A Nucleic Acid Biosensor for Gene Expression Analysis in Nanograms

Md Nazmul Islam , Mostafa Kamal Masud , Md Hakimul Haque , Md Shahriar Al Hossain , Yusuke Yamauchi , Nam-Trung Nguyen , Muhammad J. A. Shiddiky...
0 downloads 0 Views 167KB Size
Anal. Chem. 2004, 76, 4023-4029

A Nucleic Acid Biosensor for Gene Expression Analysis in Nanograms of mRNA Hong Xie, Yuan Hong Yu, Fang Xie, Yuan Zhi Lao, and Zhiqiang Gao*

Institute of Bioengineering and Nanotechnology, 51 Science Park Road, Singapore 117586, Republic of Singapore

An ultrasensitive nucleic acid biosensor for direct detection of genes in mRNA extracted from animal tissues is described. It is based on amperometric detection of a target gene by forming an mRNA/redox polymer bilayer on a gold electrode. The mRNA was directly labeled with cisplatin-biotin conjugates through coordinative bonds with purine bases in the mRNA molecules. A subsequent binding of glucose oxidase-avidin conjugates to the labeled mRNA and the introduction of a poly(vinylimidazole-co-acrylamide) partially imidazole-complexed with [Os(bpy)2(im)] (bpy ) 2,2′-bipyridine, im ) imidazole) redox polymer overcoating to the electrode allowed for electrochemical detection of the oxidation current of glucose in solution. Depending on individual genes, detection limits of subfemtograms were achieved. As compared to a sandwich-type assay, the sensitivity was improved by as much as 25-fold through the incorporation of multiple enzyme labels to the mRNA molecules. Less than 2-fold gene expression difference was unambiguously differentiated in as little as 5.0 ng of mRNA. With the greatly improved sensitivity, at least 1000-fold more sensitive than fluorescence-based techniques, the amount of mRNA needed in the assay was cut down from microgram to nanogram levels. Northern blotting,1 ribonuclease protection,2 and reverse transcription-polymerase chain reaction (RT-PCR)3,4 are the most commonly used methods in gene expression analysis. The first two methods can detect single genes at 106-107 copy levels in 10-100 µg of mRNA. If one has to quantify a number of genes in limited amounts of sample or has a need to separate only certain types of cells for analysis, for example, gene expression analysis of hematopoietic stem cells, northern blotting and ribonuclease protection techniques are not feasible. RT-PCR can theoretically amplify a single nucleic acid molecule millions of times and thus is very useful for very small sample sizes. However, RT-PCR amplification tends to introduce contamination and prolong assay * Corresponding author. Tel: +65-6874-9386. Fax: +65-6874-9341. E-mail: [email protected]. (1) Watson, J.; Gilman, M.; Witkowski, J.; Zoller, M. Recombinant DNA, 2nd ed.; W. H. Freeman and Co.: New York, 1992. (2) Chan, S. D. H.; Dill, K.; Blomdahl, J.; Wada, H. G. Anal. Biochem. 1996, 242, 214-220. (3) Cottrez, F.; Auriault, C.; Capron, A.; Groux, H. Nucleic Acids Res. 1994, 22, 2712-2713. (4) Totze, G.; Sachinidis, A.; Vettre, H.; Ko, Y. Mol. Cell. Probes 1996, 10, 427433. 10.1021/ac049839d CCC: $27.50 Published on Web 06/09/2004

