Biological mechanisms of gold nanoparticle radiosensitization

ROS and oxidative stress

Gold is believed to be chemically inert. However, growing evidence suggests that their surface is electronically active, thus catalysing chemical reactions and promoting an increase in the production of ROS (Ionita et al. 2005; Mikami et al. 2012; Ionita et al. 2007; Zhang et al. 2003). This seems to be more evident in small NPs (nanoparticles) <5 nm in diameter that present a greater surface area-to-volume ratio (Li 2006; Hvolbæk et al. 2007; Cheng et al. 2012). One of the identified mechanisms as a possible reason for cytotoxicity is through the interaction of the NP surface with O₂. In this process, donor electrons are transferred from the surface of the NPs to oxygen molecules generating superoxide, which can lead to ROS production through dismutation. This has been identified for single-component materials and for transition metals on the nanoparticle surface, such as Fe and vanadium, which take part in the formation of active sites (Li 2006).

In addition to reactive radicals on the NP surface, there are other sources of oxidative stress such as the redox groups in the coating, contaminants from the production method of non-metal NPs and oxidant-inducing properties of NPs (Li 2006; Fard et al. 2015; Tournebize et al. 2012). Oxidative stress causes damage to cell membranes, DNA and protein being identified so far as one of the major causes of NP cytotoxicity (Pan et al. 2009; Xia et al. 2006). Several reports of ROS production and oxidative stress induced by nanoparticles alone have been published as well as in combination with ionizing radiation as discussed below.

Cell cycle effects

GNPs may enhance radiosensitization by causing cell cycle disruption and inducing apoptosis. The sensitivity and consequent biological effects of radiation exposure are dependent on the cell cycle phase. Different cell cycle phases present differential radiation sensitivity with late S-phase cell being the most radioresistant and late G2 and mitosis being the most sensitive (Pawlik and Keyomarsi 2004). In response to radiation, cells activate cell cycle checkpoints in G1, S and G2 phases in order to repair genomic defects, maintaining its integrity, or prevent cell division by activating cell death mechanisms (Kastan and Bartek 2004). Materials other than gold, such as tachpyridine, have been shown to induce cell cycle arrest in G2/M phase most likely due to its metal binding activity (Turner et al. 2005). However, GNPs have been more extensively studied leading to several other reports of cell cycle distribution alterations (Geng et al. 2011; Roa et al. 2009; Kang et al. 2010; Mackey et al. 2013; Mackey and El-Sayed 2014; Ganesh Kumar et al. 2015).

So far, very few studies have been reported to analyse the effects of GNPs in the cell cycle after radiation exposure. Roa et al. (2009) found that GNPs (Glucose-GNPs, 10.8 nm) alone can promote an increase in the G2/M phase in DU-145 cancer cells. When irradiated using a Cs-137 source, G0/G1 phase has been shown to accelerate and arrest DU-145 cells in the G2/M phase. In these cells, increased expression of cyclin kinases (cyclin B1 and E) involved in the regulation of the cell cycle was found together with a decreased expression of p53 (tumour protein 53) and cyclin A.

