Anal. Chem. 2006, 78, 2055-2059
Detection of Protein Analytes via Nanoparticle-Based Bio Bar Code Technology Y. Paul Bao, Tai-Fen Wei, Phil A. Lefebvre, Hao An, Liangxiu He, Gregory T. Kunkel, and Uwe R. Mu 1 ller*
Nanosphere, Inc., 4088 Commercial Avenue, Northbrook, Illinois 60062
We describe a new format for the recently introduced bio bar code technology, which improves the dose response over 10 000-fold and thereby makes this technique analytically useful. Unlike other ultrasensitive protein detection methods, such as immuno-PCR or immunoRCA, the bio bar code technique does not employ any enzymes to achieve detection limits in the attomolar range. By sandwiching a target between a magnetic bead and an amplifier nanoparticle, a multiplicity of bar code oligonucleotides are released for each captured target analyte. These surrogate bar code targets are then hybridized to microarrays and detected with silver-amplified gold nanoparticle probes. Using PSA detection as a model, we demonstrate a linear dose response over at least 4 orders of magnitude in both target concentration and concomitant signal and a 1000-fold improvement in detection limit compared to the best ELISA system. High-sensitivity detection of proteins or other molecules is typically achieved via antibody-based sandwich assays whereby a primary antibody (Ab) is attached to a solid surface (microtiter plate or magnetic beads) and a secondary antibody is labeled with one or more detection moieties. The primary Ab serves to capture and isolate the target from the sample (e.g., through magnetic separation and washing), and the secondary Ab provides the means to generate a specific signal that correlates with the successful capture of the target. (Figure 1). The sensitivity of such an assay system is typically characterized by two parameters: the limit of detection (LOD) and the dose response. The former refers to the smallest amount of target that generates a detectable signal above background (conventionally defined as background plus 2-3 times the standard deviation of the background), and the latter describes the net amount of signal that is generated for each incremental increase in target concentration. Ideally, the doseresponse curve is linear over several orders of magnitude, starting with the lowest detectable target concentration, and has a slope near 1 (e.g. a 10-fold increase in target concentration generates a 10-fold increase in net signal). Although the sensitivity of these immunoassays strongly depends on the binding affinity of the antibodies, sensitivity is also limited by a variety of other factors that are specific to the assay configuration and the type of label and instrument system * To whom correspondence should be addressed. E-mail: nanosphere.us. 10.1021/ac051798d CCC: $33.50 Published on Web 02/14/2006
umuller@
© 2006 American Chemical Society
used for detection. Many different detection methods have been employed, but with the exception of a recent report based on nanowire field-effects,1 those based on the detection of photon energy have proven to be the most sensitive and convenient. These labels can be conjugated covalently to the secondary antibody, or they can be attached noncovalently in a later step of the assay. For routine use, the most sensitive commercially available assay systems have employed enzymes as label moieties, whereby the isolated antibody-target-enzyme complex is contacted with a substrate that is then converted to a colored, fluorescent, or luminescent compound. An example of such an enzyme-linked immunosorbent assay (ELISA) is shown in Figure 1. Detection limits for ELISAs are typically in the high picomolar to nanomolar range (i.e., 5-100 ng/mL for an average sized protein), but it may be as low as femtomolar (a few pg/mL). For example, the LOD for PSA detection by ELISA with a chemiluminescent substrate has been reported to be on the order of 3 pg/mL.2 Even more sensitivity can be achieved by either employing gold nanoparticle labels for detection3 or by adding an enzymatic amplification step.4-7 In the latter case, the captured target is first converted to a surrogate nucleic acid target that can be amplified by polymerase chain reaction (PCR) or rolling circle amplification (RCA), reaching LODs in the low femtomolar to high attomolar range. A high-sensitivity nanoparticle-based assay system that does not require any enzymatic target or signal amplification was originally introduced by Nam et al.8,9 and termed bio bar code assay. For protein detection, an amplifier nanoparticle that is coloaded with secondary Ab and double-stranded oligonucleotides is attached to the captured target. Since only one of the strands (1) Zheng, G.; Patolsky, F.; Cui, Y.; Wang, W. U.; Lieber, C. M. Nat. Biotechnol. 