Open Access
Add to BibSonomyAbstract
Effective treatment of patients with malignant brain tumors requires surgical resection of a high percentage of the bulk tumor. Surgeons require a method that enables delineation of tumor margins, which are not visually distinct by eye. In this study, the feasibility of using gold nanorods (GNRs) for this purpose is evaluated. Anti-Epidermal Growth Factor Receptor (anti-EGFR) conjugated GNRs are used to label human xenograft glioblastoma multiforme (GBM) tumors embedded within slices of brain tissues from healthy nude mice. The anti-EGFR GNRs exhibit enhanced absorption at red to near-infrared wavelengths, often referred to as the tissue optical window, where absorption from blood is minimal. To enable definition of molecular specificity and spatial accuracy of the label, the GNR absorption is compared with GFP fluorescence which is expressed by the GBM cells used here. This work demonstrates a simple but highly translational technique to classify normal and malignant brain tissue regions in open surgery applications using immunolabeled GNR contrast agents.
© 2013 Optical Society of America
1. Introduction
Glioblastoma multiforme (GBM) is the most common and aggressive form of primary brain tumor [1]. GBMs develop from the neuroepithelial glial cells, which mainly function to support and insulate surrounding neurons. The World Health Organization (WHO) has classified GBM as a class IV malignancy [2]. This classification refers to the tumor’s high rate of proliferation, increased angiogenesis, invasiveness into the surrounding tissues, and fatal outcome. For example, according to the Centers for Disease Control and Prevention (CDC), while malignant brain tumors account for just above 1% of the newly diagnosed primary tumors each year, they are estimated to be within the top ten causes of cancer related deaths each year in the United States [3]. The median prognosis for patients diagnosed with GBM is approximately 12 months, while the five-year survival rate for this disease is only 4% [4]. The poor prognosis associated with GBM is a direct effect of the tumor’s diffuse proliferation along with its resistance to traditional treatments [5,6].
Treatment for patients diagnosed with GBM normally consists of maximal surgical resection followed by radiation or chemotherapy. Studies have shown that near complete tumor resection (>90%) improves patient prognosis by preventing tumor recurrence and avoiding the need for a second procedure [7]. However, achieving a maximal extent of resection can be quite difficult due to the invasive nature of the tumor. Furthermore, tumor margins are not visually distinct compared to the surrounding brain parenchyma. Therefore, neurosurgeons require a method to delineate the margins of tumors in real time during surgical resection to maximize the extent of resection and also preserve functional brain tissue.
We propose the novel use of immunolabeled plasmonic gold nanoparticles (GNPs) as molecular contrast agents for brain tumor delineation. GNPs exhibit spectrally specific enhanced optical absorption and scattering due to the effects of local surface plasmon resonance (LSPR) [8]. The resonant peak wavelength for a GNP may be tuned by controlling factors such as the particle’s composition, size, shape, and the surrounding dielectric environment [9,10]. GNP molecular labels for specific biological markers may be designed by binding their surfaces to targeting proteins such as antibodies [10–15]. Molecular labeling with GNPs has several advantages over the use of fluorophores, the current standard in optical molecular imaging. These include biocompatibility [16,17], photostability [18], tunable optical properties [9,10,19], as well as a signal dependence on extrinsic factors including the local dielectric environment and proximity to other nanoparticles [12,20,21]. For example, by adjusting the aspect ratio of gold nanorods (GNRs), their plasmonic peak wavelength may be tuned anywhere throughout the tissue window (600-1200nm), where absorption from blood is minimal [10,19], making them ideal for imaging within thick tissues.
To target the GBM tumor cells, we use GNRs conjugated to monoclonal antibodies that target the Epidermal Growth Factor Receptor (EGFR). EGFR is classified in the ErbB receptor family. When EGFR binds onto its ligand, it induces signaling pathways responsible for cell differentiation, proliferation, survival and adhesion [22,23]. Overexpression of EGFR is indicative of various forms of malignant tumors. Approximately 40-50% of GBM tumors contain amplifications of the wild-type EGFR gene, making EGFR overexpression the most common mutation in GBMs [6,24,25]. Therefore, EGFR is an effective target for molecular labeling of GBMs. For example, 270-GBM, a human glioblastoma xenograft cell line used in this study, has been shown to overexpress this receptor [26]. 270-GBM and other GBM samples also express the mutated EGFRvIII protein [24]. The monoclonal antibodies used in this study target both the wild type and mutated forms of EGFR, allowing for effective labeling.