However, most studies have been performed in the absence of radiation and showed distinct results (Wahab et al. 2014; Arnér and Holmgren 2000; Taggart et al. 2016; Kang et al. 2010; Mackey et al. 2013; Mackey and El-Sayed 2014; Ganesh Kumar et al. 2015; Cui et al. 2014). Thioglucose-coated 14-nm gold nanoparticles promoted an increase in the G2/M cell phase, compared to the control, which lead to enhanced SK-OV-3 cell sensitivity to 6 MV X-ray exposure (Geng et al. 2011). The effects of nuclear-targeted GNPs have also been investigated. The results indicated that nanoparticles alone increase in the sub-G1 population and disruption of the G1/S transition inducing apoptosis in cancer cells (Kang et al. 2010; Mackey et al. 2013). Another study involving 30-nm NLS (nuclear localization sequence)-GNPs in human oral squamous carcinoma (HSC-3) has shown an increase in S phase and a decrease in G2/M phase sub-population. In combination with 5′-fluorouracil that is active during the S phase, these cells were chemosensitized (Mackey and El-Sayed 2014). Furthermore, bacteria-mediated anti-proliferative GNPs demonstrated G2/M arrest accompanied with the inhibition of tubulin polymerization and increased activation of caspases 8, 9 and 3, in DU145 cells, suggesting increased apoptosis levels (Ganesh Kumar et al. 2015). Despite the results demonstrating GNPs’ influence in the cell cycle, there are other reports indicating no significant interference (Pan et al. 2009; Cui et al. 2014; Butterworth et al. 2010). This has been reported with 2.7-nm tiopronin-GNPs and 1.4-nm triphenyl monosulfonate-GNPs (Pan et al. 2009; Cui et al. 2014). However, the results might be cell line dependent as found using 1.9-nm GNPs (AuroVistTM) and two different cell lines, DU-145 and MDA-MB-231 cells. An increase in the sub-G1 population of DU-145 cells was seen after 48-h incubation, whereas this was not detected in MDA-MB-231 cells (Butterworth et al. 2010). Again, the coating and size of the nanoparticles induce distinct responses in the various cell lines. The variety of concentrations, coatings, materials and cell lines makes it very hard to draw any conclusions regarding the exact mechanism of action of NPs. However, the alterations induced by GNPs in cell kinetics could be associated to the accumulation in G2/M which is known to be the most radiosensitive, thus increasing GNP-mediated radiosensitization.

DNA damage and repair

Another mechanism involved in GNP-induced radiosensitization is DNA damage and repair. Radiation itself induces double-strand breaks (DSBs) in DNA and their repair is essential for cell survival. Given the importance of DNA stability in determining cellular propagation, it is a key target for agents that attempt to halt cancer cells from dividing. Thus, cytotoxic targeting DNA agents such as cisplatin, gemcitabine and mitomycin C have been tested regarding their ability to act as radiosensitizers (Choudhury et al. 2006). The induction of DSBs has also been reported in the presence of GNP γ-H2AX foci analysis (Chithrani et al. 2010; Banáth and Olive 2003). Early DNA damage (1 h post irradiation) caused by GNPs appears to be related to its presence in the perinuclear region at the time of irradiation. However, late DNA damage (24 h post irradiation) seems to be related to other indirect processes such as radical production after interaction with water (Mcquaid et al. 2016).

Studies conducted with 50-nm citrate GNPs demonstrated increased foci number after being irradiated with 220 kVp and 6 MV energies for 4 or 24 h. The increased residual damage suggests possible inhibition or delay in DNA damage induction and/or repair which can be essential to the radiosensitization mechanism of NPs (Chithrani et al. 2010).

This has been further confirmed with 2.7-nm tiopronin-GNPs irradiated with 250 kVp X-rays. An increased residual DNA damage after being irradiated for 24 h was found in the samples where NPs were present. Contrastingly, no significant effect was observed only at 30 min post irradiation suggesting that the GNPs had no influence on the induction of DNA damage (Cui et al. 2014). Furthermore, GNPs have been found to induce DSBs in hepatocellular carcinoma cells after radiation exposure. Residual damage has also been found when the cells were irradiated in the presence of nanogold indicating influence in the repair mechanisms of the cells (Zheng et al. 2013). GNPs might not have all the same mechanisms of action and may induce distinct repair kinetics across different cell lines. For instance, BSA (bovine serum albumin)-capped GNPs induce increased γ-H2AX foci after 2 or 4 h of irradiation but no change is observed in the samples incubated for 24 h. This implies that these NPs do not influence cellular repair mechanisms (Chen et al. 2015). Other NPs such as 1.9-nm (AuroVistTM) have been reported to have no impact on the formation or repair in DSBs, neither at 1 h nor at 24 h after irradiation, in MDA-MB-231 cells (Jain et al. 2011).

