SEPARATION OF CIRCULATING TUMOR CELLS USING RESETTABLE CELL TRAPS by Xi Qin B.Eng., Zhejiang University, 2012 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Biomedical Engineering) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) January 2015 © Xi Qin, 2015ii Abstract Immunoenrichment of conventional circulating tumor cells (CTCs) may fail to capture cells with poor antigen expression. Micropore filtration is a compelling label-free alternative to separate CTCs based on their biophysical characteristics rather than biochemical characteristics. However, this strategy is prone to clogging of the filter microstructure, which dramatically reduces selectivity after processing large numbers of cells. Our group previously reported the resettable cell trap (RCT) mechanism to perform micropore filtration in a way that is resistant to clogging. We improved the selectivity of this label-free mechanism by filtering the samples multiple times on chip and dramatically improving the throughput by parallelization. The resettable cell trap device is a microfluidic mechanical constriction with adjustable apertures that can capture CTCs based on their distinct size and deformability. It can also be periodically cleared to release the trapped cells to prevent clogging. Three identical cell traps are aligned in series which improves selectivity by removing leukocytes that non-specifically adhere to the surface of microchannels. We validated this mechanism by doping UM-UC13 bladder cancer cells into diluted whole blood at a density of 1 UC13 to 1000 leukocytes. The first filtration step achieved 183-fold enrichment and 93.8% yield. The second and third traps together provided an additional enrichment of ~5 without significant change in yield. Furthermore, additional filtration steps provide even greater enrichment. In patients with metastatic castration-resistant prostate cancer (mCRPC, n=24) and localized prostate cancer (LPC, n=18), CTCs were successfully identified using the resettable cell trap device followed by single-cell spectral analysis. We additionally compared the RCT device to the CellSearch® System, the only FDA approved commercial CTC enumeration platform. The microfluidic RCT device identified 83.3% (20/24) patients with >=5 CTCs per 7.5 ml of blood with a mean of 329 counts. Within the same patient group, the CellSearch only measured >=5 CTCs in 37.5% (9/24) patients with a mean of 23 CTCs per 7.5 ml of blood. The RCT device identified significant more CTCs and positively identified more mCRPC patients than the CellSearch system. iii Preface The mechanism described in Section 2.1 was originally developed Thomas Gerhardt and further developed by William Beattie, both fellow research group members. Here, I adapted this mechanism for whole blood processing and on-chip multi-filtration. Mater wafers for producing PDMS replicas described in Section 4.1 were fabricated with the help from William Beattie in the clean room. A version of Section 3.1.3 has been published. W. Beattie, X. Qin, L. Wang, and H. Ma, “Clog-free cell filtration using resettable cell traps,” Lab on a Chip, April 8, 2014, doi:10.1039/c4lc00306c. For this publication, I conducted the experiments and analysed the data. Section 4.5 and Section 5.3 describe the identification of circulating tumor cells from patient samples. The protocol for immune-fluorescent staining was optimized by Sunyoung Park. Richard Ang developed the method for spectrum scanning using the confocal microscope. iv Table of Contents Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iii Table of Contents ......................................................................................................................... iv List of Tables ............................................................................................................................... vii List of Figures ............................................................................................................................. viii List of Abbreviations .....................................................................................................................x Glossary ........................................................................................................................................ xi Acknowledgements ..................................................................................................................... xii Dedication ................................................................................................................................... xiii Chapter 1: Introduction ................................................................................................................1 1.1 Motivation ....................................................................................................................... 1 1.2 Performance metrics ....................................................................................................... 2 1.3 Review of current methods ............................................................................................. 3 1.3.1 Biochemical methods .................................................................................................. 3 1.3.2 Biophysical methods ................................................................................................... 5 1.3.2.1 Field flow fractionation....................................................................................... 5 1.3.2.2 Hydrodynamic separation ................................................................................... 7 1.3.2.3 Filtration .............................................................................................................. 9 1.4 The clogging problem ..................................................................................................... 9 1.5 Goals of this thesis ........................................................................................................ 10 Chapter 2: Mechanism of Design ...............................................................................................12 2.1 Resettable cell trap mechanism ..................................................................................... 12 2.2 Device operation ........................................................................................................... 14 2.2.1 Multiplexing .............................................................................................................. 14 2.2.2 Operational cycle ...................................................................................................... 16 2.3 Multi-filtration .............................................................................................................. 17 2.3.1 Three identical resettable cell traps in series ............................................................ 17 2.3.2 Forward and reverse re-filtration between two traps ................................................ 18 2.4 Relationship to previous work ...................................................................................... 19 v Chapter 3: Device Characterization ...........................................................................................21 3.1 Characterization of flow speed ..................................................................................... 21 3.1.1 Introduction ............................................................................................................... 21 3.1.2 Resistance ................................................................................................................. 22 3.1.3 Flow speed calibration .............................................................................................. 25 3.2 Characterization of the trapping pressure ..................................................................... 27 3.2.1 Indirect measure of trapping pressure ....................................................................... 27 3.2.2 Pressure drop across cell traps .................................................................................. 29 3.3 Automated system ......................................................................................................... 31 Chapter 4: Materials and Methods ............................................................................................34 4.1 Device fabrication and experimental settings ............................................................... 34 4.2 Sample preparation ....................................................................................................... 37 4.3 Experimental characterization-performance matrix ..................................................... 38 4.4 Cell viability test ........................................................................................................... 39 4.5 Patient sample processing, immunofluorescence staining and enumeration ................ 39 Chapter 5: Results and Discussion .............................................................................................41 5.1 Device validation experiments ...................................................................................... 41 5.1.1 Device performance characterization using UC13 cells ........................................... 41 5.1.2 Parameter optimization for patient sample ............................................................... 43 5.1.2.1 Device validation using LNCaP cells ............................................................... 43 5.1.2.2 Optimization for working with whole Blood .................................................... 44 5.2 Discussion of the validation results .............................................................................. 46 5.2.1 Deformability ............................................................................................................ 46 5.2.2 Size and deformability based separation................................................................... 48 5.3 Enrichment and identification of candidate CTCs from patients .................................. 50 5.3.1 Distinguishing candidate CTCs from non-CTC cells ............................................... 50 5.3.2 Enrichment and identification of candidate CTCs from patients with mCRPC ....... 56 5.3.3 Enrichment and identification of candidates CTCs from patients with LPC............ 58 Chapter 6: Conclusion .................................................................................................................62 References .....................................................................................................................................63 Appendices ....................................................................................................................................69 vi Appendix A : Automatic software code.................................................................................... 69 Appendix B : PDMS device fabrication ................................................................................... 79 B.1 Fabrication of the master wafers ............................................................................... 79 Appendix C : CTC counting results for patient samples .......................................................... 83 C.1 CTC counting results for mCRPC patient samples................................................... 83 C.2 CTC counting results for LPC patient samples......................................................... 84 vii List of Tables Table 3-1 Resistance measurement of the device. ........................................................................ 24 Table 5-1 Size comparison of UC13 cells and LNCaP cells. ....................................................... 48 Table B-1 Photolithography fabrication parameters for SU-8 layers ........................................... 80 Table B-2 Photolithography fabrication parameters for SPR layer .............................................. 80 Table C-1 CTC counts in samples from patients with mCRPC................................................... 83 Table C-2 CTC counts in samples from patients with LPC......................................................... 84 viii List of Figures Figure 1-1 Process of immunomagnetic separation. ....................................................................... 3 Figure 1-2 Microfluidic affinity capture. ........................................................................................ 4 Figure 1-3 Acoustophoresis separation process. ............................................................................. 6 Figure 1-4 DLD separation process. ............................................................................................... 7 Figure 1-5 Cell separation in a spiral channel. ............................................................................... 8 Figure 1-6 Filtration-based cell separation.. ................................................................................... 9 Figure 1-7 Micro-filtration with tapered-shaped pillars. .............................................................. 10 Figure 2-1 The resettable cell trap mechanism.. ........................................................................... 12 Figure 2-2 Schematic drawing of the cell sorting implementation of a conventional rectangular membrane valve. ........................................................................................................................... 14 Figure 2-3 Photograph of the microfluidic device with 3x128 parallel traps. .............................. 15 Figure 2-4 Sample processing cycle using three resettable cell traps in series. ........................... 18 Figure 2-5 Sample processing cycle using forward and reverse re-filtration between two traps.. 19 Figure 3-1 Demonstration of the components of the cell trap device. .......................................... 22 Figure 3-2 Demonstration of the resettable cell trap device using a hydrodynamic resistor system........................................................................................................................................................ 23 Figure 3-3 Simplified hydrodynamic resistor system of the resettable cell trap device. .............. 24 Figure 3-4 Enrichment and yield performance of the RCT mechanism under different flow speeds. ........................................................................................................................................... 26 Figure 3-5 Control diaphragm in contact with the center fin of an unfilled device. ................... 28 Figure 3-6 UC13 trapping curves as a function of differential trapping pressure of devices with different actuating pressures. ........................................................................................................ 28 Figure 3-7 Calibration curve for the cell trap mechanism. ........................................................... 29 Figure 3-8 Diagram of one single trap channel. ........................................................................... 30 Figure 3-9 UC13 yield curve as a function of the applied pressure on each of the three traps on the same device. ............................................................................................................................ 31 Figure 3-10 Interface of the automatic software. .......................................................................... 33 Figure 4-1 Multilayer soft lithography fabrication process. ......................................................... 34 Figure 4-2 Connection between external pneumatic pressure controller and the falcon tube and connection between the falcon tube and the microfluidic device. ................................................ 35 ix Figure 4-3 Photo of experiment system and block diagram of experimental apparatus. .............. 36 Figure 4-4 Fluorescent microscopy pictures of the collection and waste reservoirs. ................... 38 Figure 4-5 Flow of patient samples processing, immunofluorescence staining and enumeration........................................................................................................................................................ 39 Figure 5-1 Performance of resettable cell traps in enrichment and retention of UC13 cells relative to WBCs. ....................................................................................................................................... 42 Figure 5-2 Cumulative enrichment and yield performance when filtering the cells more than 3 times. ............................................................................................................................................. 43 Figure 5-3 Occupation of pockets by RBCs with different concentration (different dilution factor). ........................................................................................................................................... 44 Figure 5-4 Yield performance with dilution factor of whole blood.............................................. 45 Figure 5-5 Yield performance with different leukocyte concentrations. ...................................... 45 Figure 5-6 Size and deformability in the resettable cell trap mechanism for cell separation. ...... 46 Figure 5-7 UC13 yield and LNCaP yield under different flow rates. ........................................... 47 Figure 5-8 Size distribution of UC13 cells and WBCs. ................................................................ 49 Figure 5-9 Micrographs of a CTC and leukocyte stained with fluorescent markers. ................... 50 Figure 5-10 Merged images of a CTC, a leukocyte and an all-positive cell with their corresponding spectral curves. ...................................................................................................... 51 Figure 5-11 Block diagram of the process of identifying candidate CTCs and non-CTC cells. .. 52 Figure 5-12 Merged image and spectrum of a typical WBC. ....................................................... 53 Figure 5-13 Merged images and spectrums of three easily distinguishable all-positive cells. ..... 53 Figure 5-14 Merged images and spectrums of three all-positive cells that are hard to distinguish........................................................................................................................................................ 54 Figure 5-15 Merged images and spectrums of three CTC candidates. ........................................ 55 Figure 5-16 Enumeration of CTCs derived from CRPC patient samples and 5 control samples. 56 Figure 5-17 Enumeration of CTCs derived from LPC patient samples processed by both resettable cell trap (RCT) device and microfluidic ratchet device. .............................................. 59 Figure 5-18 Counts of contamination WBCs left over after processing for each sample. ........... 