This article describes experiments using anti-EGFR conjugated gold nanorods (GNRs) to label solid GBM tumors embedded within brain tissue. Previously, Puvanakrishnan et al. have demonstrated the molecular specificity of anti-EGFR GNRs topically applied to solid squamous cell carcinoma tumors using widefield imaging [27]. In this study, we seek to quantify the spatial accuracy of anti-EGFR GNR labels when topically applied to brain tissues containing EGFR + 270-GBM tumors. The accuracy of the molecular label is determined by comparing the GNR absorption to GFP fluorescence from transfected malignant cells.
2. Methods
2.1 Gold nanorod synthesis and characterization
GNRs were synthesized using an adaptation of the seed-mediated method established by Nikoobakht et al. [10,11,28]. A seed solution was created by first adding 0.250 mL of 0.01 M hydrogen tetrachloroaurate trihydrate (HAuCl4ˑ3ˑH2O, Sigma–Aldrich, 520918) to 7.5 mL of an aqueous solution of 0.1 M hexadecyltrimethylammonium bromide (CTAB, Sigma–Aldrich, H9151). Gold seeds were then formed by adding 0.6 mL of cold 0.01 M sodium borohydride (NaBH4, Alfa Aesar, 13432), a strong reducing agent. This solution was gently stirred and heated until it was ready for use in nanorod synthesis.
The gold nanorod solution was formed by first adding 4 mL of 0.01 M HAuCl4 to 95 mL of 0.1 M CTAB in a separate bottle kept at 29°C in a water bath. 0.5 mL of 0.01 M silver nitrate (AgNO3, Alfa-Aesar, 11414) was added to the resulting solution. 0.64 mL of 0.1M ascorbic acid (C6H8O6, Alfa-Aesar, 11188), a weak reducing agent, was then added. After adding each of these components, the bottle was swirled to ensure even mixing. Finally, 50 μL of the previously described gold seed solution was added to the solution to begin nanorod formation. The bottle was immediately capped and inverted five times to evenly mix the seed. The solution was then left overnight in the water bath to react.
After overnight synthesis, the absorption spectrum of the nanoparticle solution was measured using a custom spectrophotometer consisting of LS-1 light source and USB2000 fiber optic spectrometer (Ocean Optics, Duendin FL). The peak resonance wavelength and corresponding 95% confidence interval were determined by fitting the extinction spectra to a Gaussian function (Fig. 1(a)). Size distributions of each batch of nanorods were measured using a transmission electron microscope (FEI Tecnai G2 Twin). Samples were prepared by placing a 10 μL drop of nanorod solution onto a formvar-coated copper TEM grid (FCF200-Cu, Electron Microscopy Sciences). The rods were allowed to settle on the grid for 15 minutes before gently removing the excess fluid with a tissue. Grids were then placed in the TEM for imaging (Fig. 1(b)).
Fig. 1 (a) TEM images of GNRs. The average nanoparticle size is 35.0 x 17.5 nm (aspect ratio = 2). Scale bar = 100nm. (b) Absorption spectra from GNR solution. The Gaussian fit to the spectra has a peak wavelength at 603nm.
2.2 Antibody-GNR conjugation
GNRs were conjugated to anti-EGFR antibodies using previously described methods [10,11,29]. Briefly, 1 ml of GNR solution was centrifuged twice (10,000 RPM for 5 minutes) and resuspended in 1 mL of 1mM NaCl. Polystyrene sulfonate (200 μL, 10 mg/mL in 1 mM NaCl, MW 18,000, Polysciences, Inc.) was added to the GNR solution. The suspension was then placed on a shaker for 20 minutes. The GNRs were centrifuged at 10,000 RPM for 5 minutes and resuspended in 1mL of 20 mM HEPES pH 7.4. 4 μL of anti-EGFR antibody (E2156, clone 225, Sigma-Aldrich, St. Louis, MO, 1.56mg/ml stock solution) were added, and the suspension was placed on a shaker for 30 minutes. The GNRs were centrifuged at 10,000 RPM for 5 minutes and finally resuspended in 50uL of PBS containing 5mg/ml BSA.
2.3 GFP cell transfection
EGFR + 270-GBM cells were transduced to express green fluorescent protein (GFP, 488nm excitation, 508nm emission) using the TransLenti Viral GIPZ Packaging System (Thermo Scientific). The pGIPZ Non-Silencing Control Vector was used to produce lentivirus containing GFP DNA from TLA-HEK293T cells. Virus containing medium was introduced into cultures of 270-GBM cells. After three days of transfection, 270-GBM cells were selected for GFP expression by introducing 0.8μg/mL of puromycin into the culture media for a three week period. Puromycin is an antibiotic normally toxic to eukaryotic cells. Infected genes also encode a puromycin resistance gene, which allows cells to survive in the presence of the antibiotic. Incubation with puromycin therefore actively selects cells with GFP expression.
