Hard Top Soft Bottom Microfluidic Devices for Cell Culture and

Apr 21, 2009 - 7.5 mm tall, with the channel features on the top face. This block was then placed in a square Petri dish, covered with PDMS prepolymer...
2 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2009, 81, 3714–3722

Articles Hard Top Soft Bottom Microfluidic Devices for Cell Culture and Chemical Analysis Geeta Mehta,† Jay Lee,‡ Wansik Cha,† Yi-Chung Tung,† Jennifer J. Linderman,*,†,‡ and Shuichi Takayama*,†,§ Department of Biomedical Engineering, Department of Chemical Engineering, and Macromolecular Science and Engineering, University of Michigan, Ann Arbor, Michigan 48108 We report fabrication and characterization of microfluidic devices made of thermoplastic and elastomeric polymers. These hard-soft hybrid material devices are motivated by the combined need for large scale manufacturability, enhanced barrier properties to gas permeation and evaporation of aqueous solutions compared to poly(dimethyl siloxane) (PDMS) devices, and compatibility with deformation-based actuation. Channel features are created on rigid polymers such as polyethylene terephthalate glycol (PETG), cyclic olefin copolymer (COC), and polystyrene (PS) by hot embossing. These “hard tops” are bonded to elastomeric “soft bottoms” (polyurethane (PU) or PDMSparylene C-PDMS) to create devices that can be used for microfluidic cell culture where deformation-based fluid actuation schemes are used to perfuse and recirculate media. The higher barrier properties of this device compared to PDMS devices enable cell culture with less evaporation and creation of hypoxic conditions. Microfluidic devices are used for myriad applications, including biotechnology, analytical technologies, heat transfer, and power generation.1-4 They enable efficient synthesis of chemicals, low consumption of expensive reagents, simultaneous execution of large numbers of analyses, enhanced control of cellular microenvironments, and integration of sample introduction to chemical separation and detection on single, miniaturized µTAS devices.1-6 Different applications demand different device properties leading * To whom correspondence should be addressed. E-mail: [email protected] (J.J.L.), [email protected] (S.T.). Phone: (734)763-0679 (J.J.L.), (734)6155539 (S.T.). Fax: (734)763-0459 (J.J.L.), (734)936-1905 (ST) (S.T.). † Department of Biomedical Engineering. ‡ Department of Chemical Engineering. § Macromolecular Science and Engineering. (1) Beebe, D. J.; Mensing, G. A.; Walker, G. M. Annu. Rev. Biomed. Eng. 2002, 4, 261–286. (2) Makamba, H.; Kim, J. H.; Lim, K.; Park, N.; Hahn, J. H. Electrophoresis 2003, 24, 3607–3619. (3) Weibel, D. B.; Whitesides, G. M. Curr. Opin. Chem. Biol. 2006, 10, 584– 591. (4) Whitesides, G. M.; Ostuni, E.; Takayama, S.; Jiang, X.; Ingber, D. E. Annu. Rev. Biomed. Eng. 2001, 3, 335–373. (5) Dittrich, P. S.; Manz, A. Nat. Rev. Drug Discovery 2006, 5, 210–218. (6) El-Ali, J.; Sorger, P. K.; Jensen, K. F. Nature (London) 2007, 442, 403– 411.

3714

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

to the use of a wide range of materials in device fabrication.1-13 One way to categorize materials used for microfluidic device fabrication is by rigidity. Rigid thermoplastics are advantageous in being readily molded by embossing or injection molding and in having structures which are more precise and stable. On the other hand, elastomers are advantageous in being flexible and useful for deformation-based valving and pumping. Since these properties are exclusive of each other, no one polymer can provide the advantages of both. What would be useful in some cases, therefore, is to create microfluidic systems composed of a combination of “hard” and “soft” polymers. All-cyclic olefin copolymer (COC), all-polystyrene (PS), allpoly(methyl methacrylate) (PMMA), all- glycol-modified polyethylene terephthalate (PETG), and all- poly(dimethyl siloxane (PDMS) microdevices are fairly common in literature.8-11 Bonding a hard top and an elastomeric bottom is less common but has been shown in the literature for bonding glass, quartz, SiO2 coated materials to elastomeric PDMS, and perfluoropolyethers.12 However, the hard materials used in these reports (glass, quartz, SiO2) are difficult or expensive to mass produce and are a challenge to work with. Rigid PMMA and elastomeric PDMS have also been bonded together to create a laminated microfluidic device for immunosensing.13 However, PDMS is porous and hence not suitable for applications that require a strong barrier to gas and moisture. Bonding of rigid materials that are readily molded with elastomeric materials that have low permeability to gas and moisture has not been shown. One specific motivation for development of hard-soft hybrid microfluidic chips is for use in deformation-based microfluidic systems. Our laboratory has established the control of fluid flow (7) Shim, J.; Bersano-Begey, T. F.; Zhu, X.; Tkaczyk, A. H.; Linderman, J. J.; Takayama, S. Curr. Top. Med. Chem. 2003, 3, 687–703. (8) Lin, R.; Burns, M. A. J. Micromech. Microeng. 2005, 15, 2156–2162. (9) Nielsen, T.; Nilsson, D.; Bundgaard, F.; Shi, P.; Szabo, P.; Geschke, O.; Kristensen, A. J. Vac. Sci. Technol., B: Microelectron. Nanometer Struct. 2004, 22, 1770–1775. (10) Steigert, J.; Haeberle, S.; Brenner, T.; Muller, C.; Steinert, C. P.; Koltay, P.; Gottschlich, N.; Reinecke, H.; Ruhe, J.; Zengerle, R.; Ducree, J. J. Micromech. Microeng. 2007, 17, 333–341. (11) Tsao, C. W.; Hromada, L.; Liu, J.; Kumar, P.; DeVoe, D. L. Lab Chip 2007, 7, 499–505. (12) Osterfeld, S. J.; Wang, S. X. US Patent 7419639, United States, 2008. (13) Ko, J. S.; Yoon, H. C.; Yang, H.; Pyo, H.-B.; Chung, K. H.; Kim, S. J.; Kim, Y. T. Lab Chip 2003, 3, 106–113. 10.1021/ac802178u CCC: $40.75  2009 American Chemical Society Published on Web 04/21/2009

