GNRs with an aspect ratio of 2.8 are synthesized via
a seed-mediated growth method previously reported by Murphy and El-Sayed.56,57
The CTAB bilayer on the surface of GNRs renders GNRs highly positively charged with a zeta potential of +45 mV, which helps maintain the stability of the nanoparticle suspension, but, at the same time, introduces cytotoxicity due to detached CTAB molecules. Both non-porous and mesoporous silica have been previously reported to encapsulate GNRs to improve their biocompatibility.42,43,58
For example, Fernandez-Lopez and coworkers have developed a robust procedure for coating GNRs with silica shells of tunable thickness.43
The key is to first transfer GNRs in water to an ethanol solution by replacing CTAB with SH–PEG so that the classic Stöber chemistry can be applied.59
Indeed, using this procedure we were able to encapsulate GNRs with silica shells ranging from 7 to 30 nm thickness. shows the transmission electron microscopy (TEM) image of GNR@SiO2
with a representative shell thickness of 7.5 nm. Although the GNRs, compared with other highly uniform spherical nanoparticles, have some size and shape variations, the coated silica shells appear uniform in thickness and smooth. For easy grafting of PEG molecules on the outer surface for optimal biocompatibility, the hydrophilic GNR@SiO2
was converted to hydrophobic first by reacting with a hydrophobic silane compound, OTMS. In spite of the three reactive sites (polymerizable), OTMS does not significantly increase the silica shell thickness, possibly due to steric hindrance caused by its long hydrocarbon chain. Nevertheless, the very thin layer of hydrocarbons efficiently converts the GNR@SiO2
hydrophobic, rendering them soluble in organic solvents such as chloroform. Previous research on semiconductor quantum dots has identified a series of simple and effective approaches based on amphiphilic polymers for solubilization of hydrophobic nanoparticles.51–54
These amphiphilic structures share a couple of common features. First, no ligand exchange is needed for nanoparticle solubilization (original hydrophobic ligands are kept and covered). Second, the amphiphilic polymers self-assemble onto inorganic nanoparticles via
multivalent hydrophobic interactions whereas the hydrophilic segments in the polymers face outward. To demonstrate the concept, we used PE-PEG in the current work. The tail chains in PE and the hydrocarbons on GNR–SiO2
interdigitate, whereas the PEG block protrudes from the surface with terminal COOH groups as reactive sites for bioconjugation. The overall yield of this multistep procedure is typically above 70%.
Fig. 2 TEM and spectroscopic characterization of caged GNRs. (a and b) TEM images of GNR@SiO2 and GNR@SiO2@PE-PEG. The silica shell is 7.5 nm, and the organic molecules on the silica shell surface are not visible because they are not electron-dense materials. (more ...)
As shown in , the resulting water-soluble GNR@SiO2@PE-PEG particles are well separated on the TEM grid, which also reflects their dispersity in solution, because if the particles were aggregated in solution, large clusters with particles piled on top of each other would be expected. To confirm the dispersity in solution, we also measured the hydrodynamic size of the particles using dynamic light scattering (DLS). In aqueous solution, the hydrodynamic diameters of GNR@SiO2 and GNR@SiO2@PE-PEG are 76 nm and 97 nm, respectively. The size discrepancy between these two samples in solution (‘dry’ sizes under TEM are similar) is likely resulted from the PEG surface coating and the negative charges on the surface of GNR@SiO2@PE-PEG (PEG terminated with COOH end group), which creates an electrical double layer surrounding the nanoparticles. It is also worth mentioning here that the DLS sizes of GNRs should be treated as an estimate because the numbers are actually expressed as GNRs’ spherical particle equivalents, spheres that have the same translational diffusion speeds as the non-spherical GNRs.
Besides the size measurements using TEM and DLS, the silica and PEG coating process can be monitored with spectroscopic measurements as well, because GNRs’ surface plasmon resonance (SPR) bands in particular the longitudinal band are sensitive to changes in local environments.18,19
As shown in , the CTAB-coated GNRs in water initially have a transverse absorption band centered at 515 nm and a longitudinal band at 685 nm. Subsequent surface modifications with an SiO2
and PEG give rise to a significant red shift of the longitudinal band to 719, 720, and 708 nm, respectively. This spectral shift is commonly observed in plasmonic materials and has been previously attributed to changes in the refractive index of the surrounding environment.60
At the same time, the transverse band remains nearly unchanged in band location, but is slightly enhanced due to stronger Rayleigh scattering arising from the increased particle size after silica encapsulation.43
Next, we systematically compared the colloidal stability of our GNR@SiO2@PE-PEG with the conventional GNR@CTAB, GNR@SiO2, GNR@PSS, and GNR@SH–PEG dispersed in a variety of aqueous solutions: DI water, NaCl solution (100 mM), phosphate buffered saline (PBS, 1×), and cell culture media OptiMEM and RPMI. The DLS measurement allows a quick assessment of colloidal stability of the GNRs with different surface chemistry. As summarized in , the original GNR@CTAB and GNR@SiO2 form aggregates in NaCl and PBS solutions as well as OptiMEM, whereas GNR@PSS are not stable in NaCl and PBS solutions. In contrast, only GNR@SH–PEG and GNR@SiO2@PE-PEG remain single under all these conditions, manifested by minor size fluctuations possibly caused by the solvent change and adsorption of solutes (e.g., proteins in cell culture media). These results indicate the importance of the pegylation layer in nanoparticle stabilization. Note that the size fluctuation of GNR@SH–PEG is slightly less than that of GNR@SiO2@PE-PEG. This is because, in contrast to the PE-PEG terminated with COOH groups, the SH–PEG used in this study are terminated with neutral methoxy groups. When GNRs are coated with SH–PEG–COOH, they show a similar level of size fluctuation to that of GNR@SiO2@PE-PEG.