2.4 Tumor growth
Athymic nude mice were anesthetized with ketamine/xylazine (100/10 mg/kg i.p.). Once anesthetized, mice were placed in a stereotaxic frame that secures the head and the surgical area was disinfected. A 1-cm vertical incision was made over the right hemisphere and the muscle and periosteum were removed. A hand drill was used to create a small hole approximately 2 mm to the right of the mid-sagittal suture and 2-3 mm to the back of the lambdoid suture. Tumor cells were injected through the hole using the z-axis of the stereotaxic frame to ensure location. 1x106 270-GFP human GBM tumor cells (1x105/μL) were injected at a depth of 5 mm below the skull surface. After cell injection, the exposed hole and skin were sealed using cyanoacrylate glue.
2.5 Tissue-nanorod incubation
After allowing the tumors to grow for two weeks, the mice were sacrificed, and their brains were sliced in half, exposing the tumor in each slice. A total of four tumor-containing slices were taken from 2 mice.
A slice of surgical gel foam(Surgifoam® Ethicon) was submerged in 50μL of 2.0x10−10M anti-EGFR GNR solution until it was saturated with nanoparticles (approximately one minute). The GNR laden gel foam was placed on top of the brain slice, where it conformed to the tissue’s surface (Fig. 2) The gel foam was left on the tissue for 20 minutes at 37°C and 5% CO2, allowing for enough time for GNRs to specifically bind to the tumor. Three PBS washes were then used to remove any unbound particles from the tissue.
Fig. 2 (a) Brain slice containing GBM270 tumor. (b) The same brain slice covered with anti-EGFR GNR laden gel foam.
2.6 Hyperspectral darkfield imaging
Tumor slices were imaged using a custom hyperspectral darkfield/fluorescence optical imaging system (Fig. 3). Briefly, for darkfield imaging, a darkfield LED ring (RL1360, Advanced Illumination) illuminates the sample from an oblique angle outside of the NA of the 10x objective (NA = 0.25) used for imaging. Therefore, only scattered light, from the tissue and nanoparticles, is collected by the objective. The scattered light is passed through an acousto-optic tunable filter-based hyperspectral imaging system (HSI-300, Gooch and Housego). The HSI-300 system filters the image with a central wavelength anywhere within the 450-800nm range, and focuses the resulting image onto an EM-CCD (Photonmax 512B, Princeton Instruments) The center wavelength of the HSI-300 was swept throughout the visible spectrum (450-800nm, 10nm steps), with an image recorded at each step. The bandwidth at each wavelength ranged from 13.2 to 33.4nm. Full spectral scans were recorded in 3.6 seconds. The resulting hyperspectral data cube displays the spectral absorption characteristics of the sample, including the spectrally distinct signatures of blood and nanoparticles. Example darkfield images are displayed in Fig. 4. The hyperspectral imaging approach detects GNR absorption with spatial resolution of approximately ~1μm.
Fig. 3 Schematic of the hyperspectral darkfield/fluorescence microscope. Dashed components are only in place during fluorescence imaging.
Fig. 4 Darkfield Images of normal and malignant tissues taken at 550nm and 620nm. (+/−) signs indicate the presence of significant absorption signal from blood or gold nanorods. Scale bar = 250 μm
For fluorescence imaging, white light from a xenon arc lamp is filtered at 470nm and illuminates the sample via epi-illumination. The resultant fluorescent light is imaged onto the HSI-300, which filters the image at the fluorophore’s peak emission wavelength (508nm). As the sample does not move between data acquisitions across the two imaging modalities, darkfield and fluorescent images correspond to the same areas, and may be used to determine the spatial accuracy of the GNR label. The system has a field of view of 0.78mm x 0.78mm. The sample was scanned vertically and/or horizontally by approximately this value for each hyperspectral scan to ensure that the new field of view did not overlap with the previous hyperspectral image. Whole fields of view were processed for tissue classification, unless a border of the tumor was detected by the GFP fluorescence. In this case one GFP + region and one GFP- region from the FOV were independently analyzed. Approximately, the entire tissue area contacted by the gel foam was imaged and analyzed. A total of 177 areas were measured from 4 brain slices.
