Adaptation of a Surface Plasmon Resonance Biosensor with

Miguel Abrantes,† M. Teresa Magone,‡,§ Lisa F. Boyd,| and Peter Schuck*,†. Molecular Interactions Resource, Division of Bioengineering and Phys...
0 downloads 0 Views 116KB Size
Anal. Chem. 2001, 73, 2828-2835

Articles

Adaptation of a Surface Plasmon Resonance Biosensor with Microfluidics for Use with Small Sample Volumes and Long Contact Times Miguel Abrantes,† M. Teresa Magone,‡,§ Lisa F. Boyd,| and Peter Schuck*,†

Molecular Interactions Resource, Division of Bioengineering and Physical Science, ORS, OD, National Eye Institute, and Molecular Biology Section, Laboratory of Immunology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland 20892

The efficient delivery of sample to surface-immobilized sites is a key element in biosensing. For a surface plasmon resonance (SPR) biosensor, this has been addressed by constant flow through a microfluidic system with a sample injection loop (Sjo1lander, S.; Urbaniczky, C. Anal. Chem. 1991, 63, 2338-2345). The present study describes an alternative mode of sample delivery without constant unidirectional flow. It was implemented on a commercial Biacore X SPR biosensor equipped with a microfluidic cartridge, but with the fluidic handling performed by an externally computer-controlled syringe pump. We demonstrate that sample volumes as low as 2 µL can be reproducibly positioned to cover the sensor surfaces, manipulated in a serial fashion, efficiently mixed by applying an oscillatory flow pattern, and fully recovered. Compared to the traditional continuous unidirectional flow configuration, we found very similar kinetic responses at high analyte concentrations and slightly slower responses at low concentrations, most likely due to depletion of analyte from the small sample volumes due to surface binding. With the antibody-antigen systems tested, binding parameters were obtained that are generally within 10% of those from conventional experiments. In the new configuration, biosensor experiments can be conducted without the usual constraints in the surface contact time that are correlated with sample volume and mass transport rate. This can translate to improved detection limits for slow reactions and can facilitate kinetic and thermodynamic binding studies. Since the introduction of optical biosensors as research tools for the characterization of reversible interactions of biological macromolecules, this method has matured into a tool that is * Corresponding author: (phone) 301 435-1950; (fax) 301 480-1242; (e-mail) [email protected]. † Molecular Interactions Resource, Division of Bioengineering and Physical Science, ORS, OD. ‡ National Eye Institute. § Department of Ophthalmology, Georgetown University Medical Center, Washington, D.C. 20007. | National Institute of Allergy and Infectious Diseases.

