Transport of Gold Nanoparticles by Vascular Endothelium from Different Human Tissues

The selective entry of nanoparticles into target tissues is the key factor which determines their tissue distribution. Entry is primarily controlled by microvascular endothelial cells, which have tissue-specific properties. This study investigated the cellular properties involved in selective transport of gold nanoparticles (<5 nm) coated with PEG-amine/galactose in two different human vascular endothelia. Kidney endothelium (ciGENC) showed higher uptake of these nanoparticles than brain endothelium (hCMEC/D3), reflecting their biodistribution in vivo. Nanoparticle uptake and subcellular localisation was quantified by transmission electron microscopy. The rate of internalisation was approximately 4x higher in kidney endothelium than brain endothelium. Vesicular endocytosis was approximately 4x greater than cytosolic uptake in both cell types, and endocytosis was blocked by metabolic inhibition, whereas cytosolic uptake was energy-independent. The cellular basis for the different rates of internalisation was investigated. Morphologically, both endothelia had similar profiles of vesicles and cell volumes. However, the rate of endocytosis was higher in kidney endothelium. Moreover, the glycocalyces of the endothelia differed, as determined by lectin-binding, and partial removal of the glycocalyx reduced nanoparticle uptake by kidney endothelium, but not brain endothelium. This study identifies tissue-specific properties of vascular endothelium that affects their interaction with nanoparticles and rate of transport.


Introduction
Nanoparticles hold great potential in biomedicine; for diagnosis or as carriers of therapeutic agents to different tissues. However, a central problem is how the nanoparticles can be selectively delivered to the target tissue. Nanoparticles in the blood stream first interact with vascular endothelium before they may cross or pass the endothelial cells and enter the tissue. Vascular endothelium in different tissues has distinctive properties including its glycocalyx, surface receptors, intercellular junctions or rate of production of transport vesicles. These distinctive properties provide an opportunity to selectively target nanoparticles. membrane of the endothelium are negatively charged, it has been proposed that cationic nanoparticles may be favoured in cell transport due to electrostatic interactions [27]. Our results confirm that the cationic PEG-amine/galactose nanoparticles are taken up much more efficiently than uncharged galactose-coated nanoparticles by the endothelium.

Mechanisms of nanoparticle internalisation by endothelium
First, we investigated the mechanism by which PEG-amine/galactose nanoparticles are transported into the cells. Nanoparticles in different cellular compartments of brain endothelium were counted and it was noted that a high proportion localised in vesicles (Fig 2A). This implies that these nanoparticles primarily enter the cells by vesicular transport, an energydependent process. In order to confirm that transport into the brain endothelium was energydependent, the cells were incubated at 4°C (Fig 2B) or in the presence of sodium azide/2-deoxyglucose ( Fig 2C). In both cases the numbers of vesicle-associated nanoparticles was significantly reduced. We considered the possibility that these inhibitors of cell metabolism could affect cell viability. To test this theory, cells were stained with propidium iodide, which only penetrates cells with damaged membranes: 95% of the cells were not stained (Fig 2D), hence we concluded that within the time-frame of the experiment, the cell viability was not affected.  In addition to nanoparticles that were localised in vesicles, some nanoparticles were present in the cytosol (Fig 2A). Immediate nanoparticle entry into the cytosol is thought to be energyindependent. It has been reported previously that gold nanoparticles (~2 nm) can enter the cytosol of cells by passively crossing the plasma membrane [28][29][30][31]. We have also previously reported that glucose-coated gold nanoparticles cross brain endothelium in vitro via the cytosolic route [17,32]. In accordance with these theories, cytosolic localisation of PEG-amine/ galactose nanoparticles was not affected by sodium azide/2-deoxyglucose treatment ( Fig 2C). On the other hand, incubation of the cells at 4°C reduced the number of cytosolic nanoparticles by about 50% (S1 Fig). This reduction is significant but smaller than the effect of low temperature on vesicular transport (Fig 2C). Passive uptake may be affected by membrane fluidity which is highly dependent on temperature-a less fluid membrane could decrease nanoparticle movement across the plasma membranes. Therefore, a change in membrane fluidity appears to be the most likely explanation for the reduction in nanoparticles localised in the cytosol at 4°C.
