Anal. Chem. 2003, 75, 5287-5291
Technical Notes
Robust Interconnects and Packaging for Microfluidic Elastomeric Chips Hao Chen, Dhruv Acharya, Arivalagan Gajraj, and Jens-Christian Meiners*
Department of Physics and Biophysics Research Division, University of Michigan, Ann Arbor, Michigan 48109-1120
A stable, rugged, and easy-to-use microfluidic platform has been developed and evaluated. The system is based on multilayer silicone elastomer chips with integrated flow channels and active components, such as pressureactuated valves. The silicone chips and stainless steel interconnect tubes are embedded in a block of an epoxy resin to give the chip and the interconnects outstanding mechanical stability. Full optical accessibility of the chip for demanding optical detection and manipulation applications, such as fluorescence and bright-field microscopy and optical tweezing, is achieved through the use of an optically transparent epoxy resin and microscope cover glasses as the packaging materials. Furthermore, this packaging technique uses a purely mechanical seal between the elastomer and cover glass, enabling applications that use chemically functionalized glass surfaces as the bottom of the flow channels. In addition, a socket was developed in which the microfluidic chips can be plugged in to provide all external connections for reagent delivery and pressure control of the integrated valves. The utility of these devices is demonstrated by showing that single DNA molecules can be attached to protein-coated walls of the flow channels and manipulated with optical tweezers. Over the past decade, applications for microfluidic devices have proliferated with a speed reminiscent of the explosive use of microelectronics after the integrated circuit was invented. While many microfluidic and microelectromechanical devices are made of the same hard materials as their electronic counterparts, namely, silicon and its oxides, soft materials such as silicone elastomers have become increasingly interesting for microfluidic applications. There are several reasons for this trend; most notably it is relatively easy and inexpensive to manufacture such devices, both on the prototype scale and in mass fabrication. Furthermore, the elasticity of the material has made it possible to create mechanically active features, such as valves and pumps, and integrate them directly into the chip without any special manufacturing require* Corresponding author. Phone: (734) 764 7383. Fax: (734) 764 5153. E-mail:
[email protected]. 10.1021/ac034179i CCC: $25.00 Published on Web 08/23/2003
© 2003 American Chemical Society
ments.1 The utility of these silicone elastomer chips has been demonstrated in a number of examples, ranging from DNA sizing2,3 and cell sorting4,5 to flow injection analysis6 and protein crystallization.7 Many more applications, in particular, in the field of microenzymology, are currently under development. This technology received a further boost when Thorsen et al.8 demonstrated the feasibility of large-scale integration of more than 2000 microvalves for multiplexing applications on a single chip. Unlike their microelectronic counterparts, soft silicone elastomer microfluidic chips are rather inconvenient to use, since they cannot simply be plugged into a socket that provides all the external connections. Instead, the external connections are made by thin metal tubes, which are stuck into holes that have previously been pierced into the elastomer chip. While the silicone does provide a tight seal around the tubes, the mechanical stability of these interconnects is nonetheless rather poor. Relatively little mechanical strain on the connections can affect the flow rates of the fluids delivered to the chips and even pull the tubes out of the silicone seal. The silicone elastomer chips are also difficult to mount on external equipment, such as a microscope, because clamping the chip results in stress that easily propagates through the material and adversely affects the performance of the micromechanical components. Therefore, the development of a packaging technique that gives the elastomer chips and interconnects the necessary stability and ruggedness is urgently needed before this technology can be used routinely in applications such as medical diagnostics or drug discovery. In this note, we present a packaging technique that allows us to stabilize microfluidic silicone elastomer chips and their interconnects by embedding them in an optically transparent epoxy resin. This technique enables us to plug these chips, like their electronic counterparts, into a socket that provides all the external (1) Unger, M. A.; Chou, H. P.; Thorsen, T.; Scherer, A.; Quake, S. R. Science 2000, 288, 113. (2) Effenhauser, C. S.; Bruin, G. J. M.; Paulus, A.; Ehrat, M. Anal. Chem. 1997, 69, 3451-3457. (3) Chou, H. P.; Spence, C.; Scherer, A.; Quake, S. R. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 11-13. (4) Fu, Y.; Spence, C.; Scherer, A.; Arnold, F.; Quake, S. R. Nat. Biotechnol. 1999, 17, 1109-1111. (5) Fu, Y.; Chou, H. P.; Spence, C.; Arnold, F.; Quake, S. R. Anal. Chem. 2002, 74, 2451-2457. (6) Leach, A. M.; Wheeler, A. R.; Zare, R. N. Anal. Chem. 2003, 75, 967-972. (7) Hansen, C. L.; Skordalakes, E.; Berger, J. M.; Quake, S. R. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 16531-16536. (8) Thorsen, T.; Maerkl, S. J.; Quake, S. R. Science 2002, 298, 580-584.
