Deterministic Lateral Displacement as a Means to Enrich Large Cells

Anal. Chem. 2009, 81, 9178–9182

Deterministic Lateral Displacement as a Means to Enrich Large Cells for Tissue Engineering James V. Green,† Milica Radisic,‡,§ and Shashi K. Murthy*,† Department of Chemical Engineering, Northeastern University, Boston, Massachusetts 02115, Institute of Biomaterials and Biomedical Engineering and Department of Chemical Engineering and Applied Chemistry, University of Toronto, Toronto, Ontario M5S 3G9, Canada The enrichment or isolation of selected cell types from heterogeneous suspensions is required in the area of tissue engineering. State of the art techniques utilized for this separation include preplating and sieve-based approaches that have limited ranges of purity and variable yield. Here, we present a deterministic lateral displacement (DLD) microfluidic device that is capable of separating large epithelial cells (17.3 ( 2.7 in diameter) from smaller fibroblast cells (13.7 ( 3.0 µm in diameter) as a potential alternative approach. The mixed suspension examined is intended to represent the content of digested rat cardiac tissue, which contains equal proportions of cardiomyocyte (17.0 ( 4.0 µm diameter) and nonmyocyte populations (12.0 ( 3.0 µm diameter). High purity separation (>97%) of the larger cell type is achieved with 90% yield in a rapid and single-pass process. The significance of this work lies in the recognition that DLD design principles can be applied for the microfluidic enrichment of large cells, up to the 40 µm diameter level examined in this work. The separation or enrichment of one of more cell types from a heterogeneous suspension is a critical initial step in many biological experiments and clinical procedures. This is especially true in the field of tissue engineering, as specific populations of cells are required to construct the desired tissue grafts. In cardiac tissue engineering, a cardiac patch constructed from cardiomyocytes harvested from the native myocardium may offer an alternative treatment for myocardial infarction and congestive heart failure. Recently, a tissue-engineered cardiac graft constructed from cardiomyocytes seeded within an alginate scaffold has been shown to prevent deterioration in cardiac function following myocardial infarction in rats.1 Furthermore, cultured cardiomyocyte monolayers have been shown to form threedimensional tissue and are beat stacked on one another.2,3 The state of the art method of cardiomyocyte enrichment involves starting with a heterogeneous suspension of digested * To whom correspondence should be addressed. E-mail: [email protected]. † Department of Chemical Engineering, Northeastern University. ‡ Institute of Biomaterials and Biomedical Engineering, University of Toronto. § Department of Chemical Engineering and Applied Chemistry, University of Toronto. (1) Dar, A.; Shachar, M.; Leor, J.; Cohen, S. Biotechnol. Bioeng. 2002, 80, 305–312. (2) Eschenhagen, T.; Zimmermann, W. H. Circ. Res. 2005, 97, 1220–1231.

