Anal. Chem. 1997, 69, 3451-3457
Integrated Capillary Electrophoresis on Flexible Silicone Microdevices: Analysis of DNA Restriction Fragments and Detection of Single DNA Molecules on Microchips Carlo S. Effenhauser,*,† Gerard J. M. Bruin, Aran Paulus, and Markus Ehrat
Bioanalytical Research, Novartis Pharma AG, CH-4002 Basel, Switzerland
Microchips for integrated capillary electrophoresis systems were produced by molding a poly(dimethylsiloxane) (PDMS) silicone elastomer against a microfabricated master. The good adhesion of the PDMS devices on clean planar surfaces allows for a simple and inexpensive generation of networks of sealed microchannels, thus removing the constraints of elaborate bonding procedures. The performance of the devices is demonstrated with both fast separations of OX-174/HaeIII DNA restriction fragments labeled with the intercalating dye YOYO-1 and fluorescently labeled peptides. Detection limits in the order of a few zeptomoles (10-21 mol) have been achieved for each injected DNA fragment, corresponding to a mass detection limit of ∼2 fg for the 603 base pair fragment. Single λ-DNA molecules intercalated with YOYO-1 at a base pair-to-dye ratio of 10:1 could be detected with an uncomplicated laser-induced fluorescence detection setup. High single-molecule detection efficiency (>50%) was achieved under electrophoretically controlled mass transport conditions in PDMS microchannels. The concept of miniaturized total chemical analysis systems (µ-TAS) continues to attract considerable attention in both industrial and academic institutions worldwide.1 This approach offers a novel way to achieve fast separations with high resolution in a miniaturized setup including sample pretreatment steps such as sample concentration, labeling, and digestion with the additional possibility of multiplexing. Particularly, capillary electrophoresis (CE) on chips or integrated capillary electrophoresis (ICE) has been in the center of interest. The success of CE in the analytical community during the last 10 years fosters hopes that ICE could follow in the footsteps of CE by adding a new level of miniaturization and, by leaning on technologies developed for the microelectronics industry, multiplication. Over the last five years, a number of functional models with impressive performance features regarding analysis speed and resolution,2-6 the integration of pre- and postcolumn reactions,7-9 and fraction collection10 have been demonstrated. However, † E-mail:
[email protected]. Fax: +41-616964504. (1) Manz, A.; Harrison, D. J.; Verpoorte, E.; Widmer, H. M. Adv. Chromatogr. 1993, 33, 1-66. (2) Harrison, D. J.; Fluri, K.; Seiler, K.; Effenhauser, C. S.; Manz, A. Science 1993, 261, 895-7. (3) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 263742.
S0003-2700(97)00391-0 CCC: $14.00
© 1997 American Chemical Society
despite these very appealing and promising developments, a number of critical issues such as the choice of the support material, detection schemes and sensitivity, injection, fluid handling, and overall cost still prevents a wider distribution of these devices in an academic and industrial environment. Early approaches to capitalize on the processing technology of the chip industry and to use silicon wafer were hampered by the unfavorable electrical characteristics of this material for electrophoretic experiments.11 However, glass and quartz can be micromachined in a similar way by using established and commercially available techniques. The optical properties of glass allow spectroscopic detection schemes such as laser-induced fluorescence and even UV detection12 to be employed. Because the bulk resistivity and dielectric breakdown field strength are sufficiently high, electrical field strengths of several thousand volts per centimeter can be readily applied in low-conductivity buffer solutions. Last but not least, there exists an arsenal of surface modification methods that can be easily transferred