We present a study on the surface-enhanced Raman scattering (SERS) properties of Ag nanoparticle island substrates (NIS) and their applications for target oligonucleotide (OND) detection. It has been found that the surface nanostructure of NIS samples can be controlled with a good degree of reproducibility, and a high SERS enhancement can be achieved when the peak extinction wavelength of NIS is tuned to a spectral window () between the excitation wavelength and the scattered Raman wavelength. The highest SERS enhancement was obtained from the NIS substrates with a nominal thickness of . Detection of target OND was performed with a sandwich format in which the target OND was hybridized both to a capture OND immobilized on the NIS substrate, and a detection OND conjugated with a Raman-active dye for SERS signal generation. We compare the detection performance of two strategies based on the use of the detection OND with or without the gold nanoparticle (Au-NP). Our results confirm that, when the detection OND is coupled to the Au-NP, a better sensitivity for the target OND detection, in terms of a wider dynamic range and a lower detection limit ( versus without Au-NP), would be achieved.
© 2009 Optical Society of America
Over the years, different assay readout strategies have been developed to achieve high sensitivity for gene detection and identification . Techniques based on surface plasmon resonance (SPR) , surface-enhanced Raman scattering (SERS) [3, 4, 5, 6, 7, 8, 9, 10], quantum dots , microcantilevers , and atomic force microscopy  have been devised for biosensing. Among them, SERS, using both the intrinsic and the extrinsic Raman signals of certain labels for sensing biomolecules, has attracted most intense research interests. Such an expansion is due to several intrinsic merits that SERS offers. For example, SERS provides a high sensitivity and has potential to be used as a multiplexed readout technique [14, 15]. These merits can be attributed to a number of features, including the SERS enhancement factor (EF), which can be optimized to a level as high as for single molecule identification . Also, when multiple labels are used for detection, the readout spectrum will be less overlapped, thus reducing the cross talk in multiplexed assays, since Raman bands are in general much narrower than those of fluorescence peaks. Moreover, in the SERS assays, it is possible to use one excitation wavelength to read multiple labels. Also, Raman signals are less susceptible to photobleaching than fluorescence [14, 15].
However, the acceptance of SERS as a general analytical tool has been hindered by the lack of well- engineered SERS substrate and systematic development of substrate technology [5, 6, 7]. Generally, SERS activities come from randomly distributed hot spots, which are primarily associated with nanosized col loidal noble metal aggregates deposited on the substrate. Nonetheless, not much information on the conditions to produce efficient SERS has been reported in the literature before.
In this work, we have optimized a strategy to produce an effective SERS platform for biosensing, which can be used for fast and sensitive detection of target DNA. Since SERS signals are essentially derived from the enhanced localized electric field associated with nanosized metal structures, it is most important to increase the density of the nanosized features and to maintain a high degree of reproduc ibility. For these reasons, we adopted a physical vapor deposition (PVD) route to prepare the metal layer because of its simplicity and the possibility of achieving high density of nanosized surface features. An investigation of the SERS properties of NIS prepared under various layer growth rates was performed systematically by studying the correlation between surface nanostructures of the NIS samples, their SERS spectra, and the EF [16, 17, 18]. Our work has revealed that the extinction spectrum or extinction maximum () of the NIS substrates is very sensitive to the surface nanostructures, which can be tuned by changing the thickness of the deposited metal, with a constant coating rate and vacuum condition of the PVD system.
For the sandwich DNA detection assay, a layer of single stranded capture OND with a thiol group at the end was first immobilized on a NIS substrate so that, upon hybridization with the target OND, the target OND would in turn associate with the detection OND with FAM at the end, through the Watson and Crick base pairing (Scheme 1) [19, 20]. Through this design, a particular Raman spectroscopic fingerprint can be identified after hybrid ization of the detection OND to the target OND according to the specific Raman-active dye [3, 5, 10, 14]. Yet, the common SERS schemes do not achieve the possible maximum EF because the detection OND are not sufficiently close to the SERS surface after hybridization with the target OND. In the present study, we adopted a new strategy that employed a detection OND capped with a gold nanoparticle (Au-NP, ) at the end. The Au-NP acted as a SERS promoter for the dye molecule when the detection OND was captured by the target and capture OND on substrate. We also compared the detection performance of the detection OND with or without Au-NP in terms of dynamic range and detection limit. Our results show that the target OND detection limit on the NIS substrate in the cases with or without Au-NP was found to be and , respectively.
