Nanoparticles with Raman Spectroscopic Fingerprints for DNA and RNA Detection

See allHide authors and affiliations

Science  30 Aug 2002:
Vol. 297, Issue 5586, pp. 1536-1540
DOI: 10.1126/science.297.5586.1536


Multiplexed detection of oligonucleotide targets has been performed with gold nanoparticle probes labeled with oligonucleotides and Raman-active dyes. The gold nanoparticles facilitate the formation of a silver coating that acts as a surface-enhanced Raman scattering promoter for the dye-labeled particles that have been captured by target molecules and an underlying chip in microarray format. The strategy provides the high-sensitivity and high-selectivity attributes of gray-scale scanometric detection but adds multiplexing and ratioing capabilities because a very large number of probes can be designed based on the concept of using a Raman tag as a narrow-band spectroscopic fingerprint. Six dissimilar DNA targets with six Raman-labeled nanoparticle probes were distinguished, as well as two RNA targets with single nucleotide polymorphisms. The current unoptimized detection limit of this method is 20 femtomolar.

A highly sensitive and selective detection format for DNA relies on oligonucleotide- functionalized nanoparticles as probes, a particle-initiated Ag developing technique for signal enhancement, and a flatbed scanner for optical readout (1). The documented detection limit for this “scanometric DNA detection” format is 50 fM, (1) and the utility of the system has been demonstrated with short synthetic strands, polymerase chain reaction products, and genomic DNA targets (2,3). However, a limitation of this approach is that it is inherently a one-color system based on gray scale. The flexibility and applicability of all DNA-detection systems benefit from access to multiple types of labels with addressable and individually discernable labeling information. In the case of fluorescence, multiple fluorophores, including quantum dots, can be used to prepare encoded structures with optical signatures that depend on the types of fluorophores used and their signal ratio within the probes (4, 5). These approaches typically use micrometer-sized probes to obtain encoded structures with the appropriate signal intensities and uniformities. In the case of molecular fluorophores, overlapping spectral features and nonuniform fluorophore photobleaching rates lead to several potential complications (4, 6, 7).

Here, we show that nanoparticles functionalized with oligonucleotides and Raman labels, coupled with surface-enhanced Raman scattering (SERS) spectroscopy, can be used to perform multiplexed detection of oligonucleotide targets (Scheme 1). Although oligonucleotides can be directly detected by SERS on aggregated particles (7, 8), the structural similarities of oligonucleotides with different sequences result in spectra that are difficult to distinguish. Therefore, researchers often use different Raman dyes to label different oligonucleotides to distinguish oligonucleotide sequences (9, 10). To realize the benefits of high-sensitivity and high-selectivity detection coupled with multiple labeling capabilities, we designed nanoparticle probes that can be used for DNA (or RNA) detection (Scheme 1). These probes consist of 13-nm-diameter Au particles functionalized with Raman dye-labeled oligonucleotides. The Raman spectroscopic fingerprint, which can be designed through choice of Raman label, can be identified after Ag enhancing by scanning Raman spectroscopy (Scheme 1). Because the SERS-active substrate in this strategy is generated immediately before the detection event, a large and reproducible Raman scattering response can be obtained.

In a typical experiment for DNA detection, a three-component sandwich assay is used in microarray format (Scheme 1). Au nanoparticles (13 ± 2 nm in diameter) modified with Cy3- labeled, alkylthiol-capped oligonucleotide strands were used as probes to monitor the presence of specific target DNA strands (11). On average, there are about 110 oligonucleotide strands on each 13-nm Au nanoparticle (12). The Cy3 group was chosen as a Raman label because of its large Raman cross section (13). A chip spotted with the appropriate 15-nucleotide capture strands was coated with a 0.6 M NaCl phosphate-buffered saline (PBS) buffer solution (10 mM of phosphate, pH 7) containing a 30-nucleotide target sequence (100 pM) in a humidity chamber at room temperature. After 4 hours, the chip was washed four times with 0.6 M NaCl PBS buffer solution to remove nonspecifically bound target. Then, the chip was treated with a 0.6 M NaCl PBS solution of nanoparticle probes (2 nM) for 1.5 hours to effect hybridization with the overhanging region of the target sequence (Scheme 1). The chip was then washed with 0.6 M NaNO3 PBS buffer solution to remove chloride ions and nonspecifically bound nanoparticle probes. The chip was immediately treated with a Ag enhancement solution (Ted Pella, Inc., Redding, California) for 8 min, rinsed with Nanopure water, and dried with a microarray centrifuge (2000g). The chip, which exhibits gray spots visible to the naked eye, could be imaged with a flatbed scanner (Expression 1600, Epson America, Torrance, California) (Fig. 1B) (1–3). The spots also were studied by a Raman spectrometer coupled with a fiber-optic probe with a 0.65-numerical aperture microscope objective (25-μm laser beam) in a 0.3 M NaCl PBS buffer solution (Fig. 1C), and each of them shows a consistent and strong Raman response at 1192 cm−1 (Fig. 1D) (Solution Raman 633 spectrometer from Concurrent Analytical, Inc., Waimanalo, Hawaii; 30 mW HeNe laser).

