Anal. Chem. 2005, 77, 2393-2399
Double-Wavelength Technique for Surface Plasmon Resonance Measurements: Basic Concept and Applications for Single Sensors and Two-Dimensional Sensor Arrays Alexander Zybin,*,† Christian Grunwald,‡ Vladimir M. Mirsky,§ Ju 1 rgen Kuhlmann,‡ Otto S. Wolfbeis,§ and Kay Niemax†
ISAS - Institute for Analytical Sciences, Bunsen-Kirchhoff-Str. 11, D-44139 Dortmund, Germany, Max-Planck-Institute for Molecular Physiology, Otto-Hahn-Str. 11, D-44202 Dortmund, Germany, and Institute of Analytical Chemistry, Chemo- and Biosensors, University of Regensburg, D-93040 Regensburg, Germany
A new technique for on-line monitoring of analyte binding to sensor surfaces by surface plasmon resonance (SPR) detection is described. It is based on differential measurements using two wavelengths provided by two diode lasers. The technique is as simple and robust as the conventional SPR detection measuring the reflected radiation at fixed incidence angle, but it has the advantage of being nonsensitive to variations of the resonance width and providing essentially higher signal/noise ratios. The paper presents the first four channel prototype system for parallel 2D-monitoring at four different spots. One channel is always used as a reference to compensate temperature fluctuations and nonspecific adsorptions. Calibration with sucrose solutions revealed an absolute sensitivity of ∆n ∼ 5 × 10-6. The new technique is tested with a biotin-streptavidin binding and with hybridization/denaturation of DNA. Biotin binding to a streptavidin monolayer is detected with a signal/noise ratio of about 5, which demonstrates the high potential of the new technique for applications in drug discovery. Applications to gene analysis are tested with short oligonucleotides of the sequences used for genotyping human hepatitis C viruses. A selective response to complementary oligonucleotides is observed. The high reproducibility in subsequent cycles of hybridization/denaturation (by formamide or by heating) points out potential applications of the technique in medical diagnostics, food industry, genomics, and proteomics too. Many biological processes are based on selective recognition and binding. Selective ligand-receptor interactions provides a discrimination in simultaneously running biochemical processes in the presence of hundreds of different substances. A key technology to observe these processes in real time was developed about 20 years ago by Liedberg et al.1 and then commercialized * To whom correspondence should be addressed. E-mail:
[email protected]. † ISAS - Institute for Analytical Sciences. ‡ Max-Planck-Institute for Molecular Physiology. § University of Regensburg. (1) Liedberg, B.; Nylander, C.; Lundstro ¨m, I. Biosens. Bioelectron. 1995, 10, I-IX. 10.1021/ac048156v CCC: $30.25 Published on Web 03/15/2005
© 2005 American Chemical Society
by the Swedish company Pharmacia, the parent company of Biacore. The measurement technology is based on the surface plasmon resonance effect (SPR).1 Currently, also several other companies offer devices for detection of (bio)molecule interaction with corresponding receptors.2 A competing method for the measurement of such reactions is the quartz crystal microbalance technique. However, the surface plasmon resonance is more reproducible and less sensitive to temperature fluctuations and mechanical instabilties.3 SPR-based technology has become an accepted and established tool for investigation of biological processes, adsorption phenomena, characterization of binding properties of antibodies and other biological molecules, gene analysis, and drug discovery. Today, SPR is used as a transducer in chemical and biological sensors with potential applications in medicine, biotechnology, environmental control, food industry, and many other fields. During the past 20 years many improvements in the SPR