NANO LETTERS
Optical Signal Comparison of Single Fluorescent Molecules and Raman Active Gold Nanostars
2009 Vol. 9, No. 11 3816-3819
Edward S. Allgeyer,† Adam Pongan,‡ Michael Browne,‡ and Michael D. Mason*,‡,¶ Department of Physics and Astronomy, Department of Chemical and Biological Engineering, UniVersity of Maine, Orono, Maine 04469, and Institute for Molecular Biophysics, Orono, Maine 04469 Received June 23, 2009; Revised Manuscript Received October 1, 2009
ABSTRACT The relevant photophysical properties of single fluorescent molecules and single SERS active surface-coated gold nanostars tagged with the Raman reporter molecule 4-mercaptopyridine are compared for imaging purposes. Mean count rate distributions are built from the single molecule/single probe level. The individually observed variance and count rates of both systems are compared as well as the behavior over multiple image acquisitions.
Currently, fluorescent molecules dominate as the accepted standard for labeling and contrast generation in a wide variety of sensing and microscopy techniques1 despite a number of well-known limitations.2 In recent years, nanoprobes (NPs) whose signal is based on surface enhanced Raman scattering (SERS) from metallic nanoparticles have been investigated as an alternative to organic fluorophores.2 SERS is now understood to result from both chemical and electromagnetic enhancement.3 The latter, electromagnetic enhancement, results from the interaction of incident radiation with the metal surface inducing conduction electron oscillations which subsequently produce a local electric field. These conduction electron oscillations are termed plasmons and have resonance modes that are sensitive to size, shape, and material.3 SERS NPs offer a number of advantages over fluorescent molecules for some imaging applications because they do not photobleach, blink, or saturate, and allow for definite spectroscopic identification with typical line widths (10-20 cm-1) many times narrower than fluorescence line widths (400-800 cm-1).4 Because of their distinct nonoverlapping “fingerprint” vibrational spectra it becomes possible to distinguish many probes simultaneously as has been demonstrated for combinatorial chemical libraries where as many as 24 Raman labeled beads have been shown to be uniquely distinguishable.5 Unlike fluorescence, the Raman process can be driven at any wavelength allowing for the use of near-infrared light * To whom correspondence should be addressed E-mail: michael.mason@ maine.edu. † Department of Physics and Astronomy, University of Maine. ‡ Department of Chemical and Biological Engineering, University of Maine. ¶ Institute for Molecular Biophysics. 10.1021/nl902008g CCC: $40.75 Published on Web 10/14/2009
2009 American Chemical Society
sources capable of improved optical penetration into biological samples while simultaneously reducing background autofluorescence (although SERS enhancement will be most effectively driven when the wavelength of incident radiation matches that of a plasmon resonance). Metal SERS NPs can be readily surface modified allowing for their use in a number of applications or chemical environments. For example, SERS-based Raman probes have been used for live cell probing and imaging,6 biomarker detection in live cells,7 tumor targeting in vivo,8 small animal whole-body imaging for living subjects,2 protein detection in tissues, multiplex DNA hybridization assays, and sandwich-binding assays.9 Furthermore, the metallic core of a SERS NP allows for identification by independent methods such as electron microscopy (TEM, SEM), which is generally not possible with organic fluorophores. Unfortunately, typical SERS probes are many times larger than standard organic fluorophores limiting their effectiveness in applications where the features of interests require a molecular size probe. While these NPs show promise, their widespread application in imaging has not yet been realized. This is, at least in part, due to the perception that Raman active NPs suffer from low signal levels that vary widely from particle-to-particle and fluctuate with their environment.10 Indeed, there are numerous studies of nanoparticle SERS systems where the data support these conclusions.11 As such, many research communities are left with the conclusion that fluorescence from established organic fluorophores are the only reliable contrast agent. Recently, a new class of star-shaped gold nanoparticles, referred to as nanostars, have been introduced that exhibit a