The mechanism of transport in vitro is correlated with observation of nanoparticle transcytosis in vivo (Fig 3). Following intracarotid infusion, nanoparticles are seen in brain microvascular endothelium, primarily in the cytosol. Within 10 minutes of infusion, the nanoparticles are also detected in cells of the brain parenchyma. Previous work has demonstrated that the nanoparticles move rapidly within the brain at speeds up to 1μm per minute.
To further analyse the mechanism of transport, brain endothelial cells were treated with inhibitors during the nanoparticle transport assay. In preliminary experiments nystatin, chlorpromazine, cytochalasin D and nocodazole were screened for their ability to inhibit the uptake of 70kDa FITC-dextran by brain endothelium. The doses of inhibitors were equivalent or  greater than those previously reported [15,17] to inhibit endocytosis in non-brain endothelium, however, only nystatin and chlorpromazine were found to significantly inhibit FITC-dextran internalisation in the brain endothelium at the doses used, and in these conditions the viability of the hCMEC/D3 cells was slightly reduced (S1 Table). Nystatin inhibits caveolar uptake and chlorpromazine inhibits uptake into clathrin-coated vesicles. The results in the nanoparticle endocytosis-assay showed a reduction in vesicular uptake of up to 50% by both inhibitors, however, the difference was not significant due to high variability in cellular uptake (S2 Fig). Treatment with chlorpromazine significantly increased the number of nanoparticles in the cytoplasm and nucleus. However, we noted that the shape of the nucleus was irregular on treated cells. It is possible therefore that chlorpromazine is affecting the permeability of the plasma and nuclear membranes as well as cytoplasmic transport systems.  Comparison of uptake by brain and kidney endothelium We then investigated the rate of nanoparticle uptake in different endothelial cells. For this purpose we selected two human endothelial cell lines, from brain and kidney. The choice of cells was based on observations in vivo which showed that nanoparticles accumulated at highest levels in the kidney, and lowest in brain 10 minutes after intravascular infusion (S3 Fig). Initially, we hypothesised that the higher rate of accumulation in the kidney was related to structural differences such as fenestrations in kidney endothelium and tight junctions in brain endothelium. Alternatively, the difference could be due to surface properties of the endothelium or their rates of transcytosis. The brain endothelial cell line hCMEC/D3 and the kidney glomerular endothelial line ciGENC are well characterised cells [33,34]. Both were immortalised by a similar process, involving transfection with SV40 large-T and both retain many of their tissue-specific properties. When we compared the uptake of PEG-amine/galactose nanoparticles in both endothelial cells, kidney endothelium showed higher nanoparticle uptake than brain endothelium, both into vesicles and cytosol (Figs 4 and 5). Both endothelial cell types showed approximately 4 times more nanoparticles in vesicles than in the cytosol. In both brain and kidney endothelium, there is an accumulation of nanoparticles on the transwell membrane, beneath the basal plasma membrane of the cells. These results imply that the nanoparticles have crossed the endothelium by cytosolic or vesicular transcytosis, and have been retained on the negatively charged filter.

The role of the endothelial glycocalyx in nanoparticle internalisation
To explain the higher uptake of the nanoparticles by kidney endothelium we considered 3 possible mechanisms: (a) the surface properties of the endothelium are different so that the initial binding of the nanoparticles to the cells varies; (b) the rate of endocytosis is comparatively low in the brain endothelium; (c) morphological features of the cell lines, such as vesicle numbers or size of the cells may differ between the endothelia.