© 2004 American Chemical Society

time. Very frequently, genes in a starting mRNA mixture may not be represented at the same level in the final RT-PCR products due to selective and nonlinear target amplification.5 Incomplete denaturation of mRNA’s secondary structure during the cDNA synthesis step can also halt the polymerase, resulting in shorter cDNA copies of target genes. These limitations affect the precision and quality of the resulting data and often provide distorted information of gene expression. Multiple replicates can help to gain confidence in such experiments, but it is not applicable to small or rare samples. Recent advances in developing bioelectronic DNA analysis open up new opportunities for molecular diagnostics and have attracted substantial research efforts.6-10 Amplified electronic transduction of nucleic acid recognition events has also been addressed recently.11-14 The inherent miniaturization of electrochemical biosensors and their compatibility with advanced semiconductor technologies promise to provide a simple, accurate, and inexpensive platform for nucleic acid assays. Despite the enormous progress made in electrochemical nucleic acid biosensors in the past 5 years, to be one step closer to the market, several important hurdles need to be overcome. The first is to test the biosensors on genomic nucleic acid from real-world samples.15 So far, most electrochemical biosensors started with relatively short synthetic oligonucleotides or with a round of PCR amplification. Another challenge will be to multiplex the electrochemical biosensors and their fabrication into useful sensor arrays. Typically, arrays of 30-100 will be needed for diagnostic purposes.15 For example, breast cancer screening requires testing for 20-30 cancer susceptibility genes plus positive and negative controls. Of the many proposed electrochemical biocatalytic schemes, only a few attempts have been made to analyze genes at mRNA (5) Baugh, L. R.; Hill, A. A.; Brown, E. L.; Hunter, C. P. Nucleic Acids Res. 2001, 29, e29. (6) Boon, E. M.; Ceres, D. M.; Drummond, T. G.; Hill, M. G.; Barton, J. K. Nat. Biotechnol. 2000, 18, 1096-1100. (7) Rodriguez, M.; Bard, A. J. Anal. Chem. 1990, 62, 1658-1662. (8) Kelley, S. O.; Barton, J. K.; Jackson, N. M.; Hill, M. G. Biconjugate Chem. 1997, 8, 31-37. (9) Kelly, S. O.; Boon, E. M.; Barton, J. K.; Jackson, N. M.; Hill, M. G. Nucleic Acids Res. 1999, 27, 4830-4837. (10) Wang, J. Nucleic Acids Res. 2000, 28, 3011-3016. (11) Caruana, D. J.; Heller, A. J. Am. Chem. Soc. 1999, 121, 769-774. (12) Patolsky, F.; Lichtenstein, A.; Kotler, M.; Willner, I. Angew. Chem., Int. Ed. 2001, 40, 2261-2265. (13) Patolsky, F.; Lichtenstein, A.; Willner, I. J. Am. Chem. Soc. 2000, 122, 418419. (14) Zhang, Y.; Kim, H. H.; Heller, A. Anal. Chem. 2003, 75, 3267-3269. (15) Drummond, T. G.; Hill, M. G.; Barton, J. K. Nat. Biotechnol. 2003, 21, 11921196.

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004 4023

Figure 1. Schematic illustration of mRNA assay using an mRNA/redox polymer bilayer model.

levels.10,15 Here we present a novel approach that allows direct detection of specific genes in mRNA extracted from animal tissues (Figure 1). Prior to the test, mRNA molecules are directly labeled with cisplatin-biotin conjugates (C-B).16,17 As shown in Figure 1, thiolated oligonucleotide capture probes (CP) and thiol molecules are immobilized on a gold electrode through self-assembly. The electrode is then exposed to the labeled mRNA in solution. Following hybridization to its complementary C-B-labeled target gene in the mRNA, glucose oxidase-avidin D (GO-A) labels are introduced to the system via avidin-biotin interaction. A redox polymer overcoating is brought to the electrode through layerby-layer electrostatic self-assembly. The redox polymer layer acts as a mediating layer for the enzymatic reaction. It electrochemically activates the enzyme labels attached to the target gene.18 In the presence of glucose, the current generated from enzymatic oxidation of glucose is detected amperometrically and it correlates directly to the target gene concentration in the sample solution. A cancer susceptibility gene, tumor protein p53 (TP53, 1182 bp), and a housekeeping gene, glyceraldehyde-3-phosphate dehydrogenase (GAPDH, 1008 bp), were selected as our target genes. The detection limit was at femtomolar levels. In practice, this sensitivity of the biosensor meets the requirement for direct detections of genes in real-world samples without a PCR step. EXPERIMENTAL SECTION Materials. Unless otherwise stated, reagents were obtained from Sigma-Aldrich (St, Louis, MO) and used without further purification. GO-As were purchased from Vector Laboratories (San Diego, CA). C-Bs (Biotin-Chem-Link) were obtained from Roche Diagnostics (Mannheim, Germany). [Osmium(2,2′-bipyridine)2Cl2]Cl ([Os(bpy)2]), [osmium(4,4′-dimethyl-2,2′-bipyridine)2Cl2]Cl, ([Os(dmbpy)2]), and [osmium(4,4′-diamino-2,2′-bipyridine)2Cl2]Cl (16) van Gijlswijk, R. P. M.; Talman, E. G.; Peekel, I.; Bloem, J.; van Velzen, M. A.; Heetebrij, R. J.; Tanke, H. J. Clin. Chem. 2002, 48, 1352-1359. (17) Hoevel, T.; Holz, H.; Kubbies, M. Biotechniques 1999, 27, 1064-1067. (18) Xie, H.; Zhang, C.; Gao, Z. Anal. Chem. 2004, 76, 1611-1617.