technology were reported. The classical “Kretschmann configuration” based on bulk optics is the most often used arrangement in commercial and laboratory devices. This technique is in particular suitable for SPR microscopy (SPR imaging) to study simultaneously up to several thousand processes on twodimensional structures4-9 which is important for modern genomics, proteomics, and drug discovery. The improvement of detection sensitivity is one of the main goals in the development of SPR biosensing. It can be improved by several actions. The first one includes the formation of an extended matrix for immobilization of receptor molecules10 and the optimization of immobilization procedures.11-15 The second (2) Baird, C. L.; Myszka, D. C. J. Mol. Recognit. 2001, 14, 261. (3) Su, X.; Zhang, J. Sens. Actuators B 2004, 100 (3), 309. (4) Homola, J.; Yee, S. S.; Gauglitz, G. Sens. Actuators B 1999, 54, 3. (5) Ro ¨thenha¨user, B.; Knoll, W. Nature 1988, 332, 615. (6) Wegner, G. J.; Lee, H. J.; Corn, R. M. Anal. Chem. 2002, 74, 5161. (7) Nelson, B. P.; Grimsrud, T. E.; Liles, M. R.; Goodman, R. M.; Corn, R. M. Anal. Chem. 2001, 73, 1. (8) Giebel, K.-F.; Bechinger, C.; Herminghaus, S.; Riedel, M.; Leiderer, P.; Weiland, U.; Bastmeyer, M. Biophys. J. 1999, 76, 509. (9) Shumaker-Parry, J. S.; Campbell, C. T. Anal. Chem. 2004, 76, 907. (10) Liedberg, B.; Lundstro ¨m, I.; Stenberg, E. Sens. Actuators B 1993, 11, 63. (11) Hermanson, G. T. Bioconjugate Technique; Academic Press: San Diego, 1995. (12) Henke, L.; Krull, U. J. Can. J. Anal. Sci. Spectrosc. 1999, 44, 61.
Analytical Chemistry, Vol. 77, No. 8, April 15, 2005 2393
level includes an optimization of the plasmon layer, for example, the use of bimetallic Ag/Au layers combining the sharp resonance (and therefore high sensitivity) of silver layers with the perfect chemical stability of gold surfaces, as reported in ref 16 and applied in ref 17. The third level of sensitivity improvement is based on the development of new detection schemes for the surface plasmon resonance or its shift and, furthermore, the optimization of data analysis. An approach suggested in refs 18-20 exploits the phase change due to the reflection near the resonance peak. Suppression of the volume effect of refractive index variation is reported.17 The simplest phase-based approach applies a dark field technique similar to that used in microscopy. It results in a drastic increase of the contrast in SPR imaging. Another method exploits the interference of the reflected beam with the reference beam to visualize the phase behavior at different areas of the surface. The resolution limit of refractive index measurements was estimated to be 4 × 10-8 assuming that temperature fluctuations do not disturb the measurements.19 However, the most often used techniques are based on the measurement of the angle shift of the reflecting intensity at fixed wavelength. The angle dependence can be measured in a wide range of the refractive index, from the critical angle of total inner reflection up to much more than the resonance angle. A fitting of the angle dependence by Fresnel equations for multilayer systems (glass, metal layer, thin organic layer, bulk phase) provides information on thickness and optical properties of the organic layer. The angle dependence can be measured by means of mechanical rotation of the prism. It is realized, for instance, in the SPR imaging instruments offered by IBIS Technologies21 and is one of the measurement modes in the SPR spectrometer BIOSUPLAR22 where parallel monochromatic laser beams are used. Mechanical movement can be avoided by angle-dependence measurements using convergent light and photodiode arrays.23,24 This concept was successfully applied in the integrated sensor for refractive index measurements25,26 commercialized by Texas Instruments. An essential improvement of sensitivity was reached by using a bicell photodetector instead of a diode array.27 This approach, however, leads to limitation of the range of refractive (13) Albers, W. M.; Vikholm, I.; Viitala, T.; Peltonen, J. In Handbook of Surfaces and Interfaces Materials, vol. 5; Academic Press: San Diego, 2001; pp 1-131. (14) Wrobel, N.; Deininger, W.; Hegemann, P.; Mirsky, V. M. Colloid Surf. B 2003, 32, 157. (15) Wrobel, N.; Schinkinger, M.; Mirsky, V. M. Anal. Biochem. 