from conventional CE to planar chips. However, the fabrication of a readyto-use glass device requires a sequence of rather cumbersome steps, such as annealing of the chip to a cover plate (typically 600 °C for several hours), into which reservoir holes have to be drilled in a time- and labor-intensive way to provide access to the channels. In our experience, the devices are prone to clogging, yet difficult to clean and very often have to be discarded after failure. Besides, the fabrication costs are presently too high in order to allow usage as single-use disposable units. Ekstro¨m et al. addressed the choice of the support material in a patent by using plastic support materials to form the necessary channel structures.13 However, the rather large channel dimensions in this early work prevented high-resolution, high-speed applications. (4) Effenhauser, C. S.; Paulus, A.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 2949-53. (5) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 1114-8. (6) Woolley, A. T.; Mathies, R. A. Anal. Chem. 1995, 67, 3676-80. (7) Jacobson, S. C.; Koutny, L. B.; Hergenro ¨der, R.; Moore, A. W.; Ramsey, J. M. Anal. Chem. 1994, 66, 3472-6. (8) Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1996, 68, 720-3. (9) Fluri, K.; Fitzpatrick, G.; Chiem, N.; Harrison, D. J. Anal. Chem. 1996, 68, 4285-90. (10) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1995, 67, 22847. (11) Harrison, D. J.; Glavina, P. G.; Manz, A. Sens. Actuators B 1993, 10, 10716. (12) Liang, Z.; Chiem, N.; Ocvirk, G.; Tang, T.; Fluri, K.; Harrison, D. J. Anal. Chem. 1996, 68, 1040-8. (13) Ekstro ¨m, B.; Jacobson, G.; Oehman, O.; Sjoedin, H. EP 0527905, 1995.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997 3451
Figure 2. Electron micrograph showing the injection region of the PDMS microdevice. Channel cross section, 50 µm (width) × 20 µm (height). The geometrical features of the channel junction result from the anisotropy of the wet chemical etching process. The small defects visible at the bottom of the channel are due to artifacts from the gold sputtering process.
Figure 1. Fabrication procedure of PDMS chips for capillary electrophoresis: (a) silicon master wafer with positive surface relief, (b) premixed solution of Sylgard 184 and its curing agent poured over the master, (c) cured PDMS slab peeled from the master wafer, (d) PDMS slab with punched reservoir holes, and (e) ready-to-use device placed on a slab of PDMS.
The shortcomings of glass devices as substrates for ICE inspired us to explore alternative materials. Casting of poly(dimethylsiloxane) (PDMS) from microfabricated masters has recently found an increasing number of applications, e.g., in microcontact printing,14 micromolding in capillaries (MIMIC),15 and microfluidic networks for surface patterning (µFN).16 In this paper, we describe the use of PDMS support material for ICE applications. By designing and building a LIF detector based on an inverted microscope, we intended to explore detection limits in combination with the use of PDMS devices. EXPERIMENTAL SECTION Fabrication of the Silicone Microchip Devices. The sequence of fabrication steps of the silicone chips is schematically depicted in Figure 1. Silicon wafers [4-in. (100)] carrying a positive surface relief served as molding templates and were manufactured at the Institut fu¨r Mikro- und Informationstechnologie (D-78052 Villingen-Schwenningen, Germany) using a wet etching process (Figure 1a). The silicon wafers were silanized in 3% (v/v) dimethyloctadecylchlorosilane/0.025% H2O in toluene for 2 h to facilitate peeling off the PDMS replica. After silanization, the wafers were thoroughly rinsed with toluene and water. (14) Wilbur, J. L.; Kumar, A.; Kim, E.; Whitesides, G. M. Adv. Mater. 1994, 6, 600-604. (15) Kim, E.; Xia, Y.; Whitesides, G. M. Nature 1995, 376, 581-4. (16) Delamarche, E.; Bernard, A.; Schmid, H.; Michel, B.; Biebuyck, H. Science 1997, 276, 779-81.
