7
Enhancement of Immunoassays Fluorescence and Detection Sensitivity Using Three-Dimensional Plasmonic Nano-Antenna-Dots Array Liangcheng Zhou, Fei Ding, Hao Chen, Wei Ding, Weihua Zhang, and Stephen Y. Chou* Nanostructure Laboratory, Department of Electrical Engineering, Princeton University, Princeton, New Jersey 08544, United States ABSTRACT: Protein detection is universal and vital in biological study and medical diagnosis (e.g., cancer detection). Fluorescent immunoassay is one of the most widely used and most sensitive methods in protein detection (Giljohann, D. A.; Mirkin, C. A. Nature 2009, 462, 461464; Yager, P.; et al. Nature 2006, 442, 412418). Improvements of such assays have many signicant implications. Here, we report the use of a new plasmonic structure and a molecular spacer to enhance the average uorescence of an immunoassay of Protein A and human immunoglo- bulin G (IgG) by over 7400-fold and the immunoassays detection sensitivity by 3 000 000-fold (the limit of detection is reduced from 0.9 × 10 9 to 0.3 × 10 15 molar (i.e., from 0.9 nM to 300 aM), compared to identical assays performed on glass plates). Furthermore, the average uorescence enhancement has a dynamic range of 8 orders of magnitude and is uniform over the entire large sample area with a spatial variation ±9%. Additionally, we observed that, when a single molecule uorophore is placed at a hot spotof the plasmonic structure, its uorescence is enhanced by 4 × 10 6 -fold, thus indicating the potential to further signicantly increase the average uorescence enhancement and the detection sensitivity. Together with good spatial uniformity, wide dynamic range, and ease to manufacture, the giant enhancement in immunoassays uorescence and detection sensitivity (orders of magnitude higher than previously reported) should open up broad applications in biology study, medical diagnosis, and others. F luorescent immunoassay (FIA) identi es a protein biomarker (i.e., targeted analyte) in solution by selectively capturing it with a capture agent immobilized on a surface of a solid carrier and then tagging it with a detection antibody labeled with uorophores. The uorophores uorescent intensity is related to the existence and the concentration of the biomarker. Current FIAs oer detection sensitivity in the range of 10 12 M (1 pM). 1,2 The uorescence signals can be enhanced by metallic nanostructures through surface plasmons, termed metal enhanced uorescence(MEF). 35 Such enhancement could signicantly increase an assays detection sensitivity and hence is very desirable in many situations such as early detection of cancers. The enhancement magnitude and uniformity of MEF critically depend on the design of plasmonic (metal) nanostructures. Previously, the best observed average uorescence enhancements in MEF are 10190-fold for small molecules 614 and 10100-fold for immunoassay, 1522 and many of them cannot achieve good uniformity over a large sample area, 18,19,21 which limits their applications. Recently, we proposed and demonstrated a new plasmonic architecture that can circumvent several key shortcomings in previous plasmonic structures and has achieved signicant improvements in enhancement and uniformity of surface enhanced Raman scattering (SERS). 23 However, a plasmonic structure with high SERS enhancements does not mean a high MEF; and in fact, it can have a low MEF, as known both experimentally 3 and theoretically. 5,7 This is because MEF and SERS are dierent physical processes: MEF goes through a real quantum state, which can be quenched (namely, go back to a lower energy level through nonradiative transition) by the same metal that enhances the uorophore excitation rates; while SERS is a process through virtual quantum states which do not have quenching. Therefore, to enhance MEF, a layer of nonmetallic material, termed spacer, must be inserted between the plasmonic structures (e.g., the metal) and the uorophores and must be optimized to balance the excitation enhancement and the quenching, through the control of the coupling between the metal and the uorophores, namely, adjusting the spacers spacing (i.e., thickness) and permittivity. Too strong coupling (i.e., a small spacing) would lead to a strong quenching; too weak coupling (e.g., a large spacing) would signicantly reduce the plasmonic eld strength and hence the uorophore excitation enhancements. Furthermore, a high MEF for small molecules does not mean a high MEF for an immunoassay, because an immunoassay (a) has multiple layers of materials and (b) might need an additional adhesion layer to reliably attach the assay to metallic Received: February 1, 2012 Accepted: April 20, 2012 Published: April 20, 2012 Article pubs.acs.org/ac © 2012 American Chemical Society 4489 dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 44894495

Enhancement of Immunoassay s Fluorescence and …chouweb/publications/219 Zhou... ·  · 2012-06-18antenna array” (D2PA), has an array of dense three-dimensional (3D) resonant

Embed Size (px)

Citation preview

Enhancement of Immunoassay’s Fluorescence and DetectionSensitivity Using Three-Dimensional Plasmonic Nano-Antenna-DotsArrayLiangcheng Zhou, Fei Ding, Hao Chen, Wei Ding, Weihua Zhang, and Stephen Y. Chou*

Nanostructure Laboratory, Department of Electrical Engineering, Princeton University, Princeton, New Jersey 08544, United States

