include a) photo excitation of fullerenes and SWNTs, causing intersystem crossing to create free electrons; b) metabolism of NPs to create redox active intermediates, especially if metabolism is via cytochrome P450s; and c) inflammation responses in vivo that may cause oxyradical release by macrophages. Other mechanisms will likely emerge as studies on NP toxicity continue.

The small size and respective large specific surface area of NPs, like those of ambient airborne UFPs, give them unique properties with respect to a potential to cause adverse effects. Certainly, as shown in studies with UFPs, chemical composition and other particle parameters are additional important effect modifiers. Results from these studies will therefore serve as a basis for future studies in the field of nanotoxicology, for example, the propensity of NSPs to translocate across cell layers and along neuronal pathways (see “Disposition of NSPs in the respiratory tract” below).

Exposure dose-response considerations. A careful evaluation of exposure-dose-response relationships is critical to the toxicologic assessment of NSPs. This includes not only questions about the dosemetric--mass, number, or surface of the particles, as discussed above--but most important, also the relevance of dose levels. For example, it is tempting, and a continual practice, to dose primary cells or cell lines in vitro with very high doses without any consideration or discussion of realistic in vivo exposures; for instance, 100 µg NSPs/mL culture medium--labeled as a low dose--is extremely high and is unlikely to be encountered in vivo. Likewise, intratracheal instillations of several hundred micrograms into a rat does not resemble a relevant in vivo inhalation exposure; both dose and dose rate cause high bolus dose artifacts. Although such studies may be used in a first proof-of-principle approach, it is mandatory to follow up and validate results using orders of magnitude lower concentrations resembling realistic in vivo exposures, including worst-case scenarios. The 500-year-old phrase “the dose makes the poison” can also be paraphrased as “the dose makes the mechanism.” The mechanistic pathways that operate at low realistic doses are likely to be different from those operating at very high doses when the cell’s or organism’s defenses are overwhelmed.

Figure 7. Some basic shapes of exposure-response or dose-response relationships. Abbreviations: H, hormetic (biphasic); L, linear (no threshold); S, supralinear; T, threshold. Prerequisites for establishing these relationships for NSPs from in vitro or in vivo studies include a sufficient number of data points, that is, over a wide range of exposure concentrations or doses; knowledge about exposure levels; and information about correlation of exposure with doses at the organismal or cellular level (an exposure is not a dose). Dose-response curves of different shapes can be extrapolated when only response data at high dose levels (indicated by dashed oval) are available. Lack of data in the low--oftentimes the most relevant--dose range can result in severe misinterpretation if a threshold or even a hormetic response is present. Consideration also needs to be given to the likelihood that the shape or slope of exposure-dose-response relationships change for susceptible parts of the population.

Figure 8. Predicted fractional deposition of inhaled particles in the nasopharyngeal, tracheobronchial, and alveolar region of the human respiratory tract during nose breathing. Based on data from the International Commission on Radiological Protection (1994). Drawing courtesy of J. Harkema.

Therefore, in vivo and in vitro studies will provide useful data on the toxicity and mode of action of NSPs only if justifiable concentrations/doses are considered when designing such studies. This approach is particularly important for the proper identification of the dose-response curve. When data are generated only at high concentrations/doses, it is difficult to determine whether the dose-response curve in question is best described by a linear (no threshold), supralinear, threshold, or hormetic model (Figure 7). Study designs should include doses that most closely reflect the expected exposure levels. A critical gap that urgently needs to be filled in this context is the complete lack of data for human or environmental exposure levels of NSPs. Furthermore, some knowledge about the biokinetics of NSPs is required in order to estimate appropriate doses. Do specific NPs reach certain target sites? If so, what are the doses, dose rates, and their persistence? Further, although it may be tempting to extrapolate from in vitro results to an in vivo risk assessment, it is important to keep in mind that in vitro tests are most useful in providing information on mechanistic processes and in elucidating mechanisms/mode of actions suggested by studies in whole animals. A combination of in vitro and in vivo studies with relevant dose levels will be most useful in identifying the potential hazards of NPs, and a thorough discussion and justification of selected dose levels should be mandatory.

