Papers dedicated to the use of gold nanoparticles (GNPs) in various fields of nanobiotechnology are numerous and cover biosensors , genomics , and the visualization  and photothermolysis  of cancer cells and other fields. These applications are based on combining two principles: (1) surface functional ization for providing the colloidal stability and bio compatibility of particles, the molecular recognition of conjugates, efficient endocytosis, etc., and (2) the excitation of plasmon resonances in the visible or near IR region in order to obtain unique optical properties . Along with the abovelisted applications that have been already tested, GNPs have been recently actively used in various fields of nanomedicine for diagnostic and therapeutic purposes [6–8]. Note that GNPs are ever more frequently administered to humans and animals in a parenteral manner. In particular, they are used as carriers for delivering drugs [9, 10], genetic material [11, 12], antigens [13, 14], and as a drug itself or diagnostic tool in the therapy of tumors [15–17] or rheumatoid arthritis [18, 19]. Two preparations, AurImmune TM and AuroLase TM, have already passedclinical trials [20, 21]. Comprehensive data on the clinical trials of the Aurasol® preparation in tablets intended for treating severe rheumatoid arthritis cases have been also published . However, burning questions on GNP biodistribution, circulation in the bloodstream, pharmacokinetics, and elimination from the body, as well as their possible toxicity throughout the whole body or cytotoxicity and genotoxicity, arose almost concurrently with their medical application. Concerns about the potential consequences of nanoparticle administration are far from groundless. For example, thorium dioxide particles 3–10 nm in size were widely used in radiography as a contrasting agent (thorotrast) in the 1930s–1950s. However, it was found [23, 24] later that these particles could accumulate and remain in the body for decades, causing adverse radiation effects. Some recent papers have reported on the toxicity of ultras mall GNPs (about 1.5 nm) and considerable GNP accumulation in the liver and spleen of animals on the background of their extremely slow elimination (see below). In the case of gold nanorods (GNRs), molecules of cetyltrimethylammonium bromide (CTAB) are the initial stabilizer [3, 6]; however, this compound is a known toxic surfactant in a free state. “Colloidal Metallic Gold Is Not BioInert” was the title of a paper  which should underline the importance of a nanometerscale size in biological effects even for an inert material such as gold. Our screening of the published data has demon strated that an explosion in the research activity on GNP biodistribution and toxicity has taken place over the last 3–4 years (Fig. 1). Since many research teams started their projects independently, experimental designs were very diverse, including the particle size and shape, functionalization methods, types of animals, doses, particle administration routes, and so on.
Correspondingly, the resulting data and conclusions on the biodistribution levels and kinetics and toxicity estimates were also diverse. Therefore, despite the fact that several reviews on the toxicity of nanomaterials have been published [25–37], there is an urgent need to systematize the new data published so far to assist researchers in estimating results and planning further experiments. This is the goal of our review, which presents original data on the size dependence of GNP toxicity along with alreadypublished data.
BIODISTRIBUTION OF NANOPARTICLES
Table 1 shows published data on GNP biodistribu tion in 1995–2010. Some general remarks are appropriate before discussing these data. First, we do not actually pretend to cover all the published papers. In particular, our table includes only GNPs and does not cover any other metals. Second, we tried to include mainly data on biodistribution after intravenous administration, although some of these papers use other administration routes. Some papers on biodistribution also involved toxicity assessments; therefore, the corresponding references can include both aspects. Finally, this table includes only parts of information that we regarded as important for comparative estimates. The type and size of particles; methods of surface functionalization; animal model (with special indication of transplanted tumors); administration doses and routes; examined tissues, organs, or cells; recording time of data on pharmacokinetics and clearance; and methods for assaying gold concentration in samples were selected as the main parameters. Our choice slightly differs from the table on published data from . Note that the data on doses are given in terms of gold concentration (μ g) administered per unit average animal weight (g) and were frequently obtained by recalculating the data on the volume of administered uspension, concentration, particle size, etc. Therefore, taking into account the variation in actual animal weight, the listed data should be regarded as approximate values that give an idea on the order of magnitude of the administered doses with an accuracy of about 30%.
