Microfluidic Immunoassay for Bacterial Toxins with Supported

In addition, the high surface-to-volume ratio of microchips enhances mass .... for quantitative fluorescence imaging of the sample chip at a resolutio...
0 downloads 0 Views 237KB Size
Anal. Chem. 2005, 77, 327-334

Microfluidic Immunoassay for Bacterial Toxins with Supported Phospholipid Bilayer Membranes on Poly(dimethylsiloxane) K. Scott Phillips and Quan Cheng*

Department of Chemistry, University of California, Riverside, California 92521

We report a heterogeneous immunoassay for cholera toxin (CT) using supported bilayer membranes (SBMs) in a poly(dimethylsiloxane) (PDMS) microfluidic device. Phosphatidylcholine membranes assembled on plasma-oxidized PDMS by vesicle fusion bring about favorable surface properties, such as improved wettability and protein resistance. Contact angle measurements show that the lipid membranes can preserve hydrophilic surfaces for hours, whereas untreated substrates rapidly undergo hydrophobic recovery. Fluorescence recovery after photobleaching performed in situ reveals that the membranes have relatively high lateral mobility. Experimental datafitting to theoretical models yields diffusion coefficients of 1.8 ( 0.7 µm2/s on PDMS and 3.4 ( 0.8 µm2/s on glass. Fluorescence studies utilizing tagged proteins show that SBMs reduce nonspecific adsorption of avidin and BSA on PDMS by 2-3 orders of magnitude, as compared to that on plasma oxidized surfaces. SBMs and their protein-resistant properties are not significantly affected by long flow times, indicating good membrane stability. These studies increase our understanding of the relationship between molecular level interactions and membrane properties, allowing for development of a rapid heterogeneous immunoassay for CT in PDMS microchips with cell surface receptor molecules. Using optimized sample injection and buffer washing conditions, microfluidic immunoassay of CT is complete within 25 min, and a dynamic range over 3 orders of magnitude with a detection limit of 8 fmol of toxin is achieved. Poly(dimethylsiloxane) (PDMS)-based microfluidic devices have attracted enormous attention in recent years as a platform for “lab-on-a-chip” systems.1-4 The microscale dimensions of these systems enable shortened analysis times concurrent with significantly reduced reagent consumption and waste generation. Fabrication and characterization of PDMS devices was recently * Corresponding author. Phone: (909) 787-2702. E-mail: [email protected]. (1) Manz, A.; Harrison, D. J.; Verpoorte, E. M. J.; Fettinger, J. C.; Paulus, A.; Ludi, H.; Widmer, M. J. Chromatogr., A. 1992, 593, 253-258. (2) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974-4984. (3) Christodoulides, N.; Tran, M.; Floriano, P. N.; Rodriguez, M.; Goodey, A.; Ali, M.; Neikirk, D.; McDevitt, J. T. Anal. Chem. 2002, 74, 3030-3036. (4) Roper, M. G.; Shackman, J. G.; Dahlgren, G. M.; Kennedy, R. T. Anal. Chem. 2003, 75, 4711-4717. 10.1021/ac049356+ CCC: $30.25 Published on Web 12/01/2004