61 x List of Abbreviations CTC Circulating Tumor Cell RCT Resettable Cell Trap mCRPC metastatic Castration-Resistant Prostate Cancer LPC Localized Prostate Cancer FFF Field Flow Fractionation DEP Dielectrophoresis DLD Deterministic lateral displacement PDMS Polydimethylsiloxane BSA Bovine Serum Albumin CK Cytokeratin EpCAM Epithelial Cellular Adhesion Molecule EMT Epithelial–Mesenchymal Transition FDA Food and Drug Administration PBS Phosphate Buffered Saline xi Glossary Enrichment The factor by which the relative concentration of target cells increases after separation. Retention (Yield) The overall percentage of target cells collected. Trapping pressure The net pressure across the control diaphragm in the resettable cell trap. It is the applied pressure to the control channel minus the pressure of the flow at that point. Critical trapping pressure The trapping pressure that allows over 90% of target cells to be captured. Actuating pressure The applied pressure to the control channel that actuates the diaphragm to achieve contact with the center fin. xii Acknowledgements It would be impossible for me to work through the project without the support of my family, friends and coworkers. Here, I want to offer my enduring gratitude to everyone who has helped me through my graduate school studies especially To my supervisor Dr. Hongshen Ma, for his patience and inspiration in guiding me on this research and teaching me how to think systematically as an engineer. To my mentor William Beattie, for his selfless help in teaching me his skills and sharing his knowledge in the microfluidic field and answering my endless questions. To my dear lab coworkers, Chao, Aline, Kerryn, Xiaoyan, Quan, Emily, Richard and the rest of the members of Multiscale Design Lab, for being such good families and friends to me. To my roommate Chao Jin, for being such an amazing person and getting over difficulties both on studies and life alongside me. To my best friend Meng Li and my parents, for encouraging, supporting and trusting me all the time. xiii Dedication This thesis is dedicated to my parents For their endless love, support and encouragement1 Chapter 1: Introduction 1.1 Motivation The separation and identification of circulating tumor cells (CTCs) from peripheral blood cells is currently gaining significant momentum in cancer research and treatment. CTCs are cells that originate from primary tumors and shed into the peripheral blood circulation. They are believed to be the seeds of metastasis and precursors for the growth of the secondary tumors [1]. Enumeration of CTCs has great clinical and prognostic value which offers a non-invasive means to diagnose tumors and monitor therapy efficacy in different tumor types [2]–[4]. CTCs are associated with overall survival and progression-free survival of patients with metastatic carcinomas [5], [6]. Specifically, previous studies have found the presence of 5 or more CTCs per 7.5 ml of peripheral whole blood to be associated with poor outcomes [7], [8]. The key challenge in isolating CTCs from whole blood is their extreme rarity in contrast with the abundance of leukocytes and erythrocytes, which are the contaminant cells in the separation process. Current techniques to separate CTCs can be broadly divided into two main categories: biochemical and biophysical methods [9]. Biochemical methods usually rely on the affinity capture of cell surface antigens such as epithelial cell adhesion molecule (EpCAM) to enrich target cells. An example of biochemical methods is the commercialized Veridex CellSearch system (Raritan, NJ, USA). However, epithelial-to-mesenchymal transition (EMT) during dissemination can trigger down-regulation of EpCAM on cancer cells and it may be impossible for CellSearch system to capture these cancer cells with inadequate EpCAM expression [10]. Biophysical methods are a group of label-free alternatives that enrich for CTCs using biophysical characteristics such as density, size, deformability and conductivity to distinguish them from other blood cells. Cells are unmodified with these physical separation processes and are still viable after separate. Thus the cells are compatible with a wider range of downstream analyses. In this chapter, I will present the performance metrics and functional requirements associated with CTC separation as well as review, discuss and compare various microfluidic approaches and their advantages and limitations. 2 1.2 Performance metrics Every cell separation mechanism can be thought of as a system with the original mixed sample as input and the purified sample as output. Common reported sets of metrics are used to compare the performance and efficacy of various cell separation techniques. They are listed as follows:  Yield: The percentage of target cells collected  Enrichment: The factor by which the relative concentration of target cells increases after separation.  Throughput: The total number of cells processed per unit time Throughput is used to determine how much time is needed to process a specific sample. While the yield and enrichment ratio are commonly used to describe how specific a cell separation mechanism is, these two metrics are not only a function of the mechanism, but is also a function of the sample used in the experiments. There are not as yet consistent standards for the validation of cell separation mechanisms and reported studies have tested different cell lines such as H1650 lung cancer cells, MCF7 breast cancer cells [11], and LNCaP prostate cancer cells [12]. LNCaP and UC13 bladder cancer cells are used in the experiments in this thesis. The intrinsic size and deformability distributions of these cell lines can vary significantly. The direct comparison of the retention and enrichment ratio of competing platforms may therefore be impractical. 3 1.3 Review of current methods 1.3.1 Biochemical methods Biochemical methods usually rely on the affinity capture of surface antigens such as EpCAM. There are two reported methods that utilize the affinity capture of CTCs: 1) Immunomagnetic separation: This method positively select cells of interest utilizing magnetic beads coated with specific antibody (i.e., most are anti- EpCAM) to bind specific proteins on cells [13]. An example of this is the CellSearch® system which is the only FDA-approved CTC enumeration technology for metastatic breast, prostate, and colorectal cancer. Figure 1-1 below describes the process of immunomagnetic separation. Figure 1-1 Process of immunomagnetic separation. Antibody conjugated magnetic beads are added into the sample and target cells with expression of specific proteins will bind to the bead. An external magnetic field is introduced to impart a force on tagged cells and direct them to be collected. Background cells will be washed away. 2) Microfluidic affinity capture: Similar to immunomagnetic separation, this method uses antibody-coated microstructures to bind CTCs. With high surface area to volume ratio, it is expected to increase the likelihood that a target CTC will come into contact with the 4 functionalized surface [14]. While the sample flows through the microstructures, target cells will bind to the antibodies bound onto the structures. Cells not bound will be washed away. Figure 1-2 shows a schematic drawing of microfluidic affinity capture with its optimized designed array of functionalized microposts and its real life application in separating model CTCs doped into whole blood [15]. Antibody conjugated microstructuresBackground CellsTarget CellsFlow DirectionFlow DirectionA B Figure 1-2 Microfluidic affinity capture. A) Schematic drawing of a mechanism of microfluidic affinity capture. Micro-pillars are conjugated with specific antibodies. Target cells will bind to the functionalized surface and background cells will simply flow away. B) Scanning electron microscope image of a captured NCI-H1650 lung cancer cell (pseudo coloured red) spiked into blood and bound to an anti-EpCAM coated micropost [15]. Biochemical methods have successfully demonstrated their strength in separating CTCs from patient samples with high throughput and high selectivity [16], [17]. However, the down-regulation or degradation of cell surface markers may lead to loss of CTCs due to the inefficient binding. This occurs in two ways: first, the heterogeneity of CTCs defines different expression levels of surface markers among different cancer types and even within the same patient influenced by disease stage [2], [18]. Second, due to the epithelial to mesenchymal transition (EMT), some fractions of CTCs, often the most aggressive subpopulations, lose expression of epithelial antigens [19]. 5 In summary, while affinity capture is currently one of the most successful CTC detection tools available. The CellSearch® system works as an excellent tool for correlating both progression and overall survival [4], [6], [20]–[22] with CTC enumeration in patients with different metastasis cancer. It is the current “gold standard” for CTC detection and is the only FDA-cleared commercial system. However, it is believed that this technique may fail to capture a subpopulation of CTCs that don’t express or only express the chosen surface marker at low levels. To compensate for the low efficiency and sensitivity of biochemical methods, development has focused on separation techniques that sort or detect CTCs based on biophysical methods such as size and deformability. 1.3.2 Biophysical methods Biophysical methods are a group of label-free alternatives that enriches for CTCs using their biophysical characteristics such as density, size, deformability and conductivity. The rapid development of soft lithography in the late 1990s has allowed microfluidic technology to thrive. Many new microfluidic platforms were developed for use in cell separation due to their low cost, potentially increased sensitivity, small sample consumption and high speed processing. Biophysical separation mechanisms can be divided into two major groups: active separation that relies on an external field to act more strongly on one cell population, and passive separation that exploits the interaction of cells with the channel structure and the fluids [23]. Active separation is also called field flow fractionation (FFF). Passive separation can be further classified by two subcategories, hydrodynamic separation and filtration. I will describe some of the most popular mechanisms for each subcategory. 1.3.2.1 Field flow fractionation Field flow fractionation relies on an external field, utilizing properties like dielectric constants, size, or density, which acts more strongly on the target cell phenotype. This group is dominated by dielectrophoresis (DEP) [24]–[26], where a dialectric particle or cell is subjected to a non-uniform electric field. DEP of cells is based on the different polarizability properties that different cell phenotypes have. These strongly depend on the cell composition and morphology. A cell’s movement is therefore decided by its relative dielectric property in the suspending 6 medium and the frequency of non-uniform electric field applied. The cell will move along the direction of the applied field if the electrical polarizability of the cell exceeds that of the suspending medium. Otherwise, it will move in the reverse direction. Another FFF method employs acoustophoresis for sorting [27], [28]. Acoustophoresis depends on three fundamental physical characteristic of the cells; its size, density and compressibility. Cells are initially positioned along the channel walls and then pass through the ultrasound area where a force is generated by ultrasonic resonance to migrate cells towards the center of the channel shown in Figure 1-3. The speeds of migration will be different for cells with different size, density and compressibility. The collection of specific cell phenotypes is enabled by the different ultimate displacement after the ultrasound area. Background CellsTarget CellsUltrasoundFlow Direction Figure 1-3 Acoustophoresis separation process. Suspension of cells is initially positioned along the two channel walls and follows through the ultrasound area. Denser and larger cells will migrate faster towards the center of the channel than less dense and smaller cells. The ultrasound area is not long enough for all cells to achieve center placement, which enables collection of cells with different lateral displacement. These methods depend on the generation of external force fields, which increase the complexity for easy control and application. Instead of applying external fields for cell interactions, passive methods exploit only the interactions of cells with structures and fluids. 7 1.3.2.2 Hydrodynamic separation Hydrodynamic separation methods are typically size-based. Popular hydrodynamic separation methods include deterministic lateral displacement (DLD) and inertial microfluidics. Deterministic lateral displacement separation uses arrays of posts to laterally deflect the cells according to their size. The microposts are carefully designed in a staggered array (Figure 1-4) with a specific critical threshold size which is determined by the gap between the post and the stagger between post rows [29]. Small cells, below the threshold size, will flow within the stream, straight through the array in the direction of the carrying fluid, while larger cells will be bumped into other streamlines at each obstacle. Figure 1-4 DLD separation process. Smaller cells, below the critical size threshold will follow the original streamline parallel to the carrying fluid. Larger cells will be bumped into other streamlines at each obstacle and thus achieve a lateral displacement which is perpendicular to the direction of the flow [29]. Applications of DLD have very high throughput but is limited by its size-based discrimination. DLD mechanisms need to be tailored for specific applications with different target cell sizes. While DLD exploits the interaction of the cells with the designed solid microstructure, inertial microfluidics discriminate cells by size, based on their interactions with the flow field. Inertial microfluidics employs fluid mechanics under low but non-negligible Reynolds number (Re>1). One successful implementation of cell separation using inertial effects takes advantages of 8 secondary Dean flows in spiral microchannels [30], [31]. Suspended cells in the spiral channel experience two competing forces: the inertial lift force driven by the shear gradient created within the channel and the Dean drag force driven by mismatch of flow velocity between channel center and channel walls caused by centrifugal forces in the spiral microchannel. The combination of the two forces causes cells to migrate to their equilibrium position across the microchannel cross-section (Figure 1-5). The equilibrium position depends on the ratio of inertial lift and Dean Drag force which varies solely with cell size. Background CellsTarget CellsSample InletBuffer InletWaste OutletCollection OutletOuter WallInner WallOuter WallInner Wall Figure 1-5 Cell separation in a spiral channel. Samples are diluted with buffer injected into the channel at the same time. Cells are subjected to a combination of inertial lift force and Dean drag force, which causes cell migration to different equilibrium positions across the channel. Equilibrium positions are solely size-dependent. Due to the nature of inertial microfluidics which demands high flow speed, these techniques involving inertial effects embrace very high throughput (~1 million cells/min [30]). As these techniques are solely sized-based, enrichment performances are relatively poor. Also, to ensure that cell-cell collisions do not interfere with the inertial focusing effect, the samples must be highly diluted, which limits the method to samples with low hematocrits. 9 1.3.2.3 Filtration Filtration methods separate cells based on a combination of size and deformability. Cells are pumped through microscale constrictions using pressure or displacement driven flow. Separation is achieved when smaller and more deformable cells can traverse freely through these constrictions while larger and stiffer cells are blocked. Current implementations of microfiltration involve a variety of designs of filter-structures within which micropillar array filters [32] and membrane filters [33] are quite popular (Figure 1-6). Flow DirectionA B Flow DirectionBackground CellsTarget Cells Figure 1-6 Filtration-based cell separation. A) Schematic drawing of a micropillar array filter. B) Schematic drawing of a membrane filter. For both designs, smaller and more deformable cells can traverse freely through these constrictions while larger and stiffer cells are blocked. Filtration is of particular interest due to its ability to sort by deformability, which could potentially distinguish cells with different phenotypes or disease status better than size only separation techniques to achieve higher enrichment performance. 1.4 The clogging problem Although microfiltration can achieve separation based on both size and deformability with high selectivity [32], [33], a key challenge is the clogging of the filter microstructures. Clogging usually occurs after processing a large number of cells and causes unpredictable changes of the device’s hydrodynamic property and results in poor performance. Retrieval of isolated target cells may be difficult due to the cells remaining trapped in the microstructures. The mechanism designed by our group solves the clogging problems by applying reverse pressure periodically to unclog the channels by pushing way some of the stuck cells. A lateral flow was used to get the cells to travel to the collection reservoirs. The pillars were designed to 10 be tapered, so that the temporally applied backward pressure won’t reverse the separation process as shown in Figure 1-7. This mechanism was named the microfluidic ratchet. The cell separation device that I developed in this thesis achieved un-clogging in an alternative manner. The separation results of patient samples will also be compared to the microfluidic ratchet device. Background CellsTarget CellsForwardBackwardBackground CellsTarget Cells Figure 1-7 Micro-filtration with tapered-shaped pillars. Smaller and less stiff cells can freely traverse the gap between the pillars or by deforming through, while larger and less deformable cells cannot. Periodically added short-time backward pressure will unclog some of the stuck cells without reversing cells already passed the constrictions. 1.5 Goals of this thesis Previously, we developed the resettable cell trap (RCT) mechanism, which uses an adjustable aperture to capture cells based on their size and deformability, and can be periodically cleared to prevent clogging [34]. It successfully separated doped cultured cancer cells from leukocytes. However, the throughput and enrichment performances were still not good enough for real life implementation of separating circulating tumor cells directly from whole blood. 11 The goal of my project was to design an improved working device based on the RCT mechanism and separate CTCs directly from patient blood samples using this device. The collected CTCs had to be viable and retrievable for further downstream analysis. My goals were as follows: 1) Design and fabricate an improved RCT mechanism 2) Experimentally characterize the performance of the device using cultured cancer cells doped into whole blood samples donated by healthy donors. 