Electrophoretic Concentration of Proteins at Laser-patterned Nanoporous Membranes in Microchips Simon Song, Anup K. Singh and Brian J. Kirby Sandia National Laboratories, Livermore, CA revised version Address correspondence to: Brian J. Kirby, Ph.D Microfluidics Department Sandia National Laboratories Livermore, CA 94550 bjkirby@sandia.gov Voice: 925 294 2898 FAX: 925 294 3020 Electrophoretic Concentration of Proteins at Laser-patterned Nanoporous Membranes in Microchips Simon Song, Anup K. Singh and Brian J. Kirby Sandia National Laboratories, Livermore, CA ABSTRACT Laser-patterning of nanoporous membranes at the junction of a cross channel in a microchip is used to integrate protein concentration with an electrokinetic injection scheme. Upon application of voltage, linear electrophoretic concentration of charged proteins is achieved at the membrane surface because buffer ions can easily pass through the membrane while proteins larger than the molecular weight cutoff of the membrane (>5.7 kD) are retained. Simple buffer systems can be used, and the concentration results evince uniform buffer ion concentration maintained throughout the process. Local and spatially-averaged concentration are increased by four and two orders of magnitude, respectively, upon injection with moderate voltages (70-150V) and concentration times (100s). The degree of concentration is limited only by the solubility limit of the proteins. The porous polymer membrane can be used repeatedly as long as care is taken to avoid protein precipitation. INTRODUCTION The development of microfluidic devices during the past decade has facilitated improvements in biological and chemical analysis.1, 2 Incorporating sample preconcentration, mixing, reaction, separation, and detection functions into microchips can facilitate automation of labor-intensive sample analysis and reduce reagent consumption. Sample preconcentration steps are ever-present components of analytical systems, since concentration of trace samples facilitates their detection. A sample preconcentration process concentrates sample analytes in a compact volume prior to detection and increases the signal-to-noise ratio, resulting in improved detectivity. The numerous liquid-phase sample preconcentration techniques performed on microchips to date can be grouped roughly into three classes (surface techniques, electrokinetic manipulation in spatially varying electric field, porous filtering) based on the transport physics involved. Surface techniques use sample adsorption to surfaces for concentration and a surface or solvent change for elution. Solid-phase extraction3-9 typically employs hydrophobic or electrostatic attraction and elutes using a change in buffer. Film electrode techniques10 use electrostatic attraction at micropatterned electrodes to capture analytes and release with changes in voltage. Programmable surface techniques use surfaces with voltage- or temperature-addressable properties11 and elute by changing the surface properties. Electrokinetic manipulation in spatially-varying electric fields has been used in a variety of forms to concentrate analytes.12, 13 The general property of these systems is that spatial variations in buffer properties lead to spatial variations in electric field and therefore concentration of analytes. Isoelectric focusing involves spatially-varying electrokinetic transport in pH gradients.14-16 Field-amplified sample stacking17-23 and isotachophoresis24, 25 involve conductivity-induced spatial electric field variations in two and three buffer systems, respectively. Temperature gradient focusing26 uses temperature gradients to establish steady-state ionic strength gradients. Porous filtering techniques concentrate analytes at semipermeable interfaces. Interfaces used at a microchip level include a silicate layer deposited between a cover and substrate of a microchip,27, 28 and an etched capillary.29 Porous filtering is inherently different from other concentration techniques in that the concentration is specific to certain size classes of analytes. Although the above techniques have been shown to concentrate analytes by factors ranging from one to five orders of magnitude, their application can be limited by fabrication complexities, buffer handling challenges, or physical uncertainties. Several of the techniques (solid-phase extraction, isoelectric focusing, field-amplified sample stacking, isotachophoresis) require spatial or temporal buffer variations. Others (film electrode technique, programmable surface technique, temperature gradient focusing) require microfabrication complexity that might not always be justified. In well-defined systems, porous filtering is a straightforward and intuitive concentration technique. Porous interfaces that are permeable to buffer ions but impermeable to sample analytes can be used to stack analytes at the interface upon application of voltage. When electrophoretic sample transport is used, the increase of sample concentration should be linearly proportional to the product of applied voltage and the length of time over which it is applied (as long as the geometry remains constant). These methods are simple to carry out, but