MICROFLUIDIC MICROARRAY FOR PATHOGENIC DNA

ANALYSIS: SINGLE-BASE-PAIR-MISMATCH DISCRIMINATION, AND MODELING/SIMULATION OF

CENTRIFUGAL FLOWS AND DYNAMIC HYBRIDIZATION

by

Lin Wang M.Sc., Northwest University, 2002

B.Sc., Ocean University of Qingdao, 1994

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

In the Department of Chemistry

Faculty of Science

© Lin Wang 2012 SIMON FRASER UNIVERSITY

Spring 2012

All rights reserved. However, in accordance with the Copyright Act of Canada, this work may be reproduced, without authorization, under the conditions for Fair Dealing. Therefore, limited reproduction of this work for the purposes of private

study, research, criticism, review and news reporting is likely to be in accordance with the law, particularly if cited appropriately.

ii

APPROVAL

Name: Lin Wang Degree: Doctor of Philosophy Title of Thesis: Microfluidic Microarray for Pathogenic DNA

Analysis: Single-Base-Pair-Mismatch Discrimination, and Modeling/Simulation of Centrifugal Flows and Dynamic Hybridization

Examining Committee: Chair: Dr. Michael H. Eikerling

Associate Professor, Chemistry

______________________________________

Dr. Paul C.H. Li, Senior Supervisor Professor, Chemistry

______________________________________

Dr. George R. Agnes, Supervisor Professor, Chemistry

______________________________________

Dr. Peter D. Wilson, Supervisor Associate Professor, Chemistry

______________________________________

Dr. Bonnie Gray, Internal Examiner Associate Professor, Engineering Science

______________________________________

Dr. Edward P.C. Lai Professor, Chemistry Carleton University

Date Defended/Approved: January 16, 2012

Partial Copyright Licence

iii

ABSTRACT

In the development of surface-based biosensors, the combination of

microfluidic technology with the DNA microarray chip has been realized with the

intersection method. The method shows the advantages of less sample usage,

multiple sample capability, enhanced hybridization signals, and reduced assay

time. In this thesis, line arrays of DNA probes were printed on a glass chip

through the microfluidic method. Target microchannels orthogonally intersected

with these line arrays, and complementary DNA molecules, upon hybridization,

were retained at the intersections as rectangular spots. Detection was achieved

through the read-out of the fluorescent labels on the targets. The high surface-to-

volume ratio in microchannels of nanolitre volume enhanced the detection

sensitivity as compared to that obtained with the bulk solution method. The spot

shape is more regular than that from the pin-spotted microarray, which is an

advantage for subsequent image processing. For diagnostic purposes, PCR

products amplified from the genomic DNA of fungal pathogens were detected

with this microfluidic intersection method. Moreover, with the aid of gold

nanoparticles, two 260-bp DNA sequences with single base-pair difference were

discriminated from each other at room temperature for the first time. In addition to

the pressure-driven flows used in rectangular chips, a centrifugal-pumping

method was employed to drive liquid movement in a CD-like microchip.

Connections from electrodes or tubing were avoided, resulting in the ease of

iv

conducting parallel hybridizations in multi-channels. For example, up to 100

samples can be analysed simultaneously with the CD-like microchip. The

centrifugal-pumping flow in the spiral channel was modeled mathematically and

simulated numerically by the computational flow dynamics (CFD) method. It was

found that the mathematical results and simulation results are consistent with the

experimental findings. Other than the flow study, the kinetics of microfluidic DNA

hybridization was also studied and simulated. The effects of probe coverage,

channel depths, assay conditions, and sample delivery rate on DNA hybridization

were investigated. The variation of signal intensity inside a hybridization spot and

from spots to spots was also discussed and compared with the experimental

findings. The proposed method provides a way to optimize both chip design and

experimental conditions during surface-based biosensor assays.

Keywords: Microfluidic DNA microarray; Centrifugal pumping; DNA diagnostics; Fungal DNA; Single-base-pair-mismatch discrimination; Gold nanoparticle; Spiral channel; Computational flow dynamics; Surface reaction kinetics.

v

DEDICATION

This thesis is dedicated to my dear parents, Yuchun, and Chelsea, without

whom it would never have been accomplished.

vi

ACKNOWLEDGEMENTS

I would like to thank my thesis advisor, Dr. Paul C. H. Li, for his support

and guidance. He has been patiently teaching me the craft and art of scientific

research in the past years. I will never forget the enjoyable experience of

conducting microfluidic research in his laboratory. In the process, his integrity, his

respect for his student's scientific liberty, and his constant encouragement have

filled me with admiration.

I deeply appreciate the inspiration and the valuable advice from the other

members of my supervisory committee – Dr. George Agnes and Dr. Peter

Wilson. Their comments on various aspects of my thesis have kept me thinking

critically on the experimental design and results. I would also like to thank Dr.

Mary-Catherine Kropinski for training me on fluid dynamics. Her support has

been my principal sources on all matters in mathematics.

I am very grateful for the kind help from Dr. William S. Davidson and his

group. I specially thank Kryztoff Lubiensky and Yvonne Lai for training me the

molecular biology techniques. I am also indebted to Dr. Dipankar Sen and his

group for the access to laboratory equipment and resources such as the confocal

laser fluorescent scanner. Besides, I deeply appreciate the valuable advice on

my project provided by Dr. Hua-Zhong (Hogan) Yu.

vii

My thanks belong also to the members of our lab, both past and present.

The excellent work of Dr. Xing Yue Peng has been the ground of my PhD

program. I want to thank Dr. Te-Chun Wu for the help on microchip fabrication,

Dr. Hong Chen for his enlightening discussions on numerical simulation.

Besides, I would also like to thank Samar Haroun and Jonathan Lee for the

proofreading, Dr. Xiujun Li, Wei Xiao, Jacky Chou, Yuchun Chen, Michael Wong,

Iryna Kolesnyk, Katrina Smyth, Abootaleb Sedighi, and Zara Sanei for their

supportive work and comments on my experimental results. I am also grateful to

Carol Koch of Agriculture and Agri-Food Canada for kindly providing the genomic

samples of plant pathogens.

For the final stages of my PhD study, I specially thank Yuchun Chen for

providing me with love and encouragement. Despite long distance between us,

my parents and my sister have always supported me in whatever and wherever I

decided to do and to go, therefore, I feel very obligated to them.

viii

TABLE OF CONTENTS

Approval .......................................................................................................................... ii Abstract .......................................................................................................................... iii Dedication ....................................................................................................................... v Acknowledgements ........................................................................................................ vi Table of Contents .......................................................................................................... viii List of Figures ................................................................................................................ xii List of Tables .............................................................................................................. xxvii List of Abbreviations .................................................................................................. xxviii List of Symbols ............................................................................................................ xxxi

1: Introduction ............................................................................................................... 1 1.1 Fundamentals of DNA microarray hybridization....................................................... 1

1.1.1 The building blocks of DNA .......................................................................... 1 1.1.2 DNA duplex structure ................................................................................... 2 1.1.3 DNA hybridization analysis .......................................................................... 3 1.1.4 Hybridization of mismatched DNA sequences ............................................. 8 1.1.5 DNA microarray and its applications .......................................................... 11

1.2 Introduction to microfluidic DNA microarray technologies ...................................... 14 1.2.1 DNA hybridization in microfluidic chambers ............................................... 16 1.2.2 Microfluidic DNA hybridization with low-density probe arrays ..................... 23 1.2.3 Microfluidic DNA microarray using centrifugal pumping ............................. 31

1.3 Research outlines ................................................................................................. 38 1.4 Dissertation structure ............................................................................................ 39

2: Microchip Fabrication and Detection Instrumentation ......................................... 43 2.1 Introduction ........................................................................................................... 43 2.2 PDMS-based Microfabrication Processes ............................................................. 45

2.2.1 Surface preparation and layer deposition processes.................................. 47 2.2.2 Photolithography processes ....................................................................... 49 2.2.3 Casting and bonding processes ................................................................. 51 2.2.4 Comparison of molding master made from SU8 or wet etching ................. 53

2.3 Surface modification to generate aldehyde-functionalized glass chips. ................. 56 2.4 Immobilization of probe DNA. ................................................................................ 58 2.5 Quantification of fluorescent image and data analysis ........................................... 59

3: Flexible Microarray Construction and Fast DNA Hybridization Conducted on a Microfluidic Chip for Greenhouse Plant Fungal Pathogen Detection ............. 60 3.1 Introduction ........................................................................................................... 60 3.2 Experimental ......................................................................................................... 63

3.2.1 Materials. ................................................................................................... 63

ix

3.2.2 Fabrication of PDMS channel plates. ......................................................... 66 3.2.3 Probe line printing, sample hybridization and result read-out by

fluorescent scanning. ................................................................................. 67 3.3 Results and discussion .......................................................................................... 69

3.3.1 Flexible probe immobilization without spotting. .......................................... 69 3.3.2 Fast hybridization of multiple DNA samples. .............................................. 73 3.3.3 Effect of probe tether length and one-base-pair-difference

discrimination. ............................................................................................ 76 3.4 Conclusion ............................................................................................................ 80

4: Gold Nanoparticle-assisted Single Base-Pair Mismatch Discrimination on A Microfluidic Microarray Device ............................................................................... 81 4.1 Introduction ........................................................................................................... 81 4.2 Experimental ......................................................................................................... 83

4.2.1 Fabrication of the PDMS-glass microchip. ................................................. 83 4.2.2 DNA samples ............................................................................................. 83 4.2.3 Depositing GNP layers on glass surface using microfluidic method. .......... 84 4.2.4 Preparation of DNA-GNPs conjugates. ...................................................... 84 4.2.5 Probe line printing, sample hybridization and result read-out by

fluorescent scanning. ................................................................................. 85 4.3 Results and discussion .......................................................................................... 85

4.3.1 GNP-modified surface and its application to microarray DNA hybridization .............................................................................................. 85

4.3.2 GNP-DNA conjugates and its application to single-base-pair discrimination ............................................................................................. 90

4.4 Conclusion ............................................................................................................ 95

5: Fungal Pathogenic Nucleic Acid Detection Achieved with a CD-like Microfluidic Microarray Device ................................................................................... 96 5.1 Introduction ........................................................................................................... 96 5.2 Experimental ......................................................................................................... 98

5.2.1 Materials .................................................................................................... 98 5.2.2 Surface modification of glass chips and the fabrication of PDMS

channel plates ........................................................................................... 98 5.2.3 Probe line array creation ............................................................................ 99 5.2.4 Sample hybridization with spiral channel plate assembly ......................... 100 5.2.5 Quantification of fluorescent image and data analysis ............................. 101

5.3 Results and discussion ........................................................................................ 101 5.3.1 Improvements in the design and fabrication of channel plates ................. 101 5.3.2 Hybridization of oligonucleotide samples on the CD-like MMA ................. 102 5.3.3 Hybridization of PCR products on CD-like MMA ...................................... 105 5.3.4 Hybridization specificity on CD-like MMA ................................................. 107

5.4 Conclusions ........................................................................................................ 110

6: Optimization of a CD-like Microfluidic Microarray Device for the Fast Discrimination of Fungal Pathogenic DNA .............................................................. 111 6.1 Introduction ......................................................................................................... 111 6.2 Experimental ....................................................................................................... 113

6.2.1 Fabrication of microfluidic devices. .......................................................... 113

x

6.2.2 Oligonucelotides and the preparation of fungal pathogenic DNA samples. .................................................................................................. 113

6.2.3 Sample hybridization and fluorescent detection. ...................................... 114 6.3 Results and discussion ........................................................................................ 116

6.3.1 Fast and parallel hybridization using continuous centrifugal flow. ............ 116 6.3.2 Enhanced sensitivity by the control of flow, channel depth and

temperature. ............................................................................................ 119 6.3.3 Enhanced detection sensitivity due to the use of Cy5 dye labels. ............ 122 6.3.4 Discrimination of PCR products with single base-pair difference. ............. 125

6.4 Conclusion .......................................................................................................... 128

7: Analysis and Modeling of Flow in Rotating Spiral Microchannels: Towards Math-aided Design of Microfluidic Systems Using Centrifugal Pumping ..................................................................................................................... 129 7.1 Introduction ......................................................................................................... 129 7.2 Chip design, microfabrication and flow measurement of the spiral

microchannels ..................................................................................................... 133 7.3 The mathematical model ..................................................................................... 134 7.4 Numerical simulation of the microflow ................................................................. 139 7.5 Results and discussions ...................................................................................... 141