PARTICLES, MODELS, AND METHODS
Figure 2 shows the main stages in modern in vivo experiments on assessing GNP biodistribution. These experiments comprise (1) the synthesis and characterization of nanoparticles with specified geometric and structural parameters; (2) the surface functionalization of nanoparticles with biocompatible ligands providing for the required properties of the bioconjugate; (3) the design of experiments in animal models, sampling, and characterization of distribution in organs, including an analysis of the accumulation and clearance kinetics; and (4) the identification and localization of particles at the cellular level, as well as their elemental and structural analyses. This scheme also shows the main parameters that characterize the stages and methods for determining these parameters. The main object for which the largest amount of data has been obtained (25 papers listed in Table 1) is naturally colloidal gold particles. The range of studied particles covers more than two orders of magnitude from the minimal size of 1.4 nm (atomic cluster Au 55)  to the maximal size of 250 nm . In addition to these particles, there are data on the biodistribution of GNRs [45, 52, 55, 69], SiO 2 /Au [49, 51, 67] and Au/Au 2 S  nanoshells, hollow nanoshells , and composites of gold nanoparticles with diameters of 5–25 nm in various nanoscale matrices [47, 62]. There are several proven methods for characterizing and standardizing the properties of nanoparticles, including transmission electron microscopy (TEM), dynamic light scattering (DLS), and optical spectros copy . TEM images are random sets of a cut offsample on a grid; therefore, two types of problems are always encountered. The first is changes in a sample during its processing and the second is providing unbiased statistics of image sets. Fortunately, the first problem doesn’t exist for GNPs if one’s interest is focused only on the size distribution of individual particles regardless of their possible aggregation within a colloid. Note that the sample should be prepared for TEM analysis according to conventional protocols (washed, suspended, etc.) to prevent any artifacts during its drying on a TEM grid (continued growth and soon). The second problem is resolvable by common statistical tools and a careful examination of all grid sites (which can be rather laborious). Experience shows that a regular rendering of 1000 particles from one site of a grid does not statistically differ from their rendering in any other randomly selected site. Assuming the abovelisted conditions, a TEM assay gives perfectly reproducible results and is now a golden standard accepted in any laboratory without exclusion around the world (according to our database, which comprises over 2000 sources from 1995 to 2010)
One doubtless advantage of DLS and optical spectroscopy is the possibility of studying the dispersion composition in situ with an averaging of the data on a large ensemble of particles within a probed volume. For example, for a typical coherence volume in a DLS assay amounting to 10 –3 cm 3, the number of analyzed particles in a typical gold colloid with an optical density of unity in a 1 cm cuvette at a resonance wave length of about 520 nm will be approximately 10 8 –10 9. It is impossible to examine such an ensemble by TEM. Second, DLS can be also used for assessing another very important parameter of particles, the zeta potential, ζ, proportional to their charge. In particular, this useful option is included in a Zetasizer Nano ZS (Malvern, United Kingdom) device.
The most popular reagent for surface functionalization (approximately 30% of all publications) is polyethylene glycol (PEG) with thiol groups (PEGSH), which forms a stable donor–acceptor bond with the surface atoms of gold particles. In several works, PEG molecules were used as linkers for further labeling with a radioactive chelator  or probe molecules for certain targets (for example, the tumor necrosis factor [9, 63]). In individual works they also used albumin, maltodex trin, polyethylene oxide, and other substances listed in Table 1. GNRs are initially produced with their surface already functionalized with CTAB molecules. To make them biocompatible, CTAB molecules are replaced with PEG molecules [45, 52] or other nontoxic ligands .
The most frequently used animal model is mice (in twothirds of all publications) and rats (six papers), while other models (pigs or humans) as well as human eyes or placenta (in vitro experiment ) were studied in only single works. Intravenous administration was reported in 26 publications and only very few works examined biodistribution after administrating particles via the trachea or digestive tract. Unfortunately, the GNP dose is among the most variable parameters in different studies. As is evident from Table 1, the values range from several hundredths of a microgram per gram of animal weight [56, 70] to 2700 μ g/g . Certainly, the last case is an exception,since the corresponding work deals with the possibilities of contrasting Xray images using nanoparticles with a diameter of 1.9 nm rather than biodistribution within a range of reasonable physiological doses. Nonetheless, despite a tremendous range of doses (over five orders of magnitude), it is evident from Fig.3 that most of the works utilized a range of 0.1–10 μ g/g, which can be regarded as the main range for comparing the data of different authors.