in PDMS microdevices using Braille tactile arrays.14-17 Braille arrays are beneficial because they allow computer programmed multidirectional perfusion or fluid recirculation in a portable format that requires no tubings or interconnects. One of the challenges that arises in many studies using PDMS microchannels, including previous studies using Braille actuation, is evaporation of liquids through microfluidic devices.16,18 Evaporation of water from the microchannels often leads to changes in the concentration and osmolality of the solutions, harming cells under culture and confounding assay readouts. Thermoplastic polymers such as PETG, COC, and PS have significantly smaller water vapor transmission rates compared to PDMS (0.75-6 versus 125 g mm/ m2-day;19 see also Supporting Information, Table S1), which makes the microfluidic devices fabricated using these materials less susceptible to evaporation of aqueous phases as compared to all-PDMS devices. In addition to permeability to water vapor, PDMS is also highly gas permeable (∼52000 cm3-mm/m2-day-atm; 19 see also Supporting Information, Table S1 for comparisons). This can be an advantage in some cases but is an impediment for study of cells under hypoxic conditions. In our previous work, for example, we have shown the modulation of oxygen tension in PDMS microbioreactors bonded with PDMS-parylene C-PDMS membranes and found that dissolved oxygen can be regulated from 20% (ambient oxygen levels) to 4.3% by operating the device at different cell densities and perfusion velocities.20 However, it was difficult to get lower oxygen tensions (of the order of 1-2%) in a microfluidic device made of gas permeable PDMS. Such low oxygen environments are needed for the biological studies of embryonic, adult stem cells, cancer, as well as for studies of the diseases of the circulatory system and bone tissue engineering. Although glass or other rigid material channels could be used for lowering oxygen concentration and increasing barrier to evaporation in microdevices, it would not be possible to have dynamic pumping and valving using deformation-based fluid channel actuation. In this report, we describe microfluidic devices made of several different combinations of rigid thermoplastics with flexible elastomeric polymers. The particular combinations of materials tested are a rigid channel layer fabricated using PETG, COC, or PS and an elastomeric membrane of polyurethane (PU) or PDMS coated with parylene C to seal the channels. We then evaluate the following: (1) ability to perform deformation-based fluid actuation, (2) strength and stability of the bond between the rigid and flexible polymers over time, (3) water permeability, particularly ability to minimize evaporation and retain aqueous solutions within the microchannels for prolonged periods to maintain mammalian cell cultures, and (4) gas permeability, particularly the ability to (14) Futai, N.; Gu, W.; Song, J. W.; Takayama, S. Lab Chip 2006, 6, 149–154. (15) Gu, W.; Zhu, X.; Futai, N.; Cho, B. S.; Takayama, S. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 15861–15866. (16) Heo, Y. S.; Cabrera, L. M.; Song, J. W.; Futai, N.; Tung, Y.-C.; Smith, G. D.; Takayama, S. Anal. Chem. 2007, 79, 1126–1134. (17) Song, J. W.; Gu, W.; Futai, N.; Warner, K. A.; Nor, J. E.; Takayama, S. Anal. Chem. 2005, 77, 3993–3999. (18) Walker, G. M.; Zeringue, H. C.; Beebe, D. J. Lab Chip 2004, 4, 91–97. (19) Massey, L. K. Permeability and Other Film Properties of Plastics and Elastomers: A Guide to Packaging and Barrier Materials, 2nd ed.; William Andrew Inc.: Norwich, NY, 2003. (20) Mehta, G.; Mehta, K.; Sud, D.; Song, J.; Bersano-Begey, T.; Futai, N.; Mycek, M.-A.; Linderman, J. J.; Takayama, S. Biomed. Microdevices 2007, 9, 123– 134.