4024

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

([Os(dabpy)2]) were synthesized from K2OsCl6 followed the proposed procedure by Lay.19 [Osmium(2,2′-bipyridine)2(1-methylimidazole)Cl]Cl2 ([Os(bpy)2(Mim)]) and [osmium(2,2′-bipyridine)2(imidazole)Cl]Cl2 ([(Os(bpy)2(im)]) complexes were prepared from [Os(bpy)2] according to literature procedures.20 The redox polymers used in this study were poly(vinylimidazole-co-acrylamide) partially imidazole-complexed with [Os(bpy)2] (I), [Os(dmbpy)2] (II), [(Os(bpy)2(im)] (III), [Os(dabpy)2] (IV), and [Os(bpy)2(Mim)] (V). Synthesis of these redox polymers was described elsewhere.21,22 A phosphate-buffered saline (PBS, pH 7.4), consisting of 0.15 M NaCl and 20 mM phosphate buffer, was used for washing and electrochemical measurements. To minimize the effect of RNases on the stability of mRNA, all solutions were treated with diethyl pyrocarbonate and surfaces were decontaminated with RNaseZap (Ambion, TX). All primers used for RT-PCR were custom-made by 1st BASE (Singapore). The primer sequences were as follows: GAPDH sense, 5′-ATGGTGAAGGTCGGTGTCAA-3′, GAPDH antisense, 5′TTACTCCTTGGAGGCCATGT-3′, TP53 sense, 5′-ATGGAGGATTCA-CAGTCGGA-3′, and TP53 antisense, 5′-TCAGTCTGAGTCAGGCCC-3′. All other oligonucleotides were custom-made by Alpha-DNA (Montreal, Canada). Oligonucleotide CPs used in this work for detecting TP53 and GAPDH genes are listed in Table 1. Messenger RNA Extraction and Labeling. Messenger RNA in rat liver tissues was extracted with a Dynabeads mRNA DIRECT Kit (Dynal ASA, Oslo, Norway) according to the manufacturer’s recommended protocol. The mRNA was directly labeled using C-B, according to the recommended procedure by Roche Diagnostics. As shown in Chart 1, C-B consists of a biotin moiety, which is bound by an aliphatic linker to a cisplatin complex. One binding site of cisplatin is covalently bound to the linker/biotin molecule, and the other site is a cleavable nitrate (19) Lay, P. A.; Sargeson, A. M.; Taube, H. Inorg. Synth. 1986, 24, 291-296. (20) Sullivan, B. P.; Salmon D.; Meyer, T. J. Inorg. Chem. 1978, 17, 3334-3341. (21) Gao, Z.; Binyamin, G.; Kim, H. H.; Barton, S. C.; Zhang, Y.; Heller, A. Angew. Chem., Int. Ed. 2002, 41, 810-813. (22) Campbell, C. N.; Gal, D.; Cristler, N.; Banditrat, C.; Heller, A. Anal. Chem. 2002, 74, 158-162.

Table 1. Oligonucleotide Capture Probes Used in the Study

capture probes

capture probes

Detection of TP53 Gene 5′-HS-(CH3)6-(A)12-ATGGAGGATTCACAGTCGGA-3′ 5′-HS-(CH3)6-(A)12-TCAGTCTGAGTCAGGCCCCA-3′ Detection of GAPDH Gene 5′-HS-(CH3)6-(A)12-TTACTCCTTGGAGGCCATGTAGG-3′ 5′-HS-(CH3)6-(A)12-ATGGTGAAGGTCGGTGTCAACGG-3′

control experiment capture probe

Chart 1. Structure of Cisplatin-Biotin Conjugate (Biotin-Chem-Link)