2828 Analytical Chemistry, Vol. 73, No. 13, July 1, 2001

routinely and widely used in many fields where molecular recognition events are of interest, among them immunology, virology, and receptor-ligand interactions (for reviews, see for example, refs 1-9). Because relatively small sample volumes are required and the measurement is suitable for automation, they also found wide application where rapid screening of many variants of the mobile reactant is of interest (e.g., ref 10). In many cases, the biosensor can provide unique insights into the equilibrium and dynamic aspects of biomolecular interactions. This appears to be true in particular for systems with very high and very low affinity, slow-to-moderate reaction rate constants, or complex multicomponent systems. During the past decade, many different experimental procedures for sample handling and enhanced detection have been developed, and significant developments in the understanding of the theoretical and practical potential as well as limitations of affinity biosensor technology for characterizing molecular interactions have taken place (see, for example, refs 9 and 11-19). The measurement is based on changes in the optical properties of the sensor surface due to binding of a mobile reaction partner (the analyte) to a surface-immobilized reaction partner. Usually, (1) Margulies, D. H.; Plaksin, D.; Khilko, S. N.; Jelonek, M. T. Curr. Opin. Immunol. 1996, 8, 262-70. (2) Malmqvist, M.; Karlsson, R. Curr. Opin. Chem. Biol. 1997, 1, 378-83. (3) Schuck, P. Annu. Rev. Biophys. Biomol. Struct. 1997, 26, 541-66. (4) Leatherbarrow, R. J.; Edwards, P. R. Curr. Opin. Chem. Biol. 1999, 3, 5447. (5) Huber, A.; Demartis, S.; Neri, D. J. Mol. Recognit. 1999, 12, 198-216. (6) Nice, E. C.; Catimel, B. Bioessays 1999, 21, 339-52. (7) Myszka, D. G. J. Mol. Recognit. 1999, 12, 390-408. (8) Slepak, V. Z. J. Mol. Recognit. 2000, 13, 20-6. (9) Winzor, D. J. J. Mol. Recognit. 2000, 13, 279-98. (10) Gomes, P.; Giralt, E.; Andreu, D. Vaccine 1999, 18, 362-70. (11) Leckband, D. E.; Kuhl, T.; Wang, H. K.; Herron, J.; Muller, W.; Ringsdorf, H. Biochemistry 1995, 34, 11467-78. (12) Schuck, P., Minton, A. P. Trends Biochem. Sci. 1996, 252, 458-60. (13) Hall, D. R.; Winzor, D. J. Anal. Biochem. 1997, 244, 152-60. (14) Schuck, P.; Millar, D. B.; Kortt, A. A. Anal. Biochem. 1998, 265, 79-91. (15) Lyon, L. A.; Musick, M. D.; Natan, M. J. Anal. Chem. 1998, 70, 5177-83. (16) Ober, R. J.; Ward, E. S. Anal. Biochem. 1999, 273, 49-59. (17) Ober, R. J.; Ward, E. S. Anal. Biochem. 1999, 271, 70-80. (18) Shank-Retzlaff, M. L.; Sligar, S. G. Anal. Chem. 2000, 72, 4212-20. (19) Rowe-Taitt, C. A.; Golden, J. P.; Feldstein, M. J.; Cras, J. J.; Hoffman, K. E.; Ligler, F. S. Biosens. Bioelectron. 2000, 14, 785-94. 10.1021/ac0100042 Not subject to U.S. Copyright. Publ. 2001 Am. Chem. Soc.

Published on Web 05/16/2001

the assumption is made that the immobilization and the vicinity of the surface do little to affect the interaction of the reaction partners, such that the time course of the signal reflects the bimolecular interaction kinetics and, in equilibrium, its thermodynamics. In the most commonly used commercial instruments (the Biacore surface plasmon resonance (SPR) sensor) with a microfluidic flow system,20 an injection loop is loaded (usually with between 25 and 120 µL), and a sample plug is chased by running buffer. In this configuration, a well-defined contact time of the analyte sample is achieved (as determined by the volume of sample in the injection loop divided by the flow rate), during which the time course of surface binding to one or more reactive and control surfaces can be observed. This is commonly followed by a washout phase, which allows the observation of the kinetics of dissociation of reversibly bound analyte from the surface. Such a constant flow system has the virtues of a very high baseline stability and the ability to expose several sensor surfaces sequentially to the same analyte. Other commercial instruments (for example, the Affinity Sensors resonant mirror21) are based on one or more cuvettes, each with a single reactive surface, that are filled with 100-200 µL of sample to allow surface binding and are rinsed to allow observation of the dissociation. The quantitative interpretation of the time course of binding is most frequently based on first-order biomolecular reaction kinetics, which predicts single-exponential binding progress. However, frequently, the experimental data show significant deviations from this model.22 In addition to complex biomolecular interaction kinetics,23,24 which can be difficult to unravel, a number of potential artifacts have been shown to be possible causes for such multiphasic surface binding progress.25-28 For example, it is well known that the mass transport rate of the analyte to the surface sites is an important factor, in particular when high densities of surface sites are used.29-33 One popular technique for avoiding transport-limited kinetics in the flow system is the use of an increased flow rate.22 However, as the molecular mass transport rate only increases with the cube root of the flow rate, a too high sample consumption frequently prevents an effective implementation of this approach, unless very short contact times and the concomitant loss of information are tolerated. Much simpler is the interpretation of the binding signal at steady state (or equilibrium), because the restriction to characterize only the thermodynamic aspects of the interaction eliminates any ambiguity of the kinetic binding mechanism or mass transport considerations. Even more independence of potential sensor(20) Sjo ¨lander, S.; Urbaniczky, C. Anal. Chem. 1991, 63, 2338-45. (21) Cush, R.; Cronin, J. M.; Steward, W. J.; Maule, C. H.; Molloy, J.; Goddard, N. J. Biosens. Bioelectron. 1993, 8, 347-53. (22) Karlsson, R.; Roos, H.; Fa¨gerstam, L.; Persson, B. Methods: Companion Methods Enzymol. 1994, 6, 99-110. (23) Lipschultz, C. A.; Li, Y.; Smith-Gill, S. Methods 2000, 20, 310-8. (24) Glaser, R. W.; Hausdorf, G. J. Immunol. Methods 1996, 189, 1-14. (25) Muller, K. M.; Arndt, K. M.; Pluckthun, A. Anal. Biochem. 1998, 261, 14958. (26) Nieba, L.; Krebber, A.; Pluckthun, A. Anal. Biochem. 1996, 234, 155-65. (27) Schuck, P. Curr. Opin. Biotechnol. 1997, 8, 498-502. (28) Minton, A. P. Biophys. J. 1999, 76, 176-87. (29) Glaser, R. W. Anal. Biochem. 1993, 213, 152-61. (30) Balgi, G.; Leckband, D. E.; Nitsche, J. M. Biophys. J. 1995, 68, 2251-60. (31) Schuck, P. Biophys. J. 1996, 70, 1230-49. (32) Schuck, P.; Minton, A. P. Anal. Biochem. 1996, 240, 262-72. (33) Vijayendran, R. A.; Ligler, F. S.; Leckband, D. E. Anal. Chem. 1999, 71, 5415-2.

