Aggregation of nanoparticles in endosomes and lysosomes produces surface-enhanced Raman spectroscopy

Abstract. The purpose of this study was to explore the use of surface-enhanced Raman spectroscopy (SERS) to image the distribution of epidermal growth factor receptor (EGFR) in cells. To accomplish this task, 30-nm gold nanoparticles (AuNPs) tagged with antibodies to EGFR (1012  per mL) were incubated with cells (106  per mL) of the A431 human epidermoid carcinoma and normal human bronchial epithelial cell lines. Using the 632.8-nm excitation line of a He-Ne laser, Raman spectroscopy measurements were performed using a point mapping scheme. Normal cells show little to no enhancement. SERS signals were observed inside the cytoplasm of A431 cells with an overall enhancement of 4 to 7 orders of magnitude. Raman intensity maps of the 1450 and 1583  cm−1 peaks correlate well with the expected distribution of EGFR and AuNPs, aggregated following uptake by endosomes and lysosomes. Spectral features from tyrosine and tryptophan residues dominate the SERS signals.


Introduction
A number of cancers are characterized by overexpression of the epidermal growth factor receptor (EGFR), a membrane protein which mediates cell growth, proliferation, and differentiation in multiple tissues. Many epithelial tumors have been found to express high numbers of EGF receptors, [1][2][3][4][5][6][7] and the receptor levels are associated with poor clinical prognosis in cancers of the bladder, 4 lung, 5 and breast. 4,6 Antibodies for EGFR tagged with fluorescent probes have been used as contrast agents to image EGFR overexpression. 8 The EGFR-induced shift in the plasmon resonance has also been used as an image contrast method. 9 Both of these methods provide at most a 10-fold intensity contrast factor.
To overcome the limited contrast, we explore the use of surface-enhanced Raman spectroscopy (SERS). By exciting a sample with laser light, energy is transferred to (Stokes process) or lost (anti-Stokes process) by the medium in an inelastic scattering process referred to as Raman spectroscopy. The scattered light is a fingerprint of the set of atomic vibrations, which is unique to the arrangement and types of atoms under study. In general, 1 in 10 6 photons is scattered inelastically in this manner. By tuning the incident laser frequency to the plasma frequency of the nanoparticle assembly, an overall enhancement of 7 to 8 orders of magnitude of the scattered signal is possible; 10 this is referred to as the surface-enhanced Raman scattering effect. Over the past few years, there has been a flurry of activity in the application of Raman spectroscopy to problems in biology. Sha et al. 11 have reviewed the field of the application of SERS to oncology. The idea of using SERS to image cancerous tissue was exploited by Kah et al. 12 A self-assembled SERS-active, gold nanoparticle (AuNP) monolayer film tagged with anti-EGFR was used to distinguish cancerous and normal cells in saliva for early diagnosis of oral cancer. Qian et al. 13 have also employed the SERS to target EGFR overexpressing cells in vivo using mouse xenograft tumor models. Approximately 1.4 to 1.5 × 10 4 reporter molecules (e.g., malachite green) are adsorbed on each 60-nm gold particle and stabilized with 3.0 × 10 4 thiol-polyethyleneglycol (thiol-PEG). Their SERS particles exhibit a signal that is 2 orders of magnitude larger than quantum dots, highlighting the power of the technique. Keren et al. 14 have demonstrated the multimodal SERS detection using antibody-tagged Oxonica nanoparticles. Kim et al. 15 developed the antibody-conjugated, silver nanoparticles-embedded silica spheres to target HER2 and CD10 on cellular membrane constructs. None of these reports have mapped the distribution and fate of the nanoparticles within a single cell. Our study is an extension of an earlier contribution 16 and aims to determine the microscopic distribution of antibody-tagged AuNPs in order to map the cellular distribution of EGFR in both normal and cancerous cell cultures.

Gold Nanoparticle Synthesis
The AuNPs used in the experiment were synthesized using the well-known method of Turkevich et al. 17 and Frens. 18 A 0.01% by weight gold chlorate solution was heated with stirring to boiling in an extremely clean round bottomed flask fitted with a refluxer in a fume hood. The flask was heated in a bath of silicon oil to ensure even heat distribution. A precise volume of 1% by weight sodium citrate was added to reduce the gold salt with continued boiling. The amount of sodium citrate added to the stirring gold solution ultimately determined the final size of the AuNPs, with smaller volumes of citrate producing larger particles.