Even if NPs demonstrate a radiosensitizing potential by promoting dose enhancement and potentially contributing to increased DNA DSB formation, the lack of consistency of cell lines, radiation sources and energies, treatment conditions and nanoparticles properties lead to incomparable results, making it difficult to draw a conclusion. The understanding of how different properties of the GNPs, irradiation conditions and biological response of various cell lines contribute to DNA damage and repair could further shed light on the underlying mechanism of GNP influence on DNA damage response.

Potential impacts of bystander effects of GNP radiosensitization

In addition to direct radiation effects, communication between cells is also very important after radiation exposure. Cells that have not been directly exposed to radiation can receive signals from irradiated ones that were in the vicinity responding in a similar way to direct exposure (Najafi et al. 2014). This process is called the bystander effect and it can occur in different cell types such as endothelial cells, fibroblasts, lymphocytes and tumour cells (Havaki et al. 2015; Prise and O’Sullivan 2009; Butterworth et al. 2013). The bystander signals involved in this process may cause altered gene expression, damage in the DNA and chromosomes, cell proliferation alterations, cell death or changes in the translation process in non-irradiated cells (Najafi et al. 2014).

The main bystander signalling molecules are reactive oxygen species or nitrogen reactive species (RNS), cytokines, miRNA (micro-ribonucleic acid) or extracellular oxidized DNA (ecDNA) (Azzam et al. 2002, 2004; Barber 2011). These are released into the surrounding environment and reach the bystander cells through passive diffusion or by binding to receptors on their plasma membrane (Barber 2011; Azzam et al. 2003). Also it can occur by direct cell-to-cell contact via gap junction intercellular communication (GJIC) (Azzam et al. 2001). Furthermore, exosomes carrying miRNA are believed to mediate intercellular signalling between tumour cells and bystander cells (Melo et al. 2016; Yang et al. 2011; Umezu et al. 2013; Sánchez et al. 2015). 21 miRNAs have been found recently to be up- or downregulated after ionizing radiation exposure. Extracellular miR-1246 (micro-ribonucleic acid 1246) in particular appears to be increasing with irradiation dose. It was found to enhance proliferation and resistance in lung cancer cells by targeting death receptor 5 (DR5) although through a non-exosome associated pathway (Yuan et al. 2014). Nevertheless, soluble miRNAs involved in bystander signalling can be generated due to ROS, RNS, cytochrome c and cytokines (Hei 2015; Shao et al. 2011; He et al. 2011).

As NPs have been shown to alter cytokine and gene expression as well as ROS production, their single presence in the tumour environment could further change the way cells respond to radiation. NPs could also mediate bystander signalling, for example, small titanium dioxide NPs induce higher levels of oxidative stress and production of inflammatory cytokines. The expression of cytokine macrophage inflammatory protein-1 alpha (MIP-1α) and high-mobility group protein 1 (HMGB1) were found to be expressed in the presence of these NPs, both in vitro and in vivo (Fujiwara et al. 2015). This may play a role in enhancing oxidative stress as HMGB1 is known to be an inflammatory macrophage-secreted cytokine, which is activated by TNFα (tumour necrosis factor alpha) and IL-1β (interleukin 1 beta) (Andersson et al. 2000; Tang et al. 2009). The latter will in turn promote the secretion of HMGB1 (Tang et al. 2009). This creates a cycle that can propagate inflammation and it might also propagate bystander effects as cytokines like TNFα and IL-1β have been shown to be elevated in bystander cells (Zhou et al. 2008). The increased production of IL (interleukin) and TNFα stimulates nitrogen oxide (NO) and ROS biosynthesis by activating NF-kB (nuclear factor kappa B) transcription factor, directly or indirectly, which in turn leads to the activation of iNOS (inducible nitric oxide synthase) and COX-2 (cyclooxygenase-2) genes’ expression (Zhou et al. 2008). iNOS controls the production of NO and COX-2 is involved in ROS production, therefore increasing oxidative stress (Najafi et al. 2014; Hei et al. 2011). Moreover, small airway epithelial cells (SAECs) exposed to GNPs are able to induce protein expression in neighbouring lung fibroblasts in co-culture systems. This study found that 47 proteins were upregulated, while 62 were downregulated in the fibroblasts receiving signals from the incubated SAECs with GNPs. Most of the proteins identified are involved in cell adhesion, extracellular matrix and cytoskeleton remodelling. Plasminogen activator, PLAU (plasminogen activator urokinase), UPA (urokinase-type plasminogen activator) and GRO-1 (growth-regulated oncogene 1) that are implicated in cell migration were found to be downregulated, potentially decreasing it. Contrastingly, proteins that promote cell adhesion such as Paxillin (PXN), breast cancer anti-oestrogen resistance 1 (BCAR1) and Caveolin-1 (Cav-1) were upregulated (Ng et al. 2015).

