Microfabricated Filters for Microfluidic Analytical Systems - Analytical

Journal of Physics: Conference Series 2006 34, 734-739 .... Trends of flow injection sample pretreatment approaching the new millennium. Zhao-Lun Fang...
0 downloads 0 Views 136KB Size
Anal. Chem. 1999, 71, 1464-1468

Microfabricated Filters for Microfluidic Analytical Systems Bing He, Li Tan, and Fred Regnier*

Department of Chemistry, Purdue University, Lafayette, Indiana 47907

Solvent and reagent filters were micromachined into quartz wafers using deep reactive ion etching to create a network of intersecting 1.5 × 10 µm channels. When placed at the bottom of reservoirs with a side exit, this channel network behaved as a lateral percolation filter composed of an array of cubelike structures one layer deep. Flow through these filters was driven by electroosmotic flow (EOF). Silanol groups at the walls of channels in the network provided the requisite charge to trigger EOF when voltage was applied laterally to the filter. Adsorption of cationic proteins in this silanol-rich matrix was controlled by the application of a polyacrylamide coating prepared by bonding N-hydroxysuccinimide (NHS)activated poly(acrylic acid) to (γ-aminopropyl)silane-derivatized filters. Subsequent reaction of residual NHS groups in the coating with 2-(2-aminoethoxy)ethanol provided channels of low charge density and adsorptivity. These lateral percolation filters were shown to be efficacious in filtering solvents containing a variety of particulate materials, ranging from dust to cells. The need to analyze increasingly smaller samples has stimulated great interest in microtechnology. Affinity biodot arrays on planar surfaces,1,2 microfabricated reaction vessels,3 capillary liquid chromatography and electrophoresis columns,4,5 and a variety of sensors ranging from microelectrochemical-6 and surface plasmon resonance-based devices to those exploiting a medley of waveguide technologies7-9 are all examples of efforts to accommodate smaller samples by miniaturization. An even more aggressive approach is to miniaturize and integrate all the components of the analytical system in a microchannel network where sample preparation, sampling, chemical reactions, separations, and detection are achieved in a (1) Fodor, S. P. A.; Read, J. L.; Pirrung, M. C.; Stryer, L.; Lu, A. T.; Solars, D. Science 1991, 251, 767. (2) Fodor, S. P. A.; Rava, R. P.; Huang, X. C.; Pease, A. C.; Holmes, C. P.; Adams, C. L. Nature 1993, 364, 555. (3) Colyer, C. L.; Tang, T.; Chiem, N.; Harrison, D. J. Electrophoresis 1997, 18, 1733. (4) Ocvirk, G.; Verpoorte, E.; Manz, A.; Grasserbauer, M.; Widmer, H. M. Anal. Methods Instrum. 1995, 2-2, 74. (5) Bruin, G. J. M.; Paulus, A. Anal. Methods Instrum. 1995, 2, 3. (6) Clark, R. P.; Hietpas, P. B.; Ewing, A. G. Anal. Chem. 1997, 69, 259. (7) Lin, V. S.-Y.; Motesharei, K.; Dancil, K.-P. S.; Sailor, M. J.; Ghadiri. M. R. Science 1997, 278, 840. (8) Schipper, E. F.; Kooyman, R. P. H.; Heideman, R. G.; Greve, J. Sens. Actuators 1995, B24-25, 90. (9) Weigl, B. H.; Lehmann, H.; Lippitsch, M. E. Sens. Actuators 1996, B32, 175.