Analytical Chemistry, Vol. 75, No. 19, October 1, 2003 5287
Figure 1. Microfabricated flow channel with a pneumatically actuated valve. The flow channel runs from the top to the bottom of the picture and is separated by a thin membrane from the perpendicular control channel. (a) When the control channel is not pressurized, flow through the flow channel is unimpeded; the valve is open. (b) When the control channel is pneumatically pressurized, it expands and pinches off the flow through the flow channel underneath; the valve is closed.
connections for reagents and control lines and in turn makes them easy to mount on external equipment. Furthermore, due to the shrinkage of the encapsulating epoxy resin, the elastomer chip is held firmly against the coverslip that seals the flow channels. Consequently, we can dispense with the customarily used annealing step that bonds the elastomer to the glass. This enables the use of surface chemistry on the glass surface because the glass can be precoated with proteins or nucleic acids before assembly, thus effectively eliminating the degradation of the protein-coated glass surface that occurs due to the direct effects of the elevated temperature or through contamination with volatile compounds that are emitted from the silicone. We demonstrate the capability to work with sensitive proteincoated cover glasses by attaching single DNA molecules inside a flow channel to a digoxigenin antibody-coated surface. As a result of the superior optical accessibility of our chips, we are able to manipulate these DNA molecules in situ with optical tweezers. In addition, we show that the integrated valves on the chip allow the rapid exchange of buffers, enabling controlled change of the chemical environment of the tethered DNA molecules for singlemolecule studies of DNA-protein interactions. EXPERIMENTAL SECTION Our microfluidic devices were manufactured in three steps. First, two-layer silicone elastomer chips are fabricated in a replication-molding technique from silicon master molds in a technique closely following the one described by Unger et al.1 The thin and soft bottom layer contains the flow channels. The flow through these channels is regulated through perpendicular control channels in the upper hard layer. When these control channels are pressurized, they expand and pinch off the flow through the fluidic channels in the soft layer underneath.1 This mechanism of action is shown in Figure 1. Steel tubing is then inserted into the chips as interconnects to access control and reagent channels in the upper and lower layers, respectively. In the second step, the chips with the interconnects are embedded in a block of optically transparent epoxy resin that gives them exceptional mechanical stability. The casting mold for this embedding process is an aluminum frame that is left in place to provide for convenient mounting on external equipment, such as an inverted microscope. Finally, the chip with its mounting frame is plugged into a socket that provides the external connections to control and reagent supply lines, giving us a compact, robust, and 5288
Analytical Chemistry, Vol. 75, No. 19, October 1, 2003
user-friendly microfluidics platform with full optical access. In the following sections, the fabrication process for these microfluidic devices and sockets is described in detail. Elastomeric Chips and Interconnects. The microfluidic chips are fabricated by replication molding of an elastomeric material from silicon masters. These master molds for both the flow channel layer and the control channel layer are manufactured in a rapid-prototyping approach.9 The design patterns for the molds are printed on transparencies with a high-resolution (3386 dpi), commercial laser typesetting printer. These transparencies, in turn, serve directly as masks for contact photolithography. Silicon wafers are spin coated with photoresist (9625, Shipley), patterned by UV photolithography and developed to yield the master molds with positive relief features. The mold for the flow channel layer is subsequently annealed at 120 °C for 30 min to temporarily melt the photoresist and round off the sharp edges of the pattern. This step is necessary to ensure proper closure of the valves.1 Typical features on the silicon wafer molds were