Gold Nanoparticle Characterization
Three techniques were used to characterize the AuNPs and their derivatives: UV-visible spectroscopy (UV-Vis), transmission electron microscopy (TEM), and dynamic light scattering (DLS). From UV-Vis spectroscopy, the approximate size, plasmon absorbance wavelength, and concentration of the nanoparticles were obtained. The UV-Vis analysis was performed using a UV-1700 (Shimadzu) or ND-1000 (Nanodrop) spectrophotometer. The concentration of nanoparticles can be calculated using the absorbance maximum: where A is the absorbance (unitless), N is Avogadro's number (6.02 × 10 23 molecules∕mole), σ is the extinction cross-section (cm 2 ), b is the path length (cm), and c is the nanoparticle concentration (NPs∕mL). The extinction cross-section for 30-nm AuNPs is 3 × 10 −11 cm 2 . The TEM was used to determine the particle size distribution. Approximately 10 to 20 μL of the nanoparticle solution was pipetted onto a copper backed carbon TEM grid and allowed to air dry overnight. A Hitachi H-8000 TEM, operating at 200 kV, was used to image the nanoparticles at a magnification of 120000×. The DLS was also used to characterize the particle size distribution. An ALV DLS/SLS apparatus was used to measure the samples. Once the diffusion coefficient (D) is known, the radius (R) of the particles can be calculated using the Stokes-Einstein equation: D ¼ k B T∕6πηR, where k B is the Boltzmann constant, T is the temperature in Kelvin, and η is the viscosity.

Antibody Conjugation to Gold Nanoparticles
To test the effectiveness of anti-EGFR antibodies, a series of experiments were run. Antibodies specific for human EGFR, produced in mice (Sigma-Aldrich, E2156, Monoclonal anti-EGFR-human, clone 225, purified mouse immunoglobin, Immunogen-partially purified EGF receptors from human A431 cells), IgG polyclonal nonspecific antibodies (Sigma Aldrich, I8765, Mouse IgG-technical grade from normal mouse serum), and antibody free solutions were incubated with nanoparticles.

Cell Preparation
Two types of cells were used: EGFR overexpressing human epidermoid carcinoma A431 cell line from ATCC and normal human bronchial epithelial (NHBE) cells from Lonza. The antibody-conjugated AuNPs (½AuNP ¼ 10 12 per mL) were then attached to the cells (½cells ¼ 10 6 per mL), leading to an approximate ratio of 10 6 AuNPs∕cell. In a biosafety cabinet, the prepared antibody solution was pipetted into a dish of cells where the cells had been grown on a coverslip. The cells were counted to give no more than 0.5 million for an experiment.
The solution was incubated with the cells for 30 min and washed with sterile 50 mM HEPES. The coverslip was flipped into a clean glass bottomed Petri dish. A polymer spacer was pressed against the coverslip to secure it. The dish was filled with HEPES buffer (pH 7.4) and loaded into a custom-made Raman sample holder.

Raman Spectra
Raman spectra were obtained with a LabRam confocal micro-Raman system (Jobin Yvon Inc., Edison, New Jersey), which consists of a grating spectrograph coupled to an Olympus BX40 microscope with a motorized x-y stage. A 632.8-nm He-Ne laser excited the Raman spectra by focusing through a 100× objective with a coverslip correction collar. The incident power at the sample is 200 μW. The Raman scattered light was collected in a confocal arrangement by the same objective and spectrally dispersed using a Horiba Jobin Yvon Raman Spectrometer onto a thermoelectrically cooled CCD detector. The system also has a video system for white light image capture, a computer to collect and process data, and a motorized stage for translation along x, y, and z.
The system was calibrated with the silicon line at 520.7 cm −1 (at least 2000 and up to 9000 counts∕s of signal was achieved). A grating of 1800 grooves∕mm, a confocal pinhole of 400 μm, and an input slit width of 200 μm were used. Spectra were collected with a 1 s integration time, averaged over two scans. A step size of 1 μm or less was utilized, typically 0.3 μm, to map the spectral features of a cell. In this confocal arrangement, the depth resolution is 2 μm. The intensity distribution in a narrow spectral range was used to produce a Raman map of the cell of interest.

Nanoparticle Preparation and Characterization
To produce gold colloidal solutions, 1 mL of 1% by weight sodium citrate was added dropwise to a boiling solution of 100 mL 0.01% (by weight) gold salt at 100°C. The reaction is complete when the solution turns a reddish color (in 5 to 10 min). The gold colloidal solution was characterized by UV-Vis spectroscopy, TEM, and DLS. An analysis of the TEM images in Fig. 1(a) reveals particle diameters of approximately 30 AE 5 nm, while the DLS histogram [ Fig. 1(b)] implies an average particle size of approximately 34 nm. An increase is expected since DLS measures the hydrodynamic radius including the citrate capping molecule. UV-Vis absorption [ Fig. 1(c)] exhibits a maximum absorption at 532 nm for the bare nanoparticles (blue curve) and 536 nm for AuNPs tagged with anti-EGFR (pink curve). The latter curve is obtained after eliminating unattached anti-EGFR by centrifugation at 3000 rcf. The 4-nm red-shift between the maxima of the absorption spectra is consistent with the attachment of anti-EGFR to the AuNPs, which is ultimately due to the increase in the index of refraction of the medium surrounding the AuNPs.