3452 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
The master wafer was then carefully cleaned with a stream of dry nitrogen and mounted in a clamp. A 10:1 mixture of Sylgard 184 (Dow Corning, Midland, MI) and its curing agent was poured over the wafer (Figure 1b). After curing of the PDMS for 4 h at 65 °C, the PDMS replica could be easily peeled off the master (Figure 1c). Holes serving as reservoirs and providing access to the channels were punched through the bulk material (Figure 1d). The device was then placed on a thin slab of PDMS in order to form a closed channel system of four equivalent walls (Figure 1e). The depth of the PDMS channels is 20 µm, the corresponding width varied between 10 and 50 µm depending on the channel layout. The angle between the side walls of the relief structures and the wafer surface is very close to 90°, however, due to the anisotropic KOH-based etching process without corner compensation, channel intersections were limited in shape by the slowest etching (111) planes. This lead to an additional geometrically defined volume of 25 pL at each corner of an intersection in the PDMS replicas, as depicted in Figure 2. The size of the channel intersection volume shown in Figure 2 is 150 pL. As can be seen, the features of the molded microchannels are very well defined, and the smoothness of the channel walls reflects the high quality of the master surface. Chip Preparation. After placing the chip on a thin slab of PDMS or on glass support, hermetically sealed microchannels were readily formed by mere adhesion without applying external force. Buffer solution and separation media were pipetted into the reservoirs, and the microchannels were filled by applying vacuum to one of the reservoirs. Alternatively, pressures up to 1 bar could be applied without the need of clamping the chip against the support. Sample solution was pipetted into one of the reservoirs, and platinum electrodes were dipped into all reservoirs. If a device became clogged, it was simply peeled off the support, rinsed with water, and used again after drying with a nitrogen gas stream. Material and Reagents. φX-174/HaeIII restriction fragments and λ-DNA were purchased from Pharmacia (Uppsala, Sweden). Tris(hydroxymethyl)aminomethane (Tris), boric acid, sodium chloride, toluene, and dimethyloctadecylchlorosilane were pur-
Figure 4. Dimensions and layout of two PDMS microchips used in this work. The figures in circles indicate the reservoir numbers: (1) sample; (2) buffer; (3) injection waste; (4) waste. The small figures denote the channel lengths and device dimensions (in mm). Channel cross sections: (a) 50 µm (width) × 20 µm (height); (b) 10 µm (width) × 20 µm (height).
Figure 3. Schematic representation of the experimental setup and the detection arrangement.
chased from Fluka (Buchs, Switzerland). YOYO-1 was obtained from Molecular Probes (Leiden, The Netherlands), hydroxypropylcellulose (HPC, average MW 100 000) from Aldrich (Steinheim, Germany), and 9-aminoacridine from ICN Biomedicals (Aurora, OH) Fluorescein-labeled peptides were synthesized in-house. DNA Labeling. A 200 µL sample of a 6.5 ng/µL solution of φX-174/HaeIII restriction fragments in 100 mM Tris/100 mM boric acid was added to 800 µL of 2.5 × 10-7 M YOYO-1 in 100 mM Tris/100 mM boric acid buffer. The resulting concentration in terms of DNA base pairs (bp) was 2 µΜ and 200 nM with respect to YOYO-1 (nominal bp-to-dye ratio, 10:1). This bp-todye ratio is in the optimal range with regard to sensitivity and separation performance according to ref 17. The dye-DNA complexes were kept for at least 30 min in the dark at room temperature before injection. λ-DNA was labeled in a similar manner with the same nominal bp-to-dye ratio as described above. Dilutions were made in 10 mM Tris/10 mM boric acid, pH 8.4. For the single-molecule detection (SMD) experiments, the background electrolytes were filtered through sterile 0.22 µm pore size filters. All solutions were vortexed for at least 30 s prior to use or further dilution. Laser-Induced Fluorescence Detection. The setup of the LIF detector is schematically shown in Figure 3. The PDMS chip assembly was mounted on the X-Y translational stage of an inverted microscope (Zeiss Axiovert 100), which also served as a platform of a confocal laser-induced fluorescence detection setup. The 488 nm output of an Ar ion laser (Omnichrome, Model 532 AP) was coupled into a single-mode fiber equipped with a focusing collimator unit (MB 04, Spindler & Hoyer, Go¨ttingen, Germany). Excitation light passing through the rear port of the microscope was reflected from a dichroic mirror (Omega 505DRLP02), and focused onto the microchannel by means of a microscope objective (17) Figeys, D.; Arriaga, E.; Renborg, A.; Dovichi, N. J. J. Chromatogr., A 1994, 669, 205-16.