7.5.1 Characterization of microflows in equiforce spiral channels with experimental data .................................................................................... 141

7.5.2 Comparison of solutions from the mathematical model with the experimental data. ................................................................................... 143

7.5.3 Validation of the assumptions in the mathematical modeling with the experimental data and simulation results ................................................. 146

7.6 Conclusions ........................................................................................................ 155 7.7 Appendix A: The spiral coordinate system ........................................................... 157 7.8 Appendix B: Simulation set-up and the study on the transversal flow with the

CFD program ...................................................................................................... 161 7.8.1 Governing equations ................................................................................ 161 7.8.2 Volume, boundary, and initial conditions in the numerical simulations ..... 162 7.8.3 The effect of mesh grid density on the computation time and

accuracy during the numerical simulations .............................................. 164 7.8.4 Validation of the assumption of negligible cross-channel flows. ............... 166

8: Modeling of Microfluidic DNA Microarray Hybridization and Its Applications to Experimental Optimization ............................................................. 168 8.1 Introduction ......................................................................................................... 168 8.2 Experimental ....................................................................................................... 171 8.3 Model construction for microfluidic DNA microarray hybridization ....................... 171

8.3.1 Navier-Stokes and continuity equations ................................................... 171 8.3.2 Mass transport and hybridization kinetics ................................................ 174 8.3.3 Non-dimensionalization of the reaction-diffusion and kinetic

equations. ................................................................................................ 179 8.3.4 Physical interpretation of the dimensionless numbers: ............................. 181

8.4 Simulation of microfluidic DNA microarray hybridization ...................................... 184 8.4.1 Numerical simulation with COMSOL program .......................................... 184

xi

8.4.2 Parameter values used during simulation ................................................ 185 8.5 Results and discussion ........................................................................................ 189

8.5.1 Bulk concentration profiles of static hybridization and microfluidic hybridization ............................................................................................ 189

8.5.2 Prediction of the detectable range of oligonucleotide targets ................... 190 8.5.3 Effects of flow rates and channel dimensions on hybridization signals ..... 192 8.5.4 Effect of probe coverage on hybridization signals .................................... 198 8.5.5 Assay conditions on the improvement of signal intensity and

specificity/selectivity ................................................................................. 201 8.5.6 Spot morphology ..................................................................................... 210 8.5.7 Signal variation from spot to spot ............................................................. 217

8.6 Conclusions ........................................................................................................ 221

9: Conclusions and Perspectives ............................................................................ 222 9.1 Concluding remarks ............................................................................................ 222

9.1.1 Microfluidic surface hybridization ............................................................. 223 9.1.2 Applications and modeling of the CD-like chip using centrifugal

pumping flow ........................................................................................... 223 9.1.3 Modeling and simulation of microfluidic hybridization kinetics. ................. 224

9.2 Outlook and research perspective ....................................................................... 225 9.2.1 Sensitivity ................................................................................................ 225 9.2.2 Selectivity ................................................................................................ 226 9.2.3 Integration and packaging. ....................................................................... 226 9.2.4 Modeling and simulation of centrifugal-pumping flows ............................. 227 9.2.5 Modeling and simulation of hybridization kinetics ..................................... 227

10: Appendix .............................................................................................................. 231 10.1 COMSOL simulation steps for the kinetics study of microfluidic microarray

hybridization ........................................................................................................ 231 10.1.1 Channel geometry construction ............................................................... 232 10.1.2 Choosing the physics models: ................................................................. 233 10.1.3 Domain Coupling: .................................................................................... 234 10.1.4 Constant and Variable Specifications ...................................................... 235 10.1.5 Subdomain equation settings: .................................................................. 239 10.1.6 Boundary conditions settings: .................................................................. 241 10.1.7 Meshing ................................................................................................... 242 10.1.8 Solver Parameters and Solver Manager .................................................. 243 10.1.9 Post-processing and visualization ............................................................ 244

Reference List ........................................................................................................... 245

xii

LIST OF FIGURES

Figure 1-1 (a) General structure of the nucleotide. The nitrogenous base and the pentose sugar are linked each other through N-glycosidic linkage. The official numbering system is also shown. (b) The two purine nitrogenous bases in DNA. (c) The two pyrimidine nitrogenous bases in DNA. ........................................................................................................... 1

Figure 1-2 (a) Watson-Crick model of a 21-bp DNA double-helix. The pentose-phosphate backbones of the two strands (shown in blue and red ribbons) form a right-handed double helix. The DNA structure was generated with the DNA modeling server by Vlahovicek et al. [2]. Image visualization was performed with the Chimera molecular modeling system [3]. (b) Chemical structure of the specific hydrogen bonding in a DNA duplex. The duplex is stabilized by hydrogen bonding between complementary base pairs and (though not obvious from this drawing) by base stacking interactions between adjacent base pairs. ...................................................................................................... 2

Figure 1-3 Schematic view of the DNA hybridization process. (a) Two complementary single-stranded DNA before hybridization. (b) Intermediate state with a few base pairs from strand collision. (c) Hybridized DNA duplex from zippering step. ................................................... 4

Figure 1-4 (a) Preparation of DNA targets from genomic DNA. Here, the genomic DNA duplexes are denatured to two single strands. One of the strands (highlighted in yellow color) are then amplified and labeled with proper tags for later hybridization analysis. (b) and (c) show the two formats of DNA hybridization analysis: (b) Target DNA is labeled. (c) Probe DNA is labeled. ................................................................................ 7

Figure 1-5 (a) Distorted DNA duplex structure from two T·G mismatches in the center. The two green arrows indicate the two T·G mismatches. (b) The sequence of the duplex shown in (a). The blue strand is 5'-CGCATTACGC-3', and the red one is 5'-GCGTGGTGCG-3'. (PDB access number: 1SNH [24]) ............................................................................ 8

Figure 1-6 Hybridization-based approaches for the discrimination of SNP sequences. (a) Targets are prepared from two SNP sequences with only one base-pair difference. (b) Fluorescence-labeled targets hybridize to probe molecules. The mismatched sequences form unstable duplexes. (c) The unstable MM duplexes are denatured and removed by high-stringency washing conditions. (d) Fluorescence signals from retained PM duplexes. .............................................................. 10

Figure 1-7 Hybridization assay with spotted probe microarray. (a) Microarray probe spotting pins used in a commercial microarrayer. The photo is

xiii

adapted from the product datasheet of SpotBot® Titan High-Capacity microarrayer. (b) Glass slide with chemically-modified surface ready for probe binding. The slide can be labeled with barcode for easy identification. (b) Microspotting probe molecules on the slide. The lower image shows surface tethered probe sequences. The individual probe sequences can be identified by the position of the corresponding feature on the regular microarray grid. (c) Hybridization. Target solutions are applied to the slide and are incubated in a heated humidified box. Labeled targets can freely diffuse over the microarray surface until they hybridize with a complementary probe (lower image). (d) Washing step. After hybridization, unbound targets are washed away with buffer solutions as shown in the lower image. (e) Microarray image analysis. Quantification of the surface bound targets is performed by measuring the fluorescence intensity of the hybridization spots using a fluorescence scanner. ............................................................. 13

Figure 1-8 (a) Diagram for sample recirculation system on the hybridization chamber. (b) Hybridization image of fluorescence-labelled target nucleotide with 0.1 mL·min-1 continuous recirculating flow on a DNA microarray for 40 min. Reproduced from [73] by permission of The Royal Society of Chemistry. (c) Chaotic mixing of the dye solutions in two PDMS microfluidic chambers facilitated by the herringbone indentations in the bridge channel. (d) Signal improvement by microfluidic mixing on a home-spotted microarray. Top: Dynamic; Bottom: static hybridization. Reprinted with permission from [74], Copyright © 2006 John Wiley & Sons, Inc. .................................................... 18

Figure 1-9 (a) Schematic showing a number of air bubbles in the top layer of the DNA biochip chamber. An energetic cavitation streaming motion created by the PZT transducer was observed in the vicinity of each bubble. (b) Top image: fluorescence image of a biochip after 2-h static hybridization; Bottom image: fluorescence image of a biochip after 2-h hybridization aided with cavitation microstreaming. Reprinted with permission from [79], Copyright © 2003 American Chemical Society. ........... 20

Figure 1-10 (a) Illustration of the coordinated pump and valve used for dynamic hybridization. Fluids are extracted and reinjected to the hybridization chamber under a protocol to induce chaotic movement inside the chamber. (b) Photographs of dye solution mixing in the hybridization chamber after 2 cycles of liquid removal and reinjection processes using a 30-µL stroke volume. The dye distribution appears to be homogeneous after 16 cycles. Reprinted with permission from [80], Copyright © 2004 Elsevier Ltd. ...................................................................... 20

Figure 1-11 (a) Schematic drawing of a straight hybridization channel and probe spots (unit: mm). The flow direction was from top to bottom. (b) Fluorescent images after 30 min hybridization. Left: hybridization results at u = 1 cm·s-1; right: hybridization results at u = 0 cm·s-1 (control). (c) Comparison of the averaged fluorescence intensities after hybridization. Reproduced from [87] by permission of The Royal Society of Chemistry. .................................................................................... 25

xiv

Figure 1-12 (a) An illustration showing scrambled discrete plugs being swept over probe rows in the serpentine channel. The black arrow denotes the shuttle flow inside the microchannel. (b) The fluorescence images are from 500-s microfluidic hybridization (top row), 500-s static hybridization (center row) or 2-h static hybridization (bottom row). The left column represents the results from 20-mer PM targets, and the right column represents the results from 20-mer MM targets with single base-pair mismatch in sequence center. Reprinted with permission from [90], Copyright © 2005 Oxford University Press. ................................... 25

Figure 1-13 (a) Fluorescent image of the hybridization of DNA probes (5 and 10 µM) spotted on activated glass slides with increasing concentration of DMSO. Reprinted with permission from [112], Copyright © 2003 Oxford University Press. (b) Fluorescence images acquired in Cy3 (green) and Cy5 (red) channels show the parallel microfluidic patterning of multiple DNA probes. Reprinted with permission from [128], Copyright © 2009 American Chemical Society. (c) A 16-channel microfluidic network was used to pattern 20-µm lines of decreasing antigen coverage onto a PDMS substrate (left line to right line) using solutions of different concentrations. The antigen was labelled with rhodamine. Reprinted with permission from [120], Copyright © 2000 American Chemical Society. (d) Cross-polarized optical image of [Mo3Se3

-]∞ nanowire patterns on glass substrate made with microchannel networks. Reprinted with permission from [131], Copyright © 2000 American Chemical Society. ............................................. 28

Figure 1-14 (a) Intersection approach for performing an immunoassay on a surface with microfluidic network. (i) Patterning different antigen molecules along the horizontal lines on a solid substrate. (ii) The area of the substrate left unpatterned during step (i) is blocked with bovine serum albumin (BSA) to prevent nonspecific binding in subsequent steps. (iii) Antibodies flowing through the channels of a second microfluidic network locally bind to the patterned antigens. (iv) Reading the rectangular binding patches reveals the amount of antibodies present in the samples. Reprinted with permission from [124], Copyright © 2001 American Chemical Society. (b) SPR difference images showing hybridization of a RNA sample onto an array of surface-bound probe DNA. Hybridization of the target RNA onto the probe DNA array was indicated by a change in the reflectivity. Reprinted with permission from [96], Copyright © 2001 American Chemical Society. ......................................................................................... 30

Figure 1-15 (a) Photograph of the microfluidic chip containing shuttle-flow channels, microvalves and micropumps. The entire chip consists of three layers, the top and middle layer were made of PDMS containing microvalves, micropumps, and shuttle flow channels; the bottom layer was a glass substrate with gel pads on which DNA probes were immobilized. The shuttle flow hybridization was realized by controlling the gas ports 1, 2 and 3 automatically. (b) Hybridization specificity assay using four serotypes of the Dengue virus under shuttle flow conditions (frequency 2 Hz). Channels 1 to 4 indicated target DNAs (200 nM) of serotypes I to IV, with repetitions in channels 5 to 8,

xv

respectively. The duration of hybridization process was 90 s and washing time was 30 s. Reproduced from [64] by permission of The Royal Society of Chemistry. .......................................................................... 32