© 2005 American Chemical Society

the subject of comprehensive reviews.5,6 New methods have been realized to construct elaborate multidimensional structures, enabling a wide range of functions, such as the study of complex fluid mechanics7 and development of assays for enzymatic activity.8 A promising application in microfluidic analysis is the detection of biological pathogens and disease-associated markers.9 Many homogeneous diagnostic and sensing applications of PDMS devices have been reported for microorganisms,10 organophosphates,11 toxins,12 and viruses.13 Parallel to these efforts is the development of heterogeneous microfluidic immunoassays with PDMS chips.14-16 In this method, ligand molecules are immobilized on the microchannel surface, and the binding activity of target molecules flowing through the substrate is detected and quantified. Flow immunoassay is attractive because it provides quantitative results, reduced sample handling, and speed. In addition, the high surface-to-volume ratio of microchips enhances mass transport, resulting in shorter assay times and increased detection sensitivity. However, the intrinsic surface properties of PDMS are a serious obstacle delaying its use in heterogeneous immunoassays. The repeating -O-Si(CH3)2- groups are significantly hydrophobic, resulting in poor sample loading and severe nonspecific protein adsorption. Several methods have been suggested to render the PDMS surface more hydrophilic, including polymer grafting, chemical functionalization, polyelectrolyte coating, and plasma oxidation. Graft polymerization involves the attachment of initiator molecules for a new polymer to the existing substrate. Methods based on irradiation as a radical source have been reported for PDMS, but they involve multiple steps and long (5) Ng, J. M. K.; Gitlin, I.; Stroock, A. D.; Whitesides, G. M. Electrophoresis 2002, 23, 3461-3473. (6) McDonald, J. C.; Whitesides, G. M. Acc. Chem. Res. 2002, 35, 491-499. (7) Kwok, Y. C.; Jeffery, N. T.; Manz, A. Anal. Chem. 2001, 73, 1748-1753. (8) Seong, G. H.; Heo, J.; Crooks, R. M. Anal. Chem. 2003, 75, 3161-3167. (9) Walter, G.; Bussow, K.; Lueking, A.; Glokler, J. Trends Mol. Med. 2002, 8, 250-253. (10) Ogata, S.; Suzuki, M.; Takahashi, T.; Nishizawa, M.; Matsue, T. Chem. Sens. 2002, 18 (Suppl. B), 85-87. (11) Wang, J.; Chatrathi, M. P.; Mulchandani, A.; Chen, W. Anal. Chem. 2001, 73, 1804-1808. (12) Ewalt, K. L.; Haigis, R. W.; Rooney, R.; Ackley, D.; Krihak, M. Anal. Biochem. 2001, 289, 162-172. (13) Kwakye, S.; Baeumner, A. Anal. Bioanal. Chem. 2003, 376, 1062-1068. (14) Dodge, A.; Fluri, K.; Verpoorte, E.; de Rooij, N. F. Anal. Chem. 2001, 73, 3400-3409. (15) Eteshola, E.; Leckband, D. Sens. Actuators, B 2001, B72, 129-133. (16) Hofmann, O.; Voirin, G.; Niedermann, P.; Manz, A. Anal. Chem. 2002, 74, 5243-5250.

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005 327

Figure 1. Structure of phospholipids and cell surface ligand GM1 (left) and the immunoassay scheme on PDMS-based SBMs (right). For immunoassay, (a) vesicle fusion to form GM1 integrated SBMs, (b) capture of CT on the membrane, (c) binding of rabbit anti-CT, (d) binding of Alexa 532-tagged goat anti-rabbit for fluorescent detection.

preparation time, and they can alter the transparent optical properties.17 Noncovalent coating with multiple layers, such as polyelectrolytes and biopolymers, has been shown to stabilize electroosmotic flow (EOF) in chip-based capilary electrophoresis (CE)18 and reduce nonspecific adsorption.19 Another widely adopted method to generate a hydrophilic surface is plasma oxidation, which produces a silicate-like network terminated with hydrophilic end groups. However, plasma-treated surfaces recover hydrophobic properties very quickly.20 Our group is interested in design and fabrication of biomimetic membranes on solid substrates for detection of pathogenic agents.21 The use of vesicles and related membranes, such as supported bilayer membranes, has great potential for sensor development. Recently, supported bilayer membranes (SBMs) have attracted considerable attention for surface modification and functionalization on glass and silicon substrates.22 Vesicle fusion on these surfaces affords a biointerface in which a single lipid bilayer is attached to the solid substrate by physical interactions or chemical bonds.23,24 Supported membranes are advantageous because they retain the protein-resistant headgroups and lipid interiors of their biological counterparts while offering convenience in handling and application. Nonspecific, irreversible adsorption of several proteins on egg phosphatidylcholine (egg PC) SBMs was reportedly reduced to below detectable limits in a study using quartz crystal microbalance (QCM) and surface plasmon resonance (SPR).25 A similar effect was observed for SBMs as a coating in capillary CE, where a substantial decrease in protein adsorption resulted.26 The well-defined membrane structure also allows for convenient integration of ligand molecules into the interface while maintaining desirable molecular conformations that would otherwise be impracticable to achieve.27,28 The fluidity of SBMs enables higher binding activity for multivalent ligands. This enhanced format for binding and subsequent detection has been demonstrated in a recent study of the binding of six protein toxins to SBM-integrated glycolipids on an optical waveguide sensor.29 Yet, compared to SBMs on glass, the development of PDMSbased SBMs has been the subject of only a few initial reports. Cremer and co-workers first used phospholipid-coated PDMS 328