3) Optimize the device for processing patient samples 4) Evaluate the viability of cancer cells after processing 5) Develop supporting software and the experiment system for automated sample processing 6) Separate and identify CTCs from blood samples from patients with Castration-Resistant Prostate Cancer (CRPC) and patients with Localized Prostate Cancer (LPC). 12 Chapter 2: Mechanism of Design 2.1 Resettable cell trap mechanism The resettable cell trap mechanism is a 2-layer PDMS structure comprising of a sample-carrying upper flow channel and a lower fluid-filled control channel. Separating these two layers is a thin flexible diaphragm that can be inflated by applying an external pneumatic pressure to control the geometry of the two microchannels. Opposing the diaphragm, the surface of the flow channel is textured with two rows of micro-pockets and a protruding center fin (Figure 2-1). These microstructures and the diaphragm combines to create an adjustable aperture that selectively trap and release target cells. The position of the diaphragm can be considered to have two states: a constricted state where the diaphragm is in contact with the textured surface to reduce the aperture of the flow channel; as well as a relaxed state where the diaphragm is deflected away from the textured surface to enlarge the aperture of the flow channel (Figure 2-1). Figure 2-1 The resettable cell trap mechanism. In the constricted state (top panel), the flow channel and diaphragm form an aperture to capture the more rigid CTCs. In the relaxed state (bottom panel), CTCs can be released and collected and this prevents clogging. Perspective SectionConstrictedRelaxedFlow Direction Target Cell Background CellControl DiaphragmFlow ChannelControl ChannelpocketsCenter FinSide Fins13 In the constricted state, the pressure in the control channel is greater than the flow channel and the diaphragm is deflected to come into contact with the center fin and side fins of the flow channel. The center fin and the two side fins acts as the mechanical stop to limit the movement of the diaphragm and flatten it to create an approximately rectangular channel on either side of the center fin. Since the top and bottom boundary of the aperture is the most parallel at the center of the channel, a flow focuser is used to center the cells upstream in the flow channels to provide the best possible selectivity to target cells [34]. Furthermore, micro-pockets lining both sides of the center fin temporary hold the larger and more rigid cells to prevent them from blocking the flow channel. This structure is capable of selectively capturing cell phenotypes with different sizes and deformabilities. In the relaxed state, the pressure in the control channel is less than the flow channel and the diaphragm is deflected away from the textured surface of the flow channel. The aperture in this state is large enough for all cells to pass through freely. By simply relaxing the diaphragm, the micro-pockets filled with captured cells can be purged to empty the recesses and the channels are reset. This ability to refresh the flow channel on demand is important to release captured cells and prevent clogging. One of the key advantages of the RCT mechanism is its ability to create an adjustable aperture with well-controlled geometry inside a microchannel. Previous adjustable mechanism relies only the basic structure of the conventional rectangular membrane micro-valves [35], which form two triangle-like openings in the two corners of the flow channel when the diaphragm is inflated to get into contact with the ceiling of the rectangular flow channel (Figure 2-2). As the result of using this irregular shape (not evenly distributed along the cross section) as the working filtration area, it is difficult to precisely correlate the opening size to the size of target cells so it lacks the ability to differ cells with similar physical properties, such as WBCs and CTCs, and is currently only used for separating particles from suspension [36]. Here we make some improvements on the basic structure of a rectangular valve by adding the center fin and the two side fins to flatten the diaphragm and make the filtration aperture approximately two regular rectangular channels. Thus it is easy to correlate the size of these two rectangular channels with the dimension of the smallest target cells in the design. 14 Working filtration area Figure 2-2 Schematic drawing of the cell sorting implementation of a conventional rectangular membrane valve. The working filtration area is two triangle-like openings formed by the inflated membrane at the two corners of the rectangular channel. This mechanism was employed by the previous version which was capable of separating microbeads with different sizes in less than 1 µm resolution by fine tuning the channel opening. Also, UM-UC13 bladder cancer cells doped into whole blood were separated from background white blood cells (WBCs) with an average enrichment of ~100 and throughput of ~900,000 cells per hour without clogging [34]. However, considering the extreme occurrence of CTCs, the new version needs to process higher volumes with less WBC contamination in order for to be used in clinical settings. 2.2 Device operation 2.2.1 Multiplexing The RCT chip multiplexes the resettable cell trap to increase sample handling and throughput. The whole cell separation device consists of 4 groups of 32 parallelized resettable cell channels, 128 channels in total (Figure 2-3). Main bifurcation channels connecting 4 groups and 4 minor bifurcation networks connecting 32 channels within each group are designed to evenly distribute cells into every single trap [37]. Five rounded valves work as on/off switches to route the sample and buffers from inlet reservoirs into collection and waste reservoirs on demand [35]. These fluidic components work together to make a functional prototype. The key of parallelization is to have a bifurcation network that evenly distributes cells into every single trap. One challenge in designing this bifurcation network is the finite size effect. For 15 example, suppose there is a uniform distribution of the lateral position of cell centres before entering the bifurcation channel. Since the centre of a cell is expelled from the channel walls in its radius, its distribution will shift considerably to the inner walls after one bifurcation. After passing through several bifurcations, its distribution will shift more and more towards the inner walls and finally cause the outmost channels in a parallel network to have very few cells. This effect is especially influential when the dimensions of channels and cells are close to each other. High pressure Buffer InletSample InletLow Pressure Buffer InletWaste ReservoirCollection ReservoirResettable Cell Traps1235 mmValve 0Valve 1Valve 2Valve 3Valve 4Flow DirectionMain bifurcationMinor bifurcation Figure 2-3 Photograph of the microfluidic device with 3x128 parallel traps. Flow channels and control channels are filled with red and green food coloring respectively. Valves are formed in the areas where 3 inlet flow channels, 2 outlet flow channels and control channels overlaps. 3x128 cell traps are formed in the areas where 128 parallelized flow channels and 3 long control channels in series. This effect can be mitigated by increasing the width of the bifurcation channels to make the cell size incomparable. In this prototype, the width of the main bifurcation channel is designed to be 500 um to alleviate the finite size effect and achieve a nearly even delivery of cells to each single channel. 16 2.2.2 Operational cycle Cell separation using this device involves a three-step cycle: 1) filtration, 2) purging, 3) collection. 1) Cells are first infused from sample inlet into the constricted cell trap. The larger and more rigid CTCs will be caught in the pockets at the trap area, while the smaller and less rigid leukocytes will traverse through this area and into the waste reservoir. 2) Buffer is infused via low pressure buffer inlet while the cell trap remains constricted. Cells captured in the trap remain captured, while contaminating background cells left over in all the channels are purged away into the waste reservoir. It takes 5-10 seconds. 3) Buffer is infused via high pressure buffer inlet into the device when all the traps are relaxed. All of the cells captured are released and direct into the collection reservoir. It takes 2-3 seconds. This releasing flow is much faster than the filtration and purging flow so as to produce high shear forces to remove cells that may have adhered to the walls of the cell traps [38], [39]. This operational cycle is performed in a periodic way. After each high shear force collection step, all cells are removed from the trap area and the whole device is reset to its initial state. This solves the clogging issue faced by many other filtration mechanisms. Clogging happens in those filters when a pore is occupied by a captured cell and remains occupied for the rest of the filtration process. This can cause unpredictable hydrodynamic property changes of the filters and decreases throughput. Our mechanism solves this problem by simply relaxing the diaphragm so that the aperture is large enough for all cells to pass through freely. Once the micro-pockets are filled up with captured cells, the channel can be purged to empty the recesses. This ability to refresh the flow channel on demand is important to keep the flow channel unclogged and reset the trap to its initial empty state for further filtration steps. This periodic refresh process prevents clogging while maintaining cell selectivity. For the validation experiments where target cell concentration is specific, the processed volume in each filtration step is determined so that on average, a total of 100 target cells are captured in the 128 channels, as to prevent obstruction of the flow channel which will dramatically decrease the filtration ability. For processing patient samples, where the CTC concentration is unknown, conservative estimates are made to determine the length/volume of the main filtration. 17 2.3 Multi-filtration It was observed in early experiments that while cancer cells were caught mostly at the beginning of constricted trap, WBCs were captured throughout the entire constricted trap. This suggested that cancer cells are captured because of mechanical constraint but contamination WBCs are caught because of non-specific adsorption to surfaces of the cell trap during filtration. Those WBCs adhered to the walls of the channels will be released with the captured cancer cells into the collection reservoir when trap is relaxed which limits the enrichment of this mechanism. Since non-specific adhesion of WBCs happens by possibility, filtering the sample through the cell trap multiple times can improve the selectivity of target cells. 2.3.1 Three identical resettable cell traps in series To do on-chip re-filtration, we cascaded three identical resettable cell traps in series on this new version as shown in Figure 2-3 and 2-4. Cells captured in the first trap will be released to move through to the second and third traps to separate the contamination WBCs. The length of the trap was also reduced to prevent non-specific adhesion of WBCs to the channel walls while still capturing target cells with high retention. Cell separation using the three cell traps in series will add two more filtration steps into the original three-step operational cycle to achieve a five-step cycle: 1) initial filtration, 2) purging, 3) trap2 re-filtration 4) trap3 re-filtration and 5) collection (Figure 2). The initial filtration using the first trap, the purging step and the collection step work exactly the same as introduced above. In step 3, low pressure buffer remains flow through the second constricted cell trap while the first trap is relaxed. Most target cells remains captured in the second trap while some of the contamination WBCs are washed way into the wash reservoir. It takes 2-3 seconds. In step 4, the third step is repeated when the third trap is constricted and the second trap relaxed. Using these three cell traps arranged in series and the five-step operational cycle, we achieve three filtrations in a row. 18 Target Cell Background CellClosed Valve Open ValveConstricted TrapRelaxed TrapWasteCollectionHPBLPB STraps1 2 3HPB WasteCollectionSTraps1 2 3LPB WasteCollectionHPBLPBSTraps1 2 3WasteCollectionHPBLPBSTraps1 2 31354Sample PressureLPB PressureHPB PressureWasteCollectionSTraps1 2 3LPB2HPBOpenOpenOpenOpenOpenOpen OpenOpenOpenOpen Figure 2-4 Sample processing cycle using three resettable cell traps in series. Cells from the inlet are filtered through three resettable cell traps before being flowed to the collection outlet. Captured cells are purged using a low-pressure buffer (LPB) flow, and collected with high-pressure buffer (HPB) flow. 1 is the initial filtration step. 2 is the purging step. 3&4 is the 2x re-filtration step. 5 is the collection step. 2.3.2 Forward and reverse re-filtration between two traps Experimental results of re-filtration using three traps in series show that at the end of the third filtration, there is further ability to improve selectivity. This can be achieved by repeated filtrations between two cell traps (Figure 2-5) using low pressure buffer fluid infused from the normal direction and reversed from the collection outlet. Theoretically, with each flow taking only seconds, many filtrations can be done in a relatively short time until the enrichment is 1. 19 Traps1 21Traps1 22Traps1 23Traps1 2N Figure 2-5 Sample processing cycle using forward and reverse re-filtration between two traps. 1 is the main filtration step. 2&3 combined is the re-filtration & purging step. 2&3 is repeated several times to achieve more filtrations. N is the collection step. LPB and HPB are switched by increasing the pressure source. 2.4 Relationship to previous work Two previous students, Tom Gerhardt and William Beattie, contributed to the development of this cell trap project. The mechanism originally came from Mr. Gerhardt’s idea. He designed this cell trap mechanism to enable chromatographic separation between target and background cells [40]. The trap was periodically constricted and relaxed to achieve different average flow speed for different cell phenotypes. The chromatographic separation nature made the throughput of this method extremely low. Mr. Beattie adopted this mechanism and redesigned the operation of this mechanism to work in a continuous separation manner [34]. He achieved an average enrichment of ~100 of cancer cells from WBCs mixtures. However, to achieve the final goal of processing clinical patient samples for separation of CTCs in a reasonable amount of time and get a pure as possible cell sample after separation for downstream analysis, a new version with much higher throughput and enrichment performance was desired. We also needed to make the new version able to process whole blood directly, to avoid any pre-processing which may potentially induce damage to the CTCs. The original contributions presented in this document are described below: 20 1) Achieved higher throughput with the new designed chip by parallelization 2) Designed the parallel traps with shorter length and corresponding operation cycle to implement on-chip re-filtration to achieve higher enrichment 3) Calibrated the relationship between trap membrane stiffness and target cell retention rate to determine the working parameters on a device-by-device basis 4) Made the mechanism work for processing whole blood by dilution 5) Characterized the efficacy of the device by separating UC13 from leukocytes in whole blood samples 6) Characterized the optimal operating conditions of the mechanism for patient sample processing by doing LNCaP cell separation experiments 7) Made the processing of bulk volume patient samples automatic without manual control of the swapping between valves 8) Successfully Separated CTCs from patient samples 21 Chapter 3: Device Characterization For the RCT mechanism, the performance of target cell yield and enrichment is determined by the combination of flow speed and channel cross-section size, which is correlated to trapping pressure. Detailed characterization of these two parameters will be laid out in the following sections. 3.1 Characterization of flow speed 3.1.1 Introduction In the microfluidic system demonstrated in this research, the fluids are driven by the application of pressure. To find the relationship between the applied pressures and flow rate and to predict the behavior of fluids in the channels, the realm where Reynolds number lies, is determined. The Reynolds number, Re is calculated by: (3.1) Where ρ is the fluid density µ is the viscosity of the fluid U is the velocity scale L is the length scale In our system, the length scale is 100 um, velocity scale is 1000 µm/s, the buffer used can be approximated to be similar to water, ρ = 103 kg/m3 and µ=10-3 Pa·s. Then Re equals to 0.1, which is <1. In this case, the flow rate through this device will be proportional to the pressure drop along its length. This constant of proportionality is the hydraulic resistance of the device. The relationship between these three parameters can be expressed as: (3.2) Where Q is the volumetric flow rate is the pressure drop across the length is the resistance 22 This resistance depends on the channel geometry and the fluid viscosity. For a rectangular channel with height h and width w, the approximation for the resistance is: ( ) (3.3) A specific resistance analysis on the resettable cell trap device is demonstrated in Section 3.1.2. 3.1.2 Resistance As introduced in Section 2.2.4, the whole cell separation device consists of 4 groups of 32 parallelized flow channels, totally 128 channels. Main bifurcation channels connect 4 groups together and micro bifurcation channels connect 32 channels within one group together as described in Figure 3-1. G1G2G3G4Front main bifurcation channel Back main bifurcation channel Main resistance channelInlet channels Minor bifurcation networks Cell trap channelsOutlet channels Channels connects bifurcation and groups Channels connects groups and bifurcation Figure 3-1 Demonstration of the components of the cell trap device. Main bifurcation channels connect 4 groups together and micro bifurcation channels connect 32 cell trap channels within one group together. 