(Zeiss LD-Achroplan 40×/0.60). The X-Y translational stage allowed us to use any point of the microchannel network as the detection volume. Fluorescence was collected with the same objective and, after passing the dichroic mirror and an interference filter (Omega 530DF30), spatially filtered through an X-Y-Z adjustable pinhole (diameter, 600 µm). The photomultiplier (Hamamatsu R1477) was mounted in an integrated detection module including HV power supply, voltage divider, and amplifier (SMT Model NV 30-1, Seefeld, Germany). An optical power meter (Model 840, Newport) was used to measure the intensity of the transmitted excitation light after passing through the chip assembly (∼0.3-0.5 mW ). A valuable practical aspect of the inverted microscope setup was the possibility of visually checking all field-controlled operations on the device through the eyepiece. This way, the actual size of the sample injection volume and any leakage effects could be easily controlled with a fluorescein test sample under actual experimental conditions. Device Operation and Data Acquisition. The two layouts of the microchip devices used in this work are depicted in Figure 4. The principle of sample injection and device operation in ICE has been described in detail before3,4 and will be discussed here only briefly. Approximately 5-10 µL of sample was pipetted into the sample reservoir 1. Special attention was given to ensure that all reservoirs were filled to the same level, in order to prevent secondary hydrodynamic flow in the device. By application of an electrical field in the order of 200-400 V/cm between reservoirs 1 (-HV) and 3 (GND), the channel intersection volume was filled with sample solution. Electrophoresis was started by applying a separation potential between reservoirs 2 (-HV) and 4 (GND). Leakage from the injection channels was prevented by inducing a small electrical field (∼50 V/cm) in both halves of the injection channel while the separation voltage was applied, thus driving the excess sample molecules toward reservoirs 1 and 3, respectively. The experiments were controlled with a LabVIEW 3 program operated on an IBM-compatible PC. The potential switching protocol (switching of high-voltage relays) was controlled by the digital output lines of a data acquisition board Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3453
(AT-MIO-16E-2, National Instruments). Potentials were set under computer control by means of an analog output board (AT-AO-6, National Instruments). The output of the detection module was smoothed with a low-pass filter (Avens Signal Equipment Corp., Model AP-255-5) and recorded by the ADC input of the data acquisition board. RESULTS AND DISCUSSION PDMS Microchips. Apart from its known capability to reproduce features of the master down to a submicrometer scale, the general physical-chemical properties of PDMS favor its use for electrophoresis experiments using optical detection schemes. Commercial PDMS is available as a product of high optical quality, transparent above ∼230 nm, and exhibits an electrical bulk resistivity sufficiently high (>1015 Ω cm) to prevent the electrical current from flowing through the bulk material. Its relatively low refraction index (n ) 1.430) reduces the amount of reflected excitation light in optical detection schemes. Although its surface free energy is not particularly high (∼22 mN/m),18 the smoothness of casted PDMS surfaces in combination with its elastomeric properties ensures good adhesion on a variety of clean and flat planar15 or curved substrates.19 A critical physical-chemical parameter in electrophoresis experiments is the coefficient of heat conductivity κ of the compartment material. PDMS shows a value of κ ) 0.15 W m-1 K-1, a value a factor of ∼10 lower than that of fused silica (1.38 W m-1 K-1) and a factor of ∼5-6 lower than most glasses (typically 0.7-1.0 W m-1 K-1). The capability of PDMS to dissipate the Joule heat was investigated by monitoring the electrical current as a function of the applied potential or field strength, respectively (Ohm’s plot). Figure 5 shows the result for a buffer solution of a relatively low electrical conductivity (0.75 mS cm-1). The positive deviation from the linear relationship at high field strengths clearly indicates insufficient heat dissipation in the microchannel, leading to an overall temperature rise of the buffer solution with an associated increase of the electrical conductivity. It is interesting to note that the deviation occurs at a value of the dissipated heat per unit length slightly below 1 W m-1, a value often regarded as an upper limit using commercial CE instrumentation. Similar experiments with buffers of higher electrical conductivity showed the same general behavior, with positive deviations occurring at lower field strengths, but at very similar values of the dissipated heat per unit length between 0.5 and 1 W