Most of the toxicity research on NSPs in vivo has been carried out in mammalian systems, with a focus on respiratory system exposures for testing the hypothesis that airborne UFPs cause significant health effects. With respect to NPs, other exposure routes, such as skin and GI tract, also need to be considered as potential portals of entry. Portal-of-entry-specific defense mechanisms protect the mammalian organism from harmful materials. However, these defenses may not always be as effective for NSPs, as is discussed below.

In order to appreciate what dose the organism receives when airborne particles are inhaled, information about their deposition as well as their subsequent fate is needed. Here we focus on the fate of inhaled nanosized materials both within the respiratory tract itself and translocated out of the respiratory tract. There are significant differences between NSPs and larger particles regarding their behavior during deposition and clearance in the respiratory tract [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/
supplemental.pdf)].

Efficient deposition of inhaled NSPs. The main mechanism for deposition of inhaled NSPs in the respiratory tract is diffusion due to displacement when they collide with air molecules. Other deposition mechanisms of importance for larger particles, such as inertial impaction, gravitational settling, and interception, do not contribute to NSP deposition, and electrostatic precipitation occurs only in cases where NSPs carry significant electric charges. Figure 8 shows the fractional deposition of inhaled particles in the nasopharyngeal, tracheobronchial, and alveolar regions of the human respiratory tract under conditions of nose breathing during rest, based on a predictive mathematical model (International Commission on Radiological Protection 1994). These predictions apply to particles that are inhaled as singlet particles of a given size and not as aggregates; the latter obviously will have larger particle size and different deposition site. In each of the three regions of the respiratory tract, significant amounts of a certain size of NSPs (1-100 nm) are deposited. For example, 90% of inhaled 1-nm particles are deposited in the nasopharyngeal compartment, only approximately 10% in the tracheobronchial region, and essentially none in the alveolar region. On the other hand, 5-nm particles show about equal deposition of approximately 30% of the inhaled particles in all three regions; 20-nm particles have the highest deposition efficiency in the alveolar region (~ 50%), whereas in tracheobronchial and nasopharyngeal regions this particle size deposits with approximately 15% efficiency. These different deposition efficiencies should have consequences for potential effects induced by inhaled NSPs of different sizes as well as for their disposition to extrapulmonary organs, as discussed further below.

Disposition of NSPs in the respiratory tract. In the preceding section we summarized data demonstrating that inhaled NSPs of different sizes can target all three regions of the respiratory tract. Several defense mechanisms exist throughout the respiratory tract aimed at keeping the mucosal surfaces free from cell debris and particles deposited by inhalation. Several reviews describe the well-known classic clearance mechanisms and pathways for deposited particles (Kreyling and Scheuch 2000; Schlesinger et al. 1997; U.S. EPA 2004), so here we only briefly mention those mechanisms and point out specific differences that exist with respect to inhaled NSPs [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

Once deposited, NSPs--in contrast to larger-sized particles--appear to translocate readily to extrapulmonary sites and reach other target organs by different transfer routes and mechanisms. One involves transcytosis across epithelia of the respiratory tract into the interstitium and access to the blood circulation directly or via lymphatics, resulting in distribution throughout the body. The other is a not generally recognized mechanism that appears to be distinct for NSPs and that involves their uptake by sensory nerve endings embedded in airway epithelia, followed by axonal translocation to ganglionic and CNS structures.

Figure 9. Pathways of particle clearance (disposition) in and out of the respiratory tract. There are significant differences between NSPs and larger particles for some of these pathways (see “Disposition of NSPs in the respiratory tract”). Drawing courtesy of J. Harkema.

Table 3

Figure 10. In vivo retention of inhaled nanosized and larger particles in alveolar macrophages (A) and in exhaustively lavaged lungs (epithelial and interstitial retention; B) 24 hr postexposure. The alveolar macrophage is the most important defense mechanism in the alveolar region for fine and coarse particles, yet inhaled singlet NSPs are not efficiently phagocytized by alveolar macrophages.