ABSTRACT: Protein detection is universal and vital in biological studyand medical diagnosis (e.g., cancer detection). Fluorescent immunoassayis one of the most widely used and most sensitive methods in proteindetection (Giljohann, D. A.; Mirkin, C. A. Nature 2009, 462, 461−464;Yager, P.; et al. Nature 2006, 442, 412−418). Improvements of suchassays have many significant implications. Here, we report the use of anew plasmonic structure and a molecular spacer to enhance the averagefluorescence of an immunoassay of Protein A and human immunoglo-bulin G (IgG) by over 7400-fold and the immunoassay’s detectionsensitivity by 3 000 000-fold (the limit of detection is reduced from 0.9 ×10−9 to 0.3 × 10−15 molar (i.e., from 0.9 nM to 300 aM), compared toidentical assays performed on glass plates). Furthermore, the averagefluorescence enhancement has a dynamic range of 8 orders of magnitudeand is uniform over the entire large sample area with a spatial variation ±9%. Additionally, we observed that, when a singlemolecule fluorophore is placed at a “hot spot” of the plasmonic structure, its fluorescence is enhanced by 4 × 106-fold, thusindicating the potential to further significantly increase the average fluorescence enhancement and the detection sensitivity.Together with good spatial uniformity, wide dynamic range, and ease to manufacture, the giant enhancement in immunoassay’sfluorescence and detection sensitivity (orders of magnitude higher than previously reported) should open up broad applicationsin biology study, medical diagnosis, and others.

F luorescent immunoassay (FIA) identifies a proteinbiomarker (i.e., targeted analyte) in solution by selectively

capturing it with a capture agent immobilized on a surface of asolid carrier and then tagging it with a detection antibodylabeled with fluorophores. The fluorophore’s fluorescentintensity is related to the existence and the concentration ofthe biomarker. Current FIAs offer detection sensitivity in therange of ∼10−12 M (1 pM).1,2 The fluorescence signals can beenhanced by metallic nanostructures through surface plasmons,termed “metal enhanced fluorescence” (MEF).3−5 Suchenhancement could significantly increase an assay’s detectionsensitivity and hence is very desirable in many situations such asearly detection of cancers. The enhancement magnitude anduniformity of MEF critically depend on the design of plasmonic(metal) nanostructures. Previously, the best observed averagefluorescence enhancements in MEF are 10−190-fold for smallmolecules6−14 and 10−100-fold for immunoassay,15−22 andmany of them cannot achieve good uniformity over a largesample area,18,19,21 which limits their applications.Recently, we proposed and demonstrated a new plasmonic

architecture that can circumvent several key shortcomings inprevious plasmonic structures and has achieved significantimprovements in enhancement and uniformity of surfaceenhanced Raman scattering (SERS).23 However, a plasmonicstructure with high SERS enhancements does not mean a highMEF; and in fact, it can have a low MEF, as known both

experimentally3 and theoretically.5,7 This is because MEF andSERS are different physical processes: MEF goes through a realquantum state, which can be quenched (namely, go back to alower energy level through nonradiative transition) by the samemetal that enhances the fluorophore excitation rates; whileSERS is a process through virtual quantum states which do nothave quenching. Therefore, to enhance MEF, a layer ofnonmetallic material, termed “spacer”, must be insertedbetween the plasmonic structures (e.g., the metal) and thefluorophores and must be optimized to balance the excitationenhancement and the quenching, through the control of thecoupling between the metal and the fluorophores, namely,adjusting the spacer’s spacing (i.e., thickness) and permittivity.Too strong coupling (i.e., a small spacing) would lead to astrong quenching; too weak coupling (e.g., a large spacing)would significantly reduce the plasmonic field strength andhence the fluorophore excitation enhancements.Furthermore, a high MEF for small molecules does not mean

a high MEF for an immunoassay, because an immunoassay (a)has multiple layers of materials and (b) might need anadditional adhesion layer to reliably attach the assay to metallic

Received: February 1, 2012Accepted: April 20, 2012Published: April 20, 2012

Article

pubs.acs.org/ac

© 2012 American Chemical Society 4489 dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−4495

(plasmonic) structures, both of them add extra materials to thespacer, hence affecting fluorescence enhancements.Therefore, it is of central importance, both fundamentally

and practically, to investigate (i) new plasmonic structures thatcan significantly enhance immunoassay’s fluorescence; (ii) theeffects of the spacer (i.e., the adhesion layers, the assay itself,and other spacers) on the properties of the fluorescenceenhancements (enhancement factor, uniformity, and dynamicrange); and (iii) the effects of MEF on the enhancement of theimmunoassay’s detection sensitivity (i.e., limit of detection,LOD). Here, we report our study of these issues, using afluorescent immunoassay of Protein A and human immuno-globulin G (IgG), on a new plasmonic structure with a self-assembled monolayer as adhesion layer.

■ EXPERIMENTAL SECTION

New Plasmonic Structure. The new plasmonic architec-ture that we developed, termed “disk-coupled dots-on-pillarantenna array” (D2PA), has an array of dense three-dimensional (3D) resonant cavity nanoantennas with denseplasmonic nanodots inside and the nanogaps that couple themetallic components (Figure 1).23 The 3D antennas greatlyincrease the efficiency in receiving and radiating light; themetallic nanodots and the nanogaps further “focus” the light tosmall regions to increase local electric fields, and the highdensities increase the average enhancement and uniformity.Particularly, the D2PA consists of a periodic nonmetallic (e.g.,dielectric or semiconductor) pillar array (200 nm pitch, ∼100nm diameter, and ∼65 nm height), a metallic disk (∼135 nmdiameter) on top of each pillar, a metallic backplane on the footof the pillars, metallic nanodots randomly located on the pillarwalls, and nanogaps between these metal components (Figure1). The disk array and the backplane (both are 55 nm thick)form a 3D cavity antenna that can efficiently trap the excitationlight vertically and laterally. The nanodots have diameters of

