Label-Free Detection of Heparin, Streptavidin, and Other Probes by

Jul 31, 2008 - File failed to load: https://cdn.mathjax.org/mathjax/contrib/a11y/accessibility-menu.js. ADVERTISEMENT · Log In Register · Cart · ACS ·...
0 downloads 0 Views 448KB Size
Download PDF
Anal. Chem. 2008, 80, 6532–6536

Label-Free Detection of Heparin, Streptavidin, and Other Probes by Pulsed Streaming Potentials in Plastic Microfluidic Channels Qiaosheng Pu, Marwa S. Elazazy, and Julio C. Alvarez* Virginia Commonwealth University, P.O. Box 842006, Richmond, Virginia 23284 In this paper, pulsed streaming potentials generated in plastic microfluidic channels are used for the label-free detection of some model analytes. The microchannels are fabricated with the commodity plastic cyclic olefin copolymer (COC), and the detection signal arises from a change in the surface charge upon analyte adsorption on the modified microchannel surface. The role of the surface modification is to confer the microchannel with a predetermined charge and a particular specificity toward the adsorption of the target analyte. In this work, several target probes displaying different levels of specificity were investigated. Heparin and streptavidin were detected by adsorption on microchannel surfaces modified with protamine and biotin, respectively, whereas bovine serum albumin (BSA) and methylene blue (MB) showed nonspecific adsorption on almost any modified or unmodified COC microchannel surface. The magnitude of the streaming potential was found to be proportional to the liquid pressure and the surface charge of the microchannel in accord with the Smoluchowski equation. Because the relative polarity of the streaming potential is determined by the surface charge, the most straightforward detection with this method occurs when the charge is reversed upon analyte adsorption. This strategy was used for the species described in this work, and the lowest concentrations detected were ∼0.01 units/mL for heparin (below clinical relevance), ∼10-9 M for BSA, and ∼10-6 M for MB. Unlike the conventional method of steady flow, in this work, the streaming potentials were measured under pulsed conditions of flow and using nonreference electrodes. This approach removes the need of special electrolytes as it is usually required when using reference electrodes, and at the same time, it mitigates the interference of electrochemical drift from the electrodes. Relative standard deviations of ∼1-2% and measuring times of ∼10 s are readily attained with this experimental setup. The on-channel modification of the surface was carried out by UV-photografting methods given the significant UV transparency of COC. This paper reports on the label-free detection of some model probes using pulsed streaming potentials in plastic microfluidic * To whom correspondence should be addressed. Fax: 804-828-8599. E-mail: [email protected].

6532

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

Scheme 1. (A) Generation of a Streaming Potential ∆E as a Result of Pressure-Driven Flow and Surface Charge at the EDL, (B) Detection of Bioprobes by Change of Surface Charge upon Adsorption, (C) Signal ∆E Obtained Using Pulsed Flow Driven by Suction, and (D) Measurement of ∆E in a Microfluidic Channel

channels built with the commodity thermoplastic cyclic olefin copolymer (COC). Streaming potentials are charge gradients spontaneously generated along the flow direction when liquids are pumped through a microhannel (Scheme 1A).1–9 These gradients, which are proportional to the liquid pressure and the surface charge of the channel, are measurable as a potential difference ∆E between two reference electrodes at the outlets of the microchannel (Scheme 1A-C).1–9 The forward motion of ions at the microscopic interface between the channel surface and the flowing liquid, usually known as electric double layer (EDL), is what causes ∆E.1–7 In this work, the analytical signal ∆E arises when the target analyte induces a change in the surface charge (1) Bellmann, C.; Klinger, C.; Opfermann, A.; Bohme, F.; Adler, H.-J. P. Prog. Org. Coat. 2002, 44, 93–98. (2) Fa, K.; Paruchuri, V. K.; Brown, S. C.; Moudgil, B. M.; Miller, J. D. Phys. Chem. Chem. Phys. 2005, 7, 678–684. (3) Hunter, R. J. Zeta Potential in Colloid Science: Principles and Applications; Academic Press: London, 1981; p 381. (4) Kirby, B. J.; Hasselbrink, E. F. Electrophoresis 2004, 25, 187–202. (5) Schweiss, R.; Welzel, P. B.; Werner, C.; Knoll, W. Langmuir 2001, 17, 4304–4311. (6) Smoluchowski, M. Z. Phys. Chem. 1918, 93, 129–131. (7) Scales, P. J.; Grieser, F.; Healy, T. W. Langmuir 1992, 8, 965–974. (8) Gusev, I.; Horvath, C. J. Chromatogr. A 2002, 948, 203–223. (9) Mattiasson, B.; Miyabayashi, A. Anal. Chim. Acta 1988, 213, 78–89. 10.1021/ac8003117 CCC: $40.75  2008 American Chemical Society Published on Web 07/31/2008

