Anal. Chem. 1998, 70, 1909-1915
Dielectrophoretic Sorting of Particles and Cells in a Microsystem Stefan Fiedler,* Stephen G. Shirley, Thomas Schnelle, and Gu 1 nter Fuhr
Institut fu¨ r BiologiesMembranphysiologie, Humboldt Universita¨ t zu Berlin, Invalidenstrasse 43, D-10115 Berlin, FRG
There are highly sensitive analytical techniques for probing cellular and molecular events in very small volumes. The development of microtools for effective sample handling and separation in such volumes remains a challenge. Most devices developed so far use electrophoretic and chromatographic separation methods. We show that forces generated by ac fields under conditions of negative dielectrophoresis (DEP) can also be used. Miniaturized electrode arrays are housed in a microchannel and driven with high-frequency ac. A laminar liquid flow carries particles past the electrodes. Modification of the ac drive changes the particle trajectories. We have handled latex particles of micrometer size and living mammalian cells in a device which consists of the following four elements: a planar funnel which concentrates particles from a 1-mmwide stream to a beam of about 50-µm width, an aligner which narrows the beam further and acts to break up particle aggregates, a field cage which can be used to trap particles, and a switch which can direct particles into one of two output channels. The electrodes are made from platinum/titanium and indium tin oxide (ITO) on glass substrates. Particle concentration and switching could be achieved for linear flow velocities up to about 10 mm s-1. The combination of this new method with highperformance optical detection offers prospects for miniaturized flow cytometry. Fast acquisition of complex data is very important for medical diagnosis and for monitoring, especially in environmental protection and chemical manufacture. Flow injection analysis in complex analytical microsystems aids real-time monitoring. Semiconductor microstructuring techniques have been used to create miniaturized processing and analysis systems. Impressive results have been obtained from miniaturized HPLC and capillary electrophoresis on a chip.1,2 Such equipment allows measurements at the single-cell and molecular levels even in flow-through mode.3-9 * To whom correspondence should be addressed. Present address: Fraunhofer Institut fu ¨ r Zuverla¨ssigkeit und Mikrointegration, Gustav-Meyer-Allee 25, Geb. 17, D-13355 Berlin, FRG. E-mail:
[email protected]. (1) Ocvirk, G.; Veerporte E.; Manz, A.; Grasserbauer, M.; Widmer, H. M. Anal. Methods Instrum. 1995, 2, 74-82. (2) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1995, 67, 22842287. (3) Eigen, M.; Rigler, R. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 5740-5747. (4) Nie, S.; Chiu, D. T.; Zare, R. N. Anal. Chem. 1995, 67, 2849-2857. S0003-2700(97)01063-9 CCC: $15.00 Published on Web 03/31/1998
© 1998 American Chemical Society