To assay the gold concentration in tissue or blood specimens, neutron activation analysis (INAA or NAA), mass spectrometry with inductively coupledplasma (ICPMS), and atomic adsorption spectrometry (AAS) have been used at approximately equal rates. The sensitivities of these three methods decrease in the order they are mentioned. The radioactive 111 In label was used [59, 68] to offer, according to the statement of the authors, the highest sensitivity among the known methods. In most cases, ICPMS yields a detection limit of about 0.001 μ g/kg , which is quite sufficient for an accurate estimate of biodistribution. The last remark in this section is related to a comparison of the biodistribution data. Unfortunately, all too often it is possible to make only a qualitative comparison for the following two reasons. First, with few exceptions, balance studies (i.e., taking into account all the routes for the distribution and elimination of particles) have not been performed. Second, the data have been presented either as a percent of the applied dose or in micrograms per gram of the organ specimen.
DEPENDENCE THAT BIODISTRIBUTION HAS ON GNP SIZE AND ADMINISTRATION ROUTE
The most important question is, how does the biodistribution kinetics depend on the nanoparticle size, dosage, and administration route? The first experiments on assessing the biodistribution of colloidal gold were performed in the 1970s–1980s using mouse  and rat [74, 75] models. It was found that parenterally administered colloid gold particles were captured by hepatocytes, secreted with bile, and excreted with feces. Presumably, work by Hardonk at al.  can be regarded as one of the first demonstrations of the size effect in an in vivo experiment with functionalized GNPs (using polyvinylpyrrolidone and BSA molecules). Namely, the maximal amount of 17 nm gold particles in hepatocytes was observed 2 h after injection, whereas 79nm particles were completely undetectable in hepatocytes. GNPs were detectable in feces for 4–12 days after administration; in addition, gold particles were identified mainly in the Kupffer cells after 6 days. Along with , Sadauskas et al.  demon strated the important role that Kupffer cells play in GNP clearance from the body in mice that intravenously received 2 and 40 nm GNPs. According to electron microscopy data, after injection, nanoparticles accumulated in liver macrophages (90%), whereas their amount in the spleen macrophages was considerably smaller (10%). Gold particles were undetectable by TEM in the other organs (kidneys, brain, lungs, adrenal glands, ovaries, and placenta); this agrees with the data of Katti et al. , who analyzed the biodistribution of radioactive colloid gold, 198 Au. The authors inferred that nanoparticles penetrated only into the phagocytic cells (first and foremost, Kupffer cells), failed to cross the placental and blood brain barriers (BBBs), and 2nm particles could be excreted with urine. Continuing their work, Sadauskas et al.  discovered that 40nm GNPs were localized to lysosomes (endosomes) and could stay there for up to 6 months. Now consider the data listed in Table 1. In 2001, Hillyer and Albrecht  made pioneering studies of the size dependence of biodistribution of the GNPs in the range of 4–58 nm administered to mice via the digestive tract as a supplement in drinking water at a concentration of 200 μ g/ml (Table 2). It was demon strated that the smallest (4nm) particles could enter via the digestive tract (being captured by enterocytes followed by their degradation in villi) and be distributed between nine studied organs. The effect of entering other organs via the digestive tract considerably decreased with an increase in the particle size (10, 28,and 58 nm), and the level of gold for 4 nm particles was maximal (about 0.075 μ g/g) in the kidneys and about 0.035–0.02 μ g/g in other organs (small intestine, lungs, stomach, spleen, and liver). The 4 nm par ticles were even detectable in the brain in a statistically significant (relative to the other sizes) amount (4.7 ng/g), which is likely to be the first experimental evidence that GNPs can pass through the BBB. Two years earlier, the same researchers published the first data  on the biodistribution of 13 nm particles after quadruple daily intraperitoneal injections into mice at a rather large dose of 20 μ g/g (our estimate for the dose for a mouse with an average weight of 28 g and 1 ml of tenfold concentrated injected suspension obtained according to Frens citrate protocol by the reduction of 0.01% HAuCl 4). The difference in the administration route and doses gave completely different distribution pattern when compared with the data of  (Table 2, last column). Most of the 13 nm particles accumulated in the liver and spleen at an amount of 1060–1400 μ g/g, i.e., a 50000 fold larger amount than in the experiment with administration via the digestive tract  in a drinking supplement. The observed concentrations in the stomach and small intestine were one order of magnitude lower (100–200 μ g/g) and in the kidneys, blood, heart and lungs, they were one more order of magnitude lower (5 – 20 μ g/ml). Note that the gold concentration in the brain (50 ng/g) exceeded the gold concentration in the experiment with 4nm particles by one order of magnitude and was 20fold for that with 10 nm particles.