PXN, a focal adhesion (FA), is an associated adapter protein that regulates cell spreading and motility. BCAR1 has been correlated with controlling the spread and motility of cancer cells through regulating FA (Machiyama et al. 2014; Miao et al. 2012; Schaller 2001). Furthermore, Cav-1 is known to mediate cancer metastasis (Brennan et al. 2012). Also Cav-1 can promote NF-kB activation and lung inflammatory response, through eNOS (endothelial nitric oxide synthase) and NO production (Mirza et al. 2010). These results might indicate airway inflammation since it can be related to increased adhesion molecules (Garrean et al. 2006; Liu et al. 2014). The elevated cell adhesion was accompanied by altered F-actin stress fibre arrangement in the cytoskeleton of the lung fibroblasts. Similarly, there was an increase in vinculin binding sites that are associated with F-actin anchoring to the cell membranes of the fibroblasts, which can potentially lead to enhanced vascular permeability as previously described (Snyder-Talkington et al. 2013; Pacurari et al. 2012; Setyawati et al. 2013). Overall, cytoskeleton remodelling and increased cell adhesion may affect lung function, thus reminding of the importance of understanding the cellular communication pathways in the presence of NPs even in the absence of radiation (Ng et al. 2015).

In addition to cellular bystander responses at the tissue and whole organism levels, a phenomenon called an abscopal effect can occur. There is a response of an organ/site distant from the irradiated one, where the cells are not close to each other, which has been observed in patients undergoing localized radiotherapy (Kaminski et al. 2005). The importance of the abscopal effect remains to be fully understood. It might have the potential to either increase cell killing or protect normal tissues (Prise and O’Sullivan 2009). The role of NPs in mediating abscopal effects has not been elucidated due to a lack of in vivo and clinical studies.

As nanoparticles may induce changes in cell signalling, leading to various responses depending on their size, shape and coating, understanding which signalling pathways are influenced could potentially help understand their mechanism of action in the bystander/abscopal and radiosensitization effects.

Uptake

Radiation absorption and dose deposition is thought to be partially reliant on the number of gold nanoparticles present within the cell, meaning that cellular uptake and distribution of GNPs will have a direct influence on the degree of radiosensitization observed (Chithrani and Chan 2007). This makes cellular uptake of GNPs an important metric in modifying sensitivity to radiotherapy. Modelling carried out to replicate GNP uptake equal to 1% mass in the cytosol demonstrated dose enhancement in both the nucleus and mitochondria, despite cytosol localization. It was concluded that the physical mechanism of dose enhancement was caused by photoelectron delocalization from the cytosol to cell organelles, meaning that the dose enhancement effects were not limited to the vicinity of the nanoparticles (McNamara et al. 2016).

Both uptake potential and blood circulation times have been shown to be closely associated with nanoparticle size. The optimal size for uptake has been found to be between 25 and 50 nm, with particles smaller than 10 nm or larger than 100 nm exhibiting a reduced uptake potential (Yang et al. 2014).