ligand. Incubation in an aqueous solution with nucleic acid templates (DNA or RNA) cleaves the nitrate, and a new complex is formed between cisplatin and the N7 position of G and A bases. The resulting coordinative compound is stable and resistant to nucleic acid denaturation. Typically, 1.0 µg of mRNA was used in each of the labeling reactions. In our experiments, a 1.0-µL aliquot of C-B was incubated at 85 °C for 30 min with 1.0 µg of mRNA at a final volume of 20 µL, and the reaction was terminated by the addition of a 5.0-µL aliquot of stop solution (Roche Diagnostics). The labeled mRNA was stored at -20 °C. CP Immobilization. Prior to CP immobilization, a gold electrode was thoroughly polished with 0.050-µm alumina slurry and sonicated in water for 10 min. It was then exposed to oxygen plasma (2.5 mA, 0.10 mbar for 5-10 min) and immersed in 50 mL of absolute ethanol for 20 min to reduce the oxide layer. A CP monolayer was formed by immersing the cleaned gold electrode in a PBS solution containing 100 µg/mL CP for 16-24 h. After adsorption, the electrode was copiously rinsed with PBS and soaked in vigorously stirred PBS for 20 min, rinsed again, and blown dry with a stream of air. The surface density of CP, assessed electrochemically using cationic redox probes according to the procedure proposed by Steel,23 was found to be in the range of (1.03-1.25) × 10-11 mol/cm2. To minimize non-nucleic acidrelated C-B and enzyme uptake and improve the quality and stability of the CP monolayer, the CP-coated electrode was immersed into an ethanolic solution of 2.0 mg/mL 1-mercaptododecane (MD) for 4-6 h. Unreacted MD molecules were rinsed off by immersing in 50 mL of ethanol under vigorous stirring for 10 min, followed by thorough rinsing with ethanol and water. The electrode was ready after air-drying. Hybridization and Enzyme Labeling. TP53 and GAPDH in the mRNA extracted from rat liver tissues were selected in this study. Purified cDNAs transcribed from the mRNAs of the corresponding genes were used as nucleic acid standards in a calibration study. A 10 mM Tris-HCl/1.0 mM EDTA/0.10 M NaCl buffer solution (TE) was used as hybridization buffer. Hybridization was carried out with 2.5-µL droplets in a 53 °C water bath for 30 min. Nucleic acid samples were denatured at 95 (cDNA) and (23) Steel, A. B.; Herne, T. T.; Tarlov, M. J. Anal. Chem. 1998, 70, 4670-4677.

5′-HS-(CH3)6-(A)12-CCTCTCGCGAGTCAACAGAAACG-3′

70 °C (mRNA) for 10 min and cooled in an ice bath before being added to the biosensor. After hybridization, the biosensor was exposed to a 2.5-µL aliquot of 5.0 mg/mL GO-A at room temperature for 30 min and soaked for 5 min in a stirred PBS solution. To ensure the maximal loading of the redox polymer, the biosensor was then exposed for at least 10 min to a 2.5-µL aliquot of 5.0 mg/mL redox polymer solution and thoroughly rinsed with PBS afterward. Microgravimetric Quartz Crystal Microbalance (QCM) Measurement. Quartz crystal resonators employed in QCM experiments were 13.6-mm-diameter 10-MHz AT-cut type with 5.11-mm-diameter gold electrodes on both sides, purchased from Fortiming Corp. (Ashland, MA). A quartz crystal microbalance (USI System) was used to monitor hybridization and GO-A loading processes. The Sauerbrey equation is used for calculating mass change at the crystal:24

∆f ) 2f02(µqFq)-1/2∆m/A

(1)

where ∆f is the measured frequency shift (in hertz) due to mass change, f0 the resonant frequency of the fundamental mode of the crystal, µq the shear modulus of quartz (2.95 × 1011 dyn/ cm2), Fq the density of the resonator (2.65 g/cm3), ∆m the mass change of the crystal, and A the electrode surface area. For the 10-MHz resonator used in this study, eq 1 predicts that a frequency change of 1.0 Hz corresponds to a mass change of 0.90 ng on the electrode. Hybridization and GO-A labeling were conducted in a QCM cell (CH Instruments Inc. Austin, TX) in a clean environment. Stable frequency responses were taken after washing off nonreacted materials and drying in a stream of air. To minimize variation from electrode to electrode, the same electrode was always used for both hybridization and GO-A labeling studies. The frequency responses were stable within (1.0 Hz in air over a period of 3-4 h. Electrochemical Measurement. Electrochemical measurements were carried out in a Faraday cage with a low-noise CH Instruments model 660A electrochemical workstation equipped with a low-current module in conjunction with a Pentium computer. Cyclic voltammetry was conducted in both PBS buffer and PBS buffer containing 40 mM glucose. Electrochemical characterization was carried out with a gold electrode (5.0 mm2). A nonleak Ag/AgCl electrode (Cypress Systems, Lawrence, KS) was used as the reference electrode and a platinum wire as the counter electrode. Amperometric measurements were carried out at 0.36 V. All potentials reported in this work were referred to the Ag/AgCl reference electrode. Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4025

Figure 2. Agarose gel electrophoretic results of unlabeled (lanes 2 and 4) and labeled (lanes 3 and 5) TP53 and GAPDH cNDAs. Gel electrophoresis was run at 10 V/cm in 90 mM Tris-90 mM boric acid-2.0 mM EDTA buffer on a 2.5% agarose gel.