related artifacts is gained by interpretation of the equilibrium competition isotherm, from which the thermodynamics of the solution interaction can be measured. However, although such variations on experimental design can be appealing from a theoretical perspective, it can sometimes be impractical to implement, because of the significantly higher sample consumption (in particular, in the competition approach) or because the extended contact times that can be required to reach a steady-state signal cannot be experimentally achieved. From the considerations described above, it is clear that sample volume, contact time, and flow rate are crucial and correlated parameters when biosensor experiments are being conducted. In particular, the maximal sample volume as defined either by the size of the injection loop or by the availability of the material can frequently be a limiting factor in the experimental design. Several attempts to solve this problem of the flow system have been described previously, including sequences of rapidly repeated injections, circumventing the sample injection loop by placing the sample into the running buffer reservoir,34 and equilibrium titration by recirculation of the sample in a modified Biacore X instrument.14 However, the later approach still requires minimal analyte volumes of the order of 200 µL. The present study is concerned with an alternative principle of sample handling, in which a small sample volume of less than 10 µL, fully recoverable, can be used for generating a virtually unlimited contact time, while at the same time maintaining sufficient mixing of the sample to generate a high mass transport rate, equivalent to that from a unidirectional flow at high rate. As will be described, such a system can be implemented easily by combining the commercial Biacore X surface plasmon resonance biosensor and its microfluidic cartridge with an externally computer controlled syringe pump. EXPERIMENTAL SECTION Materials and Modification of the Sensor Surface. Two model systems of interacting macromolecules were used. For the first, anti-human myoglobin monoclonal antibody was purchased from Biacore (Piscataway, NJ); human myoglobin was obtained from Biacore and Sigma (St. Louis, MO). Second, a soluble analogue of the MHC class I protein, H-2Dd in complex with murine β2 and motif peptide, AGPARAAAL, was expressed in bacteria, refolded, and purified as described elsewhere.35 The monoclonal antibody 34-2-12S, which binds a defined epitope on the R3 domain of H-2Dd,36 was purified from culture supernatants by protein-A Sepharose chromatography. The proteins were covalently attached to carboxymethylated dextran chips CM5 (Biacore) by amine coupling.37,38 In short, using a flow rate of 5 µL/min, a 35-µL injection of a mixture of N-ethyl-N′-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide was employed to activate the surface, followed by injection of proteins diluted into 10 mM sodium (34) Myszka, D. G.; Jonsen, M. D.; Graves, B. J. Anal. Biochem. 1998, 265, 326-30. (35) Natarajan, K.; Boyd, L. F.; Schuck, P.; Yokoyama, W. M.; Eliat, D.; Margulies, D. H. Immunity 1999, 11, 591-601. (36) McCluskey, J.; Bluestone, J. A.; Coligan, J. E.; Maloy, W. L.; Margulies, D. H. J. Immunol. 1986, 136, 1472-81. (37) Johnsson, B.; Lo¨fås, S.; Lindquist, G. Anal. Biochem. 1991, 198, 268-77. (38) Schuck, P.; Boyd, L. F.; Andersen, P. S. In Current Protocols in Protein Science; Coligan, J. E., Dunn, B. M., Ploegh, H. L., Speicher, D. W., Wingfield, P. T., Eds.; John Wiley & Sons: New York, 1999; Vol. 2, pp 20.2.1-21.