2002, 305, 135. (16) Zynio, S. A.; Samoylov, A. V.; Surovtseva, E. R.; Mirsky, V. M.; Shirshov, Y. M. Sensors 2002, 2, 62. (17) Alieva, E. V.; Konopsky, V. N. Sens. Actuators B 2004, 99, 90. (18) Grigorenko, A. N.; Beloglasov, A. A.; Nikitin, P. I.; Kuhne, C.; Steiner, G.; Salzer, R. Opt. Commun. 2000, 174, 151. (19) Kabashin, A. V.; Nikitin, P. I. Quantum Electron. 1997, 27, 653. (20) Nikitin, P. I.; Grigorenko, A. N.; Beloglasov, A. A.; Valeiko, M. V.; Savchuk, A. I.; Savchuk, O. A.; Steiner, G.; Kuhne, C.; Huebner, A.; Salzer, R. Sens. Actuators B 2000, 85, 189. (21) www.ibis-spr.nl. (22) www.biosuplar.com. (23) www.biacore.com. (24) www.ti.com/snc/products/sensors/spreeta.htm. (25) Bartholomew, D. U.; Melendez, J. L.; Taneja, H.; Yee, S.; Yung, Ch.; Furlong, C. Sens. Actuators B 1997, 38-39, 375. (26) Elkind, J. L.; Stimpson, D. I.; Strong, A. A.; Bartholomew, D.; Melendez, J. Sens. Actuators B 1999, 54, 182. (27) Tao, N. J.; Boussad, S.; Huang, W. L.; Arechabaleta, R. A.; D’Agnese, J. Rev. Sci. Instrum. 1999, 70, 4656.
2394 Analytical Chemistry, Vol. 77, No. 8, April 15, 2005
Figure 1. Single-wavelength signal by shifting of the resonance curve.
indexes and angles and, on the other hand, can hardly be used for SPR imaging. Most laboratory SPR-imaging devices are based on illumination by a parallel monochromatic beam. The angle is slightly shifted from the resonance angle so that any shift of the resonance causes a variation of the reflected intensity (Figure 1). The intensity variations are observed by means of a photodetector matrix at many positions simultaneously and are used to characterize processes at those spots.9,28 Several methods based on modulation techniques were suggested for SPR measurements. The method based on small modulation of the incidence angle by a piezo-electrical actuator was reported in ref 29. At resonance conditions, the reflectance signal detected by a lock-in amplifier is zero. Resonance shifts are measured as a deviation from zero or through a feedback signal which changes the angle of incidence. Another modulation method applies the modulation of the wavelength of the incident light beam. It was realized by means of a tunable acousto-optical filter.30 An important feature of both approaches is that no space coordinate is used for modulation. Therefore, both techniques can be considered as promising approaches for highly sensitive SPR imaging. This paper describes a new technique also based on wavelength modulation without moving parts. However, instead of one tunable light source, two lasers with different wavelengths are applied. The potential of the technique is demonstrated by simultaneous, highly sensitive label-free detection of biomolecules binding to several types of two-dimensional arrays of receptor spots including a DNA hybridization test. The high sensitivity of the new method is confirmed by the detection of biotin binding. PRINCIPLE OF DOUBLE-WAVELENGTH SPR DETECTION The most frequently used approach to obtain SPR images exploits the reflection at fixed angles slightly shifted from the plasmon resonance angle. Being highly sensitive, this technique (28) Steiner, G. Anal. Bioanal. Chem. 2004, 379, 328. (29) Berger, C. E. H.; Greve, J. Sens. Actuators B 2000, 63, 103. (30) Jory, M. J.; Bradberry, G. W.; Cann, P. S.; Sambles, J. R. Meas. Sci. Technol. 1995, 6, 1193.
Figure 2. Double-wavelength signal by shifting of the resonance curve.
