Effect of Flow and Surface Conditions on Human Lymphocyte Isolation

Langmuir 2004, 20, 11649-11655

11649

Effect of Flow and Surface Conditions on Human Lymphocyte Isolation Using Microfluidic Chambers Shashi K. Murthy, Aaron Sin, Ronald G. Tompkins, and Mehmet Toner* Center for Engineering in Medicine and Surgical Services, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts 02114, and Shriners Hospital for Children, Boston, Massachusetts 02114 Received August 2, 2004. In Final Form: September 24, 2004 Phenotypically pure subpopulations of lymphocytes can provide valuable insights into the immune response to injury and disease. The isolation of these subpopulations presents unique challenges, particularly when preprocessing incubation to attach fluorescent or antibody tags is to be minimized. This paper examines the separation of T and B lymphocytes from mixtures using microfluidic chambers coated with antibodies, focusing on flow conditions and surface chemistry. The adhesion of both cell types decreases as shear stress increases irrespective of the surface chemistry. The incorporation of poly(ethylene glycol) chains along with the antibodies on the chamber surface is shown to significantly improve the reproducibility of cell adhesion and is thus an important part of the overall system design. Furthermore, this technique is shown to be an effective way of isolating highly pure subpopulations of lymphocytes from model mixtures, even when the target cell concentration is low.

Introduction The isolation of phenotypically pure subpopulations of human lymphocytes is important in both clinical diagnostics and basic research. These cells play a crucial role in the immune response to injury and disease, and useful diagnostic information can be obtained by studying pure subpopulations. For example, the functionality of T lymphocytes is important in analyzing the progression rate of HIV infection and transplant rejection;1 and antigen-specific B lymphocytes can provide useful insights into the onset and progression of thyroid2 and skin diseases.3 The use of antibody tags on cells or on surfaces is a common way to separate lymphocyte subpopulations. Techniques that achieve separation based on size and density are generally unable to provide adequate resolution between subpopulations of lymphocytes.4 In fluorescence-activated cell sorting (FACS), fluorescent dyes are attached to cells in mixtures either directly or using antibodies that bind to surface antigens. The cells are then sorted individually based on fluorescence and light scattering. This technique can provide highly pure (95% or higher) cell populations but requires expensive equipment and has a limited throughput (∼107 cells/h).5,6 Separation using antibody-coated magnetic beads is an alternative to FACS. Magnetic-activated cell sorting (MACS) allows target cells to be processed in parallel, allowing faster separation (∼1011 cells/h) of high-purity * Corresponding author. Address: Mehmet Toner, Ph.D., Shriners Hospital for Children, 51 Blossom St., Boston, MA 02114. Phone: (617) 371-4883. Fax: (617) 371-4950. E-mail: [email protected]. (1) Collins, D. P.; Luebering, B. J.; Shaut, D. M. Cytometry 1998, 33, 249-255. (2) Leyendeckers, H.; Voth, E.; Schicha, H.; Hunzelmann, N.; Banga, P.; Schmitz, J. Eur. J. Immunol. 2002, 32, 3126-3132. (3) Leyendeckers, H.; Tasanen, K.; Bruckner-Tuderman, L.; Zillikens, D.; Sitaru, C.; Schmitz, J.; Hunzelmann, N. J. Invest. Dermatol. 2003, 120, 372-378. (4) Bauer, J. J. Chromatogr. B 1999, 722, 55-69. (5) Thiel, A.; Scheffold, A.; Radbruch, A. Immunotechnology 1998, 4, 89-96. (6) Putnam, D. D.; Namasivayam, V.; Burns, M. A. Biotechnol. Bioeng. 2003, 81, 650-665.