Analytical Chemistry, Vol. 73, No. 13, July 1, 2001

2829

acetate buffer pH 4.8, at concentrations of between 20 and 50 µg/ mL. Protein injection was stopped when the desired immobilization level was reached, and the surface was deactivated by 35 µL of 1 M ethanolamine, pH 9.0. Binding experiments were performed at 25 °C, using the standard running buffer, 10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20. For the interaction of myoglobin with the monoclonal antibody, 10 mM glycine buffer at pH 2.1 was used to strip the surface of bound analyte, whereas 50 mM phosphoric acid was used for the 34-212S monoclonal antibody surface. Biosensor and Microfluidic Sample Handling. For the surface binding experiments, a Biacore X surface plasmon resonance biosensor (Biacore) equipped with microfluidic cartridge20 was used in the “external injection mode”. The fluidic control of the instrument was disconnected, allowing it only to passively acquire data from two sensing spots at the sensor surface. The signal difference from a functionalized and a nonfunctionalized surface was recorded as the signal from specific binding to the immobilized sites. As described by Sjo¨lander and Urbaniczky, the commercial microfluidics consists essentially of a microchannel leading to two (or more) serially connected flow chambers containing the sensor surface and a microchannel leading to the drain.20 It also has an HPLC-like injection loop, which is controlled by pneumatic valves and is used in the conventional configuration for analyte injection. In our smallvolume experiments, however, it was used only during regeneration of the sensor surface for applying a short plug of acidic solution for regenerating the sensor surface. The microfluidics was extended in the following way: Polyethylene tubing (i.d. 0.381 mm, o.d. 1.09 mm) was stretched and elastically fit into the connector block previously leading to the drain. This tubing defined the new inlet and allowed convenient aspiration of sample from an Eppendorf tube. The connection of the microfluidic cartridge previously designated as the inlet was connected to a syringe pump with a stepping motor (model 402 from Gilson Inc., Middleton, WI) equipped with a 250-µL syringe and three-way valve. The syringe pump was separately computer controlled with a custom-written software that allowed the automated execution of a series of operations with control of time, volume, and flow rate, at submicroliter precision (the computer program is available from the authors on request). No hysteresis was observed at short time periods, but a slow drift of ∼1.5 µL/h was observed after more than 1 h of oscillatory flow. The total volume of the system from the inlet tubing to both flow chambers was measured by aspiration of well-defined volumes of air into the water-filled microfluidics. Observation of the response at both sensor surfaces during this process permitted monitoring the transition of the air-water meniscus across the sensor surfaces because of the large difference of refractive index. With our current inlet tubing, the center between both flow channels was measured to be 24.3 µL, with each of the flow channels switching between water and air signals within 0.1 µL, and with a 0.7-µL minimal volume for switching both flow channels. In a typical small-volume binding experiment, the microfluidics system was first rinsed and filled with running buffer. Then, sequentially, a volume of 2 µL of air, 0.3 µL of sample, 2 µL of air, 5 µL of sample, 2 µL of air, 0.3 µL of sample, and 2 µL of air was aspirated into the inlet tubing at a rate of 20 µL/min (Figure 3). 2830 Analytical Chemistry, Vol. 73, No. 13, July 1, 2001