9178

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

donor tissue followed by sequential preplating in culture flasks to allow the more adhesive nonmyocyte cells to be depleted from the supernatant.3 For digested rat cardiac tissue, which contains approximately equal numbers of cardiomyocytes and fibroblasts, one cycle of preplating over a period of 1 h results in a suspension that contains 64 ± 5% cardiomyocytes and 34 ± 16% fibroblasts while two cycles of preplating result in a suspension of 81 ± 14% cardiomyocytes and 16 ± 3% fibroblasts.4 Other than carrying out additional cycles, there are no parameters in the preplating enrichment process that can be systematically varied to obtain more efficient and fast separation. The ability to obtain high-purity cardiomyocyte fractions rapidly through preplating is particularly acute when working with limited samples of tissue and recognizing that these primary cells have limited viability in vitro. Cardiomyocyte isolation from heterogeneous suspensions has also been performed by density gradient centrifugation,5 but this technique cannot be utilized for small sample volumes and requires considerable operator skill in being able to extract multiple cell layers following centrifugation. Cardiomyocytes are the largest cells in the myocardium with an average size of 17.0 ± 4.0 µm. The average size of nonmyocytes is 12.0 ± 3.0 µm. In prior work,6 we attempted size-based enrichment of cardiomyocytes using a sieve-based microfluidic device but encountered extensive cell clogging. This paper describes an alternative approach based on the well-known principle of deterministic lateral displacement (DLD).7-9 The major advantage of this approach is that it overcomes the inherent tendency of any sieve-based designs to clog due to the ability of cells in flow to deform. DLD is a separation method in which an array of obstacles is created in a flow channel such that particles below a critical hydrodynamic diameter Dc flow along a straight path through (3) Radisic, M.; Park, H.; Shing, H.; Consi, T.; Schoen, F. J.; Langer, R.; Freed, L. E.; Vunjak-Novakovic, G. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 18129– 18134. (4) Brown, M. A.; Iyer, R. K.; Radisic, M. Biotechnol. Prog. 2008, 24, 907– 920. (5) Ling-Ling, E.; Zhao, Y. S.; Guo, X. M.; Wang, C. Y.; Jiang, H.; Li, J.; Duan, C. M.; Song, Y. J. Heart Lung Transplant. 2006, 25, 664–674. (6) Murthy, S. K.; Sethu, P.; Vunjak-Novakovic, G.; Toner, M.; Radisic, M. Biomed. Microdevices 2006, 8, 231–237. (7) Huang, L. R.; Cox, E. C.; Austin, R. H.; Sturm, J. C. Science 2004, 304, 987–990. (8) Davis, J. A.; Inglis, D. W.; Morton, K. J.; Lawrence, D. A.; Huang, L. R.; Chou, S. Y.; Sturm, J. C.; Austin, R. H. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 14779–14784. (9) Inglis, D. W.; Davis, J. A.; Austin, R. H.; Sturm, J. C. Lab Chip 2006, 6, 655–658. 10.1021/ac9018395 CCC: $40.75  2009 American Chemical Society Published on Web 10/07/2009

Figure 1. Photograph (a) and post array layout details (b) for the deterministic lateral displacement device. In (a), the cell suspension inlet is focused by the sheath fluid and separated on the basis of size by the post array (50 µm posts). The largest cells are expected to exit at outlet 6 and the smallest at outlet 2. In (b), gap spacing between posts is g, the row shift fraction is ε, and the center-to-center post spacing is λ.

the channel whereas particles with diameters larger than Dc are horizontally displaced as a function of size.7,8 This technique has been used to successfully separate submicrometer polystyrene beads with high purity and resolution as well as to separate whole blood.7,8 Furthermore, design criteria for this technology were recently defined by Inglis et al.9 enabling the creation of a sizebased separation device to suit a wide variety of applications. This paper describes a DLD device made of poly(dimethylsiloxane) (PDMS) that separates a model heterogeneous suspension of cells designed to simulate the content of digested cardiac tissue. The model system is composed of two immortalized cell lines: the H1975 epithelial cell line and the 3T3 fibroblast cell line, which are intended to represent the cardiomyocyte and nonmyocyte populations, respectively. The epithelial cell line has a size range of 10-40 µm in diameter, and these cells are first fractionated using the device presented herein to obtain a cell suspension with an average cell diameter of 17.3 ± 2.7 µm (measured using a Coulter Counter). The fibroblast cell line has an average diameter of 13.7 ± 3.0 µm. The fractionated H1975 cells and the 3T3 cells are mixed together in equal proportions to create a 50:50 mixture that is equivalent to the 50:50 cardiomyocyte/nonmyocyte suspension obtained by digesting rat cardiac tissue.4 EXPERIMENTAL SECTION Device Design and Fabrication. An array of cylindrical obstacles for cell separation by DLD was created as shown in Figure 1. For the present work, the critical hydrodynamic diameter Dc was selected to be 15 µm. The next parameter to determine is the gap spacing, g. This parameter needs to be large enough to allow the largest cell to pass through the post array without clogging; however, it should not be too large, as purity may be negatively affected. Taking this into consideration, the gap spacing was selected to be 37.5 µm as the largest cells in the present work have diameters e40 µm. (The elastic properties of the cells will allow the largest cells to pass through the post array without clogging.) The remaining post layout design parameters were determined using the design equation formulated by Inglis et al.,9 which relates the critical hydrodynamic diameter to the gap spacing, parabolic flow profile correction factor, and the row shift fraction: Dc ) 2ηgε Here, Dc represents the critical hydrodynamic diameter, η is the parabolic flow profile constant, g is the gap spacing, and ε is the row shift fraction. In the equation shown, η and ε are related by a theoretical relationship developed by Inglis et al. (Figure 3 in ref 9). ε, in turn, is related to the ratio of particle diameter to gap spacing for various modes of flow, and this