Chithrani et al. found the uptake of gold nanoparticles to be heavily reliant on their size and shape. Internalization of smaller particles is observed to be a more rapid process than that of larger particles. This is important as the uptake of nanoparticles into the cell will have a direct influence on the level of radiosensitization (Chithrani and Chan 2007). Coulter et al. reported the maximum amount of nanoparticle uptake to occur within the first few hours of exposure of cells to 1.9-nm GNPs, with a plateau being reached after 6 h. However, they also highlighted the difficulty in making direct comparisons with previous work as a result of the number of variables involved, as uptake may differ depending on the nanoparticle shape, size, coating, concentration and charge, as well as cell type (Coulter et al. 2012). These studies show cellular uptake to be dependent on concentration, time and cell type. Experiments carried out in hypoxic cell models saw reduced GNP uptake, thought to be linked to reduced energy production in the cells (Jain et al. 2014).

While GNPs will passively accumulate in tumours due to the EPR effect, their encapsulation within liposomes has been shown to result in higher cellular internalization (Maeda et al. 2003). Uncapped nanoparticles bind to various plasma proteins upon administration, resulting in a large number being internalized by macrophages and removed from circulation. Liposomes have a history of being used for drug encapsulation and delivery as their large size (100–200 nm) ensures that they can pack many GNPs within their lipid bilayers. Small nanoparticles encapsulated within liposomes show better passive accumulation within the tumour than non-encapsulated nanoparticles (Chithrani et al. 2010). The contents can then be released through a triggering technique, allowing the nanoparticles to penetrate the tumour tissue more effectively (Kneidl et al. 2014). Alternatively, nanoparticles capped with PEG (polyethylene glycol) alone will show an increase in their half-life in blood (Hirn et al. 2011), while HSA (human serum albumin)-conjugated gold nanoparticles have shown increased retention in the lungs and brain compared with both apoE (apolipoprotein E)-capped and citrate-stabilized nanoparticles (Schuffler et al. 2014).

In contrast to passive targeting, active targeting is the functionalization of the GNP surface with peptides, ligands or antibodies in order to preferentially target tumour cells (Schuemann et al. 2016), taking advantage of the overexpressed surface receptors of cancer cells. Not only does this increase the therapeutic ratio through achieving a greater nanoparticle concentration within the tumour, but also reduces the overall volume of gold required for treatment. Sykes et al. (2014) carried out an investigation into the impact that nanoparticle size has on active and passive targeting. Gold nanoparticles with diameters of 15, 30, 60 and 100 nm were prepared with the surface modified with either PEG or PEG conjugated to transferrin. Those nanoparticles modified with transferrin showed a significant increase in tumour accumulation in vivo, with accumulation in the 60-nm particles being 1.9 times greater than that in their passive counterparts. However, it is of note that while the PEG-only nanoparticles were slower, they also diffused deeper into the tumour. Popovtzer et al. (2016) successfully demonstrated significant improvement in tumour radiosensitivity using GNPs covalently conjugated to CTX monoclonal antibody in mice. This actively targeted the tumour with no evidence of early or delayed toxicity, confirmed by histological characterization of the tumour and adjacent tissue both 1 and 6 weeks post treatment. TNF has been covalently conjugated to gold nanoparticles, with the aim of the interaction between TNF and its receptor TNF-R1 causing active targeting of the tumour cells. Molecules of PEG-Thiol are interspersed between TNF molecules, and nanoparticles may be treated with a coating to increase cellular internalization.

Tumour blood vessels may also be promising targets, as sub-100-nm-diameter nanoparticles are expected to accumulate in the vasculature (Perrault et al. 2009). The important role of endothelial cells within the tumour vasculature makes them ideal targets with high potential clinical impact. However, a model generated by Berbeco et al. examining gold nanoparticles as tumour vascular disrupting agents found the boosted irradiation of endothelial cells alone was not viable due to the short range of the ionizing particles. The combination of tumour-targeting functionalization with image-guided radiotherapy could combat this by ensuring that the tumour has maximum levels of GNPs present during irradiation. Nanoparticle-mediated drug delivery to the tumour vasculature has been shown to have anti-metastatic effects and tumour gold content will give a good indication of overall tumour vascularity (Murphy et al. 2008). Shifting nanoparticle accumulation from the reticuloendothelial system (RES) to other organs will result in longer retention times, though unanticipated retention over prolonged periods of time may result in cytotoxic effects (Balasubramanian et al. 2010). This accumulation of GNPs may be achieved through the addition of peptides, allowing more site-specific delivery. For example, insulin has been used to improve the delivery of gold nanoparticles to target sites in the brain (Shilo et al. 2014), while other peptide-capped gold nanoparticles have been shown to pass through the blood–brain barrier using a number of mechanisms (Velasco-Aguirre et al. 2015). The ability to accurately target organ sites for uptake greatly enhances the radiosensitizing potential of GNPs.