∼5−20 nm, and the nanogaps between them and the nanodisksare 1−10 nm. Each pillar has about 10 to 50 nanodotsdepending upon the pillar geometry, and the pillar density is 2.5× 109 pillars/cm2. The exact diameter and height of the pillarand metal disks, which were optimized to match thewavelengths of excitation laser and fluorescence, as well asthe roles of each element of D2PA in plasmonic enhancement,have been discussed elsewhere.23

The D2PA structures were fabricated on 4 in. fused silicawafers by a novel nanofabrication approach that combinesnanoimprint (top-down) with self-aligned self-assembly(bottom-up). The pillars were patterned first in the silicawafer by nanoimprint and reactive ion etching. Then, a thingold layer was evaporated onto the wafer in a direction normalto the wafer surface, which simultaneously deposited the goldnanodisk on the pillar top, the gold backplane, and goldnanodots on the pillar sidewall. The gold deposited on thepillar sidewall is much thinner than that on the top of thenanodisks and the backplane. Such thin gold is unstable anddiffuses at the elevated evaporation temperature; and togetherwith the nonwetting property of gold on SiO2 surface, the goldself-assembles into nanodots with a small gap in between andself-aligned precisely next to the gold nanodisk. Other details ofD2PA structure and fabrication are described elsewhere.23−25

Immunoassay, Fluorophores, Adhesion Layer, andReference. The immunoassay we used to test the enhance-ment in fluorescence and detection sensitivity was a direct assayof Protein A and human immunoglobulin G (IgG), which hasbeen widely used as the simplest model assay for such testing.19

Protein-A on a solid plate surface served as the capture agent tocapture the IgG (the targeted analyte) in a solution which isalready labeled with infrared fluorescent dye (IRDye800CW(Li-COR)); hence, there was no need to use additionaldetection agent. Protein A catches human IgG through thestrong Fc (Fragment, crystallizable) binding. The concentration

Figure 1. Disk-coupled dots-on-pillar antenna array (D2PA) plate and immunoassay. (a) Schematic (overview and cross-section) of D2PA platewithout an immunoassay. D2PA has an array of dense three-dimensional (3D) resonant cavity nanoantennas (formed by the gold disks on top ofperiodic nonmetallic pillars and the gold backplane on the pillar foot) with dense plasmonic nanodots inside and couples the metallic componentsthrough nanogaps. (b) Schematic of the immunoassay on the D2PA, consisting of a self-assembled monolayer (SAM) of adhesion layer, Protein-A(as capture layer) and human-IgG prelabeled with IRDye-800cw (as prelabeled biomarker). (c) Scanning electron micrograph (SEM) of D2PA with200 nm period (overview and cross-section). The gold nanodots rested on the silica nanopillar sidewalls are clearly observed.

Analytical Chemistry Article

dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−44954490

of IgG in solution was then quantified by measuring itsfluorescent labels.Since Protein A does not bind to the metal surface well, for

the D2PA plate, an additional adhesion layer must be usedbetween the metal and Protein A. This adhesion layer will addan additional material to the spacer and may weaken theplasmonic enhancement. To limit the total spacer thicknesswhile providing good binding of Protein A to the metal (gold inour case), we used a self-assemble-monolayer (SAM) ofdithiobis succinimidyl undecanoate (DSU) as the adhesionlayer. The DSU molecule has one end of sulfide which stronglybinds to a gold surface and the other end of N-hydroxysuccinimide (NHS) ester group which binds well toProtein A’s amine group.26 The SAM was ∼1.7 nm thick withrefractive index n = 1.50. Together with the Protein A layerwhich is estimated to be 4.8 nm,27 the total spacer thickness is6.5 nm.For comparison with the D2PA plate’s immunoassay

fluorescence enhancement measurements, we used plain flatglass plates as the reference. We prepared the immunoassay onthe D2PA plates and the reference in the same manner and thesame batch.Preparation of Fluorophore Label and Immunoassay.

The human IgG (Rockland Immunochemicals) was labeledwith the infrared fluorescent dye, IRDye800CW, in house.NHS ester group on the dye molecule was coupled to theamine group on IgG by mixing the reactive dye with IgGsolution and letting them react for 2 h at 20 °C in a darkenvironment. Free dye was separated through buffer exchangeusing desalting spin columns (Pierce Zeba). Each IgG has anaverage of 1.3 IRdye800CW molecules.For coating DSU SAM on the D2PA, the plates were

immersed in a solution of 0.5 mM DSU (Dojindo, Japan) in1,4-dioxane (Sigma-Aldrich) and incubated overnight at roomtemperature in a sealed container to form the SAM spacer.After incubation, they were rinsed extensively in 1,4-dioxaneand dried with argon gas and made ready for Protein Aimmobilization.Protein A (Rockland Immunochemicals) in phosphate