Figure 3. Experimental arrangement (1, lenses; 2, beam splitters).
is affected by possible shape changes of the resonance which can be caused by unexpected modification of the surface roughness or by light absorbance. The diffracted and/or scattered light from the neighboring area around a sensing spot can also interfere with the radiation reflected from this spot and, therefore, affect the measurements. This is especially critical for miniaturized, highcapacity arrays with small spots. The double-wavelength measurement scheme (Figure 2) does overcome these problems. It is based on surface irradiation by a parallel beam composed of two laser beams with different wavelengths. After rough tuning the incidence angle to resonance, the laser wavelengths are tuned to the opposite sides of the resonance curve and fixed to positions where the slopes are steepest and the reflectivity for both wavelengths is the same; i.e., the photodetector currents are equal. The lasers are alternatively switched on and off with a frequency of a few kHz, so that always only one of the lasers is operating at a time. If the reflectivity signals for both wavelengths are equal, the photodetector produces a direct current signal. Any shift of the SPR curve results in a difference in reflectivity and in an alternative current (ac) signal at the photodiode. The value of this ac component can be used to characterize the shift and, therefore, the adsorption processes in the layer adjoined to the metal film. EXPERIMENTAL SECTION The experimental arrangement is shown in Figure 3. Radiation from two diode lasers with wavelengths of 785 nm (laser diode type: HL-7851 from Hitachi) and 830 nm (HL-8325G, also from Hitachi) is coupled by means of a beam splitter into a single mode,
polarization maintaining glass fiber. Optical Faraday isolators in front of the laser diodes (not shown in Figure 3) are used to prevent optical feedback from the different components and to keep the wavelengths and the output of the lasers stable. The total optical arrangement provides an almost perfect overlap of the beam profiles for both wavelengths behind the fiber. The combined beam is expanded to a diameter of about 2 cm and used for illumination of the prism. An optical arrangement without using the glass fiber was checked as well. Here, the alternating laser radiation behind the beam splitter was focused on a pinhole with a diameter of the order of the calculated diffraction diameter in the focus. The light which had passed the pinhole illuminated the prism. This arrangement is less expensive but it was found that it requires precise temperature stabilization unlike the fiber setup. The currents and temperatures of the laser diodes were controlled and tuned by commercial power suppliers (Profile, Germany). A homemade modulator switched the diode laser currents alternatively with a frequency of about 2 kHz. The beam reflected from the prism was expanded by a system of cylindrical and spherical lenses and directed to a “photodetector matrix” consisting of four individual Si photodiodes (OEC S-25CL). The signals of the photodiodes were amplified by lock-in amplifiers (SR 830, Stanford Research, USA) and fed via an A/D card into a computer. To compensate possible signal drift caused by variations of the laser intensities, a part of the radiation was guided by a beam splitter to another photodiode which was used for active control of the laser intensities. The photodiode signal fed into a lock-in amplifier was used in an active feedback loop to the laser diode drivers to equalize the laser intensities and to eliminate a possible, interfering ac component. Application of such active feedback control of the laser intensity in a SPR-detection arrangement with only one laser would result in an artificial signal drift because variations of the laser current also cause changes of the laser wavelength with respect to the SPR curve and, therefore, of the reflectivity at some fixed angle. The double-wavelength scheme is compensating this effect since the feedback signal is applied in opposite polarity to both lasers, which shifts both wavelengths in opposite directions. Therefore, both reflectivity signals change in the same direction, compensating signal drifts. For typical diode laser intensity fluctuations, the stabilization scheme provides practically a complete suppression of