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

channels to determine the binding kinetics of an antibody to a lipid-conjugated antigen with total internal reflection fluorescence.30 In another study, they explored a heterogeneous format for determining enzyme turnover rates and the detection of glucose.31 Nevertheless, a molecular level understanding of the factors controlling both the vesicle fusion process on PDMS and the resulting structure and stability of self-assembled lipid films is still lacking. Properties such as wettability and protein resistance in various assay conditions have not been fully characterized. Little is known about the fluidity of membranes in PDMS microchannels, as compared to those on hydrophilic SiO2 surfaces, and how it will affect biomolecular interactions. In this paper, we report a systematic study on the fabrication of robust and biocompatible SBMs on PDMS microchannels and the use of these membranes to develop a heterogeneous microfluidic immunoassay for bacterial protein toxins. Figure 1 shows the molecular structure of phospholipids and the cell surface receptor GM1 as well as the detection scheme. Incorporation of GM1 in the lipid membranes allows capture and detection of cholera toxin, an 82-kDa protein produced by Vibrio cholerae or Escherichia coli [ETEC], with high (17) Hu, S. W.; Ren, X. Q.; Bachman, M.; Sims, C. E.; Li, G. P.; Allbritton, N. Anal. Chem. 2002, 74, 4117-4123. (18) Liu, Y.; Fanguy, J. C.; Bledsoe, J. M.; Henry, C. S. Anal. Chem. 2000, 72, 5939-5944. (19) Linder, V.; Verpoorte, E.; Thormann, W.; de Rooij, N. F.; Sigrist, M. Anal. Chem. 2001, 73, 4181-4189. (20) Hillborg, H.; Gedde, U. W. IEEE Trans. Dielectr. Electr. Insul. 1999, 6, 703-717. (21) Xu, D.; Cheng, Q. J. Am. Chem. Soc. 2002, 124, 14314-14315. (22) Groves, J. T.; Ulman, N.; Boxer, S. G. Science 1997, 275, 651-653. (23) Brian, A.; McConnell, H. M. Proc. Natl. Acad. Sci. U.S.A. 1984, 81, 61596163. (24) Sackman, E. Science 1996, 271, 43-48. (25) Glasmastar, K.; Larsson, C.; Hook, F.; Kasemo, B. J. Colloid Interface Sci. 2002, 246, 40-47. (26) Cunliffe, J. M.; Baryla, N. E.; Lucy, C. A. Anal. Chem. 2002, 74, 776-783. (27) Fang, Y.; Frutos, A. G.; Lahiri, J. Langmuir 2003, 19, 1500-1505. (28) Song, X.; Shi, J.; Swanson, B. Anal. Biochem. 2000, 284, 35-41. (29) Puu, G. Anal. Chem. 2001, 73, 72-79. (30) Yang, T. L.; Jung, S. Y.; Mao, H. B.; Cremer, P. S. Anal. Chem. 2001, 73, 165-169. (31) Mao, H. B.; Yang, T. L.; Cremer, P. S. Anal. Chem. 2002, 74, 379-385.

binding affinity and specificity.32 In addition, the lateral mobility of GM1 in the bilayer is expected to increase the binding avidity by promoting multivalent interactions.33 In this study, great effort will be focused on the characterization of SBM properties on PDMS to gain a better understanding of the membrane behavior and optimize the conditions for immunoassay. The change of surface wettability will be studied by bulk contact angle measurements, and fluorescence spectroscopy will be used to evaluate protein adsorption in treated microchannels and the effect of assay conditions on membrane stability. The fluidity of the membranes will be determined by fluorescence recovery after photobleaching (FRAP). Finally, a PDMS-based immunoassay procedure will be demonstrated for fast and sensitive detection of cholera toxin. EXPERIMENTAL SECTION Materials. Phosphatidylcholine (PC), 1,2-dioleoyl-sn-glycero3-ethylphosphocholine (DOPC+) and 1-oleoyl-2-[6-[(7-nitro-2-1,3-benzoxadiazol-4-yl)amino]hexanoyl]-sn-glycero-3-phosphocholine (NBD-PC) were purchased from Avanti. Dow Corning Sylgard 184 silicon elastomer base and curing agent for fabrication of PDMS chips were obtained from a local supplier. Cholera toxin (CT), avidin Texas Red conjugate, rabbit anti-cholera serum, and anti-BSA serum were from Sigma. Bovine serum albumin (BSA), Texas Red conjugate and AlexaFluor 532 conjugated goat antirabbit IgG were purchased from Molecular Probes. Monosialoganglioside receptor GM1 was from Matreya (Pleasant Gap, PA). All other chemicals were reagent grade and were used as received. Instrumentation. A Harrick PDC-32G plasma cleaner was used for PDMS oxidation. Contact angle measurements were carried out with an in-house setup. The Amersham Typhoon 9410 scanner was used for quantitative fluorescence imaging of the sample chip at a resolution of 10 µm. A Nd:YAG laser provided excitation at 532 nm, and emission filters were centered at 610 nm for Texas Red and 555 nm for Alexa 532. In FRAP experiments, a Meridian Insight confocal laser scanning microscope (CLSM) with argon laser excitation, cooled CCD, and fluorescein emission filter was used with a 40×/0.75na Achroplan dipping objective. Contact Angle Measurements. PDMS base and curing agent were mixed at a 10:1 ratio and degassed for 1 h. The mix was poured onto clean glass slides in thin layers and cured for 1 h at 65° C. The side of PDMS in contact with the glass slides was oxidized for 30 s with a PDMS frame on top to define the test areas. Immediately after oxidation, the test areas were filled with buffer or vesicle solutions. After 1 h, the surfaces were gently washed with buffer three times and dried with a clean air stream. To avoid hydrophobic recovery related to silicate cracking, a sample area was cut from the PDMS surface just before each measurement and was placed on the drop stage of the contact angle measurement setup. This process was repeated on at least three separate substrates for every reported data point. PDMS Chip Production and Flow Injection Apparatus. The aluminum master mold for PDMS microchannel chips was (32) Van Heyningen, S. Curr. Top. Membr. Transp. 1983, 18, 445-469. (33) Lauer, S.; Goldstein, B.; Nolan, R. L.; Nolan, J. P. Biochemistry 2002, 41, 1742-1751.