1464 Analytical Chemistry, Vol. 71, No. 7, April 1, 1999

single device.10,11 Among the more appealing features of this strategy are that (i) all the unit operations are integrated, (ii) reagent consumption will be very low, (iii) sample volume will be small, (iv) analyte recovery would be maximized, (v) contamination would be minimized, and (vi) many systems could be fabricated and operated in parallel. For this concept to be realized on a wide scale, it will be necessary to miniaturize the whole analytical system. This paper examines the issue of filtration in these microfluidic systems. At present, lithographic,12 embossing, and casting processes using a micromachined mask or molds13,14 are the dominant technologies used to create the 1-100-µm objects and channels on which these integrated systems are based. The fact that microchannels of less than 20-30 µm are easily blocked by particles is a problem. These particles may be transported into the system in solvents, crystallize from samples while they are being held in reservoirs on chips, or arise from microbial growth in buffers during storage. Prefiltration would be helpful, but conventional filtration techniques require orders of magnitude greater volumes than used in sample and solvent reservoirs on chips. This approach also fails to address on-chip particulate formation. Microfiltration within the device would be a much better solution. Microfabricated filters have been described for trapping different cell types from blood,15 but the objective was to harvest cells, not prepare particle-free solvents and samples. These filters were made by microfabricating arrays of rectangular, parallel channels on chips of a width and height that would not allow particles larger than the channels to enter the channel network along the axis parallel to the chip surface. This type of filter is similar to the frited glass or membrane filter devices widely used in laboratories to harvest particulate materials. Filtration of this type will be referred to as axial percolation filtration because it occurs by percolating liquid through a filter bed along the flow axis. Axial percolation filters are generally of high cross-sectional area at the point of filtration to provide many parallel channels (10) Terry, S. C.; Jerman, J. H.; Angell, J. B. IEEE Trans. Electron. Dev. 1979, ED-26 (12), 1880. (11) Manz, A.; Miyahara, Y.; Miura, J.; Watanabe, Y.; Miyagi, H.; Sato, K. Sens. Actuators, 1990, B1, 249. (12) Thompson, L. F.; Wilson, C. G.; Bowden, M. J. Introduction to Microlithography, 2nd ed.; American Chemical Society: Washington DC, 1994. (13) Manz, A.; Harrison, D. J.; Verpoote, E. M. J.; Fettinger, J. C.; Paulas, A.; Ludi, H.; Widmer H. M. J. Chromatogr. 1992, 593, 253. (14) McCormick, R. M.; Nelson, R. J.; Alonso-Amigo, R. J.; Benvegnu, D. J.; Hooper, H. H. Anal. Chem. 1997, 69, 2626. (15) Wilding, P.; Kricka, L. J. Mesoscale Sperm Handling Devices. U.S. Patent 05427946, 1995. 10.1021/ac981010+ CCC: $18.00

© 1999 American Chemical Society Published on Web 02/24/1999

and accommodate the fact that large numbers of channels in the filter will be plugged during the filtration process. Filtration may also be achieved by lateral percolation, as occurs commonly in nature. In lateral percolation filtration, liquid enters the filter bed along the plane or face of high cross-sectional area and then migrates at a right angle through the filter bed, i.e., lateral to the point of entry. Filtrate is harvested at some point along the side of the bed. Lateral percolation filter beds in the laboratory generally have a minimum depth of at least a few hundred particles. There are no reports of the microfabrication of lateral percolation filters. This paper addresses the problem of filtering liquids in micromachined capillary systems by lateral percolation where liquids are transported by electroosmotic flow. The filter system described in this paper should apply broadly to many analytical systems on chips including those cases where liquids are driven by pressure. EXPERIMENTAL SECTION Materials. Photolithography masks, SL-4006-2C-AR3-AZ1350, and 3-in.-diameter quartz wafers, QZ-3W40-225-UP, were purchased from Hoya Corp. (Shelton, CN). Poly(acrylic acid) (MW 450 000), N-hydroxysuccinimide (NHS), (3-aminopropyl)triethoxysilane, 2-(2-aminoethoxy)ethanol, and dicyclohexylcarbodiimide (DCC) were purchased from Aldrich Chemical Co. (Milwaukee, WI). Lysozyme (chicken egg), cytochrome c (bovine heart), -Rchymotrypsinogen A (bovine pancreas), ribonuclease A (bovine pancreas), myoglobin (horse heart), and conalbumin (chicken egg white) were products of Sigma Chemical Co. (St. Louis, MO). All chemicals were used without further purification. All buffers and solutions were degassed with vacuum for 30 min to remove air trapped in solutions. Mask Fabrication. Filter chip layout was carried out on a Sun Sparc workstation (Sun Microsystem, Cupertino, CA) using ICStation of Mentor Graphics (Wilsonville, OR) at the Purdue University Solid State Laboratory. “Layout” files in the IC-Graphics format were then converted into GDS format, electronically transferred to the fabrication laboratory. The GDS file was then compiled as motion control file to generate the photomask using e-beam lithography. Filter Fabrication. The filter well microchip was fabricated in Alberta Microelectronic Center (Edmonton, AB, Canada) using procedures previously described.16,17 Photomicrography. A Nikon Inverted Eclipse TE-300 optical microscope was used to observe both the filters and filtered particles. Tungsten light illumination was used without neutral density intensity control to achieve the best imaging. A Nikon n90s 35-mm camera, attached to the front port, was used to record the image of the filter well and cells. The camera was set to manual focus, single-frame shooting, aperture priority, and centerweighted exposure mode with (2/3 stop bracketing from +0.7 over the center-weighted meter reading to get the best image. Spot meter exposure, focused on the etched filter supports, was used for recording the filter well images with a dark background. Glycine max var. Kent cells (soybean) prepared in w-38 Murashagie & Skoog media were loaded into the filter well and (16) He, B.; Regnier, F. E. Microfabricated Liquid Chromatography Columns Based on Collocated Monolith Support Structures. In press. (17) He, B.; Tait, N.; Regnier, F. E. Anal. Chem. 1998, 70, 3790.