signal drifts due to laser fluctuations. The flow cell was made from Teflon and pressed to the base of a prism. A quadratic area of 14 × 14 mm2 was homogeneously illuminated. The thickness of the flow cell was 1 mm, giving a maximum volume of 200 µL. In most measurements the whole area was divided into three independent channels, each 0.4 cm wide and 1.4 cm long. A continuous water flow with a rate of up to 1 mL/min was pumped through the cell using a peristaltic pump (Ismatec Reglo Digital MS-2/8). Prism Surface Preparation. Polished glass prisms (n ) 1.61) with a base area of 21 mm × 18 mm and a top angle of 40° were coated with 1.5 nm of thin adhesive chromium sublayer and a 47 nm gold layer. A Leybold Inficon XTC/2 metal evaporator (chrominum deposition rate ∼ 0.2 nm/s, gold deposition rate ∼ 1.5 nm/s, both at ∼10-7 mbar) was used for preparation. Although Analytical Chemistry, Vol. 77, No. 8, April 15, 2005
2395
no damage of the metal layers were observed after experiments, each metal coating was used only once. Before recoating of the prisms, all coatings were removed. Then, the prisms were rinsed with ethanol and incubated in fresh piranha solution (1/3 mixture of concentrated sulfuric acid and 25% hydrogen peroxide, v/v). Gold was removed by incubation in a solution of KI/I2 (200 mg/ mL I2 and 800 mg/mL KI in water). Chromium was removed by incubation in aqueous solution of ammonium cerium(IV) nitrate (60 mg/mL). Finally, the prisms were rinsed with water and pure ethanol and dried in a stream of high-purity nitrogen. Streptavidin-Biotin Coating. The gold layers were coated by self-assembled monolayers from thiolated compounds. The interaction surface for the biotin-streptavidin experiments was prepared by coadsorption of the biotinylated linear thiol (generously provided by A. Terfort) and mercaptoundecanol (Sigma) with molar ratio of 1:10; details are described in ref 31. The spots resistive to protein adsorption (reference spots) were formed by coating by OEG6-terminated thiol (mercaptoundecan terminated by the oligomer from 6 ethylene glycole groups, generously provided by W. Eck). The experiments with the biotin-streptavidin system were performed in the buffer containing 150 mM NaCl, 10 mM HEPES, 5 mM MgCl2, pH 7.4. The buffer was pumped through the flow cell continuously with a flow rate of 0.3 mL/min. The streptavidin concentration during the flow injection was 200 nM. Immobilization of Oligonucleotides. Single-stranded olygonucleotides were immobilized on the gold surface by a thiol-link. The borders of four spots (about 3 mm in diameter) were marked by hydrophobic ink on the prism with a gold layer. Then, about 5 µL droplets of each of the three oligonucleotides with thiol-link and (T)5-spacer (DNA-(T)5-(CH2)6-SH) at the 5′-end were deposited from a 100 µM solution in water on the corresponding surface spots and incubated for 2 h at 100% humidity. The fourth spot was uncoated and used as a reference. Former experiments with 33P labeled DNA in our laboratory had demonstrated that such deposition conditions provide a formation of almost saturated monolayer of oligonucleotides. Then, the surface was rinsed thoroughly with water and phosphate buffer and mounted to the flow cell. The same flow cell as in the biotin-streptavidin experiments was used. The sequences of DNA oligonucleotides earlier applied32 for diagnostics of human hepatitis C virus (G) and two common types of this virus (T1 and T2) were used. Their lengths were optimized to obtain similar melting temperatures. The used receptor sequences were as follows: 5′-CCAAGAAAGGACCCG-3′ (G), 5′-CTCCAGGCATTGAGC-3′ (T1), and 5′-CAACCCAACGCTACT-3′ (T2). The analytes were synthetic oligonucleotides of complementary sequence without thymin spacers. Below they will be designed as cG, cT1, and cT2, respectively. Calculated melting points of dsDNA hybridized from a receptor sequence and corresponding complementary oligonucleotide were about 39-40 °C. Hybridization experiments were performed in the same phosphate buffer (100 mM KCl, 10 mM phosphate, pH 6.5). The concentration of the complementary nucleotides in the hybridization experiments was 1 µM. Oligonucleotides were (31) Grunwald, C.; Eck, W.; Opitz, N.; Kuhlmann, J.; Wo ¨ll, C. Phys. Chem. 2004, 6, 4358. (32) Livache, T.; Fouque, B.; Roget, A.; Marchand, J.; Bidan, G. Anal. Biochem. 1998, 255, 188.