23 This device can be divided into several hydrodynamic components including inlet channels, main bifurcation channels connecting 4 groups and micro bifurcation channels connecting 32 trap channels to the main bifurcation network, 128 cell trap channels and outlet channels. This device can then be demonstrated using a hydrodynamic resistor system including the corresponding components as shown in Figure 3-2. RG1ROutRMRRB-G1RMBCFRIntRG1-BRMBCBRCT1RCT2RCT31RCT32RB-CTRCT-BRG2RB-G2RG2-BRG3RB-G3RG3-BRG4RB-G4RG4-B Figure 3-2 Demonstration of the resettable cell trap device using a hydrodynamic resistor system. Inlet channel resistance is denoted as RInt and out channel resistance as ROut. Resistance of the main bifurcation channel connected to the inlet channel is donated as RMBCF and resistance of the main bifurcation channel connected to the outlet channel is donated as RMBCB. Resistances of the channels that connect main bifurcation channels and the 4 groups of cell trap channels are denoted as RB-Gi and RGi-B (i=1, 2, 3, 4). Resistance of one group of cell trap channels is denoted as RGi. Within each group of cell trap channels, resistance of each cell trap channel is denoted as RCTi (i=1, 2…32) and the resistances of the minor bifurcation networks that connect the cell trap channels are denoted as RB-CT and RCT-B. The long rectangular channel connects the outlet channel at the end is designed to provide most of the resistance of the whole chip and the resistance is denoted as RMR. The total length of the channels that connect the main bifurcation channels and each of the 4 groups of cell trap channels is designed to be the same i.e. RB-Gi+RGi-B is the same for each group and can be denoted as RBGB. Four groups including the minor bifurcation networks are identical 24 and the resistance is the same. When the device is running, only one of the three inlet channels and one of the two outlet channels will be in use and the resistance network can be simplified as RMBC= RMBCB+RMBCB, = RBGB/4=(RB-Gi+RGi-B)/4 and = =RGi/4 as shown in Figure 3-3. R,G ROut RMRR,BGBRMBCRInt Figure 3-3 Simplified hydrodynamic resistor system of the resettable cell trap device. The value of RInt, RMBC, ROut and RMR can be can be calculated using equation 3.3 with known channel length, width and height. The only difficulty is to calculate the value of since the groups of channels are irregular shapes with minor bifurcation networks changing width along the bifurcation and the cell trap channels are textured with pockets located in the ceiling. We therefore did an experiment to measure the resistance of the chip from the inlet to the outlet without the long resistance channel at the end. According to equation 3.2, a specific pressure was applied to the chip to drive water from the inlet to the outlet of the device without the long channel. A pipette was used to measure the volume in the outlet of the device after a certain amount of time. With the resistance of the whole chip known and RInt, RMBC, and ROut calculated, the value of can be obtained. The result is summarized in Table 3-1 below. Table 3-1 Resistance measurement of the device. Parts Resistance measured (mbar*s/µL) Resistance calculated (mbar*s/µL) Whole chip without main resistance channel 516 - Inlet, outlet, main bifurcation channels and channels connecting main bifurcation channels and 4 groups - 205 Main resistance channel (RMR) 4864 4338 Sorting area ( ) - 516-205=309 25 To determine the error of this measurement method, the resistance of the main resistance channels was both measured as described above and calculated according to equation 3.3. When the calculated resistance value was used as the real value, the error of measuring by pipet was (4864-4338)/4338=11.5%. To calculate the resistance value, the height of the channel from the wafer was measure by an optical profiler from Veeco (USA) whose measurement error is less than 0.1% [41]. The width and length of the device was obtained directly from the design. However, the real value can be potentially larger than the value calculated here. A 2 - 5% shrinkage in the length of the chip was observed when fabricating PDMS replicas (consistent with other reports [42]). The shrinkage ratio of the horizontal direction was more than that of the perpendicular direction and this could make the width shrink more than the length of the main resistance channel. Thus, the real resistance value could be greater than currently calculated and therefore, the error of the measurement method is actually lower than 11.4%. Furthermore, we designed a long rectangular channel at the end of the chip. It provides over 90% of the resistance of the chip so that the pressure drop from the inlet to the cell traps is negligible. This is easier for the application of the correct trapping pressure as according to the readings from the variable pressure controller. With this, the pressure needed to drive the water-like buffer through the device to achieve desired flow rate was approximately 300 mbar and the trapping pressure was approximately 500 mbar. However, the viscosity of blood is 3~4 times that of water, therefore, the driving pressure needed for the same flow rate will be 900~1200 mbar and the trapping pressure will be over 1100 mbar. Not only is this pressure out of the range of the controller, but also the high pressure will further increase the likelihood that the bonding between the 2 PDMS layers will fail. Therefore, this long channel was discarded in the real implementation of the device. With the known resistances of all of the components, the flow speed can be predicted by the application of a certain amount of pressure. 3.1.3 Flow speed calibration Flow speed here describes the speed that cells traverse trough the cell trap area. It can be calculated as the length of the trap area divided by the time needed to go through. The time can be obtained by videoing using the camera. However, it is easier and more straightforward to use 26 flow rate instead of flow speed. Flow rate has a proportional relationship with flow speed. A throughput for processing 1 ml of blood in one hour, i.e, a volumetric flow rate of 1000 µL/h, was aimed for with this device. Under this flow rate, the required trapping pressure was first determined by flowing cultured UC13 cancer cells (as described in Chapter 4) through the device and the trapping pressure was adjusted until over 90% of incident cells were captured. To test the behavior of the device under different flow rate conditions, the trapping pressure difference was kept as constant. We tested the function of flow rate by the separation of doped UC13 cells from leukocytes in 4 different healthy donors (Figure 3-4). For each data point, the measured result shown is the average of triplicate experiments. 0 4 0 0 8 0 0 1 2 0 0 1 6 0 0 2 0 0 002 0 04 0 06 0 08 0 01 0 0 0F lo w S p e e d (L /h )EnrichmentD o n o r1D o n o r2D o n o r3D o n o r40 4 0 8 0 1 2 0 1 6 0 0 2 0 0 04 06 08 01 0F lo w S p e d (L /h )Yield (%)D o n o r1D o n o r2D o n o r3D o n o r4Figure 3-4 Enrichment and yield performance of the RCT mechanism under different flow speeds. UC13 cells doped into whole blood as a function of flow rate from 4 different donors. Each data point is the average of triplicate experiments on the same sample. The enrichment value was shown to increase with flow rate. However, a key factor limiting the enrichment is the non-specific adsorption of leukocytes to the surface of the cell trap during the filtration phase. These adsorbed leukocytes are released with the UC13 cells during the collection phase, thereby limiting the purity of the output sample. The enrichment of cancer cells relative to leukocytes improves with increasing flow speed because of the increased shear forces reduces non-specific adhesion of leukocytes [38]. At a flow rate of < 1600 µL/h, the resettable cell trap device was consistently able to obtain a yield of 88–96%. When the flow rate was raised to 2000 µL/h, however, the yield of UC13 cells drops to ~70%. Additionally, some trapped 27 UC13 cells show signs of morphology change where the previously round cells were observed to take on an elongated shape. Therefore, a flow rate of 1600 µL/h is most likely the practical limit for the RCT device to retain a reasonable yield and prevent cell damage from the high shear force. For our multi-filtration experiments, UC13 cells are expected to be trapped for longer periods of time under high shear force. To prevent damage to the cells we decreased the operational flow rate from 1600 µL/h to 1000 µL/h. 3.2 Characterization of the trapping pressure The trapping pressure is another important variable for the performance of the RCT device. It refers to the pressure difference between the control layer and the flow layer across the elastomeric diaphragm in the resettable cell trap. To achieve the same channel cross-section size under the same flow rate, the trapping pressure is different for control membranes with different thickness or stiffness. Though we tried to make the devices consistent across batches, there were many variables to consider during the fabrication process. The control layer was fabricated by spinning liquid PDMS onto a silicon wafer and baking it for a certain amount of time. As PDMS viscosity changes with temperature, heating during mixing and degassing may change the viscosity of mixed PDMS and cause a batch-to-batch thickness variation. To compensate for this problem, a method to determine the trapping pressure on a device-by-device basis was needed. 3.2.1 Indirect measure of trapping pressure Since the desired trapping pressure was determined by the diaphragm stiffness, the measurement of the stiffness of the diaphragm could therefore be used to calibrate the cell trap. The pressure needed to actuate the diaphragm to obtain contact with the center fin can be correlated to the diaphragm stiffness. This actuating pressure was then used to correlate with the trapping pressure for one specific device. Specifically, the pressure that is applied to the diaphragm is incrementally increased from 0 mbar to until contact is observed (Figure 3-5). This pressure at that contact point is recorded as the 28 actuating pressure of the specific device. By measuring the actuating pressure, the diaphragm stiffness for that individual device is quantified. The remaining question becomes how to correlate this actuating pressure to the trapping pressure. Figure 3-5 Control diaphragm in contact with the center fin of an unfilled device. The pressures applied from left to right are 109.2 mbar and 109.5 mbar. There is a distinct transit from non-clear-contact at 109.2 mbar to clear-contact at 109.5 mbar (highlighted with a red circle). The accuracy is 0.3 mbar which is quite high and it is the same as the accuracy of the variable pressure controller. To quantify this relationship, a series of experiments were performed to measure the UC13 retention curve with different trapping pressures on microfluidic devices with different actuating pressures. First, the actuating pressure for one specific device was measured and recorded. Then the UC13 retention curve for that device was obtained by flowing UC13 through this device under fixed flow rate but various trapping pressures. Plotted curves are shown in Figure 3-6. 0 1 0 0 2 0 0 3 0 04 05 06 07 08 09 01 0 0T ra p p in g p re s s u re (m b a r )UC13 yield (%)3 8 .8 m b a r A c tu a t in g P re s s u re7 7 .6 m b a r A c tu a t in g P re s s u re1 0 0 .2 m b a r A c tu a t in g P re s s u re1 2 0 m b a r A c tu a t in g P re s s u re1 5 7 .8 m b a r A c tu a t in g P re s s u re Figure 3-6 UC13 trapping curves as a function of differential trapping pressure of devices with different actuating pressures. 29 The trapping pressure at which over 90% of target cells were captured is the critical trapping pressure. The critical trapping pressure for each specific device is thus determined using Figure 3-6. Finally, a calibration curve was calculated (Figure 3-7) using the critical trapping pressure and the actuating pressure for that same device. 0 5 0 1 0 0 1 5 0 2 0 005 01 0 01 5 02 0 0A c tu a tin g P re s s u re (m b a r )Critical Trapping Pressure (mbar) Figure 3-7 Calibration curve for the cell trap mechanism. This shows the trapping pressure required to trap 90% of UC13 as function of the actuating pressure of the device. Data is based on the curves from Figure 3-6. Before every experiment described in Chapter 5, the actuating pressure of each device was measured and the calibration curved of Figure 3-7 was used to determine the desired trapping pressure. The results presented in Chapter 5 show that the UC13 yield is consistently 90±5% which further justifies the implementation of this indirect measurement. 3.2.2 Pressure drop across cell traps This new version of the device has three identical traps in parallel. After discussing the trapping pressure desired for the first trap in section 3.2.1, the trapping pressures for the other two must also be determined. As mentioned in Section 3.1.2, the high resistance long rectangular channel at the end of the chip was discarded in implementation. Since the resistance of the sorting area is close to that of the whole chip, there will be noticeable pressure drop along the trapping channel. As the trapping pressure desired for each chip is constant (determined by the diaphragm stiffness only) and it is 30 equal to the net pressure across the control diaphragm (the applied pressure of the control channel minus the flow pressure at that point), the applied pressure of the control channel needs to be dropped accordingly for traps2 and trap3. The pressure drop across two adjacent traps can be approximated by assuming the whole sorting area is a shape uniformly distributed along its length, with the same length as a single cell trap channel. This will ignore the minor bifurcation channels within each of the 4 groups. This is reasonable since the bifurcation channels have a height of 50 um and are generally wider while the trap channels have a height of 15 um and a much narrower width. According to equation 3.3, the cubic power of height will make the resistance of the minor bifurcation channels incompatible with the resistance of the trap channels. As Figure 3-8 shows, the total length of a trap channel is 3325 um while the distance between each trap is 700 um. The induced pressure drop can be calculated as: (3.4) Where is pressure drop across the chip is the resistance of the sorting area is the resistance of the whole device Since , Q is desired to be 1 mL/h, =309 mbar*s/uL (section 3.1.2), 18 mbar. 1st Trap 3rd Trap2nd Trap3325 µm700 µmFigure 3-8 Diagram of one single trap channel. The total length is 3325 um and distance between two traps is 700 mbar. The pressure drop between each trap is ~18 mbar. 31 To validate this, UC13 retention curves for each trap on the same device was obtained by experiment. The results are shown in Figure 3-9. 2 5 0 3 0 0 3 5 0 4 0 02 04 06 08 01 0 0A p p ile d p re s s u re o n tra p (m b a r )UC13 yield (%)T ra p 1T ra p 2T ra p 3 Figure 3-9 UC13 yield curve as a function of the applied pressure on each of the three traps on the same device. From Figure 3-9, there is an obvious pressure shift of 10~20 mbar between two adjacent traps. This proves the theoretical calculation of the pressure drop across each trap. However, applied pressure is not the trapping pressure and the trapping pressure is calculated as the applied pressure minus the flow pressure at that point. Thus, taking the pressure drop of flow across each trap, the trapping pressure is actually exactly the same for all of the three traps on the same device. 3.3 Automated system To prevent the clogging problem, the microstructure needs to be periodically reset. During the validation experiments, valves were manually switched to route either sample or buffer into the device to one of the collection or waste reservoirs according to the operation cycle detailed in section 2.3. When two or more target cells per channel was retained as monitored through the camera on the microscope, the device was manually reset. However, for patient sample processing, it is impossible to know how many CTCs there are or how many CTCs have been 32 captured through monitoring. The chip would still need to be periodically cleared to prevent clogging and also to prevent trapping CTCs for too long which may cause damage to the cells under high shear force. Thus the period for resetting the chip was estimated and set at 10 minute intervals. It takes hours to process one sample and switching between valves is repetitive and laborious, thus demanding an automated system. The desired system should turn the on-chip valves on/off to guide the fluids through the different routes according to the demanding operation cycles. The hardware requires that the solenoid valves are controlled by a microprocessor on a custom made pressure board. This pressure board was fully developed and the microprocessor can be controlled from a computer. Thus automated software running on the computer to give commands to the microprocessor is the only extra part needed to build the automated platform. The software was developed using Microsoft Visual C#. The interface is shown in Figure 3-10. The software can be used to control the valves individually or control the valves automatically according to the time interval setups for each step in the operation cycles. Progress bars on the bottom demonstrate the elapsed time for each step in each cycle. 33 Figure 3-10 Interface of the automatic software. Individual valves can be controlled manually or the valves will run automatically according to the time setup for each step in the operation cycles. Detailed codes are provided in Appendix A. 34 Chapter 4: Materials and Methods 4.1 Device fabrication and experimental settings The RCT device was fabricated using standard multilayer soft lithography techniques using polydimethylsiloxane (PDMS).[35], [43] Master wafers for the control and flow channels are patterned through photolithography on campus in the AMPEL Nanofabrication Facility. A schematic of the fabrication process is shown in Figure 4-1 below. UV explosurePhotomaskPhotoresistBlank Si waferFlow layer waferPDMSCured PDMSPolyurethaneCured polyurethanePDMSFlow layerWaferControl layer waferPDMSCured control layerABCPunched flow layerPunched control layerGlass slide Figure 4-1 Multilayer soft lithography fabrication process. A: Photolithography fabrication of the flow and control wafers. B: Fabrication of the master molds to produce flow layers replicas to avoid excessive use of the flow layer wafer. C: Bonding of flow layer and control layer together and then to a glass slide. Drafts are designed with Draftsight and generated to high resolution photomasks by CAD/Art Services (Bandon, Oregon). 