upon adsorption (Scheme 1B,C). The easiest detection happens when the charge of the surface is opposite to the charge of the unbound analyte because the polarity of ∆E is determined by the surface charge of the microchannel. The most significant aspect of this work is that by pumping a liquid through a microfluidic channel, label-free detection of bioprobes with adequate precision and accuracy can be attained in a matter of seconds using very simple and inexpensive instrumentation. Besides associating changes in surface charge with surface adsorption events to generate an analytical signal, the major aspect that makes this work unique is that the streaming potentials are measured under pulsed conditions of flow and using nonreference electrodes (Scheme 1C). This approach mitigates the interference of electrochemical drift from the electrodes and at the same time removes the need of special electrolytes as it would be required for reference electrodes. Although streaming potentials have been known for about a hundred years and have been mostly used to determine the charge of surfaces, to the best of our knowledge, no work has been done to employ this principle for sensing purposes in microfluidic channels.9,10 Here, we illustrate the proof of concept to detect model analytes displaying different degrees of adsorption specificity on modified surfaces. The role of the surface modification is to confer the microchannel with a predetermined charge and a particular specificity toward the adsorption of the target analyte (Scheme 1B). Heparin and streptavidin adsorbed only on COC microchannels modified with protamine and biotin, respectively, whereas bovine serum albumin (BSA) and methylene blue (MB) adsorbed nonspecifically on any modified or unmodified COC microchannel. This work is an extension of our previous efforts on making COC microchannels with different surface modifications11 and the implementation of electrochemical detection in microfluidic channels.12 The advantages that make the present approach unique relative to previous works are as follows:9,10 (i) The signal is a result of the pressure-driven flow, the adsorption affinity, and the charge on the channel surface; therefore, this method is ideally suited for COC microfluidic channels because their surface charge and adsorption affinity can be easily controlled by photografting methods.11 (ii) For detection, the analyte does not need any labeling (fluorescent, electrochemical, or radioactive) using typically tedious procedures. (iii) Because the signal acquisition is made using pulsed flow instead of the conventional steady-state conditions, accurate measurements with adequate precision can be recorded in a matter of seconds using nonreference electrodes. (iv) For irreversible adsorption (i.e., biotin-streptavidin), ∆E can be recorded with analyte-free solutions in the best conditions (pH, conductivity, etc.) to maximize the charge variation during adsorption and avoid unwanted contributions to ∆E from the sample and analyte. (v) Finally, because the instrumentation for these measurements comprises a regular vacuum pump, a high resistance voltmeter, a programmable valve, and a pressure holder for the microchannel (Scheme 1D), this detection method could lead to portable and cost-effective biosensing devices. (10) Miyabayashi, A.; Mattiasson, B. Anal. Biochem. 1990, 184, 165–171. (11) Pu, Q.; Oyesanya, O.; Thompson, B.; Liu, S.; Alvarez, J. C. Langmuir 2007, 23, 1577–1583. (12) Khalid, I. M.; Pu, Q.; Alvarez, J. C. Angew. Chem., Int. Ed. 2006, 45, 5829– 5832.

Typically, streaming potentials are measured under carefully controlled conditions of steady flow, solution conductivity, pH, using reference electrodes.1,3,5,8,13 Reference electrodes help to minimize the drift caused by electrode polarization or passivation, but specific redox species in solution are needed to stabilize their redox potential.14 Moreover, even under steady-state flow, a stable response for ∆E may take considerable time (>10 min). To avoid this limitation, we have used nonreference platinum electrodes while implementing a programmable valving system to pulse the solution flow. A valve connected to a vacuum pump and controlled by computer is used to switch the suction ON and OFF intermittently (Scheme 1D) and to produce a pulsed flow within the microfluidic channel. When the valve is in the OFF position, no suction is applied and the flow through the channel is negligible. Conversely, the ON position allows a vacuum to be applied with suction pressures up to 90 kPa and dwell times of 2 s or higher (see the Supporting Information). The streaming potential is taken as the difference in ∆E between the ON and the OFF states (see the Supporting Information). The combination of microchannels and solutions employed in this work produced resistances of about 108 Ω; therefore, we used a homemade voltage follower with an input impedance of 1012 Ω.15,16 This guaranteed that no current was drawn from the Pt electrodes while measuring ∆E (see the Supporting Information). Pulsed streaming potential values with relative standard deviations of ∼0.8 -2% are attainable in a matter of seconds despite drift in the baseline (see the Supporting Information).