2396 Analytical Chemistry, Vol. 77, No. 8, April 15, 2005
Figure 4. Signal of a sucrose solution. The sucrose solution is flowing during the time period A and B. The overshoot at the moment of injection is due to a short flow stop during injection.
Figure 5. Calibration curve with a sucrose solution.
purchased from ThermoHybaid. The hybridization temperature was 22 °C. Milli-Q water was used for buffer preparation. RESULTS AND DISCUSSION Test and Calibration of the Experimental Arrangement. The performance of the new technique was tested with aqueous sucrose solutions of defined concentrations and by the use of tabulated values of the refractive index for these solutions. Figure 4 shows the signals of both photodiodes for 0.01% sugar (∆n )1.4 × 10-5) and the difference of these two signals. Note that the signal fluctuations from both spots are well correlated so that the difference signal shows much smaller fluctuations. This demonstrates the potential for a sensitivity increase using a reference spot in measurements with selective adsorption at the surface. The remaining signal difference of both spots should be due to a residual inhomogeneity of the intensity or to inhomogeneities in the gold layer. Therefore, before quantitative measurements it is recommended to balance the sensitivities of individual spots, for example with a sugar (or salt) solution. The calibration curve for one of the spots is shown in Figure 5. The curve is linear up to ∆n ∼ 2 × 10-3. The smallest detectable refractive index variation ∆n is 5 × 10-6 (this corresponds to a sugar concentration of ∼3 × 10-3 %). The possibility of simultaneous observation of the sensing and reference spots provides a way to compensate temperature drifts.
Figure 6. Compensation of a temperature drift in the SPR signal by the measurement of a reference spot. SPR signals at four detection spots were recorded at constant buffer flow but at increasing temperature. The signal change in three spots caused by temperature variation could be completely compensated subtracting reference signal from spot four. The temperature of the flowing buffer increased from 19 to 60 °C in 7 min.
To check this option, a buffer with a slow growing temperature was pumped through the flow cell with receptors (oligonucleotides G, T1, and T2) immobilized at three receptor spots. The fourth receptor spot was used as a reference. The buffer temperature was increased from 19 °C to 60 °C in about 7 min. The temperature dependence of the refractive index led to a strong effect of the absolute signals while only a small effect on the differential signal was observed (Figure 6). It has to be noted that the temperature of the flowing buffer was increased in this experiment. Therefore, the temperature at the prism surface was not homogeneous. Nevertheless, fluctuations of the baseline were below 10-5 reflection index units (RIU). Therefore, the new technique can be used to study surface reactions even during essential temperature changes. Biotin-Streptavidin Model System. Streptavidin binding to a biotynilated surface is an established model system to investigate protein-receptor binding. Two spots with protein-resistant coating (OEG6) were used as a reference. A preliminary normalization of signals with addition of NaCl solution was performed. Figure 7 shows all four signals. The two reference curves are almost identical in the scale shown. The difference of the two signal curves (about 6%) reflects the difference in the coupling processes in both spots, probably due to nonhomogeneities of the coating. It is not surprising that the signals were high since the molecular weight of streptavidin is about 60 kD and the receptor density at the sensing spots was optimized for maximum binding. Streptavidin molecules bound to the surface in the experiment reported above have vacant binding sites which can bind biotin from solution. Biotin molecules being relatively small (244.3 D) are commonly used for testing the sensitivity of affinity methods. After rinsing the flow cell with a buffer for 20 min, a 1 µM solution of biotin in the same buffer was pumped through the cell. The signals obtained for two streptavidin-coated spots are shown in Figure 8. The reference signal was subtracted in both cases. This normalization reduced considerably temperature-induced signal fluctuations and the step caused by the difference in the refractive indexes of buffer and sample. The relative standard deviation of
Figure 7. Selective binding of streptavidin molecules to a biotinmodified surface.