relationship is plotted for various conditions by Inglis et al. (Figure 2 in ref 9). When the values of both Dc and the gap spacing are specified, the value of ε can be obtained from the parabolic flow curve in Figure 2 in ref 9. The value of η is then obtained from the parabolic flow curve in Figure 3 in ref 9. (Note that Figures 2 and 3 in ref 9 are set up to be consistent with the above design equation.) The calculated values of η and ε are 2 and 0.1, respectively. The size of the posts were then chosen to be 50 µm; this resulted in a gap spacing to post ratio of 0.75 which is within the range suggested by Inglis et al.9 The device, shown in Figure 1a, was also designed to have 6 outlets so the cell suspension could be fractionated as a function of size to the desired purity. The largest cells are expected to be displaced laterally to output 6, and the smallest cells are expected to travel straight through the device and exit at output 2. No cells are expected to travel through output 1, but output 1 is necessary in order to maintain uniform fluid flow throughout the device.8 The design and fabrication of the microfluidic devices followed previously described soft lithography techniques.10,11 Surface Treatment. Surface treatment was performed on each device by the use of solutions of ethanol (Fisher), phosphate buffered saline (PBS), and 1% bovine serum albumin (BSA). In the first step, the devices were flushed with ethanol to fill the flow channel completely without trapping air bubbles. The device was then rinsed with PBS followed by an injection of 1% BSA. The 1% BSA was allowed to adsorb in the device for 30 min before rinsing with PBS. The devices were then directly used in experiments or stored at 4 °C. Cell Culture. 3T3 fibroblasts were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 4.5 g/L glucose and L-glutamine (Mediatech), 10% fetal bovine serum (FBS; Fisher), 100 U/mL penicillin, and 100 µg/mL streptomycin (Hyclone). H1975 epithelial cells (ATCC) were cultured in RPMI 1640 with L-glutamine (Mediatech) supplemented with 10% FBS, 100 U/mL penicillin, and 100 µg/mL streptomycin. The cells were incubated in 75 cm2 tissue culture flasks at 37 °C in a humidified atmosphere with 5% CO2 and 95% air. Cells were grown to preconfluence and isolated for experiments by trypsinization using a 0.25% Trypsin-EDTA solution (Hyclone). For all experiments, cell suspensions were centrifuged at 190g and then resuspended in PBS to the desired concentration (measured using a hemacytometer). Flow Experiments. For H1975 epithelial cell fractionation experiments, the cell suspension was first run through a 40 µm nylon mesh sterile cell strainer (Fisher) and then brought to a (10) Plouffe, B. D.; Njoka, D.; Harris, J.; Liao, J.; Horick, N. K.; Radisic, M.; Murthy, S. K. Langmuir 2007, 23, 5050–5055. (11) Xia, Y. N.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37, 551–575.