In conclusion, while there are several benefits to passive targeting, it is noticeably less efficient in slow-growing models when compared to fast-growing models, since the former have more mature and intact blood vessels (Kunjachan et al. 2015). Cellular targeting can face problems in the way of tissue barriers which vascular targeting avoids by providing nanoparticles with direct access or binding to the overexpressed targets. Attacking the vasculature can also have the added benefit of affecting the numerous cancer cells that rely on it for growth (Kunjachan et al. 2014, 2015). GNP uptake has been shown to be reliant on several variables, including nanoparticle shape, size, charge and concentration, alongside the cell type. While the EPR effect results in the passive accumulation of GNPs within the tumour, encapsulation and specific coatings allow for increased cellular targeting and retention.

Imaging

A consequence of the increasing use of gold nanoparticles in biomedical applications is the need for their accurate and efficient detection in both biological and tumour samples. The ability to determine nanoparticle location provides insights of uptake pathways as well as potentially identifying nanoparticle location as a cause of cytotoxicity. Accurate imaging of nanoparticle location within a sample will help achieve precise dose deposition and may also provide understanding of mechanisms behind radiosensitization. To quantify the uptake and distribution of nanoparticles within cells, several imaging techniques can be used. The physical properties of gold nanoparticles also allow them to be used as imaging agents. Miladi et al. (2014) used 2.4-nm GNPs coated with DTDTPA (dithiolated diethylenetriamine pentaacetic acid)-gadolinium chelates to combine MRI (magnetic resonance imaging) and radiosensitizing effects. The DTDTPA coating acts in preventing the clumping of GNPs along with slowing their uptake by the RES. Osteosarcoma- and gliosarcoma-bearing rats were injected with the nanoparticles, which were then monitored using MRI. The rats were then irradiated at the point when the highest content of nanoparticles was observed in the tumour. Several approaches can be used to image the uptake and distribution of GNPs throughout the cell, including localized surface plasmon resonance, photoacoustic imaging, computerized tomography, X-ray fluorescence computed tomography and electron microscopy. Each technique will have individual advantages and limitations, so it can be worthwhile to apply multiple techniques in order to improve the reliability of diagnostics and treatments (Botchway et al. 2015).

A characteristic affecting the imaging techniques that can be used with gold nanoparticles is localized surface plasmon resonance (LSPR). This is instigated by their ability to absorb and scatter specific wavelengths of light and refers to the resonance established between incident light photons and particle surface electrons. The LSPR of a nanoparticle can provide information about its overall size and structure, as when these change, so does its resonant frequency. While LSPR has a high sensitivity and relatively low cost, it can be time consuming to set up depending on the target (Petryayeva and Krull 2011). Photoacoustic imaging (PA) involves irradiating tissue using a nonionizing short-pulsed laser beam. Exogenous contrast agents absorb this energy to produce ultrasound waves, which are received using a transducer. The mechanical acoustic waves are then converted into an electronic signal which is processed to form an image. As the photoacoustic waves are only generated within the tissue sample, there is reduced background interference. PA gives higher spatial resolution and deeper imaging depth compared to fluorescence optical imaging, while the lack of ionizing radiation also makes it a safer option than computerized tomography (CT) (Wang 2008; Pan et al. 2013; Li and Chen 2015).

The strong effect demonstrated by gold nanoparticles allows them to be used as contrast agents in photoacoustic imaging. Changing the shape and size of the nanoparticles that are being used enables the tuning of the magnitude of light being absorbed and scattered and when compared with other contrast agents, such as imaging dyes and silver nanoparticles, they demonstrate a greater absorbance (Menon et al. 2013; Huang and El-Sayed 2010).