allowed 5 min to settle. Images were taken at different focusing planes for high magnification. The images of human KB cells, prepared in Engle’s MEM with nonessential amino acids, Earle’s BSS (90%), and fetal bovine serum (10%) were taken using a similar procedure. Escherichia coli cells were prepared in Lucia burtania media containing tryptone, yeast extract, and NaCl. The 5-µm silica particles, Lichrosorb Si-100, were the product of E. Merck (Cincinnati, OH). Derivatization with (3-Aminopropyl)silane. Derivatization of internal surfaces in the chip with (3-aminopropyl)silane was achieved as previously described.16,17 Synthesis of Activated Polymer. Poly(acrylic acid) (0.2615 g, 3.63 mmol) and NHS (0.8360 g, 7.26 mmol) were placed in a dry flask and dissolved in 10 mL of DMSO. DCC (3.000 g, 14.54 mmol) in 10 mL of DMSO was then slowly added to this solution. The reaction was carried out at room temperature for 3 h. Dicyclohexylurea was removed by filtration, and the filtrate containing the activated polymer was collected. This solution was used directly to derivatize the (3-aminopropyl)silane-activated chip. Application of the Polymer Coating. Activated polymer was introduced into (3-aminopropyl)silane-activated filter wells and maintained at room temperature for 12 and 36 h, respectively. The filter wells were then flushed periodically with 2-(2-aminoethoxy)ethanol in chloroform (50%, v/v) during a 12-h reaction time at room temperature. Unreacted reagents were then washed from the filter wells with chloroform. Finally, the filter wells were washed with 10 mM phosphate buffer (pH 7.0).

RESULTS AND DISCUSSION The objective of this work was to design micromachinable filters that could be fabricated in situ on silicon and quartz wafers during the production of microfluidic analytical devices. The purpose of these filters was to prevent particles down to a few micrometers in size from entering the analytical device. Filter Architecture. On the basis of the fact that particles generally enter microfluidic analytical systems through solvent or sample reservoirs, filters were placed in the reservoirs. In an effort to increase the cross-sectional area of the filter without substantially changing the geometry of the system, a large portion of the bottom of the reservoir was converted into a lateral percolation filter. Filtrate was harvested at the side across the entire width of the reservoir to diminish the prospect that blocking a few channels would disable the whole filter. It has recently been shown that “beds” of particle-like structures and channels may be fabricated in situ and used to produce chromatography columns.16,17 Similar structures and fabrication procedures were used in this work.The design concept used to fabricate a lateral percolation filter at the bottom of a reservoir was as follows. First, a hole was either sonically or laser drilled through the top wafer to form the reservoir. Reservoirs down to 1 nL in volume can be formed by laser drilling a 50-µm-diameter hole through a 400-µm-thick cover plate. Second, this reservoir was positioned directly above the micromachined bed of the filter (Figure 1A). The channel width between filter elements used in these filters was 1.5 µm with a channel depth of roughly 10 µm in all cases. Particle filtration, percolation of liquid into the channel network of the bed, and lateral flow of liquid in the bed are illustrated in Figure 2. Obviously, the region of highest liquid flux Analytical Chemistry, Vol. 71, No. 7, April 1, 1999

1465

Figure 1. The microfabricated filter. (A) Photomicrograph of the filter. Access holes are in the cover wafer, while the bright area is the microfabricated filter bed. The large channel on the right is for connection to other microfluidic components. (B) Closeup of the filter showing extension of the microfabricated filter structures underneath the cover plate to enhance the filtration capability.