Figure 8. Biotin binding to a monolayer of immobilized streptavidin. The reference signal was subtracted in both cases.
the signal in six measurements was about 15%. This result demonstrates a very high sensitivity of the new double-wavelengths SPR technique. A difference corresponding to ∆n ) 2 × 10-6 could be measured. This is about 1 order of magnitude better than that obtained in measurements described earlier9 applying single-wavelength SPR imaging. The measured step of the relative ac component (ac/dc ratio) by detection of biotin binding was about 10-3. It was demonstrated earlier33 that relative changes in the ac component as small as 10-5 can be detected by applying the lock-in technique and properly arranged double-beam optical scheme. This value corresponds to a change of the refractive index by less than 10-7 or to layers with a thickness in the sub-picometer range if temperature drift does not limit the sensitivity. The method used for immobilization of streptavidin resulted in the formation of streptavidin monolayers. Formation of extended matrixes of receptor molecules will yield a further increase of the sensitivity, providing applications of the new SPR technique for molecules which are several times smaller than biotin, e.g., for any organic compounds. Hybridization of Oligonucleotids. The specific hybridization of complementary DNA strands is a powerful tool in medical (33) Liger, V.; Zybin, A.; Bolshov, M.; Niemax, K. Appl. Spectrosc. 2002, 56, 250.
Analytical Chemistry, Vol. 77, No. 8, April 15, 2005
2397
Figure 9. Hybridization of the oligonucleotide G-HCV-c of the sequence 5′-TTT CGG GTC CTT TCT TGG-3′ to the spot coated with the complementary olygonucleotide G-HCV of the sequence 5′TTT TTC CAA GAA AGG ACC-3′. No binding to noncomplementary oligonucleotides was detected.
diagnostics. It is used for the detection of bacteria contaminations in food and the environment, for forensic tests, in gene analysis, in investigation of gene activity and regulation, and in other fields. Here, an application of the double-wavelengths SPR technique for the hybridization test will be illustrated by modeling of diagnostics and genotyping of human hepatitis C virus. Three DNA receptors are used; one is corresponding to the common sequence of the virus (G) and the other two are to the main types of the virus (T1 and T2). An addition of the complementary DNA oligomers (cG) to the simple oligonucleotide array comprising G, T1, and T2 receptor sequences resulted in a selective response of the receptor G (Figure 9), while no responses were observed from the other receptors. Subsequent washing with buffer had no effect on the SPR signal at the receptor sequence G, indicating formation of stable cG-G double-stranded DNA. It should be pointed out that changes of the bulk refractive index had no influence on the observed SPR signals at the nucleotide concentrations used. The calculated increase in the bulk refractive index due to an addition of 1 µM of oligonucleotides with the molecular weight 4500 is below 1 × 10-6. Therefore, the observed SPR signals are caused by selective binding (hybridization) of the oligonucleotide cG to its complementary receptor G. There was no decrease of the signal in pure buffer because the desorption of this analyte at the used temperature is very slow. The high selectivity of hybridization was confirmed by an experiment where first cG and after its hybridization cT2 was added (Figure 10). The addition led to a SPR signal at the corresponding sensing spots without any mutual interference. Simultaneous addition of cG, cT1, and cT2 resulted in response of corresponding receptor oligonucleotides (Figure 11). The difference in the response kinetics indicates a difference in the rate constants of adsorption. Comparison with the rate of solution exchange demonstrates that the T1 response reached its saturated value even before the whole (or essential part of the) buffer in the flow cell was exchanged by the sample solution. The hybridization of G and T2 oligonucleotides was slower. Beside selective hybridization, we also tested denaturation of the hybridized oligonucleotides as detected by the double2398 Analytical Chemistry, Vol. 77, No. 8, April 15, 2005
Figure 10. Selective hybridization of two oligonucleotides with complementary sequences immobilized at different spots.