2. Experimental Section
Ag (99.99%) was purchased from D. F. Goldsmith (Evanston, Illinois). Glass substrates were obtained from Fisher Scientific (Pittsburgh, Pennsylvania). Tungsten vapor deposition boats were purchased from R. D. Mathis (Long Beach, California). Water () was obtained from an ultrafilter system (Milli-Q, Millipore, Marlborough, Massachusetts). All chemicals and solvents were purchased at the highest purity grade and used without further purification.
6-mercapto-1-hexanol (MCH, ), NaCl (AR grade, ), phosphate buffered solution (PBS, phosphate), ammonium hydroxide solution (30% basis), and Rhodamine 6G (Rh6G) were purchased from Sigma Aldrich (USA). Hydrogen peroxide and 96% sulfuric acid were purchased from Pierce (USA). Phosphine moiety [Bis(p-sulfonatophenyl)phenylphosphine] was obtained from STREM Chemicals (USA) and Au-NP solution was bought from Ted Pella Inc. (, USA). All OND (, PAGE purification) were obtained from Tech Dragon (Hong Kong) and diluted to a concentration of for stock (in a NaCl, pH 6.5 PBS). Capture OND p1 and detection OND p4 were modified with a thiol group (SH); detection OND p3 and p4 were labeled with Raman tag FAM. All OND were kept at and freshly prepared before use. The detailed sequence designs of OND (p1, p2, p3, and p4) are shown in Chart 1.
Chart 1. Sequences of OND used:
- p1 (Capture OND): (HS)--(C6H12)-CGC ATT CAG GAT-
- p2 (Target OND): -TAC GAG TTG AGA ATC CTG AAT GCG-
- p3 (Detection OND): FAM--TCT CAA CTC GTA-(C3H6)-
- p4 (Detection OND): FAM--TCT CAA CTC GTA-(C3H6)-(SH)-
2B. NIS Substrate Fabrication
Glass substrates were cleaned with piranha solution at for first, and then a base treatment method with () was used to render the surface hydrophilic. The metal films of Ag were deposited in a deposition system (Edwards, UK) with a base pressure of . Different thicknesses of Ag, ranging from 10 to , were coated on the substrates for further study. The deposition rate for each sample was kept at with the aid of a quartz crystal microbalance (Maxtek Inc.). Chromium ( thick) was deposited on the glass substrates prior to the deposition of Ag in order to improve adhesion. The NIS substrates were stored in a dry, high-vacuum and dark sample chamber at room temperature prior to use within 2 days.
2C. Preparation of Receptor Oligonucleotide Substrates and Hybridization
Scheme 1 illustrates our immobilization strategy via HS–Ag binding and hybridization procedures for a typical sandwich DNA detection. The Ag-coated NIS substrate was immersed in capture OND p1 solution (, ) for approximately , followed by a exposure to MCH solution (, ). The self- assembled monolayers of MCH are between the capture of ONDs and used to block direct exposure of those gaps or unoccupied substrate to target OND. [5, 21, 22]. The substrate was then washed with water several times and dried in the air. The well-prepared capture OND substrate could selectively detect target OND p2 by complementary hybridization. Subsequently, the substrate was covered with target OND p2 solution (concentration , ) and placed in a hybridization oven for with an saturated airtight container to prevent evaporation of the OND solution. Then a NaCl PBS buffer solution was used repeatedly to flush away the excess target OND p2. The substrate was then air dried and ready for complementary hybridization with the detection OND to verify the presence of target OND p2 on the substrate. For this, detection OND p3 or p4 (, ) was added, and the substrate was placed in a humidity container at for for hybridization. After hybridization, the substrate was washed copiously with water and dried by a centrifuge ( for ).
2D. Modification of Au-NPs with Probe Oligonucleotide
As shown in our sandwich assay for DNA detec tion (Scheme 1), Au-NPs modified, FAM-labeled, alkylthiol-capped OND p4 strands were used as probes to monitor the presence of specific target OND p2.
Phosphine was used to increase the stability of Au-NPs before their conjugation with detection OND p4. of NPs solution () was concentrated by a factor of 200, by precipitating the NPs using an ultracentrifuge ( for ). The pellet was resuspended in of phosphine buffer ( phosphine in distilled water) [[8, 9, 14, 23, 24, 25]. The mixture was left overnight on a rocking platform at room temperature to allow sufficient time for surfactant exchange. The final concentration of the phosphine-coated Au particles was about .