buffered saline (PBS) buffer solution (pH = 7.4, Sigma-aldrich)was dropped on the D2PA and reference plates, and each platewas incubated in a sealed container for 120 min at roomtemperature. Then, the plates were washed 3 times in washsolution (R&D systems) for 15 min each to remove theunbounded molecules. After coating Protein A, we diced theplates into 5 mm × 5 mm square pieces for testing. Thenfluorescence-labeled IgG in PBS solution was then dropped onthe Protein A layer and incubated in a sealed container for 60min at room temperature. After another washing in the samemanner, the plates were gently rinsed in streams of deionizedwater to remove any salt content. After drying with argon gas,the plates were optically measured immediately.To precisely control the concentrations of IgG on the plates,

we first prepare different concentrations of IgG in PBS solutionfrom 1 μM to 10 aM through serial dilution from stock solution(using a dispense pipet with ± 0.6% inaccuracy). Then, weprecisely dropped 3 μL of solution of each concentration onindividual square pieces (each 5 mm × 5 mm). The 3 μLvolume is chosen to make sure that the solution layer, aftercompletely wetting the plate top surface, will not spill to theplate’s back surface.Optical Measurements. The average fluorescence of the

immunoassays on the D2PA plates and the reference plates

were measured using a commercial laser scanning confocalspectrometer (ARAMIS, Horiba Jobin Yvon) with a 785 nmlaser excitation. The system uses a microscope lens to focus theexcitation laser beam normally on the sample surface and usesthe same lens to collect the generated fluorescence. The laserbeam was in a rapid raster-scanning (by a scanning galvo mirrorsystem) to homogenize the excitation over an area, termed“laser scan area”, which can be varied from a laser spot size(diffraction limited focal point) up to 100 μm × 100 μm. By astep-and-repeat of the laser scan area using an x−y stage, up to20 mm × 20 mm of the sample area can be measuredautomatically. The optical signals from a sample were sent to aspectrometer which consists of gratings and a CCD forspectrum measurement. Typically, we used a 10× objective(numerical aperture (N.A.) = 0.25), and 100 μm × 100 μmlaser scan area. For measuring single molecule fluorescence, weused another optical setup which gives 2D maps of opticalsignal over the excitation area (as discussed later).

■ RESULTS AND DISCUSSIONPlasmonic Resonance of D2PA Plate. We have

optimized the nanostructure of D2PA to make its plasmonicresonance close to the wavelengths of excitation laser (785 nm)as well as the absorption and emission peak of the fluorescentdye (IRDye-800CW) used in the immunoassay (780 and 800nm, respectively). Absorbance of the D2PA was obtained bymeasuring the transmission (T) and reflection (R) spectrumusing a white light source and calibrated to the samemeasurement performed on a glass plate (T = 94%) and silvermirror standard (R = 98%) respectively. The absorbance (1 −T − R) was found to have a resonance peak of 97% at 795 nmand a resonant full width at half-maximum (fwhm) of 145 nmfor the optimized D2PA plates without any molecular coating.After applying the immunoassay, the peak absorbance becomes98%; the resonance peak is blue-shifted slightly to 788 nm, andthe fwhm is 165 nm, slightly wider (Figure 2). A blue-shift,

rather than a common red-shift, which was also observed inanother plasmonic system,28 has been attributed to negativemolecular polarizability that destructively interferes with theoscillating polarization from the surface plasmon.28

Large-Area Average Fluorescence Enhancement over7400-Fold with Good Uniformity (±9% Variation). Thetypical fluorescence spectrum of the Protein A/IgG immuno-

Figure 2. Measured absorbance spectrum of D2PA with (blue line)and without (red line) the immunoassay being deposited. The peakabsorbance is 98% and 97%, and the resonance peak width is 165 and145 nm, respectively, with and without the immunoassay. Depositionof the immunoassay slightly blue-shifted the absorption peak from 795to 788 nm and widened the absorption wavelength range.

Analytical Chemistry Article

dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−44954491

assay on the D2PA plate and the reference (the flat glass plate)are given in Figure 3. Both spectra were measured on the assays

with 10 nM fluorescent-labeled IgG with a laser scan area of100 μm × 100 μm). The laser power and the detectorintegration time were 3 μW and 1 s for the D2PA and 212 μWand 8 s for the reference glass plate but were normalized inplotting Figure 3. Compared to the glass reference sample, theimmunoassay’s fluorescence intensity on the D2PA issignificantly enhanced. On the other hand, the fluorescencepeak wavelength is the same (800 nm) and the full-width at thehalf-maximum (fwhm) is nearly the same (∼ 30 nm) as thereference, which are due to the fact that D2PA’s plasmonicresonance has its peak optimized at 788 nm and a fwhm of∼165 nm, 5 times wider than the dye fluorescence peak, hencemaking the plasmonic enhancement factor nearly constant overthe entire wavelength range of the IRDye800CW fluorescence.