RESULTS AND DISCUSSION Stability of C-B Labeled mRNA. Instead of labeling nucleic acids using biotin-tagged nucleotides during a RT-PCR process, incubation of nucleic acid templates with C-B also leads to satisfactory labeling. As illustrated in lanes 2 and 4 in Figure 2, full-length TP53 and GAPDH cDNAs were transcripted form the extracted mRNA, indicating little degradation of the genes during the mRNA extraction process. Furthermore, the cDNA retained their integrity after reaction with C-B (lanes 3 and 5 in Figure 2). Compared to nonlabeled cDNAs, the lower mobility shifts of the reaction products suggest successful incorporation of C-B into the cDNA chains. Similarly, multiple biotin moieties were also incorporated into mRNA molecules upon incubation of the mRNA with C-B at 85 °C for 30 min. Quantitative analysis showed that 8-10% of the bases in mRNA molecules were successfully labeled.17,25 Later experiments showed that this labeling efficiency is sufficient for subsequent GO-A loading. Hybridization and Feasibility Study of Target Gene Detection. In a first hybridization test, full-length TP53 gene in the mRNA was our target gene. Prior to hybridization, the mRNA mixture was denatured at 70 °C for 10 min. Oligonucleotides with sequences complementary to TP53 gene were immobilized on the biosensor surface and served as CP. Upon hybridization at 53 °C for 30 min, TP53 mRNA from the mixture was selectively bound to its complementary CP and became fixed on the biosensor surface. Thorough rinsing with the hybridization buffer washed off all of the nonhybridization-related mRNA. A typical cyclic voltammogram of the biosensor after applying GO-A and the redox polymer overcoating is shown in curve b in Figure 3. For comparison, a voltammogram of a biosensor with noncomplementary CP is shown in curve a in Figure 3. As seen in Figure 3, considerably higher peak currents were observed for both oxidation and reduction processes at the complementary CPcoated biosensor treated by TP53 mRNA, indicating that a larger amount of redox polymer is brought to the biosensor surface, most probably due to the captured long mRNA molecules, which bring (24) Sauerbrey, G. Z. Phys. 1959, 155, 206-212. (25) Roche Molecular Biochemicals. Technical Notes: Catalog No. 1812149, 1999.

4026 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

Figure 3. Cyclic voltammograms of biosensors at 25 mV/s after hybridization with TP53 in 500 ng of C-B-labeled mRNA, incubation with GO-A and application of III overcoating. (a) Control sensor (noncomplementary CP); (b) TP53 sensor in PBS; and (c) after adding 40 mM glucose to the PBS. Conditions are detailed in the Experimental Section.

many more negative charges to the biosensor surface and create a three-dimensional microstructure on the biosensor. Treating the biosensor in the redox polymer solution results in the formation of a mixed three-dimensional mRNA/redox polymer microbilayer. Integration of oxidation or reduction current peak at slow scan rate e10 mV/s yields a redox polymer loading in terms of electroactive Os2+/Os3+ sites. Apparently, the total amount of redox polymer, (5.7-8.5) × 10-10 mol/cm2, is much higher than that of a redox polymer monolayer, ∼10-11 mol/cm2,21 suggesting that the redox polymer chains are densely “grafted” on the biosensor in the three-dimensional network configuration. It was previously shown that the minimal amount of the redox polymer needed for bioelectrocalalytic detection of nucleic acid is ∼5.0 × 10-10 mol/cm2. Obviously, the redox polymer present in the bilayer is more than sufficient. This bilayer had very rapid electronexchange processes: At scan rates up to 500 mV/s, the separation of the current peaks of the voltammetric electroreduction and electrooxidation waves was generally less than 25 mV for the bilayer containing GO-A labels. Extensive washing and potential cycling produced no noticeable changes, implying that the redox polymer is robustly bound at the biosensor surface through the formation of the electrostatic bilayer with substantial degree of overlapping of the two layers. Curve c in Figure 3 is the voltammogram in PBS containing 40 mM glucose. Obvious catalytic current was observed due to the presence of glucose oxidase in the bilayer. The noncomplementary CP-coated electrode failed to capture any TP53 mRNA and thereby no enzyme labels were able to bind to the biosensor surface. No noticeable catalytic current in voltammetry was detected. To have a better understanding of the labeled nucleic acid hybridization efficiency and GO-A loading level, a series of QCM measurements were carried out on C-B-labeled GAPDH cDNA after hybridization and after GO-A loading. The results are summarized in Table 2. As shown in Table 2, ∼16 fmol of labeled GAPDH was hybridized. This number represents ∼0.70% of the surface-bound CP actually hybridized, which is much lower than those of short oligonucleotides (20-50-mers) reported in the literature.26-28 It is not surprising that the hybridization efficiency decreases drastically with increasing length of the analyzed gene.