On average, there are about 200 OND strands attached on each Au-NP [14, 23]. Then, OND p4 () was mixed with the concentrated phosphine-coated Au-NPs () in a molar ratio of at room temperature for about . OND p4 was puri fied by centrifugation at for to remove excess reagent; the supernatant was discarded and the red oily precipitate was diluted with of NaCl pH 6.5 PBS. The prepared conjugation solution of Au-NPs with detection OND p4 should be used freshly or stored at .
2E. SERS Apparatus
We used a Renishaw MicroRaman system (, ) equipped with a objective ( 0.75 NA) and a pinhole to detect the SERS signal. The accumulation time for the Raman spectrum was typically . Five spots at the same substrate were measured for each sample to minimize measurement errors. For experiments using Rh6G, the spectrum acquired from the blank substrate was used as a reference, and the background spectrum of the substrate coated with captured OND was used in the DNA experiments.
2F. Atomic Force Microscope and UV–Visible Spectrum Measurements
An atomic force microscope (AFM, Digital Instrument, Dimension 3100) and UV–visible–NIR spectrophotometer (Hitachi U-3501) were used for surface characterization. An AFM working in tapping mode under ambient conditions was used to study the surface morphology and structural variation of NIS substrates at different silver thicknesses. The extinction spectra were acquired using a UV–visible/NIR spectrophotometer to study the localized surface plasmon resonance (LSPR) properties of the NIS substrates.
3. Results and Discussion
In this study, we tried to study (1) the physical characteristics of the NIS based on the AFM, extinction spectra, and SERS results, and (2) the applicability and efficiency of these substrates as a DNA detection platform.
3A. Morphology Study and Statistical Analysis of Distribution
An AFM study and following roughness analysis were performed to better understand the morpholog ical variation of the NIS substrates by changing the Ag thickness; the results are shown in Fig. 2. In this study, the PVD deposition conditions, such as deposition rate and vacuum were kept constant for all depositions. From Fig. 2, it can be seen that a continuous and uniform Ag nanostructured surface began to develop at a thickness of . As expected, the fine nanostructures started to disappear when the thickness went beyond , as the size of the NPs increased with the thickness of the silver layer. This can also be revealed from the roughness analysis of the AFM results, as demonstrated in Fig. 2; the roughness, i.e., the nominal size of the NPs, actually increases with the thickness from 2.625 to , while the relative size variation decreases from 78.32% to 12.64% at the same time.
To understand the effect of the thickness on the morphological variation of nanostructures, UV– visible absorption spectra of several thicknesses, i.e., 10, 30, 50, 70, 90, and , were measured, as shown in Fig. 3. As can be seen, there is a notable trend of redshift in the extinction maximum from 517.5 to when the thickness of the silver layer increased from 10 to . For the case of , the approached the upper limit of the wavelength of our UV–visible spectrophotometer. The FWHMs of the corresponding spectra of different thicknesses vary from 100 to .
A more systematic statistical study of the extinction spectra and extinction maximum of NIS substrates was conducted to explore the distribution of caused by chip-to-chip variation. In this study, the extinction spectra of 20 NIS substrates of each thickness, i.e., 10, 30, 50, 70, and , were measured and analyzed with Gaussian fitting. All parameters derived from the Gaussian fitting were reported as the average value , where σ is the half-width at half-maximum. As shown in Fig. 4, the histograms in Figs. 4A, 4B, 4C, 4D, and 4E show that the average of of thicknesses 10, 30, 50, 70, and are , , , , and , respectively. Figure 4F depicts the redshifting of the average (with corresponding values of σ shown as error bars) for different thicknesses of the silver layer. There was also a notable reduction of in the value of σ as we increased the thickness of the silver layer from 10 to . This result is consistent with the observation seen from the AFM images and roughness analysis shown in Fig. 2. For very thin layers below , the nanoislands exhibited quite observable variation in their size, whereas in the samples with larger thicknesses, the islands were more homogeneous, and the surface became more continuous.