suppress oxygen permeation through the channel materials and enable cells to create hypoxic environments by oxygen depletion. EXPERIMENTAL METHODS Materials. For hot embossing, the following three polymers were obtained: COC (Zeonor 1060R and 1020 R from Zeon Chemicals L.P., Louisville, KY); PETG (304.8 mm × 304.8 mm × 6.35 mm from McMaster Carr, Robbinsville, NJ); PS (100 mm × 15 mm sterile Petri dishes Cat. 0875712 - 2071 from Fisherbrand, Fisher Scientific, Pittsburgh, PA). For making elastomeric PU membranes, medical-grade aliphatic thermoplastic PU was used (TecoflexSG93A and SG80A, Thermedics Incorporated, Woburn, MA) as a 5% solution by weight in 95% chloroform (Sigma). Epoxy molds were made using high temperature epoxy (Epoxy Cytec Industries Inc., Conapoxy FR-180, Parts A and B). PDMS was used as 10:1 base:curing agent (Dow Corning Sylgard 184). Tridecafluoro-(1,1,2,2-tetrahydrooctyl)-1-trichlorosilane was purchased from United Chemical Technologies Inc., Bristol, PA. Negative photoresist SU-8 50 was purchased from MicroChem Corp. (Newton, MA). Fabrication of the Mold. Fabrication of SU-8 on Glass Wafer Mold. Fabrication of the mold and hot embossing process are shown in Figure 1. Briefly, a glass wafer mold was fabricated using backside diffused-light photolithography21 and conventional photolithography to form positive relief features made of SU-8 with two different heights onto a thin glass wafer (200 µm thick). First, bell shaped channels were constructed on a 30 µm SU-8 layer using backlight soft lithography for Braille display pumping and valving. This was followed by patterning the connected rectangular X region on another 200 µm SU-8 layer by conventional photolithography as shown in Figure 1A. The resulting glass wafer mold was taped to a single glass slide (75 mm × 50 mm) using doublesided tape and silanized with tridecafluoro-(1,1,2,2-tetrahydrooctyl)1-trichlorosilane. This silanized glass wafer mold was then used to create a PDMS master. Fabrication of PDMS Master (Using Soft Lithography). The glass wafer mold supported on a glass slide was stacked on six additional glass slides (also 75 mm × 50 mm) using double-sided tape between each slide. The resulting block was approximately 7.5 mm tall, with the channel features on the top face. This block was then placed in a square Petri dish, covered with PDMS prepolymer (10:1 base to curing agent (by mass) and degassed in a vacuum chamber), and then cured overnight at room temperature on a level surface (Figure 1B). It was critical that the PDMS cured on a level surface. Any deviations from levelness would be transferred to the subsequent epoxy mold and result in uneven embossing. The levelness of the surface was easily verified using a simple, liquid/air-bubble leveling device. The resulting PDMS master had the channel features set into the bottom of a “dish”, and allowed the master to be used as a molding pattern for the subsequent epoxy mold. Critical dimensions used were as follows: total height of PDMS master, 11 mm; depth of the “dish”, 7.5 mm; and PDMS height above glass wafer mold, 3.5 mm. Fabrication of Epoxy Mold. High temperature epoxy (83:100, hardener/epoxy by mass) was poured into the PDMS master “dish” (Figure 1B). The PDMS master (with epoxy) was then placed onto a level surface in a 120 °C oven. After about 5-10 (21) Futai, N.; Gu, W.; Takayama, S. Adv. Mater. 2004, 16, 1320–1323.

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3715

Figure 1. Schematic of hard top soft bottom device fabrication process. (A) SU-8 mold is made on the glass wafer by photolithography in a two step process, (B) PDMS master is made by curing PDMS over the SU-8 features, finally high temperature curing epoxy is poured onto PDMS replica to make an epoxy mold. (C) The hard top (COC, PETG, or PS) is placed on a metal sheet between the heated platens of compression press, the epoxy mold is set over the hard top, and heat and pressure are applied for a specific time, the molded hard top is removed from epoxy mold, and then bonded with either PU membrane or PDMS-parylene C-PDMS membrane by plasma oxidation and heat treatment.

min, the PDMS master was briefly removed from the oven, and a pipet tip was used to agitate the epoxy and remove all air bubbles trapped in the epoxy near the channel features (the epoxy is considerably less viscous after 5-10 min in oven, making removal of the air bubbles fairly easy). The PDMS master was then placed back into the oven for ten more minutes, after which it was checked again and any remaining air bubbles removed in the same way described above. The PDMS master was then placed back 3716

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

into the oven, and allowed to cure undisturbed for approximately 4 h. After curing the epoxy and cooling it for approximately 1 h, the hardened epoxy was carefully removed from the PDMS master. The resulting epoxy mold was checked for levelness. Hot Embossing. A 12 ton hot press with a platen size of 228.6 mm × 304.8 mm with a mold area of 203.2 mm × 254 mm and a manual hydraulic press were used for embossing channels (model 912, maximum sustained temperature 177 °C, Jackson Marking