m-1. The capability of PDMS chips to remove Joule heating is worse compared to microchannels etched into glass or quartz substrates (see, e.g., ref 20), but still sufficient enough for applications at field strengths typically used in conventional CE experiments in the order of ∼100-1000 V/cm. It is worthwhile to mention that all experiments described in this work were performed with PDMS casts without any modifications of the channel walls. The question of whether electroosmotic flow (EOF) is present under our experimental conditions has not been studied in detail. Preliminary data indicate that the EOF is significantly reduced (50%; i.e., the majority of the molecules passing the channel cross section at the detector location were registered. This relatively high SMD efficiency was supported by employing a channel geometry with a depth that is twice as large as the width, well adapted to the geometrical shape of the excitation light cone. In addition, the excitation light was adjusted to be slightly out of focus to ensure that the whole channel cross section is illuminated as uniformly as possible. The use of a focusing collimator unit allowed defocusing of the excitation light while maintaining the image plane of the microscope. The adjustment was carried out by simultaneously observing and adjusting the size of the fluorescent light spot through the eyepiece while pumping a fluorescent dye solution through the channel. As a side effect, the reduced photon flux at the microchannel reduced potential problems with the photodestruction of the dye (bleaching). The pinhole diameter (0.6 mm) used made sure that the field of view of the PMT (diameter, 600 µm/40 ) 15 µm) was slightly wider than the channel width. If we denote the coordinate along the channel with x and the propagation direction of the laser beam with z, the detection volume is in a first approximation geometrically defined by the size of the confocal pinhole (x), by the channel walls (y), and along z by the collection efficiency function of the confocal optics.27 Since (27) Qian, H.; Elson, E. L. Appl. Opt. 1991, 30, 1185-95. (28) Eigen, M.; Rigler, R. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 5740-7.
both the excitation light intensity and fluorescence collection efficiency will vary over the channel cross section, the corresponding height of the fluorescence burst signals of passing DNA molecules will also be subject to variations, as visible in Figures 8 and 9. The transit time through a 15 µm long detection volume can be calculated to be ∼77 ms at 100 V/cm, in good agreement with the observed width of the fluorescence bursts. The signalto-noise ratio exhibited in Figures 8 and 9 indicates that SMD sensitivity can also be achieved for much shorter and/or less dyeloaded DNA molecules. CONCLUSIONS A simple and inexpensive molding procedure has been successfully demonstrated to allow for the fabrication of chip-based microfluidic devices. Relief structures on the silicon master are truly reproduced on the micrometer scale. The flat surfaces adhere without bonding procedures to a variety of smooth substrates, thus forming closed microchannel systems. The optical properties of PDMS allow its use in both UV and LIF detection schemes. All relevant characteristics of PDMS render this material an excellent alternative to the commonly used glass and quartz substrates. The application of these devices in combination with a highly sensitive LIF detector enabled us to perform electrophoresis experiments on the level of single DNA molecules in chip-based microsystems for the first time. The high performance of the
electrokinetically controlled devices is also demonstrated with electrophoretic separations of samples relevant to bioanalysis (DNA fragments, peptides). The combination of electrokinetic control of picoliter sample volumes and single-molecule detection can result in a novel tool for the manipulation of single molecular objects in solution, an emerging technology with far-reaching consequences in, for example, molecular diagnostics.28 ACKNOWLEDGMENT We thank Dr. Andreas Helg and Dr. Gerolf Kraus (all Novartis Pharma AG, Basel) for several useful discussions during the fabrication of the first generation of silicone microstructures. We also thank Dr. Carlos Garcia-Echeverria (Novartis Pharma AG) for the synthesis of the fluorescein-labeled peptides, Mrs. C. Bru¨cher (Novartis Services AG, Basel) for providing the electron micrograph, and Mr. W. Bach from Institut fu¨r Mikro- und Informationstechnologie, Villingen-Schwenningen, Germany, for the fabrication of the master wafers. This article is dedicated to the memory of the late Prof. H. Michael Widmer, who died on May 25, 1997. Received for review April 11, 1997. Accepted June 12, 1997.X AC9703919 X
Abstract published in Advance ACS Abstracts, August 1, 1997.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3457