2.7 Image processing and classification
In the described mouse model, the malignant GBM cells express GFP, while healthy cells do not fluoresce. Therefore, fluorescence imaging acts as a gold standard in determining whether a given region is normal or malignant. For each region imaged, the GFP-based classification is determined by quantifying the average GFP fluorescence across the region compared to a threshold value. The threshold was determined by quantifying the distribution of GFP signal intensities through all areas examined. Non-fluorescent areas produce a narrow distribution of low GFP intensity which can be effectively separated with a proper threshold.
To assess the nanoparticle contrast, images corresponding to the GNRs’ peak resonance wavelength were selected and initially high-pass filtered (Gaussian frequency domain filter, std = 3.12 μm−1) to enhance the contrast created by GNR absorption. An intensity threshold of three standard deviations over the mean of the background signal, determined from a region containing no GNRs, was then applied to the selected image to create a map which identified the locations of bound GNRs. A second threshold was used to highlight absorbing molecules by requiring pixels before the high-pass filter to be lower than the selected background, thereby rejecting scattering objects. Areas identified as containing GNRs which were larger than 800μm2 were considered to be aggregates and were rejected, as large aggregates may likely be nonspecifically adsorbed to the tissue. Also, single pixels were removed from the map, as they are likely associated with noise. Each tissue image is classified as either normal or containing malignant cells based on the density, or area fraction, of nanoparticles in the region. Each region may then be categorized as either a (GFP+/GNR+), (GFP-/GNR-), (GFP-/GNR+) or (GFP+/GNR-). These classifications are used to calculate the kappa statistic, which measures the overall agreement between GFP and GNR classifications, throughout a range of GNR area fraction thresholds. This procedure is demonstrated in Fig. 5.
Fig. 5 Images displaying the image processing methods used to classify tissue regions as malignant ( + ) or normal (-). Average fluorescence intensities determine the true state of the region. Predictions are determined based on the area fraction of nanoparticles in the binary map. Scale bar = 250μm.
3. Results
3.1 Gold nanorod synthesis and characterization
TEM measurements (Fig. 1(a)) show that the average size of the GNRs is 35.0 ± 18.2 x 17.5 ± 5.2 nm (aspect ratio = 2.0 ± 0.9). Shorter nanorods provide enhanced absorption in the visible portion of the tissue optical window. While longer nanorods exhibit similar absorption in the NIR range, contrast in the visible range is preferable as it may be more easily understood by a surgeon observing GNR labeled tissues, especially if a color CCD is utilized in a clinical system. As previously described, the absorption spectrum of the gold nanorod solution was measured (Fig. 1(b)). The Gaussian fit to this data indicates a peak absorption wavelength of 603 ± 0.6 nm, where the error refers to the 95% confidence interval of the fit. Normally, the absorption peak of GNR solutions contain two peaks corresponding to the particles’ transverse and longitudinal plasmon resonances. For this particular GNR species, the transverse peak is masked by the longitudinal peak due to (i) the minimal redshift in the longitudinal peak compared to the transverse peak associated with the nanorods’ short aspect ratio and (ii) the broadening of the longitudinal peak due to the inhomogeneity of the nanorod population. For these reasons, only one peak associated with the GNRs’ longitudinal plasmon resonance is seen in the absorbance spectra.
3.2 Spectral contrast
Four brain slices were prepared for imaging with anti-EGFR GNRs. A total of 177 areas were imaged at 10x magnification from within and around the tumor. The hyperspectral data acquired from each site highlight the differences in spectral absorption characteristics between the blood and gold nanorods (Fig. 4). Both blood and GNRs absorb light emitted from the highly scattering thick tissue background. Absorption from the blood is detected at significant levels at lower wavelengths (~550nm), but drops off to nearly undetectable levels near the nanorods’ peak plasmon resonance wavelength (~620nm). In contrast, the nanorod absorption can be detected throughout the majority of the visible range, but becomes insignificant within in the near infrared wavelength range (>~750nm). The separation of the absorption spectra between blood and GNRs allows for the detection of bound GNRs while eliminating any confounding signal originating from the blood. In this study, this is accomplished by simply analyzing nanoparticle absorption at 620nm, where absorption from the blood is minimal. At this wavelength, GNRs are detected with an estimated SNR of 15.2 dB over fluctuations within the background.
3.3 GBM classification
GFP + sections of tissue displayed an average GNR area fraction of 2.9 ± 1.9%, while GFP- sections displayed an average GNR area fraction of 0.27 ± 0.62%. A Wilcoxon rank sum test shows that the distributions of GNR area fractions from these two populations are significantly distinct (p<0.001). An appropriate GNR area fraction threshold can be selected to effectively classify normal and GBM tissue.
