Report

Fluorescence Enhancement at Docking Sites of DNA-Directed Self-Assembled Nanoantennas

See allHide authors and affiliations

Science  26 Oct 2012:
Vol. 338, Issue 6106, pp. 506-510
DOI: 10.1126/science.1228638

Abstract

We introduce self-assembled nanoantennas to enhance the fluorescence intensity in a plasmonic hotspot of zeptoliter volume. The nanoantennas are prepared by attaching one or two gold nanoparticles (NPs) to DNA origami structures, which also incorporated docking sites for a single fluorescent dye next to one NP or in the gap between two NPs. We measured the dependence of the fluorescence enhancement on NP size and number and compare it to numerical simulations. A maximum of 117-fold fluorescence enhancement was obtained for a dye molecule positioned in the 23-nanometer gap between 100-nanometer gold NPs. Direct visualization of the binding and unbinding of short DNA strands, as well as the conformational dynamics of a DNA Holliday junction in the hotspot of the nanoantenna, show the compatibility with single-molecule assays.

Single-molecule fluorescence measurements report on kinetic processes without the need for synchronization, lifetimes of intermediates, structure, stoichiometry of subpopulations, and the choreography of biomolecular processes (1, 2). Yet, only a small number of successful commercial applications, such as real-time single-molecule sequencing and single-molecule–based super-resolution imaging, have emerged that rely on single-molecule detection (3, 4). Single-molecule fluorescence applications pose two major limitations including (i) the concentration range of the fluorescent species suitable for single-molecule detection (typically in the picomolar to nanomolar regime) and (ii) the high brightness and photostability requirements for the fluorescent marker.

The field of nanophotonics offers several solutions to overcome both the concentration limitation and the brightness requirement. Circular holes of 50- to 200-nm diameter in a metal cladding film deposited on a transparent substrate (so called zero-mode waveguides), for example, reduce the observation volumes and enable monitoring of enzymatic reactions at high substrate concentration (5). This technique led to the visualization of transcription and translation at the single-molecule level (3, 6). However, the production and handling of these nanophotonics structures is costly and serial by nature. Because molecules are not specifically placed in the center of the structures, they undergo varying levels of fluorescence quenching because of the distribution of distances to the metallic walls yielding heterogeneous signals. Furthermore, zero-mode waveguide occupancy is limited by Poissonian statistics to a fraction of the waveguides to guarantee that only one molecule is present within one hole.

Besides the exclusion of light in zero-mode waveguides, metallic nanostructures can also enhance fluorescence by creating plasmonic nanoantennas that highly enhance local fields (7, 8). Lithographically fabricated nanoantennas not only enhance the excitation field in a very small volume but can also direct single-molecule emission (9) and increase quantum yields for the detection of low–quantum yield dyes (10, 11). Again, demanding nanolithography is required, and the problem remains how to place the molecules of interest with nanometer accuracy in the hotspot of the antenna structure (7). Current approaches to fabricate fluorescence enhancing structures via self-assembly or wet chemical synthesis do not achieve the required structural control, cannot provide docking sites in the hotspot, and fail to provide a reliable stoichiometry control for particles bigger than 40 nm (1215).

In this work, we used a self-assembly scheme based on the DNA origami technique (16, 17) to construct nanoantennas with docking sites (NADS) for biomolecular assays. The DNA origami allows a precise arrangement of dimers of arbitrary size at a defined interparticle distance, a docking site for objects of interest in the plasmonic hotspot between gold nanoparticles (Au NPs), as well as specific attachment sites for surface immobilization. Based on this platform, we studied the fluorescence enhancement by Au NPs of different size. The docking sites serve for placement of single-dye molecules or biomolecular assays such as a DNA binding assay or a Holliday junction.

A three-dimensional, pillar-shaped DNA origami (Fig. 1A) was folded from one M13mp18-derived scaffold strand and 207 staple strands [see supplementary materials and methods for experimental details (18)]. The DNA origami pillar has a length of 220 nm, a 15-nm diameter consisting of a 12-helix bundle, and three extra 6-helix bundles on the base, leading to a base diameter of 30 nm. The base contains 15 biotin-modified staple strands for selective immobilization of the pillar on cover slips coated with biotinylated bovine serum albumin and neutravidin (lower-right inset of Fig. 1A). During hybridization, a dye-labeled strand (ATTO647N) and three to six capturing strands (original staple strands extended by an extra 15 bases) were additionally incorporated at the half height of the pillar (see upper-right inset of Fig. 1A and fig. S1). After immobilization, DNA functionalized Au NPs were bound to the pillar by hybridization to the capturing strands (18). We used three capturing strands per Au NP to increase the rigidity of the interaction between DNA origami and Au NPs (19). This procedure minimizes sample consumption and avoids aggregation, even for large particles (19). The DNA origami design enables the alignment of the NADS with respect to the electric field polarization.