The initial interaction of a nanoparticle with the endothelium may occur at the glycocalyx. The glycocalyx is a brush-like structure that is located on the apical side of vascular endothelial cells. Its size can range from 0.5 microns to 11 microns deep [35][36][37][38] and it contains glycoproteins and proteoglycans which consist of glycosaminoglycans attached to core polypeptides. The glycocalyx is thought to be more fully developed in vivo than in vitro [39] and it is strongly negatively-charged, particularly on brain endothelium [40]. To investigate the role of the glycocalyx, we carried out a lectin-binding assay on the kidney and brain endothelium. Lectins bind to specific carbohydrate residues on the glycocalyx ( Table 2) and they can be used to identify changes in glycocalyx composition [41]. They can also identify specialised domains on the plasma membrane [42].
The lectin Ulex Europaeus agglutin (UEA) was used as a standard marker for endothelium [43]. The other three lectins were found to bind human endothelial cells in an initial screening of a panel of 23 lectins, of which 15 bound strongly to a panel of human endothelia (S2 Table). Wheat germ agglutinin (WGA) recognises N-acetyl glucosamine, including terminal sialic acid residues, whereas peanut agglutinin (PNA) and Wisteria floribunda lectin (WFL) primarily recognise internal residues. A schematic illustration of the proposed arrangement of the major lectin binding site, based on data in this study is shown in supplementary material (S4 Fig).
Wheat germ agglutinin and peanut agglutinin have been shown to bind to microdomains of endothelium which are associated with transport vesicles ('plasmalemmal vesicles') [42]. We found that kidney endothelium exhibited a stronger binding profile to all lectins in comparison with brain endothelium (Fig 6A). This implies that the glycocalyx does indeed differ between the two cell types, and suggests that the kidney endothelium has higher numbers of domains associated with endocytosis.  To further analyse the role of the glycocalyx, enzymatic digestion was used to selectively remove carbohydrate residues. For partial removal of the glycocalyx we used neuraminidase and O-sialoglycoprotein endopeptidase (OSGEP). Neuraminidase removes sialic acid residues which contribute to the net negative charge of the glycocalyx, whereas the endopeptidase removes core polypeptides that have bound carbohydrate units. The lectin binding was changed by these enzymes (S5 Fig) confirming that elements of the glycocalyx, including the terminal sialic acid residues recognised by WGA, were removed. Chondroitinase and heparinase were also tested but they did not consistently affect lectin binding (data not shown). Neither neuraminidase nor endopeptidase treatment caused any change in the uptake of PEGamine/galactose nanoparticles either into the cytosol or vesicles of brain endothelium ( Fig 6B). However, neuraminidase produced a small reduction and endopeptidase a significant reduction in nanoparticle uptake by the kidney endothelium ( Fig 6C).
Initially, we hypothesised that removal of the negatively charged sialic acid residues by neuraminidase would increase nanoparticle uptake, by allowing the cationic nanoparticles to reach the plasma membrane more rapidly and thus enhance endocytosis. Therefore, the reduction in nanoparticle uptake following endopeptidase treatment was contrary to this hypothesis. One possible explanation is that the endopeptidase treatment disrupts microdomains (i.e. membrane subregions) which form the endocytotic vesicles, thus reducing endocytosis. The reduced nanoparticle uptake is only seen with kidney endothelium which has both higher levels of endocytosis as well as higher levels of lectin-binding sites than brain endothelium (Figs 6 and 7A).
The effect of the endothelial glycocalyx on nanoparticle uptake has not been investigated previously. The glycocalyx of erythrocytes has been shown to be crucial to the attachment of amphiphilic gold nanoparticles (around 2 nm core) to the cell membranes [44]. In this study, neuraminidase had a large effect on the nanoparticle attachment that was only slightly increased by treating with a combination of enzymes [44]. It is likely that fundamental  differences in levels of endocytosis by erythrocytes and endothelium as well as the differences in the glycocalyx and use of different nanoparticles can explain why the endothelium takes up these nanoparticles so effectively by comparison with erythrocytes. It has been previously established that there is a 20 nm gap between each proteoglycan chain on the endothelial glycocalyx [45], which may allow smaller nanoparticles to freely pass through to the membrane surface.