Figure 1. Binding of 100 nM H-2Dd/motif to immobilized monoclonal antibody 34-2-12S. The signal after a conventional injection of 10 µL at a flow rate of 5 µL/min (dashed line) is superimposed on the signal from a sample plug aspirated from the opposite direction (solid line). The sample plug was separated from the running buffer by 0.5-µL air bubbles, and the signal artifacts from the transition of the air bubbles are removed from the plot. All signals shown are net signals after subtraction of the bulk refractive index measured in a reference surface (since in the conventional flow configuration the reference surface is connected in series, a small time delay between the signals from the two surfaces results in spikes at the beginning and end of a sample injection).

This arrangement defines a 5-µL plug of sample, sandwiched on both sides between air bubbles which are subdivided by 0.3 µL of sample (see Results). This was followed by aspiration of running buffer at a volume such that the sample was centered across the sensor surfaces. To prevent the formation of a depletion zone at the functionalized surface, an oscillatory back-and-forth flow pattern was applied, with amplitudes typically of the order of 1 µL, and at flow rates of between 10 and 50 µL/min. After the measurement of the surface binding kinetics, the sample was pumped back to the inlet, where it could be recovered, and the sensor surface was rinsed at a high flow rate with a unidirectional flow through use of the three-way valve of the syringe pump and aspiration of running buffer from a reservoir. RESULTS With the goal of developing a sample handling procedure alternative to the conventional unidirectional flow chasing a fixedvolume sample plug, we performed first several initial experiments examining the robustness of the biosensor signal. First, we verified the absence of a dependence of the SPR signal on the directionality of the flow. This was accomplished by aspirating a sample plug through the new inlet tubing, separated from the running buffer by small air bubbles, and followed by a unidirectional flow (which is reversed relative to the conventional flow direction). As can be seen in Figure 1, the binding signal in both cases was very similar. The association signals virtually superimpose. It is noteworthy that the dissociation of surface-bound material proceeds faster in the reverse orientation; this may be the result of a more efficient removal of remaining analyte in the microfluidics by use of air bubbles as compared to the laminar flow. We observed that the introduction of such small air bubbles (0.2-2 µL) transiently led to artifacts during the time of their passage through the flow

Figure 2. Binding of 100 nM human myoglobin to immobilized monoclonal antibody. Signals are obtained after aspiration of a small sample volume in different configurations: 2-µL sample without mixing (dotted line), 5-µL sample without mixing (circles), 5-µL sample replicate with convection (squares), and 5-µL sample with oscillatory flow at an amplitude of 1 µL (solid line), followed by regeneration of the surface at 1300 s. Between approximately 1300 and 1400 s, artifacts from regeneration are visible in the solid line. After ∼1400 s, running buffer was applied.

chamber (e.g., at 0 s in Figure 2), but this was completely reversible and did not produce significant permanent signal offsets (offsets of 3 µL (with mixing amplitudes of >1 µL). When the observation times were extended to several hours and sample volumes of 5 µL or below were used, a small net translation of ∼1.5 µL/h could be observed. If not taken into account, this led to transient entry of the sample meniscus in the flow chamber and resulted in a series of periodic spikes (data not shown). This can be easily corrected, however, by interrupting the periodic flow and centering of the sample across the flow cells, or it can be avoided by using slightly larger sample volumes for association phases of several hours or longer. In the following, quantitative analyses of the biosensor signals are described. When working with smaller sample volumes, the effects of depletion of the total concentration of analyte molecules in the sample due to surface binding, i.e., second-order kinetics, must be considered. Using the published conversion constant for the Analytical Chemistry, Vol. 73, No. 13, July 1, 2001

2831

Figure 3. Schematic view of the small sample volume configuration using the microfluidics of the Biacore X. Running buffer is depicted as wavy areas, air as clear areas, and sample as dotted areas. The channels leading to and connecting the flow channels can hold a sample volume of Vtot ) 2-15 µL. This volume is much larger than the volume of the flow channels VFC (