Figure 11. Simultaneous examination of three different hybridization processes.
wavelengths SPR technique. A perfusion with 15 M formamide solution for 20 min led to recovery of the SPR signal to its initial value. Now the hybridization can be performed again. Figure 12 shows the response for the first hybridization (first curve) and five following responses, each after the denaturation by the formamid solution (curves 1F...5F). The last denaturation displayed in Figure 12 was performed by pumping of heated buffer (90 °C) through the cell for 10 min. Five repetitions of the whole hybridization/denaturation cycles demonstrated good reproducibility of the whole process, high stability of sensing layers, and appropriate operation of this new detection technique. Notes on the Limit of Detection. With introduction of a new SPR detection technique, there is always the question of how its sensitivity compares with existing SPR methods. However, such analysis is often difficult due to the lack of corresponding data published or complete information on the limiting factors. It should also be noted that the value of the “smallest measurable ∆n” depends on the process to be investigated (causing the ∆n), for example, on speed of the process. A reasonable comparison criterion for the detection power of different methods is the S/N ratio measured when molecules of the same substance bind the same receptor surface. Biotin binding on streptavidin might be a good candidate for such characterization, but is has not been
imaging is affected by effects such as light scattering or specularly reflected light, particularly if high-density arrays are used. Therefore, the imaging methods based on single-wavelength reflectivity measurements do not reach the sensitivity of single or few spot instruments. To our knowledge, the only publication giving detailed considerations of the minimum detectable ∆n by SPR single-wavelength imaging9 reports a limit of detection of ∆n ∼ 1.8 × 10-5 based on measurements of all spots within a 24 mm2 light beam. Our double-wavelengths measurements revealed a detection limit of ∆n ∼ 2 × 10-6 with a 6 mm2 beam.
Figure 12. Multiple hybridization/denaturation cycles. Curve 0: first hybridization of 1 µM cG oligomer with the G sensor spot; curves 1-5F: hybridization signals after successive denaturation by 15 M formamid; curve 90° (dashed): hybridization after temperature denaturation of the cG complex.
measured by the other SPR methods. For many state-of-the-art instruments the limiting factor is not the instrumental sensitivity but the temperature fluctuations and drifts. Indeed, typical values for temperature stabilization in commercial instruments are within 0.01 °C, which corresponds to a variation of ∆n ∼ 10-6 for water, the typical detection limit of SPR methods. Therefore, measurements substantially below ∆n ∼ 10-6 require additional efforts for temperature stabilization and, mainly, intelligent measures and/or new approaches. Some of them are mentioned in the introduction above. One of the most efficient ways to correct instrumental and temperature drift is the normalization of the signal using a nearby spot as a reference. This solution is especially easy to accomplish by 2-D imaging measurements. However, the sensitivity of 2-D
CONCLUSION The use of a double-wavelength approach in the SPR technique allowed us to improve the signal/noise ratio and to detect small molecules with molecular weight as typical drugs. This indicates that the technology should be applicable in so important fields as high-throughput screening. The required parallelization of the technique was also illustrated here. The use of a reference channel allowed us to reduce essentially the influence of temperature drifts and to obtain a resolution of about 2 × 10-6 refraction units. The new technique was successfully applied for hybridization analysis. It can also be applied in genomics, proteomics, medical diagnostics, and many other fields of science and industry where a real time ultra-sensitive analysis of adsorption or of analyte-receptor binding is of interest. ACKNOWLEDGMENT The authors are grateful to Drs. A. Terfort (Institut fu¨r Anorganische und Angewandte Chemie, Universita¨t Hamburg) and W. Eck (Institut fu¨r Analytische und Physikalische Chemie, Universita¨t Heidelberg) for providing us with ω-modified thiols. Received for review December 14, 2004. Accepted February 9, 2005. AC048156V
Analytical Chemistry, Vol. 77, No. 8, April 15, 2005
2399