2005, 23, 1294-301. (2) Kucera, E.; Kainz, C.; Tempfer, C.; Zeillinger, R.; Koelbl, H.; Sliutz, G. Anticancer Res. 1997, 17, 4735-7. (3) Soukka, T.; Paukkunen, J.; Harma, H.; Lonnberg, S.; Lindroos, H.; Lovgren, T. Clin. Chem. 2001, 47, 1269-78. (4) Fredriksson, S.; Gullberg, M.; Jarvius, J.; Olsson, C.; Pietras, K.; Gustafsdottir, S. M.; Ostman, A.; Landegren, U. Nat. Biotechnol. 2002, 20, 4737. (5) Sano, T.; Smith, C. L.; Cantor, C. R. Science 1992, 258, 120-2. (6) Schweitzer, B.; Roberts, S.; Grimwade, B.; Shao, W.; Wang, M.; Fu, Q.; Shu, Q.; Laroche, I.; Zhou, Z.; Tchernev, V. T.; Christiansen, J.; Velleca, M.; Kingsmore, S. F. Nat. Biotechnol. 2002, 20, 359-65. (7) Schweitzer, B.; Wiltshire, S.; Lambert, J.; O’Malley, S.; Kukanskis, K.; Zhu, Z.; Kingsmore, S. F.; Lizardi, P. M.; Ward, D. C. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 10113-9. (8) Nam, J. M.; Stoeva, S. I.; Mirkin, C. A. J. Am. Chem. Soc. 2004, 126, 59323. (9) Nam, J. M.; Thaxton, C. S.; Mirkin, C. A. Science 2003, 301, 1884-6.
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006 2055
Figure 1. Schematic comparing a magnetic-bead-based ELISA with the bio bar code technology. The enzyme used in the ELISA is HRP, which converts TMB into a colored compound absorbing at 450 nm. In lieu of HRP, the bio bar code format uses gold nanoparticles coated with single-stranded oligonucleotides (bar codes), which can be released and hybridized to a microarray for detection via a second hybridization event with nanoparticle probes.
is attached covalently to the nanoparticle, the complementary strand (DNA bar code) can be released and serves as a surrogate target, which is then detected by hybridization to an array.9 When applied to the detection of prostate specific antigen (PSA),9 a widely used tumor marker, or to ADDL peptides10 that are implicated in Alzheimer’s disease, the LODs reported were in the attomolar range. More recently, two modifications of the bar code assay were introduced. The first version uses a colorimetric detection and was applied to the analysis of Cytokines, also reporting an LOD in the attomolar range.11 In this report, as with previous publications, the dose response shown is shallow, generating only a 50% increase in signal over a 10 000-fold increase in target concentration. In the most recent version of this assay system, the bar code oligos were attached to the amplifier nanoparticle via thiol linkages and were released with dithiothreitol.12 This may be the reason for the observed increase in the dose response, although the signal change was still only 2-10fold, with a concomitant 10 000-fold increase in target concentration. However, this assay system was designed to capture a synthetic RNA target and did not use any antibodies for capture or detection, which makes a comparison to the above-discussed antibody-based assay format not possible. Given the promise of ultrahigh sensitivity for protein detection, it is critical that this assay system be rigorously examined. Moreover, to qualify as an analytical assay system, a significantly improved dose response is required. We show here that with appropriate redesign of the antibody-based assay system, we can achieve a 10 000-fold improvement in the dose response without (10) Georganopoulou, D. G.; Chang, L.; Nam, J. M.; Thaxton, C. S.; Mufson, E. J.; Klein, W. L.; Mirkin, C. A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 22736. (11) Nam, J.-M.; Wise, A. R.; Groves, J. T. Anal. Chem. 2005, 77, 6985-8. (12) Thaxton, C. S.; Hill, H. D.; Georganopoulou, D. G.; Stoeva, S. I.; Mirkin, C. A. Anal. Chem. 2005, 77, 8174-8.