Products Co., Inc., Mt. Vernon, Illinois) (Figure 1C). Before embossing a material, we had to find the appropriate embossing temperature that would allow imprinting of channel features on the substrates. If the temperature was too low, the material would not be soft enough to flow and might result in the destruction of raised features (such as the “X” region in our design which is 200 µm high, compared to the rest of the channels which are 30 µm high) on the epoxy mold as it is pressed against the relatively hard material. On the other hand, if the temperature was too high, the material could degrade. To find this optimal embossing temperature (Temboss), qualitative investigation with the polymers was performed by subjecting the materials to contact with epoxy mold at different temperature and same pressure. The Temboss of polymers used in this report are as follows: COC (Zeonor 1060R) Temboss ) 150 °C, PS Temboss ) 125 °C, and PETG Temboss ) 140 °C. The glass transition temperature (Tg) of these polymers are as follows: COC (Zeonor 1060R) Tg ) 102 °C, PS Tg ) 95 °C, and PETG Tg ) 105 °C. A silanized glass plate was centered on the metal tray with the silanized surface facing up and loaded into the press. Upon reaching Temboss, the tray was removed from the press and unembossed material was placed in the center of the silanized glass plate. The tray was loaded back into the press, and the lower platen was raised until the unembossed material was a few millimeters away from the top platen. This ensured quick and even heating of the unembossed material. After 4-5 min, the tray was quickly removed from the press, and the material was probed to verify that it was soft enough (or reached the glass transition temperature) for embossing (it deforms easily when gently poked with a wooden pick). The epoxy mold was placed face-down on the softened material, and the tray was loaded back into the press. The bottom platen of the press was raised slowly until the pressure gauze indicated the desired pressure. The pressure decreased to 0 as the material flowed around the epoxy mold. This procedure constitutes one “press”. Two-four “presses” were performed to correct for any unevenness in the epoxy mold and obtain the desired thickness: when the embossed material flows up and around the edges of the epoxy mold, enough “presses” have been made. The number of “presses” vary according to the thickness and type of material, as well as the levelness of the epoxy mold. After embossing, the glass plate, the embossed material, and the epoxy mold were removed off the tray and placed onto a cooling rack for 10-15 min after adding some weights on top to prevent the embossed material from warping as it cools. The embossed material was then separated from the epoxy mold and cut to rectangular shape using an electric saw. The holes for liquid inlet and outlet, as well as cell seeding ports, were made on the embossed material using a drill press (Figure 2A). To make PS embossed channels, a 2 mm thick PS sheet was made first from 4 layers of edge-cut Petri dishes (Fisherbrand, 100-15 mm, sterile). The sheets were melted together at 125 °C using the press to make a thicker PS sheet. Device Assembly. To assemble a device, the hole punched hot embossed top needed to be bonded to an elastomeric soft bottom (schematic shown in Figure 2B). To make 200 µm membranes from PU, 8 mL of the PU-chloroform solution (5%

PU by weight) was placed in glass molds and left covered to evaporate overnight. The membranes were cut out of the glass mold the next day and placed on silanized glass slides. PDMSparylene C-PDMS membranes were fabricated as described earlier.16,20 The hard top and soft bottom were bonded using an Expanded Plasma Cleaner & PlasmaFlo (PDC-001 (115V), 200 W input, Harrick Plasma, Ithaca, NY) with argon and oxygen gases. The ratio of argon to oxygen inside the plasma chamber was 2:1. The channel piece and membrane were plasma oxidized for 1 min at 400 mTorr. After removal from the plasma etcher, they were placed on top of each other, any remaining air bubbles were removed, and then the bonded devices were left in a 60 °C oven with some weights (∼1 lb) on top until the next day. When ready to be used, the devices were made hydrophilic again and sterilized by plasma oxidation for 1 min, and sterile Dulbecco’s Phosphate Buffer Saline (DPBS) was introduced into the channels (sealed device filled with green dye can be seen in Figure 2C). These devices were then additionally sterilized by placing under a UV-C light for ∼30 min as a precautionary step, as UV-C light (100-290 nm wavelength) is germicidal. Bond Strength Test. Measurement of burst pressure needed to break the seal between PETG and PU was performed by recording the nitrogen gas pressure range that separated the two device layers from each other. The test was performed on devices stored at room temperature and humidity, and for various time points when stored with liquid inside microchannels in a humidified incubator (37 °C). Braille Actuation. An array of 48 pin actuators adapted from a Braille display module (SC9, KGS, Saitama, Japan) was used for fluid actuation.14 The pin actuator module was controlled with a computer via Universal Serial Bus (USB) through a finger-sized stand alone custom controller circuit board (Olimex, Plovdiv, Bulgaria).14,16 The microfluidic chip interfaces with the pin actuator module by simply holding the chip in place such that the channels align with the pins which push upward closing the channel, as seen in Figure 2D. The pin movements for valving and pumping were controlled with a custom computer program written in C sharp. A stereoscope (Nikon Stereoscopic Zoom Microscope SMZ1000) was used to align the microchannels over the Braille pins. Microscopy. Pictures of the devices and channel cross sections were taken using an inverted phase contrast microscope (Nikon TE 300) and a digital CCD camera (Hamamatsu) at different magnifications. A scanning electron microscope (SEM) (Philips XL30FEG) was used for imaging the cross sections of the embossed channels. To prepare the samples for imaging, the channels were coated with carbon (SPI Supplies, Module Carbon Coater) for 2 min. Simulation of Deformation-Based Valving. To investigate the valving behaviors of the fabricated microfluidic devices, we performed simulations using finite element analysis software, ANSYS 10.0 (ANSYS Inc., Southpointe, PA). To reduce the computational time, a three-dimensional (3-D) model of half of the device with a symmetric boundary condition was constructed as shown in Supporting Information, Figure S1A. Supporting Information, Figure S1B shows the cross-sectional view of the constructed model from the symmetry plane. The model was Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3717

Figure 2. Microfluidic device made by hot embossing. (A) Hot embossed COC device, (B) overall schematic of the device showing two components: channel layer and membrane layer, (C) an actual hard top-soft bottom device with green food dye, and (D) device on the Braille array.