Figure 1-16 (a) Schematic representation of a CD microfluidic device sealed with a glass substrate for DNA hybridization. It consists of a PDMS CD slab containing twelve DNA hybridization assay units with immobilized DNA probe arrays. (b) Schematic diagram of a single DNA hybridization assay unit. (c) Hybridization specificity tests with the CD microfluidic device. Top: Dengue virus serotype 1 targets. Bottom: Dengue virus serotype 2 targets. The CD device was rotated at 22 Hz for 3 s and then stopped for 3 s during the reciprocating process, with a duration of 90 s. Reprinted with permission from [108], Copyright © 2009 Elsevier Ltd. .................................................................................................. 36

Figure 1-17 The organization structure of this dissertation. .......................................... 42

Figure 2-1 Sequence for fabrication of the microfluidic assembly using negative photoresist SU-8. (a) a plain silicon wafer (b) Si wafer coated with photoresist film; (c) wafer exposed to UV light through a photomask; (d) photoresist developed; (e) PDMS prepolymer solution casted on the SU-8 master; (f) PDMS channel plate peeled off; (g) PDMS channel plate bound to a glass chip to produce assembly. ............................ 46

Figure 2-2 Photograph of the softwall clean room facilities used for making microchip and a bionocular microscope for surface examination. .................. 47

Figure 2-3 (a)-(d) Illustration of the spin coating processes for the creation of the SU8 photoresist layer on a silicon wafer. (e) a spin coater with controller (f) rotation speed versus the thickness of the SU8 resin layer. The figure is adapted from MicroChem® product datasheet. SU8-50 and SU8-100 are different from each other in viscosity for different applications. In this thesis, SU8-50 has been used throughout. ....... 49

Figure 2-4 (a) Silicon wafer aligned with photomask under UV irradiation; (b) SU8 layer with microstructures on a 5-inch silicon wafer after the developing process. ...................................................................................... 50

Figure 2-5 Comparison of the quality of PDMS microchannels fabricated from two photomasks of different printing resolutions. (a) Photomask printed with 3368 dpi. (b) The microscopic image of the PDMS channel plate made from (a). (c) Photomask printed with 20,000 dpi. (d) The microscopic image of the PDMS channel plate made from (c)....................... 51

Figure 2-6 (a) SU8 master with weir structure around the wafer for PDMS molding. (b) Assembly of the PDMS microchannel plate with a glass disk. .............................................................................................................. 53

Figure 2-7 Sequence for fabrication of the microfluidic assembly using a positive photoresist and subsequent HF etching. (a) A clean Si or glass wafer. (b) Cr and Au masked Si wafer (c) The wafer in (b) is coated with positive photoresist. (d) The wafer in (c) is exposed to UV light through a photomask; (e) photoresist is removed; (f) The exposed metal mask is etched; (g) The exposed Si or glass etched; (h) The photoresist and metal is stripped; (i) PDMS prepolymer solution casted on the SU8

xvi

master; (j) The cured PDMS channel plate is peeled off; (k) PDMS channel plate is sealed to a glass chip to produce assembly. ....................... 54

Figure 2-8 Comparison of the Si masters and microchannels fabricated using positive and negative photoresist. (a) The reflective microscope image of the Si master fabricated from positive photoresist followed with KOH etching. (b) Surface profile of the microchannels in (a); (c) The reflective microscope image of the Si master fabricated from SU8 photoresist. (d) Surface profile of the microchannels in (c). ........................... 56

Figure 2-9 Surface modification to generate aldehyde-functionalized glass surface. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ......................................................................... 57

Figure 2-10 Covalent attachment of aminated probe DNA to aldehyde glass surface. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ......................................................................... 58

Figure 2-11 Typhoon 9410 confocal laser fluorescent scanner. ..................................... 59

Figure 3-1 The microfluidic microarray method using straight microchannels. (a) The creation of a DNA probe line array on an aldehyde-modified glass slide via straight microchannels. (b) The hybridization of DNA samples in straight channels orthogonal to the straight probe lines printed on the glass slide. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ......................................................................... 63

Figure 3-2 The images of the assembly of a 2"×2" PDMS channel plate on a 3"×2" glass slide. (a) Image of16 channels filled with blue-dye solutions in horizontal direction. (b) Image of 16 channels filled with green-dye solutions in vertical direction. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ............................................. 66

Figure 3-3 Microfluidic hybridization of target DNA strands (in red) in the parabolic liquid front to the probe DNA strands (in blue) immobilized on the glass chip surface. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ............................................................. 68

Figure 3-4 Probe immobilization. (a) Effect of ionic strength of spotting solutions and immobilization time on the ADF signal. The probe line arrays were made by incubating 0.8 µL of 25-µM ADF prepared in 1.0M NaCl (grey bar) or 0.1M NaCl (white bar) in microchannels at different durations. The slide was chemically reduced and then washed with distilled water. The fluorescent signals were measured by scanning the slide at 488nm. (b) The fluorescent image of ADF probe lines (vertical green stripes) printed on the glass slide using different immobilization times. The brighter the probe lines, the stronger the fluorescent signals. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ......................................................................................... 70

Figure 3-5 Sample hybridization. (a) The immobilization signal of various ADF probe solutions (0.8-µL) at different concentrations (10 to 400 µM) that were incubated in microchannels for 2 h. After washing, the slide was scanned at 488nm. (b) The hybridization signals resulted from the above probe lines. Complementary oligonucleotides (CD’) labelled with

xvii

Cy5 (100 nM, prepared in 1xSSC + 0.2%SDS) was hybridized to the ADF probe lines for 10 min. After washing, the slide was scanned at 633nm. (c) Overlaid dual-channel image of the same glass slide showing both printed probe lines (vertical green lines) and square hybridization patches (red) at intersections. ADF probe immobilization was achieved in duplicate at each concentration. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ........... 71

Figure 3-6 Hybridization of oligonucleotide samples to printed probe lines. (a) Fluorescent images of the hybridizations of oligonulceotides (FB’ and FD’ prepared in 1X SSC + 0.2% SDS) with probe line arrays for 10 min at room temperature. (b) Histogram showing the fluorescent intensities of hybridization versus non-specific binding at various sample concentrations. The grey bars represent the signals of samples hybridized with their complementary probe sequences, i.e., FB’ with AB, or FD’ with AD; the white bars represent the non-specific binding. The error bars describe the standard deviations of the signals from 5 or 7 hybridization patches. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ............................................. 74

Figure 3-7 Hybridization of PCR products to printed probe lines. (a) Fluorescent signals from the hybridization of 2.6-ng pre-denatured PCR products (FB’P and FD’P) at 50 ºC for 5-min flow, 30 min incubation and overnight incubation, respectively. (b) The fluorescent images corresponding to the left histogram of hybridizations of 2 samples to 7 probe lines. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ......................................................................... 76

Figure 3-8 The immobilized oligonucleotide probe with C6 or C12 tether as spaced from the glass substrate. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ............................................. 77

Figure 3-9 Hybridization of PCR products to probes having different tether lengths. (a) Fluorescent signals from hybridization of PCR products (1.4 ng and 2.6 ng) to probes with 2 different tether lengths. (b) The fluorescent image corresponding to the left histogram of hybridizations of samples to 5 probe lines. (c) Effect of probe tether length on discrimination of PCR products with one-base-pair-difference. Fluorescent signals come from the fluorescent patches of three PCR products (FB’P, FBN’P and FD’P, 1.4ng for each). The small black bars come from the non-specific binding signals of FD’P and they are too low to be seen on probe AB. The error bars describe the standard deviations of the signals from 5 duplicated tests. (d) The fluorescent image corresponding to the left histogram of hybridizations of 3 samples to 5 probe lines. In all cases, the DNA targets were hybridized by continuous-flow method for 5 min at 50 ºC. Reprinted with permission from [94], Copyright © 2007 American Chemical Society. ........... 78

Figure 4-1 Schematic diagram of the formation of the nanoscale-controlled spacing between oligonucleotide probes using GNPs. (a) Protonated amine groups on the APTES-treated glass surface. (b) GNPs, with adsorbed citrate ions, are then deposited to the aminated glass surface and form a submonolayer. (c) The remaining non-reacted

xviii

amine groups are deprotonated using a pH 11 buffer and masked by acetic anhydride (Ac2O). (d) When oligonucleotides probe DNA solutions flow through the submonolayer, the amine groups at the 5’-end of the molecules will bind only to GNPs which are well spaced out on the glass surface. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics. .............................................................. 87

Figure 4-2 Different modified surfaces for hybridization experiments. (a) Image of an APTES modified glass slide with immobilized GNP strips (b) Magnified image of the glass slide with insets showing (i) Flow through 2-µL GNP solutions, (ii) Flow through 10-µL GNP solutions. The glass slide was first treated with APTES solution to create an aminated surface. Then a PDMS channel plate was sealed against the glass slide. GNP solutions in different amounts were filled through 8 microchannels. The deposition of GNPs onto the aminated surface resulted in the pink strips. The rest of the microchannels were left empty to keep the glass surface aminated. The images were taken after washing the microchannels and peeling off the PDMS plate. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics. ....................................................................................... 88

Figure 4-3 Fluorescent images of the hybridization results from (a) 21-mer and (b) 50-mer target oligonucleotides at different modified surfaces. The targets are either complementary perfect- match (PM) or one-base mismatch (MM) in the sequence center with the probe DNA molecules. (a) Hybridization of Cy5 labelled 21-mer targets. The probe DNA lines were printed microfluidically on: aminated glass surface, on surface modified by flowing through 2-µL GNP solutions, and on surface modified by flowing through 10-µL GNP solutions (b) Hybridization of Cy5 labelled 50-mer targets. The probe DNA lines were printed microfluidically on: aminated glass surface, on GNP modified surface with Ac2O treatment to cap remaining amine groups, and on GNP modified surface without AC2O treatment. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics. ........................... 89

Figure 4-4 (a) GNP-DNA conjugates from the incubation of target DNAs with gold nanoparticles. (b) Perfectly matched target DNAs desorbed from GNPs and hybridized to the surface immobilized probes. (c) Mismatched DNAs remained bound to GNPs and were washed away through microfluidic method. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics. ........................................... 91

Figure 4-5 (a) Images of hybridized patches of perfect-matched (PM) and mismatched (MM) target oligonucleotides in triplicate. Here, the oligonucleotides were pre-incubated with GNPs (5 nm) at different ratios. (b) Discrimination ratios between PM and MM duplexes. The discrimination ratios were calculated by dividing the signal of PM DNAs with that of MM DNAs (The higher ratio, the better). (c) The fluorescent hybridization signals from the images in (a), and the results at Oligo/GNPs = 1:1 are expanded and shown in the right inset. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics. ....................................................................................... 92

xix

Figure 4-6 (a) Images of hybridized patches of perfect-matched (PM) and mismatched (MM) PCR products in triplicate. Here, the amplicons were pre-incubated with GNPs (5 nm) at different ratios. (b) Discrimination ratios between PM and MM amplicons. The discrimination ratios were calculated by dividing the signal of PM DNAs with that of MM DNAs (The higher the ratio, the better). Reprinted with permission from [95], Copyright © 2010 American Institute of Physics. ......... 94

Figure 5-1 Assemblies of the PDMS channel plates with glass disks. (a) The circular channel plate with radial microchannels. All the channels (100 in total) were filled with dye solutions; (b) The image of an equi-force spiral channel plate assembled with a glass disk. Here, 30 out of 100 spiral microchannels, three in a group, were filled with dye solutions. The scale bar represents 10 mm. .................................................................. 99

Figure 5-2 Schematic diagram of probe immobilization and sample hybridization using CD-like microfluidic microarray assemblies. (a) Probe printing procedure: an array of radial probe lines was created on an aldehyde-modified glass disk using a radial channel plate. (b) Hybridization procedure: the hybridization occurring at the intersections between the spiral channels and radial probe lines, shown as colored patches in the last circle. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd. ................................................................................................ 100

Figure 5-3 DNA hybridization on a CD-like glass chip. (a) The schematic diagram of hybridization patches formed at the intersections between 6 sample channels and 3 probe lines. (b) The fluorescent image of the entire glass disk. The 3 straight traces are resulted from the radial flow of the fluorescent marker during probe immobilization. The 3 grey spiral traces are resulted from the spiral flows of marker during sample hybridization. Hybridization results obtained in the rectangular region are expanded to give the middle inset. The right inset shows the groups of rectangular patches formed near the disk center, which are resulted from the hybridization of different concentrations of oligonucleotide samples with their complementary probe lines. The probe lines (AD or AB) were created by 40 min-incubation of 25-µM aminated oligonucleotide probes in the radial channels. Oligonucleotides hybridizations were achieved at room temperature in 3-min spinning at 700 rpm, and then dried out at 3600 rpm. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd. ......................... 103