3B. SERS Enhancement Factor Calculation
Typically, the enhancement performance can be quantitatively described by an EF of a particular adsorbate vibrational mode. The EF can be obtained by comparing the emission efficiency or signal strength per particular adsorbate molecule of the SERS sample and the normal Raman standard in terms of one vibrational mode. EF can be calculated using the following equation [15, 16, 19, 26]:27, 28, 29], using a Rh6G solution concentration of . The number of molecules contributing to the normal Raman signal measured from the standard was . The known spot size of the microscope objective was in diameter. A Rh6G (variant concentration) solution droplet () spread into a circle of about in diameter. Assuming that the solution evenly covered the NIS substrate surface after air drying and centrifugation, the number of molecules involved could be obtained by multiplying the exposed surface area by the surface density of the dried adsorbate Rh6G. In the experiment, multipoint measurements and averaging were used in order to reduce the measurement error induced by the “coffee ring” effect, which is a typical consequence of the air-dry method. We must point out that the reported collective enhancement here combines the contributions of resonance (chemical) effect and surface enhancement (electromagnetic) effect. Further experiments are needed in order to specifically isolate these two effects. In terms of the surface enhancement effect, it should be mentioned that only a portion of the molecules under excitation contribute to the measured signal from the NIS substrate because of the uneven distribution of hot spots. This suggests the actual EF for an individual nanoisland can be much higher than the observed value, which represents the average EF of a large collection of nanoislands.
3C. Correlation Among , , and
Figure 5 shows the correlation of UV–visible extinction spectra of our NIS samples with different Ag layer thicknesses and their corresponding SERS spectra from the Rh6G. The locations of for each thickness are also highlighted in the table to reveal their relative positions to and . The Raman spectra were acquired from dried Rh6G solution of concentration , using an integration time of . was the Raman-Stokes-shifted wavelength. A strong vibrational mode at , which was associated with the in-plate carbon–carbon (c–c) stretching vibration mode, was chosen for monitoring the EF in our experiments. As shown in Table 1, our results reveal that the strongest SERS intensity occurred when the extinction maximum was located between () and (), especially when the thickness of the Ag film was around . Apart from this special region, the EF did not go beyond 10 when the SERS signal was measured against the background. Out of the 20 substrates at each thickness that had been studied, the EF was typically when was between the excitation wavelength () and the Stoke-shifted Raman wavelength . The highest measured EF was about , at an Ag thickness of . The small variation of EF also suggests that the fabrication of NIS substrates using a PVD method would not introduce a large chip-to-chip or site-to-site variation in substrate surface morphology.
3D. DNA Detection using NIS
Using the optimized NIS (i.e., with an Ag layer thickness of , fabricated with a base pressure of and a coating rate of ), we performed a series of OND-detection experiments based on the sandwich assay shown in Scheme 1. The capture OND was modified to contain a thiol group (SH) to be embedded onto the Ag surface for hybridization experiments. Compared to other schemes for protein detection using avidin–biotin binding [30, 31], our approach of SH–Ag binding took full advantage of the surface ambient electromagnetic field by drastically reducing the distance between the Raman emitter and the substrate surface (results not shown). Instead of the one from the DNA itself, the Raman spectrum fingerprint of FAM labeled at the end of the detection OND was acquired. This approach offered many advantages, for example, a large pool of dyes for multiplexed detection, reduction in spectra overlapping caused by similar DNA structures, much bigger cross sections and no photobleaching. Also, labeling the end instead of the end of the OND, which is different from the configuration commonly employed by other works [8, 9, 14, 23, 24, 25], would fur ther reduce the distance effect that does not favor SERS. The use of MCH as a blocker in our scheme was also essential to prevent nonspecific binding caused by direct exposure of the silver surface to the target and detection OND.
To find the detection limit of NIS substrates for the identification of target OND, various concentrations of target OND, ranging from micromolar to nanomolar, were employed. Meanwhile, the concentrations of capture and detection OND were kept constant, i.e., , in all our experiments. At this concentration, the surface binding area would be fully occupied by the capture OND. Our results indicate that the SERS signals from the detection OND p3 did not increase further when the concentration of the target OND was greater than (data not shown), suggesting that no further binding sites were left when the concentration was approaching . Result in Fig. 6A demonstrates a continuous but nonlinear decrease of signal when the concentration of the target decreased from to . Data at were extracted from Fig. 6A and plotted in Fig. 6B. From the sigmoidal response curve depicted in Fig. 6B, the detection limit of the target OND was found to be to and the dynamic range was found in the range from to . Further increase in the concentration of the target OND beyond could not lead to a significant increase in SERS signal.
For the control experiment without the target OND, we used probe p3, which did not complementarily bind to capture OND p1. In such a condition, no SERS signal from FAM was detected (data not shown).
