NANO LETTERS
Qdot Nanobarcodes for Multiplexed Gene Expression Analysis
2006 Vol. 6, No. 5 1059-1064
P. Scott Eastman,†,| Weiming Ruan,⊥,†,§ Michael Doctolero,| Rachel Nuttall,| Gianfranco de Feo,‡,| Jennifer S. Park,# Julia S. F. Chu,# Patrick Cooke,∇ Joe W. Gray,⊥ Song Li,# and Fanqing Frank Chen*,⊥ Quantum Dots Corporation, Hayward, California 94545 (now InVitrogen, Inc., Carlsbad, California 92008), Lawrence Berkeley National Laboratory, UniVersity of California, Berkeley, California 94720, Department of Bioengineering, UniVersity of California, Berkeley, California 94720-1762, and Affymetrix Inc., 3420 Central Expressway, Santa Clara, California 95051 Received April 8, 2006
ABSTRACT We report a quantum dot (Qdot) nanobarcode-based microbead random array platform for accurate and reproducible gene expression profiling in a high-throughput and multiplexed format. Four different sizes of Qdots, with emissions at 525, 545, 565, and 585 nm are mixed with a polymer and coated onto the 8-µm-diameter magnetic microbeads to generate a nanobarcoded bead termed as QBeads. Twelve intensity levels for each of the four colors were used. Gene-specific oligonucleotide probes are conjugated to the surface of each spectrally nanobarcoded bead to create a multiplexed panel, and biotinylated cRNAs are generated from sample total RNA and hybridized to the gene probes on the microbeads. A fifth streptavidin Qdot (655 nm or infrared Qdot) binds to biotin on the cRNA, acting as a quantification reporter. Target identity was decoded based on spectral profile and intensity ratios of the four coding Qdots (525, 545, 565, and 585 nm). The intensity of the 655 nm Qdot reflects the level of biotinylated cRNA captured on the beads and provides the quantification for the corresponding target gene. The system shows a sensitivity of e104 target molecules detectable with T7 amplification, a level that is better than the 105 number achievable with a high-density microarray system, and approaching the 103−104 level usually observed for quantitative PCR (qPCR). The QBead nanobarcode system has a dynamic range of 3.5 logs, better than the 2−3 logs observed on various microarray platforms. The hybridization reaction is performed in liquid phase and completed in 1−2 hours, at least 1 order of magnitude faster than microarray-based hybridizations. Detectable fold change is lower than 1.4-fold, showing high precision even at close to single copy per cell level. Reproducibility for this proof-of-concept study approaches that of Affymetrix GeneChip microarray, with an R2 value between two repeats at 0.984, and interwell CV around 5%. In addition, it provides increased flexibility, convenience, and cost-effectiveness in comparison to conventional gene expression profiling methods.
In the past decade, high-throughput, multiplexed screening, best represented by microarray analysis, is increasingly used as an indispensable tool for research, diagnosis, and drug screening. However, conventional microarrays have limitations in flexibility, speed, cost, and sensitivity. Previously, we have reported using quantum dots in microarray analysis to improve sensitivity.1 We report here a more flexible and cost-effective high-throughput system using magnetic microbeads nanobarcoded by highly fluorescent quantum dots (also called Qdots) as an alternative expression profiling * To whom correspondence should be addressed. Dr. Fanqing Frank Chen. Life Sciences Division, Lawrence Berkley National Laboratory, MS 977R0225A, 1 Cyclotron Rd., Berkeley, CA 94720. Tel: (510) 495-2444. Fax: (510) 486-5586. E-mail:
[email protected]. † Contributed equally. ‡ Currently at NuGEN Technologies Inc., San Carlos, CA 94070. § Currently at Children’s Hospital and Research Center at Oakland, Oakland, CA 94609. | Quantum Dots Corp. ⊥ Lawrence Berkeley National Laboratory, University of California, Berkeley. # Department of Bioengineering, University of California, Berkeley. 3 Affymetrix Inc. 10.1021/nl060795t CCC: $33.50 Published on Web 04/22/2006
© 2006 American Chemical Society