Fig. 1

(A) Sketch of the DNA origami pillar with two Au NPs forming a dimer. The dye (red sphere) is located between the NPs within the central bundle of the pillar. (Upper right inset) Binding of the NP to the DNA origami structure through three capturing strands. (Lower right inset) Binding of DNA origami to the neutravidin functionalized cover slip. BSA, bovine serum albumin. (B) Numerical simulation of electric field intensity for a monomer (left) and dimer (right) for 80-nm-diameter gold NPs and an interparticle spacing of 23 nm (dimer). The incoming light was horizontally polarized at a wavelength of 640 nm. (C) Numerical simulations of the fluorescence enhancement along the gap connecting the NPs for a dye oriented in the radial direction. a.u., arbitrary units.

Plasmonic nanostructures create highly enhanced local fields, which can lead to fluorescence brightness enhancement, depending on how the excitation as well as the radiative and nonradiative rates of the fluorescent dye are influenced (20). In general, a stronger enhancement occurs for larger NPs; additionally, the enhancement is more pronounced if the spacing between the two Au NPs is reduced (21, 22). We designed a gap of 23 nm between the NPs as a compromise of strong fluorescence enhancement and sufficient space for accommodation of a biomolecular assay. We used numerical simulations to estimate the influence of single metallic NPs and NP dimers on the excitation and emission of single fluorescent dyes. Figure 1B shows the normalized electric field intensity at the equator plane of a monomer (left) and dimer (right) when illuminated with a plane wave from below at the laser excitation wavelength of 640 nm. In our arrangement, the fluorescence brightness enhancement of a dye placed in the vicinity of a NP system can be approximated by the product of the quantum yield enhancement and the normalized electric field intensity enhancement (21, 23). A single 80-nm NP, for example, is expected to induce a maximum fluorescence enhancement of 10-fold relative to the unperturbed dye, and a plateau with an enhancement around 115 is simulated for a dye with radial orientation in the gap of a dimer with 23-nm separation (Fig. 1C). Very close to the NP, the fluorescence intensity drops to nearly zero because quenching dominates (21).

We used a 640-nm pulsed diode laser with circular polarization to excite fluorescence from the ATTO647N molecules in a custom-built confocal microscope. Figure 2A shows a fluorescence image of a low concentration of self-assembled nanoantennas with capturing strands for two 80-nm Au NPs. The image exhibits spots of a broad range of intensities and sometimes fluorescence intermittencies, which indicates the presence of a single fluorescent dye for even the brightest spots. We used phosphate buffered saline as the solvent, because ATTO647N exhibits pronounced blinking and moderate photostability so that single-molecule emission is readily visualized (24). The fluorescence transients shown in Fig. 2, B to D, were recorded from a dim spot, a spot with intermediate intensity, and a bright spot, as indicated in the image of Fig. 2A. All three dye molecules showed single-step photobleaching and revealed markedly different average intensities of 3.5, 11, and 200 kHz, respectively. The 3.5-kHz intensity was the typical fluorescence count rate of ATTO647N attached to the DNA origami pillar without NPs at otherwise identical conditions. The assignment of the dim spots to DNA origami pillars without NPs is supported by the long fluorescence lifetime of 3.80 ns, typical for ATTO647N (inset of Fig. 2B). The molecules depicted in Fig. 2, C and D, are substantially brighter and exhibit reduced fluorescence lifetimes of 1.17 and 0.22 ns obtained by deconvolution with the instrument response function [see fluorescence decays in the insets of Fig. 2, C and D (18)]. These transients represent emission from NADS carrying one and two gold NPs, respectively.

Fig. 2

(A) Confocal fluorescence image of NADS with binding sites for two NPs. Three spots with different intensities are highlighted: I (no NP), II (one NP), and III (two NPs). (B to D) Corresponding intensity transients together with their fluorescence decays (insets).

To further study the effect of the NPs on the fluorescent dyes and to substantiate our assignment to the number of NPs, we analyzed the emitted intensities and fluorescence lifetimes of several images for NADS samples with one and two binding sites for 80-nm gold NPs (Fig. 3). The sample with one binding site for Au NPs exhibits two maxima in the fluorescence lifetime histogram (Fig. 3A). The long fluorescence lifetime represents DNA origami pillars without a NP, and the population with a fluorescence lifetime with a maximum at 1.0 ns represents the population with one Au NP (Fig. 3A, red data points). The sample with two NP binding sites exhibits an additional maximum at 0.2 ns that we assign to DNA origami pillars with two NPs bound. Classification of molecules and the correlation of the fluorescence intensity and the fluorescence lifetime (Fig. 3A) directly show that fluorescent dyes bound to NADS with one NP were substantially brighter than ordinary ATTO647N dye molecules. NADS with two NPs provoke an even stronger fluorescence enhancement.