To determine the optimal threshold, the kappa statistic comparing GFP and GNR classifications was evaluated for a range of nanoparticle area fraction thresholds used to make the disease state predictions (Fig. 6). Statistics were calculated using nanoparticle area fraction thresholds ranging from 0% to 7.6%. At the optimal point, corresponding to a nanoparticle area fraction of 0.43%, the kappa statistic was found to have a maximum value of 0.75, indicating substantial agreement between the GFP and GNR classification methods [30]. Except for a few statistical outliers, GFP- sections generally have GNR area fractions less than or equal to 1%. Using this measure as our threshold, we obtain a kappa statistic of 0.65, maintaining good agreement between the two classification methods. Using the threshold of 1% would effectively increase the agreement between these methods when classifying normal tissues by minimizing false positives in GNR classifications. Correctly identifying normal tissue is especially important in brain tumor delineation, where preserving functional tissue is a key concern. Therefore, reducing the overall agreement may be necessary in order to ensure that normal tissues are identified with high accuracy.
Fig. 6 Plot of the kappa statistic comparing GFP and GNR classifications as a function of GNR area fraction threshold.
Two additional brain slices were labeled with GNRs conjugated to non-specific IgG antibodies as a control. Ideally, none of these GNRs should bind onto the tissue. However, a minimal amount of particles were observed to bind to both the tumor and healthy tissue. This is most likely due to blood components interfering with the electrostatic bond between the antibody and the gold surface, causing non-specific adsorption. At the optimal nanoparticle area fraction for anti-EGFR GNRs (0.43%), the kappa statistic for comparing nonspecific GNRs and GFP is −0.04. This indicates that the probability of the nonspecific GNR and GFP classifications agreeing is almost equivalent to random chance. These control measurements demonstrate that the anti-EGFR GNRs are specifically bound to the EGFR-expressing GBM tumor.
4. Discussion
One of the greatest challenges in using GNPs for brain tumor delineation is delivery of the particles to the area of interest. The blood brain barrier (BBB) functions to protect the brain from foreign objects, allowing only the smallest nanoparticles (<10nm in diameter) to pass [31]. This makes the introduction of nanoparticles to the brain via intravenous injection very difficult. Researchers have shown that the BBB can be temporarily disrupted using focused ultrasound, allowing therapeutic drugs and nanoparticles to pass into the brain [32,33]. Even though GNPs may be temporarily allowed to pass through the BBB, the majority of particles injected intravenously are sequestered in the liver and spleen, or remain in the blood [31], leaving only a small percentage of particles which cross the BBB and can be used for brain tumor delineation. Choi et al. have also demonstrated a “Trojan Horse” strategy in which nanoparticle-loaded monocytes are chemically attracted to the necrotic cores of tumors [34]. While such a system is capable of bypassing the BBB and enabling ablation of the necrotic core, the macrophages may not migrate to the tumor margins at high enough concentrations for delineation. We have chosen to use a topical nanoparticle application using surgical gel foam in order to avoid these delivery issues.
The gel foam is usually used as an absorbable hemostat during neurosurgical procedures. It may also be soaked in thrombin before application in order to increase the hemostatic effect. This procedure provided inspiration for our method for GNR topical application. The gel foam was soaked with anti-EGFR GNR solution and applied directly onto the brain slice. Over the 20 minute incubation period, nanorod solution spreads out locally onto the brain slice. The foam allows for the delivery of GNR solution to the regions of interest without spilling over into other areas. As hemostatic gel foam is a tool already used by surgeons, this technique may easily be adapted into the surgical setting. The 20 minute incubation time used in this experiment was based on previous cell studies [10,11]. Future experiments will determine the minimum incubation time needed for optimal labeling. Reduced incubation times will make this GNR labeling procedure more efficient for surgical applications.
The kappa statistics shown in Fig. 6 was produced by analyzing 0.78x0.78mm areas of the tissue (except for frames including tumor borders, in which areas from either side of the border were analyzed). Ideally, during surgical resection, GNRs could be imaged over a larger area, allowing for delineation of larger regions of tissue simultaneously. Puvanakrishnan et al. demonstrated wide-field imaging of GNR absorbance on solid SCC tumors, although single GNRs were not resolved [27]. Therefore, it is feasible to image high densities of bound GNRs with lower resolution wide field imaging.