2056
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
significant loss in LOD. Moreover, we demonstrate the reliability and robustness of this revised assay through extensive repeat analysis. Finally, using an extensively studied protein target (PSA) we compare our assay to the most sensitive and commonly used assay format for this target. EXPERIMENTAL SECTION Bar Code Assay. Prostate specific antigen (PSA; SigmaAldrich Corp.) and 0.5 µg of magnetic beads (MB; Dynal Inc.), coated per manufacturer’s instructions with mouse anti-human PSA monoclonal antibody (Mab; Abcam, Inc. code number ab403), were mixed in a final volume of 250 µL in assay buffer (AB; 1 × PBS, pH 7.2, 0.025% Tween 20, 0.1% BSA) in 1.5-mL PCR tubes and shaken at 1200 rpm at 25 °C for 90 min, followed by addition of 1 µL (50 ng) of biotinylated goat anti-human Kallikrein 3 IgG (R&D System Catalogue no. BAF1344) and continued shaking at 25 °C for 30 min. After magnetic separation (Dynal Inc., MPC-9600), the MB were washed with 200 µL of AB, resuspended in 50 µL of AB containing 6.6 µg/µL tRNA and 1010 streptavidin-coated gold nanoparticles (either BBI Catalogue no. EM.STP15 with 15-nm diameter or BA.STP20 with 20-nm diameter) and shaken as above for 30 min. After magnetic separation, the MB were washed twice with 200 µL of AB, resuspended in 50 µL of AB containing 6.6 µg/µL tRNA and 1 nM biotinylated bar code oligo (specific 30-mer sequence containing a 30-base-long poly-dA tail), and shaken as above for 30 min. After magnetic separation, the MB were washed eight times with 200 µL of AB; resuspended in 50 µL of 98% formamide, 0.02% Tween 20, and 10 mM EDTA at pH 8.0; and the bar code oligos were released by incubation at 80 °C for 5 min. Bar Code Array Hybridization. The released bar codes were made 45% formamide, 3 × SSC, and 0.02% Tween 20 prior to
Figure 2. Detection of bar codes via array hybridization. (A) Results of three sets of experiments with two different bar codes. Blue squares indicate the normalized mean net signal intensities of eight separate hybridizations, and green squares shows the average of four independent hybridizations, both with bar code no. 1. Red squares indicate the normalized mean net signal intensities of 15 independent hybridizations with bar code no. 2. (B) The same data sets, except that 2 standard deviations of the mean background signal was subtracted from the mean net signal intensities for 95% statistical significance. In each case, separate dilutions of bar code oligo were hybridized at 40 °C for 2 h, silverstained, and scanned.
hybridization. Microarrays (CodeLink slides; GE Healthcare) were printed in-house with bar code capture oligonucleotides (complementary to specific 30-mer bar code sequence) and control sequences (noncomplementary sequence), whereby each slide received 10 arrays per slide with six repeats of each capture sequence per array. Nanosphere hybridization chambers were attached to each slide, separating each array physically. After loading 50 µL of bar codes, the slides were incubated for 90 min at 40 °C with shaking at 600 rpm. Signal probe mix ((50 µL containing 7.5 µL of 20 × SSC, 1 µL of 1% Tween 20, 5 µL of 10 nM 15-nm dT20 gold nanoparticle probe (Nanosphere, Inc.), and 36.5 µL of water)) was added, and incubation continued for 30 min. After hybridization, the slides were washed three times for 1-2 min in 0.5 N NaNO3, 0.02% Tween 20, 0.01% SDS, then twice in 0.5 N NaNO3 for 1-2 min. After a final quick wash (1-2 s) in 0.1 N NaNO3, the slides were spun dry. Bar Code Signal Detection. Equal volumes of Signal Enhancement A and Signal Enhancement B (both from Nanosphere, Inc.) were mixed and poured immediately over the slides inside plastic slide holders and incubated for 3.3 min at RT. Reactions were stopped with two washes in 5% acetic acid and a thorough rinse in H2O. Slides were spun dry, and the back of slides were carefully wiped to remove dust, salt, and other contaminants. The slides were scanned on an ArrayWorx scanner (Applied Precision) for 50-200 ms at 10-µm resolution and high dynamic range setting. Bar Code Image Analysis. Scanned images (16-bit TIFF) were analyzed with GenePix Pro v5.1 software (Axon Instruments). Mean spot intensities were first corrected for local background (mean pixel value of a similarly sized area in the vicinity of each spot) to generate raw spot intensities. The raw spot intensities of negative the control spots were then subtracted from those of the positive spots to generate net signal intensities. For normalization, the net signal intensities obtained for 5 pg PSA/ mL (bar code assay) were set to 1. HRP Assay. PSA was captured and reacted with biotinylated goat anti-human Kallikrein antibody as described above for the