Hydrodynamic diameter of the encapsulated GNRs in different media
The DLS data are also confirmed by measurements of GNRs’ SPR absorption, utilizing the sensitivity of the longitudinal bands to colloidal agglomeration (). The spectral measurements are in nearly perfect agreement with the DLS results except for GNR@SiO2. The absorption peaks of GNR@SiO2 in various solutions show relatively small changes although DLS clearly suggests aggregation. This is due to the silica shells on their surfaces which effectively create spatial separations between adjacent GNRs in clusters. Representative TEM images were also taken for GNR@CTAB, GNR@SiO2, and GNR@SiO2@PE-PEG in 100 mM NaCl solution. It shows that GNR@SiO2@PE-PEG are well dispersed but not the other two samples (). Taken together the DLS, TEM, and spectroscopic data, it is clear that direct ligand exchange with SH–PEG and encapsulation with SiO2 and PE-PEG offer the best colloidal stability for GNRs.
Fig. 3 Spectral and TEM measurements of the colloidal stabilities of GNRs with various surface chemistries. (a–e) Vis-NIR spectra of GNR@CTAB, GNR@PSS, GNR@SH–PEG, GNR@SiO2, and GNR@SiO2@PE-PEG dispersed in water, PBS buffer, 100 mM NaCl solution, (more ...)
As discussed above, GNRs’ photothermal stability is another issue that needs to be addressed for applications involving high-intensity laser illumination or extended photothermal conversion such as photoacoustic tomography, multiphoton microscopy, and optically modulated drug release. Systematic work by El-Sayed and coworkers shows that under pulsed laser irradiation, GNRs can melt into spherical particles of similar volumes or even fragment into smaller particles.7,48
High-resolution TEM also reveals that different from thermal melting, which starts from the nanomaterial surface, the photothermal melting process starts with the formation of defects inside nanorods.61
In our study, an 812 nm NIR laser was used to generate pulses of 7 ns with a repetition rate of 20 Hz, and its laser fluence was varied between 0–7.6 mJ cm−2
. To match the photon energy with GNRs’ SPR absorption, another GNR sample with a longitudinal absorption band centered at 812 nm was used, thanks to the remarkable size tunability of GNRs. As demonstrated in , the silica-caged GNRs show significantly improved photothermal stability compared with CTAB, PSS, and SH–PEG coated GNRs, in agreement with prior reports.49,50
Based on the spectroscopic measurements, GNR@CTAB and GNR@PSS are already damaged at laser fluence as low as 1.3 mJ cm−2
, indicated by a blue-shifted longitudinal peak (). GNR@SH–PEG show slightly better stability, but still follow a similar trend as GNR@CTAB (). For the silica-caged particles (GNR@SiO2
@PE-PEG), their absorptions do not change at a laser fluence of 5.1 mJ cm −2
, and drop less than 20% at a laser fluence of 7.6 mJ cm−2
after 10 min illumination (). This improved stability likely comes from two effects. First, it is well known that GNR stability is directly linked with its surface ligands.62
It has been reported that the heat relaxation time for GNRs wrapped with CTAB bilayers (also applies for the PSS-coated GNRs, because PSS are adsorbed on top of CTAB via
electrostatic interactions) is approximately 150 ps,63,64
significantly longer than the rod reshaping time (shape transformation from rods to spheres takes ca.
In contrast, the heat relaxation time of GNRs in silica shells has been estimated to be 20 ps,63,64
which is competitive with the photothermal reshaping process. Second, the silica shell is also more rigid than most organic materials such as surfactants and polymers. The high sol–gel glass transition temperature of silica (1000–1500 K) helps keep GNRs in their original shape.
Fig. 4 Comparison of photothermal stability of GNRs with various surface chemistries. (a) Peak intensity at 812 nm measured after exposure to nanosecond NIR laser at different laser fluencies. GNRs caged with silica are significantly more stable than coated (more ...)