By optimizing the nanoparticle area fraction threshold for the selected field of view, we were able to classify GBM with high agreement between GFP and GNR methods (κ = 0.75). The agreement between classification methods decreases when analyzing areas smaller than 0.78x0.78 mm. This is due to the random distribution of nanoparticles binding to the tissue surface. There are three main reasons for this randomized binding. First, the distribution of nanoparticles diffusing out from the gel foam may not necessarily be uniform. The density of nanoparticles reaching certain locations of the tissue can depend on factors such as the level of contact of the foam with that tissue area and local variations in density of GNRs in the gel foam. Second, the EGFR expression of 270-GBM cells is not uniform across the population. GBM areas not densely labeled by GNRs may be associated with cells with lower EGFR expression. Third, although 270-GBM cells in the mouse model were selected for GFP expression, GFP expression may not be constant over the population. For example, it is possible for a subset of cells to express the puromycin resistance gene without expressing GFP as well. If GNRs bound to such cells, they would produce a disagreement between the classification methods even though the label is binding to the appropriate cell receptor, since no GFP emission would be observed. A more consistent method, such as H&E staining would provide a more accurate measure of the true tumor margins, allowing for measurements of the true GNR label sensitivity and specificity, although this could not be done simultaneously with the GNR darkfield measurements. While these issues make delineation at the micron scale difficult, this label can achieve accurate classification at spatial scales which are in line with the practical limit of a surgeon’s accuracy.
The kappa statistic of this label was seen to be fairly good in this study but was limited by several factors. Due to the local diffusion of nanorods onto the surrounding tissue, the contact achieved when applying the GNR-laden gel foam is ultimately a limiting factor. Potential classification disagreements were observed at the edges of the GFP + tumor which were not properly covered with the gel foam during application, as can be seen in Fig. 2(b). This may be improved upon by using sections of foam larger than the region of interest or employing multiple sections of gel foam. Another complicating factor is that components of blood such as cysteine can displace the targeting molecules bound to the GNRs, leading to protein adsorption [35]. Such a process could cause GNRs to nonspecifically aggregate onto the tissue surface, regardless of EGFR expression, potentially causing some classification disagreements with GFP measurements. Although this was not directly measured, it is estimated that approximately 10%-15% of the images recorded contained one or more aggregates that were rejected. This may be improved upon by using a conjugation method based on bonds between the GNR and nonionic or zwitterionic molecules, to which the serum proteins do not adsorb [36]. These additional advancements could help to improve both the molecular specificity and spatial accuracy of the anti-EGFR GNR label, increasing its utility as a molecular contrast agent for brain tumor delineation.
5. Conclusion
In this study, the potential of anti-EGFR GNRs to delineate primary malignant brain tumors was evaluated. Topical application of the GNRs using surgical gel foam delivers the molecular label locally to the solid tissue, avoiding concerns with systemically delivering sufficient quantities for imaging across the blood brain barrier. A simple imaging system highlights the enhanced spectrally dependent absorption associated with the GNRs, pointing the way for a low cost, low complexity surgical tool. By imaging at a wavelength corresponding to the GNRs’ plasmon resonance, absorptive contrast from the nanoparticles is increased while minimizing absorption from blood, which can be a confounding influence. The GNR area fraction can be used to classify sub-millimeter tissue regions with substantial agreement with GFP fluorescence. This study shows that anti-EGFR GNRs label GBM tumors with high molecular specificity and spatial accuracy, indicating that immunolabeled GNRs show potential to be an effective tool for delineating brain tumor margins in vivo within the surgical setting. This translational strategy has the potential of improving our ability to maximize the extent of resection of malignant brain tumors which may in turn improve patient survival from this devastating disease.
Acknowledgments
KS acknowledges support from the NIH Training Grant (T32-EB001040) and the John T. Chambers Scholar Program (398-1026).
References and links
1. N. J. Ullrich and S. L. Pomeroy, “Pediatric brain tumors,” Neurol. Clin. 21(4), 897–913 (2003). [CrossRef] [PubMed]
2. D. N. Louis, H. Ohgaki, O. D. Wiestler, W. K. Cavenee, P. C. Burger, A. Jouvet, B. W. Scheithauer, and P. Kleihues, “The 2007 WHO classification of tumours of the central nervous system,” Acta Neuropathol. 114(2), 97–109 (2007). [CrossRef] [PubMed]
3. Cancer - United States Cancer Statistics (USCS) Data - 2009 Cancer Types Grouped by Race and Et,” http://apps.nccd.cdc.gov/uscs/cancersbyraceandethnicity.aspx.
