Anal. Chem. 2004, 76, 2055-2061
Surface Modification of the Channels of Poly(dimethylsiloxane) Microfluidic Chips with Polyacrylamide for Fast Electrophoretic Separations of Proteins Deqing Xiao, Thai Van Le, and Mary J. Wirth*
Department of Chemistry and Biochemistry, University of Delaware, Newark, Delaware 19716
The electrophoresis of proteins was investigated using poly(dimethylsiloxane) (PDMS) microfluidic chips whose surfaces were modified with polyacrylamide through atomtransfer radical polymerization. PDMS microchips were made using a glass replica to mold channels 10 µm high and 30 µm wide, with a T-intersection. The surface modification of the channels involved surface oxidation, followed by the formation of a self-assembled monolayer of benzyl chloride initiators, and then atom-transfer radical polymerization to grow a thin layer of covalently bonded polyacrylamide. The channels filled spontaneously with aqueous buffer due to the hydrophilicity of the coating. The resistance to protein adsorption was studied by open-channel electrophoresis for bovine serum albumin labeled with fluorophor. A plate height of 30 µm, corresponding to an efficiency of 33 000 plates/m, was obtained for field strengths from 18 to 889 V/cm. The lack of dependence of plate height on field strength indicates that there is no detectable contribution to broadening from adsorption. A 2- to 3-fold larger plate height was obtained for electrophoresis in a 50-cm polyacrylamide-coated silica capillary, and the shape of the electropherogram indicated the efficiency is limited by a distribution of species. The commercial capillary exhibited both reversible and irreversible adsorption of protein, whereas the PDMS microchip exhibited neither. A separation of lysozyme and cytochrome c in 35 s was demonstrated for the PDMS microchip. Microfluidic chips for electrophoretic separations have been developing rapidly in recently years. The advantages of microfuidic separation over conventional CE separation are faster analysis time, smaller device size, disposability, lower consumption of chemicals, and lab-on-a-chip versatility.1,2 Many fabrication materials are being explored, including glass,3 silica,4,5 polycarbonate,6,7 * Corresponding author address: Department of Chemistry, University of Arizona, 1306 E. University Blvd., Tucson, AZ 85721-0041. Phone: 520-626-0969. Fax: 520-621-8407. E-mail:
[email protected]. (1) Jakeway, S. C.; de Mello, A. J.; Russell, E. L. Fresenius’ J. Anal. Chem. 2000, 366, 525-539. (2) Chovan, T.; Guttman, A. Trends Biotechnol. 2002, 20, 116-122. (3) Jacobson, S. C.; Hergenroder, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 1114-1118. 10.1021/ac035254s CCC: $27.50 Published on Web 02/25/2004
© 2004 American Chemical Society
poly(methyl methacrylate),6,8 and poly(dimethylsiloxane).9 The use of polymeric materials has made the fabrication of microchips easier, because once a master is created, polymeric microchips can be made rapidly and inexpensively. Poly(dimethylsiloxane) (PDMS) is an attractive material for its excellent optical transparency, easy replica molding, nontoxicity and biocompatibility.10-12 It also has the property of sealing either reversibly or irreversibly to many materials at room temperature.13 PDMS has been used to make microfluidic devices for many applications, such as biochemical assays, capillary electrophoresis, chemical reactions, and surface patterning.9,13 PDMS has been successfully used for electrophoretic separations of amino acids14 and DNA.15 The main hindrance to protein separation in PDMS is the same as that for any material used for protein separations: surface adsorptivity. The surface of PDMS is hydrophobic in its natural state, which causes physisorption of proteins,16,17 interfering with the electrophoretic separation of proteins. Additionally, pressure