large degree of surface roughness12 and a high Raman enhancement factor.13 On the basis of this core geometry, we employ here a Raman active NP that demonstrates excellent signal strength, consistency, and environmental insensitivity. The extremely small radii of curvature present in these NPs from surface features results in strong electric field enhancement and subsequently large SERS enhancement factors per surface molecule.14 Additionally, nanostars exhibit surface plasmons resulting from the hybridization of individual tip and core plasmons giving rise to a further increase in the local electric field.3 The increased surface area, relative to a sphere of equivalent size, allows a larger number j ) of Raman active reporter molecules to be functionalized (N on their surface. In addition to the expected linear increase in the Raman signal (∝ N),8 the variations in the relative number j )/N j ) between of absorbed Raman active reporters ((Ni - N j particles are expected to be less significant as N is increased. Finally, by adding a protective surface coating to the nanostars (polyethylene glycol dithiol) the SERS signal becomes insensitive to its surroundings, even under harsh conditions, and virtually eliminates aggregation.1,8 Here we present a comparative study of relevant photophysical properties at the single molecule (SM)/single probe level. Given the implementation of an optimal NP structure, the observed signal from this new class of Raman probes can in fact exhibit characteristics superior to those seen in even the best organic fluorophores when imaged under identical conditions. This is especially true when both the peak and temporal characteristics are considered. The latter being particularly important for the biological sciences where signal degradation has long plagued quantitative imaging efforts.2 Fluorescent SM samples were prepared for imaging by spin-casting a dilute solution of Alexa Fluor 633 (Invitrogen) onto cover glass cleaned by overnight soaking in a liquid detergent solution (Decon Laboratories). Solutions of Alexa Fluor 633 were prepared daily from a stock solution of dye in UV photobleached 18MΩ DI water to minimize background fluorescence. Preparation of the SERS NPs began with gold nanostars synthesized using a modified recipe following from Kumar et al.,15 Khoury and Vo-Dinh,14 Graf et al.,16 and Silvert et al.17 Briefly, a solution of 2 mM gold nanostars were synthesized in 15 mL of dimethylformamide. From that solution, 9 mL was centrifuged for 20 min at 2000 g. After centrifugation the solution was removed, and the pellet was resuspended in 9 mL of ethanol. This process was repeated. The resulting nanostars were conjugated with a ∼1:3 molar ratio of the Raman reporter molecule 4-mercaptopyridine (4Mpy) and 5 kDa R-methoxy-omega-mercapto poly(ethylene glycol) (PEG). While being stirred vigorously, 750 mL of 22 mM 4Mpy was added dropwise and stirred for 15 min. Then 3.5 mL of 12 µM PEG was added dropwise while stirring vigorously for another 15 min. The resulting suspension was centrifuged and washed as above and finally resuspended in water. Average SERS nanoprobe core size was determined to be (93.5 ( 2) nm via TEM as seen in Figure 1a. Finally, SERS NP samples were prepared for Nano Lett., Vol. 9, No. 11, 2009
Figure 1. (a) TEM image of SERS probes nanostar core. (b) Schematic of the custom built confocal sample scanning Raman/ fluorescence microscope. (c,d) The 30 µm2 images of Alexa 633 and SERS probes, respectively.
imaging by drop casting a dilute solution on cleaned cover glass. As the PEG surface coating prevented aggregation spin coating was not necessary. Only rarely was an imaged area found to contain any SERS nanoprobe aggregates. All images and spectra were recorded on a custom built sample scanning confocal Raman/fluorescence microscope as show in Figure 1b. A 632.8 nm HeNe LASER was used for excitation of both fluorescent molecules and SERS NPs. Both fluorescence and Raman images were collected using a 647 nm long pass filter (Semrock), a slightly underfilled 1.2 N.A. UplanApo/IR water immersion objective (Olympus), and a fiber coupled, 50 µm fiber core, avalanche photodiode (Perkin-Elmer). A piezo stage (Physik Instrumente) was used for sample scanning and a monochromator coupled to a liquid nitrogen cooled CCD camera was used for spectral collection. A 600 µm fiber bundle to linear array was used to transmit light to the spectrograph and also acted as the slit for the system. All devices were controlled and synchronized using a custom built LabView program. Excitation power at the sample for Raman and fluorescence imaging ranged from 20-120 and 10-80 µW, respectively, while Raman spectra were collected with 400 µW at the sample and a 12 s integration time. Uniform regions of low density fluorescent SMs or NPs were imaged and analyzed to ensure the presence of only discrete individual molecules or nanoparticles. Example images can be see in Figure 1c,d. Fluorescence time traces were obtained and used to verify that features were indeed single molecules based on the observation of signal intermittency and single step photobleaching as seen in the insert to Figure 2a. Similar intermittency was not observed in the case of the SERS NPs. Raman spectra of individual SERS NPs were collected to verify that image features were a result 3817
Figure 2. Single molecule/single probe intensity distributions for fluorescence (a) and Raman (b) taken under identical experimental conditions. Inset for (a) shows example blinking and single step photo bleaching for imaged fluorescence features. Inset for (b) shows sample example Raman spectra for imaged SERS NP features.