The constituents of a complex particulate sample should, ideally, pass the detector site one by one. In flow cytometry, the particle jet is produced by hydrodynamic focusing in a sheath fluid. Optical signals (reflected, scattered, and fluorescent) are collected as the particles pass the detector. To sort, the jet is broken into droplets by a nozzle, and droplets containing chosen particles are electrostatically deflected. A througput of the order of 104 cells/s is common with available machines. Conventional fluorescence-activated cell sorters (FACS) suffer from the discrepancy between tool and object size. This mismatch hinders their integration with miniaturized, high-performance analytical systems. Hydrodynamic focusing has been already adapted to microfabricated structures,10,11 whereas miniaturized flow direction remains challenging. Recently, Blankenstein et al.12 scaled hydrodynamic flow splitting down to chip format by applying guiding liquid streams. Electrokinetic transport principles have also been adapted to miniaturized total analysis systems (µTAS).13-17 Optical tweezers and scissors using laser light are promising candidates for the miniaturized tool approach.18-23 However, these techniques can suffer from prob(5) Arbault, S.; Pantano, P.; Jankowski, J. A.; Vuillaume, M.; Amatore, C. Anal. Chem. 1995, 67, 3382-3390. (6) Ewing, A. G.; Gavin, P. F.; Clark R. A. In Proceedings of the 2nd International Symposium on Miniaturized Total Analysis Systems, µTAS96, Basel, 19-22 November 1996; Widmer, H. M., Ed. Anal. Methods Instrum. 1996, 141143. (7) Paras, C. D.; Kennedy, R. T. Anal. Chem. 1995, 67, 3633-3637. (8) Tan, W.; Parpura, V.; Haydon, P. G.; Yeung, E. S. Anal. Chem. 1995, 67, 2575-2579. (9) Jacobson, S. G.; Ramsey, J. M. Anal. Chem. 1996, 68, 720-723. (10) Sobek, D.; Young, A. M.; Gray, A. L.; Senturia, S. D. Proceedings of the IEEE MEMS Workshop, Pisano, A., Lang, J., Eds., Fort Lauderdale, FL, February 7-10, 1993; pp 219-222. (11) Brody J. P.; Yager, P.; Goldstein, R. E.; Austin, R. A. Biophys. J. 1996, 71, 3430-3441. (12) Blankenstein, G.; Scampavia, L.; Branebjerg, J.; Larsen, U. D.; Ruzicka, J. In Proceedings of the 2nd International Symposium on Miniaturized Total Analysis Systems, µTAS96, Basel, 19-22 November 1996; Widmer, H. M., Ed. Anal. Methods Instrum. 1996, 82-84. (13) Harrison, D. J.; Li, P.; Tang, T.; and Lee, W. In Proceedings 2nd International Symposium on Miniaturized Total Analysis Systems, µTAS96, Basel, 19-22 November 1996; Widmer, H. M., Ed. Anal. Methods Instrum. 1996, 147149. (14) Manz, A.; Effenhauser, C. S.; Burggraf, N.; Harrison, J. D.; Seiler, K.; Fluri, K. J. Micromech. Microeng. 1994, 4, 257-265. (15) Ramsey, J. M.; Jacobson, S. C.; Knapp, M. R. Nature Med. 1995, 1, 10931095. (16) Manz, A.; Graber, N.; Widmer, H. M. Sens. Actuators B 1990, 1, 244-248. (17) Murakami, Y.; Uchida, T.; Takeuchi, T.; Tamiya, E.; Karube, I.; Suda, M. Electroanalysis 1994, 6, 735-739. (18) Ashkin, A.; Dziedzic, J. M.; Yamane, T. Nature 1987, 330, 769-771.
Analytical Chemistry, Vol. 70, No. 9, May 1, 1998 1909