Classical clearance pathways. The clearance of deposited particles in the respiratory tract is basically due to two processes (Table 3): a) physical translocation of particles by different mechanisms and b) chemical clearance processes. Leaching refers to loss of elements from a particle matrix (e.g., loss of sodium from asbestos fibers due to dissolution in intra- or extracellular milieu). Chemical dissolution is directed at biosoluble particles or components of particles that are either lipid soluble or soluble in intracellular and extracellular fluids. Solutes and soluble components can then undergo absorption and diffusion or binding to proteins and other subcellular structures and may be eventually cleared into blood and lymphatic circulation. Chemical clearance for biosoluble materials can happen at any location within the three regions of the respiratory tract, although to different degrees, depending on local extracellular and intracellular conditions (pH). In contrast, a number of diverse processes involving physical translocation of inhaled particles exist in the respiratory tract and are different in the three regions. Figure 9 summarizes these clearance processes for solid particles. As discussed further below, some of them show significant particle-size-dependent differences, making them uniquely effective for a certain particle size but very inefficient for other sizes.

The most prevalent mechanism for solid particle clearance in the alveolar region is mediated by alveolar macrophages, through phagocytosis of deposited particles. The success of macrophage-particle encounter appears to be facilitated by chemotactic attraction of alveolar macrophages to the site of particle deposition (Warheit et al. 1988). The chemotactic signal is most likely complement protein 5a (C5a), derived from activation of the complement cascade from serum proteins present on the alveolar surface (Warheit et al. 1986; Warheit and Hartsky 1993). This is followed by gradual movement of the macrophages with internalized particles toward the mucociliary escalator. The retention half-time of solid particles in the alveolar region based on this clearance mechanism is about 70 days in rats and up to 700 days in humans. The efficacy of this clearance mechanism depends highly on the efficiency of alveolar macrophages to “sense” deposited particles, move to the site of their deposition, and then phagocytize them. This process of phagocytosis of deposited particles takes place within a few hours, so by 6-12 hr after deposition essentially all of the particles are phagocytized by alveolar macrophages, to be cleared subsequently by the slow alveolar clearance mentioned above. However, it appears that there are significant particle-size-dependent differences in the cascade of events leading to effective alveolar macrophage-mediated clearance.

Figure 10 displays results of several studies in which rats were exposed to different-sized particles (for the 3- and 10-µm particles, 10-µg and 40-µg polystyrene beads, respectively, were instilled intratracheally) (Kreyling et al. 2002; Oberdörster et al. 1992b, 2000; Semmler et al. 2004). Twenty-four hours later, the lungs of the animals were lavaged repeatedly, retrieving about 80% of the total macrophages as determined in earlier lavage experiments (Ferin et al. 1991). As shown in Figure 10, approximately 80% of 0.5-, 3-, and 10-µm particles could be retrieved with the macrophages, whereas only approximately 20% of nanosized 15-20-nm and 80-nm particles could be lavaged with the macrophages. In effect, approximately 80% of the UFPs were retained in the lavaged lung after exhaustive lavage, whereas approximately 20% of the larger particles > 0.5 µm remained in the lavaged lung. This indicates that NSPs either were in epithelial cells or had further translocated to the interstitium [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

Epithelial translocation. Because of the apparent inefficiency of alveolar macrophage phagocytosis of NSPs, one might expect that these particles interact instead with epithelial cells. Indeed, results from several studies show that NSPs deposited in the respiratory tract readily gain access to epithelial and interstitial sites. This was also shown in studies with ultrafine PTFE fumes: shortly after a 15-min exposure, the fluorine-containing particles could be found in interstitial and submucosal sites of the conducting airways as well as in the interstitium of the lung periphery close to the pleura (Oberdörster G 2000). Such interstitial translocation represents a shift in target site away from the alveolar space to the interstitium, potentially causing direct particle-induced effects there.

