UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Automated microfluidic device for the separation of cancer cells from blood Lin, Bill Kengli 2012

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2013_spring_lin_billkengli.pdf [ 4.31MB ]
Metadata
JSON: 24-1.0073460.json
JSON-LD: 24-1.0073460-ld.json
RDF/XML (Pretty): 24-1.0073460-rdf.xml
RDF/JSON: 24-1.0073460-rdf.json
Turtle: 24-1.0073460-turtle.txt
N-Triples: 24-1.0073460-rdf-ntriples.txt
Original Record: 24-1.0073460-source.json
Full Text
24-1.0073460-fulltext.txt
Citation
24-1.0073460.ris

Full Text

AUTOMATED MICROFLUIDIC DEVICE FOR THE SEPARATION OF CANCER CELLS FROM BLOOD  by Bill Kengli Lin  B.A.Sc., The University of Waterloo, 2009  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (Mechanical Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2012  © Bill Kengli Lin, 2012  Abstract The ability to separate cells based on biomechanical properties such as size and deformability is emerging as a potential alternative to biochemical methods for cell separation, particularly in cases where biochemical markers are unknown or expressed at low levels. The separation of circulating tumor cells (CTCs) is an example problem where this type of technology is important because the cell surface markers currently used to capture these cells are known to be unreliable. The performance of existing biomechanical cell separation techniques is currently hindered by clogging, which reduces specificity of the separation process. We previously demonstrated a microfluidic ratchet mechanism that overcomes the reversible nature of low Reynolds number flow. In this thesis, we leverage this mechanism to prevent clogging while preserving high selectivity by periodically clearing the filter microstructure to create an automated microfluidic platform that demonstrates the size and deformability-based separation of cultured human bladder UC13 cancer cells from white blood cells (WBCs). This platform has two components: the first is a size-based hydrodynamic concentrator, which performs an initial sample preparation step to reduce the sample volume while removing a fraction of the contaminant WBCs. The second is an automated cell separation device where cells are transported through a 2D array of ratcheting funnel constrictions and sorted using an oscillatory flow. We evaluate the ability of this platform for separating rare cancer cells doped into WBCs at low concentration to assess the potential of this technology for biomechanical separation of CTCs. Specifically, using a sample where cancer cells are doped into WBCs at a ratio of 1:1000, the combined system achieved a cancer cell yield of 96.0±0.1%; the outlet had a purity of 75±3%; and the population of cancer cells in the mixture was enriched by a factor 3000 (+643, -278).  ii  Preface Section 3.4 describes an off-chip pressure board controller, a C++ microprocessor code, and a Visual Basic GUI control program. The pressure board was designed by Dr. Hongshen Ma and built by Peter Woo. Initial versions of the microprocessor code and the VB program were developed by Dr. Hongshen Ma, Isaac Tang, and Peter Woo.  Sections 4.1 and 4.2 describes the process of silicon master fabrication and polydimethylsiloxane device molding. The early protocols for manufacturing were developed by Sarah McFaul, who also assisted in the fabrication of Generation 3.5 and 6 devices.  A version of Sections 2.1, 2.2.1, 3.2, and Chapters 1, 4, and 6 has been submitted for publication at Lab on a Chip, entitled ‘Highly selective biomechanical separation of cancer cells using a microfluidic ratchet mechanism and hydrodynamic concentrator’.  Figures 1.1, 1.2, 1.3, 1.4, 1.5, and 5.1 are adopted from other publications and permission is granted from their respective publishers. The photo in Figure 6.1 was taken by my colleague Will Beattie, who has kindly provided permission to use the photo.  iii  Table of Contents Abstract .................................................................................................................................... ii Preface ..................................................................................................................................... iii Table of Contents ................................................................................................................... iv List of Tables ........................................................................................................................ viii List of Figures ......................................................................................................................... ix List of Abbreviations ............................................................................................................ xv Acknowledgements .............................................................................................................. xvi Chapter 1: Introduction ........................................................................................................ 1 1.1  Definition of Performance Metrics ........................................................................... 2  1.2  Literature Review...................................................................................................... 3  1.2.1  Affinity Capture .................................................................................................... 3  1.2.2  Hydrodynamic Methods........................................................................................ 5  1.2.3  Dielectrophoresis (DEP) ....................................................................................... 7  1.2.4  Filtration Techniques ............................................................................................ 8  1.3  Ratchet Mechanism ................................................................................................... 9  1.4  Combined Hydrodynamic-Filtration System .......................................................... 11  1.5  Goals of the Thesis.................................................................................................. 11  Chapter 2: Cell Concentrator ............................................................................................. 13 2.1  Motivation and Mechanism Principle ..................................................................... 13  2.2  Modeling ................................................................................................................. 14  2.2.1  Analytical Model ................................................................................................ 14  2.2.2  Numerical Model ................................................................................................ 18 iv  2.3 2.3.1  Preliminary Concentrator Characterization ............................................................ 20 Concentrator Design Validation ......................................................................... 20  Chapter 3: Ratchet Cell Sorter ........................................................................................... 24 3.1  Microfluidic Ratchet Mechanism ........................................................................... 24  3.2  Cell Sorting Funnel Array ....................................................................................... 25  3.3  Precision Flow Control ........................................................................................... 27  3.3.1  Tree Microstructure Network ............................................................................. 27  3.3.2  Cell Infusion........................................................................................................ 28  3.3.3  Cell Extraction .................................................................................................... 30  3.4  Automation ............................................................................................................. 30  3.4.1  Off-Chip Pressure Control .................................................................................. 30  3.4.2  Automation Software .......................................................................................... 31  3.5 3.5.1  Preliminary Cell Ratchet Sorter Characterization................................................... 33 Overall Concentration Effects............................................................................. 33  Chapter 4: Fabrication and Experimental Procedures .................................................... 36 4.1  Fabrication of Silicon Masters ................................................................................ 36  4.2  Fabrication of PDMS Devices ................................................................................ 37  4.3  Sample Preparation ................................................................................................. 37  4.4  Experimental Setup and Preparation ....................................................................... 38  Chapter 5: Design Iterations ............................................................................................... 40 5.1  Generation 3 to Generation 3.5 ............................................................................... 41  5.1.1  Sorting Region .................................................................................................... 41  5.1.2  Plumbing ............................................................................................................. 42  v  5.1.3 5.2  Macro-scale Fluid Handling ............................................................................... 44 Generation 3.5 to Generation 4 ............................................................................... 44  5.2.1  Sorting Region .................................................................................................... 44  5.2.2  Plumbing ............................................................................................................. 45  5.2.3  Automation Changes ........................................................................................... 48  5.2.4  Macro-scale Fluid Handling ............................................................................... 49  5.3  Generation 4 to Generation 5 .................................................................................. 50  5.3.1  Sorting Region and Concentrator........................................................................ 50  5.3.2  Plumbing ............................................................................................................. 50  5.3.3  Macro-scale Fluid Handling ............................................................................... 51  5.3.4  Imaging ............................................................................................................... 53  5.4  Generation 5 to Generation 6 .................................................................................. 53  5.4.1  Sorting Region and Concentrator........................................................................ 53  5.4.2  Plumbing ............................................................................................................. 53  5.4.3  Macro-scale Fluid Handling ............................................................................... 55  Chapter 6: Cell Separation Results and Discussion.......................................................... 56 6.1  Performance Characterization and Uncertainty Estimation .................................... 56  6.2  Hydrodynamic Concentrator ................................................................................... 61  6.2.1  Cell Concentration Optimization ........................................................................ 61  6.2.2  Hydrodynamic Concentrator Performance Characterization .............................. 64  6.3  Ratchet Cell Sorter .................................................................................................. 64  6.3.1  Generation 5 Ratchet Sorter Performance Characterization ............................... 65  6.3.2  Generation 6 Ratchet Sorter Performance Characterization ............................... 67  vi  6.4  Combined System Characterization ........................................................................ 68  6.4.1  Generation 5 Combined System Performance .................................................... 68  6.4.2  Generation 6 Combined System Performance .................................................... 70  6.5  Discussion of Separation Results ............................................................................ 73  Chapter 7: Conclusion ......................................................................................................... 75 7.1  Summary of Results ................................................................................................ 75  7.2  Limitations .............................................................................................................. 75  7.3  Future Work ............................................................................................................ 76  References .............................................................................................................................. 77  vii  List of Tables Table 2.1 Flow ratios and overall fluid removal performance of concentrator system. ......... 17 Table 2.2 Percentage of the total amount of fluid removed at each of the concentrator mechanisms as determined from the 3D CFD simulation. ................................................... 20 Table 2.3 Experimental results for the different fluid removal resistances (corresponding to different punch holes) ............................................................................................................. 22 Table 3.1 Each parameter for the automation program. ......................................................... 33 Table 3.2 Results for yield, purity, and enrichment at different PBMC concentrations......... 35 Table 5.1 Outline of the design changes made throughout the different generations............. 40 Table 5.2 The funnel gap size distribution for Generation 3.5 and two variations of Generation 4 ............................................................................................................................ 44 Table 6.1 Sample data from a Generation 6 ratchet sorter experiment for the 1:600 doping ratio. ........................................................................................................................................ 58 Table 6.2 Concentrator yield performance at different PBMC concentrations. ..................... 61 Table 6.3 Yield and enrichment performance for different doping ratios of UC13:WBC for the Generation 5 ratchet sorter device. ................................................................................... 66 Table 6.4 Yield and enrichment performance for different doping ratios for the Generation 6 ratchet sorter device ................................................................................................................ 67 Table 6.5 Yield, purity, and enrichment performance for the Generation 5 combined system. ................................................................................................................................................. 69 Table 6.6 Yield, purity, and enrichment performance for the Generation 6 combined system. ................................................................................................................................................. 71  viii  List of Figures Figure 1.1 (A) The CTC-Chip comprised of 78,000 microposts functionalized with antiEpCAM onto which cancer cells bind through increased cell-surface interactions (Reprinted by permission from Macmillan Publishers Ltd: [Nature] [4], Copyright 2007) (B) Grooves on the ceiling of the herringbone-chip increase interactions between cancer cells and the functionalized channels by creating turbulent mixing in the flow (Reprinted with permission from [21]. Copyright 2010, Proceedings of the National Academy of Sciences). ................... 4 Figure 1.2 An example of pinch flow fractionation. Heterogeneous mixture of particles enter one inlet and a particle-free liquid is introduced at the other inlet. The flow rates of the two inlets are controlled to align particles along the desired streamline at the pinch constriction. Particles are separated according to their sizes in the expansion region. (Reprinted with permission from [22]. Copyright 2004, American Chemical Society) ..................................... 6 Figure 1.3 As a hetereogeneous sample of cells enters the channel cells are aligned towards the side walls to an equilibrium position, Xeq. At the expansion point, larger particles experience a higher lateral lift force, forcing the cancer cells into the reservoirs (Reprinted with permission from [5]. Copyright 2011, American Institute of Physics). .......................... 7 Figure 1.4 Cells of different properties can have different equilibrium positions inside the channel. With a parabolic flow profile, cells at different positions travel at different velocities and can be separated (Reprinted with permission from [27]. Copyright 2000, American Chemical Society). .................................................................................................................... 8 Figure 1.5 (1) A sample of cells is introduced at the bottom of the sorter. (2) An oscillatory pressure is applied with an upward bias, and the cells undergo irreversible separation. The smaller cells transit to the top of the device and the cancer cells are captured at the bottom. In  ix  (3), post-separated cells are purged from the device and collected off-chip ([20] - Reproduced by permission of The Royal Society of Chemistry)................................................................ 