The accumulation of gold nanoparticles in tumours due to the EPR makes them well suited for photoacoustic imaging. This helps determine the location of the tumour along with assessing the vasculature and accumulation of therapeutic agents. However, the photostability of gold nanoparticles is a potential limitation as they can change shape due to high laser energies. Zhang et al. utilized GNPs as PA agents to detect human breast cancer xenografts in mice. The nanoparticles were found to accumulate within the tumours after 5 h following injection via tail vein and a significant enhancement in signal intensity was seen. This accumulation and signal enhancement would allow gold nanoparticles to be used as both tumour contrast agents and mediators of cancer therapy (Chang et al. 1999; Zhang et al. 2009).

The high atomic number and potentially long circulation time make gold nanoparticles ideal contrast agents for CT (Cormode et al. 2014). Hainfeld et al. used 1.9-nm gold nanoparticles as a contrast agent for X-ray therapy in rats. Contrast was estimated to be ~3 times greater than that of iodine at 100 keV, a useful range for clinical CT. The agent was then observed to be excreted via the kidneys with no observed toxicity. Further work used gold nanoparticles as a contrast agent to detect smaller tumours (1.5 mm). Nanoparticles were observed to accumulate in the tumours, and micro-CT allowed for quantification of this. They also noted that there was a greater uptake of nanoparticles without any attached antibody, which they believed to be due to the smaller size.

Alric et al. used a gadolinium chelate to produce a nanoparticle that could be used for both MRI and CT imaging. The nanoparticles were found to circulate freely in blood with no adverse accumulation in the lungs, liver and spleen (Alric et al. 2008).

X-ray fluorescence computed tomography (XFCT) is an imaging technique that aims to simultaneously determine the identity, quantity and spatial distribution of elements within imaged objects. Previously, Cheong et al. had been successful in accurately identifying the location of a GNP-filled object within a small animal-sized plastic phantom, as well as quantifying the amount of GNPs present. However, at the time their technique was not yet practical for routine in vivo use (Cheong et al. 2010). More recently, Manohar et al. (2016) demonstrated the use of benchtop XFCT for imaging a small animal that had been injected with gold nanoparticles. However, they found that their set-up required further refinement before it could be used routinely.

Electron microscopy is well suited to imaging gold nanoparticles due to their high electron density. It allows for the determination of the size and shape down to 1 nm and as a result is commonly used when characterizing nanoparticles. However, transmission electron microscopy (TEM) has a complicated and time-consuming sample preparation, and individual nanoparticles are not also distinguishable if their size is below the resolution limit (Schrand et al. 2010). Scanning electron microscopy (SEM) requires coating samples in a conductive metal, usually composed of nanometer-sized clusters. The similarities between samples prepared using conventional methods and metal nanoparticles make SEM incompatible. However, Goldstein et al. (2014) developed a simplified procedure that involved chromium coating, allowing them to observe cellular uptake of GNPs.

The physical properties of GNPs allow them to be used with a wide range of imaging techniques. This not only allows for the quantification of their uptake and distribution, but also makes them potential contrast agents. The advantages and limitations of each technique indicate that it may be necessary for more than one technique to be applied to improve reliability.

Toxicity

GNPs must exhibit safe behaviour after cellular uptake to be considered potential radiosensitizers. If individual nanoparticles are found to be cytotoxic or to reduce cell viability, then they might not be suitable for clinical use. Following IV (intravenous) injection, both the biodistribution and clearance of nanoparticles are influenced by various physiological factors (bloodstream, opsonization, endothelial permeability of vessels and organs) as well as the individual physicochemical properties of nanoparticles (size, charge and surface chemistry). These physicochemical properties can be altered to control their biodistribution. It is preferable that particles are later removed through urine as this implies that there has been no degradation.

There is a level of uncertainty regarding the cytotoxicity of GNPs. While bulk gold is known to be biologically safe, functionalized GNPs have shown obvious cytotoxicity (Goodman et al. 2004). Size, concentration, cell type and treatment time are all basic parameters to be considered when examining the cytotoxicity of GNPs. Size is an important factor, as very small particles have been found to be highly toxic, and larger particles are relatively nontoxic (Pan et al. 2007).