lems with sample dilution, electrophoretic damage to sensitive cells, or elevated temperatures during particle handling. Inhomogeneous, high-frequency ac electric fields can be used to manipulate molecules and particles by dielectrophoresis.24 Positive dielectrophoresis (positive DEP) is the movement of a dielectric toward regions of high field strength; it can cause damage as particles are attracted to the electrodes. Less stress is induced, especially in living cells, when the particles move toward field minima (negative DEP). Miniaturization allows such field minima to be created on a scale suitable for handling single, small objects, such as viruses and nanometer particles.25-28 Dielectrophoretic field funnels, traps and cages, traveling-wave pumps, and a particle micromanipulator have been developed.29-33 The aim of this work was to integrate several dielectrophoretic modular elements into a miniaturized particle and cell manipulation chip. THEORY Time periodic, inhomogeneous electric fields induce polarization and subsequent movement of dielectric particles. Movement is toward regions of high field strength (positive dielectrophoresis) or low field strength (negative dielectrophoresis), depending on the driving frequency and on the dielectric properties of particles and surrounding liquid.34 Negative dielectrophoresis is desirable in most applications, as the particles are repelled from the electrodes and spend most of their time in low-field regions. Closed-field cages created by three-dimensional electrode arrangements can sustain conditions of negative DEP and have been proven useful for contactless particle confinement.30 To study real electrode arrangements, the electric field has to be calculated numerically. With the complex electric field according to (19) Buican, T. N.; Smyth, M. J.; Crissman, H. A.; Salzman, G. C.; Steward, C. C.; Martin, J. C. Appl. Opt. 1987, 26, 5311-5316. (20) Mizuno, A.; Nishioka, M.; Tanizoe, T.; Katsura, S. IEEE Trans. Ind. Appl. 1995, 31, 1452-1457. (21) Sato, S.; Inaba, H. Opt. Quantum Electron. 1996, 28, 1-16. (22) Weber, G.; Greulich, K. O. Int. Rev. Cytol. 1992, 133, 1. (23) Yao, H.; Ikeda, H.; Inoue, Y.; Kitamura, N. Anal. Chem. 1996, 68, 43044307. (24) Washizu, M.; Nanba, T.; Masuda, S. IEEE Trans. Ind. Appl. 1990, 26, 352358. (25) Fuhr, G.; Zimmermann, U.; Shirley, S. G. In Electromanipulation of cells; Zimmermann, U., Neil, G. A., Eds.; CRC Press Inc.: Boca Raton, FL, 1996; pp 259-328. (26) Schnelle, Th.; Mu ¨ ller, T.; Fiedler, S.; Shirley, S. G.; Ludwig, K.; Herrmann, A.; Fuhr, G,; Wagner, B.; Zimmermann, U. Naturwissenschaften 1996, 83, 172-176. (27) Mu ¨ ller, T.; Gerardino, A.; Schnelle, Th.; Shirley, S. G.; Bordoni, F.; De Gasperis, G.; Leoni, R.; Fuhr, G. J. Phys. D: Appl. Phys. 1996, 29, 340349. (28) Mu ¨ ller, T.; Fiedler, S.; Schnelle, Th.; Ludwig, K.; Jung, H.; Fuhr, G. Biotechnol. Tech. 1996, 10, 221-226. (29) Fuhr, G.; Fiedler, S.; Mu ¨ ller, T.; Schnelle, Th.; Glasser, H.; Lisec, T.; Wagner, B. Sens. Actuators A 1994, 41-42, 230-239. (30) Schnelle, Th.; Hagedorn, R.; Fuhr, G.; Fiedler, S.; Mu ¨ller, T. Biochim. Biophys. Acta 1993, 1157, 127-140. (31) Fuhr, G.; Mu ¨ ller, T.; Schnelle, Th.; Hagedorn, R.; Voigt, A.; Fiedler, S.; Arnold, W. M.; Zimmermann, U.; Wagner, B.; Heuberger, A. Naturwissenschaften 1994, 81, 528-535. (32) Fuhr, G.; Schnelle, Th.; Hagedorn, R.; Shirley, S. G. Cell Eng. Inc. Mol. Eng. 1995, 1, 47-57. (33) Fuhr, G.; Schnelle, Th.; Wagner, B. J. Micromech. Microeng. 1994, 4, 217226. (34) Pohl, H. A. Dielectrophoresis, 1st ed.; Cambridge University Press: Cambridge, UK, 1978.