Figure 1

Flatbed scanner images of microarrays hybridized with nanoparticles (A) before and (B) after Ag enhancing. (C) A typical Raman spectrum acquired from one of the Ag spots. (D) A profile of Raman intensity at 1192 cm−1 as a function of position on the chip; the laser beam from the Raman instrument is moved over the chip from left to right as defined by the line in (B).

Before Ag enhancing, the nanoparticle probes were invisible to the naked eye, and no Raman scattering signal was detectable (Fig. 1A). This is due to a lack of electromagnetic-field enhancement for the undeveloped nanoparticles (13 nm in diameter) in this state (7, 14, 15). Others have shown that closely spaced Au nanoparticles can give SERS enhancement (16, 17), but for DNA detection at technologically relevant target concentrations (<1 nM), nanoparticle spacings are too large to yield such effects. After Ag enhancing, the Ag particles can grow around the Cy3-labeled nanoparticle probes, leading to large Raman scattering enhancements. Typically, the spectra include both sharp Raman lines (∼15 to 30 cm−1 in spectral width) and a concomitant broad underlying continuum, as noted by Brus et al. in their studies of rhodamine 6G molecules on Ag particles (18). The Raman scattering signals arise almost exclusively from the Cy3 dye molecules immobilized on the particles; we do not observe signals from other species such as the oligonucleotides, solvent molecules, and the succinimidyl 4-(p-maleimidophenyl)butyrate on the glass surface. The Raman scattering frequency for each Raman line remains constant from experiment to experiment, deviating by less than 2 cm−1. These consistent SERS signals from the Cy3-labeled nanoparticle probes allow us to use the Raman spectrum of Cy3 as a spectroscopic fingerprint to monitor the presence of a specific target oligonucleotide strand.

Dyes other than Cy3 can be used to create a large number of probes with distinct and measurable SERS signals for multiplexed detection. To demonstrate this point, we selected six commercially available dyes with distinct Raman spectra that can be incorporated into oligonucleotides through standard automated DNA syntheses. Six types of Raman-labeled and oligonucleotide-modified Au nanoparticle probes were prepared with sequences that were respectively complementary to statistically unique 30- to 36-nucleotide sequences for (A) hepatitis A virus Vall7 polyprotein gene (HVA), (B) hepatitis B virus surface antigen gene (HVB), (C) human immunodeficiency virus (HIV), (D) Ebola virus (EV), (E) variola virus (smallpox, VV), and (F) Bacillus anthracis (BA) protective antigen gene (Fig. 2) (11).

Figure 2

(A) The Raman spectra of six dye-labeled nanoparticle probes after Ag enhancing on a chip (after background subtraction). Each dye correlates with a different color in our labeling scheme (see rectangular boxes). TAMRA, tetramethyl rhodamine. (B) Six DNA sandwich assays with corresponding target analysis systems. A10 is an oligonucleotide tether with 10 adenosine units.

Eight separate tests were carried out to evaluate the selectivity of the system and our ability to determine the number and types of strands in solutions containing mixtures of the different targets (Figs. 2 and 3). The concentrations of the target strands were kept constant for all of these experiments (100 pM each), and the hybridization conditions were as described above. In the first test (Fig. 3, row 1), all spots show the same intense gray color associated with Ag deposition. However, they can be differentiated simply by using the Raman scanning method. Once the spectroscopic fingerprint of the Ag-containing spot has been determined, the correct Raman label and, therefore, target sequence can be identified. To simplify the analysis, we assigned a color (rectangular box) to each Raman-labeled probe (Fig. 2A and Fig. 3B). In the first test (Fig. 3A), all six targets were present, and all showed strong gray-scale values when measured by means of the flatbed scanner as well as the expected Raman fingerprints. In the next seven tests, we systematically removed one or more of the targets to evaluate the suitability of this method for multiplexing. With the single-color gray-scale method one cannot determine if any cross-hybridization has occurred. However, with this “multiple color” scanning Raman method, one can carefully study the SERS spectra of each spot to determine which labels make up each spot. For the experiments described in Fig. 3, where the sequences are very dissimilar, we found that other than the expected spectroscopic probe signature for each target, there were virtually no other detectable Raman lines, indicating no cross-hybridization between different targets and probes. The SERS signal was obtained only from areas of the substrate where the Raman dye-labeled Au particles have initiated Ag formation. Therefore, this “multiple color” scanning Raman detection method does not record background signal due to Ag deposition where Au particles do not exist. This is not the case for the previous gray-scale scanometric approach, especially at ultralow target concentrations (≤50 fM) (1–3). The current unoptimized detection limit of this Raman scanning method is 20 fM (11).

Figure 3

(A) Flatbed scanner images of Ag-enhanced microarrays and (B) corresponding Raman spectra. The colored boxes correlate with the color-coded Raman spectra in Fig. 2. No false-positives or false-negatives were observed.