Table 2. QCM Data of CP-Coated Quartz Crystal Resonators after Hybridization to 3.0 pmol of C-B-Labeled GAPDH CDNA and after GO-A Labeling in 5.0 mg/mL GO-A Solution C-B-labeled GAPDH cDNA hybridization

GO-A labeling GAPDHa

CP-coated resonator

∆f (Hz)

∆m (ng)

1 2 3 4

7.2 5.9 6.3 8.5

6.5 5.3 5.7 7.7

hybridized (×1015 mol) 16 13 14 19

∆f (Hz)

∆m (ng)

GO-A loading (×1013 mol)

GO-A/base ratio

81 71 78 86

73 64 72 77

3.2 2.8 3.2 3.4

1/50 1/45 1/43 1/56

a Assuming midrange (9.0%) C-B labeling efficiency, the molecular weight of C-B-labeled GAPDH cDNA is 407.8K (330K + 1.0K × 9.0% × 864) ) 407.8K), and that of GO-A is 228K (160K + 68K).

Obviously, it is not possible to attach GO-A to every biotin moiety on the labeled gene due to steric hindrance. The QCM experiments showed that one GO-A was attached to the labeled gene per 43-56 bases, or ∼20% of biotin moieties on average. The surface coverage of GO-A loaded onto the labeled gene was found to be ∼1.2 × 10-12 mol/cm2 (geometrical area 0.20 cm2, roughness factor 1.3). This value corresponds to ∼72% of a random, densely packed glucose oxidase monolayer on the biosensor surface.29 In reality, the GO-A labels are more likely distributed throughout the three-dimensional mRNA/redox polymer microbilayer. To attain the highest sensitivity and best reproducibility, both C-B and GO-A labeling processes were optimized to ensure maximal loadings of C-B and GO-A labels. Redox polymers consisting of [Os(bpy)2] and different polymer backbones, namely, poly(vinylimidazole-co-acrylamide), poly(vinylimidazole-co-acrylic acid) and poly(vinylimidazole-co-acrylamido2-methyl-1-propanesulfonic acid), were first tested for their ability to form stable bilayers. It was observed that among the three redox polymers tested, poly(vinylimidazole-co-acrylamide) is the best in terms of stability of the bilayer and the amount of redox polymer being immobilized on the biosensor surface. Apparently, due to partial protonation of acrylamide moieties at pH 7.4, some stability reinforcement is brought to the bilayer, which brings the osmium redox centers in the proximity of GO-A. Therefore, poly(vinylimidazole-co-acrylamide) was used throughout. Various osmium-bipyridine complex-containing redox polymers were then tested for their mediating capabilities in the bilayer configuration. The results are listed in Table 3. As can be seen in Table 3, structural differences in the redox polymer overcoatings strongly affected their mediating capability. The catalytic currents of the redox polymers in which the osmium redox centers bearing two positive charges were higher than those with a single positive charge, implying that electrostatic interaction of GO-A and the osmium redox center is one of the most important factors in a successful mediation as the isoelectrical point of glucose oxidase is ∼4.0.30 It was found that the catalytic current generated with the redox polymer containing 1-substituted imidazole ligand was smaller than that of the redox polymer containing nonsubstituted imidazole, suggesting that hydrogen bonding between imino (26) Dequaire, M.; Heller, A. Anal. Chem. 2002, 74, 4370-4377. (27) Caruso, F.; Rodda, E.; Furlong, D. N.; Niikura, K.; Okahata, Y. Anal. Chem. 1997, 69, 2043-2049. (28) Satjapipat, M.; Sanedrin, R.; Zhou, F. Langmuir 2001, 17, 7637-7644. (29) Xiao, Y.; Patolsky, F.; Katz, E.; Hainfeld, J. F.; Willner, I. Science 2003, 299, 1877-1881.

Table 3. Electrochemical Characteristics of Biosensors with Different Redox Polymer Overcoatingsa redox polymer

Em (mV)

∆Ep (mV)

polymer loading (×1010 mol/cm2)

icat (nA)

I II III IV V

110 21 282 -112 291

18 21 20 27 18

6.3 5.7 7.7 6.5 8.5

136 109 242 83 184

a E ) 1/2(E m pa + Epc), ∆Ep ) Epa - Epc; conditions are as for Figure 3.