Differences in the rate of endocytosis, affecting nanoparticle uptake We next investigated whether the different transport rates of the endothelia could be primarily due to differences in their overall rate of endocytosis, rather than the rate of nanoparticle binding. The rate of endocytosis can be investigated by loading cells with FITC-dextrans of different size. We selected FITC-dextran of 10 kDa and 70 kDa size, which correspond to a diameter of 4.4 nm and 12 nm, respectively [46]. Kidney endothelium showed higher fluorescence intensity when treated with both FITC-dextrans (Fig 7A) in comparison with brain endothelium. This may mean that the kidney endothelium has larger vesicles or has more vesicles because it is a larger cell. Firstly, we investigated vesicular radius and distribution of vesicles. However, we found that the mean vesicular radius was no different (Fig 7B) and both endothelia had a similar profile of vesicle sizes (S6 Fig). Secondly, as brain and kidney endothelial cells differ morphologically, we also investigated the cell volumes, but again we found that they were not significantly different (S7 Fig). Thus we concluded that the higher fluorescence intensity of FITC-dextran is due to a higher rate of endocytosis by kidney endothelium.
Taken together, these results show that the high rate of nanoparticle uptake by kidney endothelium may be explained by its higher rate of endocytosis. This may be correlated with its high level of surface lectin-binding domains, which are associated with endocytotic vesicles. In vivo endothelial cells show even more marked tissue-specific characteristics. Specifically, brain endothelium in vivo has very low numbers of transport vesicles (caveolae) by comparison with other endothelia. While other factors may affect transport into tissues, it is nonetheless the endothelium which controls the initial uptake of nanoparticles to each tissue and hence the endothelium is likely to be the critical cell type in determining the rate of nanoparticle entry into different tissues. This is particularly true for smaller nanoparticles which are not cleared by the mononuclear phagocyte system.
In summary, we have shown that galactose/PEG-amine-gold nanoparticles are actively taken up by brain and kidney endothelium primarily into vesicles. Smaller numbers enter the cells passively via the cytosolic route, crossing the plasma membrane. The rate of uptake was higher into kidney endothelial cells in comparison with brain endothelial cells. This is best explained by the higher overall rate of endocytosis by the kidney endothelium. The glycocalyx was also different between brain and kidney endothelium. Hence the possibility that the endothelial glycocalyx restricts nanoparticle access to the cell surface was investigated. However, partial removal of the glycocalyx actually reduced the rate of nanoparticle uptake by the kidney endothelium although it had no effect on nanoparticle uptake by brain endothelium. Thus, the glycocalyx does not appear to act as a barrier to these nanoparticles. Indeed disrupting the surface proteoglycans of kidney endothelium unexpectedly reduced nanoparticle internalisation. Consequently, it appears that the rate of endocytosis, rather than cell surface binding, is the main factor accounting for the different rates of nanoparticle internalisation by the different endothelia.

Conclusions
Most research into the interactions of nanoparticles with cells has been concerned with how properties of the nanoparticles affect the cell. This study demonstrates how properties of the cell can influence nanoparticle uptake. We based our study on the observation of selective uptake of gold nanoparticles by human kidney endothelium in comparison with brain endothelium, since endothelial cells primarily determine the rate of entry of nanoparticles into different tissues. We concluded that gold nanoparticle uptake is correlated with the cell's potential for active endocytosis. In addition, the surface glycocalyx affected active uptake of nanoparticles, but only for kidney endothelium. These findings in vitro reflect the tissue distribution of these small gold nanoparticles in vivo. Our work highlights the need to carry out detailed analysis of cell behaviour, in order to assess the potential use of nanoparticles as cell-selective or tissue-targeted carriers. Detailed analysis of both endocytosis and the interactions between nanoparticles and cell surface proteoglycans and glycoproteins may improve the design of such carriers. This information can help predict nanoparticle behaviour in the body.