populations.5 A major limitation of this method is interference caused by the magnetic beads in experiments performed on the purified cell population.4 A common element of existing tagging technologies is the need for preprocessing incubation to attach the various tags (fluorescent dyes, magnetic beads, etc.) onto cell surfaces. This step can be avoided by immobilizing antibodies on surfaces, which is the principle behind cellaffinity chromatography (CAC). CAC systems can provide satisfactory throughput (108-109cells/h) with yields and purity comparable to those of FACS and MACS.6,7 Most of these systems have a packed bed design, which maximizes surface area per unit volume but also results in long residence times (on the order of 1-2 h).6-8 The limitations of long residence times can be overcome by the use of microfluidic devices for CAC. These devices offer the advantage of high surface area to volume ratios with short residence times (order of minutes or less) per device and no need for preprocessing incubation. Microfluidic cell-separation systems described in the literature include devices that are based on flow around obstacles,9 diffusion,10,11 spectral impedance,12 fluorescence-activated sorting,13 and electrophoresis.14-17 Microfluidic CAC systems, which achieve separation using surface-immobilized antibodies, have not received much attention. (7) Mandrusov, E.; Houng, A.; Klein, E.; Leonard, E. F. Biotechnol. Prog. 1995, 11, 208-213. (8) Ujam, L. B.; Clemmitt, R. H.; Clarke, S. A.; Brooks, R. A.; Rushton, N.; Chase, H. A. Biotechnol. Bioeng. 2003, 83, 554-566. (9) Huang, L. R.; Cox, E. C.; Austin, R. H.; Sturm, J. C. Science 2004, 304, 987-990. (10) Cho, B. S.; Schuster, T. G.; Zhu, X. Y.; Chang, D.; Smith, G. D.; Takayama, S. Anal. Chem. 2003, 75, 1671-1675. (11) Suh, R. S.; Phadke, N.; Ohl, D. A.; Takayama, S.; Smith, G. D. Hum. Reprod. Update 2003, 9, 451-461. (12) Gawad, S.; Schild, L.; Renaud, P. Lab Chip 2001, 1, 76-82. (13) Fu, A. Y.; Spence, C.; Scherer, A.; Arnold, F. H.; Quake, S. R. Nat. Biotechnol. 1999, 17, 1109-1111. (14) Huang, Y.; Yang, J. M.; Hopkins, P. J.; Kassegne, S.; Tirado, M.; Forster, A. H.; Reese, H. Biomed. Microdevices 2003, 5, 217-225. (15) Yang, J.; Huang, Y.; Wang, X. B.; Becker, F. F.; Gascoyne, P. R. C. Anal. Chem. 1999, 71, 911-918. (16) Fu, A. Y.; Chou, H. P.; Spence, C.; Arnold, F. H.; Quake, S. R. Anal. Chem. 2002, 74, 2451-2457. (17) Voldman, J.; Gray, M. L.; Toner, M.; Schmidt, M. A. Anal. Chem. 2002, 74, 3984-3990.

10.1021/la048047b CCC: $27.50 © 2004 American Chemical Society Published on Web 11/16/2004

11650

Langmuir, Vol. 20, No. 26, 2004

Figure 1. Flow channel geometry and shear stress profile. The relationship between shear stress (τw) and axial position (z) is given by the inset equation, where Q is the volumetric flow rate, µ the viscosity of water, h the chamber height (57 ( 1 µm), w1 the inlet width (5 mm), and L the chamber length (50 mm). The linear shear stress design is based on equations derived by Usami et al. (ref 18).

To design an effective microfluidic CAC system, it is essential to understand (1) the properties of the surfaces involved in cell isolation and (2) the effect of flow on the attachment of cells to the surfaces. This paper examines the isolation of T and B lymphocytes using antibody-coated microfluidic chambers and highlights the role of surface chemistry and flow conditions. Antibodies are attached onto the chamber surfaces using silane chemistry and avidin-biotin binding. The chamber geometry is such that the shear stress along the axis varies linearly along the chamber length (Figure 1), allowing analysis of cell adhesion over a range of shear stresses in a single experiment. The incorporation of poly(ethylene glycol) (PEG) chains along with the antibodies on the chamber surface is shown to significantly improve the reproducibility of cell adhesion and is thus an important part of the overall system design. This system is capable of isolating high-purity subpopulations of one cell type from mixed populations of T and B lymphocytes. Experimental Section Materials. 3-Mercaptopropyl trimethoxysilane was obtained from Gelest (Morrisville, PA). Ethanol (200 proof), glass slides (35 × 60 mm, no. 1), tissue culture flasks, a hemacytometer, serological pipets, and microslide field finders were purchased from Fisher Scientific (Fair Lawn, NJ). For chamber fabrication, SU-8 photoresist and developer were obtained from MicroChem (Newton, MA); silicone elastomer and curing agent were obtained from Dow Corning (Midland, MI). Phosphate buffered saline (PBS) 1× and RPMI-1640 cell culture medium were obtained from Mediatech (Herndon, VA). Fetal bovine serum (FBS) and 0.5 M ethylenediaminetetraacetic acid (EDTA) were purchased from Gibco (Grand Island, NY). Dimethyl sulfoxide (DMSO), sodium azide, lyophilized bovine serum albumin (BSA), and a glovebag for handling the moisture-sensitive silane were obtained from Aldrich Chemical Co. (Milwaukee, WI). The coupling agent GMBS (N-y-maleimidobutyryloxy succinimide ester), 3500 molecular weight cutoff dialysis cassettes, NHS-LC-LC-biotin (succinimidyl-6′-[biotinamido]-6-hexanamido hexanoate), and fluorescein-conjugated NeutrAvidin were obtained from Pierce