Figure 5-4 (a) The calibration graph of the hybridization signals from oligonucleotide FB’ (top curve) and FD’ (bottom curve). (b) The hybridization signals of FB’ samples using different solution concentrations and volumes. In all cases, the samples were hybridized to 25-μM probe lines at room temperature for 10 min and the error bars describe the standard deviations of the signals from 9 hybridization patches. The fluorescent intensities expressed have already been subtracted from the background. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd. .................................................................... 105

Figure 5-5 (a) Comparison of different flow conditions for the hybridization of oligonucleotides and PCR products at 45 ºC. From left to right the

xx

hybridization conditions were: oligonucleotide samples (FD’, solid bar) by continuous flow for 5 min, PCR products (FD’P, hatched bar) by continuous flow for 10 min, or by stop-flow incubation for 10, 30, 60 and 120 min. For continuous flow, the spun rate was 500 rpm; whereas for stop-flow incubation, the samples were first introduced at a spun rate of 700 rpm for 2 min. For oligonucleotide sample, 5-nM FD’ was used because it already gave sufficiently high intensity; for PCR products, 20-nM FD’P was used, but the signal for 10-min stop-flow hybridization of FD’P was not detected and the variation of background was shown instead. The error bars describe the standard deviations of the signals from 9 hybridization patches. (b) The effects of time and temperature on the hybridization of PCR products. FD’P (20 nM) was introduced into spiral channels by spinning at 700rpm for 2min followed by incubation in an oven for different times at different temperatures. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd. ................................................................................................ 106

Figure 5-6 The hybridization specificity shown by the plots of fluorescent intensities versus locations in the spiral channels. The samples are fluorescein-labelled oligonucleotide (FD’) and PCR products (FD’P), each at 2 concentrations. (a) The samples flow from left to right intersecting the AB and AD probe lines alternatively. (b) The samples flow from left to right intersecting various probe lines in this sequence: AD, 7 non-complementary probe lines, AB, 7 non-complementary probe lines, and AD, as depicted in the box as D, N, B, N and D, respectively. (c) The fluorescent images correspond to the hybridizations and backgrounds shown in (b). All the probe lines were created by 40 min-incubation of 25 μM aminated oligonucleotide probe. Oligonucleotide sample hybridizations were achieved by 3-min spinning and dry-out at room temperature. For PCR products, they were firstly introduced into channels by spinning and then incubated in oven at 45°C. After hybridization for 2 hours, the DNA sample solutions inside channels were spun out at 3600rpm. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd. ................................ 108

Figure 6-1 Comparison between the continuous-flow and stop-flow methods of hybridization. (a) The fluorescent images show the PCR product hybridizations conducted with two different flow methods at 42 ºC. The dark rectangular patches represent the specific binding of the complementary targets. 1.0 μL of PCR products (CB’P or CD’P) were applied to the spiral microchannels. Probe AB was used to print radial probe lines (b) Line graph showing the average fluorescent intensities along the bottom radial probe line intersected with 20 spiral sample channels depicted in (a). (c) Hybridization signal comparison of two continuous-flow methods on same microchannel plate. Here, six groups of oligo probe AB were preprinted and distributed evenly on a glass disk as shown in the inset. Sample solutions (0.2ng CB’P in 2.5xSSC+0.2%SDS), driven by either centrifugal force or vacuum suction, flowed through and intersected with these probe lines. Each bar represents signals from the average of nine hybridization patches. The bar number in the graph matches to the number as shown in the inset, indicating the position of hybridization along the spiral channels.

xxi

The error bars represented the standard deviation of hybridization signals from the nine patches. The sample volume is 0.8 μL and the rotation speed or vacuum was controlled to ensure an average 3 min of hybridization time. (d) Picture of the broken liquid column (in red) inside a 24-μm depth microchannel during vacuum suction. The inset shows the magnified picture of the area. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd. .................................................. 118

Figure 6-2 Enhanced sensitivity due to smaller channel depth and higher temperature. (a) The bar graph showing the measured fluorescence intensities from the hybridization of 1.6-ng CB’P at different rotation speeds at room temperature. The line graph shows the corresponding residence times. The error bars were calculated from three measurements. (b) The effect of channel depth on hybridization signal intensity. Different amount of PCR products (CD’P) were hybridized at 42°C for 3 min. (c) Fluorescent intensities along an AB probe line. Ten samples (CB’P) were detected on the test disk at different concentrations and each sample was conducted in duplicate. From left to right, five samples (6.4, 3.2, 1.6, 0.8, and 0.2 ng) were first hybridized at room temperature (23°C) for 3 min. After dry-out of these channels, another group of 5 samples was hybridized at 42°C for 3 min. In all cases, the volume of sample solutions was 1.0 μL. The insets showed the magnified graphs of the hybridization signals of 0.2 ng CB’P. (d) Calibration curves of hybridization signals of PCR products (0.2~6.4 ng). The error bars represented the standard deviation of hybridization signals from six patches. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd. ................................ 121

Figure 6-3 Enhanced hybridization signals of Cy5-labelled oligonucleotides due to higher hybridization efficiency and less photobleaching. (a) signal-to-noise ratio from the hybridization of different labelled oligonucleotides with the same probe lines. The results of fluorescein-labelled DNA are expanded to give the right inset picture. Here, the concentration of both oligonucleotide samples is 0.1 μM. (b) Comparison of hybridization efficiency between Cy5-labelled and fluorescein-labelled oligonucleotides. The left two bars show the normalized fluorescent intensities from the hybridization of CD’ or CD’ + FD’ (scanned at 670nm). The right two bars show the normalized fluorescent intensities from the hybridization of FD’ or CD’ + FD’ (scanned at 532 nm). In all the cases, the final concentrations of different labelled oligonucleotides are 1 µM. (c) Photobleaching effect on different dye labelled oligonucleotides. Normalized fluorescent intensity from the chip with hybridized oligos at the specific wavelength of fluorescein dye. In all the cases, 1 µL of 0.1-µM oligo samples labelled with Cy5 or fluorescein dyes were applied and hybridized at room temperature for 45min in dark place. After removal of the channel plate, the glass chip was scanned consecutively for three times and it was then exposed to room light for 4 h before it was scanned for the fourth time. The error bars here represent the standard deviations of the hybridization signals of six patches. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd. .......................................................... 124

xxii

Figure 6-4 Differentiation of Botrytis species with single-base-pair difference at various hybridization temperatures. (a) 4 groups of fluorescent images showing 9 patches of matched duplex and 9 patches of mismatched duplex. Each image was obtained from the hybridization of sample solutions in 3 spiral channels intersecting with 3 probe lines at the specified temperature. Two PCR products, perfectly matched CB’P (PM) and mismatched CBN’P (MM), were tested in the experiment. (b) Bar graphs showing the fluorescent intensities from the hybridization patches in (a). The error bars represent the standard deviations of the hybridization signals of nine patches. The connected line shows the discrimination ratio between the PM and MM species. In these discrimination experiments, 1.0-μL PCR products (1.6-ng CB’P or CBN’P) were applied in triplicate to the inlet reservoirs of 6 spiral channels Then the sample solutions flowed into the spiral channels by spun at 1500 rpm for 3 min first at 23 ºC (room temperature). After this operation, the channels were dried out at 3600 rpm and the temperature of the assembly was raised to 42 ºC. Then another batch of sample solutions was applied to another 6 spiral channels on the same chip and the process was repeated. The procedure was repeated similarly for hybridization at 52 ºC and 62 ºC. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd. ................................ 127

Figure 7-1 (a) The photograph of the equiforce spiral channel assembly before spinning. One inlet reservoir was filled with colored solutions and a piece of paper printed with grey position lines was placed underneath to facilitate the measurement of the liquid front position. (b) Magnified picture in (a) shows the image of inlet reservoirs and a microchannel partially-filled with red dye solution. Reproduced from [286] by permission of The Royal Society of Chemistry. ........................................... 134

Figure 7-2 (a) The layout of the equiforce spiral channel and the orthogonal curvilinear coordinate system used for the mathematical modeling of the liquid flow inside the spiral microchannel. (b) 3-D magnified view of the spiral microchannel in the curvilinear coordinate system. H and W are channel height and width, respectively. Reproduced from [286] by permission of The Royal Society of Chemistry. ........................................... 137

Figure 7-3 (a) A bird’s-eye view of the 3D geometry of the equiforce spiral microchannel used in CFD simulations. (b) Magnified view of the inlet region with a tilted inlet reservoir and a neck channel. Point A denotes the starting position of the spiral channel where the neck channel connects. (c) Magnified view shows the structured grid of the microchannel near point A. The cell widths range from 15 µm to 75 µm. The grid lines are as close to 90° as possible and small cell aspect ratios were employed to ensure a good modeling quality. Reproduced from [286] by permission of The Royal Society of Chemistry. ...................... 140

Figure 7-4 Liquid movement inside microchannels. (a), (b) and (c) show the 3 snapshot images of the disk spun at 1500 rpm, where 1 and 2 designate the 10th and 20th position lines, respectively. The black arrows denote the inlet reservoirs and the white arrows indicate the liquid front. The spiral trace extended as time increased, indicating the movement of the liquid front along the microchannels. (d), (e) and (f)

xxiii

represent the displacement-to-time plots showing the movement of both the liquid front and liquid rear meniscus in 3 adjacent spiral channels. The start lengths of the liquid front measurement are 7.8 mm, 9.3 mm, and 12.6 mm, respectively. Reproduced from [286] by permission of The Royal Society of Chemistry. ........................................... 143

Figure 7-5 (a) Liquid front displacement versus time at different rotation speeds. Data points represent experimental data and the solid lines are theoretical curves plotted using the mathematical model (The curves and data points were displaced horizontally for clarity). (b) Microscope images of 3 channels in a PDMS chip showing the width variation. The scale bar represents 100 μm. (c) Calculated variation of flow velocity as affected by the variations in channel width and depth using the proposed mathematical model. (d) Liquid front displacement versus time graph from two spiral-microchannel chips of different channel geometries. Data points represent experimental data and the solid lines are from the model prediction. Here, W, H, and U are channel width, channel height, and measured liquid-front velocity, respectively. Both microchips were rotated at 1500rpm. Reproduced from [286] by permission of The Royal Society of Chemistry. ........................................... 145

Figure 7-6 (a) Image showing the initial conditions used for transient flow simulation. The blue color represents the part of the channel filled with air and the pink color denotes liquid solutions. (b) Simulated liquid front displacement versus time at 1500 rpm with different start lengths. (c) comparison of the liquid front velocities at 1500 rpm with experimental data, mathematical modeling, simulation of transient filling flow and simulation of steady flow. Reproduced from [286] by permission of The Royal Society of Chemistry. ........................................................................ 148

Figure 7-7 (a) Liquid front displacement versus time from experimental data at 2400 rpm. The inset shows the liquid front velocities from fitting the experimental data in both cases. Because a longer acceleration time was required to achieve a higher rotation speed, the start length is much longer, which is around 50 mm. (b) Simulated results of the liquid front velocities at 2400 rpm from anti-clockwise and clockwise directions. Reproduced from [286] by permission of The Royal Society of Chemistry. ............................................................................................... 150

Figure 7-8 (a) Comparison of the centrifugal force (Fω, shown as a gray plane.) and the surface tension force (Fs, shown as a color surface) along the microchannel at 1500 rpm (b) Simulated the liquid front velocities at 1500 rpm from two solutions (A and B) with different contact angles on the solid substrate (c) Measured dynamic contact angle change of the two solutions from experimental observations. Reproduced from [286] by permission of The Royal Society of Chemistry........................................ 153

Figure 7-9 (a) Liquid front moving velocity under different rotation speeds. Both data from experimental data and from mathematical model prediction are shown. The error bars are from the measurement of 3 adjacent channels. (b) Pressure at the beginning of the spiral channels generated from the short liquid column in the neck section under

xxiv

different rotation speeds. Reproduced from [286] by permission of The Royal Society of Chemistry. ........................................................................ 154

Figure 7-10 Section of the spiral curve and the infinitesimal segment used for the calculation. Here the black solid curve is the spiral curve and the blue dashed lines are radial lines from the disk center. r and θ are the usual polar coordinates. Reproduced from [286] by permission of The Royal Society of Chemistry. .................................................................................. 157