In a study evaluating the pulmonary inflammatory response of TiO₂ particles, ranging from NP TiO₂ to pigment-grade TiO₂ (12-250 nm), a surprising finding was that, 24 hr after intratracheal instillation of different doses, higher doses induced a lower effect (Oberdörster et al. 1992a). This was explained by the additional finding that at the higher doses (expressed as particle surface area) of the nanosized TiO₂, ≥ 50% had reached the pulmonary interstitium, causing a shift of the inflammatory cell response from the alveolar space to the interstitium [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)]. The smaller particle size of 12 and 20 nm versus 220 and 250 nm also means that the administered particle number was more than three orders of magnitude higher for the NSPs, a factor that seems to be an important determinant for particle translocation across the alveolar epithelium, as are the delivered total dose and the dose rate (Ferin et al. 1992). Because interstitial translocation of fine particles across the alveolar epithelium is more prominent in larger species (dogs, nonhuman primates) than in rodents (Kreyling and Scheuch 2000; Nikula et al. 1997), it is reasonable to assume that the high translocation of NSPs observed in rats occurs in humans as well [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

Translocation to the circulatory system. Once the particles have reached pulmonary interstitial sites, uptake into the blood circulation, in addition to lymphatic pathways, can occur; again, this pathway is dependent on particle size, favoring NSPs. Berry et al. (1977) were the first to describe translocation of NSPs across the alveolar epithelium using intratracheal instillations of 30-nm gold particles in rats. Within 30 min postexposure, they found large amounts of these particles in platelets of pulmonary capillaries; the researchers suggested that this is an elimination pathway for inhaled particles that is significant for transporting the smallest air pollutant particles--in particular, particles of tobacco smoke--to distant organs. They also hypothesized that this “might predispose to platelet aggregation with formation of microthrombi atheromatous plaques” (Berry et al. 1977).

Table 4

Figure 11. Different forms of caveolae and cellular tight junctions function as translocation mechanisms across cell layers. Depending on particle surface chemistry, NSPs have been shown to transcytose across alveolar type I epithelial cells and capillary endothelial cells (Table 4), but not via cellular tight junctions in the healthy state (A). However, in a compromised or disease state (e.g., endotoxin exposure; B) translocation across widened tight junction occurs as well (Heckel et al. 2004). This indicates that assessing potential effects of NSPs in the compromised state is an important component of nanotoxicology. Adapted from Cohen et al. (2004).

Table 5

Figure 12. Close proximity of olfactory mucosa to olfactory bulb of the CNS. Inhaled NSP[s], especially below 10 nm, deposit efficiently on the olfactory mucosa by diffusion, similar to airborne “smell” molecules which deposit in this area of olfactory dendritic cilia. Subsequent uptake and translocation of solid NSP[s] along axons of the olfactory nerve has been demonstrated in non-human primates and rodents. Surface chemistry of the particles may influence their neuronal translocation. Copyright © the McGraw-Hill Companies, Inc. Reproduced from Widmaier et al. (2004) with permission from McGraw-Hill.

Table 6

Since then, a number of studies with different particle types have confirmed the existence of this translocation pathway, as summarized in Table 4. Collectively, these studies indicate that particle size and surface chemistry (coating), and possibly charge, govern translocation across epithelial and endothelial cell layers. In particular, the studies summarized by Mehta et al. (2004) and those performed by Heckel et al. (2004) using intravenous administration of albumin-coated gold nanoparticles in rodents demonstrated receptor-mediated transcytosis (albumin-binding proteins) via caveolae (Figure 11). These 50-100 nm vesicles, first described by Simionescu et al. (1975), form from indentations of the plasmalemma and are coated with the caveolin-1 protein. Albumin, as the most abundant protein in plasma and interstitium, appears to facilitate NP endocytosis, as does lecithin, a phospholipid: even 240-nm polystyrene particles translocated across the alveolo-capillary barrier when coated with lecithin, whereas uncoated particles did not (Kato et al. 2003). The presence of both albumin and phospholipids in alveolar epithelial lining fluid may, therefore, be important constituents for facilitated epithelial cell uptake of NSPs after deposition in the alveolar space.