must be used to fill hydrophobic channels with aqueous buffer, and this is hampered by bubble formation due to the permeability of PDMS.18 The surface can be oxidized to become somewhat (4) Jacobson, S. C.; Moore, A. W.; Ramsey, J. M. Anal. Chem. 1995, 67, 20592063. (5) Kirby, B. J.; Wheeler, A. R.; Zare, R. N.; Fruetel, J. A.; Shepodd, T. J. Lab Chip 2003, 3, 5-10. (6) Ross, D.; Locascio, L. E. Anal. Chem. 2003, 75, 1218-1220. (7) Henry, A. C.; Tutt, T. J.; Galloway, M.; Davidson, Y. Y.; McWhorter, C. S.; Soper, S. A.; McCarley, R. L. Anal. Chem. 2000, 72, 5331-5337. (8) Wabuyele, M. B.; Ford, S. M.; Stryjewski, W.; Barrow, J.; Soper, S. A. Electrophoresis 2001, 22, 3939-3948. (9) McDonald, J. C.; Duffy, D. C.; Anderson, J. R.; Chiu, D. T.; Wu, H. K.; Schueller, O. J. A.; Whitesides, G. M. Electrophoresis 2000, 21, 27-40. (10) Gawron, A. J.; Martin, R. S.; Lunte, S. M. Electrophoresis 2001, 22, 242248. (11) Ren, X. Q.; Bachman, M.; Sims, C.; Li, G. P.; Allbritton, N. J. Chromatogr., B 2001, 762, 117-125. (12) Martin, R. S.; Gawron, A. J.; Lunte, S. M.; Henry, C. S. Anal. Chem. 2000, 72, 3196-3202. (13) McDonald, J. C.; Whitesides, G. M. Acc. Chem. Res. 2002, 35, 491-499. (14) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974-4984. (15) Effenhauser, C. S.; Bruin, G. J. M.; Paulus, A.; Ehrat, M. Anal. Chem. 1997, 69, 3451-3457. (16) Linder, V.; Verpoorte, E.; Thormann, W.; de Rooij, N. F.; Sigrist, M. Anal. Chem. 2001, 73, 4181-4189. (17) Ng, J. M. K.; Gitlin, I.; Stroock, A. D.; Whitesides, G. M. Electrophoresis 2002, 23, 3461-3473. (18) Hagg, M. B. J. Membr. Sci. 2000, 170, 173-190.
Analytical Chemistry, Vol. 76, No. 7, April 1, 2004 2055
hydrophilic due to surface silanols and carboxylic acids, allowing filling to occur by capillary action, but there is gradual hydrophobic recovery over a period of hours.19 The net negative charge of the oxidized surface induces electroosmotic flow and electrostatic interactions with positively charged proteins, both of which interfere with electrophoresis.5,14,20,21 Bare PDMS is adsorptive toward proteins by virtue of its hydrophobicity. Proteins have been separated in silica capillaries that had been coated with a film of PDMS, which in turn was coated with surfactants, and efficiencies as high as 150 000 plates/m were obtained.22 Adsorption of surfactants has the advantage of being a simple way to chemically modify a surface, but they also impart a charge to the surface to give electroosmotic flow, which complicates the design of the separation and adds to Joule heating. There has been only one report, to our knowledge, of a protein separation using a PDMS chip: a negatively charged bovine anhydrase II ladder was separated in an oxidized PDMS channel, where the negative charge of the oxidized surface minimizes protein adsorption through Coulombic repulsion.14 In this same paper, the separation of positively charged proteins was achieved by adsorption of a cationic polymer to the surface of PDMS. For the separation of both positively and negatively charged proteins, the apparent efficiencies were on the order of 40 000 plates/m, which corresponds to a plate height of 25 µm. The actual efficiencies are likely much higher but are difficult to gauge because protein samples are typically heterogeneous due to posttranslational modification as well as variable positions and numbers of fluorescent labels. It would be valuable to be able to modify the surface of PDMS to make it permanently hydrophilic, neutral, and resistant to protein adsorption. This would hasten the filling of the channels with buffers, avoid electroosmotic flow to allow either cationic or anionic proteins to be separated, and allow high electric fields for fast separations. We previously demonstrated the controlled growth of polyacrylamide chains by atom-transfer radical polymerization on