composed of a microfluidic channel with a bell-shape cross-section, which was measured experimentally, and a Braille pin structure. Two sets of contact analysis were performed to simulate the interactions between the Braille pin and the membrane and the channel and the membrane, respectively. The 3-D contact and target elements, CONTA173 and TARGE170, were employed in contact analysis, and the deformation of the structure was simulated using 10-node tetrahedral 3-D solid element, SOLID92. The force boundary condition was assigned to the Braille pin to simulate its actuation motion. To simulate the large deformation of the structure, non-linear, large deformation static analysis was performed. Cell Culture. HepG2 cells (human hepatocellular carcinoma, ATCC, HB-8065) and C2C12 cells (mouse myoblasts, ATCC, CRL1772) were cultured in a medium comprising Dulbecco’s Modified Eagle’s Medium (DMEM, 11960, Gibco), 15% Fetal bovine serum 3718

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

(FBS, 10082, Gibco), 1% v/v antibiotic-antimicotic (15240, Gibco) and 1% v/v GlutaMAX2-I Supplement (35050, Gibco). MC3T3-E1 cells (mouse preosteoblast, ATCC, MC3T3-E1 subclone 4, CRL2593) were cultured in ascorbic acid free alpha-MEM medium (Minimal Essential Alpha Medium, 0010083DJ, Gibco) supplemented with 10% Fetal bovine serum, 1% v/v antibiotic-antimicotic and 1% v/v GlutaMAX2-I. Primary human dermal microvascular endothelial cells (HDMECs) were cultured in EGM-2MV (EGM2MV Microvascular Endothelial Cell Medium, CC-3202, Lonza) medium supplemented with 1% v/v antibiotic-antimicotic. The T-25 culture flasks were placed in a humidified 5% CO2 cell culture incubator. At 70-80% confluence, the cells were passaged by washing in PBS and incubated with 0.25% Trypsin/ EDTA (Invitrogen, Carlsbad, CA) for 2 min. The trypsin solution was neutralized with 15% FBS in DMEM and spun down with a centrifuge (ThermoForma, Marietta, OH) for 2

min at 4 °C and 1000 RPM. The supernatant was removed, and the pellet was resuspended in DMEM media. The pellet was reconstituted in 4 mL of DMEM media, and 1 mL of this suspension was used for creating a new plate of HepG2, C2C12, MC3T3-E1, and HDMECs. The cells were passaged every third day. When the cells were to be cultured in the PDMS microbioreactor, the pellet was reconstituted in 70-100 µL of DMEM media and then injected into the seeding ports of the chip. Cell Seeding in Devices. Fibronectin (100 µg/mL, F2006, Sigma) was pipetted into the microdevices and allowed to adsorb for 30 min, following which media (specific to the cells being seeded) was introduced in the device and the chip was placed on an array of pin actuators adapted from Braille displays for at least 1 h to peristaltically pump fluid through the channels. Cells (HDMECs, HepG2, MC3T3E1, C2C12) were seeded into the chip through the cell seeding ports. The cells were directed into the desired location inside the microdevice by using Braille pumping and valving. The cells were given 2-4 h to attach under no flow (pins up and channels valved closed) condition in a dry heat incubator (Forma 310 Series Direct Heat CO2 Incubator, Thermo electron Corporation, Marietta, OH) maintained at 37 °C and 5% CO2. Pumping was started after cells became adherent, and the chip was perfused for 12-14 h. Oxygen Measurement. A 0.2 mL volume of the oxygen sensitive dye, ruthenium tris(2,2′-dipyridyl) dichloride hexahydrate (RTDP) (93307, Fluka) (5 mg/mL in Dulbecco’s Phosphate Buffer Solution, PBS (14190250, Invitrogen)) was introduced into the inlet of the device, and pumping was continued for three more hours. The setup comprised of a microfluidic chip attached to the pin actuator array module was then taken to the microscope where lifetime measurements performed at room temperature. Cell densities in the cell culture channels in the microbioreactor chips were determined by counting the number of cells present in brightfield images taken at different positions along the length of the cell culture channel. The oxygen measurement system consisted of Nikon TS-100F microscope with fluorescence filter cube B-3A, LED light source (default is blue light with central wavelength of 470 nm (LXHLMB1D Blue Luxeon, Lumileds Lighting, LLC, CA), lock-in amplifier (SR830, Stanford Research System, Inc., CA), function generator (33220A, Agilent Technology, Inc., CA), and silicon PIN photodiode with preamplifier (PDA36A, Thorlabs, Inc., NJ). The system was controlled by a personal computer with LabVIEW graphic user interface program for operation and data acquisition through GPIB interface (GPIB-USB-HS, National Instruments, Co., TX). Supporting Information, Figure S2 shows the schematic of the Nikon TS-100F Microscope Fluorescence Intensity Detection System (FIDS). The numerical data from LabVIEW was saved in a .txt file containing values of time (in sec), intensity signal (in µV), and phase (in degree).22-25 (22) McDonagh, C.; Kolle, C.; McEvoy, A. K.; Dowling, D. L.; Cafolla, A. A.; Cullen, S. J.; MacCraith, B. D. Sens. Actuators, B 2001, 74, 124–130. (23) Szita, N.; Boccazzi, P.; Zhang, Z.; Boyle, P.; Sinskey, A. J.; Jensen, K. F. Lab Chip 2005, 5, 819–826. (24) Vollmer, A. P.; Probstein, R. F.; Gilbert, R.; Thorsen, T. Lab Chip 2005, 5, 1059–1066. (25) Zanzotto, A.; Szita, N.; Boccazzi, P.; Lessard, P.; Sinskey, A. J.; Jensen, K. F. Biotechnol. Bioeng. 2004, 87, 243–254.