fabricated at the UCR microfabrication facility on the basis of the method reported by Brooks et al.34 The channels were 200 µm × 200 µm × 2 cm with wider cylindrical ends for connections. PDMS was poured into the master mold and cured for 1 h at 65 °C. After the chip was removed from the master, the channels were plasmaoxidized for 30 s. Disposable needles attached to Teflon FEP tubing were inserted into the front connectors while regular needles were inserted in the outlets. The whole assembly was clamped down on a glass slide. A syringe pump (KD Scientific) and low-pressure sample injectors (Upchurch) were used for flow injection experiments. Vesicle Assembly and Fusion. Vesicles were prepared from chloroform stocks that were dried with a nitrogen stream to form a thin film in vials. The solid was then rehydrated with a tris buffer (10 mM with 150 mM NaCl, pH 7.4), and the probe was sonicated for 10 min, followed by 1 h incubation at 4 °C. The vesicles were injected at a flow rate of 1 mL/h for ∼2 min into each channel until the solution filled the outlet syringe wells. After a 1-hour incubation, the excess vesicles were rinsed out with the tris buffer. FRAP. Fluorescent vesicles were prepared by mixing 2% NBDPC with PC lipids. After sonication and incubation, the vesicles were purified using an ultracentrifuge (38 000 rpm for 30 min and then at 52 000 rpm for 3 h).35 The clear supernatant was collected and used for vesicle fusion in the microchannels as described above. After thorough rinsing and removal of the glass bottom, the chip was secured in a small Petri dish with a buffer level high enough to cover the channel features. The dipping objective from a Meridian Insight confocal laser scanning microscope was focused on the channel surface. The scanning mode of the CLSM was stopped, and a perpendicular line was bleached for 0.4, 0.7, or 1 s with a 488-nm argon laser. Fluorescence recovery was monitored with time, and the corrected intensity of the photobleach center vs time was plotted and fit to the model developed by Koppel.36 Protein Adsorption Studies. After vesicle assembly/rinse, the fluorescently labeled protein was injected for 1 min at a flow rate of 1 mL/h and then rinsed for 10 minutes with Tris buffer. The glass slide was then removed, and the channels were imaged on the fluorescence scanner. Two background channel intensities were collected for each chip and used for signal correction. In longer trials, a 2-h flow time was used in place of the 5 min vesicle rinse period. µFIA. To prepare toxin-binding bilayer membranes on PDMS, 5% ganglioside GM1 receptor molecules were added to the PC lipids. The functionalized vesicle solution was not kept more than 48 h. During the assay experiment, flow was maintained at 1.5 mL/h, and sample loops had a volume of 37 µL. Injection and rinse times were 2 or 5 min for each step (CT, primary, secondary fluorescent conjugated antibody). At the end of the assay, an additional 10-min rinsing was added to remove any weakly adsorbed antibodies. The chips were then imaged and quantified on the Amersham Typhoon scanner. (34) Brooks, S. A.; Dontha, N.; Davis, C. B.; Stuart, J. K.; O’Neill, G.; Kuhr, W. G. Anal. Chem. 2000, 72, 3253-3259. (35) Barenholz, Y.; Gibbes, B. J. G.; Thompson, T. E.; Carlson, F. D. Biochemistry 1977, 16, 2806-2810. (36) Koppel, D. E. Biophys. J. 1979, 28, 281-292.