The average fluorescence enhancement of the immunoassayon the D2PA plate over a reference (the glass plate) wasobtained by:

λλλ

=II

II

EF( )( )

( )fluo.D2PA

fluo.REF

Exc.REF

Exc.D2PA (1)

where IExc.REF and IExc.D2PA are the intensity of laser excitationand IFluo.REF and IFluor.D2PA is the measured average fluorescenceintensity for the reference and D2PA plate, respectively. Notethat we measured the average fluorescence enhancements usingthe same IgG concentration for both the reference and theD2PA. This is to avoid the errors caused by the effects ofdifferent IgG concentrations. Moreover, to ensure the averageaccuracy, at least, a total of 5 different laser scan areas (each 100μm × 100 μm) over a sample were measured.Using the above approach, the measured average fluores-

cence enhancement of the D2PA plate at 10 nM fluorescent-labeled IgG was 7440-fold over the reference when thefluorescence peaks are compared and 7220-fold when thefluorescence intensities are integrated over the fwhm of thefluorescence spectrum. Figure 3 shows that the averagefluorescence enhancement spectrum has much broader fwhmthan the fluorescence spectrum, which is consistent with theobserved D2PA plasmonic resonance spectrum (Figure 2). For100 nM fluorescent-labeled IgG concentration, the averagepeak enhancement is 8460-fold. The immunoassay fluorescenceenhancements observed here are 2 orders of magnitude higherthan previous plasmonic enhanced fluorescence in immuno-assays.19,22

Another important feature of D2PA is the uniformity of thegiant average fluorescence enhancement over a large area. Wemeasured the uniformity of D2PA plates by mapping thefluorescence intensities of 10 nM fluorescence-labeled IgGconcentration over the entire 5 mm ×5 mm area of the D2PAplate. We used a laser scan area (100 μm × 100 μm), termed a“tile” in mapping; hence, there are a total of 2500 tiles (50 ×50) (Figure 4a). The laser power was 3 μW, and the integrationtime per tile was 1 s. Statistics performed on the mappingmeasurements showed that the average fluorescence enhance-ment over such large sample surface is 7000-fold with avariation (defined as the variation of a Gaussian distribution) of18% or ± 9% from the mean, very uniform everywhere (Figure4b).

Figure 3. Measured area-average fluorescence intensity spectrum ofthe human-IgG labeled with IRDye800CW captured by the assay onthe D2PA (red line) and the glass plate (blue line, which is amplified1000 times to be visible at given scales), respectively. Compared withthe assay on the glass plate, the average fluorescence enhancement(dashed line) is 7440-fold at the peak wavelength of fluorescence (800nm) and 7220-fold when average over the fwhm fluorescence. Theplasmonic fluorescence enhancement factor (EF) spectrum has muchbroader fwhm than the fluorescence spectrum, which is consistent withthe observed D2PA plasmonic resonance spectrum (Figure 2).

Figure 4. Measured uniformity of fluorescence enhancement over large area. (a) Measured immunoassay fluorescence enhancement (factor) mapover a total 5 mm × 5 mm area of the D2PA. The map has a total of 2500 tiles (50 × 50), measured using each tile area (i.e., laser probe area) of 100μm × 100 μm and a step-and-repeat distance of 100 μm. (b) The corresponding histogram of the measured enhancement factor gives a Gaussiandistribution variation of ± 9%.

Analytical Chemistry Article

dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−44954492

For the laser excitation power density and excitation timeused in the above measurements, we have not observed eithersaturation or noticeable bleaching, which are essential to ensureaccurate enhancement measurements. In fact, the fluorescencesignals from both D2PA and the reference samples are found tobe linear over a wide range of laser power density and dyeconcentration, which indicate no saturation. Moreover, thefluorescence vs time measurement showed that, under the laserintensities used, even for a time period much longer than thetypical measurement time we used, there is still no noticeablebleaching.Sub-Femto-Molar Detection Sensitivity and Wide

Dynamic Range. In clinical diagnostic applications, thedetection sensitivity (i.e., the limit of detection (LOD)) anddynamic range have more practical meaning than thefluorescence enhancement factor. To test these, we measuredthe immunoassay’s fluorescence signals from the D2PA and thereference prepared with different IgG concentrations: from 1μM to 10 aM (with a series dilution factor of 10). The LOD,determined using the well-accepted standard, is the IgGconcentration corresponding to the fluorescence signal that isequal to the background optical noise plus three times of itsstandard deviation (i.e., the root-mean-squared deviation). Inour experiments, the background optical noise was obtained byperforming the exactly same optical measurements on a blanksample as the sample with IgG (i.e., the same optical setup,sample area, laser power, and integration time). The blanksample was prepared on identical substrates using the samepreparation protocol except that the normal step of droppingfluorescent-labeled IgG is replaced by dropping of pure buffersolution (i.e., no IgG).Figure 5 shows the logritham plot of the fluorescence signals

versus the fluorescent-labeled IgG concentration deposited on

the D2PA and the reference (the response curve). Error barsare the standard deviation, calculated from the measurements atfive different sample areas for each concentration. To determinethe LOD, we first used a five-parameter logistic regressionmodel to create fitting curves which allow an extrapolation ofthe data points between the measured ones. The LOD of theimmunoassay on D2PA plates was found to be 0.3 fM (3 ×

10−16 M), and the dynamic range (where the fluorescence islinear with IgG concentration) is over eight (8) orders ofmagnitude (from 1 μM to 1fM) (Figure 5). On the other hand,the LOD of identical immunoassays performed on thereference plates (planar glass plates) was found to be 0.9 nM(0.9 × 10−9 M). Therefore, the detection sensitivity is enhancedby 3 000 000-fold (over 6 orders of magnitude) on the D2PAplate compared to the glass plate. This detection sensitivityenhancement is over 2 orders of magnitude higher than theprevious work using plasmonic structures.18,22