∆E )

Pζεεο ηκ

(1)

The magnitude of ∆E is linearly proportional to the liquid pressure (P) following the Smoluchowski equation (eq 1), where η is the viscosity of the solution,  is the dielectric constant, o is the vacuum permittivity, κ is the conductivity through the channel, and ζ is the ζ-potential.1,3,4,6 The relative polarity and value of ∆E are also proportional to the sign and magnitude of the ζ-potential ζ, which is defined as the potential at the slippage plane of the liquid, and it really represents the charge density at this location.3 Adsorption of analytes on the microchannel surface can completely reverse the polarity of ∆E or the sign of ζ with respect to the underlying surface (Scheme 1B). Therefore, the ideal sensing surface is the one that contains adsorption sites for the analyte and at the same furnishes a charge inversion upon analyte adsorption. To illustrate this principle on a COC microchannel and test eq 1 under pulsed conditions, we selected BSA, heparin, and MB as target analytes. The biontin-streptavidin couple was used to demonstrate the detection with analyte-free solutions and maximize the charge variation during adsorption (see below). BSA and MB are known for adsorbing nonspecifically to virtually any surface in a wide range of pH values, while heparin can bind (13) Toshihiko, J.; Mitsuru, H.; Minoura, N.; Akihiko, T. Macromolecules 1998, 1998, 1277–1284. (14) Bard, A. J.; Faulkner, L. R. Electrochemical Methods Fundamentals and Applications, 2nd ed.; John Wiley & Sons: New York, 2001. (15) Caceci, M. S. J. Chem. Educ. 1984, 61, 935–936. (16) Holewinski, P. K. J. Chem. Educ. 1984, 61, 1108–1109.

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

6533

Figure 1. Pulsed streaming potentials for a COC-R3N-modified channel at different [BSA] in 1.5 mM buffer phosphate, pH 7.0. ∆E was recorded with BSA-free solution after flushing each [BSA] for 10 min at 24 µL/min.

surfaces containing specific receptors like protamine.17 Both BSA and heparin are negatively charged at neutral pH, while MB is a cationic dye. EXPERIMENTAL SECTION Chemicals. COC resin (Topas 8007 × 10, with the glass transition temperature of 80 °C) purchased from Ticona (Florence, KY) was used for fabricating the microfluidic chips. BSA and hexadecyltrimethylammonium bromide (CTAB) were purchased from Sigma; N-[3-(Dimethylamino)propyl]methacrylamide(DPMA)and[2-(methacryloyloxy)ethyl]-trimethylammonium chloride (75 wt % solution in water) (META) were from Aldrich; acrylamide, benzophenone, MB, phosphoric acid (85%), and sodium chloride were from Acros; acrylic acid was from Fluka; dibasic sodium phosphate and monobasic sodium phosphate were from EM Chemicals (Gibbstown, NJ). Eighteen MΩ DI water obtained from a Milli-Q purification system (Millipore, Billerica, MA) was used for all solution preparation. Fabrication of Microfluidic Chips. The fabrication of COC microchannels and the on-chip surface modification were done following procedures reported previously by our group.11 Details of the instrumentation built in our laboratory for measuring streaming potentials and the conditions for the experiments shown here are presented in the Supporting Information. RESULTS AND DISCUSSION Figure 1 shows a series of Smoluchowski plots (∆E vs P), using a COC semicircular microchannel (127 µm wide × 50 µm deep) modified with a polyacrylic film containing pendant tertiary amine groups (COC-R3N).11 The microfluidic channels were modified by UV-photografting taking advantage of the significant UVtransparency of COC.11 A 1.5 mM buffered solution of each [BSA] was flushed to allow BSA adsorption for 10 min. Afterward, ∆E was measured at increasing values of suction pressure (from 14 to 89 kPa) using 1.5 mM BSA-free buffer. Each point in Figure 1 was recorded as a pressure pulse starting at a state of no suction (0 kPa), then switching on the suction for 2 s, and finally returning to 0 kPa. The linearity (r2 ∼ 0.99) and nearly zero value of intercepts observed for the different concentrations of BSA in Figure 1 corroborate the streaming potential behavior and the (17) Milovic, N. M.; Behr, J. R.; Godin, M.; Hou, C.-S. J.; Payer, C. R.; Chandrasekaran, A.; Russo, P. R.; Sasisekharan, R.; Manalis, S. R. Proc. Natl. Acad. Sci. 2006, 103, 13374–13379.