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

329

RESULTS AND DISCUSSION Characterization of SBMs on PDMS by Contact Angle Measurements. Supported lipid membranes were fabricated on the PDMS substrate through vesicle fusion. The method has been widely used on surfaces such as glass and mica because it is simple, fast, and effective;37 however, assembly of phospholipid membranes on PDMS has not been well-characterized and -understood because of the complex structure of the polymer surface. The degree to which a membrane structure could selfassemble on PDMS, its ability to alter surface properties, and the stability of the resulting hydrophilic character were previously unknown. We first characterized the modified PDMS surfaces with contact angle measurements to directly obtain information about surface hydrophobicity. The PDMS substrate was pretreated by plasma oxidation and exposed to vesicle solutions for 1 h. Contact angle measurements were performed by the sessile drop method for both control and vesicle treated surfaces. When PDMS substrates were oxidized and exposed to Tris buffer without vesicles, an initial contact angle of 31° was observed, indicating a highly wettable surface. After 2 h, the contact angle increased to about 57°, suggesting hydrophobic recovery. This phenomenon represents a significant problem in microfluidic protein analysis because it can lead to irreproducible separations, sample loss, and high background signal. Treating the oxidized PDMS with 1 mg/ mL PC vesicle solutions for 1 h, followed by thorough rinsing, resulted in a hydrophilic surface with a similar initial contact angle of ∼30°. The treated PDMS remained highly wettable after 2 h or more, showing no significant contact angle increase. This longstanding hydrophilic property conferred by lipid assembly on PDMS is highly advantageous. It is worth noting that similar treatment of non-oxidized PDMS surfaces with vesicle solution did not change the contact angle. Apparently, PC vesicles do not fuse on native hydrophobic PDMS under the conditions reported here. To compare the SBMs with surface blocking reagents used in traditional assays, the effect of bovine serum albumin (BSA) on PDMS hydrophilicity was studied. A 1-h exposure of the oxidized PDMS surface to 1 mg/mL BSA solution resulted in an initial contact angle of 54°, which increased to 67° after 2 h. Judging from the contact angle measurements alone, BSA is clearly less effective than PC for maintaining a hydrophilic surface on PDMS. Fluidity of Lipid Membranes on PDMS. Although contact angle measurements provide an effective assessment of macroscopic properties, they are less sufficient for microscopic examination of PC-treated PDMS surfaces. A major concern is if vesicle fusion, rather than adsorption of intact vesicles, is really the source of the observed hydrophilic properties. We extended the surface characterization with FRAP techniques to investigate the degree of SBM mobility on PDMS. FRAP has been a standard approach for studying membrane fluidity.38 The measured lateral diffusion coefficients can shed light on whether a bilayer or equivalent system has resulted after vesicle solution exposure. In the FRAP experiment preparation, real assay conditions were used for assembly of membranes on PDMS microchannels. After extensive rinsing with Tris buffer, the glass coverslip was removed so that (37) Seitz, M.; Park, C. K.; Wong, J. Y.; Israelachvili, J. N. ACS Symposium Series 1999, 736, 215-230. (38) Lee, G. M.; Jacobson, K. Curr. Top. Membr. 1994, 40, 111-142.

330

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

only the PDMS part of the chip was evaluated with FRAP. During this process, the channels were kept wet at all times. Since the microchannels are very narrow, a bleaching spot on the order of 1 µm does not satisfy the requirement of having an infinite sink on all sides.39 This problem was overcome by taking advantage of a one-dimensional bleaching procedure in which a line-scan by a CLSM was used to bleach a line perpendicular to the channels. Images of the fluorescence recovery along the channel were captured at different time intervals with a CCD. The onedimensional recovery is described by Koppel in ref 36

f ) 1 - βR1(t)e-(∆x) /w0 - (1 - β)R0e-(∆x) /w0 2

2

2

2

(1)

where f is the fractional recovery, β is the mobile fraction, R is the bleached fraction, and w0 is the 1/e2 bleaching profile. By taking the fractional recovery at the minimum of the bleach line, ∆x becomes 0 and the exponential factors reduce to unity. After substituting

R1(t) )

R0

x

1+

t τd

where τd ) w02/4D into the equation, we obtained

f ) 1 - R0(1 - β) -

R0β

x1 + 4Dtw02

(2)