10 Figure 1.6 Schematic of the cell separation system. The heterogeneous sample is first reduced by the concentrator mechanism through removing a fraction of the fluid and contaminant WBCs. The remainder of sample is processed by the sorting device and can be collected from the outlet. ........................................................................................................ 11 Figure 2.1 (A) Schematic of the concentrator mechanism. The fluid removal ports remove fluid containing some contaminant cells while target cells are retained and carried downstream along the bifurcated channel. (B) Target cancer cells enter the pinch constriction from different streamlines, but are re-aligned at the constriction to streamlines adjacent to the inner side of the bifurcation channels. Length dimensions are shown in microns. (C) Numerical simulation result showing the velocity contour and streamlines at the concentrator mechanism. ............................................................................................................................. 14 Figure 2.2 Electronic-hydraulic analogy of the concentrator design. ..................................... 15 Figure 2.3 Circuit analogy diagram for a single concentrator mechanism. The outlet bifurcation channel is shown as a single combined resistance Ri. .......................................... 17 Figure 2.4 Velocity plot with streamlines suggesting cell direction after the pinch region. .. 18 Figure 2.5 CFD results showing the velocity distribution inside the concentrator. ................ 19 Figure 2.6 Design of the concentrator device used for initial characterization. Each of the punch holes represents a different hydrodynamic resistance for the fluid removal ports. ..... 21 Figure 2.7 Performance trend for the concentrator in showing yield, purity, enrichment ratio, and percentage fluid removal at various punch holes with greater numbered holes corresponding to higher removal port resistances. ................................................................. 23  x  Figure 3.1 Cell deformation through the funnel constriction in (A) forward flow and (B) reverse flow. ............................................................................................................................ 25 Figure 3.2 (A) Schematic of device design. (B) A heterogeneous mixture of WBC and UC13 cells enter the bottom row of the sorting region. (C) UC13 cells get trapped at the lower funnel rows before the critical 6 μm cut-off. (D) WBCs travel past the critical cut-off and are removed into a separate outlet. ............................................................................................... 26 Figure 3.3 Schematic showing the separation region with the 6 µm cut-off point. Cancer cells are extracted from the bottom rows while contaminant cells are extracted from the top rows. ................................................................................................................................................. 27 Figure 3.4 (A) The sieve valve consists of a deflection membrane that does not fully close, creating a gap between the top of the membrane and the top wall of the flow channel. (B) Sieve actuation blocks the cells while allowing fluid to pass through, creating a cell free zone for the valve to actuate freely.................................................................................................. 28 Figure 3.5 A custom-made cap which applies the pressure directly to the sample carried on the chip. ................................................................................................................................... 29 Figure 3.6 The pressure board setup. (1) The high pressure regulators. (2) Low pressure regulators. (3) Microcontroller and circuit board system. (4) Manifolds along with solenoid valves. ..................................................................................................................................... 31 Figure 3.7 The Visual Basic program used to perform automation and the batch cycle process, and the batch processing cycle. ................................................................................. 32 Figure 3.8 Plot showing yield, purity, and enrichment trends. Enrichment is plotted on the secondary axis (right side). ..................................................................................................... 35  xi  Figure 4.1 Composite image of a portion of the outlet after sorting. Cancer cells fluoresce green (A) while both cancer cells and WBCs fluoresce blue (B). Individual images were taken with a monochrome camera, stitched using Microsoft Image Composite Editor and artificially colored in Adobe Photoshop. ................................................................................ 39 Figure 5.1 Schematic of the Generation 3 device. Valves are denoted by the green areas, and flow channels denoted in gray ([20] - Reproduced by permission of The Royal Society of Chemistry)............................................................................................................................... 41 Figure 5.2 CAD drawings of Generation 3 and 3.5 devices. Most significant functional changes include A) increasing the valve cross sectional area, B) increasing the hydrodynamic resistance at the cell and buffer inlets, and C) separating the exit ports. ................................ 43 Figure 5.3 Generation 4 funnels are reduced by 33% with rounded trailing ends to facilitate manufacturability and improve cell handling. ........................................................................ 45 Figure 5.4 CAD drawing of Generation 4 device. Outlined are all the physical changes made from Generation 3.5. ............................................................................................................... 47 Figure 5.5 CFD analysis showing the vertical flow velocity during the downward cycle of an oscillatory-purge. .................................................................................................................... 49 Figure 5.6 A) The bottom-feed tube. B) The entire setup. ..................................................... 50 Figure 5.7 CAD layout of the Generation 5 device. The more significant change from Generation 4 is the addition of a cell concentrator. ................................................................ 52 Figure 5.8 CAD layout of the Generation 6 device. The green and blue areas are SU-8, and the red areas are SPR. ............................................................................................................. 54 Figure 5.9 Cell concentration over time when suspended in MEM and 15% Ficoll. ............. 55  xii  Figure 6.1 Brightfield and fluorescent images of two different UC13 cells. In A) the cell has an irregular shape and a weak fluorescence signal. In B) the cell has a circular morphology and a strong fluorescence. The cell found in A) is counted as an uncertain cell (photo courtesy of Will Beattie). ........................................................................................................ 58 Figure 6.2 Results showing UC13 cell retention at different PBMC concentrations. The first data point at 1 cell/µL acts as a control for UC13 retention when there are no PBMCs present. Plot is shown in logarithmic scale. ............................................................................ 62 Figure 6.3 Fluorescent images of one section of the concentrator showing non-specific adhesion of WBCs onto the PDMS walls. A) Image taken after 15 minutes of operation when no Pluronic is used. This image was taken after the sample has been flushed through the concentrator, which means that only WBCs adsorbed to the PDMS walls are shown. B) When Pluronic is used, there is minimal non-specific adhesion even at the stagnation point. This image was taken while cells were flowing through the concentrator after one hour of processing. .............................................................................................................................. 63 Figure 6.4After a sorting cycle, smaller, more deformable WBCs stained in blue travel to the higher funnel rows shown in A) brightfield, B) green fluorescence, and C) blue fluorescence. Larger and more rigid UC13 cancer cells are retained near the bottom, shown in D) brightfield, E) green, and F) blue. ........................................................................................... 65 Figure 6.5 A) Results showing average capture yield of 93% for different UC13 doping concentrations. B) Results showing enrichment for the four UC13 doping concentrations. .. 66 Figure 6.6 A) Results showing capture yields of above 95% at various UC13 doping concentrations. B) Results showing enrichment for the three UC13 doping concentrations, with poorer performance recorded in the 1:10 case. ............................................................... 68  xiii  Figure 6.7 Results showing capture yield at different UC13 doping concentrations. The yield remains consistent, even in diminishing UC13 concentrations. B) Plot for enrichment and (C) purity, Note the vertical axis in plot C) only extends to 10% to facilitate the data range. ..... 70 Figure 6.8 A) Results showing capture yield at different UC13 doping concentrations. The yield remains consistent, even in diminishing UC13 concentrations. B) Plot for enrichment and (C) purity. ......................................................................................................................... 72  xiv  List of Abbreviations BSA - Bovine Serum Albumin CTC - Circulating Tumor Cells PBS - Phosphate Buffered Saline PBMC - Peripheral Blood Mononuclear Cells PDMS - Polydimethylsiloxane WBC - White Blood Cells  xv  Acknowledgements First, I would like to thank my supervisor, Dr. Hong Ma, for providing me with the opportunity to work in his lab. I also want to acknowledge his patience in working with me. This experience has been dynamic and extremely rewarding on both the academic and personal development aspects. Secondly, I would like to thank my colleagues and friends at the Multi-Scale Design laboratory as well as in Bio-MEMS. In particular I would like to acknowledge Dr. Linfen Yu for her kindness and willingness to help with all the problems I have encountered. I would like to thank Sarah McFaul for her assistance in experimental work and proofreading this thesis. Finally, I want to thank my family for their endless love and support. I also want to thank my life-long friend Lily Man for her unwavering support in my endeavors.  xvi  Dedication  To my family and closest friends  xvii  Chapter 1: Introduction The separation of cells based on their biomechanical properties, such as size and deformability, has long been considered attractive because of the potential to avoid relying on cell surface markers, as well as the ability to extract viable cells after separation. Recently, the separation of circulating tumor cells (CTCs) has emerged as a key application for such technologies because available cell surface markers are potentially unreliable [1], and because CTCs in peripheral blood are thought to be morphologically distinct from normal WBCs. Specifically, CTCs are thought to be more similar to primary tumor cells, which are larger and less deformable than WBCs [2]. One of the key challenges of this separation problem is the rarity of CTCs relative to WBCs, which may occur at ratios as low as 1:106 and require highly selective separation processes to retain the target cancer cells while depleting the background WBCs. Microfluidic technologies present the ability to fabricate structures at the length scale of individual cells and allow precise control over the flow of minute volumes of liquid, enabling new approaches to both biomechanical and biochemical cell separation. Current techniques include affinity capture [3, 4], hydrodynamic methods [5, 6], dielectrophoresis [7, 8], and micro-filtration [9, 10]. Affinity capture utilizes microstructures functionalized with antibodies that target and bind to cancer cell surface proteins. This approach is selective, but is limited to trapping cells with a known protein expression level, which may be potentially unreliable due to the heterogeneity of CTCs [11, 12]. Hydrodynamic methods use microscale geometries to alter the flow of particles primarily based on differences in size. This approach typically has greater throughput, but has limited selectivity [5, 13, 14]. Dielectrophoresis consists of using a non-uniform electric field to isolate cells based on differences in dielectric properties. This method has shown to achieve adequate yield and purity [7, 8], but the dielectric property of CTCs is unknown and hence this method of separation may be unreliable. Filtration methods involve flowing a sample through an array of micro-scale constrictions to separate cells based on a combination of size and deformability. This approach is highly selective, but is limited by clogging. Specifically, as the filter constrictions become progressively blocked by captured cells, the overall hydrodynamic resistance of the filter changes unpredictably, altering the pressure gradient applied to the remaining open pores [15, 16]. Additionally, the persistent force on the trapped 1  cells can cause cytoskeleton remodeling and increased cytoadhesion, which further degrades filter performance [15, 17]. Finally, in many filtration techniques, captured cells cannot be extracted from the filter microstructure, limiting downstream processing options such as molecular characterization, genomic analysis, and propagation in xenograft models [18, 19]. We previously demonstrated a microfluidic ratchet mechanism that uses irreversible transport of cells through funnel shaped constrictions to sort cells based on a combination of size and deformability, while mitigating clogging by periodically clearing the filter microstructure [20]. In this thesis, we leverage this ratchet mechanism to create an automated microfluidic platform that demonstrates separation of cancer cells from WBCs. This platform has two components. The first component is a size-based hydrodynamic sample preparation stage that concentrates the sample while removing a portion of the contaminant WBCs. The second component is a cell separation stage that uses unidirectional transport of cells through a 2D array of funnel constrictions using oscillatory flow. We characterize and optimize the performance of the concentrator and sorting device individually for reliable cancer cell separation, and then demonstrate highly selective separation of cancer cells from WBCs using the combined system. Chapter 1 will first present performance metrics commonly used to assess the performance of cancer cell separation technologies, followed by an in depth technical review of emerging microfluidic platforms, and ending with the goals of this thesis.  1.1  Definition of Performance Metrics  The performance of a cancer cell separation platform is assessed using the three metrics of yield, purity, and enrichment ratio. Yield is defined as the ratio of target cells captured by the device compared to the initial amount as defined in equation 1.1.  Yield =  Target cells final Target cellsinitial  *100%  (Eq. 1.1)  Based on equation 1.1, the aim for any cancer cell separation system is to achieve 100% yield. Purity is also an important factor in rare cell separation. This metric reflects the amount of biological noise present in the final collection outlet. Purity is defined as the ratio 2  between the number of rare cells to the total number of cells at the collection port, as described in equation 1.2.    Target cells Purity =   *100% Target cells + Contaminant cells   final  (Eq. 1.2)  Finally, the enrichment ratio compares the ratio of target versus contaminant cells after separation to the ratio prior to separation, as described in equation 1.3.  Enrichment Ratio =  1.2 1.2.1  (Target cells/Contaminant cells) final (Target cells/Contaminant cells)initial  (Eq. 1.3)  Literature Review Affinity Capture  Affinity capture is a biochemical approach aimed at capturing CTCs by targeting the epithelial cell adhesion molecule (EpCAM) protein. When this approach is incorporated onto a microfluidic platform, typically microstructures functionalized with EpCAM antibodies (anti-EpCAM) are utilized to capture CTCs based on biochemical binding. Compared to the current commercial system (VeridexTM CellSearchTM), which also an EpCAM-based system, the microfluidic approach increases the probability of interaction between the cancer cells and the EpCAM antibody to improve yield. In a highly recognized study, Nagrath et al demonstrated significantly greater yield performance by using 78,000 anti-EpCAM functionalized microposts within a 970 mm2 surface [4]. Named the “CTC-Chip”, they were able to isolate >65 % of doped cancer cells regardless of EpCAM expression at approximately 50% purity [4]. They claimed the capture potential for low-EpCAM expression cells is comparable to that of high expression cells due to the increased cellsubstrate interactions characteristic of their design [4]. Utilizing the same concept of increasing surface-cell interactions, Stott et al made a herringbone chip functionalized with anti-EpCAM that intentionally creates chaotic mixing even at low Reynolds numbers [21]. The device consists of herringbone grooves at the top channel walls designed to create microvorticies which increase the interaction between CTCs and the functionalized PDMS 3  surfaces. With this new design, they achieved a yield of 91.8% for spiked cancer cells in whole blood, a 26.3% improvement over the original CTC-Chip design [21]. Figure 1.1 shows the functionality of both chip designs.  A B  Figure 1.1 (A) The CTC-Chip comprised of 78,000 microposts functionalized with anti-EpCAM onto which cancer cells bind through increased cell-surface interactions (Reprinted by permission from Macmillan Publishers Ltd: [Nature] [4], Copyright 2007) (B) Grooves on the ceiling of the herringbonechip increase interactions between cancer cells and the functionalized channels by creating turbulent mixing in the flow (Reprinted with permission from [21]. Copyright 2010, Proceedings of the National Academy of Sciences).  