silica, showing that this achieves a brush-like layer of polyacrylamide that is dense enough to avoid the proteinpolymer chain entanglement that underlies proteins adsorption.23,24 This layer of polyacrylamide allows efficient size-exclusion chromatography of proteins 23 and efficient capillary electrophoresis.25 We adapted this chemistry to PDMS and demonstrated that the chemical modification of PDMS sheets gives a thin film of polyacrylamide on its surface.19 The PDMS became hydrophilic for an indefinitely long period of time, at least for weeks, and the irreversible adsorption of the protein lysozyme was reduced by a factor of 20, as compared to the native PDMS surface.19 The experiment did not test for reversible adsorption, which can also affect electrophoresis of protein. The combined advantages of hydrophilicity and reduced irreversible adsorption of protein make this coating an attractive candidate for protein electrophoresis. (19) Xiao, D. Q.; Zhang, H.; Wirth, M. Langmuir 2002, 18, 9971-9976. (20) Dolnik, V. Electrophoresis 1999, 20, 3106-3115. (21) Ostuni, E.; Grzybowski, B. A.; Mrksich, M.; Roberts, C. S.; Whitesides, G. M. Langmuir 2003, 19, 1861-1872. (22) Badal, M. Y.; Wong, M.; Chiem, N.; Salimi-Moosavi, H.; Harrison, D. J. J. Chromatogr., A 2002, 947, 277-286. (23) Huang, X. Y.; Wirth, M. J. Anal. Chem. 1997, 69, 4577-4580. (24) Huang, X.; Wirth, M. J. Macromolecules 1999, 32, 1694-1696. (25) Huang, X. Y.; Doneski, L. J.; Wirth, M. J. Anal. Chem. 1998, 70, 40234029.
2056
Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
The purpose of this paper is to apply atom-transfer radical polymerization of acrylamide to coat the channels of PDMS microchips and study the electrophoretic behavior of proteins to determine whether the potential advantages of hydrophilicity and resistance to protein adsorption can be accrued. EXPERIMENTAL SECTION Chemicals. The separation buffer in all experiments was sodium phosphate (Aldrich) at 50 mM and pH 2.15. The labeling reagent, tetramethylrhodamine isothiocyanate (TRITC) isomer5, and bovine serum albumin conjugated with TRITC were purchased from Molecular Probes (Eugene, OR). Sylgard 184, and the curing agent were obtained from Dow Corning. Acrylamide (electrophoresis grade, 99.9%) was obtained from Fisher Scientific. CuCl (99+%) was obtained from Aldrich, and CuCl2 (anhydrous, 99+%) was obtained from Acros. Formic acid (reagent grade ACS 88+%) and tris(2-aminoethyl)amine (96%) were obtained from Acros. Tris[2-(dimethylamino)ethyl]amine, referred to as Me6TREN, was prepared from tris(2-aminoethyl)amine and stored under argon. 1-Trichlorosilyl-2-(m-p-chloromethylphenyl)ethane was obtained from Gelest (Tullytown, PA). All other solvents were from Fisher Chemicals. The proteins were obtained from Sigma. Lysozyme and cytochrome c were labeled with TRITC according to a published procedure.26 The water used in this work was purified to a resistance of 18 MΩ-cm using a commercial purification system (Barnstead E-pure). Microchip. A glass master with eight microchip patterns positively embossed was provided by Dr. J. Michael Ramsey’s group, Oak Ridge National Laboratory. The height of the pattern is 10 µm and the width at one-half height is 30 µm. Sylgard 184 PDMS prepolymer was mixed thoroughly with its cross-linking catalyst at 10:1 (w/w) and degassed by vacuum for 1 h. The mixture was then cast against the glass mold and cured at 110 °C for 8 h. The glass master was derivatized with chlorotrimethylsilane to prevent the covalent attachment of PDMS. The PDMS sheet was then peeled from the master, and microchips of 2 × 1 in. were cut from the sheet. Four 1/8-in. holes were punched on each of the channels, and four glass reservoirs were glued at these holes using 5-min epoxy. A 1 × 3-in. silica plate was used as the cover plate. Surface Modification. The details