The value of this expression is that the effect of each parameter on the fractional recovery can be easily seen. For example, the effect of a small immobile fraction (1 - β) can be seen to dominate the shape of the curve at longer times. The initial fitting parameters were obtained from spatial information in the images. Since the bleached profile is Gaussian at t ) 0, it remains Gaussian during diffusive recovery.34 Profiles across the bleach center were averaged to improve signal/noise, allowing for short integration times that minimized the extent of photobleaching. Figure 2 shows the fluorescence images of a lipid membrane that contains 2% NBD-PC and the fractional recovery intensity after photobleaching. The horizontal stripes in the images are a common artifact of slit scanning technology and are caused by minor defects in optical elements being stretched in the sweep direction of the scanning mirror. It is noteworthy that visualization of the NBD-PC by confocal microscopy confirmed that the lipid layer covers all areas of the channels. Table 1 summarizes the results obtained from nonlinear curve-fitting based on eq 2. Plasma-cleaned glass surfaces were used as a comparison, since SBMs on glass are well-characterized. The diffusion coefficient (D) was 3.4 µm2/s on glass and 1.8 µm2/s on PDMS. These values agree reasonably well with those reported for glass40 and PDMS.41 In addition, both surfaces showed complete recovery, indicated by the average β values of unity. Comparable membrane fluidity (39) Berquand, A.; Mazeran, P.; Pantigny, J.; Proux-Delrouyre, V.; Laval, J.; Bourdillon, C. Langmuir 2003, 19, 1700-1707. (40) Kalb, E.; Frey, S.; Tamm, L. K. Biochim. Biophys. Acta 1992, 1103, 307316. (41) Hovis, J. S.; Boxer, S. G. Langmuir 2001, 17, 3400-3405.

Figure 2. Top: fluorescent images of lipid membranes containing 2% NBD-PC on oxidized PDMS substrate before (1) and after bleaching (2-6) at increasing recovery times. Bottom: FRAP recovery curve for PC-treated PDMS. Table 1. FRAP Results for Lateral Diffusion Coefficients of Fluorescent Probes in PC Membranes on Glass and PDMS lipid

surface

n

β

r2

D (µm2/s)

PC

PDMS glass

12 5

1.00 1.00

0.994 0.995

1.8 ( 0.7 3.4 ( 0.8

on PDMS and glass substrates suggests that the polymer chip could be used as a simple and less expensive alternative to SBM molecular devices based on silicon and glass. Protein Adsorption on SBMs in Microchannels. A major function of hydrophilic surfaces in microfluidic analysis is to resist nonspecific adsorption. Although several reports have shown the ability of SBMs on silicon or gold surfaces to resist protein adsorption and cell adhesion,25,42 a similar study on PDMS has not been performed in a systematic manner. One reason may be due to the limitations of optical studies in microchannels, which often rely on confocal microscopy to measure fluorescence intensity. Problems in sample positioning, limited focal area, and irreproducible focal distance also make it difficult to obtain fluorescence intensity over large areas and quantify differences in multiple channels or among chips. To understand the effective(42) Groves, J. T.; Mahal, L. K.; Bertozzi, C. R. Langmuir 2001, 17, 5129-5133.

ness of SBMs on PDMS for reducing nonspecific interactions in typical FIA conditions, we used a high-resolution fluorescence scanner to investigate the adsorption of labeled proteins onto SBMs. In addition to spatial resolution, sensitivity and optically flat images, the scanner also allows a number of microchips to be imaged in minutes, making it amenable to scaling up for highthroughput applications. Figure 3a shows the fluorescence image of nonspecific adsorption of avidin in oxidized PDMS channels. Clearly, the intensity increases with protein concentration. From the plot in Figure 3b, oxidation of PDMS actually increases the irreversible protein adsorption of fluorescently labeled BSA and avidin. This coincides with a recent report showing that hydrophilic surfaces do not always confer protein resistance.43 On a molecular scale, there may be numerous hydrophobic domains that serve as the interaction sites for protein adsorption. Other factors besides hydrophobic interaction may be responsible for the increased adsorption of avidin on an oxidized surface. Given that avidin (pI ∼ 10.5) is positively charged and BSA is negatively charged (pI ∼ 4.8) in pH 7.4 buffer, the increase is probably the result of Coulombic attraction to electronegative surface functional groups on the oxidized PDMS. Figure 4a shows the fluorescence images from protein adsorption after PC vesicle fusion on PDMS. The signal from avidin was substantially reduced to nearly background level, while BSA still showed a small amount of adsorption, with an average intensity of 33 RU (Figure 4b). These values represent over 3 orders of magnitude decrease for avidin adsorption and over 2 orders of magnitude decrease for BSA. The results clearly demonstrate the effective protein resistance capabilities of the PC membranes on PDMS. For comparison, channels treated with the synthetic lipid DOPC+ were studied to reveal the effect of electrostatic interaction on adsorption. In channels treated with DOPC+, which has a positively charged headgroup, fluorescence signal from avidin was indistinguishable from background, whereas the average fluorescence for BSA increased to 145 RU, indicating enhanced nonspecific adsorption (Figure 4b). Although DOPC+ has the capacity to further suppress avidin adsorption, the zwitterionic PC lipids are more effective against nonspecific protein adsorption for both positively and negatively charged species. It should be noted that the protein concentration used here (0.1 mg/mL) is considered an upper limit for protein analysis. A typical analytical sample preparation would contain less concentrated protein, and therefore, the fluorescent signal associated with nonspecific interactions could fall below the threshold of random noise on PC treated PDMS. Although SBMs on PDMS show great potential for reduction of nonspecific signal in protein analysis, their value in microfluidic immunoassay depends on how stable they are in flow conditions. To simulate the hydrodynamic force in a flow experiment, buffer was pumped through the channels at a rate of 1 mL/h after vesicle fusion. Figure 4b shows that PC treated chips exposed to avidin after a 2-h flow time did not show a significant change in signal, as compared to those exposed to avidin immediately after assembly. If the flow stream damaged the bilayer, bare PDMS surfaces would be exposed, and protein adsorption would increase. (43) Ostuni, E.; Grzybowski, B. A.; Mrksich, M.; Roberts, C. S.; Whitesides, G. M. Langmuir 2003, 19, 1861-1872.