Despite the high capture rate achieved through affinity capture, these results were obtained using cultured cancer cells of a known EpCAM expression doped into healthy donor blood [4, 21]. However, for the case of real CTCs, it has been found that some cells do not express EpCAM, and that in many cases the EpCAM expression is reduced due to the epithelial-to-mesenchymal transition which occurs during metastatic progression [11, 12]. Consequently, the capture efficiency for the affinity capture approach may be lower for clinical blood samples. In addition, affinity capture binds the cancer cells tightly to the functionalized PDMS surface and the captured cells cannot be released for subsequent off-chip analysis. Although some analysis can be performed on-chip (expression testing of tumor-specific antigens, RTPCR, FISH assay) [4, 21], there are limitations to on-chip processing. Certain processing options such as single cellular genomic analysis and propagation in xenograft models [18,  4  19] cannot be performed. Given these limitations, other separation methods that do not depend on cell surface biomarkers are explored.  1.2.2  Hydrodynamic Methods  Hydrodynamic cell sorting separate cells based on differences in size by using microstructures to transport cells to desired streamlines through physical contact in conjunction with flow manipulation (such as contracting and expanding the flow) or by flowing cells at sufficiently high flow rates to achieve inertial focusing. A popular hydrodynamic method that utilizes physical contact is pinch flow fractionation, where a pinch constriction is used to align particles to the desired streamline where target cells can be collected in a subsequently expanded microchannel. Using this principle, Yamada et al utilized a pinched region in conjunction with controlled flow rates from two different inlets to align a heterogeneous mixture of particles along a desired streamline close to the side wall of the constriction (Figure 1.2). Subsequent collection of the particles showed a at 92% yield and small particles at 99% yield [22].  5  Figure 1.2 An example of pinch flow fractionation. Heterogeneous mixture of particles enter one inlet and a particle-free liquid is introduced at the other inlet. The flow rates of the two inlets are controlled to align particles along the desired streamline at the pinch constriction. Particles are separated according to their sizes in the expansion region (Reprinted with permission from [22]. Copyright 2004, American Chemical Society).  In inertial flow focusing, a heterogeneous sample of blood and cancer cells flows through microfluidic channels at a high Reynolds number for inertial lift forces to take effect. Two inertial lift forces act on a cell: a shear-gradient lift force FLS that draws particles closer to the channel wall, and a wall effect lift force FLW that repels particles away [5]. For a cell with a given diameter a, the two forces balance to achieve a distinct lateral equilibrium position within the channel [5, 6]. Di Carlo proposed the application of possible cell separation based on different equilibrium channel positions between cancer and blood cells [23]. Hur et al utilized this difference in lateral migration to achieve separation by introducing expansion reservoirs along a straight main channel (Figure 1.3). In their technique, cells are first aligned towards the side walls of a microchannel from the inertial lift force balance. At the expansion region, larger particles experience a greater F LS compared to the smaller blood cells, forcing the target cancer cells into the reservoirs [5]. This process  6  was able to isolate target cells at 23% yield at an enrichment of 7.06. In a different study, Hur et al was able to improve the separation performance based on cell size and deformability to achieve a much higher separation efficiency of 96% at an enrichment of 5.35 over diluted whole blood [6]. More recently, Bhagat et al utilized an inertial method coupled with a pinch region to isolate doped cancer cells at a high yield of 80% while achieving 104 enrichment over WBCs [24]. Typically, hydrodynamic methods have high throughput but lower selectivity [5, 6]. Therefore, it may be used as a first-pass sample preparation mechanism.  Figure 1.3 As a hetereogeneous sample of cells enters the channel cells are aligned towards the side walls to an equilibrium position, Xeq. At the expansion point, larger particles experience a higher lateral lift force, forcing the cancer cells into the reservoirs (Reprinted with permission from [5]. Copyright 2011, American Institute of Physics).  1.2.3  Dielectrophoresis (DEP)  Dieletrophoresis separates cells based on differences in their polarization under a nonuniform electric field. Dielectric properties depend on physiological properties such as cell membrane composition, cell size, and cytoplasm electrical properties [7, 25]. In this technique, gravitational (FG), wall lift (FLW), and DEP (FDEPz) forces are balanced to achieve different equilibrium positions for WBC and cancer cells within the channel, followed by the target collection of cancer cells with high yield and purity [26]. Utilizing this method, Becker and Gascoyne demonstrated the separation of human breast cancer cells from blood at a yield of >95% [7, 26]. Cell viability was also >80% in post-separation culture when compared with a control sample [26]. In a recent study, Moon et al combined DEP with multi-orifice flow fractionation (MOFF) to perform a dual stage separation. In the first stage, their hydrodynamic mechanism removed 87.98% of RBCs and 61.94% of WBCs while retaining 7  the cancer cells. The sample then enters the DEP section, where more of the peripheral blood cells are removed, resulting in an overall performance of 99.24% RBC and 94.23% WBC removal. Their platform achieved a yield of 75.81% at an enrichment of 162.4, which corresponds to a purity of 16.24% [8]. Since DEP utilizes an electric field to operate, a conductive material must be patterned onto the glass substrate before PDMS is attached. This step is time consuming and increases the complexity of the system. In addition, the physiological and dielectric properties of CTCs are unknown and are likely to be subject to significant variability, which may significantly compromise the selectivity of DEP based cell separation.  Figure 1.4 Cells of different properties can have different equilibrium positions inside the channel. With a parabolic flow profile, cells at different positions travel at different velocities and can be separated (Reprinted with permission from [27]. Copyright 2000, American Chemical Society).  1.2.4  Filtration Techniques  Mechanical filtration is a common approach to the isolation of cancer cells, with one of the earliest studies dating back to 1965 [28]. Microfluidic fabrication techniques can create complex structures at the characteristic length scale of single cells and therefore are well suited for filtration techniques. A common filtration approach is to etch micropores of a defined size onto a thin substrate [9, 29-31]. In a highly recognized study, Cote et al devised a parylene membrane filter that trapped CTCs based on their size [9]. Their microdevice achieved >90% of CTC recovery at 107 enrichment when using blood samples from patients with metastatic disease. Their device was able to identify 51 out of 57 patients with metastatic disease compared to only 26 detected with the CellSearchTM system [9]. In Cote’s experiment, the blood sample was fixed using formalin prior to processing, which  8  dramatically decreases the deformability of cells in the suspension and increase yield. However, fixing the sample also kills the cells, which limits the number of post-processing options. In a similar approach, Hosokawa et al created a microcavity array made of electroformed nickel to isolate cancer cells based on size and deformability [30]. Their approach is similar to that of Cote et al, but they did not fix the sample prior to sorting. In this study Hosokawa et al were able to isolate doped cancer cells at a yield of >80% at 98% viability [30]. This approach however is susceptible to cytoskeleton remodeling, changes that occur to the structure of the cell to facilitate slippage through the micropore from the persistent force acting on the cell [17]. In a more recent study, Zheng et al developed a 3D microfilter that incoporates a base structure at the bottom of each micropore to balance the forces acting on the cell to prevent cytoloskeleton remodeling and cell lysis. In their study, they achieved 86.50% yield at 1000x enrichment [10]. In all of these filtration-based approaches, the target cells cannot be extracted off the filter for further analysis. Clogging of the microfilter is also a common issue, where cells become permanently lodged onto the micropore, creating an adverse decline in filtration performance [16]. Therefore, there is a need for improved mechanical filtration techniques. One proposed way is to utilize flow in the same plane as the filter elements to isolate target cells. Tan et al designed an array of crescent-shaped isolation wells consisting of three posts positioned strategically to capture target cancer cells while reducing the effects of change in the filter hydrodynamic resistance upon cell capture. This technique is capable of isolating cancer cells at 83% efficiency while achieving a 109 enrichment factor over whole blood [32]. In their design, viable cells can be extracted off-chip by reversing the flow to push the trapped cancer cells out of the capture wells. Their approach however is still susceptible to cytoskeleton remodeling which may enable some target cells to slip through. Therefore, there is a need for a gentle mechanical system that achieves high yield and selectivity but does not induce cytoskeleton remodeling.  1.3  Ratchet Mechanism  Previously, a microfluidic ratchet was created using micro-scale funnel constrictions and this mechanism has demonstrated to enable unidirectional transport of single cells under an unbiased oscillatory flow [33]. The funnel constriction consists of an entrance region that is 9  wider than the diameter of the cell entering the constriction, and an exit gap that is smaller than the cell diameter. When cells transport through the funnel constriction, the pressure required to push the cell towards the exit increases until an instability point, known as Haines’ jump, after which the cell is transported quickly through the gap [34]. In this work, we utilize the ratchet mechanism to create a cell sorting device aimed at separating human cancer cells from peripheral WBCs. We created a sorter consisting of a 2dimensional array of ratcheting funnels with the constriction sizes diminishing from the bottom to the top rows. A sample is infused into the bottom of the sorting region and a biased oscillatory pressure separates smaller and more deformable cells to the top rows while the larger and stiffer cancer cells are retained at the lower rows (Figure 1.5). In this work we also present the design of the microfluidic plumbing required to enable automated, robust, and reliable separation.  3  2  1  Figure 1.5 (1) A sample of cells is introduced at the bottom of the sorter. (2) An oscillatory pressure is applied with an upward bias, and the cells undergo irreversible separation. The smaller cells transit to the top of the device and the cancer cells are captured at the bottom. In (3), post-separated cells are purged from the device and collected off-chip ([20] - Reproduced by permission of The Royal Society of Chemistry).  10  1.4  Combined Hydrodynamic-Filtration System  To leverage off the high throughput advantage of hydrodynamic methods and the high selectivity found in filtration systems, we propose a cell separation system that is a combination of both techniques. Our device consists initially of a hydrodynamic sample concentrator followed by a ratcheting cell sorter in tandem. The sample is first processed by the concentrator to remove the majority of the fluid and a fraction of the contaminant WBCs. This concentrated but reduced sample accumulates in an intermediate reservoir and is sent through the ratchet sorting device for highly selective separation. Finally, the target cells are collected in a reservoir suitable for image analysis (Figure 1.6).  Hydrodynamic Concentrator Sample Inlet Reservoir  Waste WBCs  Intermediate Reservoir (concentrated sample)  Waste WBCs Ratchet Cell Sorter  UC13 Collection Reservoir  Figure 1.6 Schematic of the cell separation system. The heterogeneous sample is first reduced by the concentrator mechanism through removing a fraction of the fluid and contaminant WBCs. The remainder of sample is processed by the sorting device and can be collected from the outlet.  1.5  Goals of the Thesis  The goal of this thesis is to develop an automated microfluidic system for the isolation and enrichment of cancer cells spiked into a sample of contaminant WBCs based on their mechanical differences. Specifically, we aim to:  11    Devise a high throughput on-chip hydrodynamic concentrator mechanism to decrease overall processing time by reducing the sample volume while removing a fraction of contaminant cells    Design a highly selective ratchet-based sorting device to capture cancer cells while removing the remainder of contaminant cells    Automate chip operation procedures to enable robust batch processing  12  Chapter 2: Cell Concentrator This chapter describes the design of the hydrodynamic cell concentrator portion of the microfluidic cell separation platform. The motivation to add the concentrator and the mechanism principle is described in Section 2.1. Modeling of the concentrator design is outlined in Section 2.2. This chapter concludes in Section 2.3 with preliminary optimization and characterization of the concentrator.  2.1  Motivation and Mechanism Principle  Microfluidic platforms typically have higher selectivity compared to macroscale methods in separating rare cells but at a considerably lower throughput [35]. We aim to increase the throughput of our system with the addition of an initial stage to remove the majority of the suspension fluid from the sample, as well as some of the contaminant WBCs, to present a highly concentrated sample to the ratchet sorter. This component utilizes a pinch-flow design, previously used to hydrodynamically separate particles based on size [22, 24, 36]. The concentrator, as shown in Figure 2.1, consists of a central channel that first constricts to an 18 μm pinch, and then expands into two equal channels. Each of these channels has a port connected at its outside edge after the bifurcation point that leads to a waste outlet for fluid and contaminant WBCs. The principle of this mechanism can be described as follows. Cells from the sample can be located at any lateral position in the central channel. At the pinch region, larger cells with diameters comparable to the pinch gap width are shifted to the streamline along the axial center of the channel by the constraint imposed by the pinch. After passing through the pinch, these larger cells continue to follow the central streamline to the inner wall of the bifurcated channels. On the other hand, smaller cells do not experience this alignment and continue to travel along the same streamlines as prior to the pinch (Figure 2.1B). A portion of these smaller cells, along with a fraction of the suspending fluid, are removed by the side ports on the outside edge of the bifurcated channels. The amount of fluid removed is delineated by a critical streamline dividing the fluid into an inner fraction that is retained and an outer fraction that is removed (Figure 2.1A). After fluid and cell removal, the collection channels converge downstream to carry the reduced sample on to a subsequent concentrator mechanism. The complete concentrator consists of four identical concentration mechanisms in series. 13  A  200  Sample inlet B  Contaminant cells Target cancer cells 18 Fluid/ contaminant cell removal port Target cells retained  Sample outlet  o  90  Fluid retained  Fluid removed  Critical Streamline  C  (m/s) (m/s)  0.01 0.009 0.008 0.007 0.006 0.005 0.004 0.003 0.002 0.001 0  Figure 2.1 (A) Schematic of the concentrator mechanism. The fluid removal ports remove fluid containing some contaminant cells while target cells are retained and carried downstream along the bifurcated channel. (B) Target cancer cells enter the pinch constriction from different streamlines, but are re-aligned at the constriction to streamlines adjacent to the inner side of the bifurcation channels. Length dimensions are shown in microns. (C) Numerical simulation result showing the velocity contour and streamlines at the concentrator mechanism.  2.2 2.2.1  Modeling Analytical Model  The fraction of fluid removed was calculated first using an analytical model. In this model, the fluid system was modeled using the electronic-hydraulic analogy (where pressure corresponds to voltage, flow rate to current, and hydrodynamic resistance to electrical resistance). The hydrodynamic resistance of each channel was calculated and modeled using a circuit diagram as shown in Figure 2.2. For model simplicity, the hydrodynamic resistance of the pinch regions was not considered. We anticipate that this simplification underestimates 14  the resistance of the concentrator at the pinch region, which reduces the estimated amount of fluid removed through the side channels.  I  II  IV  R3  R3  R3  R3  III  P0 R0  P1  R1  P2  R2  P3  R1  P4  Q0  R4  Q4 R3  R3  R3  R3  Figure 2.2 Electronic-hydraulic analogy of the concentrator design.  