Murthy et al. Biotechnology (Rockford, IL). NH2-terminated PEG (PEG-NH2) of 5000 and 20 000 molecular weight was obtained from Nektar Therapeutics (Huntsville, AL). Biotinylated mouse anti-human anti-CD5 and anti-CD19 were purchased from Serotec Antibodies (Raleigh, NC). Human mature naı¨ve B-lymphoblast (Raji) and immature T-lymphoblast (Molt-3) cell lines were purchased from American Type Culture Collection (Manassas, VA). Orange [5(and 6-)-(((4-chloromethyl)-benzoyl)amino)tetramethyl-rhodamine, CMTMR] and green [5-chloromethylfluorescein diacetate, CMFDA] cell tracker dyes were obtained from Molecular Probes (Eugene, OR). Solution Preparation. A 4% (v/v) solution of 3-mercaptopropyl trimethoxysilane in ethanol (solution 1) was prepared under a nitrogen atmosphere inside a glovebag. Each 50 mg batch of GMBS was dissolved in 0.5 mL of DMSO (solution 2). The lyophilized NeutrAvidin was restored with distilled water as recommended by the manufacturer (solution 3). Each 50 mg batch of NHS-LC-LC-biotin was dissolved in 1 mL of DMSO (solution 4). Chamber Design. The geometry of the chambers is based on the theory of Hele-Shaw flow between parallel plates and design equations derived by Usami et al.18 The shape of the chambers is such that the shear stress along the axis of the chamber decreases linearly along the chamber length, as shown in Figure 1. Chamber Fabrication. Microfluidic flow chambers were assembled at the BioMEMS Resource Center at the Center for Engineering in Medicine (Massachusetts General Hospital). A high-resolution transparency (CAD/Art Services, Inc., Poway CA) was generated from an Auto-CAD file created in-house. This transparency was used to generate a negative master. Briefly, a silicon wafer was coated with SU-8 to a thickness of approximately 57 µm. With the transparency overlaid, the wafer was then exposed to 365 nm, 11 mW/cm2 UV light from a Q2001 mask aligner (Quintel Co., San Jose, CA). Unexposed SU-8 was then removed using developer. Feature height was verified using a Dektak surface profiler (Veeco Instruments, Santa Barbara, CA). Silicone elastomer and curing agent were mixed (10/1 ratio) and poured on top of the wafers and allowed to cure overnight in an oven at 65 °C. Holes for the inlet and outlet ports on the poly(dimethylsiloxane) (PDMS) replicas were punched out with a blunt-nosed needle. PDMS replicas and glass slides were cleaned with an oxygen plasma (100 mW, 1% oxygen, 30 s) in a PX-250 plasma chamber (March Instruments, Concord, MA) and then immediately placed in contact to bond the surfaces irreversibly. Chambers were baked at 65 °C for 2 min following bonding. Surface modification was carried out immediately after the bake step to allow covalent bond formation between the glass surface and silane molecules. Surface Modification. The major steps of the surface modification procedure are: (a) surface pretreatment with silane, (b) attachment of the coupling agent (GMBS) to the silane, (c) attachment of avidin to GMBS, and finally (d) attachment of biotinylated antibody, biotinylated PEG, or PEG-antibody mixture. This procedure was carried out as follows: Chambers were flushed with solution 1 and allowed to react at room temperature for 30 min. Unreacted silane was removed by flushing with ethanol. The chambers were then flushed with a 0.28% (v/v) mixture of solution 2 and ethanol and allowed to react for 15 min. After flushing with ethanol, the chambers were flushed with a 0.1% (v/v) mixture of solution 3 and PBS and stored overnight in a refrigerator. The chambers were flushed with PBS and used in flow experiments or modified further as follows. For antibody coating, the anti-CD5 and anti-CD19 stocks were diluted 1:10 by volume with PBS containing 1% (w/v) sodium azide and 0.09% (w/v) BSA and flowed through the chambers at room temperature. After a 15 min period, the chambers were flushed with PBS to remove unattached antibody. For coating chamber surfaces with PEG, the PEG-NH2 was dissolved in PBS and biotinylated by mixing with solution 4 such that the molar ratio of PEG-NH2 to NHS-LC-LC-biotin was 1:1. After incubating for 2 h at 4 °C, unreacted biotin was removed by dialyzing overnight against PBS inside a refrigerator (4 °C). The (18) Usami, S.; Chen, H. H.; Zhao, Y. H.; Chien, S.; Skalak, R. Ann. Biomed. Eng. 1993, 21, 77-83.