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

331

Figure 3. (a) Fluorescence image of oxidized PDMS microchannels after exposure to Texas Red-tagged avidin. The avidin concentrations are, from right to left, 0.1, 0.05, 0.02, and 0.01 mg/mL. (b) Fluorescence intensity for 0.1 mg/mL avidin and BSA on oxidized and unoxidized PDMS.

Figure 4. (A) Fluorescence image of PDMS microchannels after exposure to Texas Red-tagged avidin. From left: PC-treated, DOPC+treated, oxidized only, and background channel. Flow rate was 1 mL/h. 37 µL of 0.1 mg/mL avidin was injected, followed by 10-min rinsing, except in the background channel. (B) Effect of experimental conditions on fluorescence intensity of PDMS microchannels. From left: (a) immediately tested after rinsing, PC-treated; (b) immediately tested after rinsing, DOPC+-treated; and (c) tested after a 2-h flow period at 1 mL/h, PC-treated.

From Figure 4b, we conclude that the SBMs obtained on PDMS have extended stability suitable for development of heterogeneous microfluidic immunoassays. Fluorescence Immunoassay of CT in PDMS Microfluidic Device. The full coverage and fluidity of the PC bilayer membrane on PDMS, long-standing stability in flowing conditions, and resistance to protein adsorption make PC-treated PDMS a promising substrate for flow-based assays aiming at high sensitivity. Therefore, a microflow immunoassay (µFIA) for cholera toxin was developed by incorporating the CT receptor GM1 into PC vesicles. GM1 is an amphiphilic cell membrane-bound molecule with a carbohydrate headgroup and a ceramide tail, allowing it to be easily reconstituted in vesicle membranes. After spontaneous fusion on the PDMS surface, it is thought to be distributed between both leaflets of the membrane. While the hydrophilic choline headgroup of PC suppresses nonspecific binding of proteins, the GM1 receptor binds specifically to cholera toxin. This property makes functionalized phospholipid bilayers an attractive sensing interface. A major goal of our PDMS-based microfluidic approach is to decrease analysis time. After vesicle fusion and thorough rinsing with Tris buffer, 37 µL of CT solution was injected into the channels at a flow rate of 1.5 mL/h, followed by alternating rinsing 332

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

steps (5 min) with injection of primary and Alexa 532-tagged secondary antibodies (2 min each). Thus, the total time for injection steps was only 6 min, and the total rinsing time was ∼20 min. The slightly longer rinsing times were chosen because the dead volume of the entire system, including tubing, injectors, and connectors, was much larger than the volume of the microchannels. Optimization studies showed that reconstitution of ∼5 mol % GM1 in PC gave optimal signal strength (data not shown). These results are reasonable, considering that receptor concentrations of 10% and higher in SPBs have been found to increase domain formation rather than homogeneous distribution.44 The antibody optimization experiments were carried out in a two-step procedure. First, the tagged secondary IgG dilution was fixed while the primary antibody was varied. Then the primary antibody dilution was fixed at the optimum while the tagged secondary IgG concentration was varied. The results showed that the best signal/ background values were obtained with high concentrations of primary antibody and lower concentration of secondary antibody. This is probably due to the chip design, which limited the washing efficiency and might allow a small amount of tagged secondary antibody to remain in unwashed pockets after rinsing. (44) Yuan, C. B.; Johnston, L. J. Biophys. J. 2001, 81, 1059-1069.

The standard deviation in analysis is excellent considering that a microfluidic immunoassay on silica was reported to have up to 30% RSD46 due to hydrodynamic variability. Although this problem (caused by differences in needle position) was also large in our case because of manual insertion, we managed to achieve an average 10% RSD. The ∼25-min assay time is at least several hours faster than conventional stopped flow immunoassays, which typically consume an hour for each incubation period and often have to be run overnight to achieve maximum sensitivity.

Figure 5. Fluorescence intensity as a function of CT concentrations on the SBM-functionalized PDMS chips. The flow rate is 1.5 mL/h.