3D structures are formed on a silicon wafer by sequentially depositing 2D layers with photoresist. The flow layer master wafer consists of five layers: four layers with rectangular cross-sections made of SU-8 negative photoresist (Microchem, Newton, Massachusetts) and one layer of round cross-section made of SPR positive photoresist (DOW Canada, Calgary, Alberta) for the valves. The control layer master wafer consists only one layer 35 made of SU-8. PDMS masters are fabricated against the flow layer master wafer to make polyurethane molds. Mass production of the PDMS replicas of the flow layer are casted from the polyurethane molds instead of the master wafer to avoid breaking the fragile wafer which takes a lot of effort to produce. PDMS replicas against the master molds yield the flow layers while control layers are made by spinning a thin PDMS layer on the control layer wafers. The two layers were joined when brought into contact after oxidization in an oxygen plasma chamber (Harrick Plasma, Ithaca, NY). A 0.5 mm OD punch (Harris Unicore, Ted Pella Inc., Redding, CA) was used to create the on-chip ports. The devices were then plasma bonded to a 25x75 mm glass slide (Fisher Scientific). Detailed fabrication procedures will be included in Appendix B. For all channels, fluids are introduced into the microfluidic device from 15 ml polypropylene falcon tubes (BD Biosciences, Mississauga, Canada) through Tygon microbore tubing with 0.02 inch ID (Cole-Parmer, Montreal, Canada) and then a 0.017 inch ID stainless steel needle (New England Small Tube, Litchfield, NH) connected to the device (Figure 4-2). Custom machined capFalcon tubeTubingsNeedlesRCT deviceTubing Figure 4-2 Connection between external pneumatic pressure controller and the falcon tube and connection between the falcon tube and the microfluidic device. 36 The flow is actuated via external pneumatic pressure through custom machined caps that fit the falcon tubes (Figure 4-2). The pneumatic pressure source for both the sample and buffer infusion is from a 4-channel microfluidic flow control system (MCFS-Flex, Fluigent, France). The control valves on the device chip are activated by a custom designed system consisting of on-off pressure valves and a MSP430 microprocessor (Texas Instruments), which provides easy and flexible programming abilities to meet different automation requirements. Prior to use, device channels were slowing flushed with 0.2% Pluronic F-127 (Sigma-Aldrich, St. Louis, Missouri, USA) in PBS for surface passivation for 20 min. Fluid outlets can be punched with 6 mm OD punch to form on-chip reservoirs or with 0.5 mm OD punch to collect the fluids into a 96-well plate (Thermo Fisher Scientific, Rochester, NY, USA) or a 15 ml tube through the needle and tubing. The whole experiment setup is shown as Figure 4-3. 123451.Computer 3.Pressure Controller2.Pressure Valve Controller4.Microfluidic device5.Microscope with camera Figure 4-3 Photo of experiment system and block diagram of experimental apparatus. The computer controls the pressure delivered to the microfluidic chip and monitors the progress of experiment through the camera. 37 4.2 Sample preparation Validations of the RCT device were performed using whole blood doped with UM-UC13 bladder cancer cells and LNCaP prostate cancer cells. Whole blood was drawn from 20 healthy donors into 6 ml EDTA-coated collection tubes (Becton-Dickinson, Franklin Lakes, NJ, USA). Whole blood was then stained with Hoechst 33342 (Invitrogen, Carlsbad, CA, USA) in a concentration of 8 µg/ml and diluted to a concentration of 2 million leukocytes per ml with PBS. UC13 bladder cancer cells were cultured in complete minimal essential medium (CMEM): minimum essential medium Eagle (MEM) (Life Technology, Carlsbad, CA, USA) supplemented with 10% (v/v) fetal bovine serum (FBS) (Life Technology), 1% sodium pyruvate (Life Technology), 1% L-glutamine (Life Technology), 1% MEM non-essential amino acids (Life Technology), and 1% penicillin streptomycin (Fisher Scientific). LnCaP cells were cultured in RPMI 1640 media (Life Technology) containing 10% (v/v) FBS, 2mM L-glutamine and 1% penicillin streptomycin. Both two cell lines are incubated at 37 °C in a humidified environment with 5% CO2. Cells were detached from the 15 ml culture flask (Sigma-Aldrich) through the following steps: 1) waste culture medium was take out using pipet 2) cells were washed with 2 ml PBS 3)cells were kept 1 ml Trypsin (Life Technology) for 2 mins 4) the bottom of the flask wash was washed with 2 ml medium (CEME for UC13, RPMI for LNCaP) 5)all fluids was taken out from the flask and put into a 15 ml falcon tube and centrifuged at 980 rpm for 5 mins 6) supernatant was take out and cells were re-suspended in 2ml medium Cancer cells were stained with calcein AM (Life Technology) in a concentration of 2 µM for 30 mins. Then cells were centrifuged at 980 rpm for 5 mins, re-suspended with 2 ml medium and a small portion of the cell fluids was taken out and diluted to count the concentration of cells. Cancer cells with known concentration were doped into the already stained and diluted whole blood at a ratio of 1:1000 cancer cells to leukocytes. The mixed sample processed in each cell separation trial contained a minimum of 100 cancer cells. Each processed sample contained ~100 000 leukocytes. 38 4.3 Experimental characterization-performance matrix As described in Chapter 1-2, there are two main characteristic experiment results that are used to measure the performance of RCT device. The first is yield, which here, is defined as the retention rate of target cancer cells. The second is enrichment, which is defined as the ratio of target cancer cells to background cells (leukocytes) in the collection reservoir divided by the same ratio of the input sample. After each experiment, the number of cancer cells in both collection and waste reservoirs were counted, while leukocytes were counted in the collection reservoir. Cancer cells were identified by the green fluorescence of the Calcein AM stain and leukocytes were identified by the fluorescent blue Hoescht stain. Fluorescent microscopy pictures were taken using an inverted microscope (Nikon ECLIPSE Ti) and camera (QImaging, Surrey, BC, Canada). Numbers of cells in the pictures were manually counted. Examples of the fluorescent pictures of both collection and waste reservoirs are shown in Figure 4-4. UC13 Collection WBC Collection UC13 Waste Figure 4-4 Fluorescent microscopy pictures of the collection and waste reservoirs. Cancer cells were stained green and WBCs were stained blue. Cells in each reservoir were counted manually and used to calculate yield and enrichment. 39 4.4 Cell viability test Viability of captured cells was tested using the Live/Dead Viability Assay kit (Life Technology). The kit tests cell membrane integrity and includes two staining components-calcein AM which is green and ethidium homodimer-1which is red. Stained cells will turn red from green under fluorescent microscope when the membrane integrity is damaged. To do cell viability test, UC13 cells were incubated in a 2 mL solution of 2 μM calcein AM and 1 μM ethidium homodimer-1 for 30 minutes. Then the UC13 cells were processed through the RCT device and collected in a well of the 96-well plate. Cells directly taken from the original unprocessed sample were put into another well. Live and dead cells in each well were counted under the fluorescent microscope. Passing UC13 through the RCT device resulted in a small drop of the viability from 97.41% to 97.14%. 4.5 Patient sample processing, immunofluorescence staining and enumeration Metastatic castrate resistant prostate cancer (mCRPC) patients (n=24) and localized prostate cancer (LPC) patients (n=18) were recruited to the study. Each patient with mCRPC had failed their initial therapy, mostly abiratirone or enzalutamide. The patients with LPC were untreated and were going to take the prostatectomy. The blood was draw right before the surgery. The patients were treated at the Vancouver Prostate Centre in Vancouver, Canada. Blood was drawn from these patients into 6ml EDTA-coated blood collection tubes (BD). A volume of 1 ml of blood was pipetted into 15 ml falcon tube and diluted with 1 ml PBS. The diluted sample was directly processed using the RCT device. The whole process is shown in Figure 4-5. 1 ml PBS 1 ml Whole blood 2 ml SampleStaining and washingDilutionRCT processingStainingScanning and identification Figure 4-5 Flow of patient samples processing, immunofluorescence staining and enumeration. 40 Cells were collected into a 15 ml falcon tube through needle and microbore tubing. The enriched cell fraction was washed with 1 ml PBS by centrifugation at 400 g for 5 min and then fixed in 1ml of 3% paraformaldehyde (PFA, Sigma, USA) for 15 min, after which the cells were washed again. Cells were permeabilized in 1 ml of 0.5% Tween20 for 10 min, washed with PBS and blocked by incubation in 3% BSA (Sigma-Aldrich, St. Louis, Missouri, USA) in PBS for 30 min. Every step was conducted at room temperature. Cells were immunofluorescently stained with cytokeratin (CK) using Pan-Keratin (C11) Mouse mAb-Alexa Fluor® 488 conjugated (Cell Signaling Technology, Danvers, Massachusetts, USA), for EpCAM using EpCAM (VU1D9) Mouse mAb-Alexa Fluor® 594 conjugated (Cell Signaling Technology), and for CD45 using anti-human CD45-APC (Biolegend, San Diego, California, USA) with 0.625 µg/ml, 0.525 µg/ml, 0.36 µg/ml concentrations respectively in PBS supplemented with 3% BSA and kept at 4℃ overnight. Stained cells were washed 3 times with PBS to remove floating superfluous antibodies. After the last centrifuge, cells were re-suspended in 40ul PBS and stained with DAPI using VECTASHIELD® Mounting medium with DAPI (Vector Laboratories, Burlingame, CA, USA) in a concentration of 0.075 µg/ml. All cells were transferred into a well of a Corning® 384-well high content image plate (Sigma-Aldrich) and centrifuged down at 400 g for 2 min. The well was automatically scanned at 40X magnification with a confocal microscope (LSM 780, Carl Zeiss, Oberkochen, Germany) and Zen software (Carl Zeiss). Spectrum analysis of single cells was conducted manually to identify the presence of CTC candidates in the captured image. CTCs were considered DAPI+/CK+/EpCAM+or-/CD45- while WBCs were considered DAPI+/CK-/EpCAM-/CD45+. 41 Chapter 5: Results and Discussion 5.1 Device validation experiments The two goals of the device validation experiments are to characterize device performance under different circumstances and to develop optimal parameters for processing patient samples. 5.1.1 Device performance characterization using UC13 cells For the first goal, to characterize device performance, UM-UC13 bladder cancer (UC13) cells were used as target cells and were doped into whole blood. Compared to WBCs, UC13 cells have significantly different deformability while the size distributions overlap. The detailed distributions are shown in section 5.2. The overlapping in size makes it difficult to separate UC13 cells from WBCs with high selectivity for mechanisms that are size based only. The RCT mechanism separates cells based on size and deformability. Thus, UC13 cells will be a good phenotype for validating the device. Samples were prepared by doping UC13 cells into diluted whole blood at a ratio of ~1:1000 UC13: leukocytes. The optimal flow rate was previously determined to be 1600 µL/h after which there is significant drop of yield as well as a change in the shape of the trapped cancer cells, indicating potential damage to the cells (refer to detail in Section 3.1.3). During the multi-filtration experiment, UC13s will be trapped for a longer period of time under high shear force. To prevent damage to cells, the operational flow speed was decreased from 1600 µL/h to 1000 µL/h. The trapping pressure was determined for each device using the way described in Section 3.2.1. The key performance metrics are yield and enrichment. The yield refers to the capture rate of target cells relative to the total number of processed target cells. The enrichment is defined as the ratio of target cancer cells to the background cells in the collection reservoir divided by the same ratio in the input sample. The RCT device increased enrichment without sacrificing the yield by doing on-chip multi-filtrations. To determine the enrichment and yield of our RCT device during its multi-filtration, UC13 cells and WBCs filtered away in each of filtration steps and cells collected in the final collection step were directed to different wells and the number of UC13 42 cells and WBCs were counted respectively in each of the wells. Enrichment and yield values for each of the filtration steps can be calculated from these cell numbers. For multi-filtration using three identical cell traps in parallel, as described in Section 2.3.1, the first trap acts as the initial filtration trap. After processing a total of 15 samples, we found that that the first trap processed 2x106 cells/hour with an average of 183-fold enrichment and 93.8% yield (Figure 5-1A). Cells captured in the first trap were released and re-filtered through the second and third traps. From the three tests with the three traps, the second and third traps together provided an additional enrichment of about 5-fold without additional change in the yield (Figure 5-1B). The average enrichment of the third trap is 1.4, still above 1. S a m p le sFold enrichment02 0 04 0 06 0 08 0 0Aa p le sYield (%)6 07 08 09 01 0 0Fold cumulativeenrichmentT r a p 1 T r a p 2 T r a p 31 01 0 01 0 0 0D o n o r 1D o n o r 2D o n o r 3BCumulative yield (%)T r a p 1 T r a p 2 r a p 36 07 08 09 01 0 0D o n o r 1D o n o r 2D o n o r 3 Figure 5-1 Performance of resettable cell traps in enrichment and retention of UC13 cells relative to WBCs. A: results of main filtration step from 15 samples. B: cumulative results of 3-trap serial filtrations. 43 For multi-filtration using forward and reverse re-filtration, as described in Section 2.3.2, the improvement in enrichment and yield of these additional filtrations is shown in Figure 5-2. Using this way, we achieved up to 5 filtrations in a row in the experiment until the additive enrichment of the last filtration is very close to unity, which suggests we are close to the limitation of improving enrichment by re-filtration. Fold cumulativeenrichmentPass1Pass2Pass3Pass4Pass51 0 01 0 0 01 0 0 0 0D o n o r 4D o n o r 5Cumulative yield (%)Pass1Pass2Pass3Pass4Pass50 .70 .80 .91 .0D o n o r 4D o n o r 5 Figure 5-2 Cumulative enrichment and yield performance when filtering the cells more than 3 times. The performance of our device is highly donor dependent but the trend of improvement is the same for each donor. These results show that the leukocytes that are captured in our device because of non-specific adhesion can be depleted by multiple re-filtrations. We achieved an average enrichment of ~900 after three filtrations. The enrichment and yield performance of RCT device rival previously reported label-free separation techniques [31], [44]–[46]. 5.1.2 Parameter optimization for patient sample 5.1.2.1 Device validation using LNCaP cells We hypothesized that CTCs from patients with metastatic castration-resistant prostate cancer (mCRPC) would be more deformable than cultured cancer cells. CTCs are highly invasive and greater invasiveness is correlated with greater deformability [47], [48]. Therefore, to optimize parameters for processing sensitive patient samples, we used androgen-sensitive human prostate adenocarcinoma LNCaP cells. Though cultured LNCaP cells have similar size distributions as the cultured UC13 cells, we found during experiments that higher trapping 44 pressures were required to capture LNCaP cells with the same flow speed (further illustrated in detail in section 5.2). This implies that LNCaP cells are more deformable than UC13 cells. LNCaP would therefore be a better mimic of the real CTCs than UC13 cells. Indeed, we found that the LNCaP cells are softer and the flow rate limit that they can withstand before they are damaged was 600 µL/h, much less compared to the limit for UC13 cells, which was 1600 µL/h. The trapping pressure under 600 µL/h flow rate to get a ~90% yield was found to be 200 mbar higher than that for UC13 cells. Multiple filtrations worked the same way as for the UC13 cells. The only difference was that the enrichment of the first main filtration step is much lower with an average of 83-fold, due to the slower flow speed and smaller channel openings. Extra filtrations gave an average enrichment improvement of ~5. We therefore applied the same parameter settings for the LNCaP cells (600 µL/h flow rate) to patient samples processing. 5.1.2.2 Optimization for working with whole Blood We further observed that high concentrations of red blood cells (RBCs) had a negative influence on the yield of target cells as RBCs aggregate together in the storing pockets (Figure 5-3) and thereby prevents the target cells from getting trapped. Whole Blood1.5X Dilution3X DilutionPockets Figure 5-3 Occupation of pockets by RBCs with different concentration (different dilution factor). This situation gets mitigated by diluting whole blood. 45 To avoid this, we diluted the whole blood, which significantly improved the device performance and yield from an average of 74.3% in whole blood to an average of 90.2% when blood was diluted 3 times (Figure 5-4). 0 2 4 6 802 04 06 08 01 0 0D ilu t io n fa c to rLNCaP yield (%)T e s t 1T e s t 2T e s t 3 Figure 5-4 Yield performance with dilution factor of whole blood. Yield is improved back to 90% with 3X dilution of whole blood. To determine that this improvement in yield was not influenced by the diluted leukocyte concentrations, we separated leukocytes from whole blood, resuspended them at various concentrations and added LNCaP cells to each suspension at a ratio of 1:1000 LNCaP cells to leukocytes. The lack of correlation between the yield and leukocyte concentration confirm that the performance is not necessarily related to leukocyte concentration (Figure 5-5). However, dilution of the sample means sacrifice of the time and throughput needed to process the sample. We therefore implemented a 2 times dilution to balance the overall yield and throughput. 0 2 4 6 8 1 005 08 09 01 0 0W B C c o n c e n tra tio n (1 06 /m l)LNCaP yield (%) Figure 5-5 Yield performance with different leukocyte concentrations. 46 5.2 Discussion of the validation results 5.2.1 Deformability Deformability is an intrinsic cellular property of the cell. It enables the cell to squeeze through tight spaces, such as the capillaries in the body. In cell separation applications, where deformability is used, cells can pass through constrictions that are smaller than the cells themselves. Our group developed a method to measure single cell deformability by using microfluidic micropipette aspiration [49], where deformability was determined by the pressure required to push single cells through a funnel. Difference in cell size was also taken into account using this mechanism. Other group members in our lab (Guan Quo and Lin Wang) have quantified the deformability of UC13 and WBCs using microfluidic micropipette aspiration, which shows that UC13 cells are less deformable than WBCs [50]. With our RCT device, deformation occurs when a cell goes through the trapping area. Trapping of the cell is determined by its deformability as well as the magnitude of the shear force from the surrounding fluid. Whether a cell can traverse the trapping area is determined by a combination of the cell’s deformability and size as well as the size of the opening of the channel. Figure 5-6 gives an intuitive view of how size and deformability work in this mechanism for cell separation. Small and soft cellLarge and stiff cellLarge and soft cellSmall cell Figure 5-6 Size and deformability in the resettable cell trap mechanism for cell separation. When certain shear force is applied and the channel opening is set, large and stiff cells cannot deform enough to traverse the channel while large and soft cells can deform more to pass through. Small and soft cells only need to deform a little to pass and cells smaller than the channel opening can traverse freely without deformation. 