6534

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

agreement with eq 1 under pulsed conditions, despite using nonreference electrodes. At all pressure values, the highest positive ∆E was observed when no BSA was present in solution, indicating that the COC-R3N surface has a positive charge density at the EDL region. A gradual increase in the BSA concentration (from 3 to 20 µg/mL) shows a steady reduction in ∆E until it eventually reaches its minimum negative value at the highest BSA concentration used (20 µg/mL). This effect is caused by the adsorption of the anionic BSA on the polyacrylic film, which gradually reverses the positive charge density of COC-R3N into a negative value (Scheme 1B). The adsorption of BSA on bare and modified COC was confirmed by fluorescence imaging using fluorolabeled BSA.11 Despite the adsorption on both bare COC (negative ζ) and COC-R3N (positive ζ), only the latter surface is suitable for BSA detection because ζ is reversed during BSA adsorption. Similarly, the detection of the cationic MB is maximized by using a negative surface (see the Supporting Information). Limits of detection were about 1 µM for MB and 1 µg/mL (∼1 nM) for BSA. Fits to Langmuirian and exponential isotherms were observed for MB and BSA, respectively, including the calculation of adsorption constants (see the Supporting Information). Heparin is a glycosaminoglycan widely used as an injectable anticoagulant whose detection is critical during cardiopulmonary bypass surgery (CPB) because high heparin doses can lead to bleeding problems in patients.18,19 Accurate and quick monitoring of blood heparin can help in finding the minimal protamine dose to neutralize the anticoagulant activity of heparin at the end of the CPB.18,19 The clinical significance of this is reflected in the many reports available for sensing heparin17,20–23 and because protamine overdose is known to cause complications and toxic effects.18,19 Figure 2 shows the adsorption isotherm for heparin on a COC microchannel modified with protamine. Protamine adsorbs spontaneously on COC almost irreversibly (Figure S8, Supporting Information), so after each heparin reading in Figure 2, washing was done before a new layer of protamine was adsorbed to regenerate the surface. Control experiments without protamine regeneration or no protamine at all confirm negligible adsorption of heparin on pristine COC (Figure S9, Supporting Information). The apparent affinity constant determined after fitting is 0.03 units/mL, which is not far from the value of 0.12 units/mL previously reported.17 The positive charge of the adsorbed protamine is gradually reversed to negative streaming potential values as the concentration of anionic heparin increases. The linear region for heparin detection in Figure 2 ranges from ∼0.01 to 0.10 units/mL, which is below the clinically relevant concentrations. Heparin doses given to patients during CPB typically go from 2 and 8 units/mL and an order of magnitude lower for postsurgical therapy.17 Because the immobilization of protamine occurs by spontaneous irreversible adsorption on the (18) DeBois, W. J.; Liu, J. L.; Lee, L. Y.; Girardi, L. N.; Mack, C.; Tortolani, A.; Krieger, K. H.; Isom, O. W. Perfusion 2003, 18, 47–53. (19) Lillo-de louet, A.; Boutoyrie, P.; Alhenc-Gelas, M.; Le Beller, C.; Gautier, I.; Aiach, M.; Lasne, D. J. Thromb. Haemost. 2004, 2, 1882–1888. (20) Cai, S.; Dufner-Beattie, J. L.; Pretwich, G. D. Anal. Biochem. 2004, 326, 33–41. (21) Guo, J.; Amemiya, S. Anal. Chem. 2006, 78, 6893–6902. (22) Xiao, K. P.; Kim, B. Y.; Bruening, M. L. Electroanalysis 2001, 13, 1447– 1453. (23) Buchanan, S. A.; Kennedy, T. P.; MacArthur, R. B.; Meyerhoff, M. E. Anal. Biochem. 2005, 346, 241–245.