about 10 µm in height and 100-130 µm in width, as measured with a surface profilometer. The master molds for each layer are treated with trimethylchlorosilane in the gas phase for ∼7 min to silanize their surfaces and prevent unwanted adhesion of the elastomer during the casting process. Both layers are made of a poly(dimethylsiloxane) (PDMS)-based two-component silicone elastomer (RTV-615, General Electric) using different ratios of the two components A and B. The upper hard layer is cast as a slab of 4-5-mm thickness from a 5:1 mixture of RTV 615 A and B. The lower soft layer is cast with a spin coater (P-6708 D, Specialty Coating Systems) at 2000 rpm for 35 s at a thickness of ∼30 µm from a 25:1 mixture of the RTV components. Both layers are cured in their molds for 90 min at 80 °C. Then the top layer is removed from the mold, and access holes for external connections to the control channels are pierced into the elastomer slab with 20-gauge hypodermic needles. Special care is taken to ensure that these holes are perpendicular to the surface, as these holes determine the alignment of the interconnects, which are inserted at a later stage. The two layers are then carefully superimposed and further baked overnight at 80 °C to create a monolithic elastomere chip. After removal from the mold, additional access holes for the flow channels are punched into the chip. Then 25-mm-long pieces of 23-gauge stainless steel hypodermic tubing are inserted by hand into the access holes to provide the interconnects for the flow inlets and outlets of the channels to be connected to the external pressure lines and reagent supply reservoirs. The bottom surface of the elastomer chip with the exposed flow channels is then rinsed off with methanol to hydrophilize the surface and placed on a microscope cover glass that has been thoroughly cleaned in piranha solution (3:1 solution of H2SO4 and 30% H2O2) for 10 min. The cover glass is slightly larger than the silicone chip to provide an exposed rim for the subsequent packaging in a block of epoxy resin. It is worth emphasizing that no further baking step to bond the cover glass to the elastomer chip is necessary, as the subsequent embedding in the epoxy resin holds both parts together. This allows the use of chemically functionalized, e.g., protein- or nucleic acid-coated cover glasses, as these coatings (9) Duffy, D. C.; McDonald, J. C.; Schueller, J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974-4984.
are not degraded by the elevated temperatures of the annealing process or by volatile components that are emitted from the silicone in such a process. Because the manual process that punches the access holes in which the interconnect tubing is inserted is tedious and somewhat unreliable, we have tried an alternative fabrication process in which a special casting chamber was used to create the upper hard elastomer block with the channels and access holes in one step. This casting chamber has the silicon master mold as its bottom and a lid with protruding stainless steel rods as molds for the access holes on top. After the lid is carefully aligned with the bottom mold to ensure that the access holes are correctly positioned with respect to the channels, the casting chamber is filled with the elastomer and cured as described above. This technique reduces the manual labor to punch the access holes into the elastomer and ensures that these holes are perfectly perpendicular to the bottom surface of the chip. However, the difficulties with the alignment of the lid with respect to the silicon mold were such that this method did not speed up the overall manufacturing process for the small number of laboratory prototype chips that we produced. Nonetheless, we remain confident that with an improved engineering effort aimed at resolving the alignment problems a reliable all-in-one casting mold that is suitable for large-scale manufacturing can be devised. Embedding the Chip and Interconnects. The microfluidic chip with the inserted interconnect tubes is embedded in a block of an optically transparent two-component epoxy resin that cures at room temperature. For this process, we use a casting mold, which consists of two square horseshoe aluminum pieces that are screwed together. The bottom of this casting chamber is