Nanoparticle synthesis, characterisation and toxicity assessment
Gold nanoparticles were synthesized essentially as described previously [15]. Briefly, the organic ligand coating for this gold nanoparticle was comprised of two ligands: 1-thiohexaethylene glycol-17-ammonium acetate (Galchimia, Spain) and 2-thioethyl α-D-galactopyranoside as the disulphide (Galchimia, Spain). A 50:50 ligand starting ratio was used at the time of reduction of HAuCl 4 with NaBH 4 in 90% methanol and 10% water, with a 3-fold molar excess of total ligands to gold. After 2 hr shaking the nanoparticles were purified from unbound ligands and reactants by washing them 4x with excess water using a Millipore 10kDa ultra-centrifugation device. The nanoparticles were resuspended in water acidified to pH 6 with acetic acid and stored at 4°C prior to use.
The nanoparticle size was analysed using a transmission electron microscope (TEM). The nanoparticle solution was air-dried on a copper mesh grid slot and viewed on a JEM 1400 (JEOL, Japan) at an acceleration voltage of 80 kV and magnification of 80,000X. The images were analysed using ImageJ. The characteristic atomic mesh of the nanoparticles was observed on a high resolution TEM JEM 2100 (JEOL, Japan) at an acceleration voltage of 150 kV and magnification of 150,000X. The absorption spectrum of these nanoparticles was determined using a SPECTROstar Nano spectrophotometer (BMG Labtech, Germany).
The toxicity of PEG-amine/galactose gold nanoparticles was assessed on cells grown in 96-well plates. The nanoparticles were incubated with brain endothelial cells (hCMEC/D3) for 48 hrs at concentrations of up to 75 μg/ml at 37°C, with 4 replicates per treatment. The cells were then washed 2x with HBSS (Hank's balanced salt solution, Sigma) and MTT [3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] was added to the cultures for 3.5 hrs at 37°C. The solution was aspirated and 100 μl of DMSO (dimethylsulfoxide) was added to each well. The plate was then shaken for 10 min and the absorbance was read at 540 nm on a plate reader OPTIMA FluoStar (S8 Fig).
The conditionally immortalized glomerular endothelial cells (ciGENC) were supplied by Simon Satchell [33]. They were propagated in EBM-2 MV supplemented according to the manufacturer's protocol, with the exception of growth factor VEGF, which was omitted from the medium. The medium was changed every 2-3 days. The propagating cultures were grown at 33°C. For experiments the cultures were kept at 33°C until they were 90% confluent, then they were kept at 37°C for 3-5 days in order to develop kidney glomerular endothelial phenotype before they were used. In each case, the cells were kept in a humidified atmosphere with 5% CO 2 . In all assays the different endothelial cells were cultured in identical conditions. Cell viability following antibiotic treatment was measured by trypan blue exclusion. The hCMEC/D3 cells (10 5 ) were stained with 0.04% trypan blue in PBS and the percentage of stained (non-viable) cells counted. The background percentage of non-viable cells in untreated cultures was subtracted from the values stated.
To assess the active uptake involvement in transport of PEG-amine/galactose gold nanoparticles, the hCMEC/D3 cells were cultured for 3 hrs at 4°C or for 2 hrs with inhibitors of active metabolism NaN 3 (10 mM) + 2-deoxyglucose (50 mM) and processed for TEM (see below). The possible cytotoxic effect of NaN 3 + 2-deoxyglucose on cells was assessed by staining cells with propidium iodide after 2 or 4 hr treatment. Digitonin (30 μg/ml for 15 min) was used as a positive control for cell death, and fluorescence intensity was measured on a Facscan, channel FL2 set at 400V gain: 10,000 events were analysed from the gated cell population.