The total amount of fluid removed by the system can be determined from the difference between volumetric flow rates Q0 and Q4. All the resistance values can be calculated using equation 2.1:  RH   12 L , wh wh3  (Eq. 2.1)  15  where  is the viscosity of the buffer fluid; L , w , and h are the length, width, and height of the channel, respectively. The pressure P0 is known and the pressures values before and after each concentrator mechanism (P1 to P4) can be calculated using the following relationships:  P4 =  R4||R3||R3  P3 R4||R3||R3  R1  (Eq. 2.2)  P3 =  (R4||R3||R3  R1)||R3||R3  P2 ((R4||R3||R3  R1)||R3||R3)  R2  (Eq. 2.3)  P2 =  [(R4||R3||R3  R1)||R3||R3  R2]||R3||R3  P1 ([R4||R3||R3  R1)||R3||R3  R2]||R3||R3)  R1  (Eq. 2.4)  P1 =  ([R4||R3||R3  R1)||R3||R3  R2]||R3||R3  R1)||R3||R3  P0 (([R4||R3||R3  R1)||R3||R3  R2]||R3||R3  R1)||R3||R3)  R0  (Eq. 2.5)  The amount of fluid removed at the ith concentrator mechanism (i = 1, 2, 3, 4) is a function of the flow resistances and pressure drops across the main channel and fluid removal channels as defined by:  Qi  Pi  Pi 1  RR   QR ,i  Pi  Ri  (Eq. 2.6)  where Qi is the combined flow rate at the output bifurcation channels; QR,i is the total flow rate at the fluid removal channels; Pi and Pi+1 are the pressures at the inlet and outlets of each concentrator mechanism; and Ri and RR are the hydrodynamic resistances of the bifurcation and fluid removal channels, respectively (Figure 2.3). Pressures values P1 to P4 are proportional to P0, so the amount of fluid removed is independent of the applied pressure P0. In other words, the system removes the same amount of fluid regardless of flow rate.  16  Pi (QR,i)/2 RR  RR Qi  Ri  Pi+1 Figure 2.3 Circuit analogy diagram for a single concentrator mechanism. The outlet bifurcation channel is shown as a single combined resistance Ri.  Maintaining a constant flow ratio of Qi/QR,i for each concentrator in the series ensures consistent retention of target cells across the entire system. A flow ratio in the range of 6 to 8 (Table 2.1) was experimentally verified to retain nearly 100% of the cancer cells which corresponds to an RR value approximately 50 times greater than Ri. At these flow ratios, the amount of predicted fluid removal ranges from 21.2% to 9.9% of the total volume from the first to last concentrator mechanism, respectively. The total amount of fluid removed is progressively less at each concentrator because the fluid volume is diminished after each, reducing the absolute amount of fluid the subsequent concentrator can remove under the same flow ratio conditions. This model predicts that the entire system is capable of removing 61.1% of the fluid passing through the device. However, since the pinch regions are not modeled, the flow rate in the bifurcated channels is higher than the real-life scenario, which contributes to an expected underestimation of the amount of fluid removed. Table 2.1 Flow ratios and overall fluid removal performance of concentrator system.  Concentrator Index (i) 1 2 3 4 Total  Flow Rate Ratio Qi/QR,i 7.4 6.7 8.2 7.8 -  % Removal Calculated 21.2 18.2 11.8 9.9 61.1  17  2.2.2  Numerical Model  To further validate the design performance prior to prototype construction, a 2D and 3D simulation are performed in CFD using COMSOL Multiphysics 4.2. The cell suspension was modeled as a homogeneous Newtonian fluid. Velocity and streamline behavior in the mechanism is confirmed visually using the 2D simulation model, shown in Figure 2.4. Velocity results for the 3D simulation are shown in Figure 2.5, and the percentage of fluid removed at each concentrator mechanism is listed in Table 2.2. The total amount of fluid removed is simulated to be 69.1 %, which is higher than the analytical model. This difference is expected since the analytical model underestimates the resistance of the concentrator at the pinch region.  (m/s) 0.01 0.009 0.008 0.007 0.006 0.005 0.004 0.003 0.002 0.001 0  Figure 2.4 Velocity plot with streamlines suggesting cell direction after the pinch region.  18  (m/s)  1E-3  9E-4  8E-4  7E-4  6E-4  5E-4  4E-4  3E-4  2E-4  1E-4  0  Figure 2.5 CFD results showing the velocity distribution inside the concentrator.  19  Table 2.2 Percentage of the total amount of fluid removed at each of the concentrator mechanisms as determined from the 3D CFD simulation.  Concentrator  I II III IV  2.3 2.3.1  Estimated Percentage of Fluid Removed [%] 30.2 20.2 11.9 6.7  Preliminary Concentrator Characterization Concentrator Design Validation  We characterized the concentrator performance at different fluid removal port resistances to determine the final design configuration. Figure 2.6 shows the design for the test concentrator, which includes the ten different resistance options for the fluid removal ports. Each device is punched with a single hole to the left and the right, with the selection of the hole corresponding to the desired resistance for a particular experiment. Using a mixture of large and small microbeads at a ratio of 1:10, approximating the size range of WBCs and UC13 cells respectively, we tested the concentrator according to the performance metrics of yield, purity, enrichment, and percentage fluid removal. The primary goal of the concentrator is to remove as much fluid as possible (>60%) and to maintain a high yield (>90%). The fraction of WBCs removed is an advantageous result of the design but is not a primary design criterion. The 2nd, 4th, 6th, and 8th punch holes are tested to generate a trend for the performance (Figure 2.7). The performance is then further validated using a real cell sample.  20  Hole #2 #4 #6 #8  Figure 2.6 Design of the concentrator device used for initial characterization. Each of the punch holes represents a different hydrodynamic resistance for the fluid removal ports.  Looking at the results (Table 2.3), we can see that the percentage of fluid removal is reduced as the resistance of the fluid removal port increases. The fluid removal fraction decreases from 83% at the 2nd to 52% at the 4th hole, which is already below the design criterion of 60% removal. This trend continues to diminish to 17% for the 8th hole. This relationship is almost inversely proportional to the yield, with the 8th punch hole retaining 100% of the target beads, while the 2nd hole retains only 23%. The enrichment and purity values also degrade with increasing fluid removal channel resistance, with the enrichment factor ranging from 6 to 2 and purity values from 38% to 17%. To achieve desired amount of fluid removal, we further investigated the performance characteristics of the 3rd punch hole. We also anticipated that using real cells would give a higher yield since UC13 cells (17.3±2.8 µm) have a larger diameter compared to the microbeads (15.5±1.5 µm), making UC13 cells more comparable to the size of the pinch constriction (18μm), which could potentially enable more selective hydrodynamic focusing. Repeating the experiment with a WBC and UC13 cell sample at the 3rd hole, the device was capable of removing 78% (+6%,-8%) of the fluid while retaining 99% of the UC13 cells. However, the enrichment performance using WBCs decreased from an interpolated value of ~5.4 from the microbead experimental results to a measured value of ~1.2 when using real cells (corresponding to 24±4% of contaminant cells removed). We think that because the size of WBCs has a larger average size and distribution (10.3±1.8 µm) compared to the small 21  microbeads used (6.33±0.37 µm), more WBCs reside within the critical boundary of the retained fraction and are carried downstream. From this result, we determined that the 3 rd punch hole, which consists of side removal channels 75 µm wide by 32 000 µm long to give the hydrodynamic resistance R3 in the analytical model (Figure 2.2) satisfies the design criteria of >90% UC13 retention and >60% fluid removal. This final configuration is capable of increasing the throughput of the entire system by an approximate factor of 5. Table 2.3 Experimental results for the different fluid removal resistances (corresponding to different punch holes)  (+/-)  Punch  Yield  Hole #  (%)  2  23  +/- 3  38  4  82  +/- 2  6  89  8  100  (+/-)  Enrichment  (+/-)  % Fluid  (+/-) %  Ratio (ER)  ER  Removal  Removal  +/- 6  6  +2,-1  83  +/- 1  33  +/- 3  5  +/- 1  52  +6,-7  +/- 4  22  +/- 3  3  +1  44  +6,-7  +/- 3  17  +/- 2  2  <1  17  +12,-15  Yield (%)  Purity (%)  Purity (%)  22  9  90% 80%  Percentage  70% 60%  Yield  8  Purity  7  % fluid removed Enrichment  6  50%  5  40%  4  Enrichment Ratio  100%  30% 3 20%  2  10% 0%  1 2  4 6 Punch-hole Number  8  Figure 2.7 Performance trend for the concentrator in showing yield, purity, enrichment ratio, and percentage fluid removal at various punch holes with greater numbered holes corresponding to higher removal port resistances.  23  Chapter 3: Ratchet Cell Sorter This chapter describes the ratchet cell sorting section of the microfluidic chip. The ratcheting funnel mechanism and the funnel shape design are described in Section 3.1. The configuration of the cell sorting funnel array is outlined in Section 3.2. The precision flow control required for robust and consistent cell sorting is explained in Section 3.3. Section 3.4 outlines the steps required for automated sorting using the ratchet sorting mechanism. This chapter concludes in Section 3.5 with preliminary characterization of the cell ratchet sorter to determine the optimal operational conditions.  3.1  Microfluidic Ratchet Mechanism  The main parameters of the funnel microstructure are the funnel depth, pore size, and the shape of the funnel taper. The funnel microstructure is 25 µm to accommodate the size of the cancer and WBCs in suspension. The pore size is defined as the distance between adjacent funnels, and the taper has a shape according to the equation y   (kx 2  W0 ) , with  k  2000 m1 chosen to provide the optimal pressure asymmetry [20]. With the assumption that a cell behaves like a liquid-drop with a constant cortical tension TC , the critical pressure required to push a cell through the funnel constriction can be modelled using the LaplaceYoung equation:   1 1  P  TC     Ra Rb   (Eq. 3.1)  Radius Ra is defined as the leading radius of the cell as the cell enters the funnel gap, and Rb is the trailing radius, as shown in Figure 3.1. The pressure required to push the cell through the constriction along the direction of the taper (forward direction) is smaller than the pressure required to push the cell in the direction against the taper (reverse direction). Therefore, under a constant oscillatory pressure application, the cells will exhibit the ratcheting effect as they undergo unidirectional translation in the forward direction [33].  24  A  B  ΔP  2W0 Rb Ra x  y  Ra  Rb ΔP  Figure 3.1 Cell deformation through the funnel constriction in (A) forward flow and (B) reverse flow.  3.2  Cell Sorting Funnel Array  The cell sorting device is composed of an array of microfluidic ratchet mechanisms that enable unidirectional transport of cells under oscillatory flow [33]. Supporting microchannels include inlet microchannels, target cell and waste outlets, and oscillation flow control microchannels (Figure 3.2). Mixed cell samples are infused from the inlet microchannel, and the separated cells are purged from the sorting region using a cross flow from the buffer inlet. The flow rate of the infusion and purging process are controlled using elongated microchannels at the inlet and outlet to provide a dominant hydrodynamic resistance to dictate the flow rate for a given applied pressure. The sorting region consists of a 2dimensional array of ratcheting funnel microstructures arranged in 32 successive rows, with 128 funnels in each row. The funnels in successive rows are aligned with each other to generate a linear streamline from the bottom to the top of the sorting region. The funnel array has a gap opening of 18 µm at the bottom row, and this size decreases with successive rows to 2 µm at the highest row. The cut-off funnel size for the separation of UC13 cancer cells and WBCs was determined experimentally to be 6 µm. Cells that cannot transit beyond the cut-off are directed to an on-chip collection reservoir, while cells that are able to move beyond the cut-off are directed into a waste reservoir (Figure 3.3). In order to sort cells effectively, the cells must be deformed through each funnel constriction with approximately the same force. To ensure an approximately equal hydrodynamic force, oscillation flow  25  through the funnel array is distributed using a tree microchannel network, which sequentially bifurcates the single entrance stream into 128 equal streams. Sample cells are introduced into the sorting region in batches with a process controlled using membrane microvalves [37]. Initially, a plug of fluid containing sample cells are infused into the bottom row of the sorting region using a horizontal flow created by opening valves V1 to V4 with V5 and V6 closed. Near the end of the infusion process, the sieve valve [38, 39] is closed to create a cell-free section prior to closing V3 to stop the inletting completely. Next, the cell separation process begins by closing valves V1 to V4, and then introducing an oscillatory vertical flow using a fluidic H-bridge created using V5 and V6. Specifically, an upward flow can be generated using V5 closed and V6 open, while a downward flow is generated with V5 open and V6 closed. The upward and downward flow rates are controlled using the pressure applied at the forward and reverse oscillation ports respectively. Finally, the separated cells are extracted using a horizontal flow introduced with valves V1 and V2 open to flush the cells into the collection outlets.  Figure 3.2 (A) Schematic of device design. (B) A heterogeneous mixture of WBC and UC13 cells enter the bottom row of the sorting region. (C) UC13 cells get trapped at the lower funnel rows before the critical 6 μm cut-off. (D) WBCs travel past the critical cut-off and are removed into a separate outlet.  26  Contaminant Cell Exit Cancer Cell Exit  6 µm cut-off  Figure 3.3 Schematic showing the separation region with the 6 µm cut-off point. Cancer cells are extracted from the bottom rows while contaminant cells are extracted from the top rows.  3.3 3.3.1  Precision Flow Control Tree Microstructure Network  A key functional requirement of this device is to push the cells through each of the 128 funnel constrictions with roughly the same amount of force, which corresponds to an equal flow profile at each of the constrictions. To divide the flow equally across 128 gaps, a tree microchannel network is designed to sequentially bifurcate the single entrance stream into 128 equal streams. The flow velocity in the sorting region is controlled predominantly by the oscillation channels which are in series with the bifurcation network. The channels are 150 µm wide by 25 µm tall by 15 400 µm long. With an applied oscillatory pressure of ~20 kPa, the flow rate is calculated by the following equation:  Q  P RH  (Eq. 3.2)  The hydrodynamic resistance of the channel, RH , is calculated using equation 2.1. The target flow rate inside the sorting region is estimated to be on the order of ~200 µm s-1.  27  3.3.2  Cell Infusion  During cell infusion, membrane microvalves [37] are used to control the flow of the cell suspension into the sorting region. The infusion consists of opening valves V1 to V4 with V5 and V6 closed. A sieve valve [38, 39] placed upstream of the microvalve which controls the cell infusion channel to prevent cell crushing through repeated actuation. The sieve valve has a geometrical dimension of 75 µm wide by 25 µm tall, which is inadequate to fully seal the flow channel upon valve actuation, shown in Figure 3.4. The gap remaining at the top of the flow channel allows fluid to transport through while blocking the cells, effectively creating a cell-free region after the sieve. An applied pressure of 170 kPa was determined experimentally to achieve the suitable membrane deflection. During automated operation, the sieve actuates for 2 seconds first, followed by actuation of the main sealing valve.  A  B  S1  Figure 3.4 (A) The sieve valve consists of a deflection membrane that does not fully close, creating a gap between the top of the membrane and the top wall of the flow channel. (B) Sieve actuation blocks the cells while allowing fluid to pass through, creating a cell free zone for the valve to actuate freely.  The flow velocity of the cells during the infusion phase is controlled using elongated microchannels at both the buffer and cell inlet/outlets. These elongated channels provide a dominant hydrodynamic resistance that reduces the sensitivity of flow due to variations in applied pressure. For experimentation, the buffer pressure is set at 210 mbar, and the cell infusion pressure is set at 115 mbar. It has been determined experimentally that small variations in the applied pressure on the range of ±5 mbar do not have a noticeable effect on 28  the flow rate, indicating that the additional hydrodynamic resistance has achieved its design intent. After whole blood processing, the sample is re-suspended in less than 100 µL of culture medium. The small volume of fluid cannot be delivered using traditional fluid handling methods that consist of millilitre-sized sample container tubes. Another disadvantage of using external fluid handling components is non-specific adhesion of cells to the inner lining of the container tubes and hoses. This sticking will inhibit cells from entering the microfluidic chip, and it is difficult to quantify the exact number of cells lost. Therefore, an on-chip cell handling solution is devised to allow effective processing of small samples that are less than 100 μL. At the cell inlet, a 6 mm diameter reservoir with a vertical dimension of 5 mm tall is created on the PDMS chip. Cells are injected into this small reservoir with a pipette and a custom-made pressure cap is press-fitted onto the top of the reservoir, as seen in Figure 3.5. The cap is attached a pressure source and pressure can be applied with the same level of precision as previously done with Falcon tubes. This design also alleviates the problem of flow blockage due to compressible air bubbles occasionally found at the interface between the fluid in the fluid delivery needle and the microchannel.  