of the surface modification are described in our previous work.19 A T10 × 10/OES UV/Ozone Cleaning System (UVOCS Inc., PA) was used to generate the ozone plasma to oxidize the surface of PDMS. The PDMS microchip was placed at a distance of 7.5 mm from the UV lamp, which was a low-pressure quartz-mercury vapor lamp with emission lines at 254 and 185 nm and an average UV intensity of 10 mW/cm2. The chip was covered with a glass microscope slide to block direct exposure to the UV to prevent bulk damage to the PDMS. The channels were facing down and were oxidized for 15 min, which we previously determined to be the optimal time of exposure.19 Immediately after oxidation, the PDMS microchip was sealed with a 1-trichlorosilyl-2-(m-p-chloromethylphenyl)ethanederivatized glass microscope slide and placed in a Schlenk flask. A small vial containing 0.2 mL of 1-trichlorosilyl-2-(m-p-chloromethylphenyl)ethane was put into the flask at the same time. The flask was filled with argon, a vacuum of 10 mmHg was applied, (26) Brinkley, M. Bioconjugate Chem. 1992, 3, 2-13.
Figure 1. Schematic of PDMS microfluidic chip, with silica cover plate, reservoirs, position of microscope objective for fluorescence excitation and detection, and electronics.
and the flask was put into an oven at 90 °C for 0.5 h to vaporize the silane for diffusion into the microchannels for reaction with the surface of the PDMS channels. The flask was then cooled to room temperature and the PDMS was peeled from the glass slide and sealed by a silica plate coated with polyacrylamide. The procedure of coating polyacrylamide on the silica slide was described previously.27 For coating the PDMS channels, the sealed microchip was charged into a Schlenk flask and cycled between vacuum and argon three times. One of the reservoirs was filled with the polyacrylamide reaction solution using a syringe. Because of the hydrophilic nature of the polyacrylamide-coated silica surface, the aqueous solution of reagents was able to fill the channels by capillary force within 0.5 h. The aqueous solution of reagents for atom-transfer radical polymerization ATRP of acrylamide was prepared according to a published procedure,19 which is briefly described here. A solution of 0.3 M acrylamide was made oxygen-free by purging with argon, and the reagents were added with the concentrations ratios of acrylamide/Cu(I)Cl/Cu(II)Cl/ Tris[2-(dimethylamino)ethyl]amine ) 100:1:0.1:1. The polymerization reaction was allowed to proceed for 10 h. The microchip was then washed with DI water, dried under nitrogen flow, and then sealed again with the same polyacrylamide-silica plate. The chip was then ready for use in electrophoresis or for storage. Instrumentation. A schematic diagram of the experimental setup is shown in Figure 1. The channels in the PDMS chip are tapered to open out to the reservoirs. The injection and separation channels near the T-intersection are 30 µm wide. The distances from the beginnings of taper to the T-intersection are 1.0 cm from reservoirs A, B, and C and 3.5 cm for reservoir D. The separation path length from the T-intersection to the detection region toward reservoir D is 3.5 cm. A silica cover plate was used to minimize deformation of the soft PDMS. A plexiglass cover, not shown, was used to hold the electrodes in place and prevent accidental electric shock. A control station was built that had four high voltage supplies (Ultra Volt, Inc., Ronkonkoma, NY; model no. 10A12-P4C) and four high-voltage relays (Kilovac Corporation, Santa Barbara, CA; K81c Series). The high voltage supplies, as well as the relays, were controlled independently by a computer with the Labview program (version 6i, National Instruments, Austin, TX). The output of each power supply could be varied from 0 to 4000 (27) Xiao, D. Q.; Wirth, M. J. Macromolecules 2002, 35, 2919-2925.