3E. Effect of Conjugation of Au-NPs and Oligonucleotide on DNA Detection
As the NIS substrate could not detect the target OND at a concentration smaller than , we tried another method to increase the detection limit further for practical application purpose. Works from other groups indicate that formation of a capture-OND–target-OND–detection-OND triplex is essential for single molecule detection [15, 19, 26, 27]. Therefore, in our triplex system, Au-NPs were conjugated to the detection OND. It is noteworthy that each Au-NP, having a diameter of , is capable of carrying detection OND [14, 23]. Under this situation, both the number of Raman emitters and the possibility of the dye being present in the vicinity of a hot spot increased. These would enhance the SERS signal and the detection capability of the system.
Our experimental results indeed confirmed this enhancement in terms of the SERS intensity. As shown in Figs. 7A and 7B using of capture and detection OND again, a smaller target OND concentration at was detected. The sigmoidal curve in Fig. 7B also revealed that both the dynamic range and the detection limit were amplified when Au-NPs were attached to the end of the detection OND. Compared to the results in Fig. 6, where no Au-NPs was employed, this scheme offered a wider dynamic range from to , with a detection limit extended to subfemtomolar concentration. Now, it becomes quite clear that the combined approach of using Au-NPs and NIS substrates is very useful for DNA biosensing in light of their high EF, simplicity, and consistency.
In this investigation, we have accomplished two goals. First, through a systematic study of deposition conditions and spectral characteristics of NIS substrates, we demonstrated that it is necessary to correlate the excitation wavelength, extinction maximum, and Raman-scattered wavelength in order to optimize the SERS performance of the NIS substrate for the detection of target OND [16, 17, 18]. It has become clear that NIS substrates offer practical and consistent SERS surfaces for a wide range of detections. Our experimental results using adsorbate Rh6G reveal that high signal-to-noise SERS spectra can be obtained when the extinction maximum wavelength of the silver-coated substrate is properly tuned to an spectral window between the excitation wavelength and the Stokes-shifted wavelength through varying the silver film thickness. The experimental SERS EF was typically from to , with the highest level being , which occurs at a silver thickness of . Second, we have shown that a combined approach based on the use of NIS substrates and Au-NPs, which results in the formation of duplex and triplex configurations among the captured OND and target OND, can provide stronger SERS signals. Our experiments also demonstrate that both the dynamic range and the detection limit can be optimized by using Au-NP for the detecting OND. The dynamic range achieved in the present case is from to , while the detection limit is . The approach we report here renders that SERS is useful for cases where the target DNA concentration is extremely low. We also anticipate that our approach can be extended to a multiplexed platform for detecting disease-related genes by using different Raman-active dyes.
We acknowledge J. Li and Daniel H. C. Ong for helpful discussions. This work has been supported by a Research Grants Council (RGC) grant under Competitive Earmarked Research Grant (CERG) project 411907. Partial funding support from the Shun Hing Institute for Advanced Engineering, The Chinese University of Hong Kong (CUHK), and a research studentship from CUHK for W. Yuan are also gratefully acknowledged.