Fluorescence Enhancement of Single Molecule Fluo-rophore at Hot Spot up to 4 × 106-Fold. To explore thepotential in further enhancing immunoassay’s fluorescence anddetection sensitivity, we measured the fluorescence enhance-ment of the immunoassay from a single labeled IgG moleculewhich was placed at a “hot spot” of D2PA (namely, the regionwhere the local electric field is the strongest). Such singlemolecule fluorescence can be visible when the IgG moleculesare far apart from each other (i.e., a very low IgGconcentration) and a sensitive CCD camera is used.Particularly, we used an IgG concentration of 100 pM to

study single molecule fluorescence, which gives an averagedistance between two immobilized IgG about 420 nm. Wemapped the two-dimensional fluorescence of the immunoassayusing an inverted microscope (Nikon, USA) with a 40×objective lens (N.A. = 0.6). A 785 nm laser beam was expandeduniformly to illuminate a 50 μm × 50 μm area on D2PA plates.Images were continuously collected by an electron multiplyingcharge-coupled device (EM-CCD, Andor) of 512 × 512 pixelresolution (hence, ∼390 nm per pixel for the given laserscanning area). The CCD pixel size oversamples thefluorescence intensity distribution imaged at optical diffrac-tion-limit (0.8 μm determined by Rayleigh criterion).From the fluorescence imaging of 100 pM fluorescent-

labeled IgG on the D2PA plate (Figure 6), we observed distinctfluorescence “bright spots” that were randomly distributed in auniform background. The fluorescence intensity of individualbright spot as a function of time was shown to have a binarystepwise behavior (Figure 6b), which indicates that a singlemolecule at or near a D2PA’s hot spot first emits fluorescenceand then gets bleached.29

To estimate the fluorescence enhancement factor for a singlemolecule at a hot spot, gHotspot, we used two methods.30 Formethod one, gHotspot is the ratio of the single moleculefluorescence signal at a “hot spot” of D2PA, SHot.Spot, to theaverage fluorescence signal per molecule on the referencesample (which is equal to the area-average fluorescenceintensity on the reference sample, Iref.Avg, divided by the averageIgG molecules per unit area on the reference sample, nref.Avg.):

=gS I

I n I( / )Hot.SpotHot.Spot Exc.Ref

Ref.Avg Ref.Avg Exc.D2PA (2)

where IExc.D2PA and IExc.ref are the excitation intensity for theD2PA and reference plates, respectively. According to Figure6a, SHotspot = 1200 counts, Iref.Avg = 3088 counts/μm2, nref.Avg =7.22 × 105 molecule/μm2, Iref.Exc = 1.74 mW, and IExc.D2PA = 110μW. We found the fluorescence enhancement is gHotspot = 4.4 ×106, which is 3 orders of magnitude larger than most of thereported fluorescence enhancement for a single molecule in the“hot spot”.31

For the second method, the average fluorescence intensityper molecule for the reference was deducted from the average

Figure 5. Fluorescence intensity vs IgG concentration on D2PA(squares) and glass plate reference (circles). The squares and circlesare measured data, and the curves were the fittings using a five-parameter logistic regression model to allow an extrapolation of thedata points between the measured ones. The limit of detection (LOD)of D2PA and glass plate was found to be 0.3 fM and 0.9 nM,respectively, giving an enhancement of LOD of 3 000 000-fold.

Analytical Chemistry Article

dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−44954493

fluorescence intensity per molecule for the D2PA plate(ID2PA.Avg/nD2PA.Arg) divided by the fluorescence enhancementfactor (EF). The division of EF scales the signal from the D2PAplate to a regular glass plate:

=gS

I n( / )/EFHot.SpotHot.Spot

D2PA.Avg D2PA.Avg (3)

For ID2PA.Avg = 19 counts, EF = 7220, and nAvg∼7.22molecule/μm2, we found gHotspot = 3.28 × 106. Both methodsgave consistent results for calculating the single moleculefluorescence enhancements. The average of the two methodsgives gHotspot ∼ 4 × 106.Origin of Giant Assay Fluorescence and Detection

Sensitivity Enhancement and Uniformity. We believethere are three primary reasons for the observed largefluorescence and detection sensitivity enhancements. The firstand the most important is the unique D2PA plasmonicstructure. The second is a proper ultrathin spacer layer, and thethird is the possibility that the D2PA structure mightconcentrate the biomarkers into the hot spots of the D2PA.The plasmonic structure D2PA enhances the fluorescence

significantly through four key factors. (1) The 3D antennasarray is extremely efficient in receiving excitation light andradiating fluorescent light. As already shown in Figure 2, themeasured absorbance of the optimized D2PA is ∼97% at theexcitation laser wavelength of 785 nm. A good light absorber isalso a good radiator. (2) The small metallic dots and the smallgaps in the D2PA can strongly focus light to small regions tosignificantly enhance local electric fields, as already demon-strated in the SERS study.23 The smaller the dots and the gaps,the stronger will be the focusing (and local electric fieldenhancements).32,33 However, it is also well-known that smallmetallic structures of subwavelength size are extremely poor