platform with accuracy and sensitivity matching and exceeding the microarrays. Quantum dots are CdSe/ZnS core/shell nanocrystal fluorescence emitters with high quantum yield and extreme photostability.2 Their emission wavelengths can be tuned through size variations, while maintaining narrow bandwidth, and still be excited using a single blue or UV light source.2 Previously, in a proof-of-concept paper, Han et al.3 prepared 1.2 µm polystyrene microspheres containing three colors of quantum dots in controlled ratios. Each type of quantum dot in the bead has 10 intensity levels, thus creating a Qdot code with ∼103 combinatorial possibilities. Each type of coded polymer bead is conjugated to an oligonucleotide DNA so that the DNA probe can be identified by the corresponding code. The coded beads were used for detection of the Cascade Blue-labeled oligo DNAs at the single-bead level through hybridization in solution.3 Here we have developed a different nanobarcoded bead platform (Figure 1) that can not only identify but also accurately quantify the gene expression variations in a high-
Figure 1. Schematics of Qdot nanobarcoded microbead system for high-throughput gene expression analysis. (A). Pseudocolor picture of the microbeads embedded with Qdots, real color image is in Figure S1C. (B). Example spectra of the beads coded with different mixture of Qdots (Figure S1A and B). (C). Construction of the nanobarcoded microbeads. Each bead has a distinctive ratio of four different Qdots, allowing identification by a characteristic spectral nanobarcode. The transcript-specific oligonucleotide probes are conjugated to the bead surface. Therefore, each spectral-barcoded bead detects a specific oligomucleotide determined by the probe. (D) Gene expression monitoring and quantification sandwich assay. The nanobarcoded microbead-attached oligo probes capture biotinylated cRNA sample through hybridization, the cRNA is further sandwiched by the 655 nm streptavidin Qdots (or 705 nm, 800 nm) to be quantified. The gene expression variation is measured by fluorescence levels upon imaging with the Mosaic scanner (Quantum Dot Corp.).
throughput and multiplexed format. We used 8-µm-diameter magnetic beads (Figure 1C) as the core to achieve convenient manipulation and automation during liquid handling, and to 1060
lower the background. One major advantage of using the magnetic core is that unbound cRNAs or reporter fluorophores can be washed off easily when these microbeads are Nano Lett., Vol. 6, No. 5, 2006
subjected to a magnetic field. Four different sizes (and thus four different fluorescent colors) of Qdots, with emissions at 525, 545, 565, and 585 nm, that is, 20 nm spacing between the peak wavelengths, are mixed with a polymer and coated onto the microbeads to generate a nanobarcoded microbead termed as QBeads (Figure 1B). The nanobarcoding scheme in previous publications3 was changed in a way that 12 intensity levels for each of the four colors were used. Moreover, we fixed the total light intensity level at 12 so that I525 nm + I545 nm + I565 nm + I585 nm ) Imax, where each individual Qdot intensity (Ix) can have 12 increments of Imax/12. As a result, all QBeads have the same total light intensity and noise levels when excited at 405 nm, and there are no weak emitting beads that are affected more by the noise than the others. This is one of the issues encountered in the previous barcoding scheme, where, for example, barcode (1,1,1) has 1/10th the signal-to-noise ratio of barcode(10,10,10), while the reporter color still introduces same level of noises, though the issue is not critical. The total combinations of nanobarcodes can be calculated as N-1
Cbarcode )
(NCi ‚M-1CN-i-1) ∑ i)0
where N is the total number of Qdot colors used for barcoding, M is the total levels discernible within each color, and i is the number of color(s) that is(are) blank. Using four colors, 455 genes can be theoretically monitored at the same time (Supporting Information Table S1, Figure S1D), although in this current proof-of-principle study, only 100 genes and 20 calibrator sequences are used. In this proof-of-principle report, gene-specific oligonucleotide probes are conjugated to the surface of each spectrally nanobarcoded bead to create a multiplexed panel, and biotinylated cRNAs are generated from sample total RNA and hybridized to the gene probes on the microbeads, (Figure 1C). A fifth streptavidin Qdot (655 nm or infrared Qdot emitting at 705 and 800 nm, also named IR-Qdot) binds to biotin on the cRNA, acting as a quantification reporter. The biotinylated cRNA is sandwiched between the nanobarcoded microbead and the streptavidin Qdot reporter (Figure 1D). The nanobarcoded beads are captured at the bottom of a multiwell plate as a “random array”, as described for metallic nanobarcodes.4 The 655 nm streptavidin-Qdot, compared to the Cascade Blue dye, is much brighter and does not photobleach, eliminating the photobleaching issue during gene level quantification. Target identity was decoded based on the spectral profile and intensity ratios of the coding Qdots (525, 545, 565, and 585 nm) (Figure 1D). At the same time, the intensity of the 655 nm Qdot reflects the level of biotinylated cRNA captured on the beads and provides the quantification for the corresponding target gene. Quality control for the nanobarcode system is easy because each individual nanobarcode can be made in a batch of one gram quantity, and each gram will be enough to perform at least 109 assays. The system shows absolute detection sensitivity of 106 target molecules without T7 amplification of target genes (Figure 2A), and a sensitivity of e104 target Nano Lett., Vol. 6, No. 5, 2006