1910 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998
E B(b,t) r ) Re[(E Bre(b) r + iE Bim(b)) r exp(iωt)]
(1)
with radian frequency ω, we use a finite difference method. If the fluid in the cage is dielectrically homogeneous, a timeindependent Laplace equation for the complex potential of the electric field, E B ) -∇φ b, will result. The time-averaged dielectrophoretic force acting on a particle of radius R can be calculated in dipole approximation for known electric field distribution as E B(t):
〈F BDEP〉 ) 2πl0R3 ×
[
B Re(fCM)∇
∑{(E µ
∑
Im(fCM)
µ,ν
{(
Ere µ
re 2 µ)
2 + (Eim µ ) } +
∂Eim µ ∂ν
-
Eim µ
)}
∂Ere µ ∂ν
b eν
]
(2)
with µ, ν ) x, y, z (see also ref 35). The Clausius-Mosotti factor, fCM, reflecting the dielectric properties (permittivity and conductivity σ) of particle (index p) and surrounding liquid (index l), becomes, for a homogeneous sphere,
fCM )
˜ p - ˜ l ˜ p + 2˜l
(3)
with the complex permittivity ˜ ) σ + iω0. The first term in eq 2 is proportional to the gradient of the mean square value of the electric field strength; the second one is proportional to the gradient of the field phase. Hence, individual manipulation of single particles requires field inhomogeneities with dimensions similar to those of the particles. For the ac case, without gradient in the phase, the conditions for negative dielectrophoresis are fulfilled for
Re(fCM) < 0
(4)
i.e., if the absolute value of the complex permittivity of the particle is lower than that of the suspending liquid. Due to the high permittivity of water, polymer particles and living cells show negative dielectrophoresis at high-field frequencies. To estimate the field strength necessary for dielectrophoretic particle manipulation in streaming suspensions, model calculations have been made. The dielectrophoretic forces experienced by a latex sphere in a 0.5 MV/m field (10 V for 20-µm electrode gap) in water are in the range of some tens of nanonewtons down to piconewtons. This is illustrated for an idealized geometry of a semiinfinite plate capacitor. The electric field distribution can be calculated analytically using the theory of complex functions. Particles of radius R showing negative DEP that are brought by streaming toward the capacitor boundary are focused to the middle plane. There, they experience a maximum repelling force (perpendicular to the boundary of the capacitor) that can be given in dipole approximation as
FDEP )
27π2Urms2 R 3 Re(fCM) 32 a l
()
(5)
where a stands for the distance of the plates. In our case (applied
Figure 1. Experimental setup. (A) Schematic drawing of the equipment used to manipulate microparticles and to observe their motion. (B) A close-up view of the encapsulated chip and its connections.
voltage of 10 V and an electrode distance of 20 µm), the repelling force varies from about 0.2 pN for a 3.4-µm latex sphere to 15 pN for the 15-µm latex bead. These values are greater than or comparable to the hydrodynamic force the particles experience because of the streaming. According to Stokes’ law and our maximum streaming velocity of 1 cm/s, the hydrodynamic force is in the range of about 0.3 pN for the smaller and 1.4 pN for the larger one. Theoretically, the smallest single particle that could be aligned at the highest fluid velocity in our structure, having an angle of 18° (see, e.g., Figure 3) between the streaming and capacitor, is about 2 µm in size. Thus, passive hydrodynamic streaming of a particle should be influenced or even stopped. For undisturbed suspensions, the formation of dense particle aggregates according to field force distribution and the possible “freezing” of them in a trap has been already shown.35 EXPERIMENTAL SECTION Latex Particles. Commercially available latex samples of different particle diameters (0.46, 3.4, 6, and 15 µm) were suspended in water containing 0.001% Tween 20 (SERVA). The solution conductivity was 100 µS/cm. Cells. Confluent L929 mouse cell (DSMZ GmbH, Braunschweig, FRG) monolayers were trypsinized according to common practice and resuspended in the original growth medium (RPMI, supplemented with 5% FCS, 1.3 S/m). Cell suspensions were prepared freshly to minimize adhesion. Electrode Design and Drive. Titanium (50-80 nm) and platinum (100-200 nm) layers were sequentially sputtered (Multicoat GmbH & Co. KG, Zittau/Sa, Germany) onto AF 45 glass wafers (DESAG, Gru¨nenplan, Germany). These chips were annealed at 450 °C. Metal-coated glass chips, 9 × 9 mm2, and ITO-coated ones (about 400-nm layer thickness, also DESAG) were structured using the 248-nm line of a KrF excimer laser system (Exitech EX-PS-750, Long Hanborough, Oxford, UK). A positive mask technique36 produced precise ablation in the central parts of the multipole electrode arrays.37 Bond pads and peripheral electrodes were structured by direct scribing with the laser (35) Fiedler, S.; Schnelle, Th.; Fuhr, G. Microsyst. Technol. 1995, 4, 1-7. (36) Fuhr, G.; Shirley, S. G. Lambda Physik Science Report No. 7, 1996; pp 6-7. (37) Fiedler, S.; Shirley, S. G.; Schnelle, Th. Sens. Mater. 1997, 9, 141-148.