Figure 3.5 A custom-made cap which applies the pressure directly to the sample carried on the chip.  29  3.3.3  Cell Extraction  During cell extraction, membrane microvalves [37] are used to control the flow of the cells out of the sorting region and into the respective collection channels. The extraction consists of opening valves V1 and V2, while V3 to V6 remain closed. Elongated microchannels added at the buffer outlet provide a dominant hydrodynamic resistance to reduce the flow sensitivity in the vertical direction to enable a straight purge of the sorted cells into their respective outlets. At the two outlet ports, two 2.5 mm diameter reservoirs are created on the PDMS chip. The sorted cell suspension enters the reservoir from the bottom of the reservoirs, and an inverted microscope is used to manually count the sorted cells.  3.4 3.4.1  Automation Off-Chip Pressure Control  A custom pressure board is used for the pressure control of the on-chip valves and oscillation ports. The pressure board was originally designed and built by Dr. Hongshen Ma, Peter Woo, and Isaac Tang. The pressure board consists of two high pressure (0-60 psi) regulators (OMEGA, Stamford, CT) and two low pressure (0-5 psi) regulators (OMEGA). Each regulator feeds into a manifold that divides the inlet air into separate ports, which are controlled by a normally closed solenoid valve (Pneumadyne, Plymouth, MN). The solenoid valves are controlled by a microcontroller (Texas Instruments, Dallas, TX). A diagram of the setup is outlined in Figure 3.6.  30  2 4 3  1 Front  Back  Figure 3.6 The pressure board setup. (1) The high pressure regulators. (2) Low pressure regulators. (3) Microcontroller and circuit board system. (4) Manifolds along with solenoid valves.  3.4.2  Automation Software  The electronic hardware of the pressure board consists of a Texas Instrument MSP430 microcontroller attached to a circuit board that enables independent activation of each solenoid valve. The microcontroller code is written originally by Isaac Tang and Peter Woo using C, and new functions have since been added to achieve automation for the current device design. Specifically, code that allows for the simultaneous actuation of multiple solenoid valves at once was added. The revised code was uploaded to the microcontroller through Texas Instrument’s Code Composer Studio V4. Interfacial control of the system is done with a Visual Basic GUI that allows users to deliver pressure at the specified port. The original Visual Basic program was written by Isaac Tang and modified by Peter Woo. The VB program sends pre-defined signals to the microprocessor which in turn activates the desired solenoid valve(s). For this project, the VB code was updated to include an automation section, along with buttons to achieve multiple port activation for convenience. A screenshot of the automation code is shown in Figure 3.7.  31  Figure 3.7 The Visual Basic program used to perform automation and the batch cycle process, and the batch processing cycle.  For automation, key parameters were added for the user to define the automation cycle. The device operation is based on batch processing which takes around one minute to complete each cycle. The heterogeneous cell mixture first enters the device during the cell infusion phase. The system then undergoes oscillation, and the cells are irreversibly separated, and reach their respective steady state funnel rows. Cancer cells are trapped at lower funnel rows while contaminants move to the top. After the cells are sorted, they are purged from the device. A process diagram of the automation cycle is shown in Figure 3.7, and the functionality of each automation parameter is listed in Table 3.1. The throughput of the current device, denoted as the 1x device, can be calculated by the geometry of the infusion region. At a geometrical dimension of 3177 µm long by 115 µm wide by 25 µm tall, the volume of fluid that the region can intake per batch is 9.13 nL. Each batch takes approximately one minute to process, which results in approximately 0.5 µL of sample processed per hour.  32  Table 3.1 Each parameter for the automation program.  Parameter Flow Entry Time Total Oscillation Time  Definition Duration cells are injected into the device Duration cells undergo sorting  Sieving Time  Duration the sieve is actuated  Forward Flow Time  Duration cells are pushed into the funnels during oscillation Duration cells are released from the funnels in reverse during oscillation Duration sorted cells are purged from the device  Reverse Flow Time  Total Purge Time  3.5 3.5.1  Additional Notes Tailored to the speed of the flow Dependent on the amount of time required for cells to reach steady state position Dependent on the applied sieve pressure and injection flow speed Set at 4 seconds  Set at 1 second  Dependent on the speed of the purge flow  Preliminary Cell Ratchet Sorter Characterization Overall Concentration Effects  We first characterized the ratchet cell sorting device using an older version of the design to determine the concentration that gives the optimal balance between performance and throughput. The critical fundamentals of the final design (Gen6) presented in the latter portion of the thesis do not change from the design presented in this section (Gen5), so the trends from this characterization hold valid. These design changes are outlined in Section 5.4. For this characterization, we doped cancer cells into peripheral blood mononuclear cell (PBMC) samples at a concentration of 1:100, and tested the device at concentrations ranging from 2500 to 20,000 cells/µL. We evaluate the performance on the metrics of capture yield, purity, and enrichment. Initially, the sample is infused into the device and sorted using an oscillatory pressure of 40 kPa which results in an approximate speed of x μm/s for cells within the sorting area. The oscillatory flow consists of a 4 s forward pressure and a 1 s reverse pressure. After 35 s of oscillation, the sorted sample is purged from the device and counted manually in the collection reservoir using an optical microscope. Table 3.2 and Figure 3.8 show the performance across different PBMC concentrations. The results indicate that capture efficiency is independent of the contaminant cell concentration, with an average 33  capture efficiency of 93±1%. On the contrary, the purification performance of the device decreases linearly with PBMC concentration. The percentage of PBMC removed ranges from 84±1% to 76±2% from the lowest to highest concentrations respectively which corresponds to purity values of 7% (+2%,-1%) to 4±1% and enrichment ratios of 6 to 4. This reduction in purification performance occurs primarily due to non-specific adhesion of PBMCs at the funnel structures. Adhered cells act as barriers for some PBMCs to travel to higher funnel rows which reduces the selectivity during oscillation. We found that PBMC sticking worsens with increasing concentration and therefore reduces the enrichment performance. With this characterization result, we are confident that the design is capable of achieving high yield across different concentrations, whereas enrichment performance decreases with increasing concentration. Therefore, we determined that a concentration of 10,000 cells/μL is a good balance between performance and throughput, and we used this concentration for subsequent characterization experiments.  34  PBMC Concentration  YIeld (%)  (cells/µL)  Yield (%)  (+/-) Yield (%)  Purity (%)  (+/-) Purity (%)  Enrichment  (+/-)  Ratio (ER)  ER  2500  94  +/- 1  7  +2,-1  8  +/-1  5000  87  +/- 1  7  +2,-1  7  +/-1  10 000  95  <1  6  +2,-1  6  +/-1  20 000  95  <1  4  +/- 1  5  +/-1  100%  10  90%  9  80%  8  70%  7  60%  6  50%  5  40%  Yield  4  30%  Purity  3  20%  Enrichment  2  10%  1  0%  0 0  5000  10000  15000  Enrichment Ratio  Table 3.2 Results for yield, purity, and enrichment at different PBMC concentrations.  20000  Concentration of PBMCs (cells/µl) Figure 3.8 Plot showing yield, purity, and enrichment trends. Enrichment is plotted on the secondary axis (right side).  35  Chapter 4: Fabrication and Experimental Procedures The device is fabricated using standard techniques of photolithography and multilayer soft lithography. The desired pattern is first formed on a silicon wafer and used as a master for replication in polydimethylsiloxane (PDMS) silicone. The PDMS device is then treated with plasma and bonded to a glass slide. Fluid handling is controlled through means of off-chip pressure controllers as described in Chapter 3. This chapter describes the details of the fabrication, sample preparation, and experiment process.  4.1  Fabrication of Silicon Masters  Two silicon masters are fabricated, the first forming the structure of the device (known as the flow wafer) and the second defining the channels used to form valves for fluid handling (known as the control wafer). Patterns of the required features are first designed using DraftSight (Dassault Systems, Velizy-Villacoublay, France) which are then translated onto commercially printed optical photomasks. To form the flow wafer, the features of the microstructure are imprinted onto a silicon wafer using three photolithographic layers. In the first layer, the silicon wafer is coated with SU-8 negative photoresist (Microchem, Newton, MA) and spun at a speed of 3000 rpm for 30 s. This wafer is baked on a 95°C hotplate for 5 minutes followed by UV exposure through the optical photomask (Advance Reproductions, Andover, MA). The wafer is then baked at 65, 95, then 65°C for 1, 4, and 1 minutes respectively. The patterned wafer is developed using SU-8 developer (Microchem, Newton, MA) and washed with isopropanol. A second photomask (CAD/Art Services, Brandon, OR) is aligned to the first pattern and a second set of SU-8 features are patterned and developed following the same procedures. To harden the patterned SU-8 microstructures, the wafer is placed on a hot plate and the temperature is ramped from 40oC at 15 oC every 10 minutes until 165°C is reached. The wafer is then baked for 30 minutes at 165oC. After baking the wafer is ramped down at a rate of 50 oC every 10 minutes and removed when the wafer reaches 65 oC. A third layer of SPR220-7.0 photoresist (Microchem, Newton, MA) is then spun onto the wafer at 600 rpm for 50s. The edge bead is removed from the wafer manually with a clean wipe and the wafer is softbaked for 1, 5, and 1 minutes at 65, 95, and 65°C respectively. A third photomask (CAD/Art Services, Brandon, OR) is aligned to the previous set of patterns and the wafer is exposed with UV and developed using MF319 developer 36  (Microchem, Newton, MA). Finally, the wafer is baked on a hotplate at 95°C for 5 minutes to allow the SPR220-7.0 photoresist layer to reflow. The resulting height of each layer is approximately 25 μm. The control wafer is fabricated with a single layer of SU-8 using a separate patterned mask following the same SU-8 fabrication procedures.  4.2  Fabrication of PDMS Devices  The PDMS devices are fabricated using standard multi-layer soft lithography techniques [37]. The flow layer is formed using a 5:1 base to hardener ratio of Sylgard 184 silicone (Dow Coring, Midland, MI), and the control layer is formed using a 20:1 ratio. The flow and control layers are bonded together by diffusion in the oven at 65oC for 2.5 hours. The bonded flow and control layers are subsequently attached to a glass slide following 90 seconds activation in air plasma (Harrick Plasma, Ithaca, NY).  Inlets and outlets are punched  manually using 0.5, 3.5, and 5 mm punches (Harris, Redding, CA).  4.3  Sample Preparation  Experiments were conducted using UC13 bladder cancer cells and white blood cells (WBCs), or leukocytes, from healthy donors. Whole blood was drawn from healthy donors with informed consent into 6 mL EDTA blood collection tubes. Gradient density centrifugation was performed to isolate nucleated cells, which includes the entire fraction of leukocytes in the buffy coat [40, 41]. In a 15 mL centrifuge tube, 2 mL of whole blood was carefully layered over 2 mL of Histopaque 1119 (Sigma-Aldrich, St. Louis, MO). The tube was centrifuged at 400g for 40 minutes. The WBC layer was then transferred into a 15 mL tube containing 10 mL of HBSS without Ca2+ and Mg2+ solution (Invitrogen, Grand Island, NY) and centrifuged at 200g for 10 minutes. The supernatant was removed and the cells were washed again with HBSS. UC13 bladder cancer cells were cultured in MEM solution with the addition of 10% (v/v) fetal bovine serum, 1% L-glutamine, 1% MEM Non-Essential Amino Acids, 1% Sodium Pyruvate (Invitrogen), and 1% Penicillin Steptomycin (Fisher Thermo Scientific, Waltham, MA), and incubated at 37oC in a humidified environment with 5% CO2. For doped experiments, the number of WBCs and UC13s were first counted using a hemocytometer, and the desired number of cells of each phenotype added to a tube (to give the desired WBC:UC13 fraction). The tube was then centrifuged and the supernatant was 37  removed; the cells were re-suspended in the correct amount of MEM media (to yield the desired concentration) with 5% bovine serum albumin (BSA), 15% w/v Ficoll PM400 (Sigma-Aldrich, St. Louis, MO). 0.2% Pluronic (Invitrogen) was used for the latest generation (Gen6) of experiments. 15% Ficoll PM400 increases the density of the media to keep the cells in suspension while BSA and Pluronic prevent nonspecific adsorption of cells to the microstructure [42]. Cancer cells were measured to have a size of 17.3±2.8 µm, and WBCs were measured to be 10.3±1.8 µm. Cancer cells were stained with calcein AM (Invitrogen) and WBCs were stained with Hoechst 33342 (Invitrogen). Microparticles (Bangs Laboratories, Fishers, IN) of sizes 15.5±1.5 µm and 6.33±0.37 µm were used for particle experiments. Microparticles were suspended in deionized water with 15% Ficoll PM400 with 0.5% Tween 20 (Invitrogen).  4.4  Experimental Setup and Preparation  Fluid handling for the experiment is performed using a commercial pressure controller (Fluigent, Paris, France), and the custom made pressure board and software described in Section 2.1. Microscopy is performed using an inverted microscope (TS-100, Nikon, Tokyo, Japan), and a CCD camera (DS-2MBW, Nikon, Tokyo, Japan). For each experiment, the PDMS microfluidic devices were filled with MEM culture media with 15% Ficoll PM400 and 5% BSA (and 0.2% Pluronic for Gen6 experiments), then incubated for 30 minutes to prevent non-specific adsorption at the channel walls. The initial count for the number of UC13 and WBCs is determined manually using a hemacytometer. After separation, PBS is added to the collection reservoirs to dilute the Ficoll concentration to allow cells to settle at a single focal plane at the bottom of the reservoir. A microscope with encoded stage is used to take fluorescence images of the collection reservoir with a 4X objective. The resulting images are stitched together using Microsoft Image Composite Editor (Figure 4.1). The number of target and contaminant cells is then counted manually from the composite images. The performance metrics of yield, purity, and enrichment is calculated from the count values.  38  Figure 4.1 Composite image of a portion of the outlet after sorting. Cancer cells fluoresce green (A) while both cancer cells and WBCs fluoresce blue (B). Individual images were taken with a monochrome camera, stitched using Microsoft Image Composite Editor and artificially colored in Adobe Photoshop.  39  Chapter 5: Design Iterations This chapter describes the different design iterations from Generations 3 to 6. Generation 3 was originally designed by Sarah McFaul. I am primarily responsible for the design, fabrication, and testing of all subsequent designs, denoted as Generations 3, 3.5, 4, 5, and 6. Table 5.1 summarizes the features found in each design, with each described in detail in the rest of the chapter.  Table 5.1 Outline of the design changes made throughout the different generations  Generation 3.5  Generation 4  Generation 5  Generation 6  Sorting Region   Same as Gen 3               Hydrodynamic resistance added for cell & buffer inlet Improved valve sealing Separate cell & buffer channel valves Separate exit ports    Same as Gen 3             Increased rows  Same as Gen 4 from 12 to 16 Tailored funnel gap sizing to isolate UC13 cancer cells Funnel length reduced by 33% Rounded funnel trailing end Plumbing (aside from Sorting Region)    Increased rows from 16 to 32  More modular  Increased design spacing at cell outlet ports Increased hydrodynamic  Tailored resistance at cell & collection of buffer inlets and target cells outlets Improved valving Tailored collection of target cells Addition of imaging area Macro-scale Fluid Handling    Separate hydrodynamic resistance added for each funnel row pair Revised tailored collection of target cells  Use of bottom feed tube Use of shaker Addition of sieve     On-chip sample storage On-chip concentrator      Same as Gen 5  40  5.1  Generation 3 to Generation 3.5  The previous generation device, denoted as Generation 3, was originally designed by Sarah McFaul to demonstrate the isolation of cells of different phenotypes based on cellular differences in size and rigidity [20]. A schematic of the Generation 3 device is outlined in Figure 5.1.  Figure 5.1 Schematic of the Generation 3 device. Valves are denoted by the green areas, and flow channels denoted in gray ([20] - Reproduced by permission of The Royal Society of Chemistry).  During Generation 3 device operation, cells were first infused through the cell inlet with buffer simultaneously. After a sufficient heterogeneous cell sample was loaded, valves V1 and V2 were actuated to enclose the cells in the separation area. Pressure was applied in the forward and reverse oscillation channels, with oscillatory flow created by the offset actuation between valves V3 and V4. Forward flow was achieved by opening V4 while having V3 closed and reverse flow was achieved by opening V3 while closing V4. Oscillation consisted of 3 seconds of forward flow to push cells through the funnel constrictions, followed by 1 second of reverse flow to release the cells from the funnels. In Generation 3, the sorting process was controlled manually using an older version of the Visual Basic program interface.  