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

9179

concentration of 5 × 105 cells/mL. This suspension (0.6 mL) was then loaded into a syringe and connected to the device inlet using tygon tubing. The flow of cells into the device was controlled by a Harvard Apparatus PHD 2000 syringe pump mounted such that the orientation of the syringe was vertical.12 To ensure that 100% of the cell suspension flowed through the syringe and no cell loss occurred due to cell settling within the syringe, a steel ball bearing was placed inside the syringe and moved up and down by a motorized external magnet throughout the course of the flow.13,14 The cell suspension was flowed through the device at a flow rate of 200 µL/min. The sheath fluid (PBS) was flowed in through the two side inlets at a flow rate of 500 µL/min through each inlet. Samples were collected from each of the six outlets and centrifuged at 190g. The output suspensions were then resuspended in PBS, and cell concentrations were measured with a hemacytometer. Fractions 4 and 5 were then pooled and measured for size using a Beckman Z2 Coulter Counter. This suspension was utilized for the heterogeneous suspension experiments described below. For the heterogeneous suspension experiments, the H1975 cell suspension obtained by the fractionation process described above was mixed with 3T3 fibroblasts to create a mixed suspension containing equal concentrations of both cell types (50:50 mixture). Prior to this addition, the 3T3 fibroblasts were labeled with a green cell tracker dye, as described previously.10 The 50:50 mixture was then diluted to a volume of 0.6 mL using PBS and loaded into a syringe at a concentration of approximately 5 × 105 cells/mL and flowed through the device at a flow rate of 200 µL/min with PBS sheath fluid at 500 µL/min through each inlet. The concentration of cells obtained from each outlet was measured using a hemacytometer in conjuction with fluorescence microscopy (to discern the unlabeled H1975 cells from the fluorescently labeled 3T3 cells). Statistics and Data Analysis. The cell count measurements reported represent average values over five repetitions, and the error represents the standard errors of the mean (standard deviation/(n)1/2 where n ) 5). RESULTS AND DISCUSSION Surface treatment with BSA was carried out in order to prevent cell clogging between posts or cell adhesion to the posts. Previous DLD devices have been made by silicon etching,7-9 and in these instances, clogging occurred only due to cells or particles being larger than the gap spacing between posts. This type of clogging can be easily avoided by creating a gap spacing larger than the largest particle or cell being run through the device. DLD devices made of PDMS, however, have been reported to clog severely due to cell adhesion to the posts.15-17 This problem was addressed in the present work by BSA blocking. The difference between a PDMS device treated with BSA and one that is not can be seen (12) Plouffe, B. D.; Radisic, M.; Murthy, S. K. Lab Chip 2008, 8, 462–472. (13) Cooper, R.; Lee, L. P. Lab Chip 2007, online “Chips & Tips” article. (14) Green, J. V.; Kniazeva, T.; Abedi, M.; Sokhey, D. S.; Taslim, M. E.; Murthy, S. K. Lab Chip 2009, 9, 677–685. (15) Chang, S.; Cho, Y. H. Lab Chip 2008, 8, 1930–1936. (16) Inglis, D. W.; Davis, J. A.; Zieziulewicz, T. J.; Lawrence, D. A.; Austin, R. H.; Sturm, J. C. J. Immunol. Methods 2008, 329, 151–156. (17) Zheng, S.; Yung, R.; Tai, Y.-C.; Kasdan, H. IEEE Proc. MEMS 2005, 851854.

9180

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

Figure 2. Brightfield image of a device (a) without surface treatment and (b) with a BSA-adsorbed surface following an experiment with H1975 cells, demonstrating effective prevention of clogging with large cells (up to 40 µm diameter) in the flowing sample.

Figure 3. Average cell size for each outlet following the H1975 epithelial cell fractionation. The number of H1975 cells input into the device was approximately 3 × 105 (input cell concentration was 5 × 105 cells/mL). Outlet 1 is not shown as no cells were found. Error bars denote standard errors based on five independent experiments.

in Figure 2. BSA was chosen for this application because of its well-documented use in microfluidic cell separation devices;18,19 however, other antiadhesive molecules such as poly(ethylene glycol) can also be utilized to play a similar role.20 Using these devices treated with BSA, the H1975 cell fractionation was performed with the purpose of creating a cell suspension with a size distribution equal to that of cardiomyocytes. The results of these experiments, specifically the cell size distribution in each outlet and the total number of cells emerging from each outlet, are shown in Figures 3 and 4, respectively. To mimic the cardiomyocyte population in digested native myocardium tissue, (18) Cheng, X.; Irimia, D.; Dixon, M.; Sekine, K.; Demirci, U.; Zamir, L.; Tompkins, R. G.; Rodriguez, W.; Toner, M. Lab Chip 2007, 7, 170–178. (19) Nagrath, S.; Sequist, L. V.; Maheswaran, S.; Bell, D. W.; Irimia, D.; Ulkus, L.; Smith, M. R.; Kwak, E. L.; Digumarthy, S.; Muzikansky, A.; Ryan, P.; Balis, U. J.; Tompkins, R. G.; Haber, D. A.; Toner, M. Nature 2007, 450, 1235–U1210. (20) Murthy, S. K.; Sin, A.; Tompkins, R. G.; Toner, M. Langmuir 2004, 20, 11649–11655.

