Modern biological sciences seek to elucidate and understand the functions of the thousands of different molecules inside living cells and describe changes in their concentrations as a function of disease. Proteins represent one of the most important subsets of molecules. Interactions between proteins are currently understood to be the drivers of biological change. Unfortunately, methods to measure protein interactions at the scale required to gain a comprehensive picture of the state of an entire cell are currently lacking.
Surface-plasmon resonance,1 a noncontact and label-free optical technique that measures changes in refractive index at a metal-solution interface, has long been used to measure up to hundreds of protein interactions at a time. But to date, such approaches have been limited by the requirement that one of the two binding partners must always remain the same: see Figure 1(A) and 1(B). Existing microfluidic approaches have increased the throughput and parallelism of such measurements, but have not addressed this fundamental limitation.
Figure 1. Surface-based detection with (A) a single probe type and single target, (B) multiple targets and a single probe, and (C) multiple types of probes and multiple targets.
Using a popular technique known as soft lithography, we can microfabricate channels, chambers, and valves the thickness of a human hair or smaller (<100μm). The technique, first introduced by the Quake group,2,3 has gained wide acceptance for its ease of fabrication, fast turnaround time, and the inexpensive material it typically uses, poly(dimethylsilozane) (PDMS). Our approach uses a complex series of PDMS microfluidic valves to form tiny isolated chambers, each with a volume of only 700pl (7×10−10l). By forming an array of 264 such chambers on a substrate and addressing the valves using well-established multiplexing techniques, our systems can both isolate a range of target molecules immobilized on the surface and probe molecules in solution: see Figure 1(C). The inputs required to control the array scale much more slowly than the number of chambers. We use only 20 inputs to address more than 1000 microvalves across 264 chambers. This ‘element addressability’ is a key concept distinguishing our work from that of predecessors.
We have demonstrated our device by performing a series of common protein-interaction assays. The first was serial dilution, which to date has been performed manually at great time expense and with the inherent errors of pipetting. Our device contains a serial-dilution network, i.e., a series of connected channels with chaotic advection micromixers first introduced by the Whitesides group,4 to dilute an input sample stream five times up to a factor of 32. We next demonstrated immobilization of a protein, human α-thrombin, to gold surfaces in our device's chambers. By introducing an antibody with affinity for the thrombin, we observed the binding kinetics of a 1:1 Langmuir interaction. By performing the experiment simultaneously at different concentrations, we successfully determined the binding affinity of the antibody for its target.
Another critical feature of any parallel microfluidic system is that experiments in one chamber should not affect different experiments conducted in an adjacent chamber. To quantify this, we conducted an experiment in which we immobilized two different proteins into different sets of chambers on the device. We then introduced a mixture of two antibodies, each of which had been fluorescently labeled with different dyes. After performing binding reactions in all chambers and washing, we measured the concentrations of fluorescent antibodies. Our data showed that the binding of an antibody to its noncognate target and the nonspecific binding of the antibody in the device outside the chambers are small compared to the total binding detected within the chambers. These results bode well for use of such systems in clinical settings.
In summary, we have fabricated and tested a microfluidic system for detection of protein interactions that combines the advantages of small volumes, parallelism, and element addressability. The resulting system is capable of on-chip serial dilutions, measurement of binding affinities, and sample recovery, a set of features as yet unavailable in other research or commercial systems. Future work will seek to expand the range of molecules that may be interrogated with such systems, including small molecules, nucleic acids, and weakly bound protein complexes. We hope that microfluidic surface-plasmon-resonance systems can become a valuable tool in the search for understanding in biology and medicine.
University of British Columbia
Eric Lagally is an assistant professor in the Michael Smith Laboratories and Department of Chemical and Biological Engineering. His research group focuses on microfluidic molecular diagnostics.
4. X. Jiang, J. M. K. Ng, A. D. Stroock, S. K. W. Dertinger, G. M. Whitesides, A miniaturized, parallel, serially diluted immunoassay for analyzing multiple antigens, J. Am. Chem. Soc. 125, no. 18, pp. 5294-5295, 2003. doi:10.1021/ja034566+