light absorbers and radiators.34,35 Hence, only the small dotsand gap alone without antenna will not make a goodfluorescence enhancer, since it only can concentrate a smallportion of the incoming photons while most of the photons arethrown away, and it cannot efficiently radiate the fluorescencegenerated in the near field into the far field. (3) The effectivecoupling between the D2PA’s antennas and the nanostructuresthrough the nanogaps between them make the D2PA effectivein both receiving and radiating light and in locally focusing thelight to small spots. (4) The high densities of antennas, dots,and gaps allow a larger percentage of targeted molecules to benear the hot spots, hence increasing the fluorescence’s averageenhancement and uniformity and reducing the performancesensitivity to the device geometry variations. Since the finalplasmonic enhancement is a product of all four factors, anyplasmonic structures that lack one of the four factors could endup a poor plasmonic enhancer, which is exactly the problemsuffered by the most previous plasmonic structures and theexact reason why the D2PA, which improves all four factorstogether, is superior.To maximize the overall MEF fluorescence, the proper

ultrathin spacer layer plays a key role in balancing thefluorescence excitation and quenching by the same metal. Ina separate experiment where the only spacer between a D2PA’smetal and a fluorophore is a SiO2 layer (i.e., without any assay),we found that a 5 nm SiO2 thickness offers the best balancebetween MEF and quenching.30 Considering the total spacerlayer thickness for our D2PA’s assay is 6.5 nm (does notinclude IgG) and the effective permittivity is 3, this spacer has atotal effective dielectric distance very close to the 5 nm SiO2spacer. Therefore, our choice of self-assembled adhesion layer,DSU, offers a proper spacing for MEF.As to the possibility that the D2PA structure may

concentrate the biomarkers (targeted analytes) in a solutioninto the hot spots (i.e., large local electric field region), it is justa speculation and needs further experimental proof. Wespeculate three possible reasons for such concentration. (1)The drying of liquid on the D2PA surface will occur firstoutside the pillar sidewall, but the last is inside the gaps. Hence,the liquid movement during the drying may bring biomarkersfrom other locations into the gaps. (2) The local built-inelectric field in D2PA could moves biomarkers (if they are polarmolecules) to hot spots. (3) The SAM adhesion layer, DSU,adheres only to the gold not to the SiO2, which could led tomore Protein A and hence more fluorescent-labeled IgG ongold nanodots than the SiO2 sidewalls that are not covered bythe gold.There are two more important experimental facts that need

discussion. (1) The measured immunoassay fluorescence signalintensity does not drop as fast as the biomakers’ concentration(i.e., not in 1:1 ratio). This, known to the fluorescenceimmunoassays, is believed to be caused by the fact that thebonding of IgG to Protein A during the incubation and the lossof the IgG bonded on Protein A during the washing may bedifferent for different IgG concentrations. (2) The detectionsensitivity (LOD) enhancement by D2PA (3 000 000) is over400 times higher than the fluorescence enhancement (7400) at10 nM fluorescent-labeled IgG. We suspect this might be dueto biomarkers in a solution being concentrated into the hotspots of D2PA.It should also be pointed out that the fluorescence

enhancement by plasmonic structures is known to depend onthe intrinsic quantum efficient (QE) of a dye: stronger

Figure 6. Single molecule fluorescence of IRDye800CW labeled IgGon D2PA plate. (a) 2D fluorescence image of 50 μm × 50 μm area of aProtein A/IgG immunoassay on a D2PA plate with an IgGconcentration of 10−10 M. Distinct “bright spots” are visible. (b)Fluorescence vs time of a single bright spot. The binary stepwisebehavior indicates that the fluorescence is from a single dye moleculeplaced at a hot spot (large electric field location) of D2PA. Comparedwith the immunoassay on the glass reference, the single moleculefluorescence at a hot spot is enhanced by 4 × 106-fold.

Analytical Chemistry Article

dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−44954494

enhancement for a lower QE. At a low QE, the fluorescenceenhancement is inversely proportional to the QE. TheIRDye800cw dye has a quantum efficiency of 7%, similar tothe QE of the dyes used in other assay experiments. Even whenthe difference of the different dyes’s QEs is scaled, the observedaverage fluorescence enhancements are still 2 orders ofmagnitude higher than previous experiments. Finally, we needto point out that the large fluorescence enhancement factor ofsingle fluorophor at a hot spot observed in our experimentindicates that we might be able to further increase the assaydetection sensitivity significantly if we selectively place thebiomarkers into the hot spots.

■ CONCLUSIONSWe measured the fluorescence and detection sensitivity of animmunoassay with Protein A and the human-immunoglobulinG (IgG) tagged with near IR dye (IRDye800CW) on a newplasmonic structure, D2PA, with a self-assembled monolayer(SAM) of molecular spacer. We observed the averagefluorescence of the immunoassay on D2PA is enhanced by7400-fold and the detection sensitivity by 3 000 000-fold (thelimit of detection is reduced from 0.9 nM to 0.3 fM), comparedto identical assays performed on glass plates. Furthermore, theaverage fluorescence enhancement has a dynamic range of 8orders of magnitude (from 100 nM to 1 fM) and is uniformover the entire large sample area with a spatial variation of ±9%. We observed that, when a single molecule fluorophore isplaced at a “hot spot” of D2PA, its fluorescence is enhanced by4 × 106-fold, which indicates the potential to further increasethe average enhancements and the detection sensitivitysignificantly. The observed enhancements are orders ofmagnitude higher than previously reported. The large enhance-ments are attributed to the unique 3D architecture of the D2PAthat can overcome some key shortcomings in current plasmonicstructure design, as well as the use of the thin SAM molecularspacer. We suspect that D2PA may concentrate biomarkers in asolution into the hot spots of the plasmonic structure, whichcan further improve the enhancement. Our fabrication methodof D2PA plates is simple, inexpensive, and scalable. Togetherwith good spatial uniformity, wide dynamic range, and ease tomanufacture, the giant enhancements in the immunoassay’sfluorescence and detection sensitivity should have broadapplications in biology study, medical diagnosis, and manyothers.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: [email protected].