Figure 4. Number of cells emerging from each outlet following the H1975 epithelial cell fractionation. The number of H1975 cells input into the device was approximately 3 × 105 (input cell concentration was 5 × 105 cells/mL). Outlet 1 is not shown as no cells were observed. Error bars denote standard errors based on five independent experiments.

outlets 4 and 5 were pooled and measured for size using a Coulter Counter. The H1975 cells in this combined suspension had an average diameter of 17.3 ± 2.7 µm which is comparable to the size of rat cardiomyocytes (17.0 ± 4.0 µm). This suspension of H1975 cells was combined with a suspension of 3T3 fibroblasts with the same concentration to create the 50:50 mixture representative of rat cardiac tissue digestate. Following the creation of the model cell suspension, the ability of the device shown in Figure 1 to separate the two populations was examined. The composition of the starting suspension and the measured composition of suspensions emerging from each of the outlets is shown in Figure 5. Outputs 4, 5, and 6 contained 94.7 ± 1.0%, 98.2 ± 1.0%, and 99.6 ± 0.4% H1975 cells, respectively. On the other end of the spectrum, outlets 2 and 3 contained 97.4 ± 0.3% and 81.4 ± 0.8% 3T3 cells, respectively. Varying the cell input composition also had no effect on purity or yield. If outputs 4, 5, and 6 were pooled, a suspension containing approximately 97.0 ± 1.0% H1975 cells would be obtained with approximately 90% of the H1975 cells flowed through the device recovered. This level of purity, achieved by only a single pass of the suspension through the microfluidic device, is markedly higher than that achieved for cardiomyocytes following two cycles of preplating (81.0 ± 14.0%4). Furthermore, the high level of yield in the DLD device makes the separation process efficient and attractive from the standpoint of small sample volumes arising from limited tissue availability. The viability and function of the cells obtained from all of the outputs were assessed by a Trypan blue assay (>97% of cells were viable) and by plating these cells in culture medium for a period of 24 h, respectively. At the end of the plating period, both cell types were observed to spread normally (results not shown). These observations indicate that the separation process, which occurs in only 3 min, does not adversely affect the cells, despite the relatively high flow rate. Another consideration for cell separation applications is the length of time over which a device can be operated without losing performance in terms of purity and yield. No clogging was

Figure 5. Outlet composition following the separation of H1975 epithelial cells and 3T3 fibroblasts. The total number of cells input into the device (as a mixed suspension containing equal numbers of H1975 and 3T3 cells) was approximately 3 × 105 (input cell concentration was 5 × 105 cells/mL). Outlet 1 is not shown as no cells were found. Error bars denote standard errors based on five independent experiments.

observed for all experiments where the input suspension concentration was held at 5 × 105 cells/mL (homogeneous H1975 cells or mixed suspensions with 3T3 cells), and the device could be run indefinitely (order of several hours or longer) without any loss of performance. This long run time is significant given that these devices are not designed to be washable and reusable. However, with higher input concentrations, specifically 1 × 106 cells/mL or higher, clogging was observed. With increased input concentrations, a greater number of cells will try to pass through the small space between any two posts, thereby causing clogging. For this reason, larger and more deformable cells will have a lower concentration threshold above which clogging occurs. Since the lateral displacement process inherently involves collisions with the obstacles in the flow path, another challenge associated with greater cell deformability is reduced accuracy in predictions of cell displacement based on the obstacle layout, which would result in lower separation efficiencies. The quantity of cells required for tissue engineering can range from relatively small (on the order of 106 cells) for functional tissue “patches”3 to relatively large (on the order of 109 cells) for whole organ replacements.21 If a large number of cells is required, the device described herein can be operated in a highthroughput mode by multiplexing, i.e., operating a large number of the same devices in parallel with an appropriate input cell concentration. This is the preferred mode for high throughput because the throughput of each individual device is limited by the device volume and input cell concentration, as discussed above. Using higher flow rates will provide a higher throughput, but only to a point, as stronger collisions between cells and the obstacles may cause greater cell (21) Ott, H. C.; Matthiesen, T. S.; Goh, S. K.; Black, L. D.; Kren, S. M.; Netoff, T. I.; Taylor, D. A. Nat. Med. 2008, 14, 213–221.