Human Lymphocyte Isolation biotinylated PEG solution was then flowed through the chambers at room temperature, and untethered PEG was removed by flushing with PBS after 15 min. For coating chambers with PEG/ antibody mixtures, the biotinylated PEG was mixed in a 1/1 molar ratio (PEG/antibody) for anti-CD19 and 1/100 for anti-CD5. These mixtures were flowed through the chambers, and unreacted PEG and antibody were removed by flushing with PBS after 15 min. Flow Experiments. Raji and Molt-3 cells were cultured in 150 cm2 tissue culture flasks at 37 °C in a humidified atmosphere with 5% CO2/95% air. The cells were incubated in RPMI-1640 supplemented with 10% fetal bovine serum, 200 U/mL penicillin. The cell suspension was centrifuged at 150g for 5 min and then resuspended in PBS to remove dead cells and cell debris. After centrifuging again at 150g for 5 min, the cells were resuspended in a 1 mM EDTA solution in PBS to obtain a concentration of approximately 106 cells/mL (measured using a hemacytometer). For experiments with Raji/Molt-3 model mixtures, the cells were stained with orange and green cell tracker dyes, respectively. For this step, the cell suspension drawn from the culture flask was centrifuged at 150g for 5 min, resuspended in 10 mL of phenol red-free RPMI-1640 containing 4 mM of cell tracker dye and incubated for 1 h at 37 °C. Centrifugation and resuspension in PBS/EDTA were performed as described above. For flow experiments, the cell suspension was flowed into the chambers at 10 µL/min for 15 min using a Harvard Apparatus PHD 2000 syringe pump (Holliston, MA). PBS was then flowed through the chamber at 125 µL/min for 5 min. Cell adhesion was measured by placing a field finder under the chambers and counting cells at select points along the device axis using a Nikon Eclipse TE2000 inverted microscope (Nikon, Japan). For each point, three measurements were made, corresponding to three 1 mm2 squares in that vicinity, and averaged. Cell mixtures were examined and photographed using a SPOT digital camera (Diagnostic Instruments, Inc., Burlingame, CA) attached to the Nikon microscope. Images were obtained at 10× magnification using fluorescein (480 ( 30 nm/535 ( 40 nm) and rhodamine (540 ( 25 nm/605 ( 50 nm) excitation/emission filters. All flow experiments were performed at room temperature. Atomic Force Microscopy (AFM). Images were obtained using a Nanoscope III instrument (Digital Instruments, Santa Barbara, CA) operating in tapping mode, with a standard 117 µm silicon cantilever. Glass slides were modified as described above and, following surface modification, washed in distilled water and dried in air prior to AFM analysis. Statistics and Data Analysis. In each cell adhesion experiment, cell counts were obtained for seven different shear stresses. Each experiment was repeated 6 times. The data shown in Figures 2 and 4 represent cell counts averaged over 6 repetitions, and each error bar represents the standard error of the mean (standard deviation/xn, where n ) 6). Error bars for the data in Figure 5 also represent standard errors with n ) 6.

Results and Discussion Linear shear stress chambers of the type shown in Figure 1 offer the ability to examine cell adhesion over a range of shear stresses in a single experiment. These chambers can therefore be used to identify optimal surface and flow conditions prior to the design of a real system that operates at a single shear stress. The methodology used in this study consists of the following steps: surface modification of the microfluidic chambers to attach avidin onto the chamber surfaces, further modification of the surfaces with biotinylated antibodies and/or biotinylated PEG, and evaluation of cell adhesion on these surfaces. In designing a fluidic system that captures target cells from a mixture of several cell types using immobilized antibodies, it is important to ensure that adhesion of nontarget cells is minimized. It is also necessary to minimize artifacts that may result from cell-cell interactions. The attachment of PEG on the microfluidic chamber surfaces was considered as a way to meet these requirements, based on its well-known biological nonadhesiveness.19-24 A series (19) Lee, S. W.; Laibinis, P. E. Biomaterials 1998, 19, 1669-1675.