V, switched to earth ground through a relay or held at open-circuit to prevent current flow. The outputs of the power supplies were connected to four platinum electrodes, which were inserted in the four reservoirs of the microfluidic chip. Excitation, detection, and imaging of fluid flow in the channels were accomplished using an inverted Leica DM-IRM microscope (Leica, Wetzlar, Germany) with TRITC filter cube (Omega Optical, Brattleboro, VT). The filter cube consisted of two filters (535DF35, 590DF35) and a dichroic mirror (570DRLP) that allowed the passage of excitation and emission beams. A 10× objective with a numerical aperture of 0.25 was used for focusing light from a mercury lamp (12 V, 100 W, Leica, Wetzlar, Germany) into the microfluidic channel and for collecting fluorescence. Fluorescence images were captured by an ICCD camera (I-Pentamax-512EFTGIII, Roper Scientific, Trenton, NJ) with thermoelectric cooling, for which Winview 32 (Roper Scientific) was used as the software. To monitor the injection and detection processes, movies were obtained for acquisition times of 0.1 s/frame. To obtain electropherograms, movies were obtained for acquisition times of 1.0 s/frame, and the data were exported as text files for analysis by a program written in-house for MATLAB 6.5. The program calculates and plots fluorescence intensity as a function of time for selected pixels. Electrophoresis of Proteins. All solutions were degassed for 10 min before use. For microchip injections, floating sample loading was used,28 in which the sample was driven by electrophoretic migration from the analyte reservoir (A) to the analyte waste reservoir (B), while the buffer and waste reservoirs (C and D, respectively) were held at open circuit. For separation, reservoir C was held at high voltage; experiments varied this from 80 to 4000 V, the voltages for reservoirs A and B were 57% that of reservoir C, and reservoir D was earth-grounded. In switching between loading and separation, the voltage at C was held at 4000 for 0.4 s to ensure that the same sample length was injected regardless of separation voltage, after which it was set to its desired level for the separation. For detection, the light from the mercury lamp was focused to a spot 3.5 cm downstream from the T-intersection. For the capillary electrophoresis, a Beckman P/ACE System MDQ (Beckman, Fullerton, CA) was used. A BioCAP linear polyacrylamide (LPA)-coated capillary (60 cm total length, 50 cm separation length, 50 µm i.d., 360 µm o.d.; Bio-Rad, Hercules, CA; assembled in a Beckman eCAP catridge) was used for the electrophoresis of proteins. The column was rinsed with pure water for 2 min and then rinsed with phosphate buffer (pH 2.15) for 4 min. The injection time was 5 s at 1.0 psi. RESULTS AND DISCUSSION A comparison was made between PDMS microchips that were uniformly chemically modified with polyacrylamide and those chemically modified only in the channels. First, a PDMS microchip modified uniformly over its surface was sealed to a silica plate, also modified with polyacrylamide over its entire surface. The two specimens were found to seal tightly and reversibly after light compression. When the channels were filled with aqueous buffer, the seal was gradually permeated by water from the channels and eventually exhibited leakage during electrophoresis. Second, the (28) Jacobson, S. C.; Hergenroder, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113.
Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
2057
Figure 2. Fluorescence images of the injected zone in the PDMS microchip, surface-modified by polyacrylamide, at varying times after application of the separation voltage. The width of the injected zone is 120 um.