One would like to be able to use such detection systems to differentiate single nucleotide polymorphisms (SNPs), and in the case of gene expression studies, one would like access to RNA detection with single-spot signal ratioing capabilities. Nanoparticle probes heavily functionalized with oligonucleotides exhibit extraordinarily sharp thermally induced denaturation transitions that lead to substantially higher selectivity than conventional molecular fluorophore probes and microscopic bead probes in DNA detection (1, 19). However, thus far the behavior of these probes in the context of RNA detection has not been explored. To further test the selectivity of this Raman-based system and its ability to identify SNP targets, we chose two RNA targets that can bind to the same capture-strand DNA but have a single-base mutation in the probe-binding regions (target 1:T1, normal; target 2:T2, single-base difference) (Scheme 2). Therefore, two DNA-functionalized nanoparticle probes (probe 1:P1, probe 2:P2), which differ in sequence and Raman label, are required to differentiate these two RNA target strands (Scheme 2). Seven separate tests were performed to show not only how the two targets (T1 and T2) can be differentiated, but also how mixtures of the two targets can be analyzed in semiquantitative fashion.

Figure 4

(A) Typical flatbed scanner images of microarrays hybridized with nanoparticles (1) before and (2) after stringency wash but before Ag enhancing, and (3) after Ag enhancing. (B) Raman spectra (from 1550 to 1750 cm−1) from the stained spots at different ratios of target 1 and target 2: (a) 1:0; (b) 5:1; (c) 3:1; (d) 1:1; (e) 1:3; (f) 1:5; (g) 0:1. The full Raman spectra from 400 to 1800 cm−1 are shown in the supporting text. (Inset) Profile of Raman intensity ratio (I 2/I 1) versus target ratio (T2/T1), whereI 1 is the Raman intensity at 1650 cm−1 (from probe 1, TMR labeled);I 2 is the Raman intensity at 1588 cm−1 (from probe 2, Cy3 labeled).

Scheme 1
Scheme 2

In each of these tests, a slide was treated with a 0.3 M NaCl PBS buffer solution containing T1 and T2 in different ratios (total concentration = 1 nM) in a humidity chamber. After 2 hours, the chip was washed with a 0.3 M NaCl PBS buffer to remove nonspecifically bound target. Then, the chip was treated with nanoparticle probes (P1 and P2 at 1:1 ratio, 2 nM total concentration) for 1.5 hours to effect hybridization with the overhanging region of the target sequences (Scheme 2). The chip was washed with 0.3 M NaNO3 PBS buffer solution to remove chloride ions and nonspecifically bound nanoparticle probes. There are four possible hybridization modes, namely, T1:P1, T2:P2, T1:P2, and T2:P1(Scheme 2). When the chip was developed by Ag enhancing without a previous stringency wash, the Raman measurements on the gray spots, which correspond to different solution target ratios, yielded nearly identical spectra in all seven experiments; these spectra also are almost identical to those obtained for a sample containing a 1:1 ratio of probe 1 and probe 2 (11). These data show that probe 1 and probe 2 are bound to the spots on the chip in equal amounts, regardless of the target composition on the spot.

To identify the target composition on the spots, one must apply a salt- or temperature-based stringency wash (1, 19). Accordingly, we used a salt stringency wash (8 mM NaCl PBS buffer) to selectively denature the imperfect duplexes (T1:P2 and/or T2:P1) (Scheme 2, C and D) but not the duplexes formed from the perfectly complementary oligonucleotides (T1:P1 and/or T2:P2) (Scheme 2, A and B). After stringency wash and subsequent Ag staining, the Raman analysis of the gray spots can be used to readily identify the target composition on the spots by their spectral fingerprints. In tests where only pure RNA target 1 or 2 is present, only signals for probe 1 or 2, respectively, were observed (compare Fig. 4B, “a” and “g”) (11). In the case of mixtures, signals for both probes (I 1 at 1650 cm−1 from probe 1 and I 2 at 1588 cm−1 from probe 2) were detected, and the intensity ratios are proportional to the ratios of the two targets in each experiment (Fig. 4B, inset).

Compared with fluorescence-based chip detection, this nanoparticle-based methodology offers several advantages. The ratio of Raman intensities can be extracted from a single Raman spectrum with single-laser excitation. Second, the number of available Raman dyes is much greater than the number of available and discernable fluorescent dyes (7, 9). Indeed, a Raman dye can be either fluorescent or nonfluorescent, but a minor chemical modification of a dye molecule can lead to a new dye with a different Raman spectrum even though the two dyes exhibit virtually indistinguishable fluorescence spectra (7). Therefore, this fingerprinting method offers potentially greater flexibility, a larger pool of available and nonoverlapping probes, and higher multiplexing capabilities than conventional fluorescence-based detection approaches. Finally, the method incorporates all of the previous advantages of Au- nanoparticle based detection, including several orders of magnitude higher sensitivity and many orders of magnitude higher selectivity than the analogous molecular fluorescence-based approach (1,19). When considered with previous demonstrations of the unique properties of nanoparticle probes (1–4,19–27), a strong argument is being made for nanoparticles as the next-generation labeling technology for biodiagnostic research.

Supporting Online Material

Materials and Methods

Schemes S1 and S2

Figs. S1 to S3

  • * To whom correspondence should be addressed: camirkin{at}


View Abstract

Navigate This Article