molecules detectable with only one round of T7 amplification (Figure 2B), a level that is better than the 105 number achievable with a high-density microarray system, and approaching the 103-104 level usually observed for quantitative PCR (qPCR) (Figure 2). If two rounds of T7 amplification are used, then the system can reach sensitivity level better than qPCR. The relative quantification level for gene expression is at single gene copy per cell (Figure 2A-C). The QBead nanobarcode system has a dynamic range of 3.5 logs, although not as good as the 6 logs achievable by qPCR, it is at least equal to or better than the 2-3 logs observed on various microarray platforms (Figure 2B). Because the hybridization reaction is performed in the liquid phase, it can be completed in 1-2 hours, at least 1 order of magnitude faster than microarray-based hybridizations. Detectable fold change is lower than 1.4-fold according to spike-in experiments, showing high precision even at close to single copy per cell level (Figure 2C). Reproducibility for this proof-ofconcept study approaches that of Affymetrix GeneChip microarray, with an R2 value between two repeats at 0.984 (Figure 2D, 3A, 3B, S6), and inter-assay CV around 5%. To further demonstrate the nanobarcode system, we used the Affymetrix high-throughput automations (HTA) platform, utilizing the Human U133A2.0 GeneChip technology5 for comparison studies. The GeneChip platform is a reliable gene expression system and is currently accepted as the industry standard. We investigated transforming growth factor-β1 (TGF-β1)-induced transcription changes in human bone marrow mesenchymal stem cells (MSCs) with the nanobarcode-based system and the HTA system. Upon treatment with TGF-β1 for 3 days, the MSCs have a more spread out and myoblast-like morphology, showing a significant increase in actin filament bundles compared to untreated MSC samples (Supporting Information Figure 3A and B). Expression levels of 100 selected target genes were examined, and 20 housekeeping genes were also included as calibrators. Because only 92 of the 100 genes in the nanobarcode system are present in the Affymetrix HTA U133A2.0 GeneChip, we focused on these 92 for cross-platform comparison. An amount of 100 ng total RNA were used in the nanobarcode system, and 2 µg of total RNA in the HTA system, with both amount the current lower limit for T7 amplificationaided assays for the respective systems. Therefore, the nanobarcodes use significantly less RNA than the Affymetrix microarray because of increased sensitivity. Under the nanobarcode system, three sets of control or TGF-β1-treated MSCs (treatment replicates) were employed in the test, and every set was then hybridized to the nanobarcodes three times (hybridization replicates), thus creating nine independent measurements. Relative fluorescence measurements throughout the 92 selected genes reflect highly reproducible gene expression levels among hybridization triplicates in a single MSC control treatment analysis (Supporting Information Figure S6A) and in a single TGF-β1-treated assay (Figure 3A). Likewise, the relative fluorescence levels between the three different control treatment replicates were also at comparable levels (Figure S6B), and similar results were obtained in the three different TGF-β1-treated replicates 1061
Figure 2. Sensitivity, dynamic range, and reproducibility of the Qdot nanobarcode bead. (A) Target detection sensitivity without T7 amplification step. 106 copies of target oligo can be detected. (B) Target detection sensitivity with T7 amplification step, less than 104 copies of target genes can be detected. RNA of three different genes are spiked into 100 ng of total RNA mixture and measured. The dynamic range is about 3.5 orders of magnitude in log. (C) Spike experiments indicate single copy per cell sensitivity, 1.4-fold change can be detected with more than 95% confidence. (D) Reproducibility between samples, two hybridization replicates show a correlation of more than 0.98 R2 value.