Figure 3. (a) Sample AFM image of drop cast SERS nanoprobes showing single probe features. (b) Extinction spectrum of 4Mpy tagged PEG coated SERS nanoprobes.
of the expected 4Mpy Raman signal18 as seen in the insert to Figure 2b. Atomic force microscopy (Schaefer XE 100) was performed on SERS NP samples to verify that the optically imaged diffraction limited spots were in fact single probe features, as can be seen in Figure 3a. Although not used for data collection NP samples were prepared under identical conditions to those used for data collection and Raman imaged. After optically imaging the samples (data not shown) AFM was performed. AFM showed that the SERS NP samples are overwhelmingly composed of discrete single probes with only a small number of probes not well spatially separated. An extinction spectrum (Ocean Optics USB4000) of the SERS NPs is shown in Figure 3b for further characterization. Images were recorded for a series of excitation intensities and then statistically analyzed using a custom MatLab script. The average background count rate per pixel was determined for each excitation power and subtracted. SM/NP features were identified by examining those pixels with intensities at 3818
Figure 4. (a) Integrated count rate variance as a function of excitation power and (b) normalized integrated mean count rate as a function of excitation power. Error bars shown on (a) are the standard deviation for each population.
least six standard deviations above the average pixel value for each image. A fixed rectangular region-of-interest (ROI) was created around each bright pixel and a two-dimensional Gaussian fit was used to find feature centers. The count rate within each ROI was integrated. The integrated count rate as well as the fit parameters for each feature were recorded. Features with fitted 1/e2 radii greater than would be expected for this imaging system and features giving poor fit results, that is, aggregates and features not well spatially separated, were discarded. From this data, the mean integrated count rate for SM/NP features was computed for each power as well as the variance in the integrated count rate as a measure of the consistency between probes. Additionally, the mean total count rate was recorded for multiple scans of the same region at different locations and averaged assessing the question of long-term stability of each probe system. Examples of the experimentally obtained intensity distributions for each probe system are shown in Figure 2a,b. The variance and the normalized mean count rate for each distribution were calculated and are shown as a function of excitation power in Figure 4. From Figure 4a, we see that the variance in integrated count rate from the NPs increases with increasing excitation power as would be expected for a system limited only by photon counting statistics. In contrast, the variance of the SMs follows no predicable trend, likely due to the presence of fluorescence intermittency and bleaching. Furthermore, as the excitation power is increased we observe that the SMs reach saturation around 60 µW (21kW/cm2), as seen in Figure 4b, whereas the SERS NPs deliver continually increasing signal over a wider range of powers. Unlike fluorescent molecules that exhibit significant photobleaching, SERS NPs exhibit negligible loss in signal over time. Figure 5 shows the relative mean count rate for multiple Nano Lett., Vol. 9, No. 11, 2009
same saturation conditions observed for organic molecular fluorophores potentially allowing for greatly improved signal. We also find that SERS NPs exhibit superior temporal behavior free of intermittency and bleaching. In addition to the other potential advantages, this suggests that SERS NPs may be better suited for long-term or repeated imaging studies than their fluorescent counterparts.
Figure 5. Normalized integrated count rates as a function of scan number for both fluorescence and SERS NPs.