The relationship between fluorescence intensity (or lifetime) anddissolvedoxygenconcentrationisdescribedbytheStern-Volmer equation:26-29 I0/I ) 1 + Kq[O2] or τ0/τ ) 1 + Kq[O2], where I0 or τ0 ) uninhibited sensor dye intensity or lifetime (i.e., 0% oxygen), I or τ ) sensor dye intensity or lifetime at oxygen level [O2] and Kq is the Stern-Volmer quenching constant. The system was calibrated by testing devices with just the dye and no cells and no oxygen for uninhibited sensor lifetime. The oxygen was removed by using oxygen quenching sodium thiosulphate supersaturated solution. Statistical analyses were performed using a Student’s t test with R ) 0.01. RESULTS AND DISCUSSION Hot Embossing of Hard Top Devices. A representative photograph of the hot embossed COC is observed in Figure 2A. Channel features were replicated onto the PETG, COC, and PS “hard tops” with good reproducibility for channels with different heights and shapes. All channels except the “X” region were bell shaped and 30 µm high, while the “X” region was rectangular and 200 µm high. SEM images of the top and cross-sectional views of the fabricated microfluidic channels demonstrate the high fidelity of the embossing processes, general smoothness of the features, and the bell-shape cross-sectional geometries that facilitate deformation-based fluid actuation (Supporting Information, Figure S3). Bonding of Hard Top Devices with Soft Elastomeric Membranes. Monolithic devices made of just COC, PETG, PS, or PDMS have been bonded together to form strong long-lasting devices using a wide range of methods.8,9,11,30-35 On the basis of these reports, we tried various ways to bond hard tops (COC, PETG, PS) to PU or PDMS membranes. Methods that resulted in an unstable bond between hard top and soft bottom included air or oxygen plasma treatment, chemical grafting on hard top, UV polymerization between hard and soft components, and thermal sealing. Because of the straightforward procedures and some bonding provided by oxygen plasma treatment,2,32-34,36 we further optimized this process. We successfully bonded devices by surface treating all components with 1 min of oxygen and argon (2:1) plasma, bringing the surfaces together, and heating in a 60 °C oven overnight with small weights (less than 1 lb) pressing (26) Gerritsen, H. C.; Sanders, R.; Draaijer, A.; Levine, Y. K. J. Flouresc. 1997, 7, 11–16. (27) Sud, D.; Zhong, W.; Beer, D. G.; Mycek, M. A. Opt. Express 2006, 14, 4412–4426. (28) Urayama, P.; Zhong, W.; Beamish, J. A.; Minn, F. K.; Sloboda, R. D.; Dragnev, K. H.; Dmitrovsky, E.; Mycek, M. A. Appl. Phys. B: Lasers Opt. 2003, 76, 483. (29) Zhong, W.; Urayama, P.; Mycek, M.-A. J. Phys. D: Appl. Phys. 2003, 36, 1689. (30) Cameron, N. S.; Roberge, H.; Veres, T.; Jakeway, S. C.; Crabtree, H. J. Lab Chip 2006, 6, 936–941. (31) Fredrickson, C. K.; Xia, Z.; Das, C.; Ferguson, R.; Tavares, F. T.; Hugh Fan, Z. J. Microelectromech. Syst. 2006, 15, 1060–1068. (32) Bhattacharyya, A.; Klapperich, C. M. Lab Chip 2007, 7, 876–882. (33) Mair, D. A.; Rolandi, M.; Snauko, M.; Noroski, R.; Svec, F.; Frechet, J. M. J. Anal. Chem. 2007, 5097–5102. (34) Mizuno, J.; Farrens, S.; Ishida, H.; Dragoi, V.; Shinohara, H.; Suzuki, T.; Ishizuka, M.; Glinsner, T.; Shoji, S. Proceedings of 2005 International Conference on MEMS, NANO and Smart Systems (ICMENS’05), IEEE 2005, 1–4. (35) Pal, R.; Sung, K. E.; Burns, M. A. Langmuir 2006, 22, 5392–5397. (36) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974–4984.

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3719

Figure 3. Braille actuation in COC-PU hard top-soft bottom devices (A and B) as compared to all PDMS device (C and D). Green dye-filled microfluidic channels stay atop white circular Braille pins as the pins move up (B and D) or down (A and C).

on the structure. Devices sealed in this way had stable bond strength at room temperatures for up to several months under dry conditions. There is abundant evidence in the literature of bonding of monolithic devices made of COC, PETG, PS, or PDMS, whose bond strengths are larger in comparison to our values for a measure of bond strength (burst pressure) for the hybrid hard top soft bottom devices.11,32-34 The technique presented here is a two step process: plasma bonding with argon and oxygen followed by heat treatment.32,37,38 At high humidity (for example, as in the humidified incubators), the device bonding lifetime was limited to about 14 days because of gradual delamination of the bottom elastomeric membrane from the hard top. To ensure that the channels are hydrophilic, the devices were reoxidized as a complete, bonded device for 2 min prior to use. The burst resistance of PETG-PU devices (used for mammalian cell culture and oxygen measurements) stored at room temperature for up to a month was above our measurement limit of ∼120 kPa (longer time periods at room temperature were not tested). After storing these dissimilar bonded devices in a humidified incubator at 37 °C, the burst pressure changed from ∼80 kPa after 24 h, to ∼60 kPa after 48 h, to ∼40 kPa after 72 h, and ∼20 kPa after 168 h. If PETG-PU microdevices were filled with aqueous solutions before storing them in the humidified incubator (37 °C), burst pressures were much lower than for devices stored without liquid, and the devices often delaminated after about 2 weeks of storage. We note that many bonds often reported to be irreversible,36 such as PDMS plasma bonded to PS or glass, will also often delaminate upon extended periods of exposure to moisture. Braille Actuation and Perfusion in Hard Top-Soft Bottom Devices. Analysis of fluid flow with Braille peristaltic pumping and valving was performed to confirm and validate perfusion in hard top-soft bottom devices. Devices were filled with green food dye and placed on top of a Braille array with channels aligned with respect to Braille pins. Figure 3 compares valving of COC(37) Millare, B.; Thomas, M.; Ferreira, A.; Xu, H.; Holesinger, M.; Vullev, V. I. Langmuir 2008, 24, 13218–13224. (38) Malek, C. K.; Thuillier, G.; Blind, P. Microsyst. Technol. 2004, 10 (10), 711–715.