groups and carboxyl groups on glucose oxidase in the vicinity of the FAD redox centers further facilitates electron exchange. Under identical experimental conditions, the voltammetric catalytic current generated at the biosensor employing a III overcoating was the highest and was measured to be ∼3-fold higher than that observed at the biosensor with a IV overcoating. Therefore, to attain the highest current sensitivity, III was used in subsequent experiments. In another test, after hybridization with TP53 in 30 ng of mRNA, the completed biosensor was immersed in PBS. The oxidation current in amperometry increased by 5.3 nA at 0.36 V (vs Ag/AgCl) upon addition of 40 mM glucose to the PBS solution (Figure 4, curve b). In a control experiment where noncomplementary CPs were immobilized on the biosensor surface, only a 0.19 nA increment was observed (Figure 4, curve a). The amperometric results agreed well with the cyclic voltammograms obtained earlier and confirmed again that TP53 was successfully detected from the mRNA mixture with high specificity, considering that there are tens of thousands of genes in the mRNA pool. In a similar way, as shown in Figure 4, curve c, GAPDH gene was also successfully detected in the mRNA by changing CPs immobilized on the biosensor surface. Since the two genes are similar in length, similar sensitivity is anticipated. A much higher copy number of GAPDH, ∼3-fold of TP53, was found in the mRNA judging from the signals of TP53 and GAPDH. It was found that the amperometric response sensitivity increases with increasing glucose concentration in the range of 0-40 mM and starts to level off beyond 40 mM. Therefore, 40 mM glucose was used in the amperometric detection of the target genes. (30) Trudeau, F.; Daigle, F.; Leech, D. Anal. Chem. 1997, 69, 882-886.

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4027

Figure 4. Amperometric responses of biosensors after TP53 and GAPDH in 30 ng of C-B-labeled mRNA hybridized with (a) noncomplementary and (b (TP53), c (GAPDH)) complementary capture probes, incubation with GO-A and application of III overcoatings. Poised potential: 0.36 V, 40 mM glucose. Conditions are detailed in the Experimental Section; smoothing was applied after each run. The arrow indicates the time at which glucose was injected into PBS.

Detection of Target Genes with One-Base Mismatch. The specificity of the biosensor for detection of target genes was further evaluated in 5.0 ng of mRNA by replacing fully complementary CP with CP in which one of the bases was mismatched. As shown in Figure 5, the current increments for the perfectly matched sequences were 0.47 and 0.95 nA for TP53 and GAPDH, whereas, for one-base-mismatch CP, the increments dropped to 0.25 and 0.34 nA, respectively, slightly higher than 0.19 nA observed for a totally noncomplementary sequence (control biosensor), readily allowing discrimination between the perfectly matched and mismatched genes. Calibration Curves for Target Genes. Since most of the cancers are in one way or another associated with cancer susceptibility genes, techniques that could offer sensitive detection and accurate quantification of these genes will help to facilitate earlier diagnosis and prognosis.31 In this study, C-B-labeled cDNAs were used as standards and diluted to different concentrations with TE buffer before use. Analyte solutions with different concentrations of cDNA, ranging from 0.10 fM to 2.0 pM, were tested. For control experiments, noncomplementary CPs were used in the biosensor preparation. As depicted in Figure 6, the current increased linearly with cDNA concentration within this range. The dynamic ranges for TP53 and GAPDH were 1.5-400 and 1.0-400 fM with detection limits of 0.50 fM, corresponding to 0.47 and 0.40 fg of TP53 and GAPDH, respectively. Taking the sample volume into consideration, as few as 800 copies of TP53 and GAPDH cRNAs were successfully detected using the proposed method. For the detection of genes in mRNA, it was found that the lowest amount of mRNA needed for a successful detection is ∼1.5 ng. Considering that there are more than 30 000 genes in this mRNA pool, the actual detectable limit for each specific gene is at subfemtograms on average, which is in good agreement with the calibration curve study (see above). To the best of our knowledge, this is the lowest reported amount of genomic nucleic acid detected electrochemically. The relative errors associated with the mRNA assays on individual genes were generally less than 20% in the concentration range of 2.0-300 fM. Compared to the previous results of sandwich-type assays,18,22,26 the sensitivity 4028