Anti-EGFR-Tagged Gold Nanoparticles
The anti-EGFR-tagged AuNPs were aggregated by adding 0.01 M PBS (pH 7.4) to the gold colloidal solution. Figure 2 is the Raman spectra of the point indicated in the optical image of an aggregate (inset). The data were highly reproducible if the laser power, under the    19 reported the SERS spectra of anti-EGFR. The peaks reported by them are comparable with those found here ( Table 1). The absence of an amide I band near 1670 cm −1 is striking, as this feature is normally prominent in the nonSERS Raman spectra of proteins. This band requires the presence of the entire backbone in the enhancement area in a parallel orientation. Its absence implies that only the side chain groups were enhanced. Proteins are expected 23 to bind to the nanoparticles through those amino acids which form complexes with gold; among them sulfur containing cysteine, nitrogen containing histidine and tryptophan, and the basic amino acids lysine and arginine, at high pH above their isoelectric point (pI). The acids, aspartic and glutamic acids, would also be expected to adsorb when deprotonated, as would tyrosine through the oxygen atom side group following the loss of hydrogen. At the pH used in these experiments, 7.4, the deprotonated acids would be expected to bind to the nanoparticles, but not the bases. The results outlined in Table 1 confirm this hypothesis. The tentative peak assignments are taken from a variety of sources. Cysteine, histidine, tryptophan, Fig. 3 SERS spectra taken from four locations inside and outside the A431 cell indicated in Fig. 4(a). The intensity is in units of counts per second (CPS).
tyrosine, aspartate, and glutamate peaks dominate the spectra below 1800 cm −1 , while the highfrequency peaks are likely due to C─H and CH 3 vibrations of these amino acids. In addition, among these amino acids, few characteristic peaks appear below 400 cm −1 . The appearance of low-frequency features arises from the binding of sulfur, nitrogen, and oxygen to the AuNP; for which the magnitude of charge transfer decreases in the same order, sulfur-nitrogen-gold. This has the effect of reducing the contribution of charge-transfer effects to the resonance enhancement in SERS, reducing the strength of these modes, in the order of 302, 397, and 386 cm −1 . Six modes exhibit very strong enhancement, corresponding to vibrational frequencies of 1009 cm −1 (phenylalanine or tryptophan), 1131 cm −1 (symmetric C─C stretch), 1267 cm −1 (C─H bending modes), 1380 cm −1 (carbonyl symmetric stretch), and 2935 cm −1 (C─H stretch).