beam. Electrical connections between the planar electrodes on the chip and a carrier board were made by gold wire and silverfilled epoxy. The connections were strengthened with conventional epoxy. Two chips were mounted face to face with an lateral alignment precision of about 2 µm. One chip had metal electrodes, and the other had a mirror image pattern in ITO. Small strips of 30-µmthick polymer spacer (Sikadeflex HS 11/90; Sika Werke GmbH, Leipzig, Germany) were arranged to form a channel in the sandwich across the electrode array. Stainless steel capillaries were connected at both ends of the channel to allow connection to PTFE tubing. The sandwich was sealed with a photocuring glue (multipurpose adhesive 7543, 3M Dental Products, St. Paul, MN) and strengthed with epoxy. The mounted microsystem was encapsulated with a transparent, single-compound silicone glue (Elastosil E 43, Drawin Vertriebs GmbH, Ottobrunn, Germany). Pressure differentials of a few centimeters to a few tens of centimeters of water produced streaming velocities in the working range; flow was regulated by adjusting the heights of inlet and outlet reservoirs. The electrodes were energized by rectangular pulses (typically 1 MHz) from an HP 8111 A pulse function generator (HewlettPackard, Palo Alto, CA). Before use, the microchannel systems were carefully wetted to avoid bubbles. After use, systems were washed with detergent by applying slight pressure or cleaned by a short incubation with a solution of hydrogen peroxide and sulfuric acid (1:4). Thorough rinsing with water followed each cleaning step. Particle movement was observed with a conventional microscope or with additional epifluorescence equipment (Zeiss Axioscop), supplemented with a camera and video system. For the experimental setup, see Figure 1A. Figure 1B shows the carrier (microscope slide size) with a silicone-packed hybrid chip, the observation window, and individual connecting wires. RESULTS AND DISCUSSION Figure 2 gives a schematic representation of the electrode arrangement in the device. As particles move through the channel, they encounter (in order) a planar funnel (A), an aligner Analytical Chemistry, Vol. 70, No. 9, May 1, 1998
1911
Figure 2. (A) Schematic drawing of the electrode system used. Each area pictured here represents a Pt/Ti electrode of the lower chip and a corresponding ITO electrode of the upper chip. The arrow indicates the direction of flow. A, The funnel; B, the first aligner; C, the cage; D, the second aligner; E, the switch; F, the outlet channels which, in this design, are integrated with the switch. (B) Closer view of the chip sandwich microsystem. Two chips, carrying matching electrode arrays, are mounted face to face and sealed by spacer material to form a flat, parallel-sided channel. The electrodes were structured by excimer laser in indium tin oxide (ITO) as shown or in Pt/Ti layers on glass. This prototype contains electrically unconnected areas of ITO between the elements. Each electrode was driven individually to find optimum conditions. Inlet and outlet capillaries are connected at the left and right sides of the channel. The chip sandwich is mounted on printed circuit board material and packed in silicone, leaving a central observation window.