sample regions each imaged multiple times, using a modest 10 µW (3.5 kW/cm2) excitation power and a 1 msec dwell time per pixel. Each region was imaged four times and the integrated count rate for each image was computed and normalized with respect to that region’s initial integrated count rate. For each time point, the normalized count rates across multiple regions were averaged. Although very small, the error bars in Figure 5 present the propagation of the counting uncertainty through the normalization and averaging process. As can be seen in Figure 5, when imaged, the Alexa Fluor 633 dye’s integrated count rate decreases by slightly greater than half after only four image acquisitions while the NPs, imaged under identical conditions, exhibits no loss in signal strength. Interestingly, the SERS NP signal actually shows a slight increase in signal over the same time series. While the mechanism behind this increase is unclear, similar behavior has been observed for SERS NPs in different conditions.19 Although fluorescent dyes are well established for many imaging techniques, SERS nanoprobes can overcome some known limitations of fluorescent dyes. Despite previous challenges, the nanostar-based SERS NP presented here eliminates the traditional challenges of signal consistency and strength and demonstrates the viability of SERS probes for imaging applications. By comparing populations at the single probe level we find that the count rate variance from a SERS NP is at least as good as that observed for Alexa Fluor 633. Whereas at lower surface concentration only “hot particles” are observed, by increasing the number of Raman reporters on each particle and using a nanostar core we show that signal consistency is comparable to a highly regarded fluorescent dye. This NP system does not suffer from the
Nano Lett., Vol. 9, No. 11, 2009
Acknowledgment. This material is based upon work supported by the National Science Foundation under Grant 0722759. The authors thank Sarah M. Sterling for her helpful laboratory advise, Travis J. Gould for advise regarding data analysis, and the Institute for Molecular Biophysics. References (1) Mulvaney, S.; Musick, M.; Kearting, C.; Natan, M. Langmuir 2003, 19, 4784–4790. (2) Keren, S.; Zavaleta, C.; Cheng, Z.; de la Zerda, A.; Gheysens, O.; Gambhir, S. S. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 5844–5849. (3) Hao, F.; Nehl, C. L.; Hafner, H. J.; Nordlander, P. Nano Lett. 2007, 7, 729–732. (4) Kneipp, K.; Kneipp, H.; Bohr, H. G. Single-Molecule SERS Spectroscopy. In Surfaced-Snhanced Raman Scattering - Physics and Applications; Springer-Verlag: Berlin, 2006; Vol. 103. (5) Fenniri, H.; Ding, L.; ribbe, A.; Zyrianov, Y. J. Am. Chem. Soc. 2001, 123, 8151–8152. (6) Kneipp, J.; Kneipp, H.; Rajadurai, A.; Redmond, R. W.; Kneipp, K. J. Raman Spectrosc. 2009, 40, 1–5. (7) Lee, S.; Kim, S.; Jaebum, C.; Shin, S. Y.; Lee, Y. H.; Choi, H. Y.; Seunghan, H.; Kang, K.; Oh, C. H. Anal. Chem. 2007, 79, 916–922. (8) Qian, X.; Peng, X.-H.; Ansari, D. C.; Yin-Goen, Q.; Chen, G. Z.; Shin, D. M.; Yang, L.; Young, A. N.; Wang, M. D.; Nie, S. Nat. Biotechnol. 2008, 26, 83. (9) Lutz, B.; Dentinger, C.; Sun, L.; Nguyen, L.; Zhang, J.; Chmura, A. J.; Allen, A.; Chan, S.; Knudsen, B. J. Histochem. Cytochem. 2008, 56, 371–379. (10) Nie, S.; Emory, S. Science 1997, 275, 1102–1106. (11) Pieczonka, N. P. W.; Aroca, R. F. ChemPhysChem 2005, 6, 2473– 2484. (12) Nehl, C. L.; Liao, H.; Hafner, J. H. Nano Lett. 2006, 6, 683–688. (13) Hrelescu, C.; Sau, T. K.; Rogach, A. L.; Jackel, F.; Feldmann, J. Appl. Phys. Lett. 2009, 94, 153113. (14) Khoury, C. G.; Vo-Dinh, T. J. Phys. Chem. C 2008, 112, 18849– 18859. (15) Kumar, P. S.; Pastoriza-Santos, I.; Rodrı´guez-Gonza´lez, B.; Garcı´a de Abajo, F. J.; Liz-Marza´n, L. M. Nanotechnology 2008, 19, 015606. (16) Graf, C.; Vossen, D. L. J.; Imhof, A.; van Blaaderen, A. Langmuir 2003, 19, 6693–6700. (17) Silvert, P.-Y.; Herrera-Urbina, R.; Tekaia-Elhsissen, K. J. Mater. Chem. 1997, 7, 293–299. (18) Zhang, L.; Bai, Y.; Shang, Z.; Zhang, Y.; Mo, Y. J. Raman Spectrosc. 2007, 38, 1106–1111. (19) King, M. D.; Khadka, S. K.; Craig, G. A.; Mason, M. D. J. Phys. Chem C 2008, 112, 11751–11757.
NL902008G
3819