PDMS microchip chemically modified with polyacrylamide only in its channels was found also to seal tightly and reversibly after light compression to the PDMS microchip. In this case, there was no leakage of water during electrophoresis. In both cases, the filling of the channels occurred spontaneously within several minutes by filling one reservoir with buffer, which is attributed to the hydrophilicity of the surfaces after modification with polyacrylamide. It is concluded that surface modification only in the channels puts the hydrophilicity where it is needed without inviting water transport where it is not desired. The performance of the chemically modified PDMS microchip was studied by the electrophoresis of bovine serum albumin (BSA) labeled with TRITC. BSA was chosen because it tends to adsorb significantly to a variety of surfaces. The pH of the buffer was set to 2.15 to maximize the speed of electrophoresis. Figure 2 shows fluorescence images of the zones in the channels as a function of time after injection. These images establish that spatially narrow zones are injected cleanly into the channel. The widths of the zones are somewhat larger than the 30-µm width of the Tintersection due to diffusion into the separation channel during sample loading. This increases the contribution of the injected width to the plate height, as described by eq 1,
linj2/12 Hinj ) L
(1)
which gives the relation between of the width of the injected zone, l, for a rectangular zone and the plate height, Hinj, for separation length, Linj. Since linj ) 120 µm, as per Figure 2, then Hinj ) 0.35 µm for the separation length of 3.5 cm. This is a satisfactorily low contribution; therefore, no further attempts to reduce the injected width were made. Figure 3 shows the electropherograms of the BSA sample for separation voltages ranging from 1 to 4 kV applied over a distance of 4.5 cm. For each peak, the fluorescence returns to baseline, indicating no detectable irreversible adsorption within the noise of the baseline. Three replicate injections were made at each applied voltage, with no indication of irreversible adsorption. Peak tailing can be seen on these electropherograms. Peak tailing can be caused by a variety of factors, one of which is from reversible adsorption, which is the factor most relevant to this investigation. Peak tailing associated with reversible adsorption would occur 2058 Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
Figure 3. Electropherograms of TRITC-labeled bovine serum albumin in the PDMS microchip, surface-modified by polyacrylamide, at varying applied voltages over 4.5 cm and a separation length of 3.5 cm. The electropherograms are artificially offset for display.
Figure 4. The same electropherograms as in Figure 3, except that these are plotted vs time‚kV to make the peaks nearly coincide, allowing comparison of widths and shapes. The individual data points are shown to clarify the limits in knowing the shape of the peak.
either if there were a rare site that is so adsorptive that it saturates, or if there were extremely slow desorption. In either event, peak tailing from reversible adsorption would cause either the width or the asymmetry of the zone to increase with voltage. Figure 4 shows the electropherograms replotted on a scale normalized for voltage. The peaks nearly line up, as expected. The small discrepancies in the exact positions could be due to changes in channel resistance from the flexibility of the PDMS. It is apparent from this plot that the tailing does not increase with migration rate. In fact, the peak shapes and widths on this normalized scale are constant within experimental error. This indicates qualitatively that reversible adsorption does not contribute to the electrophoretic zone. A more quantitative assessment of the contribution of adsorption/desorption to zone width can be achieved by plot of the plate height, H, vs migration velocity, u. These are shown in the lower portion of Figure 5. The plate height was computed for these
Figure 6. White-light images at various positions of the glass master used to mold the microfluidic channel.
Figure 5. Van Deemter plots for capillary electrophoresis (b) and microchip electrophoresis (O), where plate height is plotted vs migration velocity of TRITC-labeled bovine serum albumin.
asymmetric peaks as if the width at half-height were that for a Gaussian peak. The result is that H ) 30 ( 4 µm, independent of voltage. The plate height normally varies with migration velocity, u, according to eq 2.