Fig. 3

(A) Scatter plot of fluorescence intensity versus lifetime with a corresponding lifetime histogram of the ATTO647N-labeled DNA origami pillar with binding sites for one (monomer) and two (dimer) particles. Au NPs with 80-nm diameters were used. Based on the lifetime, three populations are identified. The mean fluorescence enhancement (normalized) for different NP sizes, together with numerical simulations assuming a radial or tangential orientation of the dye’s dipole, is shown for a monomer (B) and a NP dimer (C). Error bars indicate the SD for the measurements and consider the size distribution of NPs for the simulations.

For quantifying the fluorescence brightness enhancement, we sorted the spots from the intensity lifetime plots by the number of NPs into separate intensity histograms as indicated in Fig. 3A. The average intensity of the population without NPs served as an internal reference to which the intensity of the other populations was normalized. This possibility for internal referencing and the ability to separate subpopulations were also the reasons why we did not aim for 100% binding efficiency of NPs to NADS, although we obtained higher binding efficiencies by longer incubation with higher NP concentrations (see fig. S2). Analogous results of the average fluorescence enhancement with specified NP diameters of 20, 40, 60, 80, and 100 nm (experimentally determined diameters were 19.4 ± 1.7, 43.5 ± 4.6, 61.5 ± 7.0, 80.1 ± 7.4, and 105 ± 10.4 nm; see fig. S3 for particle characterization) are shown in Fig. 3B for monomers and in Fig. 3C for dimers. For the smallest NPs, a reduction in the fluorescence brightness is observed, whereas for the bigger NPs, a clear enhancement occurred (21). For the 100-nm NPs, the average fluorescence brightness enhancement is 8-fold for the monomer and 28-fold for the dimer, with some dimers exhibiting more than 100-fold enhancement (see fig. S4). The distribution of enhancement factors is ascribed to the inhomogeneity of the Au NPs in size, shape, and distance, as well as to a possible deviation of the DNA origami pillar from vertical orientation on the surface.

To compare the measured enhancement with theoretical expectations, we performed numerical simulations (Fig. 1, B and C). These simulations determined the fluorescence enhancement for a radial and tangential orientation of the dye’s transition dipole (Fig. 3, B and C). In the first case, we observed a slight increase in the relative quantum yield accompanied by a drastic reduction in the fluorescence lifetime (fig. S5). For the tangential orientation, the relative quantum yield was dramatically decreased, whereas the lifetime was only slightly reduced. Because of the nature of the experiment, the dipole moment of the fluorophore was randomly oriented and rotating. ATTO647N attached to DNA had a typical rotational correlation time of 0.8 ns (25), which is of similar magnitude as the decay rates in the NADS. Thus, the probability of absorption and decay was biased toward radially oriented dyes, and a quickly rotating dye would show fluorescence enhancement resembling the results of the radial rather than the tangential simulation, with the simulation for the radial (or tangential) orientation representing an upper (or lower) enhancement limit. This upper limit is in good agreement with the highest measured values for each NP diameter, and we expect that the average measured values fall between the simulations for the radial and tangential dye orientation (Fig. 3, B and C).

We demonstrate the biocompatibility of the NADS by a transient DNA binding as well as a single-molecule fluorescence resonance energy transfer (FRET) assay. First, we observed binding and unbinding of single-stranded DNA (ssDNA) to the hotspot created by the NP dimer (26). For this reason, we modified the previous DNA origami pillar by replacing the ATTO647N dye inside the DNA bundle by five ssDNA strands (nine base pairs long) attached to the central bundle in the hotspot (Fig. 4A and fig. S1). We visualized binding and unbinding of single ATTO655-labeled counterstrands at a concentration of 100 nM (26). Figure 4B shows a fluorescence intensity transient in which the background of the freely diffusing, complementary strands has been subtracted, and the remaining intensity is normalized to the intensity of a single ATTO655 dye in the absence of NPs (the ordinate directly reports on the fluorescence enhancement). Short fluorescence bursts report on transient binding of the labeled ATTO655 strand to one of the protruding DNA strands in the hotspot of the NADS, showing that biomolecular interaction can be visualized at the docking sites of the NADS.

Fig. 4

(A) Sketch of the modified DNA pillar for the DNA binding assay. ssDNA sequences modified with ATTO655 (red) at 100 nM transiently hybridize with complimentary sequences (green) at the hotspot between the NPs. (B) Fluorescence intensity transient showing several binding and unbinding events on a DNA origami pillar with two NPs. Background was subtracted, and the intensity was normalized to the fluorescence without NPs. (C and D) Enlarged views reveal the magnitude of enhancement, as well as the effect of rotating linear excitation polarization. (E) Sketch of the DNA origami dimer with an incorporated HJ labeled with green and red dyes. Fluorescent intensity and FRET transients show oscillations of the HJ for a pillar with no NPs (F) and a NP dimer (G). The green, red, and brown transients represent donor excitation/donor emission, donor excitation/acceptor emission (FRET), and acceptor excitation/acceptor emission, respectively. Fluorescence decays of acceptor excitation/acceptor emission are shown in the insets.