47 The shear force is proportional to the fluid velocity and therefore to the volumetric flow rate when the Reynolds number is <1. Even though UC13s are less deformable than WBCs, when flow rate is increased, UC13s will deform more and will finally traverse through the trapping area. Furthermore, too high a shear force will damage the cell membrane and decrease cell viability. To test the ability of UC13 cells to deform under various shear forces, UC13 cells were processed at varying flow speeds. The size of the channel opening was fixed by adjusting the control pressure to maintain a constant differential trapping pressure. The size of the channel opening was also larger than what was optimized in section 3.1.3 and was achieved by dropping the trapping pressure. Yield performance for each flow rate was calculated after each experiment. As shown in Figure 5-7, less UC13 cells were captured under higher flow speeds because more cells can squeeze through the constrictions. F lo w ra te (L /h )Yield (%)0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 02 04 06 08 01 0 0U C 1 3L N C a P Figure 5-7 UC13 yield and LNCaP yield under different flow rates. Yield decreases with high flow rate due to the higher shear force. LNCaP cells were observed to be more deformable than UC13s, even though their size distribution is similar (Table 5-1). Therefore, a higher trapping pressure (smaller channel opening size) was needed to achieve the same yield as UC13s with the same flow speed setup. 48 The same experiments were conducted to evaluate the deformability of LNCaP cells under different shear forces. As shown in Figure 5-7, the yield decreases faster with the increasing shear forces as compared to UC13 cells. Table 5-1 Size comparison of UC13 cells and LNCaP cells. A total of over 100 cells for each phenotype are measured. UC13 LNCaP Average Size in diameter /um 16.88 17.27 Standard deviation of size /um 2.02 1.88 5.2.2 Size and deformability based separation A separation method based solely on size is binary. Cells that are larger than the threshold size will be discriminated from cells that are smaller. These methods usually have poor specificity performance especially when there is a large overlap of the size distribution of target cells and background cells. For our device, the separation is based not only on size but also on deformability, so better performance can be expected as compared to the size-only separations. To prove that deformability is a functional biomarker in cell separation, we can compare the performance of our device to a solely size-based separation method using the exact same cell type in a mixture. Since the upper limit of performance of any solely size-based separation can be theoretically calculated from the size distribution of the cell mixtures, we started with measuring the sizes of both UC13 and WBCs. The size distribution shown in Figure 5-8 was achieved by manually measuring over 100 cells for each cell phenotype using a calibrated 20x objective lens on the Nikon ECLIPSE Ti microscope using NIS Element Basic software (Nikon). 49 Figure 5-8 Size distribution of UC13 cells and WBCs. A total of over 100 cells for each phenotype are measured. As there is a certain size distribution overlap between UC13 and WBCs, it is difficult to separate UC13 cells with a high selectivity based on size alone. According to the size distribution (mean and standard deviation known), the performance of a size-only separation method for both yield and enrichment can be calculated when a threshold size is chosen. By referring to the Z-chart of the Gaussian distribution (Figure 5-8), for a 90% yield of UC13 cells (mean diameter 16.88 µm and standard deviation 2.02 µm), we chose a threshold size of 14.28 µm. For this size, the WBC depletion rate is 99.65% (WBCs have mean diameter of 10.5 µm and standard deviation of 1.4 µm). The enrichment from an original 1:1000 UC13 to WBC doping ratio mixture will be 0.9/((1-0.9965)*1000))/(1/1000) = 257. With our RCT device, the average enrichment after the 3-step filtration can be over 1000 for a yield higher than 90%. This proves that our device is based on size and deformability, where deformability adds a selectivity capacity to the cell sorting mechanism. 00.050.10.150.20.250.30.350.40.450510152025306 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25UC13 WBCUC13 Normal Distribution WBC Normal DistributionNumber of cellsPercentageof CellsDiameter/um50 5.3 Enrichment and identification of candidate CTCs from patients 5.3.1 Distinguishing candidate CTCs from non-CTC cells After processing each patient sample using the RCT device, enriched CTCs were identified using single-cell spectral analysis after immunostaining. CTCs were defined as DAPI+ CK+ EpCAM+/- and CD45- while leukocytes were identified as DAPI+ CK- EpCAM- and CD45+ as represented in Figure 5-9. CTCWBCDAPI CKBBEpCAM MergedCD45 Figure 5-9 Micrographs of a CTC and leukocyte stained with fluorescent markers. Single cells in enriched samples were spectrally analyzed using the Zeiss LSM 780 system. This system incorporates a photomultiplier tube (PMT) detector which has 1.8-fold higher quantum efficiency than conventional PMT detectors to achieve higher sensitivity. It can excite fluorophores with up to eight different lasers and read 34 channels simultaneously with parallel spectral detection. Based on the antibodies we use and their different conjugated flouorophores, CK-Alexa 488 has a peak at 529nm, EpCAM-Alexa 594 has a peak at 617nm, and CD45-APC has a peak at 660nm in this system. According to these distinct emission wavelengths, the spectral analysis using the Zeiss system will allow us to distinguish potential CTCs from leukocytes. To see the spectrum of a cell, we need to make a circle surrounding the whole cell in the merged scanned images. The Zeiss software takes the average of fluorescence intensity of each pixel within the circle and displays the averaged fluorescence signals in the spectrum format. According to the definition of CTCs, the spectrum of a candidate cell in our system must meet the following criteria to be considered a CTC: 1) Must have DAPI signal to be considered as a 51 cell 2) Must have one peak for CK at 525 nm 3) May or may not have a peak for EpCAM at 617 nm 4) Must not have a peak for CD45 at 660 nm. Cells that do not meet the above criteria are not considered a CTC. A typical spectrum of a WBC in our system is defined as a cell that meets the following criteria: 1) Must have DAPI signal to be considered as a cell 2) Must have a peak for CD45 at 660 nm 3) Must not have one peak for CK at 525 nm 4) Must not have a peak for EpCAM at 617 nm. The spectral curves of an example CTC and an example WBC are shown in Figure 5-10A. A typical spectrum of CTCs has two distinct peaks, one for CK at 525 nm and one for EpCAM at 617 nm, whereas a typical spectrum of leukocyte has only one clear peak for CD45 at 660 nm. All+CTCWBC10 umCTCWBCAll+ CellA BDAPI CK EpCAM CD45 Figure 5-10 Merged images of a CTC, a leukocyte and an all-positive cell with their corresponding spectral curves. CTCs are defined as DAPI+ CK+ EpCAM+/- and CD45- while leukocytes are identified as DAPI+ CK- EpCAM- and CD45+. An all-positive cell displays all four positive signals. During experiments, we found cells that cannot be classified into these two groups only. A certain proportion of cells were found to be positive for all 4 immuno-stains in the patient samples, consistent with previously reported literature [51]. They have DAPI signal and have peaks for CK at 525 nm, for EpCAM at 617 nm and for CD45 at 660 nm. They are called the all-52 positive cells (Figure 5-10). As the nature of these cells is not yet established, they are not counted as CTC cells. The merged images of these cells look like potential CTCs. The only way we can tell them from CTCs is the presence of a CD45 peak at 660 nm. CD45 proteins are situated on the cell surface. By marking a big circle that surrounds the whole cell, the CD45 signal on the 2 dimensional edges of the cells will be averaged by the whole area of the cell. Sometimes the CD45 signal is sufficiently strong to show a visible peak when the spectrum is taken from the entire cell. Other times, the CD45 peak is weaker and can only be visible on the spectrum on the edge of the cell. In this case, on the whole cell spectrum, the weak CD45 peak may be covered by the relatively strong peak of EpCAM, which may lead us to falsely consider that specific cell as a CTC. This phenomenon makes it really hard to distinguish CTCs candidates as well as all-positive cells. To see the CD45 signals more clearly, we therefore decided to mark small circles around the edge of those cells. The steps used to identify CTCs include the following: 1. Look for green (CK) or dark red (EpCAM) fluorescence in merged images and then circle the whole cell to see its spectrum. 2. Check that the circled cells are DAPI+ CK+ EpCAM+/-. If there is obvious CD45 peak, this cell is not considered a CTC. If no obvious CD45 peak, continue to the next step. 3. Draw small circles around the edge of the suspicious cells to check the spectrum of these sub-cellular regions. If any of the spectrums have an obvious CD45 peak, this cell is also not considered as a CTC. If none of the spectrums has obvious CD45 peak, this cell is considered a CTC. The above steps can be simplified as the block diagram shown in Figure 5-11. DAPI+ CK+ and EpCAM+/- ?Any of the spectrums has obvious CD45 peak?Obvious CD45 peak?Sub-cell spectrums of the whole edge of the cell Non-CTCYesNo Whole cell spectrum of cellsith green or dark red color YesYesCTC candidateNoNo Figure 5-11 Block diagram of the process of identifying candidate CTCs and non-CTC cells. 53 We therefore classified cells into three groups: typical WBCs, all-positive cells and CTCs. Examples from each group are be shown in Figure 5-12, 5-13, 5-14 and 5-15. Cells in the same patient always have the same pattern for both all-positive cells and CTC candidates. They can look green and red (both strong signals for CK and EpCAM), or green (stronger CK than EpCAM) or dark red (stronger EpCAM than CK). Cells with the same numbering in Figure 5-13, 5-14 and 5-15 are from the same patients. Figure 5-12 Merged image and spectrum of a typical WBC. WBC is DAPI positive and CD45 positive. Figure 5-13 Merged images and spectrums of three easily distinguishable all-positive cells. These cells are DAPI CK, EpCAM positive and have an obvious CD45 peak when circling around the whole cell. 54 Figure 5-14 Merged images and spectrums of three all-positive cells that are hard to distinguish. These cells are DAPI CK, EpCAM positive and have no obvious CD45 peak when circling around the whole cell. Obvious CD45 peak shows up when circling small area on the edge of the cells. From top to bottom, the first cell has strong signals for both CK and EpCAM; the second cell has stronger signal for CK than EpCAM thus the whole cell looks more green; the third cell has stronger signal for CK than EpCAM thus the whole cell looks more dark red. 55 Figure 5-15 Merged images and spectrums of three CTC candidates. These cells are DAPI CK, EpCAM positive and have no obvious CD45 peak when circling around the whole cell and no obvious CD45 peak when circling small area on the edge of the cells. From top to bottom, the first cell has strong signals for both CK and EpCAM; the second cell has stronger signal for CK than EpCAM thus the whole cell looks more green; the third cell has stronger signal for CK than EpCAM thus the whole cell looks more dark red. 56 5.3.2 Enrichment and identification of candidate CTCs from patients with mCRPC Blood samples from 24 patients with mCRPC were processed using the RCT device. After processing, the CTCs were viable, unattached, and compatible with standard cellular analysis methods. For each patient, a parallel 7.5 ml of blood was analyzed using the Veridex CellSearch™ system. AB Figure 5-16 Enumeration of CTCs derived from CRPC patient samples and 5 control samples. A: Number of CTCs identified following resettable cell trap (RCT) or CellSearch® enrichment. D: Grouped results of RCT device and CellSearch System. Data is displayed with mean ± standard deviation. P value is calculated by parametric paired T-test analysis. 57 CTCs counts from the resettable cell trap device were scaled to numbers per 7.5 ml to compare with the CellSearch system. The RCT device identified 83.3% (20/24) patients as exhibiting >5 CTCs per 7.5ml of blood. The numbers varied between patients, from 0 to 2145, with a mean of 329 per 7.5ml of blood. In the same patient group, the CellSearch analysis revealed >=5 CTCs in 37.5% (9/24) patients. The numbers ranged from 0 to 281 with a mean of 23 CTCs per 7.5 ml of blood. Detailed counting results for each patient are shown in Appendix C1. Control samples from five healthy donors were processed with the RCT device. Scanned images of sorted cells from the healthy controls were mixed blindly with the images of patient samples and then counted. Among the five healthy blind tests, only one donor had a count of 1 CTC from 1 ml of blood processed. As shown in Figure 5-16, significantly more CTCs were identified using our approach as compared to the CellSearch platform (p=0.0056) showing that our RCT device is more sensitive. Our results are similar to other biomechanical (size and deformability) based separations which show that more CTCs are captured compared to the EpCAM affinity capture of the CellSearch System [52], [53]. The improved result may be attributed to two facts. First, as CellSearch is dependent on the expression of the epithelial-lineage marker EpCAM, it may not be able to identify CTCs with weak expression. During the spectral analysis of our study, we found high heterogeneity of expression levels (intensity of the spectrum) of markers for CTCs between patients. In some patient samples, EpCAM expression is weaker than CK while the opposite is true for other samples. There are also patient samples with both weak CK and weak EpCAM expression. Previous reports affirm heterogeneity of biological (expression level of surface antigens) properties existing in CTCs from diverse cancer origins, different subtypes and even the CTCs in the same patient[2], [54]. Down-regulation of EpCAM on cancer cells occurs during dissemination into the blood stream [55] where the epithelial-to-mesenchymal transition (EMT) triggers it [56]. For those samples with inadequate EpCAM expression, successful capture of CTCs may be impossible with the CellSearch system. 58 Second, improvement results obtained using our RCT device over CellSearch system is based on the microscope system used. The Zeiss system we used has a better sensitivity than that of the CellSearch system, which is sufficient to pick up weak expression of CK. This system is also accurate enough to automatically separate the overlapping curves of spectrum. The CellSearch system classifies captured cells based on an operators’ judgment of the fluorescent images. Operator inconsistencies of image interpretation may lead to CTC misclassification [57]. The operator might also mistakenly eliminate the correct CTCs because they either fail to identify weak CK signal or the emission spectrum of CK (576 nm) spill over into the CD45 (660 nm) signal such that only CD45 emits fluorescence. These results compete well with other reported methods that process CRPC samples and compare to CellSearch platform. Other methods are either based on EpCAM affinity capture [58]–[60] using EpCAM coated micro-structures which increase contact between CTC and surface and thereby improve efficiency or are label-free methods based on the physical properties of CTCs [61], [62] or even hybrid methods that combine both EpCAM affinity and the physical properties [63]. Unlike most label-free microfluidics chips, our RCT device can process whole blood samples with a dilution factor of only 2 [45], [64]. There is no further processing of the blood sample such as lysis of RBCs [51], [65] or fixation [61], [66] where the addition of chemical buffers might effect the viability of the CTCs. Captured CTCs are easily retrieved from the device in the collection reservoirs of the device for easy enumeration or further downstream analysis. 5.3.3 Enrichment and identification of candidates CTCs from patients with LPC As mCRPC refers to prostate cancer that is resistant to medical or surgical treatments, and has spread to other parts of the body, localized prostate cancer (LPC) refers to prostate cancer that is diagnosed by a biopsy and appears to be only located completely inside the prostate gland. Blood samples from 18 patients with LPC were processed using the RCT device using the same processing set-ups and downstream spectral analysis as used for the CRPC samples. For each 59 patient, a parallel sample of 1 or 2 ml of blood was analyzed using the microfluidic ratchet device as stated in section 1.4. CTC candidate counts from both devices are shown and compared in Figure 5-17 below. AB Figure 5-17 Enumeration of CTCs derived from LPC patient samples processed by both resettable cell trap (RCT) device and microfluidic ratchet device. 60 After processing, sorted cells from both devices were stained and scanned using the same exact staining protocol and scanning procedure with the same microscope system. As shown in Figure 5-17, the RCT mechanism tends to capture more CTCs than the microfluidic ratchet device indicating that the RCT mechanism is generally more sensitive. Detailed counting results for each patient are shown in Appendix C2. This difference arises from differences in the principle of mechanisms. For the RCT device, when processing patient samples sufficient pressure is used to inflate the diaphragm to be in contact with the center fin which makes the opening size of the channel approximately 5 um. For the ratchet device, the constriction between each tapered shaped post is designed to be approximately 6 um. When the constriction size is smaller, it is instinctive that this filter will have better yield performance in order to trap smaller CTCs. However, the improved sensitivity is actually achieved by sacrificing the selectivity. We did observe more contamination of WBCs in the sample processed by the RCT device than that of the microfluidic ratchet device. The counts of contamination WBCs left over after processing for each sample are shown in Figure 5-18 in the same order as the samples shown in Figure 5-17 (CTC counts from high to low.) 61 5 7 3 2 5 1 5 4 4 1 3 8 4 0 5 2 1 8 2 4 3 5 4 9 2 1 2 8 4 3 4 6 5 0 5 3 C 1 C 2 C 3 C 4 C 505 0 0 01 0 0 0 01 5 0 0 02 0 0 0 02 5 0 0 0S a m p le ID N O .WBC counts after prococessing Figure 5-18 Counts of contamination WBCs left over after processing for each sample. There are 18 patient samples and 5 control samples. For each sample, 1 ml blood was processed. To count the total number of cells in each well, cell numbers in three 500 umX500 um squares were counted manually and then scaled to the whole area of the round well with 3.67 mm diameter. Only 66.7% (12/18) of the patient samples were identified to have CTCs with an average CTC count of 10.6 per sample (80 CTCs per 7.5ml blood). This is much less than the average count from mCRPC patient samples (329 per 7.5 ml blood). Though circumstances may change among patients, there is a sense that CTC enumeration may correlate with different stages of cancers. 