Figure 2. Pulsed streaming potentials as a function of [heparin] on a COC-protamine-modified microchannel. The plot was obtained through the following sequence: (A) Protamine was adsorbed by flushing 0.1 mg/mL of protamine in 10% phosphate buffer saline solution (PBS) at ∼7 µL/s for 14 s. (B) Protamine-free solution (10%) in PBS was flushed, and the measured ∆E was labeled as “adsorbed protamine”. (C) Another load of protamine was adsorbed as in step A to compensate for any protamine loss during the ∆E reading. (D) Heparin was flushed in 10% PBS, and the measured ∆E was labeled as “heparin after protamine”. (E) Washing was done with 10% PBS at ∼7 µL/s for 14 s three times. (F) A new cycle A-E was started for the next [heparin]. The conductivity and pH for all solutions were 1750-1850 µS/cm and 7.4, respectively.

COC microchannel, the time for each data point in Figure 2 takes no longer than 15 s, implying a fairly quick time for analysis. This aspect is clinically significant as the rapid monitoring of heparin concentration can aid on assessing a patient’s state in a timely manner during surgery.18,19 Nevertheless, further studies exploring other immobilization strategies as well as more selective heparin receptors like antithrombin are under way in our laboratory. Depending on the kinetics of the adsorption-desorption equilibrium in comparison to the time scale of the measurement (5-10 s), the binding of the analyte to the modified surface can be considered as reversible or irreversible. Reversible adsorption implies that ∆E can be measured in the presence of the analyte (continuous monitoring), whereas irreversible binding means that ∆E can be measured after the analyte has been adsorbed and using analyte-free solution (analyte-free detection). This consideration, which is made purely on the basis of the pulsing and flow rates available with our setup, in the end determines the most suitable mode of detection. In any case, the magnitude of the signal really depends on the actual amount of adsorbed analyte that changes the charge under continuous or analyte-free monitoring. It was found that heparin and MB were best detected by continuous monitoring, whereas BSA and streptavidin were detected in analyte-free conditions. To illustrate the advantage of detection with analyte-free solutions, we studied the adsorption of streptavidin to a COC microchannel modified with biotin on the surface. The use of the biotin-streptavidin binding for detection and immobilization of numerous biological targets is widespread in modern biotechnology. The complex is considered to be irreversible as a result of its high binding constant (∼1015). Figure 3 shows a bar plot with pulsed streaming potentials for two different COC microchannels. Microchannels 1 and 2 were first modified with biotin (yellow) and then incubated with streptavidin (red) to observe the change on the surface charge upon adsorption (inset of Figure 3). To rule out the possibility of nonspecific adsorption, we exposed both microchannels to a

Figure 3. Pulsed streaming potentials for a COC-biotin-modified microchannel incubated for 1 h with 0.5 mg/mL streptavidin 1× PBS after (1) and before (2) exposure to 0.5 mg/mL BSA 1× PBS. ∆E was recorded with streptavidin-free buffer at pH 3.3 and 0.59 mM phosphate.