Langmuir, Vol. 20, No. 26, 2004 11651

Figure 2. Adhesion of Raji cells to surfaces coated with avidin and PEG. Error bars denote standard errors for each point based on six repetitions.

of experiments (Figure 2) was performed to determine which molecular weight would provide the greatest reduction in nonspecific adhesion. Two different molecular weights, 5000 and 20 000 g/mol, were examined and compared on chamber surfaces coated with avidin. The results shown in Figure 2 are for Raji cells; Molt-3 cells show similar adhesion trends. The data in Figure 2 indicate that using PEG results in significant reduction in nonspecific adhesion, with higher molecular weight providing better results. The biological nonadhesiveness of PEG can be explained in terms of steric stabilization and solution properties of PEG in water.23 The steric stabilization effect has two components: an elastic term and an osmotic term.23,25 When a cell or protein approaches the PEG-grafted surface, the total number of conformations available to the PEG chains is reduced, resulting in a loss in configurational entropy. This loss is reflected in the elastic term. The osmotic term describes the repulsive force generated by the compression or interpenetration (or both) of the PEG chains by the approaching protein or cell. The unique behavior of PEG, however, cannot be explained by steric stabilization alone. The second mechanism for nonadhesiveness is related to the solubility of PEG in water. Due to the structural similarity of the PEG repeat units and water molecules, chains of PEG are easily accommodated in water lattices without steric hindrance. PEG chains in aqueous solutions are therefore hydrated and highly mobile and will tend to prevent any approaching cells or proteins from contacting the underlying surface. All of (20) Malmsten, M.; Emoto, K.; Van Alstine, J. M. J. Colloid Interface Sci. 1998, 202, 507-517. (21) Prime, K. L.; Whitesides, G. M. J. Am. Chem. Soc. 1993, 115, 10714-10721. (22) Prime, K. L.; Whitesides, G. M. Science 1991, 252, 1164-1167. (23) Lee, J. H.; Lee, H. B.; Andrade, J. D. Prog. Polym. Sci. 1995, 20, 1043-1079. (24) Sofia, S. J.; Merill, E.; Harris, J. M. In Poly(Ethylene Glycol): Chemistry and Biological Applications; Zalipsky, S., Ed.; American Chemical Society: Washington, DC, 1997; Vol. 680, pp 342-360. (25) Holmberg, K.; Bergstro¨m, K.; Stark, M.-B. In Poly(Ethylene Glycol) Chemistry: Biotechnical and Biomedical Applications; Harris, J. M., Ed.; Plenum Press: New York, 1992; p 303.

11652

Langmuir, Vol. 20, No. 26, 2004

Murthy et al.

Figure 3. AFM images of glass slides coated with avidin (a,d), anti-CD5 (b), anti-CD5 and PEG (c), anti-CD19 (e), and anti-CD19 and PEG (f). For all micrographs, the image area is 1 µm × 1 µm.

the above effects, namely, configurational entropy loss, osmotic repulsion, and chain mobility, are enhanced by increasing chain length. Longer PEG chains would thus be expected to show a greater resistance to protein adsorption and cell adhesion, consistent with the results shown in Figure 2. Based on these results, the 20 000 molecular weight PEG was selected for the cell-capture experiments using microfluidic chambers. The attachment of biotinylated antibodies to avidincoated surfaces was verified using AFM. Figure 3 shows AFM images of glass surfaces coated with avidin, antibody, and a combination of PEG and antibody. These samples were prepared following the same surface modification procedure used for the flow chambers and then washed in distilled water and dried in air prior to AFM analysis. The large features present in the avidin-coated surfaces remain after further surface modification. These features are approximately 10 nm in height and 50 nm in diameter. It is hypothesized that these are clusters of avidin molecules or small salt crystals. The smallest features observed in Figure 3b,e are 4-10 nm in height and approximately 18-25 nm in diameter. These dimensions are consistent with estimates from the literature of the height (7 nm) and diameter (∼10 nm) of IgG antibodies immobilized on surfaces as measured by AFM and X-ray experiments, respectively.26,27 It is therefore postulated that these features are single, immobilized anti-CD5 or anti-CD19 molecules. The approximate antibody surface density, as observed in Figure 3b,e, is 60 molecules/µm2. Figure 3c,f shows glass surfaces coated with a combination of antibody and 20 000 molecular weight PEG. In air, the long PEG chains would tend to be in a tightly coiled or overlapping conformation, resulting in the relatively smooth surfaces (the root-mean-square (rms) roughness is 1.6 nm in Figure 3c and 0.9 nm in Figure 3f). In these images, it is not possible to distinguish between the PEG (26) Silverton, E. W.; Navia, M. A.; Davies, D. R. Proc. Natl. Acad. Sci. U.S.A. 1977, 74, 5140-5144. (27) Dong, Y. Z.; Shannon, C. Anal. Chem. 2000, 72, 2371-2376.