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSWe thank Dr. Haige Lu and Dr. Yueming Li from MemorialSloan-Kettering Cancer Center for valuable discussion on assaypreparation and Defense Advanced Research Project Agency(DARPA) for funding and support. S.Y.C originated the idea ofuse of the D2PA structures for immunoassay fluorescence anddetection sensitivity enhancement and designed and directedthe research. L.C.Z and S.Y.C designed and performed theexperiments of immunoassay preparation and optical character-ization. H.C, W.D, F.D, and S.Y.C designed and fabricated theD2PA samples. W.H.Z. contributed to the measurements of

single fluorophore at a hot spot. L.C.Z and S.Y.C contributed todata analysis.

■ REFERENCES(1) Giljohann, D. A.; Mirkin, C. A. Nature 2009, 462, 461−464.(2) Yager, P.; et al. Nature 2006, 442, 412−418.(3) Weitz, D.; Garoff, S.; Gersten, J.; Nitzan, A. J. Chem. Phys. 1983,78, 5324−5338.(4) Das, P.; Metiu, H. J. Phys. Chem. 1985, 89, 4680−4687.(5) Fort, E.; Gresillon, S. J. Phys. D: Appl. Phys. 2008, 41, 013001.(6) Wokaun, A.; Lutz, H. P.; King, A. P.; Wild, U. P.; Ernst, R. R. J.Chem. Phys. 1983, 79, 509−514.(7) Anger, P.; Bharadwaj, P.; Novotny, L. Phys. Rev. Lett. 2006, 96,113002.(8) Kuehn, S.; Hakanson, U.; Rogobete, L.; Sandoghdar, V. Phys. Rev.Lett. 2006, 97, 017402.(9) Tam, F.; Goodrich, G. P.; Johnson, B. R.; Halas, N. J. Nano Lett.2007, 7, 496−501.(10) Zhang, J.; Fu, Y.; Chowdhury, M. H.; Lakowicz, J. R. Nano Lett.2007, 7, 2101−2107.(11) Bakker, R. M.; et al. New J. Phys. 2008, 10, 125022.(12) Bek, A.; et al. Nano Lett. 2008, 8, 485−490.(13) Gerard, D.; et al. Phys. Rev. B 2008, 77, 045413.(14) Bardhan, R.; Grady, N. K.; Cole, J. R.; Joshi, A.; Halas, N. J. ACSNano 2009, 3, 744−752.(15) Attridge, J. W.; Daniels, P. B.; Deacon, J. K.; Robinson, G. A.;Davidson, G. P. Biosens. Bioelectron. 1991, 6, 201−214.(16) Sokolov, K.; Chumanov, G.; Cotton, T. M. Anal. Chem. 1998,70, 3898−3905.(17) Dostalek, J.; Knoll, W. Biointerphases 2008, 3, Fd12−Fd22.(18) Matveeva, E.; Gryczynski, Z.; Lakowicz, J. J. Immunol. Methods2005, 302, 26−35.(19) Zhang, J.; Lakowicz, J. R. Optics Express 2007, 15, 2598−2606.(20) Nooney, R. I.; Stranik, O.; McDonagh, C.; MacCraith, B. D.Langmuir 2008, 24, 11261−11267.(21) de la Escosura-Muniz, A.; Parolo, C.; Merkoci, A. Mater. Today2010, 13, 17−27.(22) Tabakman, S. M.; et al. Nat. Commun. 2011, 2, 466.(23) Li, W.-D.; Ding, F.; Hu, J.; Chou, S. Y. Opt. Express 2011, 19,3925−3936.(24) Chou, S.; Krauss, P.; Renstrom, P. Science 1996, 272, 85−87.(25) Chou, S. Y.; Krauss, P. R.; Renstrom, P. J. Appl. Phys. Lett. 1995,67, 3114−3116.(26) Wagner, P.; Hegner, M.; Kernen, P.; Zaugg, F.; Semenza, G.Biophys. J. 1996, 70, 2052−2066.(27) Erickson, H. P. Biol. Proced. Online 2009, 11, 32−51.(28) Zhao, J.; et al. J. Am. Chem. Soc. 2007, 129, 7647−7656.(29) Ambrose, W.; Goodwin, P.; Martin, J.; Keller, R. Phys. Rev. Lett.1994, 72, 160−163.(30) Zhang, W.; et al. Nanotechnology 2012, in press.(31) Kinkhabwala, A.; et al. Nat. Photonics 2009, 3, 654−657.(32) Schuller, J. A.; et al. Nat. Mater. 2010, 9, 193−204.(33) Nie, S. M.; Emery, S. R. Science 1997, 275, 1102−1106.(34) Fromm, D. P.; Sundaramurthy, A.; Schuck, P. J.; Kino, G.;Moerner, W. E. Nano Lett. 2004, 4, 957−961.(35) Farahani, J.; Pohl, D.; Eisler, H.; Hecht, B. Phys. Rev. Lett. 2005,95, 017402.

Analytical Chemistry Article

dx.doi.org/10.1021/ac3003215 | Anal. Chem. 2012, 84, 4489−44954495