Thus, both studies unambiguously confirmed that particles with a diameter of 10 nm or smaller could pass through the BBB. Now let us discuss the data on the GNP biodistribution after intravenous administration. Another important issue in these experiments is the kinetics of GNP circulation in the bloodstream. In one of the first experiments, Darien et al.  measured the distribution of 16nm particles stabilized with albumin after intravenous administration in pigs (with a weight of 20 kg and more) at a dose of about 10–20 μ g/g 5–6 h after injection. INAA detected the largest amount of particles (270 μ g/g) in the lungs followed by the liver (88 μ
g/g) and blood plasma (1 μ g/g). The size effect in the biodistribution 24 h after intravenous GNP administration in rats in a wide range of sizes (10, 50, 100, and 250 nm) was studied for the first time by De Jong et al.  using ICP MS (Table 3). It was demonstrated that nanoparticles of all sizes accumulated to the greatest degree in the liver and spleen in terms of both absolute values and percentage of dose, which on the average exceeded 30%. Such a high percent rate evidently indicates the effect of redistribution and the accumulation in these particular two organs during the circulation of particles in the bloodstream. An evident difference was observed between the accumulations of 10nm particles and particles of larger sizes. Only 10 nm particles were identified in the kidneys, testicles, thymus, heart, lungs, and brain in addition to the liver and spleen.
Administration route and doseWater ad libitum, 200 μ g/ml, 7 days Intraperitoneal administration, 4 × 20 μ g/g ConcentrationAfter 12 h, ng/gAfter 3 days, μ g/g Size, nm 4 102813 Blood 6.8 0.77 0.4715 Brain 4.0 2.10.700.5
Lungs 328.6 0.805 Heart 157.3 1.55.5 Kidneys 75 176.223 Spleen 207.0 141400 Liver 212.81.2 1060 Small intestine 39 14 215150 Stomach 22 2038229 An analogous experiment for GNPs with a diameter of 15, 50, 100, and 200 nm was performed by the team led by Makino  but using another model. The biodistribution in mouse organs was determined by ICP MS 24 h after the intravenous administration of the GNPs suspended in sodium alginate solution
Note that, as we see it, a dose of 1 g/kg = 1 mg/g (usedin ) at a volume of injected suspension of about 0.1 ml (3 ml/100 g ) is not physiological and difficult to implement from the technical standpoint. The mass concentration of suspension should be about 30 mg/0.1 ml = 300 mg/ml, which is approximately 5000fold higher than the standard concentration of sol obtained from 0.01% HAuCl 4 according to the Frens protocol with an optical density of ~1 in a 1 cm cuvette at a resonance wavelength of 520 nm. Comparing the data of  and  demonstrates a certain similarity in that a considerable accumulation of particles of all sizes is, on the average, observed in the liver. However, these two studies also display considerable distinctions. In particular, accumulation according to  was not as dominant as in the case of a rat model . In , 15 nm particles displayed a considerable accumulation in the kidneys and lungs and only the largest GNPs (200 nm) were prevalent in the spleen along with the liver. The third evident distinction in the data of  and  is in the GNP level in the blood. The GNP content in the blood for rats 24 h after injection was quite comparable with the contents in the main accumulating organs ; however, in the case of mice , the content in blood was one order of magnitude lower than in the accumulating organs. Finally, both studies demonstrated the ability of small 10–15nm particles to penetrate through the BBB; however, 50nm GNPs were undetectable in the rat brain  yet were observed in the mouse brain . First and foremost, these differences can be associated with a drastic distinction in the doses, animal models, and protocols for preparing the particles.Note that, according to , the colloids changed their color upon passing into phosphate buffer, which evidently suggests an aggregation of particles and a change in circulation pattern. It is clear that these data on biodistribution were inevitably distorted to a certain degree by aggregation.