5.1.1  Sorting Region  The sorting region remained the same from Generation 3 to 3.5, as the same optical mask was used. We used SU-8 photoresist to create the sorting region microstructures. Different 41  variations of the device were fitted onto a single silicon wafer, each with differences in the funnel size distribution. For Generation 3.5, we did characterization tests to determine the funnel sizes suitable for trapping different cell types.  5.1.2  Plumbing  Several changes in the plumbing were made to Generation 3 to solve some issues found in the previous design, as shown in Figure 5.2. First, the valves that seal the cell and buffer inlet and outlet channels were found to be inadequate to provide a tight seal. The valves were 200 µm wide and the flow channels were 100 µm wide, resulting in a valve overlap of 200 x 100 µm. It was determined experimentally that this valve overlap was inadequate, as a large valve pressure exceeding 210 kPa was required to fully seal the channels. This high pressure occasionally caused the valve membrane to rupture. This configuration was also incapable of performing independent purging after separation because with V1 and V2 opened, a new batch of cells would enter and subsequently exit the sorting region prematurely without being sorted. For Generation 3.5, two independent valves were placed at the flow channels before and after the bifurcation channels. The flow channel and valve intersection areas were increased to 200 µm x 200 µm to improve sealing. However, during the sorting operation it was discovered that without the valves blocking each individual bifurcation channel, cells would enter the bifurcation channels to bypass the sorting region during oscillation. It was also discovered that the buffer and cell channels were very sensitive, and fine adjustments on the order of ±0.1 mbar had to be made for each experiment to achieve the desired flow speeds. This sensitivity also had an adverse effect on the flow direction of the cells during post-separation purging. During purging, a 1 s forward and reverse oscillation was applied to dislodge cells from the funnels. A non-linear purge caused cells to transverse between rows to undo separation. We tackled this issue by adding extra hydrodynamic resistance. However, preliminary experiments revealed that although the flow sensitivity was reduced, it was insufficient to achieve a consistently straight purge. Lastly, in Generation 3 all the funnel rows converged into one outlet. This arrangement did not allow the independent extraction of different cell phenotypes. To achieve independent collection, the exit ports were separated in Generation 3.5.  42  Generation 3  Generation 3.5  A) Increased valve area C) Separate Exit Ports  B) Increased hydrodynamic resistance  Figure 5.2 CAD drawings of Generation 3 and 3.5 devices. Most significant functional changes include A) increasing the valve cross sectional area, B) increasing the hydrodynamic resistance at the cell and buffer inlets, and C) separating the exit ports.  43  5.1.3  Macro-scale Fluid Handling  Aside from non-functional channel routing changes to accommodate the key changes in the new design, the fluid and cell handling methods used for Generation 3.5 remained the same.  5.2 5.2.1  Generation 3.5 to Generation 4 Sorting Region  The sorting region was tailored to accommodate isolation of cancer cells. First, the sorting area was elongated from 12 to 16 rows to provide a spatial buffer between the cancer and waste collection rows to achieve higher isolation purity. The gap sizing increments for the Generation 3.5 device was linear for cell characterization purposes, whereas in Generation 4 the funnel gap sizing was changed to allow the targeted isolation of UC13 cells. The comparison in the distribution of funnel gap sizes between Generations 3.5 and 4 are shown in Table 5.2. Characterization experiments performed using Generation 3 devices found PBMCs to be trapped mainly at the 6 µm funnel constriction [20], whereas cancer cells were trapped at funnels larger than 15 µm. With this data, a 7 µm funnel row cut-off was selected for Generation 4.  Table 5.2 The funnel gap size distribution for Generation 3.5 and two variations of Generation 4  Bottom  Top (for Gen 3.5)  Top (for Gen4)  Funnel Gap Sizing (µm) Generation 3.5 Generation 4 15 20 14 18 13 16 12 15 11 14 10 12 9 11 8 10 7 7 6 6 5 6 4 5 5 4 3 2  44  The funnel shape was also changed in Generation 4 to facilitate easier manufacturability. In Generation 3.5, it was observed during experiments however that cells only contact approximately the top 1/3 of the funnel upon entry. The fabrication of long funnels was also difficult, so we reduced the funnel length by 33%. This reduction in aspect ratio minimizes the chances of structural distortions due to thermal stresses during the hard bake process. In addition, the trailing ends of the funnels were also rounded to allow gentle handling of the cells during sorting. The changes in funnel configuration are shown in Figure 5.3.  Generation 3  Generation 4  Figure 5.3 Generation 4 funnels are reduced by 33% with rounded trailing ends to facilitate manufacturability and improve cell handling.  5.2.2  Plumbing  Numerous plumbing changes were made to Generation 4 to resolve the issues found in Generation 3.5, as shown in Figure 5.4. First, to prevent cells from bypassing the funnel rows through the buffer bifurcation network, we placed a long valve to seal each bifuricated channel at the inlet and outlets. A 150 x 200 µm valve overlap was utilized with satisfactory results. However, it was discovered that during prolonged experiments, repeated actuation of the cell inlet valve crushed cells between the valve and channel ceiling. As a result, cell debris accumulated over time to eventually clog the flow channel. To alleviate this problem, a sieve with a 75 x 200 µm overlap with the flow channel was placed upstream of the cell inlet valve to create a cell-free zone prior to valve actuation. The buffer and cell channels remained sensitive to input pressure variations even with added hydrodynamic resistance as seen in Generation 3.5. To further reduce flow sensitivity, 45  more hydrodynamic resistance was added at both the inlets and outlets. This increase in hydrodynamic resistance successfully reduced flow sensitivity to ±5 mbar during cell infusion and purging. The sorted cells are collected in two separate reservoirs, with the cut-off set at the 7 µm funnel row. An imaging area was also added to the design to allow real-time counting of the sorted cells. However, due to technical computing difficulties encountered during real-time imaging, this feature was removed in the next generation design.  46  Elongated Sorting Area Further Increased hydrodynamic resistance  Improved valving Added Imaging Area  Further Increased hydrodynamic resistance  Added Sieve  Tailored Collection (Cut-Off at 7 um)  Figure 5.4 CAD drawing of Generation 4 device. Outlined are all the physical changes made from Generation 3.5.  47  5.2.3  Automation Changes  In Generation 3.5 experiments, we applied an oscillation of of 1s forward and reverse flow during purging to dislodge cells from the funnels. However, it was first discovered that cells at the top left and bottom right corner of the sorting region were not being released from the funnels, leading to cell buildup over time. To investigate into this issue, we modeled the system using a simplified model in computational fluid dynamics to understand the flow dynamics during oscillatory purge. The bifurcation channels were modeled as thin channels, and funnels were modeled as thin rectangular elements. The number of funnels and rows was reduced to speed up simulation time. Figure 5.5 shows the vertical flow velocity component during a downward oscillation while applying a cross flow from the buffer inlet. From the results, one can see that flow is almost stagnant at the top left and bottom right corners, while the flow is high at the bottom left and top right corners. This flow profile is inverted during the upward oscillation cycle, where the top left and bottom right corners experience high vertical flow. Cells in the top left and bottom right corners are not cleared from their respective funnels during purging, leading to cell accumulation and eventual clogging. resolve this issue, the oscillation component was removed during purging. Instead, the total oscillation time was set to end on a reverse flow cycle (to push cells out of their respective funnels), followed by a cross flow. This setting ensured a straight purge of the cells into the correct collection outlets.  48  Figure 5.5 CFD analysis showing the vertical flow velocity during the downward cycle of an oscillatorypurge.  5.2.4  Macro-scale Fluid Handling  Due to the small volume of the sample processed during this generation of experiments (~200 µL), it was not preferable to use the conventional 15 mL Falcon tube (BD Biosciences, East Rutherford, NJ) because it was difficult to aspirate the sample without creating air bubbles in the Tygon tubing. As a result, a 2 mL pierce centrifuge column (Fisher Thermo Scientific, Waltham, MA) was used with success. The centrifuge columns included a filter at the bottom of the tube that had a ~30 µm pore size. This was potentially problematic for our application, so the filter on each column was removed prior to experimentation. A blunt tip Luer-LokTM needle (BD Biosciences) was press fitted onto the bottom of the column and the Tygon tube slid over the needle tip. The column had a smaller diameter compared to the traditional 15 mL falcon tubes, which enabled easier small volume fluid manipulation. The bottom-feed design eliminated the air bubble issue, and also enabled samples to be added during the middle of an experiment without the need to re-flush the line. A picture of the setup is shown in Figure 5.6.  49  A)  B)  Figure 5.6 A) The bottom-feed tube. B) The entire setup.  5.3 5.3.1  Generation 4 to Generation 5 Sorting Region and Concentrator  The sorting region remained the same from Generation 4 to 5, as the same optical mask was used. From Generation 4, experiments revealed the 7 µm cut-off was too conservative and many contaminants were entering the cancer collection ports. Therefore, the cut-off was revised to 14 µm in Generation 5. This change however did not require a new mask for the sorting region because we revised the plumbing layer to achieve the same functionality. We added the hydrodynamic concentrator to this design generation. The concentrator consists of an inlet reservoir, followed by a sieve and inlet valve. During operation, the cells are infused through the concentrator directly into the sorting region during each batch cycle. This operation is tricky and requires delicate coordination, so we opted for a modular design in the next generation.  5.3.2  Plumbing  A diagram of the layout for the Generation 5 design is shown in Figure 5.7. There have been small changes made to the SPR plumbing layer. The cut-off was changed to the 14 µm funnel  50  row, and the separate outlets now direct the cells into adjacent reservoirs to facilitate postseparation analysis.  5.3.3  Macro-scale Fluid Handling  Further improvements were made in blood pre-processing during the experimentation stage for Generation 5, so the sample could be reduced to approximately 20 µL. This sample volume enabled a realistic processing time, but was extremely tricky to handle. The pierce centrifuge column used for Generation 4 devices could not be used since 20 µL could barely fill up the volume inside the Tygon tube. To solve this, an on-chip cell sample storage solution was incorporated into Generation 5, as described in Section 3.3.2.  51  WBC Outlet Reservoir  UC13 Outlet Reservoir  Tailored Collection (Cut-Off at 12 um)  Cell Inlet Reservoir  Concentrator Mechanism  Figure 5.7 CAD layout of the Generation 5 device. The more significant change from Generation 4 is the addition of a cell concentrator.  52  5.3.4  Imaging  Two 3.5 mm holes were punched at the cell outlet ports, and the cells accumulate in these reservoirs over the duration of the experiment. After the experiment, PBS was injected into the reservoirs to dilute the Ficoll PM400 to allow the cells to settle to a single focal plane at the bottom of the well. Approximately 10 pictures were taken in a grid-like sequence for each reservoir, and the pictures were stitched together using Microsoft Image Composite Editor to form the entire image of the reservoir. Target and contaminant cells were then counted manually.  5.4 5.4.1  Generation 5 to Generation 6 Sorting Region and Concentrator  We designed a new chrome optical mask for the Gen6 design, which separates the concentrator and sorter into two modular mechanisms with an intermediate reservoir in between (Figure 5.8). The sorting region now has 32 individual funnel rows, with the spatial distribution ranging from 18 µm at the lowest rows to 2 µm at the highest. The design intent for Gen6 is to improve the enrichment performance by creating a large spatial buffer between the WBC and UC13 collection ports, and also by reducing the cut-off (6 μm). Creating a larger spatial buffer reduces the likelihood of WBCs from entering the cancer cell collection port if there are discrepancies in the flow causing the WBCs to skew during purging. The reduced cut-off size also allows the application of a higher oscillation pressure to drive WBCs higher up in the sorting region without a high risk of losing cancer cells, hence creating a more effective separation.  5.4.2  Plumbing  SU-8 photoresist is used for the majority of the microstructures in Gen6 primarily due to its ease of development during microfabrication, as shown in Figure 5.8. Only microvalve areas are made from SPR photoresist to facilitate proper valve sealing.  53  WBC collection ports  UC13 collection  Figure 5.8 CAD layout of the Generation 6 device. The green and blue areas are SU-8, and the red areas are SPR.  54  5.4.3  Macro-scale Fluid Handling  In previous generation experiments cells were suspended in MEM media and 5% BSA. Using the media with BSA alone, we observed cell sedimentation over time. This was problematic as the experiment would not give consistent data that could be normalized by time. To solve this problem we added 15% Ficoll PM400 to increase the density of the suspension fluid to keep the cells evenly distributed within the media. We have validated that the concentration of cells remain fairly consistent over time (Figure 5.9). In addition, we added 0.2% Pluronic F-127 to the suspension media to further reduce nonspecific adhesion of cells to PDMS channel walls.  16000  PBMC concentration (cells/μl)  14000 12000 10000 8000 6000 4000 2000 0 0  10  20  30  40  50  60  70  Running Time (minutes) Figure 5.9 Cell concentration over time when suspended in MEM and 15% Ficoll.  55  Chapter 6: Cell Separation Results and Discussion This chapter reports the experimental characterization of the Generation 5 and 6 device to separate UC13 cells from WBCs. Section 6.1 describes the metrics used to characterize the performance of the cell separation platform along with the methods to estimate of uncertainty of the results. Section 6.2 presents the characterization of the hydrodynamic concentrator. Section 6.3 describes the characterization of the ratchet cell sorter at different cancer cell doping concentrations for both Generations 5 and 6. The hydrodynamic concentrator and ratchet sorter are then linked together in tandem and the performance of the entire device is characterized in Section 6.4. A comparison of our system with other microfluidic platforms, along with our system’s practical advantages, is presented in Section 6.5.  6.1  Performance Characterization and Uncertainty Estimation  The performance of cell separation devices is evaluated using 1) the capture yield (Eq. 6.1), which is the percentage of target cells retained by the device; 2) the enrichment ratio (ER, Eq. 6.2), which is the ability of the device to amplify the concentration of target cells relative to contaminant cells; and 3) purity (Eq. 6.3), which is the fraction of target to contaminant cells in the outlet. These performance metrics are used frequently in the field as a standard methodology in assessing the performance of a rare cell separation device [4-6, 8-10, 20, 30, 32, 43, 44]. Specifically for our experiments, these metrics are defined as:  Yield =  ER =  UC13 final UC13initial  *100%  UC13/WBC final UC13/WBCinitial  UC13   Purity =   *100%  UC13 + WBC  final  (Eq. 6.1)  (Eq. 6.2)  (Eq. 6.3)  56  Prior to the experiment, UC13 cells are stained using the green fluorescence dye calcein (Invitrogen), while WBCs are stained using the blue fluorescence dye Hoechst 33342 (Invitrogen) to facilitate counting. The initial count for the number of UC13s (UC13 initial) and WBCs is determined manually using a hemacytometer. After separation, PBS is added to the collection reservoirs to dilute the Ficoll concentration to allow the cells to settle to a single focal plane at the bottom of the reservoir. A microscope with an encoded XY stage is used to take fluorescence images of the collection reservoir with a 4X objective, and the resulting images are stitched together using Adobe Photoshop. The number of UC13s (UC13final) and WBCs (WBCfinal) is counted manually from the composite images. The WBC and UC13 count is used to calculate the performance yield, purity, and enrichment for each experimental trial. The uncertainty of our characterization method arises from the hemocytometer count uncertainty and manual count uncertainty. The hemocytometer count uncertainty is typically considered to be 15% [45]. The manual count uncertainty arises from the heterogeneous nature of cells and their response to staining. As shown in Figure 6.1, some cells do not fluoresce at the same intensity as others when observed under identical conditions using a microscope. These differences likely results from differences in cell morphology or cytochemistry. Additionally, reduce fluorescence occurs for a small fraction of WBCs because differences in buoyant mass of these cells prevent them from settling to the same focal plane as other cells at the bottom of the reservoir. While counting for each cell type, we estimate the count error using the number of cells that have an uncertain phenotype, which is denoted as UC13uncertain and WBCuncertain. After quantifying the uncertainty, we calculated the high and low limits for yield, purity, and enrichment using equations 6.4 to 6.11. Specifically, when calculating the upper limit of the cell count we add one-half of the uncertainty value to the counted value in the numerator while subtracting one-half of the uncertainty value from the denominator. In calculating the lower limit, we did the opposite by subtracting one-half of the uncertainty value in the numerator while adding one-half of the uncertainty value in the denominator. If identical quantities appear in both the numerator and denominator (e.g. Eq. 6.5), the same uncertainty calculation as performed in the numerator is repeated for the measured value in the denominator. We then determined the differences between the high/low values and the 57  calculated value to estimate the corresponding error. A sample calculation is outlined below (Table 6.1), with the data obtained from one of the Generation 6 ratchet sorter characterization experiments.  