3720

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

PU hard-soft hybrid material chips versus all soft chips constructed of PDMS. The Braille pin actuator can successfully valve a device made of a rigid top layer and elastic bottom layer. There were no leaks observed with repeated pumping and valving actions (over several days). When comparing hard top-soft bottom channel valving to that in the all-PDMS channels, the extent of channel closure and deformation is lower in the hard-soft hybrid material channels (see the area of closed microchannel in 3B compared to 3D). The deformation in hard top-soft bottom channels was further compared to a device made of all PDMS channels by simulations (model shown in Supporting Information, Figure S1). Figure 4 shows the simulated vertical displacement contours of the microfluidic devices when actuated by Braille pins. In the PDMSPDMS microfluidic channel, both the channel and the membrane are deformed to seal the channel, while in the COC-PDMS microfluidic channel only the membrane is deformed. Moreover, because of the higher Young’s modulus of the COC, the Braille pin causes less displacement in hardtop channel (ca. 55 µm) as compared to the PDMS channel (ca. 90 µm). The smaller displacement makes the alignment between the device and the Braille pin more critical. For instance, when the Braille pin is aligned 10% off center (32 µm) along the channel width, the PDMS-PDMS channel can still be completely sealed by Braille pin actuation for complete valving (Figure 4A). However, in COCPDMS channels, the corner of the channel will not be contacted with the membrane (Figure 4B), which starts to lead to incomplete valving (maximum non-contact gap size is 0.3 and 0.9 µm for 10 and 20% offset, respectively). For studies in which isolation of cells or analytes in particular compartments is important, or when it is necessary to know the flow rates in microchannels with very high precision, accurate alignment will be essential. However, good alignment can be achieved by using a stereoscope to position the microchannels over the Braille pins. Additionally, for many applications, including applications shown in this manuscript, some uncertainty in alignment can be tolerated since the leakage is still rather small.

Figure 4. Finite element analysis of vertical displacement (y-direction) of a channel in (A) PDMS-PDMS microfluidic device, and (B) hard top COC-PDMS microfluidic device with the 0% and 10% channel width misalignment of the Braille pin. The maximum non-contact gap size for 5% (16 µm), 10% (32 µm), and 20% (64 µm) offset is 0, 0.3, and 0.9 µm respectively (channel width is 320 µm). MN and MX refer to minimum and maximum vertical displacement of the channel (in µm), respectively, from the maximum (red color) at center of the Braille pin to the minimum (dark blue color) at furthest periphery where least amount of pressure is felt. Notice the higher vertical displacement (UY µm) in the case of PDMS-PDMS microchannel compared with COC-PDMS microchannel at both 0% and 10% offset. The maximum displacement is observed at the point where center of the Braille pin pushes against the microchannel. Table 1. Dissolved Oxygen Concentrations in the Microfluidic Devicesa oxygen concentration (%) cell type material

HepG2

C2C12

MC3T3-E1

PETG-PU COC-PU PS-PU All-PDMS

1.06 ± 0.02 1.25 ± 0.03 2.78 ± 0.03 3.49 ± 0.04

2.14 ± 0.02 2.36 ± 0.04 4.04 ± 0.03 4.97 ± 0.05

3.1 ± 0.04 3.27 ± 0.04 4.97 ± 0.04 5.88 ± 0.06

a

Mean ( SD.

Cell Culture in a Recirculation Loop in Hard Top-Soft Bottom Devices. To verify that our devices could be used for cell culture, we cultured primary human dermal microvascular cells (HDMECs) in a small amount of media (∼1 µL) inside a recirculation loop in the hard top-soft bottom device for 12-24 h outside the incubator. Cell survival is ∼100% and the cell morphology is healthy, as seen in Supporting Information, Figure S4. This is a significant improvement over previous work from our laboratory where because of evaporation we could only perform the outside culture of HDMECs in all-PDMS devices for less than 2 h.16 Oxygen Tension in Hard Top-Soft Bottom Devices. We quantified cell respiration-mediated oxygen depletion in our hard top-soft bottom devices for three cell lines at low flow rates and high cell densities to obtain a lower bound on the dissolved oxygen concentration inside microfluidic devices. Real-time measurement of fluorescence signal intensity and phase based on Nikon TS100F microscope with stage for Braille display systems was used for dissolved oxygen measurements in this report. Table 1 shows the lowest oxygen tension recorded in the hard top-soft bottom channels with three different cell lines (HepG2, C2C12, and MC3T3-E1) at a low average velocity of 5.4 × 10-7 m/sec and high average cell density of 1.54 × 109 cells/m2.