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

9181

deformation (and diminished efficiencies as described above) and also cell damage. The ultimate strength of this DLD device may be in separation of cells from small volumes, such as cardiac biopsies for diagnostic purposes, that would otherwise be difficult to achieve using standard methods such as preplating. The performance of the DLD device can be compared with the conventional preplating process used in cardiac tissue engineering in terms of yield and throughput. Each cycle of preplating requires at least 1 h for a suspension containing approximately 108 cells22 (in a 15 mL volume), and some of the desired cardiomyocytes are invariably lost by adhesion to the culture flask surface, along with the undesired nonmyocytes. Two cycles of preplating are required to obtain a cardiomyocyte purity level of 81.0 ± 14.0%.4 With the flow and inlet concentration conditions utilized in the present work, the DLD device can process 105 cells/min. The ability to run a large number of devices in parallel means that the time required for separation of a large number of cells is only limited by the extent of multiplexing. For example, in order to process 108 cells and obtain 97% purity of H1975 cells, one could either carry out a rapid 10 min single-pass separation with 100 devices in parallel (an order of magnitude in time faster than conventional preplating) or a 100 min separation with 10 devices in parallel (with a similar total time as two cycles of preplating). It is important to note that in order to obtain 97% or purity with the preplating technique, it would be necessary to employ at a minimum 3 or 4 cycles of preplating (∼4 h). During this time, the viability of nonadherent cardiomyocytes will significantly decrease and although purity will go up, cell viability will be compromised. Running microfluidic devices in a parallel configuration is, thus, clearly superior to conventional preplating. The DLD device described in this study was able to maintain high purity and yield while separating on the basis of size. Furthermore, the issue of clogging that has been encountered with some PDMS-based microfluidic DLD separation devices15,17 was avoided by BSA adsorption. Size-based enrichment was accomplished with only a single pass of the starting suspension (22) Carrier, R. L.; Papadaki, M.; Rupnick, M.; Schoen, F. J.; Bursac, N.; Langer, R.; Freed, L. E.; Vunjak-Novakovic, G. Biotechnol. Bioeng. 1999, 64, 580– 589.

9182

Analytical Chemistry, Vol. 81, No. 21, November 1, 2009

through the device, and the device had only a single type of post array as opposed to multiple regions with different parameters of post size, spacing, and offset.8,16 The significance of this work lies in the recognition that DLD design principles can be applied for the microfluidic enrichment of large cells, up to the 40 µm diameter level examined in this work. While large rigid particles have been examined,16 to our knowledge, no other reports exist of cell fractionation by DLD at these large size scales; this distinction is important because of the wide range of deformability that large cells can exhibit. An important caveat in the present work is that some level of sample preparation is needed for this device to function properly, whether the sample consists of the model cell mixture considered herein or primary cardiac cells. Broken cells need to be removed so that intracellular material does not adhere to posts and cause clogging. DNA fragments commonly found in the primary cell isolates upon enzymatic digestion also need to be removed (using DNAase I, for example). Also, other larger contaminants (such as dust particles) that may be present need to be removed as they will also disrupt the cell path between obstacles. While these contaminants can generally be removed by centrifugation, additional filtration may be needed with an in-line syringe filter. ACKNOWLEDGMENT We gratefully acknowledge financial support from the National Heart Foundation, a program of the American Heart Association Foundation (Grant H2007-017 to S.K.M.), the National Science Foundation (Grant 0827868 to S.K.M.), and the Canadian Natural Sciences and Engineering Research Council (Discovery Grant RGPIN 326982-06 to M.R.). We also thank Prof. Mehmet Toner for access to the Beckman Z2 Coulter Counter and Melissa Brown for providing size measurements of the primary cardiac cells. SUPPORTING INFORMATION AVAILABLE Protocol for the isolation and size measurement of cardiac myocytes and fibroblasts. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review August 14, 2009. Accepted September 22, 2009. AC9018395