(B), a field cage (C), a second aligner (D), a switch (E), and outlet channels (F). Particle behavior at each element is described below. (A) The Planar Funnel. This element consists of two metal electrodes on one chip, angled so that they converge from the full width of the channel (about 1 mm) to a gap of 35 µm. Each metal electrode has a corresponding ITO electrode on the other chip. A metal electrode and its ITO counterpart are driven in antiphase. This produces two field barriers at an angle of 18° to the flow. Particles enter across the full width of the channel and are deflected by the barriers into the central region (Figure 3A). This enhances particle concentration by a factor of about 30 and produces a broad, centralized particle beam. The angle of the funnel electrodes is 18°, but this is not critical. The connectors to the cage system are angled at 45°, and funneling can also occur here when the drive amplitude is suitably chosen (Figure 3B). Particles which are very much smaller than the exit width of the funnel leave the device in two distinct beams, which then travel nearly parallel to each other (Figure 3C). Each beam originates from the tip of one of the electrode pairs. 1912 Analytical Chemistry, Vol. 70, No. 9, May 1, 1998
Figure 3. Particle funneling from streaming aqueous suspensions by dielectrophoresis. (A) Concentration of latex beads (o.d. 15 µm) by the funnel. Only the bottom plane Pt/Ti electrodes are visible, and the dark triangular regions on the centerline are unremoved (unconnected) areas of metal. Streaming is indicated by an arrow. Particle are guided by the angled electrodes to the 35-µm-wide exit gap. At this point, particle concentration has been enhanced by a factor of about 30 over the starting concentration. (B) Alignment of latex beads (o.d. 3.4 µm) using the connectors of an octopole cage. Flow in this figure is from left to right. Particles showing negative dielectrophoresis are funneled by the angled connectors into the central region, where they would be held at higher drive amplitude. Particles leave the underdriven cage in three jets, two of which are guided by the downstream connectors. (C) The trajectories of particles leaving the funnel (indicated with arrows). This figure has been computer processed. Much of the variance of intensity of each pixel between successive video frames derives from moving particles. An image has been generated from the variance of 15 successive frames (to show moving particles) and superimposed on the mean image (to show the electrodes). In this case, the particles were sufficiently small (latex 3.4 µm) that they formed two jets, one from each electrode pair. The aligner, over which the jets are traveling, is here unenergized.
(B) The Aligner. This consists of two metal comblike structures with angled upstream and downstream ends. Each metal electrode has an ITO counterpart. The device is energized as a quadrupole (a metal electrode and the diagonally opposite ITO electrode are connected to the same phase). The angled ends of the device compress the incoming broad beam further;
Figure 4. Force distribution inside multiple closed quadrupole field cages and their use to align particles. (A) Calculated surface of equal inward force for one single unit. The mathematical modeling is discussed in the theoretical section. Size is given in micrometers. (B) Collection of latex particles from a starting dense suspension. Particles have been suspended between top and bottom planes of a “sandwich” formed by arranging identically shaped gold electrodes over each other. After energizing the array with 1-MHz signals of alternating polarity at the corresponding four electrodes, the otherwise undisturbed particle suspension arranges into the shown loose agglomerates. Particle size, 3.4 µm; conductivity of aqueous solution, 100 µS/cm. (C) Trapped latex particles, 15-µm diameter. Single particles have been confined from a particle jet which leaves the funnel (Figure 3A,C). Streaming direction is from left to right. The drive is 1 MHz, with a Pt electrode and the diagonally opposite ITO one connected to phase 1 and the other two electrodes to phase 2. The video shows that the particles move through the structure in a series of jerks and leave it in a single jet. (D) Closer view of the comblike aligner element. Precise superposition of symmetric top and bottom dipoles allows concentric arrangement of latex sphere pairs inside “closed” quadrupole cages. Larger particles would be held individually. In this frame, particles are held stationary in a left to right liquid stream. (E) Particles leave the aligner in a single beam (see arrow) which, due to the low Reynold’s number, can travel considerable distances, here about 0.4 mm, even when the flow is disturbed and the jet deflected by a minor obstruction in the channel. This image has been processed (as Figure 3C) from eight successive video frames.
the exit beam is little wider than the width of an individual particle. With streaming rates of