formed by a strip of adhesive tape. The chip is placed inside the chamber such that the cover glass is protected by the adhesive tape, as shown in Figure 2. Then the epoxy resin (Tra-bond 2115,Tra-con Inc.) is poured into the chamber to a level of 5-7.5 mm above the elastomer chip. The epoxy adheres well to the exposed rim of the cover glass and to the protruding steel interconnect tubes, but not to the inert PDMS surface. As the epoxy resin cures, it shrinks slightly and forces the elastomer to seal tightly against the cover slip. After 10 h of curing at room temperature, the epoxy hardens to form an extraordinarily stable encapsulation around the silicone chip and holds the interconnect tubes firmly in place. At this point, the adhesive tape can be removed from the bottom of the casting chamber to reveal a pristine glass surface with direct optical access to the flow channels for microscopy or optical micromanipulation applications. It is also possible to remove the epoxy block with the embedded chip from the casting frame at this time, in particular if the aluminum frame has been coated with a silicone sealant prior to casting. We found, however, that it is convenient to leave the chip in the casting frame and use the frame to mount the microfluidic device on an inverted microscope that is used for most of our applications. For this purpose, the casting and mounting frame features appropriate holes for direct attachment to the microscope stage. Before the microfluidic device with the integrated valves can be used, it is necessary to fill the control channels for the valves with water. This priming is necessary as the silicone elastomer is slightly permeable to most nonpolar gases, and pressurizing the control lines with air would lead to a penetration of air through
Figure 2. (a) Sketch of the casting chamber with the adhesive tape seal on the bottom and the elastomer chip with attached cover glass and inserted interconnect tubes. (b) Photograph of the casting chamber and elastomer chip before assembly and filling with epoxy resin.
the valve membranes into the flow channels, which in turn results in undesirable air bubbles in the flowing liquid. Thorsen et al.,8 demonstrated that the priming of the valves can be achieved by pressing water into the control channels and forcing the air that is trapped in the dead ends into the elastomer bulk. This technique still works in our case where the elastomer is tightly sealed in the epoxy packaging block. We presume that the air is drained from the elastomer through the flow channels that are not filled with liquid at the time of the priming. We also found it advantageous to reduce the amount of air that is trapped in the control channels by evacuating them first and then backfilling them with water under pressure in excess of 140 kPa through a three-way valve. This significantly reduces the time required for priming the valves. Sockets for External Connections. With the interconnect tubing held firmly in place by the epoxy block, we can build a socket for the chip that provides the external connections for reagents and pressure lines. Our design places eight connectors for the pins from the chip in a square pattern on a brass socket body. This enables us to have a large clearance hole in the middle of the socket, where the active elements and other main features of the microfluidic chip are located. This gives us optical access from both sides to the central region of the chip. Each of the connectors on the socket, as shown in Figure 3, consists of a short (10 mm) piece of Tygon microbore tubing with an inner diameter of 0.76 mm, which is glued with a two-component epoxy glue from the backside into the body of the socket. The pins of the chip fit Analytical Chemistry, Vol. 75, No. 19, October 1, 2003
5289
Figure 3. Schematic diagram of the socket for the external reagent and pressure connections. The chip with the embedded stainless steel interconnect tubes is inserted into the conical holes on the bottom of the socket. On the inside, a tight seal is provided by short pieces of Tygon microbore tubing, which is glued into the chip with a twocomponent epoxy glue and attached to stainless steel terminal tubes on the other side. A large central hole provides optical access to the chip from this side.