Transmission electron microscopy to quantify the localisation and number of nanoparticles in cells
The rate of transport of PEG-amine/galactose gold nanoparticles into and across cells was assessed by quantification of the number of nanoparticles in cells by transmission electron microscopy (TEM). The cells were grown on transwell inserts (1 cm 2 , Corning Costar), the nanoparticles (8 μg/ml) were applied to the top chamber in EBM-2 MV medium supplemented with 2.5% FBS. The time of incubation was 2-3 hrs. After the incubation, both chambers were washed 3x in HBSS. The cells were fixed in 2.5% glutaraldehyde for 1 hr at room temperature. The fixative was removed, both chambers washed 3x in PBS (phosphate buffer saline) and stored in phosphate buffer at 4°C.
The samples were processed for TEM (all incubations were performed at room temperature and all solutions were applied to both the insert and the bottom chamber). Firstly, the cells were permeabilized in 0.01% Triton x100 for 15 minutes on a rocker. The inserts were washed 3x in PB (phosphate buffer) and silver enhanced. Silver enhancement (Aurion) was prepared according to the manufacturer's protocol and applied to the insert for 45 min. 3 washes in distilled water followed. The cells were then treated with 1% osmium tetroxide for 30 min. The insert was washed 3x in PB and removed from the well; the membrane was cut in 2 pieces of about 2 x 3 mm each. The membranes were then gradually dehydrated in ethanol: 30% and 50% for 5 min, 70% and 90% for 10 min, 100% for 10 min twice, 100% with molecular sieve for 10 min. The membranes were incubated in a 50:50 mixture of 100% ethanol and Epon resin overnight and then in fresh resin twice for 2 hrs. The embedding took place at 60°C for 48 hrs. Resin blocks were microsectioned to 80 nm sections using a Diamond knife (Diatome, Switzerland). The sections were collected onto pioloform film-covered copper grids. They were stained in 1% uranyl acetate (30 min) followed by lead citrate (10 min). Sections were viewed on a JEM 1010 (Jeol, Japan) at an acceleration voltage of 80 kV. The cells were observed at magnification of 20,000x which allowed sufficient magnification to view the nanoparticles. The nanoparticles were counted and sorted into categories "vesicles" or "cytosol" during viewing on the electron microscope. The length of insert which was viewed (usually around 1000 microns in length) was measured with low magnification of 150X and the number of nanoparticles was recalculated to obtain number of nanoparticles per micron of insert.

Characterisation and modulation of the endothelial glycocalyx
In order to characterise the endothelial glycocalyx, cells were stained using biotinylated lectins. hCMEC/D3 cells or ciGENC cells were cultured in a 96 well plate for 3 and 4 days respectively. The cells were washed twice in HBSS, fixed in 0.1% glutaraldehyde (in PBS) for 15 min. They were washed twice in PBS, blocked in 0.05M Tris/HCl for 20 min and washed 3x in wash buffer (0.05% Tween20 in PBS). Lectins were added for 1 hr at concentrations: Ulex Europaeus agglutinin 10 μg/ml, Wheat germ agglutinin 5 μg/ml, Peanut agglutinin 20 μg/ml, Wisteria floribunda lectin 20 μg/ml; all prepared in diluent (5 mg/ml BSA and 0.01% Tween20 in PBS). After the incubation, the cells were washed in wash buffer 3x. Streptavidin peroxidase (1:700 in 5 mg/ml BSA and 0.01% Tween20 in PBS) was added to the cells for 1hr, then washed in wash buffer 3X and once in PBS. The chromogen solution containing 100 μg/ml tetramethylbenzidine in 0.1M sodium acetate/citric acid was prepared and hydrogen peroxide added to a final concentration of 1/3000. This solution was added to the cells for 10-15 min resulting in blue colour; the reaction was stopped by adding 20 μl of 30% H 2 SO 4 per well. The absorbance was read at 450 nm on a plate reader OPTIMA FluoSTAR. Data are expressed as mean absorbance from replicate wells, of a representative experiment, which had been repeated 3 times.