bar code assay. After magnetic separation, the beads were resuspended in 100 µL of AB containing avidin-horseradish peroxidase (HRP) (200× diluted stock; R&D Systems Catalogue no. 890803) and incubated at room temperature for 30 min with shaking at 1200 rpm. The beads were washed four times in AB and two times in AB without BSA. The beads were then resuspended in tetramethylbenzidin (TMB) substrate (100 µL) and incubated for 12.5 min in the dark. The reaction was stopped by adding 100 µL of 2 M HCl, and the absorbance was measured with a SPECTRAFluor Plus (Tecan) at 450 nm. Net signal intensities were determined by subtracting the absorbance value of the negative controls (no target), and for normalization, the net signal intensity values at 400 pg/mL were set to 1. RESULTS AND DISCUSSION To realize the full potential of the bio bar code assay for quantitative and ultrasensitive detection of biomedical analytes, it is critical to develop all assay steps such that the overall assay provides a good dose response over several orders of magnitude without compromising the LOD. To achieve this, we have modified the bio bar code assay and broken it functionally into three steps for individual analysis and optimization. As shown in Figure 1, the detection of bar codes (step 3) is via hybridization to a slidebased array whereby the bar codes are first captured by slidesurface-immobilized capture probes and then detected via a second hybridization event that anneals an oligonucleotide-functionalized gold nanoparticle probe to the captured bar code. The gold nanoparticle probes are then enlarged by a brief exposure to silver solution, which results in deposition of elemental silver around the gold particles. This dramatically increases the ability of the gold probes to scatter light, allowing their detection at a density of 0.0025 particles/µm2 with low resolution optics and without the need for laser excitation.13,14 Consequently, on the order of 100 (13) Storhoff, J. J.; Marla, S. S.; Bao, P.; Hagenow, S.; Mehta, H.; Lucas, A.; Garimella, V.; Patno, T. J.; Buckingham, W.; Cork, W. H.; Mu ¨ ller, U. R. Biosens. Bioelectron. 2004, 19, 875-83. (14) Wang, Z.; Lee, J.; Cossins, A. R.; Brust, M. Anal. Chem. 2005, 77, 57704.
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
2057
Figure 3. Comparison of bio bar code to ELISA assay. Prostate specific antigen (PSA, MW 34 000) was serially diluted in buffer and served as target in seven separate bio bar code detection assays, each with 2-6 independent repeats (total N ) 22), and four independent ELISA assays. The normalized relative net signal intensities are represented by different symbols for each set of assays (bio bar code ) blue doseresponse curve, ELISA ) red dose-response curve). The inset shows the dose-response curve for the bio bar code assay after subtracting 2 standard deviations of the mean background signal from each data set.
hybridization events per 200-µm-diameter spot provide a detectable scatter signal above background, allowing the detection of 10 00020 000 bar code molecules per hybridization on a routine basis (Figure 2). Importantly, this hybridization-based detection of oligonucleotides has a linear dose response over several orders of magnitude, with a slope near 1, providing the basis for quantitation. We think that this assay is an extremely important quality control measure of the nanoparticles and bar codes used, since with the appropriate design, it allows quantitation of the number of bar codes per nanoparticle that can be released and recaptured and thereby sets a theoretical limit for the LOD of the assay. The efficiency of step 2, the conversion of captured target to a surrogate target with simultaneous amplification (i.e., release of bar codes), depends on the size and type of amplifier nanoparticle, its oligonucleotide loading density, and the efficiency of its attachment to the antibody-target sandwich. In the current format, we have chosen 15- and 20-nm gold nanoparticles that have been coated with ∼40 or ∼70 streptavidin molecules, respectively, per particle. We estimate that no more than two of the four biotin-binding sites are accessible per surface-adsorbed molecule, potentially providing for an up to 80-150-fold amplification. In theory, therefore, the release of 20 000 bar codes could be achieved by roughly 150-250 captured nanoparticles, potentially providing for detection of fewer than 500 targets. The limiting parameters here are similar to those of a typical ELISA, such as the dissociation constants of primary and secondary antibody, number of active antibody molecules per magnetic bead, concentration of beads, reaction time, and a variety of other factors that are known to have a significant effect on the overall assay 2058 Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