Figure 5 shows the fluorescence intensity vs CT concentration. The response has a dynamic range over nearly 3 orders of magnitude, with saturation occurring at concentrations higher than ∼2 µg/mL. A detection limit of 8 fmol of CT was obtained on the basis of the S/N at the lowest concentration tested. This corresponds to a concentration of ∼210 pM, an almost 2 orders of magnitude improvement over another protein immunoassay on PDMS15 and lower than a GM1 array immunosensor for CT on glass.45 It is significant that this detection limit was achieved with a fluorescent secondary antibody without enzyme amplification. We attribute the higher sensitivity to improved control of lipid membranes on PDMS as a result of systematic surface characterization. The large inlet and outlet ports have about the same surface area as the entire channel and were likely the limiting factor for detection at lower concentrations. An attempt was made to increase the molecular interaction time by decreasing the flow rate to 1 mL/h, but the sensitivity actually decreased, possibly due to less efficient mass transport (data not shown). Two control experiments were performed to verify that the signal observed was from the specific interaction of CT with the GM1 receptors. In the first experiment, 1 µg/mL CT was injected into PC-treated channels without GM1 receptors, followed by binding assay with antibodies. The fluorescence intensity was similar to those in the blank experiments with GM1 receptor-functionalized membranes but without CT injection. This verified that the signal in the CT assay was not the result of nonspecific adsorption to the membrane. In the second control, an immunoassay was performed with 1 µg/mL BSA and rabbit anti-BSA serum on PC-GM1-treated channels. The result was also similar to the background level in the blank experiments and confirmed that BSA does not bind nonspecifically to the GM1 receptors or the membrane. (45) Rowe-Taitt, C. A.; Cras, J. J.; Patterson, C. H.; Golden, J. P.; Ligler, F. S. Anal. Biochem. 2000, 281, 123-133.

CONCLUSIONS The assembly of SBMs on PDMS is an inexpensive and easy method for surface functionalization, as compared to multistep protocols, such as grafting or multilayers. The results shown above imply that self-assembly of phospholipids to form SBMs is an ideal alternative for treatment of PDMS for several reasons. First, the hydrophilic property is surprisingly stable. As shown in the control experiments, BSA as a blocking reagent does not come close to matching the hydrophilic characteristic of PC-treated surfaces. We noticed that even months after treatment, the lipid-treated channels easily load with water when dipped into a drop, whereas untreated channels are extremely resistant to filling. This property will not only improve wetting and reduce bubbles in pressurebased analysis, but is also ideal for capillary action-based devices that utilize hydrophilic/hydrophobic properties to control fluid flow. Second, PDMS surface treatment can substantially reduce protein adsorption. We have shown that plasma oxidation results in a hydrophilic surface but actually increases nonspecific adsorption, as compared to native PDMS. In contrast, PC-treated surfaces show little or almost background level fluorescence signal for 0.1 mg/mL avidin and BSA exposure. Since this concentration is much higher than levels typically encountered in analysis, it should be possible to use the SBMs for most chip-based techniques. Electrostatic contributions were found to play an important role in the interactions between lipid headgroups and charged proteins. The positively charged lipid DOPC+ is only ideal for selective rejection of positively charged proteins. The zwitterionic PC, on the other hand, has the most ideal resistance against a wide range of proteins. In addition, the protein resistance property for SBMs is stable against flowing streams, making the membranes suitable for longer periods of analysis. We speculate that the laminar flow profile in the channels does not present much damage to the SBMs. Turbulent flow, however, could lead to some degree of membrane damage. This is supported by the observation of high background at the needle inlet regions where turbulent flow may have occurred. The immunoassay developed here makes several significant improvements over previous work. It detects much lower concentrations of proteins, as compared to other reported immunoassays on PDMS. The analysis can be completed in less than 30 min, and the reproducibility of the measurements is relatively high. Future work will focus on array patterning to obtain higher throughput and better detection limits and use of new lipid (46) Yakovleva, J.; Davidsson, R.; Lobanova, A.; Bengtsson, M.; Eremin, S.; Laurel, T.; Emneus, J. Anal. Chem. 2002, 74, 2994-3004.

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

333

materials to increase the durability and functionality of the membrane interface.

assistance with the CLSM and the financial support from UC Riverside and Eli Lilly Young Analytical Chemist Grant.

ACKNOWLEDGMENT The authors thank Dr. Tim Kingan from the UCR Genomics Institute for helpful suggestions and criticism. We also thank Dr. David Carter of the UCR Center for Plant Cell Biology for his

Received for review April 30, 2004. Accepted October 14, 2004.

334

Analytical Chemistry, Vol. 77, No. 1, January 1, 2005

AC049356+