Before applying this new series of GNRs to tumor cell targeting and photothermal therapy, two additional properties need to be checked first: (i) whether GNR@SiO2
@PE-PEG are toxic to cells; and (ii) whether the GNR@SiO2
@PE-PEG maintain their capability of heating up the surrounding media when illuminated with a CW laser source. In contrast to pulsed laser that specifically heats GNRs and their very close proximity (10 nm), CW laser illumination warms the surrounding media and is thus frequently used in hyperthermia treatment of tumors. To probe the cytotoxicity and put it in context with other surface coating chemistry, we compared the cytotoxicity of GNR@SiO2
@PE-PEG with GNR@CTAB, GNR@SiO2
, GNR@PSS, and GNR@SH–PEG towards a prostate cancer cell line, LNCaP. We chose LNCaP because of its well characterized in vivo
behavior and tumor biology such as profile of surface receptor expression, which can be used to develop specific targeting strategies.52
Plot of dose-dependent cytotoxicity () shows that after 24 hour exposure to cells GNR@SiO2
@PE-PEG are essentially non-toxic in the concentration range of 0–0.6 nM, similar to GNR@SiO2
, GNR@PSS, and GNR@SH–PEG. The only surface coating that is highly toxic to cells is CTAB, in agreement with the literatures.36–46
To investigate the photothermal conversion capability of the new GNR@SiO2
@PE-PEG, we compare it with the original GNR@CTAB and control samples of water and a spherical gold nanoparticle solution. Under the same laser illumination condition, GNR@SiO2
@PE-PEG and GNR@CTAB show virtually identical rate of temperature increase, which is approximately one order of magnitude higher than those of spherical particles and pure water ().
Fig. 5 Characterization of cytotoxicity and capability of photothermal conversion of GNR@SiO2@PE-PEG. (a) Dose-dependent cytotoxicity of GNR@SiO2@PE-PEG in comparison with GNR@CTAB, GNR@PSS, GNR@SH–PEG, GNR@SiO2 in LNCaP prostate tumor cells. In the (more ...)
With the systematic characterization studies on stability, biocompatibility and photothermal conversion discussed above, we proceeded to demonstrate specific targeting and thermal ablation of tumor cells using the new GNRs. Photothermal therapy of tumors using targeted plasmonic materials has the potential to improve current clinical tumor ablation approaches (e.g.
, laser, radio frequency, and high-intensity focused ultrasound or HIFU) that lack specificity, and to open new opportunities for treating small metastasis, and tumors with poorly defined boundaries or embedded in vital regions. Pioneer work by Halas et al
. using gold nanoshells has produced encouraging results using cultured cells and lab animals, and the nanoshell agent is currently under clinical trials.66
GNRs have become an important alternative due to their higher absorption efficiency per unit volume. For specific targeting, we conjugated a RNA aptamer (A10) that recognizes a prostate specific membrane antigen (PSMA),67
one of the most specific biomarkers expressed in prostate tumor epithelial cells and attractive targets for imaging and therapeutic interventions (e.g.
, ProstaScint® scan).68
To evaluate specific targeting, PSMA-positive LNCaP cells and PSMA-negative PC3 cells were incubated with the GNR@SiO2
@PE-PEG with or without the aptamer followed by exposure to a NIR laser (803 nm, 10 W cm−2
). Cell viabilities were quantified using a dual-color labeling technique with calcein AM and ethidium homodimer 1 (EthD-1), where red fluorescence from EthD-1 indicates dead cells and green fluorescence from calcein AM indicates live cells. Control experiments (), in which either NIR laser illumination or GNRs are absent, show that the cell viabilities are essentially unaffected (>95% live cells). When NIR illumination and GNR–aptamer bioconjugates are combined, majority of the LNCaP cells are killed with the cell viability dropping below 4% (). In contrast, under the same experimental conditions, 99% of PSMA-negative PC3 cells survive the treatment (), indicating remarkable targeting specificity.
Fig. 6 Photothermal therapy of tumor cells. (a–d) When either NIR laser illumination or GNRs targeted with A10 aptamers are missing, the viabilities of LNCaP cells remain high, 95.2% and 97.5% respectively. (e and f) Combined treatment of NIR laser illumination (more ...)
Besides the targeting specificity and effective photothermal therapy, another important feature worth mentioning is the spectral insensitivity of GNR@SiO2
@PE-PEG to aggregation. When plasmonic nanoparticles enter cells through endocytosis (one of the most common pathways for nanoparticle uptake), they often form clusters in endosome and lysosome, which leads to aggregation-induced spectral shifting. For example, when GNR@SH–PEG (targeted with the same PSMA aptamer) enter cells, the spectrum shifts by approximately 20 nm (Fig. S1
, ESI†). The unpredictable nature of the aggregation-induced spectral shift renders photothermal therapy difficult because the nanoparticle absorption maximum no longer overlaps with the laser (thus new light source will have to be identified for effective treatment and minimized side effect). In contrast, because of the silica layer, even GNR@SiO2
@PE-PEG particles form aggregates during the endocytosis process, the embedded GNRs are still spaced out by a distance at least twice of the silica shell thickness, which helps maintain the original spectrum.