Figure 7-11 Comparison of the simulation results using two meshes with different resolutions. (a) low-density mesh, cross-channel resolution is same as that used in the manuscript (5 × 2 cells). (b) high-density mode, the cross-channel resolution of the computational grid is 20×12 cells. (i) computational grids showing the cross-section of the microchannel. (ii) the color map of the down-channel flow velocities (us) at the cross-section of the microchannel.(iii) the color map of the transversal flow velocities (un) at the cross-section of the microchannel. The center black lines in high-resolution grid results are the symmetry plane. .............. 166

Figure 7-12 (a) Schematic diagram showing the position of the cross-section part of the microchannels used for secondary flow analysis. (b) The flow vector map combining un and uz. (c) The color map of us on the cross section. (d) The color map of un on the cross section. (e) The color map of uz on the cross section. The center black lines in (b) to (e) are the symmetry plane used during simulation. ................................................ 167

Figure 8-1 (a) The configuration of a PDMS plate with a rectangular microchannel filled with solutions. The bottom wall of the microchannel is immobilized with 3 probe strips. (b) The schematic diagram showing the simplified 2-D geometry of the above microchannel used for later numerical simulation. The two diagrams are not drawn to scale. ................. 173

Figure 8-2 (a) Schematic representation of proposed heterogeneous DNA hybridization model. The probe DNA molecules are shown as blue wave-like ribbons in the center region of the glass substrate. The target DNA molecule is a red twisted ribbon with a red pellet at one end which represents the fluorescence label. The hybridized DNA are shown as duplexes in the probe region. (b) The three kinetic processes used in the later model construction regarding the heterogeneous hybridization of target DNA: k3 and k-3 represent bulk-to-surface hybridization and dessociation of targets to the bulk phase; k2 and k-2 represent surface-to-surface hybridization and dissociation of the nonspecifically adsorbed targets; and ka and kd represent the reversible nonspecific adsorption and desorption of the targets to the surface. ....................................................................................................... 175

Figure 8-3 An example of meshed 2-D geometry of the microchannel. The grid closed to bottom wall was meshed with much higher density because it is close to the reactive surface with probe spots.......................................... 184

Figure 8-4 (a) Concentration profile of static hybridization. (b) Concentration profile of microflow hybridization. The average flow velocity is 1 mm/s and the flow direction is shown as a blue arrow. In both (a) and (b), an array of 3 probe spots has been incorporated in the model. The probe

xxv

region width and spacing are at 200 µm. The sample concentrations at inlets are set to 10 nM and the hybridization time is 3 min.The bottom rainbow bars denote the bulk concentration scale. (c) Kinetic curves showing the change of the hybridization fraction vs. time during the static or microflow hybridization in 15 min. .................................................. 190

Figure 8-5 (a) The image of hybridization patches by flowing through 1-µL oligonucleotide samples at different concentrations in 12 min. The hybridization patches from two vertical repeated probe lines are shown. (b) The normalized hybridization signal intensities from the experimental results in (a). The error bars come from the standard deviation of 5 hybridization patches. The simulated hybridization signals are also shown for comparison. ....................................................... 192

Figure 8-6 (a) The fluorescent image of microchannel hybridization results at different flow rates. Six flow rates, from left to right, 2.0, 1.0, 0.5, 0.3, 0.2, and 0.1 µL/min have been applied. The hybridization image in the right most is from 30-min stop-flow hybridization. 1 µL of 10-nM oligonucleotide samples were applied to each microchannel. (b) The relative hybridization intensities from the left image. The error bars are from the standard deviation of 5 hybridization patches. ............................... 194

Figure 8-7 (a) The simulated results of the relative bulk target concentrations adjacent to the probe regions. (b) The simulated hybridization kinetics at different flow conditions. The hybridization time for the stop-flow method is 30-min. In the dynamic flow method, four flow speeds, from left to right, 0.01 to 1 mm/s have been simulated and the corresponding Pe values are from 10 to 10000. 1 µL of 10-nM oligonucleotide samples were applied to each microchannel. ..................... 196

Figure 8-8 (a) The simulated hybridization intensities in microchannels of 2 different depths. Here, 2 types of microchannels, 75 µm and 24 µm, are simulated and the hybridization time is 3 min in all cases. The Pe number of both types of microchannels is 1000. In stop-flow method, the hybridization time is set to 30 min. (b) Experimental results from flow hybridization of 1-µL samples in the 2 types of microchannels. ............ 197

Figure 8-9 (a) The experimental and simulation results of the effect of probe coverage on hybridization signals. Top curve with triangle labels shows the simulated hybridization signals at different probe coverage; Bottom curve with square labels shows the experimental hybridization signals at different probe coverage. (b) The simulated kinetic curves of hybridization at different probe densities. .................................................... 201

Figure 8-10 Sequence structures showing (a) a perfect-matched duplex (B-B’) and (b) a one-base-pair-mismatched duplex (B-NB’) from two 21-mer oligonucleotides. ......................................................................................... 203

Figure 8-11 The experimental hybridization results from the microfluidic hybridization of 6 groups of DNA. Each group represents two targets with one-base difference at the sequence center and the probe molecules are complementary to one of them. The 12 DNA solutions were hybridized in separated microchannels but under the same conditions at ~50°C and in 2.5x SSC for 3 min. ........................................... 205

xxvi

Figure 8-12 The simulated effect of the 2 Damköhler numbers (Da3 and Dn) on relative hybridization fraction after (a) 3-min, (b) 15-min, and (c) 60-min microfluidic hybridization. The corresponding dimensionless dissociation equilibrium constants (K3') are also shown as dashed lines in the 3 graphs. The colorscale bar beside each graph denotes the level of the relative hybridization fraction. .................................................... 207

Figure 8-13 The kinetic curves of the microchannel washing procedure at different dissociation constants. In all the cases the Da values were set to 1, which is a typical value for 20-mer duplex formation. ........................... 210

Figure 8-14 (a) Top: The fluorescence image of the hybridization results from pin-spotted DNA microarray. Bottom: The fluorescence intensity along the yellow dashed line in the top image. (b) Top: The fluorescence image of the stop-flow hybridization results from channel-spotted DNA microarray. Here, the sample channel was in horizontal and crossed two vertical probe lines. Bottom: The fluorescence intensity along the yellow dashed line in the top image. ............................................................ 211

Figure 8-15 (a) The simulated relative bulk concentration profile of target DNA at the bottom wall with a probe spot. Here, both static and dynamic hybridizations are simulated. For the latter method, 5 different flow rates were tested resulted in 5 Pe/Da ratios. (b) The simulated hybridization fraction profiles at the bottom wall with different hybridization methods. In all cases, the hybridization time is set to 3 min. ............................................................................................................. 215

Figure 8-16 The simulated hybridization fractions along the bottom wall with a probe region. The width of the probe region varies from 50 µm to 1mm. (a) Static hybridization. The insets show the experimental results from the hybridization of oligonucleotide samples with complementary probe lines with a width of 50 µm. The concentration of the oligonucleotide samples is 100, 10, and 1 nM, respectively, from left to right. (b) Dynamic hybridization. ................................................................................ 216

Figure 8-17 Simulated graphs showing the evolvement of bulk concentration profiles with a group of 3 probe spots. (a) After 1-min static hybridization. (b) After 15-min hybridization. (c) Simulation results showing the evolvement of the discrepancy of hybridization fractions on the 3 probe spots at different time. ......................................................... 218

Figure 8-18 The experimental results from the microfluidic hybridization of oligonucleotide samples at 4 concentrations. (a) Samples are prepared with 1x SSC. (b) Samples are prepared with 4xSSC. 0.1-nM targets were not detected in (a). The bottom numbers represent the probe row number. (c) and (d) show the simulated profile of bulk concentration during hybridization in 4x SSC and 1x SSC buffer, respectively. ................. 220

Figure 10-1 The 2-D geometry used in the COMSOL simulation. ................................ 233

xxvii

LIST OF TABLES

Table 1-1 Microfluidic DNA microarray hybridization conducted using the pressure-driven flow method. ........................................................................ 21

Table 1-2 Microfluidic DNA microarray hybridization conducted using centrifugal pumping. ....................................................................................................... 34

Table 3-1 Oligonucleotides and PCR products used in this study .................................. 65

Table 6-1 Oligonucleotide primers used in this study ................................................... 114

Table 7-1 Comparison of the simulation results from two different mesh densities. ..... 165

Table 8-1 The calculated free energies of 6 groups of DNA duplexes. ........................ 204

xxviii

LIST OF ABBREVIATIONS

2-D Two-dimensional

3-D Three-dimensional

AB 21-mer oligonucleotide probe designed for the detection of Botrytis cinerea. The 5’-end of the sequence is labelled by an amine group linked with a –(CH2)6– spacer

ALB 21-mer oligonucleotide probe designed for the detection of Botrytis cinerea. The 5’-end of the sequence is labelled by an amine group linked with a longer –(CH2)12– spacer

AD 22-mer oligonucleotide probe designed for the detection of Didymella bryoniae. The 5’-end of the sequence is labelled by an amine group linked with a –(CH2)6– spacer

ADF 22-mer oligonucleotide probe designed for the detection of Didymella bryoniae. The 5’-end of the sequence is labelled by an amine group linked with a –(CH2)6– spacer. The 3’-end of the sequence is labelled by a fluorescein molecule

APTES 3-Aminopropyltriethoxysilane

APTMS 3-Aminopropyltrimethoxysilane

bp Base pair(s)

cDNA Complementary DNA synthesized from a messenger RNA template

CB’ 21-mer oligonucleotides complementary to the sequence of probe AB, the 5’-end of the molecule is labelled with Cy5 dye

CB’50 Cy5 dye labelled 50-mer oligonucleotides. The central 21 bases are complementary to the sequence of probe AB

CB’P Cy5 dye labelled 264-bp PCR products amplified from genomic DNA of Botrytis cinerea

CBN’P Fluorescein labelled 264-bp PCR products amplified from genomic DNA of Botrytis squamosa, there is one base-pair difference in the

xxix

center from the sequence of CB'P

CCD Charge-coupled device

CD Compact disk

CD’ 21-mer oligonucleotides complementary to the sequence of probe AD, the 5’-end of the molecule is labelled with Cy5 dye

CD’50 Cy5 dye labelled 50-mer oligonucleotides. The central 22 bases are complementary to the sequence of probe AD

CD’P Cy5 dye labelled 259-bp PCR products amplified from genomic DNA of Didymella bryoniae

CE Capillary electrophoresis

CFD Computational fluid dynamics

Cy5 A fluorescent cyanine dye. The number of methine groups in its polymethine chain is 5

Da Damköhler number

DMSO Dimethyl sulfoxide

DNA Deoxyribonucleic acid

dsDNA Double-stranded DNA

EOF Electroosmotic flow

FB' Fluorescein labelled 21-mer oligonucleotides complementary to the sequence of probe AB

FD' Fluorescein labelled 22-mer oligonucleotides complementary to the sequence of probe AD

FB'P Fluorescein labelled 264-bp PCR products amplified from genomic DNA of Botrytis cinerea

FBN'P Fluorescein labelled 264-bp PCR products amplified from genomic DNA of Botrytis squamosa, there is one base-pair difference in the center from the sequence of FB'P

FD'P Fluorescein labelled 259-bp PCR products amplified from genomic DNA of Didymella bryoniae

xxx

GNP Gold nanoparticle

MEMS Microelectromechanical systems

MM Mismatch

MMA Microfluidic microarray assembly

ODE Ordinary differential equation

Oligo(s) Oligonucleotide(s)

PBS Phosphate Buffered Saline

PCR Polymerase chain reaction

PDE Partial Differential Equation

PDMS Polydimethylsiloxane

Pe Péclet number

PM Perfect match

PMT Photomultiplier tube

Re Reynolds number

RNA Ribonucleic acid

rpm Revolutions per minute

SDS Sodium dodecyl sulphate

SNP Single-nucleotide polymorphism

SSC Sodium chloride-sodium citrate buffer

ssDNA Single-stranded DNA

xxxi

LIST OF SYMBOLS

C Bulk-phase concentration of targets.

C0 Initial bulk-phase concentration of targets.

Cw Bulk-phase concentration of targets at bottom wall.

CDX Bulk-phase concentration of the duplex part in PCR product samples.

CLC Bulk-phase concentration of the loose-coiled target strands in PCR product samples.

CSC Bulk-phase concentration of the super-coiled target strands in PCR product samples.

dH Hydraulic diameter of the microchannel.

Ds Two-dimensional (i.e., surface phase) diffusion coefficient.

D Three-dimensional (i.e., bulk phase) diffusion coefficient.

Da1 Damköhler number of the target adsorption process.

Da2 Damköhler number of the surface-to-surface hybridization process.