1. E. P. Diamandis and T. K. Christopoulous, eds., Immunoassay (Academic, 1996).
2. G. J. Wegner, H. J. Lee, and R. M. Corn, “Characterization and optimization of peptide arrays for the study of epitope- antibody interactions using surface plasmon resonance imaging,” Anal. Chem. 74, 5161–5168 (2002). [CrossRef] [PubMed]
5. M. Culha, D. Stokes, L. R. Allain, and T. Vo-Dinh, “Surface- enhanced Raman scattering substrate based on a self- assembled monolayer for use in gene diagnostics,” Anal. Chem. 75, 6196–6201 (2003). [CrossRef] [PubMed]
6. X. Zhang, J. Zhao, A. V. Whitney, J. W. Elam, and R. P. Van Duyne, “Ultrastable substrates for surface-enhanced Raman spectroscopy: overlayers fabricated by atomic layer deposition yield improved anthrax biomarker detection,” J. Am. Chem. Soc. 128, 10304–10309 (2006). [CrossRef] [PubMed]
7. X. Zhang, M. A. Young, O. Lyandres, and R. P. Van Duyne, “Rapid detection of an anthrax biomarker by surface- enhanced Raman spectroscopy,” J. Am. Chem. Soc. 127, 4484–4489 (2005). [CrossRef] [PubMed]
8. J. Driskell, K. M. Kwarta, R. J. Lipert, and M. D. Porter, “Low-level detection of viral pathogens by a surface-enhanced Raman scattering based immunoassay,” Anal. Chem. 77, 6147–6154 (2005). [CrossRef] [PubMed]
9. D. S. Grubisha, R. J. Lipert, H. Park, J. Driskell, and M. D. Porter, “Femtomolar detection of prostate-specific antigen: an immunoassay based on surface-enhanced Raman scattering and immunogold labels,” Anal. Chem. 75, 5936–5943 (2003). [CrossRef] [PubMed]
10. S. E. J. Bell and N. M. S. Sirimuthu, “Surface-enhanced Raman spectroscopy (SERS) for sub-micromolar detection of DNA/RNA mononucleotides,” J. Am. Chem. Soc. 128, 15580–15581 (2006). [CrossRef] [PubMed]
12. G. Wu, R. H. Datar, K. M. Hansen, T. Thundat, R. J. Cote, and A. Majumdar, “Bioassay of prostate-specific antigen (PSA) using microcantilevers,” Nat. Biotechnol. 19, 856–860 (2001). [CrossRef] [PubMed]
13. V. W. Jones, J. R. Kenseth, M. D. Porter, C. L. Mosher, and E. Henderson, “Microminiaturized immunoassays using atomic force microscopy and compositionally patterned antigen arrays,” Anal. Chem. 70, 1233–1241 (1998). [CrossRef] [PubMed]
16. C. L. Haynes and R. P. Van Duyne, “Plasmon-sampled surface-enhanced Raman excitation spectroscopy,” J. Phys. Chem. B 107, 7426–7433 (2003). [CrossRef]
17. A. D. McFarland, M. A. Young, J. A. Dieringer, and R. P. Van Duyne, “Wavelength-scanned surface-enhanced Raman excitation spectroscopy,” J. Phys. Chem. B 109, 11279–11285 (2005). [CrossRef]
18. G. C. Schatz and R. P. Van Duyne, Handbookof Vibrational Spectroscopy, J.P. M.R. ChalmersGriffiths, eds. (Wiley, 2002), Vol. 1, pp. 759–774.
19. K. Kneipp, H. Kneipp, I. Itzkan, R. R. Dasari, and M. S. Feld, “Ultrasensitive chemical analysis by Raman spectroscopy,” Chem. Rev. 99, 2957–2976 (1999). [CrossRef]
20. L. A. Gearheart, H. J. Ploehn, and C. J. Murphy, “Oligonucleotide adsorption to gold nanoparticles: a surface-enhanced Raman spectroscopy study of intrinsically bent DNA,” J. Phys. Chem. B 105, 12609–12615 (2001). [CrossRef]
21. R. Levicky, T. M. Herne, M. J. Tarlov, and S. K. Satija, “Using self-assembly to control the structure of DNA monolayers on gold: a neutron reflectivity study,” J. Am. Chem. Soc. 120, 9787–9792 (1998). [CrossRef]
23. G. L. Liu, Y. Yin, S. Kunchakarra, B. Mukherjee, D. Gerion, S. D. Jett, D. G. Bear, J. W. Gray, A. P. Alivisatos, L. P. Lee, and F. F. Chen, “A nanoplasmonic molecular ruler for measuring nuclease activity and DNA footprinting,” Nat. Nanotechnol. 1, 47–52 (2006). [CrossRef]
24. C-C. You, A. Chompoosor, and V. M. Rotello, “The biomacromolecule-nanoparticle interface,” Nanotoday 2, 34–43 (2007). [CrossRef]
26. K. Kneipp, Y. Wang, H. Kneipp, L. T. Perelman, I. Itzkan, R. R. Dasari, and M. S. Feld, “Single molecule detection using surface-enhanced Raman scattering,” Phys. Rev. Lett. 78, 1667–1670 (1997). [CrossRef]
30. M. Howarth, D. J-F. Chinnapen, K. Gerrow, P. C. Dorrestein, M. R. Grandy, N. L. Kelleher, A. El-Husseini, and A. Y. Ting, “A monovalent streptavidin with a single femtomolar biotin binding site,” Nat. Methods 3, 267–273 (2006). [CrossRef] [PubMed]
31. P. R. Langer, A. A. Waldrop, and D. C. Ward, “Enzymatic synthesisof biotin-labeled polynucleotides: novel nucleic acid affinity probes,” Proc. Natl. Acad. Sci. USA 78, 6633–6637 (1981). [CrossRef]