The average oxygen concentration at 25 °C is evaluated from the differences in the lifetime of the dye at normal room oxygen and that at downstream of cells in the devices. The lowest oxygen tensions in the hard top-soft bottom devices are significantly different (p < 0.01) from that in a PDMS device for all cell types. Microdevices made of PETG-PU have the lowest oxygen tension, followed by COC-PU, and last PS-PU, in agreement with the oxygen permeabilities of these materials listed in Supporting Information, Table S1. HepG2, C2C12, and MC3T3-E1 cells remained viable in these hypoxic conditions over long periods of time, up to 7 days (data not shown). Although dense, rapidly growing bacterial cultures often become hypoxic, monolayer mammalian cell cultures rarely deplete oxygen to such low levels in ambient atmosphere cultures. Such low oxygen tensions are useful for culture and study of embryonic and adult stem cells to control their differentiation capabilities39-46 and also for the study of ischemia injuries in specific organs.47-50 The hard top-soft bottom microfluidic devices offers a highly controlled in (39) Okazaki, K.; Maltepe, E. Regener. Med. 2006, 1 (1), 71–83. (40) Gassmann, M.; Fandrey, J.; Bichet, S.; Wartenberg, M.; Marti, H. H.; Bauer, C.; Wenger, R. H.; Acker, H. Proc. Natl. Acad. Sci. U.S.A. 1996, 93 (7), 2867–2872. (41) Bauwens, C.; Yin, T.; Dang, S.; Peerani, R.; Zandstra, P. W. Biotechnol. Bioeng. 2005, 90 (4), 452–461. (42) Kurosawa, H.; Kimura, M.; Noda, T.; Amano, Y. J. Biosci. Bioeng. 2006, 101 (1), 26–30. (43) D’Ippolito, G.; Diabira, S.; Howard, G. A.; Roos, B. A.; Schiller, P. C. Bone 2006, 39 (3), 513–522. (44) Cameron, C. M.; Harding, F.; Hu, W. S.; Kaufman, D. S. Exp. Biol. Med. 2008, 233 (8)), 1044–1057. (45) Niebruegge, S.; Bauwens, C. L.; Peerani, R.; Thavandiran, N.; Masse, S.; Sevaptisidis, E.; Nanthakumar, K.; Woodhouse, K.; Husain, M.; Kumacheva, E.; Zandstra, P. W. Biotechnol. Bioeng. 2009, 102 (2), 493–507. (46) Fehrer, C.; Brunauer, R.; Laschober, G.; Unterluggauer, H.; Reitinger, S.; Kloss, F.; Gu ¨ lly, C.; Gassner, R.; Lepperdinger, G. Aging Cell 2007, 6 (6), 745–757. (47) Shi, H.; Liu, K. J. Front. Biosci. 2007, 12, 1318–1328. (48) Buja, L. M. Cardiovasc. Pathol. 2005, 14 (4), 170–175. (49) de Perrot, M.; Liu, M.; Waddell, T. K.; Keshavjee, S. Am. J. Respir. Crit. Care Med. 2003, 167 (4)), 490–511.

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

3721

vitro model for creating hypoxic conditions by slow perfusion, reoxygenating by higher perfusion, as well as creating oxygen gradients20 along the length of a channel simulating physiological environments for certain cell types under various physiological and pathologic conditions. CONCLUSIONS This report describes versatile, cell culture-compatible, hard top-soft bottom microfluidic devices which combine the advantages of elastomeric all-PDMS devices (ability to perform deformation-based fluid actuation schemes) with the advantages of rigid plastic devices (low vapor and oxygen permeability, dimensional stability, and access to scalable manufacturing processes). We were able to imprint channel features reproducibly on PETG, COC, and PS hard tops with great fidelity by hot embossing. The channel features on hard tops were bonded to elastomeric but low permeability soft membranes to create microchannels. The bonding between dissimilar hard tops and soft bottoms was achieved by a combination of oxygen and argon plasma followed by heating. The resulting devices had lower evaporation as witnessed by outside incubator heated culture of HDMECS for 12 h and lower oxygen permeability as measured by cell respiration-mediated oxygen depletion, compared to all-PDMS devices. (50) Karhausen, J.; Ibla, J. C.; Colgan, S. P. Cell Mol. Biol. 2003, 49 (1)), 77– 87.

3722

Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

The methods for manufacture of these hard top-soft bottom devices described here are straightforward and are designed to minimize use of expensive or uncommon equipment. Although the applications described in this manuscript focused on cellular manipulation and analysis, the methods and devices should be useful for a broad range of applications where deformation-based microfluidic systems need to be combined with lower gas permeability and precise rigid channel structures. SUPPORTING INFORMATION AVAILABLE Further details are given in Table S1 and Figures S1-S3. This material is available free of charge via the Internet at http://pubs.acs.org. ACKNOWLEDGMENT We thank Brian Johnson and Prof. Mark Burns, Department of Chemical Engineering, Univ. of Michigan for use of clean room facilities. We also thank Dr. Nobuyuki Futai for his helpful suggestions throughout this work. This material is based upon work supported by the U.S. Army Research Laboratory and the U.S. Army Research Office under contract/grant number DAAD1903-1-0168 and the National Science Foundation (BES-0238625). Received for review October 14, 2008. Accepted March 31, 2009. AC802178U