from the front side of the socket into these pieces of tubing, which provides a tight seal akin to an O-ring. To make the insertion of the pins into the tubing as easy as possible, even if the pins are slightly bent or otherwise misaligned, the holes on the front side of the socket are conically shaped to guide the pins into the tubing seals. On the backside of the socket, the tubing is attached to metal pipes that in turn are connected to external reagent reservoirs and computer-controlled pressure lines. The connections to this socket can be flexibly configured to deliver any combination of control pressure for the valves, provide reagent for the fluidic inlet, or accept products from the chip. The number of connections can easily be scaled up, and larger chips can be accommodated as well. In Situ Attachement of Single DNA Molecules. The cover glasses that form the bottom of the flow channels are precoated with polyclonal digoxigenin antibodies from sheep (Roche Diagnostics) through passive adsorption. For this aim, piranha-cleaned cover glasses were incubated with a 25 µg/mL antibody solution in PBS (8 mM Na2PO4, 1.5 mM KHPO4, 2.7 mM KCl, 130 mM NaCl, pH 7.4) for 3 h. The antibody-coated cover glass is then sealed onto the two-layered microfluidic elastomer chip and, together with the interconnects, embedded in the epoxy resin as described above. End-functionalized DNA is prepared by enzymatically incorporating digoxigenin- and biotin-labeled nucleotides into the ends of λ-bacteriophage DNA (Gibco).10 Fluorescent streptavidin-coated polystyrene microspheres of 0.65-µm diameter (Bangs Laboratories) are attached to the biotin-labeled end of the DNA molecule through overnight incubation at room temperature in TE-NaCl buffer (10 mM Tris-Cl, 200 mM NaCl, 0.1 mM EDTA, pH 8.0). The concentration of DNA was chosen such that on average less (10) Zimmermann, R. M.; Cox, E. C. Nucleic Acids Res. 1994, 22, 492-497.
5290 Analytical Chemistry, Vol. 75, No. 19, October 1, 2003
Figure 4. Photographs of assembled microfluidics platform. (a) The chip in its aluminum mounting frame is ready to be plugged into the socket with the connections to the external reagent and pressure delivery system and the central hole for bright-field illumination. (b) The microfluidic chip is mounted on the stage of an inverted microscope and optical tweezers workstation, and all external connections are made through the socket on top.
than one DNA molecule binds to each microsphere. The DNAmicrosphere constructs are then introduced into the digoxigen antibody-modified microfluidic channels using the socket connections for reagent delivery. Once the channels are filled, the device is left to incubate overnight, allowing the digoxigenin-labeled end of the DNA molecule to bind to the antibody-coated cover glass. Any unattached DNA molecules and microspheres are then flushed out of the flow channels with a buffer solution, leaving only microspheres that are tethered to the bottom of the flow channel through a DNA molecule. The Brownian motion of the tethered microsphere can be observed in fluorescence microscopy. The microsphere also provides a handle for micromechanical manipulations of the DNA molecule with optical tweezers.11 RESULTS AND DISCUSSION Figure 4 shows the complete system with the embedded silicone microfluidic chip, its mounting frame, and the socket for the external reagent and pressure delivery system attached to the stage of an inverted microcope and optical tweezing station. The central region of the chip is accessible through the cover glass for observation in epifluorescence microscopy and manipulation with optical tweezers. The large central hole in the socket and the transparency of the embedding epoxy resin allow bright-field illumination from the backside of the chip. The mechanical stability of the chip and its interconnects, as well as the attachment to the external infrastructure, is excellent. Chips can be plugged into the socket with the external connections in a matter of seconds, and the exchange of chips is easy. This fulfills a key prerequisite for transferring this technology into the realm of routine applications. With this mechanical stability, the resultant flow rates in the channels are stable as well and not affected by strain on the connecting tubing or motion of the microscope stage. Chips can be repeatedly plugged in and unplugged, without any degradation or detachment of the interconnects. The bond between the epoxy resin and the interconnect tubing is excellent; (11) Wang, M. D.; Yin, H.; Landick, R.; Gelles, J.; Block, S. M. Biophys. J. 1997, 72, 1335-1346.