sensitivity.15,16 Most critical here, however, is the potential for background generated by nonspecific binding of amplifier nanoparticles to the magnetic beads. With a typical input of 1010 nanoparticles to achieve efficient binding to target-bound secondary Ab, a 99.999 999% reduction of input particles is required to achieve a signal-to-background ratio above 2 near the LOD of the assay. To achieve this, we had to minimize the input of magnetic beads by a factor of 300 as compared to the amount used by Nam at al.9 to give background levels that would allow detection of a signal above background noise. In addition, we did not co-load the amplifier nanoparticles directly with antibodies and bar codes. Instead, we have chosen streptavidin-coated nanoparticles, which have more biotin-binding sites per nanoparticle than the antibodybinding sites that co-loaded particles provide. This improves the binding kinetics for docking of nanoparticles to the biotinylated secondary antibody. The so-formed magnetic bead-targetnanoparticle complexes can now be washed extensively to remove nonspecifically bound nanoparticles before biotinylated bar codes are added. Using commercially available antibodies, magnetic beads, and streptavidin-coated 15- or 20-nm nanoparticles as carriers for bar codes, we have compared our version of the bio bar code technology with an optimized colorimetric ELISA technique using PSA as a model system (Figure 1). PSA in decreasing concentrations was captured by magnetic beads and contacted with the biotinylated secondary antibody before magnetic separation, washing, and detection either by the HRP-based colorimetric (15) Ekins, R. P.; Chu, F. W. Clin. Chem. 1991, 37, 1955-67. (16) Saviranta, P.; Okon, R.; Brinker, A.; Warashina, M.; Eppinger, J.; Geierstanger, B. H. Clin. Chem. 2004, 50, 1907-20.
ELISA or by the bio bar code assay system. The averaged results of several experiments are shown in Figure 3 and demonstrate that this technique offers a better than 1000-fold improvement in the detection limit over that achievable with the enzymatic signal amplification power of horseradish peroxidase, even with chemiluminescent substrate (in our hands as well as others’,17 chemiluminescent substrates for HRP do not improve detectability; data not shown). Importantly, the dose response is over 10 000-fold better than what has been reported previously8,11 and provides the expected change in signal with concomitant changes in target concentration over a 10 000-fold range. Note that the slope of this dose-response curve is significantly steeper than that of most “high-sensitivity” protein assays.1,9,11,14,18,19,20 Furthermore, our results demonstrate an absolute detection limit on the order of 1-10 fg PSA/mL (high attomolar). However, the detection limit is obviously a function of the background and its noise (i.e., standard deviation) and the statistics applied to the (17) Siddiqui, J.; Remick, D. G. J. Immunoassay Immunochem. 2003, 24, 27383. (18) Grubisha, D. S.; Lipert, R. J.; Park, H. Y.; Driskell, J.; Porter, M. D. Anal. Chem. 2003, 75, 5936-43. (19) Liang, R. Q.; Tan, C. Y.; Ruan, K. C. J. Immunol. Methods 2004, 285, 15763. (20) Niemeyer, C. M.; Adler, M.; Wacker, R. Trends Biotechnol. 2005, 23, 20816. (21) Adler, M.; Wacker, R.; Niemeyer, C. M. Biochem. Biophys. Res. Commun. 2003, 308, 240-50. (22) Niemeyer, C. M.; Wacker, R.; Adler, M. Nucleic Acids. Res. 2003, 31, e90. (23) Bao, Y. P.; Huber, M.; Wei, T.-F.; Marla, S.; Storhoff, J.; Mu ¨ ller, U. R. Nucleic Acids Res. 2005, 33 (2) e15.
calculation of LOD. When we subtract from the data shown in Figure 3 not only the mean background but also the mean background plus 2 standard deviations of that background, most data points below 10 fg are lost, and the LOD increases to ∼10 fg PSA/mL (inset). Either way, these results compete well with the LODs reported for immuno-PCR,20-22 but without the need for enzymes or thermocycling, and appear significantly better than those reported for PSA detection by immuno-RCA.7 Since we have used relatively small nanoparticles for the amplification step with moderate amplification power, the power of this technique is currently based mostly on the high sensitivity with which we can detect released bar codes. Further improvements in the LOD should be possible if larger particles can be used for bar code loading and if those particles can be efficiently captured by the antibody sandwich without concomitant increase in background. Combined with the obvious and previously demonstrated multiplexing power of array hybridizations,23 these results suggest that the bio bar code technology can become a simple and robust multiplex assay system for ultrahigh sensitivity detection of proteins and other target analytes. COMPETING INTEREST STATEMENT The authors declare a competing financial interest.
Received for review October 7, 2005. Accepted January 19, 2006. AC051798D
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
2059