(Figure 3B). Most importantly, the gene expression data obtained with nanobarcodes shows high correlation with the Affymetrix Genechip microarray data. We calculated the TGF-β1-treated/control ratio for individual genes as fold changes, which is the ratio of the relative 655 nm Qdot fluorescence of the genes in TGF-β1-treated cell to that of the untreated cells. When the fold changes for all 92 genes are compared, the correlation coefficient is 0.86 between the nanobarcodes and Affymetrix GeneChip systems (Figure 3D). Because many genes have small fluctuations with no meaningful variations, we set thresholds to 10%, 20%, and 30% changes from the baseline. This allows us to reduce the noise in the biological system by filtering out the genes that changed minutely (e.g., 0.90 (Figure 3D). While we did not observe 100% correlation, the level of correlation is similar to what we observe when we compare qPCR with HTA, and is within a satisfactory range. For one of the genes that does not agree 1062
between the Affymetrix HTA GeneChip and Qdot nanobarcode system, the nanobarcode results preliminarily show better correlation to the qPCR and immunostaining results, although further confirmation is needed (Figure S5). Judging from the above data, the nanobarcodes yield highly consistent and reproducible results. Compared with the Affymetrix Genechip microarray, the bead system offers improved hybridization speed, increased sensitivity, reduction in sample amount, and increased flexibility, and the data correlate closely with the microarray platform. First, hybridization is limited by diffusional transport in a solid surface in microarray. The magnetic microbeads could have accelerated hybridization by allowing more uniform mixing in the liquid phase, and the washing step is cleaner and faster with the magnetic capturing. DNA microarray is typically performed in triplicate. Because the nanobarcode microbeads can be assayed at 20 000 per second, it would be reasonable to perform 50 or 100 replicate beads per gene in a single well experiment. Furthermore, though only 1/20 of the cRNA sample is used, the bead system yields a very high correlation Nano Lett., Vol. 6, No. 5, 2006
Figure 3. (A) High reproducibility between technical replicates of TGF-β1-treated cells. The data for technical replicates from control cells are shown in Figure S6A. (B) High reproducibility between treatment replicates. The data for treatment replicates from control cells are shown in Figure S6B. (C) Fold-change comparison of gene expression of TGF-beta1-treated to untreated cells as measured by both the Affymetrix GeneChip array and nanobarcoded bead platform. (D) Correlation coefficients of variance at different cutoff threshold between the Affymetrix GeneChip array and nanobarcode platform. The gene expression data in C was used. Gene names for the panel are presented in Table S1.
coefficient compared with the GeneChip microarray, suggesting a much higher sensitivity. Finally, the GeneChips or other microarrays are fixed assays. Each GeneChip has its own set of 100 lithography masks, which are timeconsuming and expensive to make. There are recent developments of flexible microarrays, but expensive instruments are required.6,7 Yet, the Qdot nanobarcode offers a flexible assay, different Qbeads can be “dialed in” to screen for different genes based on the results of a previous assay, and are inexpensive to assemble on the fly with a simple robotics system. Thus, the costs for the bead assay could be much lower with only the relevant genes to study; a set of 100 genes can have comparable price to multiplexed qPCR assay, yet with much higher multiplexity. Illumina has developed a randomly assembled array of beads in wells and coded the beads with oligo zipcodes. Even though this approach has been shown to have whole genome-encoding capacity, there requires a decoding hybridization that has to be performed for each individual chip.8 Luminex uses dual color bead assays involving antibodies, enzymes, toxins, and nucleic acids; however, the photoinstability of dye in the beads makes quantification less reproducible, and the multiplexity is difficult to expand, due to the broad bandwidth of the conventional fluorophores used in the Luminex beads.9 Nano Lett., Vol. 6, No. 5, 2006