Figure 2. Lateral percolation in microfabricated filter. It is seen that as liquid vertically enters an array of microfabricated cubes attached to an underlying substrate and is drawn laterally to the sides of the array through an interconnecting channel network, particles larger than the 1.5-µm channels separating the cubes will be excluded, similar to axial filters. It should also be noted that as the entrance to a channel is blocked by a particle, liquid is still capable of flowing under the particle and multiple alternative routes are still available for liquid to enter the bed and migrate through the filter. The significance of this is that the illustrated lateral percolation filter will have a specific filter capacity (particles filtered/cm2 at the filter face) equivalent to high-capacity axial filtration systems.

will be at the point where the reservoir wall meets the filter bed. It is important that the filter bed be slightly larger than the bottom of the reservoir so that lateral flow may occur in multiple directions. If or when particulates accumulate near the reservoir wall, liquid will flow laterally through the bed from further away to exit the reservoir. It is this lateral flow from further out in the filter bed that increases the filter capacity and allows a large portion of the bottom of the reservoir to be used as a filter. Another feature designed into these filters was the requirement that filtrate must pass through a tortuous channel network during migration toward the filter exit. This discriminates against nonspherical particles. Although asymmetrically shaped particles may be able to enter a channel along its longitudinal axis, they are more likely to be trapped at turns in the channel system. In this regard, lateral percolation filters have traits of an axial filtration bed. A channel network with 90° angles was used in this work because it was most easily fabricated. Although it would be possible to fabricate even higher degrees of channel tortuosity, 1466 Analytical Chemistry, Vol. 71, No. 7, April 1, 1999

this variable was not examined. Extending the bed beyond the exit from the reservoir in an enclosed channel (Figure 1B) further enhanced filtration. Particulates were removed from the microfabricated filters described below by extensive washing of reservoirs from the top, as is done with fritted glass filters. However, a capacity for backflushing was also incorporated into the design. This was achieved by placing an inlet channel at the bottom of the reservoir such that liquid could be aspirated into chambers from the side and create turbulent flow across the face of the filter as liquid exited out the top of the reservoir. Fabrication of Filters. Forming the channels in the filter shown in Figure 1 and 1B is best achieved with anisotropic gasphase etching. Anisotropic etching is generally directed vertically into the surface of the wafer and achieves aspect ratios of 30/1 or greater. This is critical in the construction of microfilters for several reasons. One is that channel width may be precisely selected and maintained during etching. The second is that width is essentially independent of depth in anisotropically etched channels. Channel depth, i.e., depth of the filter bed, generally ranged from 8 to 10 µm in quartz wafers. It is difficult to etch deeper in quartz with current reactive ion etching technology.17 Beds with deeper channels were fabricated in silicon wafers but were not examined as filters. Reservoir fabrication may be achieved in two ways. The most obvious is to machine the reservoir in the cover wafer before it is bonded to the lower filter bed wafer. The only difficulty with this predrilling approach is that alignment can vary in the range of (10 µm as a result of wafer migration during bonding. This problem could be circumvented by drilling the reservoir after bonding, but unfortunately, the drilling process is so harsh that the delicate filter network is damaged when the drill breaks through the cover wafer. For this reason, all filters were fabricated using predrilled reservoirs. Surface Pacivation. Because these filters were fabricated in quartz, the walls of the channel network were rich in surface silanols. Silanols are well-known to adsorb cations18 and a wide (18) Cheng, K.; Zhao, Z.; Lamb, J. D. J. Chromatogr. 1995, 706, 517. (19) VanOrmann, B.; McIntire, G. J. Microcolumn Sep. 1989, 1, 289.

Figure 3. Surface pacification reactions. It should be noted that polymeric structures I-V are represented by the structure of the monomer unit in the polymer instead of the more common structural formula notation used to describe polymers.