To demonstrate that the fluorescence bursts were related to the nanoantenna, we rotated linear excitation polarization at a frequency of 50 Hz using an electro-optical modulator. The zoom in Fig. 4C reveals binding events with a relative intensity enhancement of 60-fold. A further zoom in Fig. 4D shows the effect of the rotating incident electric field polarization. Whereas a single fluorophore wiggling and rotating around the linker commonly exhibits fluorescence independent of the excitation polarization, the plasmon coupling depends strongly on the relative orientation of the incident light field, NP, and dye. At the hotspot of the dimer, an enhancement of the electric field occurs when the incident electric field is parallel to the dimer (Fig. 1B), leading to the highest fluorescence signal. As the polarization rotated, the plasmon coupling between the NPs decreased, as did the enhancement of the electric field and the fluorescence signal until the perpendicular orientation was reached, leading to a minimum.

In a single-molecule FRET assay, we visualized fluctuation of a Holliday junction (HJ) (27) bound to the hotspot of a dimer NADS (Fig. 4E and fig. S1). The HJ was directly incorporated to the DNA origami structure as in (28) and was labeled with Cy3 and Cy5 dyes (18). The anticorrelated intensity changes of the green (donor excitation/donor emission) and red (donor excitation/acceptor emission) detection channels shown in Fig. 4F directly reflected the HJ conformational dynamics in the presence of 100 mM magnesium (27). In addition, we used the intensity transient of direct acceptor excitation to estimate fluorescence enhancement as well as fluorescence lifetime (inset in Fig. 4F). Transients with a NP dimer bound such as the HJ depicted in Fig. 4G showed similar fluctuations at much higher fluorescence count rates. Despite faster bleaching at higher count rates, switching between the two conformations occurred at a similar rate of 3 Hz.

NADS created fluorescence enhancement in ultrasmall hotspots in the zeptoliter range that could be used for single-molecule detection at elevated dye concentrations. Additional time-gated detection making use of the difference of the fluorescence lifetimes inside and outside of the hotspot could almost fully recover the signal-to-noise ratio at close to micromolar concentrations (fig. S6). Alternatively, very fast kinetics—for example, in protein folding—could be resolved if the signal was large enough (29). Increasing the excitation power to only 7 μW (10 kW/cm2) yielded fluorescence signals of up to 10.6 MHz, allowing the direct visualization of 10-μs blinking events (see fig. S7).

Self-assembled nanoantennas with docking sites based on DNA origami scaffolds represent an inexpensive and versatile platform to study plasmonic effects of metallic NP systems. The approach offers a simple solution to one of the urgent needs, that is, the ability to couple optical sources to nanoantennas. We have studied the dependence of the fluorescence intensity and lifetime of single dyes placed in the vicinity of gold NP monomers and dimers of varying sizes. We achieved fluorescence enhancement of up to 117-fold for 100-nm dimers that enables higher count rates in single-molecule applications and relaxes the requirements for single-molecule–compatible fluorescent dyes. Our results are in good agreement with numerical simulations and show that substantial fluorescence enhancement can be achieved, even at an interparticle distance of 23 nm that allows for the accommodation of biomolecular assays. The reduction of the hotspot size far beyond diffraction-limited dimensions and the improved signal-to-noise ratio pave the way for sensor applications and nanoscale light control and extend the concentration range of single-molecule measurements toward the biologically relevant micromolar regime.

Supplementary Materials

www.sciencemag.org/cgi/content/full/338/6106/506/DC1

Materials and Methods

Figs. S1 to S9

Table S1

References (3034)

References and Notes

  1. See the supplementary materials on Science Online.
  2. Acknowledgments: We thank A. Gietl, J. J. Schmied, D. Grohmann, T. Liedl, and F. Stefani for fruitful discussion; A. Tiefnig for sample preparation; F. Demming for assistance with the numerical simulations; and S. Carregal Romero for the transmission electron microscopy images. This work was carried out at the NanoBioScience group, Braunschweig University of Technology, and was supported by a starting grant (SiMBA) of the European Research Council, the Volkswagen Foundation, and the Center for NanoScience. Technische Universität Braunschweig, G.P.A., and P.T. have filed a provisional patent application, EP1260316.1., on the described method of creating hotspots using self-assembled DNA origami.
View Abstract

Navigate This Article