results presented to date have shown that the increase of sample concentration is a non-linear function of sample-injection time and voltages for unknown reasons.27, 28 This paper presents a porous membrane sample preconcentration system that is straightforward to use with pinched electrokinetic injections,30 allows concentration up to the protein solubility limit, and affords a linear response to the time-voltage product. We have implemented a technique for in-situ fabrication of nano-porous polymer membranes on microchip, used previously for microchip dialysis of proteins.31 In the present studies, a membrane is laser-patterned at the junction of a cross channel in microchips. By implementing the electrokinetic pinched injection, sample analytes (charged proteins) concentrate at the surface of membrane and the concentrated sample can be eluted downstream. In the following sections, the fabrication method and membrane characteristics are presented, then, the performance of the preconcentration system is explored in detail. EXPERIMENTAL SECTION Microchip fabrication and surface functionalization The fabrication procedures are based on previous work and are summarized briefly here. Fused silica (Corning 7980) microchips with cross channels were fabricated in-house using standard photoresist, UV patterning, and isotropic HF wet-etch steps.31 Following thermal bonding, the silica channel surfaces were prepared for covalent attachment of polymers by coating with 3-trimethoxysilylpropyl acrylate through acid-catalyzed hydrolysis and condensation.32 Channels were first conditioned by 30-minute flushes with 1 M HCl then 1 M NaOH, then rinsed with deionized water and dried. The microchannels were then incubated for 30 minutes with a solution consisting of a 2:2:1 (vol) mixture of deionized water, glacial acetic acid, and 3-trimethoxysilylpropyl acrylate. The channel was then rinsed thoroughly with deionized water and 1-propanol. Membrane fabrication Zwitterionic polymer membranes with molecular weight cutoff <5.7 kD were fabricated inside microchannels using laser-patterning phase-separation techniques specified previously31 and summarized briefly here. A 355 nm laser beam was used to excite a photoinitiator in a monomer/solvent solution, leading to polymerization and phase-separation of a thin membrane that was covalently bonded to the acrylate-functionalized silica channel surface. The irradiated region was defined by shaping the beam into a 40?m x 300 ?m laser sheet by imaging an adjustable slit with cylindrical and spherical optics. The resulting membranes were 50?m-thick (Figure 1). Laser exposure time was typically 3.5 minutes. After polymerization, the channels were flushed thoroughly with water. Electroosmosis was suppressed in the silica channels by coating with linear polyacrylamide32 after membrane fabrication. Channels were coated by placing the microchip 0.5” above a 30W, 365nm-cutoff UV lamp (Model EW-97605-00, Cole-Parmer, Vernon Hills, IL) for 30 min while the channel was incubated with a solution of 5mg/mL acrylamide in deionized water containing 200ppm hydroquinone (Sigma-Aldrich, St. Louis, MO) and 0.2 mg/mL 2,2’ azobis(2-methylpropanimidamide)dihydrochloride (Wako Chemicals, Richmond, VA). The channel was then rinsed thoroughly with water. The entire coating process, including filling and rinsing the channel, was repeated twice. Analytes and instrumentation Two Alexa-Fluor-488-labeled proteins were used as model analytes: 0.5 ?M bovine serum albumin (66kD, pI=4.9, Molecular Probes) in a pH 7.5, 10 mM phosphate buffer and 0.5 ?M phosphorylase b (94 kD, pI=6.3, Sigma-Aldrich) in a pH 9.3, 10 mM borate buffer. Protein concentration increases were quantified with a fluorescence microscope (Olympus, model IX70) and 12 bit, mono, cooled CCD camera (CoolSNAPHQ, Roper Scientific, Inc., Tucson, AZ) at 20 ms exposure time. Background signal from dark current and ambient light was subtracted before analysis. Photobleaching was minimized through the choice of fluorescent label and by illuminating the field of view only briefly at each measurement point. In order to inject and elute analytes by electrophoresis, voltage (20-150V) was applied across the membrane using a high voltage source33 (?ChemLab?? Sandia National Laboratories, Livermore, CA). The detailed voltage configuration is described in the following sections. Sample loading, concentration, and injection A cross channel is used with a membrane laser-patterned at the intersection (Figure 2a). Each channel is connected to its own reservoir. All the channels and reservoirs are filled with the same buffer, but the buffer to the left of the membrane also contains the fluorescently-labeled proteins. To load proteins, voltage (15-150 V) is applied for 20-100 seconds between the sample and sample waste reservoirs (Figure 2b), and proteins above the membrane MWCO are retained at the surface of the membrane. A pinch voltage (0.45 HV) is applied to the buffer and waste channels during this step to define the sample plug shape and to minimize diffusional loss of the proteins away from the membrane caused by the high concentration gradient. Finally, 100 V is applied to the waste reservoir