The lower size limit in the papers considered above was still rather high (about 10 nm) except for , where they studied the contrast in X ray images with large doses of 1.9nm particles. In , the authors used INAA to study the biodistribution of the smallest GNPs (1.4 nm) in comparison with the commonly used colloidal 18nm GNPs (Table 5). The 1.4 nm GNPs were used as a complex of Au 55 with sulfonated triphenylphosphine ligand molecules. The selection of this object was determined by its ability to irreversibly bind to Bform DNA  and cause a toxic effect in experiments with cell cultures (see below).
In this case, the effect that size has on distribution is evident according to accumulation in the liver and kidneys. The main accumulation target for 1.4nm GNPs was the liver, whereas 18 nm GNPs circulated in the blood for a longer time and accumulated mainly in the liver, skin, and kidneys. A considerable amount of these particles was distributed throughout the animal body (in the carcass, according to ).
It is interesting to compare the data of  and  for particles of similar sizes (15 and 18 nm) and the same type of animals. A certain similarity is evident in the predominant accumulation in the liver; however, the ratios of accumulation in various organs display considerable differences. In particular, comparable accumulation levels in the liver, kidneys, and lungs according to  are not confirmed by the data obtained in . In addition, the penetration through the BBB was unobservable in  for both the smallest 1.4 nm particles and 18nm GNPs. One of the most comprehensive studies (in both the number of examined organs and duration of the time interval) of the biodistribution of unmodified 20nm GNPs was performed by Balasubramanian et al. . The authors found the gold concentration (using ICP MS) in 28 organs and the feces and urine of rats for 2 months after a single intravenous GNP administration at a very low dose of 0.01–0.015 μ g/g. In agree ment with most published data, it was shown that the nanoparticles predominantly accumulated in the liver (50–70 ng/g) and spleen (8–10 ng/g). Smaller amounts of gold were found in the kidneys (5 ng/g) and testicles (0.6 ng/g). Unlike the other works considered above, the lungs contained very low amounts of gold, and even trace GNP amounts were undetectable in the brain.
A comparison of this fact with the detection of GNPs smaller in diameter in the brain suggests that the GNP penetration through the BBB is critically sizedependent, with the upper boundary for penetration being about 20 nm. Sonavano et al.  postulated a mechanism underlying the size dependence of the BBB penetration. Almost 100% of the surface area of the capillary basal membrane is covered by thesynaptic nerve endings of astrocytes. These nerve endings are separated from the capillary endotheliumby a distance of about 20 nm. Consequently, in principle, GNPs of smaller sizes can pass through this clear ance. Another mechanism is connected with the deliv ery of doxorubicin to the brain using nanoparticles, as was studied by Petri et al. . It was assumed that thistransfer could involve interaction with apolipo protein AI, which attached to the nanoparticle surface via the SRBI receptor localized to the BBB. The apolipoprotein adsorption on GNPs can enhance the penetration of GNPs through the BBB. Although the need for further studies is evident, we would like to support the hypothesis on the critical 20nm size for overcoming the BBB with the data of a recent paper  on the possibilities of GNP penetration through the mouse hematoretinal barrier. Particles 100 nm in size were undetectable in retinal tissues, whereas 20 nm particles were found in almost all the retinal layers, including neurons (70%), endothelium (17%), and glial cells (8%).
We conclude this section with one general inference from all the examined data, namely, that the liver is undoubtedly the most vulnerable organ from the standpoint of accumulation and slow clearance kinetics (the spleen is vulnerable to a lesser degree). This specific feature of biodistribution can lead to acute inflammatory process in the liver .
DEPENDENCE OF BIODISTRIBUTION ON SURFACE FUNCTIONALIZATION
Undoubtedly, the functionalization of the GNP surface is one of the key things, along with the particle size, that determines the fate of the particles in the body of an animal after intravenous administration.
First and foremost, the unstabilized GNPs will aggregate at the moment of injection under conditions of elevated ion concentration, inevitably leading to differences in the GNP circulation kinetics in the bloodstream. Moreover, the use of specialized stabilizers with probe molecules can increase GNP accumulation in a target organ, for example, a tumor.