The sensitivities in the Luminex system are not comparable to those observed with the QBeads. In the paper by Naciff et al.,10 one attomole sensitivity (or 6 × 105 copies of target) was achieved, while the nanobarcode system here can detect down to 1 × 104 copies, at least 1 order of magnitude more sensitive than the Luminex platform. Furthermore, Naciff et al.10 have only done a 20-plex, while the nanobarcodes can easily reach several hundreds with our conservative coding scheme (Figure S1, Supporting Information). If an aggressive coding scheme is adopted, as described by Han et al.,3 then the nanobarcodes can reach >104 combinations, even with the current four-color coding technology. Though only ∼100 genes were examined in this model assay, the four-color nanobarcoding scheme has the potential to monitor 455 genes simultaneously. The scale of genes analyzed in the bead system can be further expanded significantly by incorporating quantum dots of more colors and/or more intensity levels of each dot (Figure S1D). Qdots with emissions from 445 nm to 665 nm with 20 nm spacing can be used for coding, if infrared Qdot such as 705 nm and 800 nm are used for quantifications in place of the 655 nm Qdot used in this study. For instance, 8-12 Qdot colors combined with 12 intensity levels would yield between 50 388 to 1 352 078 codes (Figure S1D), enough to encode 1063
all the human genes (Figure S1D). It is worth noting that the Qdots with emission wavelengths lower than 505 nm have to be synthesized with ZnSe, ZnCdSe, or CdTe, rather than CdSe. One additional way to increase multiplexity is by increasing the number of the reporter colors. In addition to the 655 nm reporter Qdot used here, we have also used infrared (IR) Qdots with 705 or 800 nm emissions in other configurations of the sandwich assay (data not shown). The cRNA can have multiple labelings, such as digoxigenin, in addition to biotin. In this case, biotin can be quantified with streptavidin-705 nm Qdot and digoxigenin can be quantified with 800 nm Qdot, allowing two-color quantification in one well, introducing another dimension of multiplexity when multiplex reporter Qdots can be used simultaneously. Unlike dual color labeling in spotted microarray,11 which yields relative ratios between the two channels, both 705 and 800 nm channels yield absolute quantification. In conclusion, we have presented the first demonstrative study on a robust nanobarcoded microbead system that has broad applications in biology, chemistry, and medical diagnosis, and can potentially code more than one million combinations. The nanobarcode technology has the potential to impact genomics, proteomics, combinatorial chemistry, and other high-throughput screening assays. Nanobarcodes could be used as an alternative to perform single nucleotide polymorphisms (SNPs) genotyping, gene expression analysis, and comparative genomic hybridization. Another immediate application will be multiplexed protein quantification with picograms per milliliter protein detection sensitivity (study in progress). The oligonucleotide probes can be replaced by small molecule library, peptides, peptoids, aptamers, antigens, antibodies, proteins, microRNA probes, phages, or even bacteria, where the nanobarcode system provides the identification for the library content, and the reporter Qdot provides quantification.
1064
Acknowledgment. We thank Dr. Hang Cheng, Mr. Jeremy Semeik, and Ms. Pao-lin Shen for help with the Figures. The work is supported by NIH grants HL078534 and DARPA. This work was performed under the auspices of the U.S. Department of Energy, at the University of California/Lawrence Berkeley National Laboratory under contract no. DE-AC03-76SF00098. Supporting Information Available: This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Gerion, D.; Chen, F.; Kannan, B.; Fu, A.; Parak, W. J.; Chen, D. J.; Majumdar, A.; Alivisatos, A. P. Anal. Chem 2003, 75, 4766-4772. (2) Alivisatos, A. P. Nat. Biotechnol. 2004, 22, 47-52. (3) Han, M.; Gao, X.; Su, J. Z.; Nie, S. Nat. Biotechnol. 2001, 19, 631635. (4) Nicewarner-Pena, S. R.; Freeman, R. G.; Reiss, B. D.; He, L.; Pena, D. J.; Walton, I. D.; Cromer, R.; Keating, C. D.; Natan, M. J. Science 2001, 294, 137-141. (5) Ding, L.; Stilwell, J.; Zhang, T.; Elboudwarej, O.; Jiang, H.; Selegue, J. P.; Cooke, P. A.; Gray, J. W.; Chen, F. F. Nano Lett. 2005, 5, 2448-2464. (6) Tesfu, E.; Maurer, K.; Ragsdale, S. R.; Moeller, K. D. J. Am. Chem. Soc. 2004, 126, 6212-6213. (7) Nuwaysir, E. F.; Huang, W.; Albert, T. J.; Singh, J.; Nuwaysir, K.; Pitas, A.; Richmond, T.; Gorski, T.; Berg, J. P.; Ballin, J.; McCormick, M.; Norton, J.; Pollock, T.; Sumwalt, T.; Butcher, L.; Porter, D.; Molla, M.; Hall, C.; Blattner, F.; Sussman, M. R.; Wallace, R. L.; Cerrina, F.; Green, R. D. Genome Res. 2002, 12, 1749-1755. (8) Gunderson, K. L.; Kruglyak, S.; Graige, M. S.; Garcia, F.; Kermani, B. G.; Zhao, C.; Che, D.; Dickinson, T.; Wickham, E.; Bierle, J.; Doucet, D.; Milewski, M.; Yang, R.; Siegmund, C.; Haas, J.; Zhou, L.; Oliphant, A.; Fan, J. B.; Barnard, S.; Chee, M. S. Genome Res. 2004, 14, 870-877. (9) Keij, J. F.; Steinkamp, J. A. Cytometry 1998, 33, 318-323. (10) Naciff, J. M.; Richardson, B. D.; Oliver, K. G.; Jump, M. L.; Torontali, S. M.; Juhlin, K. D.; Carr, G. J.; Paine, J. R.; Tiesman, J. P.; Daston, G. P. EnViron. Health Perspect. 2005, 113, 1164-1171. (11) Shalon, D.; Smith, S. J.; Brown, P. O. Genome Res. 1996, 6, 639-645.
NL060795T
Nano Lett., Vol. 6, No. 5, 2006