and antibody molecules. While Figure 3b,e confirms the attachment of biotinylated antibody molecules to avidincoated glass surfaces, it also indicates that the antibodies do not cover the surfaces entirely (i.e., there are regions of exposed avidin). Adding PEG (Figure 3c,f) along with the antibody results in almost complete surface coverage. The results of flow experiments using chambers coated with antibodies and PEG as well as chambers coated with only antibodies are shown in Figure 4a,b and Figure 4c,d, respectively. Each data set on these graphs represents a flow experiment with a single cell type; no mixtures were used. The data indicate preferential adhesion of the target cell type, namely, the CD5+ Molt-3 cells in Figure 4a,c and the CD19+ Raji cells in Figure 4b,d. The adhesion of these target cells decreases with increasing shear stress, as expected. However, in the absence of PEG (Figure 4c,d), there is significant scatter in almost all of the data points. This behavior could be the result of inadequate coverage of the surface by antibodies, leaving regions of exposed avidin (as seen in Figure 3b,e). These regions would be susceptible to nonspecific cell adhesion, as shown in Figure 2. The scatter in the data in Figure 4c,d could then be explained by the fact that the location and size of these regions would vary between experiments. When biotinylated-PEG chains are attached to the surface along with the biotinylated antibodies, there would be fewer and smaller regions of avidin. Furthermore, the long PEG chains will screen the cells from the underlying surface, thereby reducing significantly the variation between experiments that produced large error bars in Figure 4c,d. In the presence of PEG, the only mechanism for cell adhesion to the surface is by specific binding with the antibodies. The above mechanism, however, fails to explain why the total adhesion of the target cell type (Molt-3 in Figure 4c and Raji in Figure 4d) is lower when PEG is absent compared to the case when PEG is coimmobilized along with the antibody (Figure 4a,b). This difference could be due to effects of PEG on the orientation of antibody

Human Lymphocyte Isolation

Langmuir, Vol. 20, No. 26, 2004 11653

Figure 4. Variation of cell attachment with shear stress in chambers coated with antibodies and PEG (a,b) and chambers coated with antibodies only (c,d). Each data set represents a distinct experiment perfomed with only one cell type. Error bars denote standard errors for each point based on six repetitions.

molecules on the chamber surface. Due to its proteinrepulsive character, it is likely that the presence of PEG would limit the orientation of antibody molecules to conformations that produce minimal repulsion. If these conformations are such that they facilitate optimal bond formation between the antibodies and cell surface antigens, the result would be an overall increase in the number of target cells adhered in the presence of PEG. The adhesion behavior shown in Figure 4 can be used as a basis for separation of Raji and Molt-3 model mixtures. Mixtures containing approximately equal numbers of Raji and Molt-3 cells (the total concentration of the mixture was 106 cells/mL) were introduced into the linear shear stress chambers and then subjected to shear flow. Figure 5 shows fluorescence micrographs of cells remaining after the shear flow in chambers coated with PEG (20 000 molecular weight) and antibody (anti-CD5 in Figure 5a-c and anti-CD19 in Figure 5d-f). The images were obtained at different positions along the chamber axis (corresponding to high, medium, and low shear stress), and the green and red colors correspond to Molt-3 and Raji cells, respectively. In Figure 5a-c, the CD5-positive Molt-3 cells show increasing adhesion with decreasing shear stress, consistent with Figure 4a. No Raji cells are observed except