to inject analytes for downstream analysis (Figure 2d), while the sample reservoir is allowed to float and the other reservoirs are grounded. In order to evaluate the performance of the sample preconcentration system, sample analyte concentration was measured near the membrane for various injection times and voltages. Images of the sample analytes stacking at the membrane surface were recorded and a spatially-averaged concentration factor was calculated by integrating over a measurement window placed near the membrane as shown in the inset of Figure 3. Since the protein is localized to a thin region adjacent to the membrane during the loading step, this measurement was performed after allowing the proteins to diffuse away from the membranes for a prescribed time interval (7 s). The integration window is a 21 by 21 pixel square whose size is comparable to the channel width—spatially-averaged concentration factors in this window are thus representative of the effective sample concentration observed downstream as compared to an electrokinetic injection alone without concentration. Further, allowing the sample to diffuse eliminated self-quenching and CCD saturation errors. The pixel intensity in the window was averaged and compared to a similar average before concentration. As will be noted later, local protein concentration values are greatly in excess of the spatially-averaged value. RESULTS AND DISCUSSION Membrane characteristics Mechanical and species-diffusion characteristics of the polymer membrane have been examined extensively. The molecular weight cutoff corresponding to the pore size of the membrane used in this work has been estimated to be below 5.7 kD, by demonstrating that insulin (5.7 kD) and a number of proteins do not cross the membrane while small tracer species (<0.5 kD) do.31 The membranes also maintain their properties upon exposure to a wide range of solvent conditions such as pH (2-9.3) and polarity (water, acetonitrile, C1-C3 alcohols). The membrane shown in Figure 1 has been used in more than 120 realizations over three months without any observable changes in properties. The zwitterionic and hydrophilic polymer membrane shows no measurable protein adhesion after flushing with water and isotonic phosphate-buffered saline. The membranes have been tested with differential pressures up to 1 bar, so vacuum pressure can be applied to manipulate fluids as part of the fabrication or testing processes. The electrical attributes of the membrane confirm that the charge-carrying ions in solution freely pass through the membrane and that no electric breakdown or ion depletion effects are present. The electrical resistances of equivalent channels with and without the laser-patterned membrane are equal within measurement accuracy (5%), consistent with the assertion that the membrane is permeable to molecules below 0.5 kD.31 Furthermore, currents through the system are temporally steady and linear with voltage. Electrophoretic concentration of proteins The mechanical and electrical properties of the membranes allow simple electrophoretic concentration and injection of proteins (Figure 2) to be performed with a single buffer as part of a pinched electrokinetic injection, as described in the experimental section. Electrophoresis stacks proteins against the semipermeable membrane while the charge-carrying ions pass essentially unimpeded. The negatively-charged proteins concentrate at the lower corner of the membrane (Figure 2c) because the lower corner is at a higher potential. By placing the membrane at an angle to the equipotential lines of the system and thereby inducing an asymmetric protein distribution, the majority of the protein sample is injected onto the running column (W in Figure 2d) and only a small amount is pulled back into the sample channel when voltages are switched. Protein retention at the membrane is minor. The spatially-averaged concentrations of both BSA and phosphorylase b observed in the measurement window can be increased by over two orders of magnitude (Figure 3). The maximum concentration factors demonstrated are about 130 for BSA and 160 for phosphorylase b. The injection time varied between 20-100 seconds for both cases. The voltage applied to the sample waste reservoir ranged from 15-70V for BSA and from 40-150V for phosphorylase b. The corresponding electric field strengths ranged from 7.5-35 (V/cm) for BSA and from 20-75 (V/cm) for phosphorylase b. The concentration increases linearly with the time-voltage product (Figure 3). Since the proteins are carried towards the membrane by electrophoresis and then retained at the membrane surface, the rate at which proteins are concentrated (i.e., the slope of the concentration factor vs. time-voltage product) relates directly to their electrophoretic mobility (?ep). The concentration rates inferred from the slopes of the linear fits to the BSA and phosphorylase b data (r2=0.92) are 16.7kV-1s-1 and 11.3kV-1s-1, respectively, leading to a ratio of the concentration rates of approximately 1.5. From an approximate calculation of the charge states of BSA and phosphorylase b in the experimental buffer systems,34 combined with the relation ?ep~q/Ms (0.33