Da3 Damköhler number of the bulk-to-surface hybridization process.

Dn Damköhler number of the dissociation process of the hybridized duplexes on the surface.

en Unit vector in n direction of the proposed orthogonal curvilinear coordinate system.

es Unit vector in s direction of the proposed orthogonal curvilinear coordinate system.

er Unit vector of radial distance in the cylindrical coordinate system.

eθ Unit vector of azimuth in the cylindrical coordinate system.

f Volume fraction of fluid in partially filled grids during the CFD simulation.

Fs Centrifugal force.

Fω Surface tension force.

g Gravitational constant.

h Scale factor of the proposed orthogonal curvilinear coordinate system in the s direction.

H Height of the microchannel.

J Mole flux of target DNA molecules in bulk solution.

k Degree of spirality of the spiral microchannel.

xxxii

k Unit vector in z direction of the proposed orthogonal curvilinear coordinate system.

kn Empirical nucleation constant.

ka Kinetic association constant for nonspecific adsorption of target strands.

kd Kinetic dissociation constant for nonspecific adsorption of target strands.

k2 Kinetic association constant for surface-to-surface hybridization.

k-2 Kinetic dissociation constant for surface-to-surface hybridization.

k3 Kinetic association constant for bulk-to-surface hybridization.

k-3 Kinetic dissociation constant for bulk-to-surface hybridization.

kdx Kinetic association constant for bulk-phase renaturation of PCR products.

k-dx Kinetic dissociation constant for bulk-phase renaturation of PCR products.

ksc Kinetic association constant for bulk-phase formation of the secondary structures of PCR products.

k-sc Kinetic association constant for bulk-phase formation of the secondary structures of PCR products.

KD' Dimensionless dissociation equilibrium constant for nonspecific adsorption of targets.

K2' Dimensionless dissociation equilibrium constant for surface-to-surface hybridization of targets.

K3' Dimensionless dissociation equilibrium constant for bulk-to-surface hybridization of targets.

KDX Dissociation equilibrium constant for bulk-phase PCR products.

KSC Dissociation equilibrium constant for the super-coiled part in PCR products.

l Liquid front displacement along the microchannel.

L Length of the microchannel.

n Across-channel distance in the proposed orthogonal curvilinear coordinate system.

NA Avogadro’s number.

Pe Péclet number of the microflow.

p Fluid pressure.

P0 Fluid pressure at the inlet of the spiral channel.

Pm Fluid pressure at the meniscus of liquid front.

Ps Fluid pressure from the surface tension of meniscus of liquid front.

r Radial distance from the rotation center to the liquid front.

r0 Initial radial distance from the rotation center to the start of the spiral channel.

r Average radial distance from the rotation center to the liquid column.

Δr The radial extent of the liquid column subject to centrifugal force.

xxxiii

R Position vector to a point described by (n, s, z) system within the microchannel.

Re Reynolds number of the microflow.

Rp Radius of each probe molecule occupied on the glass surface.

Rt Gyration radius of a DNA molecule in solution.

Rads Adsorption rate of DNA targets.

Rbhy Reaction rate of bulk-to-surface hybridization.

Rshy Reaction rate of surface-to-surface hybridization.

s Distance along the spiral channel in the proposed orthogonal curvilinear coordinate system.

t Time.

u Flow velocity in x direction (along the channel).

u Vector field of fluid velocity.

U Characteristic flow speed, measured as the average liquid front speed from the experiment.

us Fluid speed in the n-direction of the proposed orthogonal curvilinear coordinate system.

un Fluid speed in the s-direction of the proposed orthogonal curvilinear coordinate system.

uz Fluid speed in the z-direction of the proposed orthogonal curvilinear coordinate system.

W Width of the microchannel.

z Distance perpendicular to the spiral channel in the proposed orthogonal curvilinear coordinate system.

α Angle between the radial direction and the tangent of the spiral channel.

α0 Initial angle between the radial direction and the tangent of the spiral channel at the start position.

γ Surface tension coefficient of the liquid.

Δ Laplacian operator.

ΔT A customized operator similar to the Laplacian operator.

ΔG Free energy of duplex formation.

ΔΔG Free energy difference between MM duplex formation and PM duplex formation.

ε1 Ratio of bulk-phase targets to maximum adsorbable targets.

ε2 Ratio of surface adsorption capacity relative to hybridization capacity.

ε3 Ratio of bulk-phase target numbers to maximum hybridizable target numbers.

φ Ratio between the surface diffusion coefficient (Ds) and the bulk diffusion coefficient (D).

xxxiv

η Surface concentration of non-specifically adsorbed target DNA molecules.

ηmax Maximal number of the non-specifically adsorbed target DNA at a unit area of the glass surface.

θ Contact angle of the fluid on either PDMS or glass surface.

Θ Surface concentration of the hybridized duplexes.

Θ0 Probe density, equal to the maximal surface concentration of hybridized duplexes.

κ Curvature of the spiral microchannel.

ξ Strand complexity in nucleotide numbers (nt).

Γ Strand length in nucleotide numbers (nt).

σ3 Persistence length of the target DNA in bulk solution.

µ Dynamic viscosity of the fluid.

ν Kinematic viscosity of the fluid.

ω Angular speed.

Ω Angular velocity field.

ρ Fluid density.

χ3 Reaction success probability.

1

1: INTRODUCTION

1.1 Fundamentals of DNA microarray hybridization

1.1.1 The building blocks of DNA

Deoxyribonucleic acid (DNA), which is found primarily in the nucleus of a

cell, serves as the repository of genetic information. DNA is a linear polymer built

from different monomeric units called nucleotides. A nucleotide consists of three

chemical units including a nitrogenous base, a pentose sugar (2-deoxyribose),

and a phosphate group. The nitrogenous bases in DNA include two purines,

adenine (A) and guanine (G), and two pyrimidines, thymine (T) and cytosine (C).

The structures of the nucleotides and the four nitrogenous bases are listed in

Figure 1-1.

Figure 1-1 (a) General structure of the nucleotide. The nitrogenous base and the pentose sugar are linked each other through N-glycosidic linkage. The official numbering system is also shown. (b) The two purine nitrogenous bases in DNA. (c) The two pyrimidine nitrogenous bases in DNA.

(c) (b)

Pyrimidine Purine

O

HOH

HH

HH

OP-O

O-

OBase

1'

2'3'

4'

5'

(a)

N

NH

NH2

O

HN

NH

O

O

N

N NH

N

NH2

HN

N NH

N

O

H2N

Adenine Guanine Thymine Cytosine

2

1.1.2 DNA duplex structure

In its natural state, DNA is a double helix consisting of two polynucleotide

strands, as discovered by James Watson and Francis Crick in 1953 [1]. The two

strands wrap around each other in a helical "twisted ladder" structure, or double

helix, as shown in Figure 1-2(a). The covalent sugar-phosphate backbone is on

the outside of the double helix with alternating deoxyribose and phosphate

groups, and the nitrogenous bases are located in the centre (Figure 1-2(b)).

Figure 1-2 (a) Watson-Crick model of a 21-bp DNA double-helix. The pentose-phosphate backbones of the two strands (shown in blue and red ribbons) form a right-handed double helix. The DNA structure was generated with the DNA modeling server by Vlahovicek et al. [2]. Image visualization was performed with the Chimera molecular modeling system [3]. (b) Chemical structure of the specific hydrogen bonding in a DNA duplex. The duplex is stabilized by hydrogen bonding between complementary base pairs and (though not obvious from this drawing) by base stacking interactions between adjacent base pairs.

(b) (a)

N

NN

N

HN

O

O

P O

O

O

-ON

HN

ONO

O

P O

O

-O N

N

NO

NH

NO

O

P O

O

-ON

O

O

H3C

NO

OH

P O

OH-O

H

H

H

H

H

N N

N

NHN

O

O

PO

O

O

O-

N

HN

O

NO

O

PO

O

O-

N N

NO

NH

N

O

O

PO

O

O-

N

O

O

H3C

N

O

OH

PO

OH

O-

H

H

H

H

H

A

C

T

G

A

T

C

G

5’-end

5’-end

3’-end

3’-end

3

The most remarkable feature of the Watson-Crick structure is that the two

DNA strands are held together by the formation of specific hydrogen bonds

between nitrogenous bases. There are only two types of base pairings found in

the DNA duplex: A·T and G·C. The geometries of these complementary A·T and

G·C base pairs, the so-called Watson-Crick base pairs, are shown in Figure

1-2(b). The double helix structure is also stabilized by interactions between the

flat aromatic rings of adjacent base pairs [4]. Other combinations of bases (e.g.

A·G or T·T) destabilize the double helix structure [5].

1.1.3 DNA hybridization analysis

The formation of DNA duplexes from two complementary polynucleotide

strands is called hybridization or annealing. As depicted in Figure 1-3, the 2-step

process starts with the intermolecular collision between two strands to form an

intermediate state, in which a few base pairs have formed (also known as

nucleation). This is followed by the second step that the rest of the

complementary sequences join together much like a zipper closing [6, 7].

Because the second step is rapid and is in the order of 106-107 base pairs per

second, the first step of intermediate formation is the rate-limiting step during

hybridization [8, 9].

4

Figure 1-3 Schematic view of the DNA hybridization process. (a) Two complementary single-stranded DNA before hybridization. (b) Intermediate state with a few base pairs from strand collision. (c) Hybridized DNA duplex from zippering step.

The kinetics of DNA hybridization is dependent on the properties of the

DNA sequences and various experimental factors. In the former case, the

hybridization rate constants of DNA (in aqueous solutions) are inversely

proportional to the number of nucleotides of non-repeating sequences for the

given DNA [10]. Therefore, the hybridization rate of two shorter complementary

oligonucleotides is usually much faster than the hybridization of longer DNA

strands. In terms of experimental factors, the DNA hybridization rate is affected

by temperature, viscosity, and ionic strength. First, the apparent rate constant

shows a “bell-shaped” dependence on temperature, with a broad flat maximum at

temperatures that are from 15 to 30°C below the melting temperature [10, 11].

Second, because the DNA molecules are brought together by diffusion before the

intermediate formation [7], the hybridization rate decreases in viscous solutions

[10]. Third, DNA-DNA hybridization is also highly dependent on the ionic strength

of bulk solutions. The two negatively charged DNA strands strongly repel each

other at low salt (in NaCl) concentrations, and adding salt shields this repulsion

Zippering

Fast

Nucleation

Slow

(a) (b) (c)

5

[11]. At a constant temperature, the hybridization rate is proportional to the cube

of [Na+] in low-salt concentration (< 0.2 M), although this strong dependency

diminishes at higher [Na+] [12-14]. Moreover, it should be noted that during the

heterogeneous hybridization between solution-phase DNA and surface-

immobilized DNA, the hybridization rate can be slower due to the additional steric

effect involved in the surface reaction [15, 16].

DNA hybridization is a reversible process. For example, when the duplex

DNA solution is heated above a characteristic temperature, the double-helix

structure collapses and the two complementary strands dissociate into two

flexible single-stranded DNA. This duplex dissociation process is also known as

denaturation or melting. The characteristic temperature is named as the melting

temperature (Tm). Depending on the sequence properties (described as chain

length, GC content, and mismatch percentage) and the solution conditions (pH

and ionic strength), the double-stranded DNA can have a very distinct Tm value,

as reported elsewhere [17, 18]. Tm is thus an important thermodynamic

parameter in devising DNA hybridization assays, such as probe/primer design,

the selection of temperature in hybridization and washing steps, as well as the

development of methodology to identify DNA with single base-pair mismatch [19-

21].

Because of the high-strength, sequence-specific base pairing between

complementary DNA strands during hybridization, DNA molecules themselves

are the perfect set of agents to identify particular DNA sequences. Therefore, the

principle of DNA hybridization analysis is to use a known, well-characterized

6

population of DNA molecules (i.e. probes) to search for complementary

sequences within an unknown population of DNA samples of biological or

medical interest (i.e. targets). Here, probes are short single-stranded DNA

molecules prepared either from synthetic oligonucleotides (< 100 bases) or from

DNA fragments (< 1000 bases) [19]. On the other hand, target DNA samples are

frequently made by amplifying cell-based DNA extracts using polymerase chain

reaction (PCR) technology. Because the PCR products obtained are usually

double-stranded, the DNA targets have to be denatured to the single-stranded

format by heating before being exposed to probes. With proper labels on the

probe or the target, the DNA duplexes formed during hybridization could be

identified and quantitated.