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

of the genomic nucleic acid assay was greatly improved by adopting the multiple enzyme labeling scheme, and the result was comparable to that obtained with short synthetic oligonucleotides of 20-50 bases in length. In the sandwich-type assays reported earlier,18,22,26 the ratio of enzyme label and target nucleic acid molecule was fixed at unit. The amount of CP immobilized on the sensor surface and hybridization efficiency determine the amount of target nucleic acid bound to the surface and thereby the amount of enzyme labels despite the size of the genes. Moreover, much lower overall hybridization efficiency is expected owing to the dual-hybridization strategy employed in those systems. It is, therefore, awkward in analyzing genomic nucleic acids since most of them are several hundred to several thousand bases in length. In fact, much lower current sensitivity was observed for longer nucleic acid molecules.18 On the other hand, in our proposed model, multiple C-B labels on a single nucleic acid chain greatly increased the enzyme label loading, amperometric responses from enzymatic reaction were improved accordingly, and hence the sensitivity and detection limit of the nucleic acid biosensor were substantially enhanced when working with real-world samples. For example, for a 1000-base-long nucleic acid, if there is one enzyme label per 50 bases, the overall signal could be increased by 20-fold. It was found that by labeling the nucleic acid molecules with multiple enzyme molecules using C-B the sensitivity was increased by 25-fold compared to the sandwichtype assay.18 The sensitivity obtained in this work is comparable to that observed with the short synthetic oligonucleotide (50-mer) of the sandwich-type approaches,18,22,26 indicating that the enzyme/ base ratio has very small variation for both short synthetic oligonucleotides and genomic nucleic acid samples, which, in turn, generated analytical signals of similar sensitivities. The enzyme label/base ratio was estimated to be in the range of 1/43-56 depending on the length of the nucleic acid molecules. Theoretically, if this ratio keeps unchanged for all genes, the same current sensitivity per base should be obtained for all genes. At the same molar concentration, the sensitivity should be roughly proportional to the number of bases in the gene. Indeed, higher sensitivity, ∼13%, was obtained with TP53 due to 17% more bases in it. Expression Analysis of TP53 and GAPDH in mRNA Extracted from Rat Liver Tissues. With the much improved sensitivity and precision, the nucleic acid biosensor allowed us to identify genes that differ less than 2-fold in expression between two conditions. It is very difficult to detect 2-fold difference in gene expression using current technology. One can only reliably discriminate gene expression differences of more than 3-fold.32 In many cases, the expressions of many of the most interesting genes may only differ a little between different conditions. To determine the ability in detecting minute changes in gene expression of real-world samples, we performed multiple assays on both TP53 and GAPDH with 5.0 ng of mRNA. Expression levels of the genes were mimicked by adding various amounts of the mRNA to the test solution. Figure 7 clearly shows that the proposed assay can unambiguously differentiate less than 2-fold gene expression difference. As can be seen in Figure 7, amperometric responses of TP53 and GAPDH were completely resolved from those of the original test solution when 80% of the original quantity of mRNA (31) Martin, K. J.; Sager, R. Adv. Exp. Med. Biol. 1998, 451, 1-7. (32) Gullans, S. R. Nat. Genet. 2000, 26, 4-5.

Figure 5. Amperometric responses of biosensors after hybridization in 5.0 ng of C-B-labeled mRNA with capture probes (1) complementary and (2) one-base mismatch to TP53 (A) and GAPDH (B). Conditions as for Figure 4. The arrows indicate the time at which glucose was injected into PBS.

Figure 6. Amperometric responses of biosensors to C-B-labeled TP53 and GAPDH cDNAs at different concentrations. Conditions as for Figure 4.

was added. This allows a greater accuracy in the identification of differentially expressed genes and cuts down the need for running too many replicates. With the greatly improved sensitivity, at least 1000-fold more sensitive than those of fluorescence-based assays,2-5 the proposed method also significantly cuts down the amounts of mRNA needed from micrograms to nanograms. CONCLUSIONS The electrochemical nucleic acid biosensor described here is rapid, ultrasensitive, and nonradioactive and is able to directly detect genes in mRNA without involving PCR amplification. By employing the cisplatin-biotin conjugates, mRNA was directly labeled with biotin moieties in a one-step nonenzymatic reaction. Specific genes were detected amperometrically with high sensitivity and specificity. By labeling mRNA sample with multiple enzyme molecules, the sensitivity was increased by 25-fold compared to the sandwich-type assay. Full-length mRNAs of TP53 and GAPDH were selectively detected at femtomolar levels. Less than 2-fold gene expression difference was successfully differentiated. The lowest detectable amount of a specific gene was found to be ∼800 copies in as little as 1.5 ng of mRNA. Developments in progress

Figure 7. Dependence of amperometric responses (6 duplicates) on (b) TP53 and (4) GAPDH expression. The expression levels were mimicked by spiking various amounts of the mRNA to a test solution containing 5.0 ng of C-B-labeled mRNA. Conditions as for Figure 4; i0 and i denote signals of the original test solution and those with different C-B-labeled mRNA spikes, respectively.

include extension of the assay described here to include the most frequent mutations in cancer susceptibility genes. Such an assay will provide a rapid approach for the identification of high-risk mutations in these genes. Integrating 50-100 individual sensors into an array and incorporating the array into a fully automated microelectromechanical system, from tissue digestion and sample preparation to nucleic acid isolation and quantification, will provide faster, less expensive, and simpler solutions for molecular diagnosis, particularly for early cancer diagnosis, point-of-care, and field uses. Such an integrated system is being undertaken in our laboratory. ACKNOWLEDGMENT The authors acknowledge financial support from IBN/A*STAR. Received for review January 27, 2004. Accepted April 15, 2004. AC049839D Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4029