4. N. R. Smoll, K. Schaller, and O. P. Gautschi, “Long-term survival of patients with glioblastoma multiforme (GBM),” J. Clin. Neurosci. 20(5), 670–675 (2013). [CrossRef] [PubMed]
5. P. Schmalz, M. Shen, and J. Park, “Treatment resistance mechanisms of malignant glioma tumor stem cells,” Cancers 3(4), 621–635 (2011). [CrossRef]
6. S. K. Ray, ed., Glioblastoma (Springer New York, 2010).
7. G. E. Keles, B. Anderson, and M. S. Berger, “The effect of extent of resection on time to tumor progression and survival in patients with glioblastoma multiforme of the cerebral hemisphere,” Surg. Neurol. 52(4), 371–379 (1999). [CrossRef] [PubMed]
8. S. A. Maier, Plasmonics: Fundamentals and Applications (2007).
9. K. L. Kelly, E. Coronado, L. L. Zhao, and G. C. Schatz, “The optical properties of metal nanoparticles: the influence of size, shape, and dielectric environment,” J. Phys. Chem. B 107(3), 668–677 (2003). [CrossRef]
10. K. Seekell, H. Price, S. Marinakos, and A. Wax, “Optimization of immunolabeled plasmonic nanoparticles for cell surface receptor analysis,” Methods 56(2), 310–316 (2012). [CrossRef] [PubMed]
11. K. Seekell, M. J. Crow, S. Marinakos, J. Ostrander, A. Chilkoti, and A. Wax, “Hyperspectral molecular imaging of multiple receptors using immunolabeled plasmonic nanoparticles,” J. Biomed. Opt. 16(11), 116003 (2011). [CrossRef] [PubMed]
12. M. J. Crow, K. Seekell, J. H. Ostrander, and A. Wax, “Monitoring of receptor dimerization using plasmonic coupling of gold nanoparticles,” ACS Nano 5(11), 8532–8540 (2011). [CrossRef] [PubMed]
13. K. Sokolov, M. Follen, J. Aaron, I. Pavlova, A. Malpica, R. Lotan, and R. Richards-Kortum, “Real-time vital optical imaging of precancer using anti-epidermal growth factor receptor antibodies conjugated to gold nanoparticles,” Cancer Res. 63(9), 1999–2004 (2003). [PubMed]
14. A. Wax and K. Sokolov, “Molecular imaging and darkfield microspectroscopy of live cells using gold plasmonic nanoparticles,” Laser Photon. Rev 3(1-2), 146–158 (2009). [CrossRef]
15. S. Kumar, J. Aaron, and K. Sokolov, “Directional conjugation of antibodies to nanoparticles for synthesis of multiplexed optical contrast agents with both delivery and targeting moieties,” Nat. Protoc. 3(2), 314–320 (2008). [CrossRef] [PubMed]
16. C. J. Murphy, A. M. Gole, J. W. Stone, P. N. Sisco, A. M. Alkilany, E. C. Goldsmith, and S. C. Baxter, “Gold nanoparticles in biology: beyond toxicity to cellular imaging,” Acc. Chem. Res. 41(12), 1721–1730 (2008). [CrossRef] [PubMed]
17. E. E. Connor, J. Mwamuka, A. Gole, C. J. Murphy, and M. D. Wyatt, “Gold nanoparticles are taken up by human cells but do not cause acute cytotoxicity,” Small 1(3), 325–327 (2005). [CrossRef] [PubMed]
18. R. A. Sperling, P. Rivera Gil, F. Zhang, M. Zanella, and W. J. Parak, “Biological applications of gold nanoparticles,” Chem. Soc. Rev. 37(9), 1896–1908 (2008). [CrossRef] [PubMed]
19. P. K. Jain, K. S. Lee, I. H. El-Sayed, and M. A. El-Sayed, “Calculated absorption and scattering properties of gold nanoparticles of different size, shape, and composition: applications in biological imaging and biomedicine,” J. Phys. Chem. B 110(14), 7238–7248 (2006). [CrossRef] [PubMed]
20. M. J. Crow, K. Seekell, and A. Wax, “Polarization mapping of nanoparticle plasmonic coupling,” Opt. Lett. 36(5), 757–759 (2011). [CrossRef] [PubMed]
21. A. Curry, G. Nusz, A. Chilkoti, and A. Wax, “Substrate effect on refractive index dependence of plasmon resonance for individual silver nanoparticles observed using darkfield microspectroscopy,” Opt. Express 13(7), 2668–2677 (2005). [CrossRef] [PubMed]