Figure 1-4(a) depicts a schematic diagram of the preparation of

probe/target DNA for the hybridization test. With the knowledge of the genomic

DNA sequence, the probe DNA sequence (short 21-base oligonucleotide) is

selected based on the design criteria of specificity, melting temperature, and

structural stability [22]. The DNA probes are then synthesized chemically and

functional groups such as labels may be added during the synthesis. The

genomic DNA extracts from cell cultures are first denatured by heat and then

amplified by PCR to make the DNA target samples (260 bp). Fluorescence labels

can be integrated into the target sequence during the DNA amplification step.

The analysis for a particular DNA sequence is based on the premise that the

probe can find its complementary target in the sample of interest through

hybridization. Either labeled targets (Figure 1-4b) or labeled probes (Figure 1-4c)

7

can be used to identify the duplex, and the former format is adopted in this thesis.

In earlier solution-phase hybridization assays, the probe-target duplexes are

separated from the reaction mixture by chromatography and are quantitated by

the labels on the duplex [23]. The modern microarray technology relies on the

solid-phase hybridization for the ease of separation and this method will be

discussed in Section 1.1.5.

Figure 1-4 (a) Preparation of DNA targets from genomic DNA. Here, the genomic DNA duplexes are denatured to two single strands. One of the strands (highlighted in yellow color) are then amplified and labeled with proper tags for later hybridization analysis. (b) and (c) show the two formats of DNA hybridization analysis: (b) Target DNA is labeled. (c) Probe DNA is labeled.

(b)

(a)

Target

Probe

Target

Probe

Genomic DNA

(c)

... ACTTTGTTGCTTTGGCGAGCTGCCTTCGGGCCTTGTATGCTCGCCAGAGAATACCAAAACTCTTTTTATTAATGTCGTCTGAGTACTATA ... ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

... TGAAACAACGAAACCGCTCGACGGAAGCCCGGAACATACGAGCGGTCTCTTATGGTTTTGAGAAAAATAATTACAGCAGACTCATGATAT ...

1) Heat denaturation 2) PCR amplification 3) Labelled with fluorescence tags at the end

... TGAAACAACGAAACCGCTCGACGGAAGCCCGGAACATACGAGCGGTCTCTTATGGTTTTGAGAAAAATAATTACAGCAGACTCATGATAT ...

Selected region for target recognition

DNA targets ready for hybridization analysis Fluorescence

labels

Synthetic oligonucleotide probes

CGCCAGAGAATACCAAAACTC

8

1.1.4 Hybridization of mismatched DNA sequences

Other than the hybridization of complementary DNA sequences in

perfectly matched (PM) situation, DNA sequences with a few mismatched (MM)

base pairs can hybridize too. For example, in Figure 1-5, the base thymine lining

up opposite to a guanine group would be a mismatch (The T·G mismatch). As

long as most of base pairs in the duplex do match and form hydrogen bonds, the

strands will hybridize, albeit giving rise to a distorted and less stable duplex.

Figure 1-5 (a) Distorted DNA duplex structure from two T·G mismatches in the center. The two green arrows indicate the two T·G mismatches. (b) The sequence of the duplex shown in (a). The blue strand is 5'-CGCATTACGC-3', and the red one is 5'-GCGTGGTGCG-3'. (PDB access number: 1SNH [24])

The study of DNA duplexes with mismatched base pairing has significant

implications. In particular, during studies of single nucleotide polymorphism

(SNP) and point mutation, there is only a single nucleotide variation at a specific

location in the genome (the variation can lead to the development of disease in

some cases) [25-28]. It is thus important to recognize and reliably discriminate

the target sequence from its peers with only single base difference.

(a) (b)

C C C A

T T

A C G C

G C G T

G G

T G C G

9

Hybridization under proper conditions (e.g. temperature) is a common

method used to achieve discrimination between the DNA sequences with single

base-pair difference [25]. The approach is based on revealing differences in the

stability of perfectly-matched (PM) and mismatched (MM) DNA duplexes,

because the presence of MM base pairs significantly reduces the overall stability

of the duplex. This can be observed through a reduction of the Tm of MM

duplexes as compared to that of PM duplexes [29, 30]. The destabilization effect,

or reduction of Tm, is determined by a number of factors including the probe

length (e.g. 20-mer or 50-mer), the mismatch type (e.g. T·G vs. T·T), etc. [14,

31]. For instance, the Tm is estimated to decrease by 1°C for every 1%

mismatch in the duplex (e.g. 1-bp mismatch in a 100-mer probe) [30]. Therefore,

the reduction of duplex Tm due to single base-pair mismatch diminishes for longer

duplexes. Hence it is necessary to use short oligonucleotide probes (typically ~20

bases) to increase the percent of mismatched region for the successful

discrimination of MM DNA targets [13].

10

Figure 1-6 Hybridization-based approaches for the discrimination of SNP sequences. (a) Targets are prepared from two SNP sequences with only one base-pair difference. (b) Fluorescence-labeled targets hybridize to probe molecules. The mismatched sequences form unstable duplexes. (c) The unstable MM duplexes are denatured and removed by high-stringency washing conditions. (d) Fluorescence signals from retained PM duplexes.

By selecting proper hybridization conditions, PM duplexes can be

distinguished from MM duplexes. In hybridization applications, the degree of

mismatching tolerated in duplex formation is called “stringency” [20]. Under

conditions of low stringency (low temperature or high salt concentration), even

mismatched strands will be able to bind and give a hybridization signal, although

it is a false positive one. Under conditions of high stringency (high temperature or

low salt concentration), only PM strands will hybridize and stay together.

Therefore, by carefully controlling test conditions to an acceptable level of

stringency, MM targets can be removed from binding to probes and only PM

targets give hybridization signals in a reproducible manner (Figure 1-6).

(b) (a)

High-stringency washing

(c)

AT

GT

AT

T

Fluorescence analysis

ATAT

GCGC

T

SNP specific probes

Two sequences with single base-pair difference

(d)

11

1.1.5 DNA microarray and its applications

DNA microarray can be defined as a miniaturized, ordered arrangement of

different probe DNA populations fixed onto a solid surface at precisely defined

positions in a grid format, enabling the analysis of numerous genes in parallel by

specific hybridization [14, 32]. The concept of DNA microarray was evolved from

the Southern blotting technology using the solid-phase hybridization method

developed in the early 1990s [33]. Compared to the solution hybridization test,

where the probe-target duplexes have to be isolated by chromatography [34],

microarray hybridization is more efficient in the removal of un-hybridized DNA

molecules and in the reduction of reagent consumption due to the miniaturized

architectures. Another significant advantage of the microarray method is that the

DNA identification processes can be completed in a parallel manner. DNA

microarrays have been widely used in many applications in molecular biology,

biochemistry, and biotechnology fields, as reviewed elsewhere [35-39].

In general, there are two formats of DNA microarray, depending on the

probe arraying technologies: in situ synthesized or spotted. The first format

involves light-directed in situ synthesis of oligonucleotide probe arrays, which is

pioneered by Affymetrix GeneChip® technology [40, 41]. The technology

features high probe density, where more than 6 million different oligonucleotide

populations can be integrated on a chip [42]. Another cheaper and more flexible

format is to spot probe DNA molecules onto a functionalized glass slide [43]. The

probe DNA molecules can be either DNA fragments with known identities [44] or

pre-synthesized oligonucleotides [45]. As shown in Figure 1-7a, a robotic arrayer

12

with multiple pins is used to simultaneously pick up and deposit DNA solutions.

The probe molecules can also be spotted by inkjet methods [46] or more

recently, by microfluidic methods. The latter approach will be further discussed in

the following sections.

When DNA microarray is used in a hybridization assay, a test sample,

which is an aqueous solution containing fluorescently labelled target DNA, is

applied to the microarray slide and allowed to hybridize to the probe molecules

(Figure 1-7c). After the slide is washed to remove nonspecific hybridization

(Figure 1-7d), it is read in a high-resolution laser scanner (Figure 1-7e). The

signal emitted from each spot is then collected to produce an image, where

differences in fluorescent labels and signal intensities result in different colors

with different brightness, respectively (Figure 1-7f). Image-processing programs

are used to find the spots and quantify them [47]. The hybridized target can also

be measured chemiluminescently, electrochemically, or radiochemically with

different labels [35, 48]. The hybridization signal intensity of a microarray spot

depends on the target abundance, but also on the particular probe-target binding

affinity.

13

Figure 1-7 Hybridization assay with spotted probe microarray. (a) Microarray probe spotting pins used in a commercial microarrayer. The photo is adapted from the product datasheet of SpotBot® Titan High-Capacity microarrayer. (b) Glass slide with chemically-modified surface ready for probe binding. The slide can be labeled with barcode for easy identification. (b) Microspotting probe molecules on the slide. The lower image shows surface tethered probe sequences. The individual probe sequences can be identified by the position of the corresponding feature on the regular microarray grid. (c) Hybridization. Target solutions are applied to the slide and are incubated in a heated humidified box. Labeled targets can freely diffuse over the microarray surface until they hybridize with a complementary probe (lower image). (d) Washing step. After hybridization, unbound targets are washed away with buffer solutions as shown in the lower image. (e) Microarray image analysis. Quantification of the surface bound targets is performed by measuring the fluorescence intensity of the hybridization spots using a fluorescence scanner.

Although the DNA microarray technology was developed only very

recently, numerous important applications have already been developed, and

their impact on biomedical research and diagnosis is dramatic. Here, tracking

gene transcripts and assaying DNA variation (genotyping) are considered as the

(b) (c) (d) (e) (f)

Scanning Washing Hybridization Spotting

(a)

14

main applications of the DNA microarray method [14, 39]. The first application is

used to compare gene expression profiles of closely related cell types, or of

identical cells that are exposed to different environments, such as the presence

of a drug [49, 50]. In terms of DNA variation assays, microarray hybridization

could help in tracking disease genes. It can also be used to identify

microorganisms, and this is important in pathogen control, safe food production,

and agricultural industry [51]. Moreover, vigorous efforts have been made to

identify and catalogue single nucleotide polymorphism (SNP) markers [20]. In this

thesis, the second application of microarray hybridization has been explored for

the identification of fungal pathogen DNA. At last, it should be noted that in

addition to transcript profiling and genotyping, microarrays can also be used in

studying protein-DNA interactions [52], DNA sequencing [53], and many other

applications [39].

1.2 Introduction to microfluidic DNA microarray technologies

With the rapid growth of microelectromechanical systems (MEMS),

microfluidic technology has developed quickly over the past two decades. By

combining the fields of microfluidics and DNA microarrays, the advantages of

both fields can be exploited simultaneously [54-56]. Since microfluidics deals with

the transfer and control of a small amount of fluids in microscale flow

configurations, one obvious advantage of using a microfluidic system is the

dramatic reduction in the sample volume. Instead of handling a volume of the

milliliter scale, a microfluidic chamber could have a volume as little as 1 picoliter

[56]. Samples such as proteins and DNA extracts from a few selected cells are

15

scarce or unavailable in large amounts; microfluidics therefore provides a way to

analyze these materials efficiently. The small volumes of the microfluidic systems

also make it possible to develop compact and portable lab-on-a-chip devices that

can be manufactured at low cost by mass production.

The second advantage of using microfluidic technology is that the

microarray hybridization kinetics of target nucleic acid can be accelerated. In

conventional DNA microarray hybridization, samples, which are typically >30μL,

are incubated for ~12h with the probe microarrays on a glass slide by covering

the solution with a thin coverslip [19]. At first, this static hybridization is

characterized by a reaction-limited process. After an initial time period, target

depletion renders the hybridization process diffusion-limited [57, 58]. Because

diffusion coefficients (D) for DNA (21–6000 bp) in aqueous solutions are on the

order of 10-7 cm2·s-1 [59], the average travel distance ( L Dt= ) of the target

nucleic acid solely due to diffusion in 24 h is ~1 mm. Considering that the

horizontal length scale of a microarray is on the order of a few centimeters,

hemispherical depletion volumes form around each probe [60]. Pappaert et al.

found that for an analysis time of 24 h the maximum binding efficiency was less

than 0.2 %, and a six-day analysis would be required for an efficiency of 2% [61].

On the other hand, recent microfluidic technology is characterized by the

advantage of the reduction of the diffusion distance. Moreover, liquid movement

in microchannels enhances the mass transport of target molecules and

accelerates the hybridization of target molecules with probes on the substrate

surface.