A  B  Figure 6.1 Brightfield and fluorescent images of two different UC13 cells. In A) the cell has an irregular shape and a weak fluorescence signal. In B) the cell has a circular morphology and a strong fluorescence. The cell found in A) is counted as an uncertain cell (photo courtesy of Will Beattie).  Table 6.1 Sample data from a Generation 6 ratchet sorter experiment for the 1:600 doping ratio.  Collected  Total  (+/-)  WBC  (+/-)  Yield  UC13  UC13  UC13  contaminants  WBC  (%)  (%)  (%)  27  28  3  25  3  96  97  96  Purity  Puritymax Puritymin  (%)  (%)  (%)  52  56  49  (+)  (-)  Purity  Purity  (%)  (%)  4  3  Yieldmax Yieldmin  Enrichment Ratio (ER) 617  ERmax  ERmin  717  571  (+)  (-)  Yield  Yield  (%)  (%)  0.2  0.2  (+)  (-)  ER  ER  100  46  58  High limit of yield: For ratchet cell sorter & combined system:  Yield max  1 UC13 final  UC13uncertain 2 = *100% 1 UC13total  UC13uncertain 2  (Eq. 6.4)  Note: Since a relatively small fraction of the UC13 cells are lost, the estimated uncertainty for the UC13total is identical to the estimated uncertainty for UC13final.  For concentrator characterization:  Yield max =    UC13 final   1    UC13 final  UC13uncertain  2  Collected *100% 1 1     UC13uncertain  +  UC13 final  UC13uncertain  2 2 Collected   Lost  (Eq. 6.5)  Low limit of yield: For ratchet cell sorter & combined system:  Yield min  1 UC13 final  UC13uncertain 2 = *100% 1 UC13total  UC13uncertain 2  (Eq. 6.6)  59  For concentrator characterization:  Yield min  1    UC13 final  UC13uncertain  2  Collected = *100% 1 1     +  UC13 final  UC13uncertain   UC13 final  UC13uncertain  2 2  Collected   Lost  (Eq. 6.7)  High limit of purity:  Purity max  1   UC13+ UC13uncertain   2 =  *100% 1 1  UC13+ UC13uncertain + WBC  WBCuncertain   2 2  final  (Eq. 6.8)  Low limit of purity:  Purity min  1   UC13  UC13uncertain   2 =  *100% 1 1  UC13  UC13uncertain + WBC+ WBCuncertain   2 2  final  (Eq. 6.9)  High limit of enrichment:  Enrichment max  1 1   (UC13+ 2 UC13uncertain ) / (WBC  2 WBCuncertain )  final =  UC13/WBCinitial  (Eq. 6.10)  Low limit of enrichment:  Enrichment min  1 1   (UC13  2 UC13uncertain ) / (WBC  2 WBCuncertain )  final = UC13/WBC  initial  (Eq. 6.11) 60  6.2  Hydrodynamic Concentrator  6.2.1  Cell Concentration Optimization  We initially characterized the concentrator across different cell concentrations to determine the optimal operating conditions using UC13 cancer cells as target cells and peripheral mononuclear blood cells as a model of the contaminant cells. We spiked 100 cancer cells into each microliter of the sample in each trial, and we tested PMBC concentrations from 500 to 40,000 cells/µL. Table 6.2 and Figure 6.2 outline the retention of the target cancer cells across different PBMC concentrations. With increasing concentration, the retention decreases from 97±1% to 79±2% in the highest to lowest concentration cases respectively. This trend is expected since we have observed an increase in non-specific adhesion of PBMCs onto the PDMS channel walls at higher concentrations. The build-up of PBMCs along the channel walls can adversely influence the flow trajectory of incoming cancer cells, causing some of them to be “bumped into” the waste removal channels. The results indicate a steep drop in the retention performance at concentrations higher than 10,000 cells/µL. For this reason, we determined that a concentration below 10,000 cells/µL is optimal for this mechanism to achieve high capture efficiency.  Table 6.2 Concentrator yield performance at different PBMC concentrations.  PBMC  Collected  Collected  Lost  Lost  UC13  UC13uncertain  UC13  UC13uncertain  500  430  40  15  8  97  +/-1  1000  228  40  21  10  92  +2,-3  5000  364  40  34  10  91  +/-2  10 000  500  100  34  10  94  +1,-2  20 000  261  40  43  10  86  +/-2  40 000  421  40  113  20  79  +/-2  Concentration (cells/µL)  Yield (%)  +/- Yield (%)  61  UC13 Yield (%)  100% 95% 90% 85% 80% 75% 70% 65% 60% 55% 50% 45% 40% 35% 30% 25% 1  10  100  1,000  10,000  100,000  Total PBMC Concentration (cells/µL) Figure 6.2 Results showing UC13 cell retention at different PBMC concentrations. The first data point at 1 cell/µL acts as a control for UC13 retention when there are no PBMCs present. Plot is shown in logarithmic scale.  It is worthwhile to note that in this set of experiments, we did not yet use 0.2% Pluronic (Invitrogen), so there was more non-specific adhesion of the cells at the PDMS walls. Figure 6.3 shows a comparison between a section of the concentrator processing a sample of the same cell concentration with and without Pluronic. In the case without Pluronic, cell buildup started to occur after 15 minutes of device operation, especially at areas with increased probability of non-specific adhesion (Figure 6.3A). In experiments with Pluronic, we observed very minimal build-up even after one hour of continuous device operation at the same flow rate. Therefore, Pluronic was used for all subsequent experiments.  62  A  Flow Direction  Cell Build-Up  B  Zero Cell Build-Up  Figure 6.3 Fluorescent images of one section of the concentrator showing non-specific adhesion of WBCs onto the PDMS walls. A) Image taken after 15 minutes of operation when no Pluronic is used. This image was taken after the sample has been flushed through the concentrator, which means that only WBCs adsorbed to the PDMS walls are shown. B) When Pluronic is used, there is minimal non-specific adhesion even at the stagnation point. This image was taken while cells were flowing through the concentrator after one hour of processing.  63  6.2.2  Hydrodynamic Concentrator Performance Characterization  The performance of the cell concentrator was characterized experimentally using a mixture of WBCs and UC13 cells at a ratio of 10:1 and at a WBC concentration of 2500 cells/μL with the addition of 0.2% Pluronic. We expect the yield to be independent of the doping concentration, because the pinch region focuses the UC13 cells regardless of the total number of cells. After experimentation, we found the hydrodynamic concentrator to be capable of retaining 99% of the cancer cells, while removing 78% (+6%, -8%) of the fluid and 24±4% of the contaminant WBCs. This yield is higher compared to the trend in the previous results when no Pluronic was used, which is congruent to our predictions. The percentage fluid removal measured is slightly greater than the modeling prediction. This discrepancy is likely a result of errors in the estimate of hydrodynamic resistance of the microchannels.  6.3  Ratchet Cell Sorter  To sort the cell sample using the microfluidic ratchet mechanism, the sample is first infused into the device and an oscillatory flow is applied in the direction of the ratchet. The oscillatory flow is biased towards the direction of the taper, where flow along the direction of the taper is applied for 4 s, and against the direction of taper for 1 s. After 7 cycles of oscillation, the sample is separated with WBCs transported to the top of the funnel array (Figure 6.4A-C) and UC13 cells retained in the lower rows (Figure 6.4D-F). Following separation, the cells are purged from the device into their respective collection reservoirs, and then counted manually using an optical microscope at the end of each experiment.  64  Figure 6.4After a sorting cycle, smaller, more deformable WBCs stained in blue travel to the higher funnel rows shown in A) brightfield, B) green fluorescence, and C) blue fluorescence. Larger and more rigid UC13 cancer cells are retained near the bottom, shown in D) brightfield, E) green, and F) blue.  6.3.1  Generation 5 Ratchet Sorter Performance Characterization  We evaluated the performance of the Generation 5 device using cell samples with UC13 cells spiked into WBCs at ratios of 1:10, 1:100, 1:500, and 1:1000. As shown in Table 6.3 and Figure 6.5, the capture yield is consistent with an average of 93% across the range of cancer cell concentrations, and appears to be independent of the UC13 cell concentration. Although the yield performance is higher than most other methods [9, 15, 30, 32, 46], we observed that some of the smaller cancer cells can transverse past the critical cut-off and are removed through the waste outlet. Therefore, we predicted that a smaller pore size for the cut-off could further increase the yield. On the other hand, the enrichment performance of this device was poor and inconsistent. An average enrichment ratio of 5 was not representative of the high-selectivity capability of the ratchet mechanism. From the initial experiments, we realized that due to the large funnel size of the cut-off (14 µm), we were not able to apply sufficient oscillation pressure to enable effective ratcheting of all the WBCs, in particular the larger ones. An increase in oscillation pressure would enable all the WBCs to ratchet, at the expense however of losing smaller UC13s. Therefore, it was challenging to select an optimal  65  pressure that preserved the yield that also removed sufficient WBCs to obtain a consistent and highly pure sample at the collection outlet. In addition, the cell sorting area in the Generation 5 device had 16 closely spaced funnel rows, where the spatial distance between the WBC and UC13s after separation was not very large. Therefore, any skewing of the extraction flow during cell purging resulted in WBCs entering the incorrect outlet.  Table 6.3 Yield and enrichment performance for different doping ratios of UC13:WBC for the Generation 5 ratchet sorter device. Doping  UC13  UC13  (+/-)  WBC  (+/-)  Yield  Collected  Lost  UC13  Contaminants  WBC  (%)  1:10  359  20  10  1166  200  95  1:100  61  3  4  1003  340  1:500  22  5  4  1214  1:1000  5  0  2  2487  Concentration UC13:WBC  A  (+/-)  Enrichment  (+/-)  Ratio (ER)  ER  <1  3  <1  95  <1  6  +/-1  200  81  <1  9  +2,-1  600  100  <1  2  +/-1  Yield (%)  B 100  15  E n r ic h m e n t R a t io  Y ie ld ( % )  80  60  40  20  10  5  0  0 1  :1  0  0  0 :5 1  1  :1  :1 1  0  0  0  0 0 :1 1  0  0  0 0 1  :5  0 :1 1  1  :1  0  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  Figure 6.5 A) Results showing average capture yield of 93% for different UC13 doping concentrations. B) Results showing enrichment for the four UC13 doping concentrations.  66  6.3.2  Generation 6 Ratchet Sorter Performance Characterization  We characterized the improved version of the ratchet cell sorter in the Generation 6 device performance using cell samples with UC13 cells spiked into WBCs at ratios of 1:10, 1:100, and 1:600. As shown in Table 6.4 and Figure 6.6, the capture yield is consistent across these experiments, with an average of 96%. Enrichment is also relatively consistent at doping ratios of 1:100 and 1:600. At the doping ratio of 1:10 however, the enrichment was found to be only 59x (+9, -3). This degradation in performance likely results from WBCs occasionally becoming trapped behind UC13 cells at the entrance to a pore, preventing the WBCs from moving up the funnel rows when high concentrations of UC13 cells are present. Therefore, at doping ratios of 1:100 or smaller, device performance is highly selective and independent of the UC13 cell concentration, which shows promise in the detection and enrichment of rare cells at diminishing concentrations.  Table 6.4 Yield and enrichment performance for different doping ratios for the Generation 6 ratchet sorter device Doping  UC13  UC13  (+/-)  WBC  (+/-)  Yield  Collected  Lost  UC13  Contaminants  WBC  (%)  1:10  164  8  3  28  5  95  1:100  97  3  2  19  5  1:600  27  1  3  25  3  Concentration UC13:WBC  (+/-)  Enrichment  (+/-)  Ratio (ER)  ER  <1  59  +9, -3  97  <1  511  96  <1  617  Yield (%)  +102, -50 +100, -46  67  A  B  100  800  E n r ic h m e n t R a t io  Y ie ld ( % )  80  60  40  20  0  600  400  200  0  1  :1  0 1  :1  0  0 1  :6  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  1  :1  0 1  :1  0  0 1  :6  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  Figure 6.6 A) Results showing capture yields of above 95% at various UC13 doping concentrations. B) Results showing enrichment for the three UC13 doping concentrations, with poorer performance recorded in the 1:10 case.  6.4 6.4.1  Combined System Characterization Generation 5 Combined System Performance  We first evaluated the performance of the complete Generation 5 system consisting of the hydrodynamic concentrator and ratchet sorter. The sample was prepared at a concentration of 2500 WBCs per microliter, which is concentrated to approximately 10,000 WBCs per microliter using the concentrator. The sorting device then processed the remaining sample using operational parameters optimized previously. The effective yield for the combined system is calculated as the product of the yield of the concentrator and the ratchet sorter. The average effective yield for the combined system is 93% (Table 6.5, Figure 6.7A). The complete system shows an enrichment ratio of 2±1 and 6 (+1, -2) for doping ratios of 1:500 and 1:1000 UC13 cells to WBCs, respectively. This yield and enrichment performance results in a final cancer cell purity of less than 1% for both doping concentrations. This enrichment performance is poor and a final purity <1%, which is too low to be useful for applications such as circulating tumor cell separation.  68  (+/-) WBC  WBC Contaminants  (+/-) UC13  Sorter  UC13 Lost  Sorter  UC13 Collected  (+/-) UC13  Concentrator  UC13 Lost  Concentrator  UC13 Collected  (UC13:WBC)  Concentration  Doping  Table 6.5 Yield, purity, and enrichment performance for the Generation 5 combined system.  1:500  74  1  1  15  0  4  3103  800  1:1000  74  1  1  8  1  2  1441  300  Conc.  (+/-)  Sorter  (+/-)  Total  (+/-)  Yield  Yield  Yield  Yield  Yield  Yield  (%)  (%)  (%)  (%)  (%)  (%)  1:500  99  <1  100  <1  99  <1  0.5  1:1000  99  <1  89  +/-1  88  +/-1  0.6  UC13: WBC  Purity (%)  (+/-) Purity  ER  (%)  +/-0.1 +0.2, -0.1  (+/-) ER  2  +/-1  6  +1,-2  69  B  A  100  E n r ic h m e n t R a t io  8  Y ie ld ( % )  80 60 40 20  6  4  2  0  0  1  :5  0  0 1  :1  0  0  0  1  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  :5  0  0 1  :1  0  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  C  10  P u r it y ( % )  8 6 4 2 0  1  :5  0  0 1  :1  0  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  Figure 6.7 Results showing capture yield at different UC13 doping concentrations. The yield remains consistent, even in diminishing UC13 concentrations. B) Plot for enrichment and (C) purity, Note the vertical axis in plot C) only extends to 10% to facilitate the data range.  6.4.2  Generation 6 Combined System Performance  Finally, we evaluated the performance of the complete Generation 6 system combining the ratchet cell sorter and the hydrodynamic concentrator. We utilized the same sample preparation parameters and experimental procedures as the previous Generation 5 trial but  70  with the addition of 0.2% Pluronic in the suspension, and tested the new design at different doping ratios. The average effective yield for the combined system is 97% and is independent of the UC13 doping ratio (Table 6.6, Figure 6.8A). The complete system shows an enrichment ratio of 822 (+198, -141) and 3000 (+643, -278) for doping ratios of 1:100 and 1:1000 UC13 cells to WBCs, respectively. This yield and enrichment performance results in a final cancer cell purity of 89±2% and 75±2% for the two doping concentrations. While the yield performance only increased slightly compared to the previous design, the enrichment performance increased significantly. The ability of this device to run at an adequate oscillation pressure to enable successful ratcheting of almost all WBCs while retaining even the smaller UC13s contributes to a ~500x increase in enrichment performance compared to the previous design, raising the final output purity from <1% to 75±2% for the 1:1000 case.  (+/-) WBC  WBC Contaminants  (+/-) UC13  Sorter  UC13 Lost  Sorter  UC13 Collected  (+/-) UC13  Concentrator  UC13 Lost  Concentrator  UC13 Collected  (UC13:WBC)  Concentration  Doping  Table 6.6 Yield, purity, and enrichment performance for the Generation 6 combined system.  1:100  74  1  1  74  0  5  9  3  1:1000  74  1  1  24  1  1  8  2  Conc.  (+/-)  Sorter  (+/-)  Total  (+/-)  Yield  Yield  Yield  Yield  Yield  Yield  (%)  (%)  (%)  (%)  (%)  (%)  1:100  99  <1  100  <1  99  <1  89  +/-2  822  1:1000  99  <1  96  <1  95  <1  75  +/-2  3000  UC13: WBC  Purity (%)  (+/-) Purity  ER  (%)  (+/-) ER  +198, -141 +643, -278  71  A  B  4000  E n r ic h m e n t R a t io  100  Y ie ld ( % )  80 60 40 20  3000  2000  1000  0  0  1  :1  0  0 1  :1  0  0  0  1  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  :1  0  0 1  :1  0  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  C  100  P u r it y ( % )  80 60 40 20 0  1  :1  0  0 1  :1  0  0  0  D o p in g C o n c e n t r a t io n ( U C 1 3 : W B C )  Figure 6.8 A) Results showing capture yield at different UC13 doping concentrations. The yield remains consistent, even in diminishing UC13 concentrations. B) Plot for enrichment and (C) purity.  The robustness of the platform is also confirmed across multiple experimental trials. The device consistently removes >99% of the contaminant WBCs while retaining >95% of the target UC13 cells under optimized operating conditions. Specifically, optimal experimental conditions mainly involves of having a high quality WBC and UC13 sample. UC13 samples would ideally come from a culture flask that is 3 days from passaging, when  72  the cells are the healthiest and most uniform in size. However, due to resource constraints, we typically used cells that were 3-4 days old. On a few occasions where 4 day old cells were used, some cells were too small and observed to traverse past the cut-off. Therefore, we discarded results when there was an abundance of these small 4-day-old cancer cells. For WBCs, a good sample consists of cells that are uniformly distributed without cell aggregates. However, WBC quality is harder to control since both donor and sample preparation techniques have an influence on the quality of the final sample. WBC clumps are problematic for our device because they can be trapped at the lower funnel rows and be collected at the cancer cell outlet, hence skewing the enrichment and purity results. We typically discarded the results from experiments when there was an over-abundance of WBC clumps. Through the process of using blood from the same small group of donors, along with careful and systematic sample preparation, we were able to achieve consistent results for the amount of WBC removed (>99%) and UC13 yield (>95%) across all experiments. To calculate the throughput of our cell separation device requires estimating the time required to process a sample in both the hydrodynamic concentrator and subsequently the ratchet sorter. The concentrator reduces the overall sample volume as well as the total number of contaminant cells. The throughput of this device is approximately 0.5 μL/min. The throughput of the ratchet sorter is limited by the size of the funnel array. When cells are infused into the sorting area, the maximum amount of fluid processed in that batch corresponds to the channel volume at the bottom of the sorting area. This volume corresponds to an hourly processing capacity of approximately 10,000 cells per hour using a maximum input concentration to the ratchet sorter of 20,000 cells/μL (achievable with the addition of Pluronic to prevent non-specific adhesion). The throughput can be enhanced through parallelization of the ratchet sorter, with an estimated performance of 160,000 cells per hour for the maximum device length (16x longer) capable of fitting onto a standard 3”x2” microscope slide.  6.5  Discussion of Separation Results  The primary advantage of this device over other filtration-based separation systems is its extreme high selectivity, which can be evaluated using the parameters of capture yield and enrichment ratio. The average capture yield of our combined microfluidic platform is 97%. 73  In comparison, other filtration technologies typically report yields of 80-90% [9, 15, 30, 32, 46]. Furthermore, this yield can be obtained while simultaneously achieving high purity of the isolated sample. In our experiments where UC13 cells are doped into WBCs at a concentration of 1000:1, >99% of the contaminant WBCs were removed, leading to a 3000 (+643, -278) fold enrichment of the UC13 population. The capability to simultaneously achieve both extremely high yield and enrichment makes this approach very promising for rare cell separation applications such as the separation of circulating tumor cells. Another important advantage of our system is the ability to extract the separated cells and refresh the filter microstructure after each batch of cells is processed. This capability enables filtration conditions to remain constant over time. In other words, the Nth batch of cells to enter the device is processed in the same manner as the first batch, therefore preserving high filtration selectivity. Experiments spanning several hours have shown no progressive clogging or degradation in the performance of the device over time. Furthermore, this resettable filter reduces the duration of time each target cell is pushed against the micropore of a filter, typically lasting less than 60 s compared to other filtration techniques where the cells remain in the pore for the entire duration of the filtration process. Reducing cell exposure to persistent forces ensures gentle handling of rare and potentially fragile cells while preserving their inherent physiological properties. Moreover, the ability to extract cells off the filter element makes this system cascade-able for both on-chip and off-chip downstream processing. Finally, since cell separation in our device is achieved through active deformation of cells by the microstructure, resulting separation based on a combination of size and deformability, we are able to achieve better yield and purity than devices that separate cells solely based on size separation, such as hydrodynamic flow fractionation and inertial microfluidics. Using a cutoff of 6 μm ensures that even very small cancer cells are not able to pass through while large WBCs easily traverse through this pore into higher rows of the device. Differences in deformability are likely to be a more important factor than size in the separation of CTCs from WBCs because of the variation in the size of CTCs [47] that may overlap with the size scale of WBCs. Our system is therefore potentially useful for applications such as this where cells have overlapping size distributions but can be distinguished based on deformability. 74  Chapter 7: Conclusion 7.1  Summary of Results  We developed a microfluidic device for automated label-free separation of rare cancer cells from WBCs. This device consists of a hydrodynamic concentrator and a ratchet cell sorter. The concentrator prepares the sample for the cell sorter by removing the majority of the suspending fluid to increase the throughput of the cell sorter. Specifically, the concentrator removes 24% of unwanted peripheral WBCs and 78% of excess fluid, while retaining 99% of the target cancer cells. The ratchet sorter then removes >99% of the unwanted WBCs while retaining an average of 96% of the target cancer cells. Using a sample where cancer cells are doped into WBCs at a ratio of 1:1000, the combined system achieved a cancer cell yield of 96.0±0.1%; the outlet had a purity of 75±3%; and the population of cancer cells in the mixture was enriched by a factor 3000 (+643, -278). The throughput of the overall device is limited by the ratchet sorter at 0.5 μL per hour, or equivalently 10,000 cells per hour, but can be increased in the future through parallelization. The ability to separate cells with high yield and selectivity based on size and deformability shows promise as a biomechanical cell separation method that is an alternative to biochemical cell separation. Cells captured using our platform can also be extracted easily off-chip or undergo further processing using a microfluidic platform, making this device versatile to be stand-alone or coupled with other technologies.  7.2  Limitations  Although the throughput of our cell separation system has been increased substantially with the addition of the hydrodynamic concentrator, the overall throughput is limited by the capacity of the ratchet sorter. The ratchet sorter is capable of operating at 0.5 μL per hour at a maximum concentration of 20,000 cells per microliter. This throughput corresponds to an hourly processing capacity of approximately 10,000 cells per hour, which is too small for many applications. This processing speed can be enhanced through parallelization of the ratchet sorter, which is the next step in the development pipeline.  75  7.3  Future Work  The work presented in this thesis shows a proof-of-principle for the separation of cultured cancer cells from peripheral WBCs based on biomechanical properties. Currently, the throughput of this platform is approximately 10,000 cells per hour. To make this platform suitable for processing clinical samples, we must increase the throughput of the system significantly. We plan to do this by multiplexing the ratchet sorter device by increasing the length of the 2D array by a factor of 16. We will also investigate into the possibility to make the device into a single layer device. The removal of membrane microvalves will make the device simpler and easier to fabricate with PDMS and other potentially more robust polymer materials such as PMMA and COC. A single layer device will require the ratchet sorter to perform continuous processing, which will require further research work. Finally, we will further optimize our sample preparation technique and perform experiments on samples that more closely simulate patient blood samples. Currently, UC13 cancer cells are doped at the desired concentration into a WBC sample after removing the red blood cells by centrifugation. Since CTCs are already inside whole blood in patient samples, we will need to investigate the cancer cell yield by doping cancer cells into whole blood in order to ensure the cancer cell population is preserved by our sample preparation procedure. Finally, after parallelizing the device throughput, improving the device design, and optimizing the sample preparation technique, we will then compare the performance of the microfluidic system to the commercial CellSearchTM system using doped samples, and eventually, samples obtained from cancer patients.  76  References 1.  2. 3. 4. 5. 6. 7. 8.  9.  10. 11. 12. 13. 14. 15. 16. 17.  18.  19. 20.  Mego, M., et al., Circulating Tumor Cells (CTCs) and Epithelial Mesenchymal Transition (EMT) in Breast Cancer: Describing the Heterogeneity of Microscopic Disease. Cancer Research, 2009. 69(24): p. 649s-650s. Yu, M., et al., Circulating tumor cells: approaches to isolation and characterization. Journal of Cell Biology, 2011. 192(3): p. 373-382. Hoshino, K., et al., Microchip-based immunomagnetic detection of circulating tumor cells. Lab on a Chip, 2011. 11(20): p. 3449-3457. Nagrath, S., et al., Isolation of rare circulating tumour cells in cancer patients by microchip technology. Nature, 2007. 450(7173): p. 1235-U10. Hur, S.C., A.J. Mach, and D. Di Carlo, High-throughput size-based rare cell enrichment using microscale vortices. Biomicrofluidics, 2011. 5(2). Hur, S.C., et al., Deformability-based cell classification and enrichment using inertial microfluidics. Lab on a Chip, 2011. 11(5): p. 912-920. Gascoyne, P.R.C., et al., Dielectrophoretic separation of cancer cells from blood. Ieee Transactions on Industry Applications, 1997. 33(3): p. 670-678. Moon, H.S., et al., Continuous separation of breast cancer cells from blood samples using multi-orifice flow fractionation (MOFF) and dielectrophoresis (DEP). Lab on a Chip, 2011. 11(6): p. 1118-1125. Lin, H.K., et al., Portable Filter-Based Microdevice for Detection and Characterization of Circulating Tumor Cells. Clinical Cancer Research, 2010. 16(20): p. 5011-5018. Zheng, S.Y., et al., 3D microfilter device for viable circulating tumor cell (CTC) enrichment from blood. Biomedical Microdevices, 2011. 13(1): p. 203-213. Lee, J.M., et al., The epithelial-mesenchymal transition: new insights in signaling, development, and disease. Journal of Cell Biology, 2006. 172(7): p. 973-981. Mikolajczyk, S.D., et al., Detection of EpCAM-Negative and Cytokeratin-Negative Circulating Tumor Cells in Peripheral Blood. J Oncol, 2011. 2011: p. 252361. Hur, S.C., et al., Deformability-based cell classification and enrichment using inertial microfluidics. Lab Chip, 2011. 11(5): p. 912-920. Sun, J., et al., Double spiral microchannel for label-free tumor cell separation and enrichment. Lab Chip, 2012. 12(20): p. 3952-60. Zheng, S.Y., et al., 3D microfilter device for viable circulating tumor cell (CTC) enrichment from blood. Biomed Microdevices, 2011. 13(1): p. 203-213. Wakeman, R.J. and C.J. Williams, Additional techniques to improve microfiltration. Separation and Purification Technology, 2002. 26(1): p. 3-18. Yap, B. and R.D. Kamm, Cytoskeletal remodeling and cellular activation during deformation of neutrophils into narrow channels. Journal of Applied Physiology, 2005. 99(6): p. 2323-2330. Dong, X., et al., Patient-Derived First Generation Xenografts of Non-Small Cell Lung Cancers: Promising Tools for Predicting Drug Responses for Personalized Chemotherapy. Clinical Cancer Research, 2010. 16(5): p. 1442-1451. Hannemann, J., et al., Quantitative High-Resolution Genomic Analysis of Single Cancer Cells. Plos One, 2011. 6(11). McFaul, S.M., B.K. Lin, and H. Ma, Cell separation based on size and deformability using microfluidic funnel ratchets. Lab Chip, 2012. 12(13): p. 2369-76. 77  21.  22.  23. 24.  25. 26.  27. 28. 29. 30.  31.  32.  33. 34.  35. 36. 37. 38. 39.  Stott, S.L., et al., Isolation of circulating tumor cells using a microvortex-generating herringbone-chip. Proceedings of the National Academy of Sciences of the United States of America, 2010. 107(43): p. 18392-18397. Yamada, M., M. Nakashima, and M. Seki, Pinched flow fractionation: Continuous size separation of particles utilizing a laminar flow profile in a pinched microchannel. Analytical Chemistry, 2004. 76(18): p. 5465-5471. Di Carlo, D., Inertial microfluidics. Lab on a Chip, 2009. 9(21): p. 3038-3046. Bhagat, A.A.S., et al., Pinched flow coupled shear-modulated inertial microfluidics for high-throughput rare blood cell separation. Lab on a Chip, 2011. 11(11): p. 18701878. Voldman, J., Electrical forces for microscale cell manipulation. Annual Review of Biomedical Engineering, 2006. 8: p. 425-454. Becker, F.F., et al., Separation of Human Breast-Cancer Cells from Blood by Differential Dielectric Affinity. Proceedings of the National Academy of Sciences of the United States of America, 1995. 92(3): p. 860-864. Wang, X.B., et al., Cell separation by dielectrophoretic field-flow-fractionation. Analytical Chemistry, 2000. 72(4): p. 832-839. Seal, S.H., A Sieve for the Isolation of Cancer Cells and Other Large Cells from the Blood. Cancer, 1964. 17: p. 637-42. Desitter, I., et al., A new device for rapid isolation by size and characterization of rare circulating tumor cells. Anticancer Res, 2011. 31(2): p. 427-41. Hosokawa, M., et al., Size-Selective Microcavity Array for Rapid and Efficient Detection of Circulating Tumor Cells. Analytical Chemistry, 2010. 82(15): p. 66296635. Vona, G., et al., Isolation by size of epithelial tumor cells - A new method for the immunomorphological and molecular characterization of circulating tumor cells. American Journal of Pathology, 2000. 156(1): p. 57-63. Tan, S.J., et al., Versatile label free biochip for the detection of circulating tumor cells from peripheral blood in cancer patients. Biosensors & Bioelectronics, 2010. 26(4): p. 1701-1705. Guo, Q., S.M. McFaul, and H.S. Ma, Deterministic microfluidic ratchet based on the deformation of individual cells. Physical Review E, 2011. 83(5). Haines, W.B., Studies in the physical properties of soil V The hysteresis effect in capillary properties, and the modes of moisture distribution associated therewith. Journal of Agricultural Science, 1930. 20: p. 97-116. Dharmasiri, U., et al., Microsystems for the Capture of Low-Abundance Cells. Annual Review of Analytical Chemistry, Vol 3, 2010. 3: p. 409-431. Roberts, B.W. and W.L. Olbricht, The distribution of freely suspended particles at microfluidic bifurcations. Aiche Journal, 2006. 52(1): p. 199-206. Unger, M.A., et al., Monolithic microfabricated valves and pumps by multilayer soft lithography. Science, 2000. 288(5463): p. 113-116. Marcus, J.S., W.F. Anderson, and S.R. Quake, Microfluidic single-cell mRNA isolation and analysis. Analytical Chemistry, 2006. 78(9): p. 3084-3089. Singhal, A., C.A. Haynes, and C.L. Hansen, Microfluidic Measurement of AntibodyAntigen Binding Kinetics from Low-Abundance Samples and Single Cells. Analytical Chemistry, 2010. 82(20): p. 8671-8679. 78  40.  41. 42.  43. 44. 45. 46. 47.  Lara, O., et al., Enrichment of rare cancer cells through depletion of normal cells using density and flow-through, immunomagnetic cell separation. Experimental Hematology, 2004. 32(10): p. 891-904. Pantel, K. and R.H. Brakenhoff, Dissecting the metastatic cascade. Nature Reviews Cancer, 2004. 4(6): p. 448-456. Wong, I. and C.M. Ho, Surface molecular property modifications for poly(dimethylsiloxane) (PDMS) based microfluidic devices. Microfluidics and Nanofluidics, 2009. 7(3): p. 291-306. Cote, R.J., et al., Detection of Occult Bone-Marrow Micrometastases in Patients with Operable Lung-Carcinoma. Annals of Surgery, 1995. 222(4): p. 415-425. Tan, S.J., et al., Microdevice for the isolation and enumeration of cancer cells from blood. Biomedical Microdevices, 2009. 11(4): p. 883-892. Absher, M., Hemocytometer Counting. Tissue Culture: Methods and Applications 1973, New York: Academic Press. Tan, S.J., et al., Microdevice for the isolation and enumeration of cancer cells from blood. Biomed Microdevices, 2009. 11(4): p. 883-892. Marrinucci, D., et al., Case study of the morphologic variation of circulating tumor cells. Human Pathology, 2007. 38(3): p. 514-519.  79  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0073460/manifest

Comment

Related Items