H)
linj2 l2det 2D + +A+ + Cu 12L 12L u
(2)
The first two terms are the voltage-independent contributions from nonzero injection and detection lengths, and the third term, A, is the voltage-independent contribution due to any irregularities that give rise to differences in path length. Voltage-dependent terms are from diffusion, D, and kinetics of adsorption and desorption, which are contained in the C term. There can also be a steep rise at high applied voltage due to Joule heating, which is not accounted for in eq 2. The van Deemter plot of Figure 5 shows no evidence of a change in plate height with migration velocity for the microchip, indicating that effects of diffusion are negligible compared to effects of other broadening contributions. The diffusion coefficient for moderately sized proteins, such as serum albumin, is 10-6 cm2/s or less, which would contribute less than 1 µm to H for any reasonable migration velocity. More pertinent to the problem at hand, the flat van Deemter plot establishes that effects of adsorption/desorption are negligible for the PDMS microchip chemically modified with polyacrylamide. The plate height of 30 µm is rather large, but not unusually large for a protein sample. An H of 30 µm for this 3.5-cm separation length corresponds to 33 000 plates/m. The van Deemter plot indicates that this plate height is due to a process that is independent of migration rate. The first three terms of eq 2 are independent of migration rate; therefore, they are possible contributions to the plate height. The injected width has already been eliminated as a possibility. The detected width can readily be eliminated: the detection window itself is comparable to the injection length. The integration time of the detector is 1 s, which can be estimated from Figure 2 to give a detection length that is 5-fold larger than the injection length. This is still a negligible
contribution to H. Since there is no packing material in the channel, for the A term to contribute 30 µm to the plate height would require frequent and large imperfections in the channel shape. Imperfections in the glass master used to mold the channel are evident under examination by white-light microscopy, as illustrated in Figure 6. Since the average channel depth is only 10 µm, visible imperfections could be an indication of significant local variations in channel depth. This possible contribution to plate height was not investigated quantitatively. Another possible origin of the flat van Deemter plot and the high intercept is that the softness of the PDMS allows undulations in the width of the channel, giving rise to unequal path lengths for migration in the microchannel. The effect would be exacerbated by the resulting variations in channel resistance. A large number, n, of excursions of length l from the average path length would contribute a variance to σA2, according to the random walk model.
σA2 ) l2n
(3)
Many small excursions give a Gaussian zone. The asymmetry of the zone indicates that there would be fewer and larger excursions from the average path length if channel irregularities were the origin of the A term. One more possible contribution to the flat van Deemter plot and the high intercept is that the protein sample itself is impure. The effect of the distribution would be to give an apparent large A term. If the peak width were due entirely to a distribution of species, then the zone width would be proportional to zone position. In this limit, the apparent value for N would be independent of voltage and separation length.
Napp(V) )
() tm σ
2
) constant ) L/Happ
(4)
Happ would then be proportional to separation length in this limit. We refer to these as apparent N or H because these values would be a property of the sample, not the separation channel. The possibility of a distribution can be tested by capillary electrophoresis, in which the longer capillary length would either resolve Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
2059
Figure 7. Capillary electropherogram of the same sample of bovine serum albumin and the same buffer as used for the microchip electrophoresis. A 60-cm silica capillary coated by conventional polymerization with polyacrylamide was used, with a 50-cm separation length and an applied voltage of 10 kV.
the distribution of species or give the same value of (tm/σ), or N1/2, as the microchip. Figure 7 shows a capillary electropherogram of the same BSA sample, run with the identical buffer solution, using a 50-cm separation length and an applied voltage of 10 kV over a 60-cm distance. The relatively low voltage was chosen to minimize any contribution from symmetric broadening due to reversible adsorption. The electropherogram of Figure 7 exhibits low separation efficiency, but only slight peak asymmetry that is considerably less than that for the microchip. It is possible that the distribution of species is rather symmetric with respect to migration rate, and this was tested by varying voltage to determine whether there is an A term for the capillary separation. The results are included in the plot of Figure 5 as the solid circles. The A term for the capillary separation is large, which indicates that there is a substantial distribution of species. If the A term were entirely due to the distribution, then the corresponding value of N due to A, i.e., L/A, would be the same for the capillary and microchip. For the capillary, L/A ) 8000, which is significantly larger than L/A ) 1300 for the microchip. It is concluded that the sample has a large distribution of species, but that this distribution does not explain the large A term for the microchip. The most likely explanation for the A term for the microchip remains to be the irregularities in the channel geometry that were discussed earlier. The electropherogram of Figure 7 also exhibits an increase in baseline after the peak, indicating that some BSA remains irreversibly adsorbed to the capillary. This increase in baseline was reproducible and occurred in all capillary electropherograms. It illustrates the disadvantage of conventional polymerization of acrylamide, which gives a film of low density, allowing irreversible entanglement of protein and polymer. The van Deemter plot for the capillary, included in Figure 5, clearly indicates a C term for 2060 Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
Figure 8. (a) Electropherogram demonstrating rapid separation of a mixture of TRITC-labeled proteins, lysozyme and cytochrome c, in the PDMS microchip whose surface is modified with polyacrylamide through atom-transfer radical polymerization. Separation voltages: A, 2280 v; B, 2280 V; C, 4000 V; D, ground. (b) Electropherograms for a mixture of TRITC-labeled lysozyme and cytochrome c using (a) microchip electrophoresis using a field strength of 888 V/cm and a separation length of 3.5 cm and (b) capillary electrophoresis using a field strength of 300 V/cm and a separation length of 50 cm. Fluorescence detection in the visible range was used for the microchip separation and UV absorbance was used for detection for the capillary separation.