62 Chapter 6: Conclusion Immunoaffinity is the conventional method to isolate CTCs with relatively high enrichment. However, due to the heterogeneous expression of EpCAM on CTCs within even the same patient, this method may fail to capture CTCs with weak expression below a certain threshold. As a label-free method, micropore filtration presents a compelling and more sensitive alternative that enriches for CTCs using size and deformability characteristics rather than biochemical characteristics. But a key challenge for this method is clogging. Additionally, a much purer sample is always demanded for downstream analysis. We improved a previously reported RCT cell separation mechanism which is capable of avoiding clogging by periodically resetting the microstructures. On-chip multi-filtration is integrated in this new version to achieve higher selectivity. Also, throughput is improved by multiplexing four single units. This new chip demonstrated successful high-sensitivity separation of CTCs from 2x diluted whole blood from patients with mCRPC and LPC. The RCT device positively identified more mCRPC patients based on the clinically defined threshold of >=5 CTC per 7.5 ml than the CellSearch system. This method also identified significant numbers of CTCs in some of the patients where the CellSearch system found zero CTCs. Finally, the cost for the CellSearch system is high because of capital investment, reagent consumption and specialist training [64]. Our RCT microfluidic device is more sensitive and potentially less expensive than the CellSearch system and can detect and trap CTCs with low EpCAM expression as well as identify patients with low CTC burden. 63 References [1] L. Norton and J. Massagué, “Is cancer a disease of self-seeding?,” Nat. Med., vol. 12, no. 8, pp. 875–878, Aug. 2006. [2] A. van de Stolpe, K. Pantel, S. Sleijfer, L. W. Terstappen, and J. M. J. den Toonder, “Circulating tumor cell isolation and diagnostics: toward routine clinical use,” Cancer Res., vol. 71, no. 18, pp. 5955–5960, Sep. 2011. [3] E. S. Lianidou, A. Strati, and A. Markou, “Circulating tumor cells as promising novel biomarkers in solid cancers,” Crit. Rev. Clin. Lab. Sci., vol. 51, no. 3, pp. 160–171, Jun. 2014. [4] X.-L. Ma, Y.-Y. Li, J. Zhang, J.-W. Huang, H.-Y. Jia, L. Liu, and P. Li, “Prognostic Role of Circulating Tumor Cells in Patients with Pancreatic Cancer: a Meta-analysis,” Asian Pac. J. Cancer Prev. APJCP, vol. 15, no. 15, pp. 6015–6020, 2014. [5] M. Wallwiener, S. Riethdorf, A. D. Hartkopf, C. Modugno, J. Nees, D. Madhavan, M. R. Sprick, S. Schott, C. Domschke, I. Baccelli, B. Schönfisch, B. Burwinkel, F. Marmé, J. Heil, C. Sohn, K. Pantel, A. Trumpp, and A. Schneeweiss, “Serial enumeration of circulating tumor cells predicts treatment response and prognosis in metastatic breast cancer: a prospective study in 393 patients,” BMC Cancer, vol. 14, no. 1, p. 512, 2014. [6] A. Kulasinghe, C. Perry, L. Jovanovic, C. Nelson, and C. Punyadeera, “Circulating tumour cells in metastatic head and neck cancers,” Int. J. Cancer J. Int. Cancer, Aug. 2014. [7] F.-B. Wang, X.-Q. Yang, S. Yang, B.-C. Wang, M.-H. Feng, and J.-C. Tu, “A higher number of circulating tumor cells (CTC) in peripheral blood indicates poor prognosis in prostate cancer patients--a meta-analysis,” Asian Pac. J. Cancer Prev. APJCP, vol. 12, no. 10, pp. 2629–2635, 2011. [8] G. Lurje, M. Schiesser, A. Claudius Hoffmann, and P. M. Schneider, “Circulating Tumor Cells in Gastrointestinal Malignancies: Current Techniques and Clinical Implications,” J. Oncol., vol. 2010, p. e392652, Nov. 2009. [9] C. Jin, S. M. McFaul, S. P. Duffy, X. Deng, P. Tavassoli, P. C. Black, and H. Ma, “Technologies for label-free separation of circulating tumor cells: from historical foundations to recent developments,” Lab. Chip, vol. 14, no. 1, pp. 32–44, Jan. 2014. [10] T. M. Gorges, I. Tinhofer, M. Drosch, L. Röse, T. M. Zollner, T. Krahn, and O. von Ahsen, “Circulating tumour cells escape from EpCAM-based detection due to epithelial-to-mesenchymal transition,” BMC Cancer, vol. 12, p. 178, 2012. [11] V. Murlidhar, M. Zeinali, S. Grabauskiene, M. Ghannad-Rezaie, M. S. Wicha, D. M. Simeone, N. Ramnath, R. M. Reddy, and S. Nagrath, “A Radial Flow Microfluidic Device for Ultra-High-Throughput Affinity-Based Isolation of Circulating Tumor Cells,” Small, p. n/a–n/a, Jul. 2014. [12] C. Huang, S. M. Santana, H. Liu, N. H. Bander, B. G. Hawkins, and B. J. Kirby, “Characterization of a hybrid dielectrophoresis and immunocapture microfluidic system for cancer cell capture,” ELECTROPHORESIS, vol. 34, no. 20–21, pp. 2970–2979, Nov. 2013. [13] S. Miltenyi, W. Müller, W. Weichel, and A. Radbruch, “High gradient magnetic cell separation with MACS,” Cytometry, vol. 11, no. 2, pp. 231–238, Jan. 1990. [14] J. P. Gleghorn, E. D. Pratt, D. Denning, H. Liu, N. H. Bander, S. T. Tagawa, D. M. Nanus, P. A. Giannakakou, and B. J. Kirby, “Capture of circulating tumor cells from whole blood of prostate cancer patients using geometrically enhanced differential immunocapture (GEDI) and a prostate-specific antibody,” Lab. Chip, vol. 10, no. 1, pp. 27–29, Jan. 2010. 64 [15] S. Nagrath, L. V. Sequist, S. Maheswaran, D. W. Bell, D. Irimia, L. Ulkus, M. R. Smith, E. L. Kwak, S. Digumarthy, A. Muzikansky, P. Ryan, U. J. Balis, R. G. Tompkins, D. A. Haber, and M. Toner, “Isolation of rare circulating tumour cells in cancer patients by microchip technology,” Nature, vol. 450, no. 7173, pp. 1235–1239, Dec. 2007. [16] S. Riethdorf, H. Fritsche, V. Müller, T. Rau, C. Schindlbeck, B. Rack, W. Janni, C. Coith, K. Beck, and F. Jänicke, “Detection of circulating tumor cells in peripheral blood of patients with metastatic breast cancer: a validation study of the CellSearch system,” Clin. Cancer Res., vol. 13, no. 3, pp. 920–928, 2007. [17] M. Lin, J.-F. Chen, Y.-T. Lu, Y. Zhang, J. Song, S. Hou, Z. Ke, and H.-R. Tseng, “Nanostructure Embedded Microchips for Detection, Isolation, and Characterization of Circulating Tumor Cells,” Acc. Chem. Res., Aug. 2014. [18] A. A. Powell, A. H. Talasaz, H. Zhang, M. A. Coram, A. Reddy, G. Deng, M. L. Telli, R. H. Advani, R. W. Carlson, J. A. Mollick, S. Sheth, A. W. Kurian, J. M. Ford, F. E. Stockdale, S. R. Quake, R. F. Pease, M. N. Mindrinos, G. Bhanot, S. H. Dairkee, R. W. Davis, and S. S. Jeffrey, “Single cell profiling of circulating tumor cells: transcriptional heterogeneity and diversity from breast cancer cell lines,” PloS One, vol. 7, no. 5, p. e33788, 2012. [19] N. Krawczyk, F. Meier-Stiegen, M. Banys, H. Neubauer, E. Ruckhaeberle, and T. Fehm, “Expression of stem cell and epithelial-mesenchymal transition markers in circulating tumor cells of breast cancer patients,” BioMed Res. Int., vol. 2014, p. 415721, 2014. [20] D. F. Hayes, M. Cristofanilli, G. T. Budd, M. J. Ellis, A. Stopeck, M. C. Miller, J. Matera, W. J. Allard, G. V. Doyle, and L. W. Terstappen, “Circulating tumor cells at each follow-up time point during therapy of metastatic breast cancer patients predict progression-free and overall survival,” Clin. Cancer Res., vol. 12, no. 14, pp. 4218–4224, 2006. [21] J. S. de Bono, H. I. Scher, R. B. Montgomery, C. Parker, M. C. Miller, H. Tissing, G. V. Doyle, L. W. Terstappen, K. J. Pienta, and D. Raghavan, “Circulating tumor cells predict survival benefit from treatment in metastatic castration-resistant prostate cancer,” Clin. Cancer Res., vol. 14, no. 19, pp. 6302–6309, 2008. [22] S. J. Cohen, C. J. Punt, N. Iannotti, B. H. Saidman, K. D. Sabbath, N. Y. Gabrail, J. Picus, M. Morse, E. Mitchell, and M. C. Miller, “Relationship of circulating tumor cells to tumor response, progression-free survival, and overall survival in patients with metastatic colorectal cancer,” J. Clin. Oncol., vol. 26, no. 19, pp. 3213–3221, 2008. [23] I. Cima, C. Wen Yee, F. S. Iliescu, W. Min Phyo, K. Hon Lim, C. Iliescu, and M. Han Tan, “Label-free isolation of circulating tumor cells in microfluidic devices: Current research and perspectives,” Biomicrofluidics, vol. 7, no. 1, p. 11810, 2013. [24] X.-B. Wang, J. Yang, Y. Huang, J. Vykoukal, F. F. Becker, and P. R. C. Gascoyne, “Cell Separation by Dielectrophoretic Field-flow-fractionation,” Anal. Chem., vol. 72, no. 4, pp. 832–839, Feb. 2000. [25] A. Alazzam, I. Stiharu, R. Bhat, and A.-N. Meguerditchian, “Interdigitated comb-like electrodes for continuous separation of malignant cells from blood using dielectrophoresis,” Electrophoresis, vol. 32, no. 11, pp. 1327–1336, Jun. 2011. [26] S. Shim, K. Stemke-Hale, A. M. Tsimberidou, J. Noshari, T. E. Anderson, and P. R. C. Gascoyne, “Antibody-independent isolation of circulating tumor cells by continuous-flow dielectrophoresis,” Biomicrofluidics, vol. 7, no. 1, Jan. 2013. 65 [27] X. Ding, Z. Peng, S.-C. S. Lin, M. Geri, S. Li, P. Li, Y. Chen, M. Dao, S. Suresh, and T. J. Huang, “Cell separation using tilted-angle standing surface acoustic waves,” Proc. Natl. Acad. Sci., vol. 111, no. 36, pp. 12992–12997, Sep. 2014. [28] P. Augustsson, C. Magnusson, M. Nordin, H. Lilja, and T. Laurell, “Microfluidic, Label-Free Enrichment of Prostate Cancer Cells in Blood Based on Acoustophoresis,” Anal. Chem., vol. 84, no. 18, pp. 7954–7962, Sep. 2012. [29] J. A. Davis, D. W. Inglis, K. J. Morton, D. A. Lawrence, L. R. Huang, S. Y. Chou, J. C. Sturm, and R. H. Austin, “Deterministic hydrodynamics: taking blood apart,” Proc. Natl. Acad. Sci. U. S. A., vol. 103, no. 40, pp. 14779–14784, Oct. 2006. [30] S. S. Kuntaegowdanahalli, A. A. S. Bhagat, G. Kumar, and I. Papautsky, “Inertial microfluidics for continuous particle separation in spiral microchannels,” Lab. Chip, vol. 9, no. 20, pp. 2973–2980, Oct. 2009. [31] J. Sun, M. Li, C. Liu, Y. Zhang, D. Liu, W. Liu, G. Hu, and X. Jiang, “Double spiral microchannel for label-free tumor cell separation and enrichment,” Lab. Chip, vol. 12, no. 20, p. 3952, 2012. [32] S. J. Tan, R. L. Lakshmi, P. Chen, W.-T. Lim, L. Yobas, and C. T. Lim, “Versatile label free biochip for the detection of circulating tumor cells from peripheral blood in cancer patients,” Biosens. Bioelectron., vol. 26, no. 4, pp. 1701–1705, Dec. 2010. [33] S. Zheng, H. Lin, J.-Q. Liu, M. Balic, R. Datar, R. J. Cote, and Y.-C. Tai, “Membrane microfilter device for selective capture, electrolysis and genomic analysis of human circulating tumor cells,” J. Chromatogr. A, vol. 1162, no. 2, pp. 154–161, Aug. 2007. [34] W. Beattie, X. Qin, L. Wang, and H. Ma, “Clog-free cell filtration using resettable cell traps,” Lab. Chip, vol. 14, no. 15, p. 2657, 2014. [35] M. A. Unger, H. P. Chou, T. Thorsen, A. Scherer, and S. R. Quake, “Monolithic microfabricated valves and pumps by multilayer soft lithography,” Science, vol. 288, no. 5463, pp. 113–116, Apr. 2000. [36] S.-B. Huang, M.-H. Wu, and G.-B. Lee, “A tunable micro filter modulated by pneumatic pressure for cell separation,” Sens. Actuators B Chem., vol. 142, no. 1, pp. 389–399, Oct. 2009. [37] L. Saias, J. Autebert, L. Malaquin, and J.-L. Viovy, “Design, modeling and characterization of microfluidic architectures for high flow rate, small footprint microfluidic systems,” Lab. Chip, vol. 11, no. 5, pp. 822–832, Mar. 2011. [38] A. Jain and L. L. Munn, “Determinants of Leukocyte Margination in Rectangular Microchannels,” PLoS ONE, vol. 4, no. 9, p. e7104, Sep. 2009. [39] S. K. Murthy, A. Sin, R. G. Tompkins, and M. Toner, “Effect of flow and surface conditions on human lymphocyte isolation using microfluidic chambers,” Langmuir ACS J. Surf. Colloids, vol. 20, no. 26, pp. 11649–11655, Dec. 2004. [40] T. Gerhardt, S. Woo, and H. Ma, “Chromatographic behaviour of single cells in a microchannel with dynamic geometry,” Lab. Chip, vol. 11, no. 16, pp. 2731–2737, Aug. 2011. [41] M. Zecchino, W. Mozer, and K. Loken, “Accurate Thick Film Measurement with Optical Profiling,” Veeco Instrum. Inc. [42] S. W. Lee and S. S. Lee, “Shrinkage ratio of PDMS and its alignment method for the wafer level process,” Microsyst. Technol., vol. 14, no. 2, pp. 205–208, Feb. 2008. 66 [43] D. C. Duffy, J. C. McDonald, O. J. A. Schueller, and G. M. Whitesides, “Rapid Prototyping of Microfluidic Systems in Poly(dimethylsiloxane),” Anal. Chem., vol. 70, no. 23, pp. 4974–4984, Dec. 1998. [44] H.-S. Moon, K. Kwon, K.-A. Hyun, T. Seok Sim, J. Chan Park, J.-G. Lee, and H.-I. Jung, “Continual collection and re-separation of circulating tumor cells from blood using multi-stage multi-orifice flow fractionation,” Biomicrofluidics, vol. 7, no. 1, Jan. 2013. [45] S. Zheng, H. K. Lin, B. Lu, A. Williams, R. Datar, R. J. Cote, and Y.-C. Tai, “3D microfilter device for viable circulating tumor cell (CTC) enrichment from blood,” Biomed. Microdevices, vol. 13, no. 1, pp. 203–213, Feb. 2011. [46] S. C. Hur, A. J. Mach, and D. Di Carlo, “High-throughput size-based rare cell enrichment using microscale vortices,” Biomicrofluidics, vol. 5, no. 2, Jun. 2011. [47] V. Swaminathan, K. Mythreye, E. T. O’Brien, A. Berchuck, G. C. Blobe, and R. Superfine, “Mechanical stiffness grades metastatic potential in patient tumor cells and in cancer cell lines,” Cancer Res., vol. 71, no. 15, pp. 5075–5080, Aug. 2011. [48] J. Guck, S. Schinkinger, B. Lincoln, F. Wottawah, S. Ebert, M. Romeyke, D. Lenz, H. M. Erickson, R. Ananthakrishnan, D. Mitchell, J. Kas, S. Ulvick, and C. Bilby, “Optical Deformability as an Inherent Cell Marker for Testing Malignant Transformation and Metastatic Competence,” Biophys. J., vol. 88, no. 5, pp. 3689–3698, May 2005. [49] Q. Guo, S. Park, and H. Ma, “Microfluidic micropipette aspiration for measuring the deformability of single cells,” Lab. Chip, vol. 12, no. 15, pp. 2687–2695, Aug. 2012. [50] Q. Guo, S. Park, and H. Ma, “Microfluidic micropipette aspiration for measuring the deformability of single cells,” Lab. Chip, vol. 12, no. 15, p. 2687, 2012. [51] B. L. Khoo, M. E. Warkiani, D. S.-W. Tan, A. A. S. Bhagat, D. Irwin, D. P. Lau, A. S. T. Lim, K. H. Lim, S. S. Krisna, W.-T. Lim, Y. S. Yap, S. C. Lee, R. A. Soo, J. Han, and C. T. Lim, “Clinical validation of an ultra high-throughput spiral microfluidics for the detection and enrichment of viable circulating tumor cells,” PloS One, vol. 9, no. 7, p. e99409, 2014. [52] S. J. Tan, R. L. Lakshmi, P. Chen, W.-T. Lim, L. Yobas, and C. T. Lim, “Versatile label free biochip for the detection of circulating tumor cells from peripheral blood in cancer patients,” Biosens. Bioelectron., vol. 26, no. 4, pp. 1701–1705, Dec. 2010. [53] H. K. Lin, S. Zheng, A. J. Williams, M. Balic, S. Groshen, H. I. Scher, M. Fleisher, W. Stadler, R. H. Datar, Y.-C. Tai, and R. J. Cote, “Portable Filter-Based Microdevice for Detection and Characterization of Circulating Tumor Cells,” Clin. Cancer Res., vol. 16, no. 20, pp. 5011–5018, Oct. 2010. [54] A. M. Sieuwerts, J. Kraan, J. Bolt, P. van der Spoel, F. Elstrodt, M. Schutte, J. W. M. Martens, J.-W. Gratama, S. Sleijfer, and J. A. Foekens, “Anti-Epithelial Cell Adhesion Molecule Antibodies and the Detection of Circulating Normal-Like Breast Tumor Cells,” J. Natl. Cancer Inst., vol. 101, no. 1, pp. 61–66, Jan. 2009. [55] C. G. Rao, D. Chianese, G. V. Doyle, M. C. Miller, T. Russell, R. A. Sanders, and L. W. M. M. Terstappen, “Expression of epithelial cell adhesion molecule in carcinoma cells present in blood and primary and metastatic tumors,” Int. J. Oncol., vol. 27, no. 1, pp. 49–57, Jul. 2005. [56] O. Gires and N. H. Stoecklein, “Dynamic EpCAM expression on circulating and disseminating tumor cells: causes and consequences,” Cell. Mol. Life Sci. CMLS, Aug. 2014. [57] J. Kraan, S. Sleijfer, M. H. Strijbos, M. Ignatiadis, D. Peeters, J.-Y. Pierga, F. Farace, S. Riethdorf, T. Fehm, L. Zorzino, A. G. J. Tibbe, M. Maestro, R. Gisbert-Criado, G. Denton, 67 J. S. de Bono, C. Dive, J. A. Foekens, and J. W. Gratama, “External quality assurance of circulating tumor cell enumeration using the CellSearch® system: A feasibility study,” Cytometry B Clin. Cytom., vol. 80B, no. 2, pp. 112–118, Mar. 2011. [58] S. Wang, K. Liu, J. Liu, Z. T.-F. Yu, X. Xu, L. Zhao, T. Lee, E. K. Lee, J. Reiss, Y.-K. Lee, L. W. K. Chung, J. Huang, M. Rettig, D. Seligson, K. N. Duraiswamy, C. K.-F. Shen, and H.-R. Tseng, “Highly Efficient Capture of Circulating Tumor Cells by Using Nanostructured Silicon Substrates with Integrated Chaotic Micromixers,” Angew. Chem. Int. Ed., vol. 50, no. 13, pp. 3084–3088, Mar. 2011. [59] T. W. Friedlander, V. T. Ngo, H. Dong, G. Premasekharan, V. Weinberg, S. Doty, Q. Zhao, E. G. Gilbert, C. J. Ryan, W.-T. Chen, and P. L. Paris, “Detection and characterization of invasive circulating tumor cells (ictcs) derived from men with metastatic castration resistant prostate cancer (mCRPC),” Int. J. Cancer, p. n/a–n/a, 2013. [60] S. L. Stott, C.-H. Hsu, D. I. Tsukrov, M. Yu, D. T. Miyamoto, B. A. Waltman, S. M. Rothenberg, A. M. Shah, M. E. Smas, G. K. Korir, F. P. Floyd, A. J. Gilman, J. B. Lord, D. Winokur, S. Springer, D. Irimia, S. Nagrath, L. V. Sequist, R. J. Lee, K. J. Isselbacher, S. Maheswaran, D. A. Haber, and M. Toner, “Isolation of circulating tumor cells using a microvortex-generating herringbone-chip,” Proc. Natl. Acad. Sci., vol. 107, no. 43, pp. 18392–18397, Oct. 2010. [61] H. K. Lin, S. Zheng, A. J. Williams, M. Balic, S. Groshen, H. I. Scher, M. Fleisher, W. Stadler, R. H. Datar, Y.-C. Tai, and R. J. Cote, “Portable filter-based microdevice for detection and characterization of circulating tumor cells,” Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res., vol. 16, no. 20, pp. 5011–5018, Oct. 2010. [62] F. Farace, C. Massard, N. Vimond, F. Drusch, N. Jacques, F. Billiot, A. Laplanche, A. Chauchereau, L. Lacroix, D. Planchard, S. Le Moulec, F. André, K. Fizazi, J. C. Soria, and P. Vielh, “A direct comparison of CellSearch and ISET for circulating tumour-cell detection in patients with metastatic carcinomas,” Br. J. Cancer, vol. 105, no. 6, pp. 847–853, Sep. 2011. [63] E. Ozkumur, A. M. Shah, J. C. Ciciliano, B. L. Emmink, D. T. Miyamoto, E. Brachtel, M. Yu, P. Chen, B. Morgan, J. Trautwein, A. Kimura, S. Sengupta, S. L. Stott, N. M. Karabacak, T. A. Barber, J. R. Walsh, K. Smith, P. S. Spuhler, J. P. Sullivan, R. J. Lee, D. T. Ting, X. Luo, A. T. Shaw, A. Bardia, L. V. Sequist, D. N. Louis, S. Maheswaran, R. Kapur, D. A. Haber, and M. Toner, “Inertial Focusing for Tumor Antigen–Dependent and –Independent Sorting of Rare Circulating Tumor Cells,” Sci. Transl. Med., vol. 5, no. 179, pp. 179ra47–179ra47, Apr. 2013. [64] E. Sollier, D. E. Go, J. Che, D. Gossett, S. O’Byrne, W. M. Weaver, N. Kummer, M. Rettig, J. Goldman, N. Nickols, S. McCloskey, R. Kulkarni, and D. D. Carlo, “Size-Selective Collection of Circulating Tumor Cells using Vortex Technology,” Lab. Chip, Aug. 2013. [65] L. Chinen, C. Mello, E. Abdallah, L. Ocea, M. Buim, N. Mingues, J. Gasparini Junior, M. Fanelli, and P. Paterlini, “Isolation, detection, and immunomorphological characterization of circulating tumor cells (CTCs) from patients with different types of sarcoma using isolation by size of tumor cells: a window on sarcoma-cell invasion,” OncoTargets Ther., p. 1609, Sep. 2014. [66] R. Riahi, P. Gogoi, S. Sepehri, Y. Zhou, K. Handique, J. Godsey, and Y. Wang, “A novel microchannel-based device to capture and analyze circulating tumor cells (CTCs) of breast cancer,” Int. J. Oncol., vol. 44, no. 6, pp. 1870–1878, Jun. 2014. 