Raman Spectra of Cancerous and Normal Cells
A 0.5 mL of 10 6 per mL A431 and NHBE cells was pipetted into glass bottom Petri dishes and left to attach overnight (for A431 cells) or for 2 days (for NHBE cells) in an incubator (5% CO 2 , 37°C). 0.5 mL of the anti-EGFR-tagged AuNPs, at a concentration of 5 × 10 11 per mL, was added to the cells 30 min prior to the Raman experiments.
Raman spectra were collected by a point mapping scheme, with a step size of 0.3 μm along x and y. At each point, spectra are collected for 1 s, averaged over two scans. Figure 3 shows the Raman spectra following excitation at four locations of the cell, indicated in Fig. 4(a). All peaks in the spectra are due to SERS. The simple Raman spectrum without SERS is basically the same as the background noise (spectra 4). It is clear that the SERS spectra collected at different locations vary. Figure 4(b) is a SERS image obtained by mapping the intensity of the 1583 cm −1 peak, while Fig. 4(c) is a SERS image obtained by mapping the intensity of the 1450 cm −1 peak.
The  Fig. 3. This implies that the environment in the immediate vicinity of the AuNPs differs. Within endosomes, the EGFR is broken down into protein fragments which would produce significantly different spectra. When the image of any vibrational mode (e.g., 1450 or 1583 cm −1 ) in a given Raman spectrum is mapped, one observes little change in the Raman image (i.e., its spatial intensity distribution). While we have presented only two examples, this observation appears to be generally true. Selection of any other mode in a given spectra gives rise to a similar image, implying that the Raman image is consistent with the distribution of nanoparticle aggregates. Table 2 lists the observed Raman shifts for each of points 1 to 3 in Fig. 3. Most of the peaks are common to anti-EGFR, as listed in Table 1. There are additional peaks which can be mostly attributed to the appearance of tyrosine and tryptophan residues, and C─C and C═O stretches of amino acids. New peaks appear at points internal to the cell, where the lysozyme would begin breakdown of the protein, freeing those that are already bound to the nanoparticle to make an even stronger connection and undergo reorientation. The conclusion can be drawn that there is a  Fig. 4(d). The spectra are very similar. LLB53A refers to the sample nomenclature used in our lab (LL-Leanne Lucas' book B, page 53 sample A). The spectral signature of the anti-EGFR antibody is present outside the cell, confirming that the nanoparticles retain their anti-EGFR coating. Table 2 Peak positions, relative strength-ST, and tentative assignments for the SERS spectra in the vicinity of nanoparticles, located within or on the surface of A431 cells. Peak positions obtained from Lorentzian fits (very strong-vs, strong-s, moderate-m, weak-w, and very weak-vw). Trp-tryptophan, Tyr-tyrosine, Phe-phenylalanine, Arg-arginine, and Pro-proline. Point  rearrangement of the amino acids upon endocytosis to make tryptophan and tyrosine residues, which both contain ring systems, more prominent in the spectra. One can estimate the enhancement by considering the Raman signal generated by a bulk sample of anti-EGFR and the reduction in the number of anti-EGFR molecules contained within an AuNP aggregate. The numerical value of the enhancement factor is determined by comparing the Raman intensity per molecule in the presence or absence of nanoparticles with all other conditions being constant (e.g., laser power). Its value is given by E ¼ ðI SERS ∕I normal Þ × ðN normal ∕N SERS Þ, where N SERS is the number of analyte molecules in the illuminated volume in the presence of nanoparticles and N normal is the number of molecules in the illuminated volume in the absence of nanoparticles. We estimate that the enhancement ranges from 10 4 to 10 7 . The optical image [ Fig. 4(d)] clearly shows an aggregate of AuNPs, adhering to the Petri dish, that was not removed by the washing process. The spectra (Fig. 5) are very similar to the SERS anti-EGFR spectra of Fig. 2. The similarity of the spectra indicates that the nanoparticles retain their anti-EGFR coating. Other than this artifact, the Raman images and spectra are featureless and weak. Figure 6 is a plot of the Raman spectra from the cells in Figs. 7(a) and 7(d). These spectra are completely featureless and are comparable with the background intensity of the cell nucleus spectrum (Fig. 3, point 4).   intensity map at 1450 cm −1 for the nonspecific antibody-tagged AuNPs is shown in Fig. 7(e) and the 1583 cm −1 map is shown in Fig. 7(f). These maps are also similar, however, the spectrum at 1583 cm −1 seems to have slightly higher background intensity. The reduced intensity of the normal cell spectra

Conclusions
We have targeted EGFR in EGFR-overexpressing cells of the A431 human epidermoid carcinoma cell line (one million per milliliter) and in NHBE cells by binding anti-EGFR-tagged 30-nm AuNPs. We observed dramatic SERS effects in the A431 cells. As expected, control experiments with NHBE cells and with bare or nonspecific antibody-tagged AuNPs bound to A431 cells produced insignificant SERS enhancement. The spectra are associated with residues which have functional groups which associate with gold, including the sulfur containing cysteine and nitrogen containing histidine and tryptophan. Aspartic and glutamic acids would also be expected to adsorb when deprotonated, as would tyrosine through the oxygen atom side group following the loss of hydrogen. Evidence for rearrangement of these amino acids upon uptake into lysosome and endosomes is evident when comparing the SERS spectra of anti-EGFR-coated nanoparticles with spectra from cells. Tyrosine and tryptophan residues dominate. Raman signal enhancements are observed for A431 cells and reflected the AuNP distribution. AuNPs are located on the surface as well as on the interior of the cell, while being excluded from the nucleus. The EGFR is recycled by the cell through the process of endocytosis and is not found in the nucleus. Moreover, EGFR is known to distribute nonuniformly on the cell surface. Thus, the SERS map is consistent with the expected distribution of EGFR. The SERS enhancement factor estimated from our data is between 10 4 and 10 7 . Studies are underway to optimize the AuNP size and laser excitation wavelength. The NHBE cells do not show signal enhancements. A contrast factor of at least 100∶1 is estimated, which means that a SERS image should have orders of magnitude more contrast than a typical fluorescence image or one generated by white light scattering from nanoparticles. A SERS approach to imaging EGFR distribution is, therefore, very promising.
research has been focused on developing optical imaging and spectroscopy techniques for improving early cancer detection. He has published over 123 refereed journal papers, 1 book, 8 book chapters, and has 20 granted patents.
Patrick W. K. Lee is a professor and Cameron Chair in Basic Cancer Research at Dalhousie University. He received his PhD degree in biochemistry from the University of Alberta in 1978. He is the author of more than 100 journal articles. His current research focuses on oncolytic viruses, oncogenes, and tumor suppressors.
Kevin Cecil Hewitt is an associate professor of physics at Dalhousie University. He received his BSc degree in physics and biology from the University of Toronto and a PhD degree in physics from Simon Fraser University, in 2000. He is the author of more than 30 journal articles and has written 1 book chapter. His current research interests include Raman spectroscopy, holography, and simulated Raman spectroscopy.