22. M. D. Marmor, K. B. Skaria, and Y. Yarden, “Signal transduction and oncogenesis by ErbB/HER receptors,” Int. J. Radiat. Oncol. Biol. Phys. 58(3), 903–913 (2004). [CrossRef] [PubMed]
23. R. I. Nicholson, J. M. Gee, and M. E. Harper, “EGFR and cancer prognosis,” Eur. J. Cancer 37(Suppl 4), S9–S15 (2001). [CrossRef] [PubMed]
24. H. K. Gan, A. H. Kaye, and R. B. Luwor, “The EGFRvIII variant in glioblastoma multiforme,” J. Clin. Neurosci. 16(6), 748–754 (2009). [CrossRef] [PubMed]
25. D. M. Peereboom, D. R. Shepard, M. S. Ahluwalia, C. J. Brewer, N. Agarwal, G. H. J. Stevens, J. H. Suh, S. A. Toms, M. A. Vogelbaum, R. J. Weil, P. Elson, and G. H. Barnett, “Phase II trial of erlotinib with temozolomide and radiation in patients with newly diagnosed glioblastoma multiforme,” J. Neurooncol. 98(1), 93–99 (2010). [CrossRef] [PubMed]
26. M. J. Crow, G. Grant, J. M. Provenzale, and A. Wax, “Molecular imaging and quantitative measurement of epidermal growth factor receptor expression in live cancer cells using immunolabeled gold nanoparticles,” Am. J. Roentgenol. 192(4), 1021–1028 (2009). [CrossRef] [PubMed]
27. P. Puvanakrishnan, P. Diagaradjane, S. M. S. Kazmi, A. K. Dunn, S. Krishnan, and J. W. Tunnell, “Narrow band imaging of squamous cell carcinoma tumors using topically delivered anti-EGFR antibody conjugated gold nanorods,” Lasers Surg. Med. 44(4), 310–317 (2012). [CrossRef] [PubMed]
28. B. Nikoobakht and M. A. El-Sayed, “Preparation and growth mechanism of gold nanorods (NRs) using seed-mediated growth method,” Chem. Mater. 15(10), 1957–1962 (2003). [CrossRef]
29. X. Huang, I. H. El-Sayed, W. Qian, and M. A. El-Sayed, “Cancer cells assemble and align gold nanorods conjugated to antibodies to produce highly enhanced, sharp, and polarized surface Raman spectra: a potential cancer diagnostic marker,” Nano Lett. 7(6), 1591–1597 (2007). [CrossRef] [PubMed]
30. A. J. Viera and J. M. Garrett, “Understanding interobserver agreement: the kappa statistic,” Fam. Med. 37(5), 360–363 (2005). [PubMed]
31. W. H. De Jong, W. I. Hagens, P. Krystek, M. C. Burger, A. J. A. M. Sips, and R. E. Geertsma, “Particle size-dependent organ distribution of gold nanoparticles after intravenous administration,” Biomaterials 29(12), 1912–1919 (2008). [CrossRef] [PubMed]
32. K. Hynynen, N. McDannold, N. A. Sheikov, F. A. Jolesz, and N. Vykhodtseva, “Local and reversible blood-brain barrier disruption by noninvasive focused ultrasound at frequencies suitable for trans-skull sonications,” Neuroimage 24(1), 12–20 (2005). [CrossRef] [PubMed]
33. K. F. Bing, G. P. Howles, Y. Qi, M. L. Palmeri, and K. R. Nightingale, “Blood-brain barrier (BBB) disruption using a diagnostic ultrasound scanner and definity in mice,” Ultrasound Med. Biol. 35(8), 1298–1308 (2009). [CrossRef] [PubMed]
34. M.-R. Choi, K. J. Stanton-Maxey, J. K. Stanley, C. S. Levin, R. Bardhan, D. Akin, S. Badve, J. Sturgis, J. P. Robinson, R. Bashir, N. J. Halas, and S. E. Clare, “A cellular Trojan Horse for delivery of therapeutic nanoparticles into tumors,” Nano Lett. 7(12), 3759–3765 (2007). [CrossRef] [PubMed]
35. T. A. Larson, P. P. Joshi, and K. Sokolov, “Preventing protein adsorption and macrophage uptake of gold nanoparticles via a hydrophobic shield,” ACS Nano 6(10), 9182–9190 (2012). [CrossRef] [PubMed]
36. A. K. Murthy, R. J. Stover, W. G. Hardin, R. Schramm, G. D. Nie, S. Gourisankar, T. M. Truskett, K. V. Sokolov, and K. P. Johnston, “Charged gold nanoparticles with essentially zero serum protein adsorption in undiluted fetal bovine serum,” JACS135(21), 7799–7802 (2013).