68 [67] D. R. Gossett, W. M. Weaver, A. J. Mach, S. C. Hur, H. T. K. Tse, W. Lee, H. Amini, and D. Di Carlo, “Label-free cell separation and sorting in microfluidic systems,” Anal. Bioanal. Chem., vol. 397, no. 8, pp. 3249–3267, Aug. 2010. 69 Appendices Appendix A : Automatic software code The following C# code was created to give command to the MSP430 microprocessor to control the solenoid valves on the custom pressure board. using System; using System.Collections.Generic; using System.ComponentModel; using System.Data; using System.Drawing; using System.Linq; using System.Text; using System.Windows.Forms; using System.Diagnostics; using System.Threading; namespace WindowsFormsApplication1 { public partial class Form1 : Form { public List ValveList; public List StatusStripList; public bool[] valveState; public Stopwatch timeMaster; public Stopwatch totalTimer; public int currentCycle; public int valveIndex; public List ModeList; public Form1() { InitializeComponent(); startFormInterface(); timeMaster = new Stopwatch(); totalTimer = new Stopwatch(); System.Diagnostics.Process.GetCurrentProcess().PriorityClass = ProcessPriorityClass.High; ValveList = new List(); StatusStripList = new List(); valveState = new bool[18]; currentCycle = 1; valveIndex = 0; btnStart.Enabled = false; tmrController.Enabled = false; ModeList= new List(){"Processing","Purging","Collecting"}; tspbProcessBar.Maximum = System.Convert.ToInt32(System.Convert.ToDouble(txtProcessTime.Text) * 60000); 70 tspbPurgeBar.Maximum = System.Convert.ToInt32(System.Convert.ToDouble(txtPurgeTime.Text) * 60000); tspbCollectBar.Maximum = System.Convert.ToInt32(System.Convert.ToDouble(txtCollectTime.Text) * 60000); tspbProcessBar.Value = System.Convert.ToInt32(System.Convert.ToDouble(txtProcessTime.Text) * 60000); tspbPurgeBar.Value = System.Convert.ToInt32(System.Convert.ToDouble(txtPurgeTime.Text) * 60000); tspbCollectBar.Value = System.Convert.ToInt32(System.Convert.ToDouble(txtCollectTime.Text) * 60000); ValveList.Add(PortA0); ValveList.Add(PortA1); ValveList.Add(PortA6); ValveList.Add(PortA3); ValveList.Add(PortA4); ValveList.Add(PortA5); ValveList.Add(PortB0); ValveList.Add(PortB1); ValveList.Add(PortB2); ValveList.Add(PortB3); ValveList.Add(PortC0); ValveList.Add(PortC1); ValveList.Add(PortC2); ValveList.Add(PortC3); ValveList.Add(PortC4); ValveList.Add(PortC5); ValveList.Add(PortC6); ValveList.Add(PortC7); StatusStripList.Add(statusStripProcess); StatusStripList.Add(statusStripPurge); StatusStripList.Add(statusStripCollect); foreach (CheckBox valveHandlerToSet in ValveList) { valveHandlerToSet.Click += new System.EventHandler(this.valveCheckBoxClick); valveHandlerToSet.Enabled = false; } reset_allvalves(); } private void activeValve(char portName, char portNumber, int index)//open Port { serCOM.Write(new byte[] { 255, (byte)'1', (byte)portName, (byte)portNumber }, 0, 4); //resetAll valveState[index] = true; ValveList[index].BackColor = System.Drawing.Color.Pink; ValveList[index].Checked = true; } private void releaseValve(char portName, char portNumber,int index)// close port 71 { serCOM.Write(new byte[] { 255, (byte)'2', (byte)portName, (byte)portNumber }, 0, 4); //resetAll valveState[index] = false; ValveList[index].BackColor = System.Drawing.Color.LightGreen; ValveList[index].Checked = false; } private void startFormInterface() { updateComList(); } private void updateComList() { cmbPortSelect.Items.Clear(); cmbPortSelect.Items.AddRange(System.IO.Ports.SerialPort.GetPortNames().ToArray()); cmbPortSelect.SelectedIndex = 0; } private void cmbPortSelect_DropDown(object sender, EventArgs e) { updateComList(); } private void btnPortConnect_Click(object sender, EventArgs e) { if (serCOM.IsOpen == false) { btnPortConnect.Text = "Disconnect"; serCOM.PortName = cmbPortSelect.Text; try { serCOM.Open(); // tmrController.Enabled = true; //reset_allvalves(); timeMaster.Reset(); totalTimer.Reset(); this.tsslMainStatus.Text = "Serial Port Opened."; this.statusStripMain.BackColor = System.Drawing.Color.LightBlue; btnStart.Enabled = true; foreach (Control controlColorSet in ValveList) { controlColorSet.Enabled = true; controlColorSet.BackColor = System.Drawing.Color.Transparent; } } catch (Exception b) { this.tsslCollectStatus.Text = "ERROR, " + b.Message; this.statusStripMain.BackColor = System.Drawing.Color.Red; 72 this.btnPortConnect.Text = "Connect/Disconnect"; foreach (Control controlColorSet in ValveList) { controlColorSet.Enabled = false; controlColorSet.BackColor = System.Drawing.Color.Yellow; } } } else { try { //reset_allvalves(); tmrController.Enabled = false; serCOM.Close(); timeMaster.Stop(); totalTimer.Stop(); this.tsslMainStatus.Text = "Serial Port Closed, Board Reset (" + (System.Convert.ToDouble(totalTimer.ElapsedMilliseconds / 60000).ToString())+ "mins)"; this.statusStripMain.BackColor = System.Drawing.Color.LightGray; foreach (Control controlColorSet in ValveList) { controlColorSet.Enabled = false; controlColorSet.BackColor = System.Drawing.Color.Transparent; } } catch (Exception b) { this.tsslCollectStatus.Text = "ERROR, " + b.Message; this.statusStripMain.BackColor = System.Drawing.Color.Red; this.btnPortConnect.Text = "Connect/Disconnect"; foreach (Control controlColorSet in ValveList) { controlColorSet.Enabled = false; controlColorSet.BackColor = System.Drawing.Color.Yellow; } } btnPortConnect.Text = "Connect"; } } private void reset_allvalves() { if (serCOM.IsOpen == true) { serCOM.Write(new byte[] { 255, (byte)'7', (byte)'0', (byte)'0' }, 0, 4); //resetAll for (int i = 0; i < 17; i++) { valveState[i] = false; } 73 foreach (Control controlColorSet in ValveList) { controlColorSet.Enabled = true; controlColorSet.BackColor = System.Drawing.Color.Transparent; } } } private void active_deviceValves() { if (serCOM.IsOpen == true) { activeValve('A', '0', 0); activeValve('A', '1', 1); activeValve('A', '6', 2); activeValve('A', '3', 3); activeValve('A', '4', 4); activeValve('B', '0', 6); } } private void btnStart_Click(object sender, EventArgs e) { if(btnStart.Text == "Start/Resume") { btnStart.Text = "Running... Stop"; active_deviceValves(); btnReset.Enabled = false; tmrController.Enabled = true; timeMaster.Start(); totalTimer.Start(); cmbPortSelect.Enabled = false; //btnSearch.Enabled = false; btnPortConnect.Enabled = false; txtProcessTime.Enabled = false; txtPurgeTime.Enabled = false; txtCollectTime.Enabled = false; btnStart.ForeColor = Color.Green; this.tsslMainStatus.Text = "Start running"; this.statusStripMain.BackColor = System.Drawing.Color.Green; foreach (Control controlColorSet in ValveList) { //controlColorSet.BackColor = System.Drawing.Color.Transparent; controlColorSet.Enabled = false; } switch (mode) { case 0: startProcessCycle(); 74 break; case 1: startPurgeCycle(); break; case 2: startCollectCycle(); break; } } else { active_deviceValves(); btnStart.Text = "Start/Resume"; btnReset.Enabled = true; tmrController.Enabled = false; timeMaster.Stop(); totalTimer.Stop(); cmbPortSelect.Enabled = true; //btnSearch.Enabled = true; btnPortConnect.Enabled = true; txtProcessTime.Enabled = true; txtPurgeTime.Enabled = true; txtCollectTime.Enabled = true; btnStart.ForeColor = Color.Black; foreach (Control controlColorSet in ValveList) { //controlColorSet.BackColor = System.Drawing.Color.Transparent; controlColorSet.Enabled = true; } this.tsslMainStatus.Text = "Stopped. Current " + (System.Convert.ToDouble(timeMaster.ElapsedMilliseconds / 1000).ToString()) + "s elapsed. Total " + (System.Convert.ToDouble(totalTimer.ElapsedMilliseconds / 60000).ToString()) + "mins elapsed."; this.statusStripMain.BackColor = System.Drawing.Color.Pink; } } private void btnReset_Click(object sender, EventArgs e) { tspbProcessBar.Maximum = System.Convert.ToInt32(System.Convert.ToDouble(txtProcessTime.Text) * 60000); tspbPurgeBar.Maximum = System.Convert.ToInt32(System.Convert.ToDouble(txtPurgeTime.Text) * 60000); tspbCollectBar.Maximum = System.Convert.ToInt32(System.Convert.ToDouble(txtCollectTime.Text) * 60000); tspbProcessBar.Value = System.Convert.ToInt32(System.Convert.ToDouble(txtProcessTime.Text) * 60000); tspbPurgeBar.Value = System.Convert.ToInt32(System.Convert.ToDouble(txtPurgeTime.Text) * 60000); 75 tspbCollectBar.Value = System.Convert.ToInt32(System.Convert.ToDouble(txtCollectTime.Text) * 60000); timeMaster.Reset(); // mode = 0; this.tsslMainStatus.Text = "Parameters reset."; } private void startProcessCycle() { active_deviceValves(); timeMaster.Stop(); tmrController.Enabled = false; releaseValve('A', '4', 4); activeValve('B','0',6); // trap1 timerDelay.Enabled = true; timerDelay.Interval = 5000; } private void startPurgeCycle() { timeMaster.Stop(); tmrController.Enabled = false; //closePort('B', '0'); // trap1 //delay here activeValve('A', '1',1); timerDelay.Enabled = true; timerDelay.Interval = 2000; } private void startCollectCycle() { timeMaster.Stop(); tmrController.Enabled = false; activeValve('A', '4', 4); // activeValve('A', '0', 0); releaseValve('B', '0', 6);//trap1, delay here timerDelay.Enabled = true; timerDelay.Interval = 5000; } private void timerDelay_Tick(object sender, EventArgs e) { switch (mode) { case 0: releaseValve('A', '1', 1); break; case 1: releaseValve('A', '0', 0); break; case 2: releaseValve('A', '3', 3); 76 releaseValve('A', '2', 2); break; } timerDelay.Enabled = false; timeMaster.Start(); tmrController.Enabled = true; } private void valveCheckBoxClick(object sender, EventArgs e) { if (serCOM.IsOpen == true) { var valveToAct = (CheckBox)sender; char temp = valveToAct.Name.Substring(4, 1).First(); if (temp=='A') if (valveToAct.Name.Substring(5, 1).First()=='6') valveIndex=2; else valveIndex = valveToAct.Name.Substring(5, 1).First() - 48; else if (temp == 'B') valveIndex = valveToAct.Name.Substring(5, 1).First() + 6 - 48; else valveIndex = valveToAct.Name.Substring(5, 1).First() + 10 - 48; if (valveState[valveIndex] == true) { releaseValve(valveToAct.Name.Substring(4, 1).ToArray()[0], valveToAct.Name.Substring(5, 1).ToArray()[0], valveIndex); } else { activeValve(valveToAct.Name.Substring(4, 1).ToArray()[0], valveToAct.Name.Substring(5, 1).ToArray()[0], valveIndex); } } } //} static public int mode = 0; private void tmrController_Tick(object sender, EventArgs e) { long currentTime = timeMaster.ElapsedMilliseconds; this.tsslMainStatus.Text = "Cycle " + currentCycle.ToString() + ": " + ModeList[mode] +" "+ (timeMaster.ElapsedMilliseconds / 1000).ToString("") + "s, (Total running: " + (totalTimer.ElapsedMilliseconds / 60000).ToString() + "mins)"; for (int i = 0; i < 3; i++) { if(i==mode) this.StatusStripList[i].BackColor = System.Drawing.Color.Yellow; else this.StatusStripList[i].BackColor = System.Drawing.Color.LightGray; 77 } if (mode == 0) { if ((this.tspbProcessBar.Maximum - 50) - currentTime >= 0) { this.tspbProcessBar.Value = (int)(this.tspbProcessBar.Maximum - currentTime); } else { timeMaster.Restart(); mode = 1; startPurgeCycle();//start purging } } else if (mode == 1) { if ((this.tspbPurgeBar.Maximum - 50) - currentTime >= 0) { this.tspbPurgeBar.Value = (int)(this.tspbPurgeBar.Maximum - currentTime); } else { timeMaster.Restart(); mode = 2; startCollectCycle(); } } else { if ((this.tspbCollectBar.Maximum - 50) - currentTime >= 0) { this.tspbCollectBar.Value = (int)(this.tspbCollectBar.Maximum - currentTime); } else { timeMaster.Restart(); mode = 0; currentCycle++; startProcessCycle(); } } } private void Form1_FormClosing(object sender, FormClosingEventArgs e) { if (serCOM.IsOpen == true) { 78 reset_allvalves(); serCOM.Close(); timeMaster.Stop(); } } } } 79 Appendix B : PDMS device fabrication B.1 Fabrication of the master wafers Flow layer wafers: 1. The wafer is cleaned with acetone, methanol and isopropanol. 2. Dry the wafer with nitrogen, and then bake it at 95℃ for 5 minutes for dehydration 3. A layer of SU-8 photoresist is deposited on the wafer surface on a spinner with spin speed determined by the demanded thickness of the layer. 4. Bake the wafer at 65℃ 95℃ and 65℃ for 2, 5 and 2 minutes. 5. Load the wafer into a Canon aligner (Canon PLA-501F, Mississauga, Ontario) with the photomask positioned already. Align the features on the wafer from previous depositions with the photomask. 6. Expose the wafer to UV light through the photomask for 60-90 seconds. 7. The wafer is ejected from the mask aligner and baked at 65℃ 95℃ and 65℃ for 2, 5, and 2 minute respectively. 8. Develop the wafer in SU-8 developer (Microchem) 9. Clean the wafer with isopropanol and dry it with nitrogen. 10. Steps 3-9 are repeated to produce the four layers with SU-8 photoresist 11. Spin a thin layer of HMDS (Sigma-Aldrich) onto the wafer at 4000 rpm for 40 seconds, and then allow it to evaporate. 12. Spin SPR onto the wafer 13. Bake the wafer at 65℃ 95℃ and 65℃ for 1, 3 and 1 minute(s). 14. Leave the wafer to sit at room temperature for 15 minutes 15. Load the wafer into mask aligner, align, and expose for six 30 second bursts, with a 60 second interval between each burst. 16. Leave to sit at room temperature for 5 hours 17. Develop the wafer in MF 24A photoresist developer (DOW) 18. Rinse with DI water, then dry with nitrogen 12. Bake the wafer at 65℃ 95℃ and 65℃ for 1, 5, and 1 minute(s) respectively to reflow and photoresist and round the features 80 Control layer wafers: 1. The wafer is cleaned with acetone, methanol and isopropanol. 2. Dry the wafer with nitrogen, and then bake it at 150℃ for 5 minutes for dehydration 3. A layer of SU-8 photoresist is deposited on the wafer surface on a spinner with spin speed determined by the demanded thickness of the layer. 4. Bake the wafer at 65℃ 95℃ and 65℃ for 2, 5 and 2 minutes. 5. Load the wafer into a Canon aligner 6. Expose the wafer to UV light through the photomask for 90 seconds. 7. The wafer is ejected from the mask aligner and baked at 65℃ 95℃ and 65℃ for 2, 5, and 2 minute respectively. 8. Develop the wafer in SU-8 developer (Microchem) 9. Clean the wafer with isopropanol and dry it with nitrogen. Table B-1 below shows the parameters used to fabricate the four SU-8 layers of flow layer wafers and the control layer wafers. Table B-2 shows the fabrication parameters for the SPR layer. Table B-1 Photolithography fabrication parameters for SU-8 layers Layer Dehydration bake time (min) Dehydration bake temperature (℃) Photoresist Spin time (s) Spin speed (rpm) Softbake time (min) Softbake temp (℃) Exposure time (s) Post-exposure bake time (min) Post-exposure bake temp (℃) Measured thickness (μm) Flow 1 5 150 SU-8 3010 30 2400 2,5,2 65,95,65 60 2,7 65,95 10.7 Flow 2 1,5,1 65,95,65 SU-8 3005 30 2500 2,5,2 65,95,65 60 2,5,2 65,95,65 5.5 Flow 3 2,5,2 65,95,65 SU-8 3025 30 3000 2,5,2 65,95,65 60 2,5,2 65,95,65 10 Flow 4 2,5,2 65,95,65 SU-8 3025 30 1750 2,7,2 65,95,65 90 2,7,2 65,95,65 49.8 Control 5 200 SU-8-3025 30 2500 2,5,2 65,95,65 90 2,7 65,95 29 Table B-2 Photolithography fabrication parameters for SPR layer Layer HMDS spin time (s) HMDS spin speed (rpm) Photoresist Spin time (s) Spin speed (rpm) Soft bake time (min) Soft bake temp (℃) Wait time (min) Exposure time (s) Wait time (hrs) Reflow bake time (min) Measured thickness (μm) Flow 4 50 400 SPR 220-7 50 625 1,3,1 65,95,65 15 6x 30s 5 1,5,1 28 peak 81 Fabrication of polyurethane molds 10. Prepare Sylgard 184 PDMS at 10:1 base to hardener (Momentive Performance Materials) 11. Mix the samples manually with a tongue depressor 12. Form a 1.5 cm high dish around the flow layer wafer using a folded tinfoil sheet 13. Pour the mixed PDMS onto the master silicon wafer 14. Vacuum the wafer until air bubbles are brought to the surface 15. Bake for 2 hours in a 65℃ oven to cure the PDMS 16. Carefully remove the PDMS from the master wafer and trim the edges of the cured PDMS 17. Tape the PDMS to a rubber dish with feature facing up 9. Prepare polyurethane (Smooth-Cast ONYX, Fiber-Tek, Burnaby, BC) according to the manufacturer’s instructions 10. Pour the polyurethane over the PDMS in the rubber dish and wait 2 hours to allow the polyurethane to cure 11. Separate the polyurethane mold, PDMS and rubber dish Fabrication of PDMS devices from polyurethane molds 18. Prepare Sylgard 184 at 10:1 base to hardener 19. Mix the samples manually with a tongue depressor 20. Cap the samples with Parafilm and cap it, then mix in the centrifuge mixer (mix for 2 mins 30s, degas for about 1 minute 45s) 21. While the samples are mixing, clean the molds of debris by scraping away any PDMS left over and blowing the surface with gas. 22. Once the mixing is complete, pour the 10:1 PDMS into the molds, then vacuum for 20 minutes 23. Center the control layer on the wafer spinner chuck 24. Spin the 10:1 PDMS onto the control layer wafer according to this recipe 25. Place the coated wafer into the oven, bake for 2 hours 26. Remove the flow mold from the vacuum chamber, bake for 2 hours 27. Remove the cured devices from the oven and cut the flow layer out of the mold 28. Punch the 3 flow layer inlets with a 0.5 mm OD punch and clean the surface of the flow layer with scotch tape 82 29. Put the flow layer devices and the control waters into plasma chamber and plasma treat for 1 minute 15s 30. Bond them together carefully 31. Place the device in the oven for about 15 minutes 32. Remove the bonded devices from the oven and peel up the thin layer from the edges of the control layer wafer 33. Carefully peel off the bonded device and trim the device edges 34. Punch the outlet wells and all the remaining valve control holes and trap holes 35. Clean the surface of the device and the glass slide with scotch tape 36. Put the bonded devices and glass slides into plasma chamber and plasma treat for 1 minute 15s 37. Bond them together carefully 38. Bake the bonded device for about 15 minutes 83 Appendix C : CTC counting results for patient samples C.1 CTC counting results for mCRPC patient samples Table C-1 CTC counts in samples from patients with mCRPC Patient ID No. RCT counts / 7.5 ml CellSearch counts / 7.5 ml 63 2145 0 68 932 0 75 930 5 16 795 4 52 547.5 0 46 337.5 55 57 315 12 8 315 9 65 262.5 4 58 232.5 2 44 210 0 37 187.5 1 22 150 281 68 127.5 0 61 97.5 3 76 90 0 57 82.5 91 70 82.5 12 55 45 0 66 7.5 2 4 0 41 84 C.2 CTC counting results for LPC patient samples Table C-2 CTC counts in samples from patients with LPC Patient ID No. RCT counts / 7.5 ml CellSearch counts / 7.5 ml 52 0 1 60 0 0 77 0 32 Mean 329 23 Patient ID No. RCT counts / ml CellSearch counts / ml 57 51 332 35 3 51 22 3 54 19 0 41 15 25 38 14 1.5 40 14 13.5 52 8 8.5 18 6 0 24 4 15.5 35 2 5.5 49 1 2.5 21 0 028 0 0 43 0 085 Patient ID No. RCT counts / 7.5 ml CellSearch counts / 7.5 ml 46 0 3 50 0 0 53 0 0Mean 10.6 4.7