variety of polymers,19 particularly proteins.20 This presents a significant problem when dealing with biological samples, in terms of both sample loss and alteration of the ζ potential at surfaces within the filter. Reduction of ζ potential by adsorbed polymers decreases the rate of electroosmotic flow in microchannels. One solution to this problem in capillary electrophoresis has been to apply a surface coating that controls surface fouling and electroomotic flow.21 A similar strategy was used in this work. The objective was to control protein adsorption without eliminating electroosmotic flow. Application of a covalently bonded organic coating with a low carboxyl content did this. It has been demonstrated in capillary zone electrophoresis that weakly cationic coatings can be prepared that adsorb neither cationic nor anionic proteins while still having sufficient ζ potential to provide electroosmotic pumping.22 Surface bonded polyacrylates with various functionalities have been shown to be of broad efficacy in capillary electrophoresis for the control of both protein adsorption and electroosmotic flow.23,24 Polyacrylates have been widely applied to surfaces in microchannel systems by in situ polymerization.23-25 Unfortunately, it is difficult to characterize covalently bonded polymers and control thickness when the coating is formed at the surface by in situ polymerization. The strategy used here was to coat the surface with a preformed acrylate polymer of known molecular weight distribution. NHS-derivatized polyacrylate II was synthesized by activation of 2-300-kDa poly(acrylic acid) I (Figure 3). Based on NMR and elemental analysis, carboxyl derivatization in these polymers was nearly complete (data not shown). NHSactivated poly(acrylic acid) (II) was then coupled through amide bond formation to the quartz surface of filter wells that had been derivatized with (γ-aminopropyl)silane (III). NHS-activated polymers obtained by both polymerization of activated polymer and activation of the preformed polymer were used with no apparent difference. The results reported below were obtained with preformed polymer. It was reasoned that residual NHS activated (20) (21) (22) (23) (24) (25)

Zhao, Z.; Malik, A.; Lee, M. L. Anal. Chem. 1993, 65, 2747. Towns, J. K.; Bao, J.; Regnier, F. E. J. Chromatogr. 1992, 599, 227. Mechref, Y.; ElRassi, Z. Electrophoresis 1995, 16, 617. Cobb, K. A.; Dolnik, V.; Novotny, M. Anal. Chem. 1990, 62, 2478. Hjerten, S. J. Chromatogr. 1985, 347, 191. Nakatani, M.; Shibukawa, A.; Nakagawa, T. Electrophoresis 1994, 15, 177.

carboxyl groups would probably reside in the coating IV after the polyacrylate was bonded. Residual NHS-activated carboxyl groups were derivatized with 2-(2-aminoethoxy)ethanol to form the fully coated surface V. 2-(2-Aminoethoxy)ethanol was chosen on the basis of observations that in the case of poly(acrylic acid) it produces a very hydrophilic coating and that simple amides derived from ammonia and aminoethanol are hydrolytically less stable than the 2-(2-aminoethoxy)ethanol derivative.26 Coating Characterization. Microfabricated filters are so small that it is very difficult to characterize the coating directly. For this reason, the coating was characterized indirectly; i.e., it was applied to other silanol-rich surfaces of higher surface area that were more amenable to characterization. The assumption in this indirect characterization strategy is that the coating would be generally the same on all silanol-rich surfaces. Although it is recognized that not all surfaces are silylated with equal efficacy, it will be shown below that the number of functional groups in the bonded polymers was so much larger than the density of organosilanes that differences in silylation efficiency would play little role in the characteristics of the coating. This conclusion is based on the fact that preformed polymeric ion-exchange coatings applied to both porous polystyrene and silica supports through different coupling chemistries have been shown to have identical chromatographic characteristics.27 This study shows that when surfaces are coated with large polymers they loose their original characteristics and take on those of the polymer. The very high surface area of porous silica facilitates elemental analysis of organic coatings. Elemental analysis of a 1000-Å pore diameter (30 m2/g) silica support to which III had been applied under the same conditions used to derivatize the filter showed an organosilane density of 4 µmol/m2. Although silica is generally thought to have a silanol density of 8 µmol/m2,28 the bonding density of an organosilane monolayer is thought to be in the range of 6 µmol/m2.29 This leads to the conclusion that surface silylation in this study approached a monolayer. After coupling NHSactivated poly(acrylic acid) II to the support in a 12-h reaction, elemental analysis showed the acrylate residue density in the polymeric surface coating IV to be 70 µmol/m2. Extending the reaction time to 36 h increased the density of acrylate monomer residues to 350 µmol/m2 in the polyacrylate coating IV. This means that the number of acrylate groups in the coating exceeds the number of alkylamine groups on the surface by ratios of 17/1 and 88/1, respectively, and that there is a large number of residual NHS-activated carboxyl groups in the coating IV. Assuming that a 6 µmol/m2 organosilane film is a monolayer,29 it may be concluded that polymer molecules in the 70 and 350 µmol/m2 functional group density films are not laying flat on the surface. The final step in the preparation of the surface layer V was to add 2-(2-aminoethoxy)ethanol. Attempts to totally derivatize all residual NHS and carboxyl groups with 2-(2-aminoethoxy)ethanol using DCC were not successful based on the magnitude of electroosmotic flow in electrophoretic evaluation. There are apparently a small number of residual carboxyl groups in all coatings (26) Chiari, M.; Micheletti, C.; Fazio, M.; Righetti, P. G. Electrophoresis 1995, 16, 1451. (27) Vondruska, M.; Sudrich, M.; Mladek, M. J. Chromatogr. 1976, 116, 457. (28) Snyder, L. R.; Kirkland, J. J. Introduction to Modern Liquid Chromatography; John Wiley & Sons: New York, 1974. (29) Berthod, A. J. Chromatogr. 1991, 549, 1.