As the first example, compare the data of one of the first experiments with pigs  and a recent study  using the same model with particles of approximately the same size range (15–20 nm). The main difference in the experiment in  is in the dose (2 μ g/g), duration of kinetics observation (170 h), and the use of other surface stabilizers (gum arabic and maltose). It has been demonstrated in  that the particles coated with gum arabic and maltose displayed different accumulation patterns in the blood, tissues (liver and lungs), and urine; moreover, the difference reached 50% and higher. In particular, the GNPs coated with gum arabic predominantly accumulated in the liver, whereas particles coated with maltose accumulated in the lungs. Bergen et al.  compared the biodistribution kinetics for unstabilized particles with diameters of 50, 80, 100, and 150 nm with particles conjugated with various stabilizers (PEG and PEG + galactose), which provided a total positive (only for 50nm particles) or negative GNP charge (Fig. 4). Unfortunately, theexperiment on assessing the effect of a charge on 50nm GNP biodistribution was not described in this paper. The level of GNPs in the mouse blood 2 h after injection at a dose of about 0.5 μ g/ml was maximal for particles coated with PEG. In terms of the ratio of gold (ng) to protein (mg) amounts, the level of gold was about 5–15 ng/mg for the PEGcoated particles with sizes of 50, 80, and 100 nm and considerably smaller (about 0.25 ng/mg) for the 150nm particles. The adsorption of albumin and other plasma proteins on the particles partially stabilizes them, which explains the low yet detectable level of GNP in the blood (on average, it is approximately two orders of magnitude lower than for the GNPs protected with PEG mole cules). However, the level of gold in the blood for the GNPs conjugated with galactose + PEG complex was minimal, being three and more orders of magnitude lower than PEG without galactose The authors explained this unexpected result by the fact that the galactose molecules served as a target for liver hepatocytes. However, this conclusion is completely attributable only to conjugates of 50 nm particles (Fig. 4), because the ratio of the gold concentration in hepatocytes to the gold concentration in non parenchymatous cells (mainly Kupffer and sinus endothelial cells) for them was about 2.5. Moreover, only for these particles was the level of gold in hepato cytes 16 fold higher for the conjugate GNPs 50 + galactose–PEGSH than in the control, GNPs 50 + PEG SH. Thus, these results unambiguously demon strate in vivo the specific delivery of a complex of 50 nm particles with surfaces modified by galactose–PEG SH to hepatocytes (but not to the Kupffer cells). The role that surface PEG modifiers play in biodistribution and the increased accumulation in a trans planted tumor was studied by the example of 20, 40, and 80nm GNPs . The modifiers also contained a surface radioactive chelator with the label (111 In).
The authors demonstrated that the characteristic halftime of clearance of the particles from mouse blood at a dose of 0.04–4 μ g/g was less than 1 min for 80 nm GNPs, about 10 min for 40 nm, and about 30–40 min for 20 nm particles. Moreover, the 80 nm GNPs entered the liver and spleen as early as after 10 min, being undetectable in the blood, kidneys, bladder, and intestine, whereas the 20 nm particles circulated in the blood longer; accumulated in the liver and spleen to a lesser degree; and were detectable in the heart, kidneys, and intestine. In addition, only 20 nm particles accumulated in the tumor tissue, which is natu rally explainable by longer circulation and the effect of retention in the tumor with an elevated blood supply.
Either unmodified GNPs or their conjugates with BSA, PEG, or their derivatives were used in the papers considered above. Goel et al.  studied the biodistribution of 33nm GNPs coated with tumor necrosis factor α (TNF α) and PEGSH. The experiments were performed with healthy mice and mice with a transplanted solid tumor (Fig. 5). The authors demonstrated that the GNPs coated with PEG predominantly accumulated in the liver and spleen of healthy mice; note that high concentrations in the liver and spleen were recorded after 24 h and then remained almost constant for almost 3–4 months. A slow clearance kinetics was also observed in the kidneys, although at a considerably lower total average level (approximately five to tenfold). Trace amounts of gold were detectable in the lungs. The particles coated with TNF α also accumulated in the tumor. However, despite biospecific functionalization, the maximal accumulation level in the tumor (after 12 h) was almost one order of magnitude lower than in the liver and spleen and comparable only to the level in the kidneys and lungs. One of the most important observations in this work is that a considerable GNP content in the liver was recorded even 120 days (!) after injecting 125 μ g (6.25 μ g/kg) GNPs (Fig. 5b). Thus, both the rapid elimination of particles from the blood and longterm retention in the body are connected with the function of the hepatobiliary system.