a 5-mm-thick epoxy layer gives the interconnects enough stability to withstand rough handling. The seal between the flow channels and the cover glass is good because the epoxy packaging block presses the elastomer against the cover glass. This eliminates the need to anneal this interface to produce a more stable bond. In fact, the flow channels of chips that are assembled and encapsulated at room temperature can withstand pressures of up to ∼140 kPa before the elastomer detaches from the coverslip. This compares to a pressure of ∼30 kPa at which nonembedded chips that rely purely on the adhesion between the glass and silicone elastomer fail when not enhanced by an additional annealing step, which, however, prevents the use of most chemically functionalized surfaces. Thus, this technique opens up the possibility of using chemically functionalized cover glasses as the bottom of the flow channels, enabling a wide range of experimental applications in which it is desirable to combine surface chemistry with microfluidics, such as a combination of conventional DNA microarrays and a microfluidic reagent delivery system. As an example of the utility of our microfluidic platform, we report the successful micromechanical manipulation of single DNA molecules inside the flow channel using optical tweezers. The DNA molecules were attached on one end to the digoxigenincoated bottom surface of the flow channel and on the other end to a polystyrene microsphere. An Nd:YAG (λ ) 1064 nm) laser beam was focused to a diffraction-limited spot inside the flow channel with an oil-immersion microscope objective (Zeiss Plan Neofluar 100 × 1.3 oil). The DNA-tethered polystyrene microspheres were trapped in this focal spot, and the DNA molecule was stretched by moving the microfluidic chip on the microscope stage with respect to the laser focus, exploiting our ability to mount the microfluidic chips easily and sturdily on the microscope. We were able to stretch the DNA molecule to an extension of 16.0 µm, which corresponds to a force of 33 pN. These experiments demonstrate that protein-coated cover glasses can be used as substrates for the microfluidic chips to enable applications that require in situ surface chemistry in the flow channels. It further shows that single-molecule experiments requiring superior optical accessibility of the flow channels are feasible. Since the main advantage of performing single-molecule experiments on a microfluidc chip is that a molecule can be studied under rapidly changing chemical conditions, we demonstrate a rapid buffer exchange in a reaction area on the chip. The chip has three reagent supply channels and one reaction channel leading to an outlet. Each of these four channels can be closed independently with a pressure-actuated valve. For the demonstration experiment shown in Figure 5, two of the supply channels were filled with a red dye solution, while the third one was filled with a green dye. When the appropriate valves on the chip were opened, dye solution flowed through the reaction channel. By repeatedly opening and closing the valves that control the flow of the dye solutions, we created an alternating stream of red and green dye solutions in the reaction channel. We anticipate that this technique will be very useful for a variety of single-molecule studies, such as DNA-protein binding experiments, in which the
Figure 5. Photograph showing the buffer exchange in a reaction channel on the chip. The reagent supply channels 1-3 converge into the reaction channel 4. Each of the flow channels can be independently controlled by a perpendicular, pressure-actuated control channel. (a) Supply channels 1 and 2 have been filled with a red dye solution, and supply channel 3 contains a green dye solution. The control channel that switches the fluid flow in channel 3 is visualized by filling it with a black dye solution. (b) shows the chip after several cycles of alternating the flow of the two dyed buffer solutions from channels 1 and 3 into the reaction channel by opening and closing the appropriate valves. In this way, a molecule attached to the bottom of the flow channel can be exposed to a rapidly changing chemical environment.
mechanical properties of the same molecule under different buffer conditions are investigated. Our microfluidic system satisfies all the requirements for such experiments, as we have the ability to attach single DNA molecules to the bottom of the flow channels, full optical access with a high numerical aperture objective to the central region of the chip, and a reliable fluid delivery system. In conclusion, we have demonstrated an ultrastable, versatile, and easy-to-use microfluidic platform for demanding optical singlemolecule experiments. Our system allows the incorporation of active elements such as valves and pumps on the chip that makes multilayer soft lithography so attractive for a wide range of applications. Furthermore, we maintain a clean glass surface at the bottom of the flow channels, which can be chemically functionalized, and we have full optical access to both sides of the chip. The stability of the interconnects and socket for external connections make the chips sufficiently rugged for routine applications in the research laboratory and beyond. ACKNOWLEDGMENT The authors thank J. L. Bourjaily for the design and testing of the casting chamber to make elastomer chips with integrated access holes, S. R. Quake and co-workers for helpful discussions, and S. Burns and R. Kennedy for access to their cleanroom. This work was supported by the National Institutes of Health Grant GM 65934-01, the Research Corporation, and the Alfred. P. Sloan Foundation. Received for review February 21, 2003. Accepted July 17, 2003. AC034179I
Analytical Chemistry, Vol. 75, No. 19, October 1, 2003
5291