for approximately 5 cells in Figure 5c out of a total of about 290 cells. A similar trend is observed for the CD19positive Raji cells in Figure 5d-f. No Molt-3 cells are observed except for 1 cell in Figure 5f out of a total of about 200 cells. The adhesion of small numbers of nontarget cells is consistent with the nonspecific binding in Figure 4a,b. Similar results were obtained with Raji/Molt-3 model mixtures of significantly unequal concentrations: 90% Raji + 10% Molt-3 and 90% Molt-3 + 10% Raji. At high and medium shear stress, the only cells adhered to the antibody-coated chamber surface were the target cells (i.e., only Molt-3 cells on anti-CD5 and only Raji cells on antiCD19). Nontarget cells were observed at the low shear stress positions, but their number was small (3% or less) relative to that of the target cells. A comparison of the captured cell compositions at the lowest shear stress position for the different cell mixtures is shown in Figure 6. The results shown in Figures 5 and 6 images indicate the potential of antibody-coated microfluidic chambers to isolate highly pure subpopulations of lymphocytes from model mixtures. The purity of the captured subpopulation remains high (>97%, Figure 6) even when the concentra-

11654

Langmuir, Vol. 20, No. 26, 2004

Murthy et al.

Figure 5. Cell adhesion as a function of shear stress in anti-CD5-coated devices, photographed at (a) 10.9, (b) 9.8, and (c) 5.2 dyn/cm2; and in anti-CD19-coated devices, photographed at shear stress values of (d) 13.9, (e) 12.3, and (f) 9.2 dyn/cm2. Green and red colors correspond to Molt-3 and Raji cells, respectively. The area covered in each image is 1 mm × 1 mm.

the former; T cells relative to B cells in the latter) is drastically reduced. While the present system matches FACS and MACS in terms of purity of captured cells, it has not been designed for high-throughput operation. The tools developed herein, however, can be used to design a high-throughput microfluidic system. This can be done by identifying an optimal shear stress level and determining how much surface area would be required to captured the desired number of cells. Conclusions

Figure 6. Comparison of initial cell mixture composition with the composition of cells remaining in the microfluidic chamber after flushing, for chambers coated with anti-CD5 (a,b) and anti-CD19 (c,d). Measurements were made at the lowest shear stress position (5.2 dyn/cm2 for anti-CD5 and 9.2 dyn/cm2 for anti-CD19).

tion of the target cells in the starting mixture is low. This type of system could potentially be employed to isolate pure subpopulations of T and B lymphocytes from whole blood to study immune system disorders. The ability to isolate target cells present in low concentrations could be relevant for diseases such as AIDS and DiGeorge syndrome,28 where the proportion of one lymphocyte type relative to another (CD4+ cells relative to CD8+ cells in (28) Kuby, J. Immunology; W. H. Freeman: New York, 1997; pp 511-515.

This paper demonstrates the role of surface and flow conditions in the isolation of T and B lymphocytes from model mixtures using microfluidic chambers. The use of linear shear stress chambers allows the analysis of cell adhesion over a range of shear stresses in a single experiment. The attachment of high-molecular-weight PEG to the surface is shown to improve adhesion characteristics significantly and is thus an important element of the overall system design. Separation of highly pure populations of T and B lymphocytes from mixtures is achieved over a short time period and without preprocessing incubation. High-purity subpopulations can also be obtained from model mixtures where the concentration of the target cell type is low. The methodology described in this work, along with the surface and flow conditions employed, could potentially be used in the isolation of lymphocyte subpopulations as well as other cell types (such as dendritic cells) from whole blood. Acknowledgment. We gratefully acknowledge all the investigators of the Inflammation and Host Response to Injury Large Scale Collaborative Project for helpful discussions. We thank Dr. Alexander Revzin for help in

Human Lymphocyte Isolation

developing the surface modification protocol and Mr. Octavio Hurtado for technical support with the microfabrication procedures. This work was partially supported by the National Institute of General Medical Sciences under Grant No. U54 GM062119 (Inflammation and Host Response to Injury Large Scale Collaborative Project) and the National Institute of Biomedical Imaging and Bioengi-

Langmuir, Vol. 20, No. 26, 2004 11655

neering under Grant No. P41 EB002503 (BioMEMS Resource Center). This work made use of MRSEC Shared Facilities (Center for Materials Science and Engineering, MIT) supported by the National Science Foundation under Award Number DMR-0213282. LA048047B