the capillary electrophoresis. This shows that significant reversible adsorption occurs for the silica capillary coated with conventionally polymerized acrylamide. The silica capillary coated with the conventional linear polyacrylamide film thus shows both reversible and irreversible adsorption of BSA, in contrast to the PDMS microchip. The performance of the PDMS microchip might be better demonstrated with a pure sample of protein, but real samples tend to have distributions of species due to posttranslational modification as well as the fluorescence labeling. The advantage of short separation length is high speed. A fast separation of two proteins is demonstrated for this microchip in Figure 8a. This shows separation of a mixture of lysozyme and cytochrome c, both labeled with TRITC. The two proteins are nearly baseline-resolved, and the two zones are completely past the detector in 35 s. A separation of the same two proteins is shown for capillary electrophoresis in Figure 8b using an applied voltage of 15 kV, which is the manufacturer’s recommendation for the instrument and capillary. UV absorbance was used for detection, revealing that there are both labeled and unlabeled species, in contrast to
the bovine serum albumin sample. The proportion of cytochrome c was increased relative to that for the microchip separation to observe the unlabeled species better in the electropherogram. The results show that the capillary does not perform much better in the separation of the labeled proteins, but it is better in revealing that there is a distribution. Resolution increases with the square root of the number of theoretical plates; therefore, one would only expect an improvement factor of 2 in resolution for the capillary vs the microchip. The microchip provides a factor of 40 improvement in analysis time with only a factor of 2 loss in resolution. The ability to chemically modify PDMS with polyacrylamide for resistance to protein adsorption could be enabling technology for a variety of applications. The low adsorptivity of the chemically modified microchannel enables this route of fast microchip separation as a tool for 2D methods. Other potential advantages include coating of PDMS as a cover plate for glass or other microchips; coating of PDMS channel walls for electrochromatographic systems, as well as for any use of PDMS as a cover plate; and possibly medical applications, such as coating of PDMS catheters. CONCLUSIONS Surface modification of PDMS microchips with polyacrylamide, grown on the surface through atom-transfer radical polymerization,
enables rapid filling of the microchannels through capillary action. Surface modification of only the inner channels of the PDMS microchip enables reversible sealing without leakage. The surface modification also reduces both reversible and irreversible adsorption of bovine serum albumin, as compared to a commercial silica capillary coated with a conventionally polymerized layer of polyacrylamide. A separation of lysozyme and cytochrome c in 35 s is demonstrated for the PDMS microchip, also showing no evidence of adsorption. ACKNOWLEDGMENT This work was supported by the Department of Energy under the grant DE-FG02-91ER14187 and the National Institutes of Heath under grant R01 GM065980-01. We thank Dr. J. Michael Ramsey and Dr. Stephen Jacobson for the glass mold used to make the PDMS microchip.
Received for review October 23, 2003. Accepted January 18, 2004. AC035254S
Analytical Chemistry, Vol. 76, No. 7, April 1, 2004
2061