solution of BSA (blue) before (microchannel 1) and after (microchannel 2) streptavidin incubation. All ∆E values were recorded with streptavidin-free solution at pH 3.3 buffered with 0.59 mM phosphate. Both biotinlated COC microchannels (yellow) showed similar increases in streaming potentials after streptavidin adsorption (red): from +0.010 ± 0.001 to +0.150 ± 0.007 V for microchannel 1 and from +0.009 ± 0.000 to 0.180 ± 0.001 V for microchannel 2. Incubation with BSA for either microchannel showed little change in the magnitude of the streaming potential even if it was before (from +0.010 to +0.010 V) or after (from +0.180 to +0.170 V) streptavidin adsorption. Because both BSA and streptavidin have isoelectric points of ∼5.8 and ∼5.0, respectively,24 their adsorption on the microchannel should produce an addition of positive charge to the surface when recording ∆E at pH 3.3. However, despite employing the same concentration (0.5 mg/mL) and incubation time (1 h), only streptavidin increased the streaming potential of the biotinlated microchannels by ∼94% (Figure 3). In contrast, by recording ∆E with a solution at pH 7.7, the values of ∆E were about -0.140 V and only varied in ∼3% throughout the same sequence shown in Figure 3 (Figure S10, Supporting Information). The analyte-free detection mode has great potential for bioprobe detection using immunoreagents because a significant number of these antibody-antigen complexes are irreversible in the time scale utilized here. No surface regeneration is needed because the COC chips are disposable and can be mass produced. Besides, they require very minute amounts of sample. CONCLUSION Streaming potentials have been known for almost a hundred years since the introduction of the Smoluchowski equation.6 During this time, these measurements have been primarily made for determining ζ-potentials in surface characterization.1–4,8 Although there have been attempts to develop detection methods based on steady-state streaming potentials,9,10 the intrinsic difficulties derived from using eq 1 have posed a real challenge. For instance, contributions to ∆E from adsorption, polarization, and Nernst potentials on the indicator electrodes can affect ∆E, especially if redox species are present. Likewise, changes in η, , (24) Shi, Q.; Zhou, Y.; Sun, Y. Biotechnol. Prog. 2005, 21, 516–523.

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

6535

κ, and ζ (eq 1) during a measurement can contribute to ∆E as well. These contributions, which may originate from the analyte itself or from other species in solution, are always present in some extent and are difficult to remove. However, with the adaptation to a plastic microfluidic platform, the automated pulsed flow, and the detection modes implemented here, some of these factors are minimized although not completely removed. For example, the pressure pulses help to separate the contribution from electrochemical drift (Nernst potentials) on the indicator electrodes because the streaming potential is proportional to the pressure. The nonconducting nature of the COC channel also helps to mitigate the effect of Nernst potentials, which otherwise would be more pronounced on a metallic channel.25,26 Finally, for analytes that adsorb irreversibly in the time scale of the measurement (i.e., biotin-streptavidin conjugates or immunoprobes), detection with analyte-free solutions could reduce unwanted analyte and sample contributions to eq 1. This benefit stems from the fact that at the moment of detection the analyte is only present bound to the surface, while the other sample components are washed away before detection. Overall, this work is the first demonstration of label-free detection of analytes using pulsed streaming potentials in a plastic microfluidic platform. Despite the fact that the pressure-driven flow is carried out under pulsed conditions, the Smoluchowski equation is still an acceptable model. This detection approach shares similarities with other surface-based detection techniques such as surface plasmon resonance and quartz crystal microbalance. The analytical signal, which is induced by the analyte upon surface adsorption, is a change in the surface charge. Some (25) Duval, J. F. L.; Huijs, G. K.; Threels, W. F.; Lyklema, J.; van Leeuwen, H. P. J. Colloid Interface Sci. 2003, 260, 95–106. (26) Duval, J. F. L.; van Leeuwen, H. P. J. Phys. Chem. B 2003, 107, 6782.

6536

Analytical Chemistry, Vol. 80, No. 17, September 1, 2008

advantages that we envision over those techniques are the fact that this approach is not restricted to a unique surface (gold) and there is no need for lasers or special optics to obtain the signal. Additionally, the streaming potential approach is ideally suited for microfluidics because the signal is a result of the liquid flow, the microchannel adsorption affinity, and its surface charge, which can be modified by UV-photografting methods.11 This proof of concept, involving analytes with different degrees of adsorption specificity, shows that this method has acceptable precision with lowest detectable concentrations between ∼10-9 and ∼10-6 M, depending on the analyte. Nevertheless, before these advantages can be fully realized for potential applications in diagnostics or biodefense, problems like selectivity and nonspecific adsorption, which are ubiquitous in any surface-based sensor, must be addressed. To this end, current efforts using more specific and selective receptors such as the ones based on immunoaffinity are under way in our laboratory. ACKNOWLEDGMENT We thank VCU, The Jeffress Memorial Trust of Virginia (Grant J-754), The NSF (CHE-0645494), and The Egyptian Bureau of Education for their generous support of this work. SUPPORTING INFORMATION AVAILABLE Details of the instrumentation built in our laboratory for measuring streaming potentials and control experiments. This material is available free of charge via the Internet at http:// pubs.acs.org. Received for review February 14, 2008. Accepted June 24, 2008. AC8003117