Rejman et al. (2004) reviewed a number of different endocytic pathways for internalization of a variety of substances, including phagocytosis, macropinocytosis, clathrin-mediated endocytosis, and caveolae-mediated endocytosis. They found in nonphagocytic cells in vitro that internalization via clathrin-coated pits prevailed for latex microspheres < 200 nm, whereas with increasing size up to 500 nm, caveolae became the predominant pathway. However, as shown in Table 4, surface coating of NSPs with albumin clearly causes even the smallest particles to be internalized via caveolae. The presence of caveolae on cells differs: they are abundant in lung capillaries and alveolar type l cells but not in brain capillaries (Gumbleton 2001). In the lung, during inspiratory expansion and expiratory contraction of the alveolar walls, caveolae with openings around 40 nm disappear and reappear, forming vesicles that are thought to function as transport pathways across the cells for macromolecules (Patton 1996). Knowledge from virology about cell entry of biologic NSPs (viruses) via clathrin-coated pits and caveolae mechanisms should also be considered (Smith and Helenius 2004) and can shed light on the mechanism by which engineered NPs may enter cells and interact with subcellular structures.

Evidence in humans for the translocation of inhaled NSPs into the blood circulation is ambiguous, with one study showing rapid appearance in the blood and significant accumulation of label in the liver of humans inhaling ⁹⁹Tc-labeled 20-nm carbon particles (Nemmar et al. 2002a), whereas another study using the same labeled particles reported no such accumulation (Brown et al. 2002). Taking into consideration all of the evidence from animal and human studies for alveolar translocation of NSPs, it is likely that this pathway also exists in humans; however, the extent of extrapulmonary translocation is highly dependent on particle surface characteristics/chemistry, in addition to particle size. Translocation to the blood circulation could provide a mechanism for a direct particle effect on the cardiovascular system as an explanation for epidemiologic findings of cardiovascular effects associated with inhaled ambient UFPs (Pekkanen et al. 2002; Wichmann et al. 2000) and for results of clinical studies showing vascular responses to inhaled elemental carbon UFPs (Pietropaoli et al. 2004). In addition to direct alveolar translocation of NSPs, cardiovascular effects may also be the corollary of a sequence of events starting with particle-induced alveolar inflammation initiating a systemic acute phase response with changes in blood coagulability and resulting in cardiovascular effects (Seaton et al. 1995).

Once NSPs have translocated to the blood circulation, they can be distributed throughout the body. The liver is the major distribution site via uptake by Kupffer cells, followed by the spleen as another organ of the reticuloendothelial system, although coating with polyethylene glycol (PEG) almost completely prevents hepatic and splenic localization so that other organs can be targeted (Akerman et al. 2002). Distribution to heart, kidney, and immune-modulating organs (spleen, bone marrow) has been reported. For example, several types of NPs, ranging from 10 to 240 nm, localized to a significant degree in bone marrow after intravenous injection into mice (Table 5). Such target specificity may be extremely valuable for drug delivery; for example, drug delivery to the CNS via blood-borne NPs requires NP surface modifications in order to facilitate translocation across the tight blood-brain barrier via specific receptors (e.g., apolipoprotein coating for LDL-receptor-mediated endocytosis in brain capillaries) (Kreuter 2001, 2004; Kreuter et al. 2002). Such highly desirable properties of NPs must be carefully weighed against potential adverse cellular responses of targeted NP drug delivery, and a rigorous toxicologic assessment is mandatory [Supplemental Material available online (http://ehp.niehs.nih.gov/members/2005/7339/supplemental.pdf)].

Neuronal uptake and translocation. A translocation pathway for solid particles in the respiratory tract involving neuronal axons is apparently specific for NSPs. Respective studies are summarized in Table 6. This pathway was described > 60 years ago, yet it has received little or no attention from toxicologists. This pathway, shown in Figure 9 for the nasal and tracheobronchial regions, comprises sensory nerve endings of the olfactory and the trigeminus nerves and an intricate network of sensory nerve endings in the tracheobronchial region. These early studies concerned a large series of studies with 30-nm polio virus intranasally instilled into chimpanzees and rhesus monkeys (Bodian and Howe 1941a, 1941b; Howe and Bodian 1940). Their studies revealed that the olfactory nerve and olfactory bulbs are, indeed, portals of entry to the CNS for intranasally instilled nanosized polio virus particles, which could subsequently be