Zhang et al. summarized that while high gold concentrations cause an obvious decrease in cell viability, low gold concentrations do not appear to influence viability. Nanoparticles with a diameter of 15 nm were found to be nontoxic up to 75 µg/ml, though cell viability was obviously affected at concentrations of gold >150 µg/ml. At 600 µg/ml, cell viability was reduced to 41.8%, compared to 93.9% at 18.75 µg/ml (Zhang et al. 2009).

A study by Connor et al. found no difference in growth rates between untreated cells and cells exposed to 25 mM concentration of 18-nm nanoparticles over the course of 5 days. The uptake and localization of the nanoparticles in the cell were confirmed by TEM, leading to the conclusion that nanoparticles are not inherently toxic to human cells. However, it was noted that the determination of whether the nanoparticles are modified by their environment is important, as this may result in significant variation to their clinical applications (Connor et al. 2005).

In 2004, Hainfeld et al. (2004) carried out work that involved injecting mice with 2.7 g Au/kg per body mass. These mice survived over a year without showing any obvious clinical effects, suggesting that in this case the nanoparticles were biocompatible. Further work by Hainfeld et al. (2013) showed that irradiation with 11.2-nm nanoparticles increased long-term survival (over 1 year) of mice by 50%, while none of the mice receiving the same treatment without nanoparticles survived longer than 150 days. As the LD50 (lethal dose at 50%) of these nanoparticles was >5 g Au/kg, a dose of 4 g Au/kg was used, showing that by using the appropriate dose of nanoparticles, toxicity can be averted and lifespan can be greatly improved.

The effects of 5- and 15-nm GNPs on mouse fibroblast cells were examined by Coradeghini et al. to provide data on the toxic potential of different sized nanoparticles. Using a colony forming efficiency (CFE) assay, the in vitro toxicity of gold nanoparticles was tested at the concentrations of 10–300 µM and at the times of 2, 24 and 72 h. Significant cytotoxicity was only seen in cells treated with 5-nm nanoparticles at a concentration over 50 µM and an exposure time of 72 h, with no significant cytotoxicity observed in the 15-nm nanoparticles. TEM imaging showed that cellular internalization occurred for both sizes of nanoparticles (Coradeghini et al. 2013).

Stefan et al. examined the effects of 12- and 22-nm chitosan-capped gold nanoparticles on rats treated with lipopolysaccharide (LPS). Brain and liver tissue reactivity was assessed following 8 days of administration. They found that while the body weight of the treated rats did not change significantly, the ratio between liver and body weights significantly increased with both nanoparticle sizes, especially the 22-nm ones, suggesting potential liver toxicity. The 22-nm nanoparticles also experienced a significant decrease in their brain-to-body weight ratio, suggesting possible brain damage. Dark-field imaging showed the agglomeration of nanoparticles within cytoplasmic cellular regions as a potential cause of this damage. This was not seen when using the 12-nm nanoparticles (Stefan et al. 2013).

While it remains a challenge to accurately estimate the cellular response to a given nanoparticle size, there are general trends which can be trusted. While both the cell type and surface properties of the nanoparticle play a role in nanoparticle uptake, smaller nanoparticles are more likely to be passively internalized, though are also more likely to have a cytotoxic effect (Shang et al. 2014).

As there is increasing interest in using nanomaterials in both research and a clinical setting it is important that nanoparticle cytotoxicity can be tested in a fast and efficient manner. Karlsson et al. used the in vivo assay “ToxTracker” with a panel of metal oxide and silver based nanoparticles. This comprises of a panel of mouse embryonic stem cells, each containing a GFP (green fluorescent protein)-tagged reporter for a distinct cellular signalling pathway. In this way it is possible to identify DNA damage caused by direct DNA interaction, oxidative stress and general cellular stress (Karlsson et al. 2014). If this assay was adapted for GNPs it may elucidate their cytotoxic properties.