Endothelial cells were also treated with enzymes for 2 hrs, to remove parts of the glycocalyx. The enzymes used were 50 mU/ml neuraminidase or 25 μg/ml endopeptidase. After the treatment, the cells were washed twice in HBSS and either stained with lectins or treated with nanoparticles (as described above).

Analysis of rate of endocytosis
The cells were treated with FITC-dextran of 70 kDa or 10 kDa in order to test the rate of endocytosis. The cells were grown in 12-well plates and FITC-dextran was applied at 0.2 mg/ml for 1 hr. The cells were washed with ice-cold HBSS 5x over 5 min; they were trypsinized, collected and washed 2x in HBSS + 0.1 mg/ml BSA. Then they were resuspended in PBS and analysed using flow cytometry (BD FacScan, channel FL1, 460#x2014;500V: 10,000 events analysed from the gated cell population). Data are expressed as the mean of the median fluorescence values from 3 independent experiments.

Analysis of vesicular diameter and cell volume
TEM sections of hCMEC/D3 and ciGENC cells were used to analyse vesicular diameter and cell volume. Each section corresponded to a well with a single treatment. As each treatment had 3 wells, one was randomly picked for this analysis and 3 independent experiments were analysed per cell type, resulting in 3 sections analysed altogether. At 20,000x magnification on JEM 1010 (Jeol, Japan), about 100 vesicular diameters were measured per section in cells that were randomly picked. The data set was analysed with Microsoft Office Excel 2010 and Graph-Pad Prism. Similarly, to obtain volumes of brain cells in comparison to kidney cells, 3 independent experiments were analysed for each cell type. At magnification of 2,000X on JEM 1010 (Jeol, Japan) 10 images per treatment were taken randomly. The cell area was measured using ImageJ. The data were analysed as a distribution histogram and means of each independent experiment were calculated and analysed by GraphPad Prism.

In vivo study of nanoparticle tissue-distribution
For the animal study, 4 adult male Wistar albino rats weighing 250-300 g were used (Vivarium from Bogazici University, Istanbul, Turkey) for testing nanoparticle distribution.
This work was approved by the Local Ethics Committee for Animal Experimentation of Istanbul University (2014/06).
The animals were anaesthetized with chloral hydrate (360 mg/kg, i.p.) and locally anaesthetized with bupivacaine hydrochloride. After the incision in the neck, the left common internal and external carotid arteries were exposed. The external carotid was ligated and the artery was retrogradely catheterized. Each animal was infused with warm 50 μg (100 μl) of nanoparticle solution (prepared in water) through the exposed external carotid artery over 2 min using an infusion pump (Harvard Apparatus, Infusion/Withdrawal pump, Model No: 901A). The nanoparticles were allowed to circulate for 10 min and the animals were administered with sodium pentothal (50 mg/kg, i.p.) and chloral hydrate (100 mg/kg, i.p.) before perfusion/fixation procedure. The animals were perfused transcardially with 0.9% saline (75 ml), and fixed with 200 ml of 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M PB for 15 min. The brain, liver, kidney and lung were collected and stored for the analysis.
To analyse the amount of gold in the organs, ICP-MS (inductively-coupled plasma massed spectrometry) was used. Boiling HNO 3 was used to digest brain, liver, kidney and lung for 5-10 min. After diluting the samples to 5 ml volume, they were run in 1% HCl with Iridium standard (10 ng/ml) on NexION ICP-MS (PerkinElmer, USA). The ICP-MS produced data on the amount of gold per gram of tissue. In order to account for the differences in individual size (body weight) of the animals, these values were compensated according to the weight of each animal, resulting in relative levels of gold in each tissue.