Analytical Chemistry, Vol. 71, No. 7, April 1, 1999

1467

Figure 4. Particle capture by the filter: (A) 5-µm silica particles trapped by the microfabricated filter; (B) accumulation of particles and other buffer residue trapped by the filter; (C) capture of large eukaryotic plant soybean cells on the filter; (D) accumulation of human KB cancer cells on the surface of the filter; (E) capture of E. coli cells.

prepared by this synthetic route that resist derivatization. This fact was exploited to produce electroosmotic flow in coated filters. Electrophoretic examination of 75-µm fused-silica capillaries to which the polyacryly(2-(2-aminoethoxy)ethanol) coating V had been applied showed that from pH 5-10 the surface still had a net negative charge but that electroosmotic flow (EOF) had been reduced more than 80% at pH 7. Although part of the remaining charge may derive from residual silanol groups at the surface, it is more likely due to underivatized carboxyl groups in the coating. Actually a small amount of EOF is necessary to transport analytes out of the reservoir. Variation in EOF with this coating did not exceed 10% in four trials. This was deemed to be acceptable for analytical systems in which the rate of sample transport to the next step in the analytical process is not a critical issue. In those cases where it might be critical, further optimization of the coating process would probably reduce variations in EOF. All proteins in a test mixture consisting of lysozyme, cytochrome c, ribonuclease A, R-chymotrypsinogen, myoglobin, and conalbumin were recovered at pH 7. It is assumed that capillaries in the microfabricated filter are similar to the open tubular capillaries used in the electrophoretic evaluation. Operation. Filter efficacy was examined with three types of particles: sized inorganic particles, dust, and cells. The best imaging of particles trapped on filters was achieved using tungsten light microscopy. Filtration of 5-µm silica particles is seen in Figure 4A. Based on the fact that particle diameter is more than 3 times the average channel diameter, the efficiency of the filter is not surprising. Unfiltered solutions dispensed into the solvent wells on the chip from laboratory glassware provided a more difficult test. After a polyacrylamide-coated V filter was used for a few days, particle accumulation was seen on the face of the filter (Figure 4B). These particles are presumably from particle-contaminated

1468 Analytical Chemistry, Vol. 71, No. 7, April 1, 1999

glassware. Perhaps more significant is the fact that no particles are seen in the channel network beyond the bottom of the filter. Particles were clearly trapped totally or not at all. Capture of large eukaryotic plant and animal cells and even the smaller prokaryotic E. coli was possible (Figure 4C-E). Although individual cells of E. coli could partially penetrate the channels, channel tortuosity seemed to be a major aid in their capture. CONCLUSIONS It may be concluded that lateral percolation filters composed of 1.5 × 10 µm channels intersecting at a 90° angle may be micromachined into quartz wafers in situ. When placed at the bottom of solvent or reagent reservoirs ranging down to 1 nL in volume, it is possible to filter cells and dust from liquid entering the channel network on the wafer. It is further concluded that surface coatings play an important role in diminishing surface fouling, enhancing the recovery of cationic proteins, and controlling electroosmosis in these silica-based filters. Polyacrylamide coatings reduced electroosmotic flow more than 80% while increasing the recovery of positively charged proteins. ACKNOWLEDGMENT The authors gratefully acknowledge the gift of soybean cells and human KB cells from Ms. Ann Taylor, Mr. Zhenfan Yang, and Prof. Phillip Low of the Purdue Chemistry Department, and the gift of E. coli cells from Anissa Buckner and Prof. Minou Bina of the same department. This research was funded by NIH Grant 35421 and PerSeptive Biosystems. Received for review September 10, 1998. Accepted January 8, 1999. AC981010+