16

Microfluidics also offers the advantage of multi-sample capabilities on one

single chip. Conventional microarray experiments usually employ one sample on

one glass slide [19]. However, in the applications of genetic mutation analysis,

clinical diagnostics or microorganism identifications, direct comparison between

different samples on the same chip would be preferable because the quality of

slides with probe arrays varies from batch to batch [36]. Microfluidics allows for

the delivery of controlled volumes of samples and reagents to the DNA

microarray. By integrating multiple channels onto one chip, high-throughput

multisample analysis has been achieved [62-65].

In the development of microarray hybridization assays, many different

parameters such as temperature, ionic strength, and washing buffer conditions,

need to be optimized. The microfluidic method has the capability of handling

multiple samples (e.g. with different concentrations or different ionic strengths) as

well as accurate control in liquid flow and temperature, and such a method has

been developed for automated selection of optimal assay parameters [66].

Applications involving the combination of microfluidic technology with DNA

microarray hybridization are listed in Table 1-1. The table summarizes the

sample properties (sample types, sample number, sample volume and

concentrations), chip properties (chip materials and chip bonding methods), and

flow and hybridization conditions of various experiments.

1.2.1 DNA hybridization in microfluidic chambers

In conventional microarray hybridization protocols, DNA sample solutions

are loaded directly onto the glass slide surface and are incubated for hours. To

17

overcome the problem of the diffusion barrier during DNA hybridization, several

methods have been employed. For instance, some early improvements utilized

electric fields to concentrate the negatively charged molecules at probe zones.

By integrating microelectrodes onto the microarray substrate, negatively charged

DNA molecules in sample solutions were electrically attracted to the probe area

(positively charged microelectrodes) and the hybridization process was thus

accelerated [67-69]. In another approach, a microfluidic chamber was used to

cover the area with spotted probes; sample DNA solutions were then delivered to

the chamber and hybridized with the probes [70]. When the microarray area was

small and the chamber covered only a small part of the glass slide, simple

pressure-driven flow could be used [71, 72] as the solution containing the target

nucleic acids was able to move throughout the chamber and encounter every

single probe. When the microarray area is larger, however, the total sample

volume was increased and the efficiency of the hybridization was reduced. In

order to solve this problem, several approaches have been developed to improve

the efficiency of the chamber hybridization process and reduce the total sample

volume. For example, Lee et al. proposed a recirculating microfluidic device for

the hybridization of oligonucleotides to DNA microarrays [73]. The probe array

has 676 features and its total area is 62.4 mm2. The circulatory flow was

generated through a peristaltic pump connected to both ends of the

microchamber (Figure 1-8a). For the results from complementary DNA targets

shown in Figure 1-8b, the sample volume was 100 µL and hybridization time was

completed at 2 h, as compared to 16 h in the conventional static method [19].

18

Quake and co-workers constructed a more complicated PDMS device for larger

hybridization chambers (270 mm2) [74]. As shown in Figure 1-8c, the bifurcating

channel design equalized the liquid flowing into the chambers and the

herringbone indentations in the bridge channel promoted dynamic mixing.

Signals from their microfluidic hybridization method are over 3-4 fold higher than

those of the static hybridization method (Figure 1-8d).

Figure 1-8 (a) Diagram for sample recirculation system on the hybridization chamber. (b) Hybridization image of fluorescence-labelled target nucleotide with 0.1 mL·min-1 continuous recirculating flow on a DNA microarray for 40 min. Reproduced from [73] by permission of The Royal Society of Chemistry. (c) Chaotic mixing of the dye solutions in two PDMS microfluidic chambers facilitated by the herringbone indentations in the bridge channel. (d) Signal improvement by microfluidic mixing on a home-spotted microarray. Top: Dynamic; Bottom: static hybridization. Reprinted with permission from [74], Copyright © 2006 John Wiley & Sons, Inc.

(b) (a)

(d) (c)

19

For a microfluidic chamber with a size comparable to a standard

microscope slide, liquid profiles have to be carefully designed to achieve an

equally distributed coverage of sample solutions over the microarray area as well

as minimizing the amount of air trapped during filling. Different methods have

been developed to induce convective microflows inside chambers. Adey et al.

integrated two air-driven chambers on the chip and the pneumatic system

created a convective flow of hybridization fluid across the microarray slide [75].

Similar reciprocal flow was achieved by employing a mechanical rotator [60]. In

addition to these methods, shear-driven microflows from moving the coverslip

back and forth or circularly were also explored [61, 76, 77]. Moreover, Yuen et al.

proposed a closed-loop microfluidic device consisting of two interconnected

reaction chambers. Circulatory microflows in chambers were generated by the

rotation of the magnetic stirring bars in the end reservoirs [78]. Liu et al.

introduced a method called cavitation microstreaming (Figure 1-9), where

acoustic microflows accelerated chamber hybridization by up to 5 fold in reaction

kinetics [79]. Other groups have proposed designs to increase hybridization

efficiency in large chambers [80-82]. For example, McQuain et al. reported a

coordinated pump and valve system to construct a chaotic mixer to enhance

microarray hybridization, as shown in Figure 1-10. It was reported that the

number of targets hybridized in dynamic hybridization was ~3-fold more than that

in the static hybridization, and the reaction time was shortened from 24 h (static)

to 4 h (dynamic) [80].

20

Figure 1-9 (a) Schematic showing a number of air bubbles in the top layer of the DNA biochip chamber. An energetic cavitation streaming motion created by the PZT transducer was observed in the vicinity of each bubble. (b) Top image: fluorescence image of a biochip after 2-h static hybridization; Bottom image: fluorescence image of a biochip after 2-h hybridization aided with cavitation microstreaming. Reprinted with permission from [79], Copyright © 2003 American Chemical Society.

Figure 1-10 (a) Illustration of the coordinated pump and valve used for dynamic hybridization. Fluids are extracted and reinjected to the hybridization chamber under a protocol to induce chaotic movement inside the chamber. (b) Photographs of dye solution mixing in the hybridization chamber after 2 cycles of liquid removal and reinjection processes using a 30-µL stroke volume. The dye distribution appears to be homogeneous after 16 cycles. Reprinted with permission from [80], Copyright © 2004 Elsevier Ltd.

(b) (a)

(b) (a)

21

Table 1-1 Microfluidic DNA microarray hybridization conducted using the pressure-driven flow method.

Ref. Number of probe features

Sample volume

(μL)

Sample concentration

Sample type Number

of samples

Hybridization time and

temperature

Chip materials*

Substrate materials

Chip bonding* Flow method and applications

[71] High 95 >500 copies

in vitro transcribed 1.6

kb RNA 1 20 min at

37°C PC PC Screws and clamp

Pneumatic controlled flow, integrated sample extraction, amplification and washing facilities.

[80] 102 30-50 N/A 725bp PCR amplicons 1 10 or 60 min

at 42°C Acrylic Glass Rubber

gasket with bolts

Syringe pumping, chaotic mixing chamber design.

[76] 1760 5 8µM cDNA 1 60 min at 42°C

borosilicate glass Glass Clamp Shear driven by a rotating chamber, compared with a

diffusion-driven hybridization system.

[63] 16 10 N/A PCR amplicons 32 2 h at 50°C PMMA Glass PDMS Consists of 32 individually addressable array reaction chambers.

[77] 360 5 to 209 N/A PCR amplicons 1 30 or 60min at 42°C

borosilicate glass Glass Clamp Shear driven by a rotating chamber, found that velocity

field disturbs the binding process at the binding site.

[83] 90 200 5pM oligonucleotides 1 15 min at 50°C

PS, PC, PMMA, or

PP

PS, PC, PMMA, or

PP

Adhesive tape

Syringe pumping, oscillation flow, investigated for attaching oligonucleotide probes on four different types of plastic surfaces.

[74] 9500 70 0.8 or

1.6 ng/µL

cDNA 1 2h at 52°C PDMS glass Reversible bonding

Peristaltic pumping enhanced hybridization signal from microfluidic chaotic mixing.

[61] 216 7.5 or 20 1ng/µL cDNA fragments 1 10 to 30 min at 42°C

borosilicate glass Glass Weight Induction of a convective flow using a shear-driven

microchannel flow system.

[75] 6912 60 N/A PCR amplicons 1 Overnight at 42°C plastic Glass Laminating

adhesive Pneumatic control, integrating two air-driven bladders that continuously mix the hybridization fluid.

[79] high 45 10nM oligonucleotides or PCR amplicons up to 4 2 h at 37°C PC Glass Double-

sided tape Microflow induced from acoustic microstreaming.

[78] 96 150 0.12-6 pg/µL cDNA 1 2 h at 42°C PDMS or

plastic Glass Double-sided tape Fluid circulation inside the chamber by magnet stirring.

[72] 1000 10 10 nM oligonucleotides 1 5 h at 30°C PC PC Polyolefinic or PP foil.

Syringe pumping, online monitoring of microarray hybridization on a polymer chip.

[73] 676 400 1µM oligonucleotides 1 Various, up to 2h Glass Glass Mylar layer

(PET)

Peristaltic pumping, recirculation flows in the chamber, hybridization kinetics and the effects of chamber configuration were studied.

22

[84] low 6 50pM oligonucleotides 1 up to 6min PDMS glass Reversible bonding

Syringe pumping, characterization of DNA hybridization kinetics in a microchannel.

[85] 5 2.8nL to 460 nL

50pM to 500

pM oligonucleotides 1 20 to 120 sec PDMS glass Reversible

bonding Electrokinetic delivery of nanoliter sample volumes and rapid removal of nonspecific adsorption.

[86] 4 2 down to 10 pM

oligonucleotides or PCR amplicons 1 2 min at 50°C PMMA PMMA Heat

annealing

Syringe pumping, detected point mutations in oncogene fragments using universal zip code arrays placed into plastic microchannels.

[87] 8 N/A N/A oligonucleotides or 1.4 kb ssDNA 1 30 min PMMA glass N/A

Pneumatic control, oscillation flow, hybridization efficiency improved by introducing velocity and extensional strain rate.

[88] 9 N/A N/A oligonucleotides or 1.4 kb ssDNA 1 30 min at

42°C glass glass Clamp Pneumatic control, oscillation flow, combined with hot-region temperature and cold-region flow effects provided additional efficiency in DNA hybridization.

[89] 16 25 N/A PCR amplicons up to 4 30 min at 50°C PC PC Adhesive

tape Plastic chip, integrated air pump to allow oscillation of the hybridization mixture.

[90] 50 1 50pM-90nM oligonucleotides 1 500s at RT PMMA glass Clamp Syringe pumping, oscillation flow, discrimination of

single-base-pair-mismatch.

[91] 8 0.3 10pM oligonucleotides 8 5min at 22°C PDMS glass Reversible bonding

Vacuum suction, intersection format, proposed theoretical models to predict hybridization rates for both microfluidic and conventional microarrays.

[92] 8 0.5 0.1-2 μM oligonucleotides 8 30 min at 20-

40°C PDMS PC Reversible bonding

Intersection format, fast surface activation of plastic substrate.

[93] 16 N/A N/A oligonucleotides or PCR products 16 1h at 55°C PDMS PMMA or

PC Reversible bonding

Hydrostatic flow, intersection approach hybridization on plastic chip, parallel detection of two low-abundant DNA point mutations in the oncogenes.

[94]ǂ 16 1 1.4 ng/μL

oligonucleotides or PCR amplicons 16 5 min at 50°C PDMS glass Reversible

bonding Vacuum suction, intersection approach hybridization, discrimination of two pathogen PCR amplicons.

[95] ǂ 16 1 N/A oligonucleotides or PCR amplicons 16 RT PDMS glass Reversible

bonding

Vacuum suction, intersection approach hybridization, room temperature discrimination of two pathogen PCR amplicons.

[96] 6 1 N/A 1.7kb RNA N/A 1 h at 27°C PDMS gold Reversible bonding

Pressure pumping, intersection format, SPR imaging measurements of microfluidic hybridization channels.

* N/A: data not available; PS: polystyrene; PC: polycarbonate; PMMA: poly(methylmethacrylate); PP: polypropylene; PET: Polyethylene terephthalate; RT: room temperature; SPR: surface plasmon resonance. ǂ: The work was done by me as given in Chapter 3 (for ref. [94]) or Chapter 4 (for ref. [95]).

23

1.2.2 Microfluidic DNA hybridization with low-density probe arrays

Although the microfluidic system offers significant improvement over the

conventional bulk solution method for microarray hybridizations, two concerns

arise from the use of large chamber hybridization on high-density microarrays. As

described in the previous section, liquid flows over the large chamber have to be

carefully designed to achieve an equally distributed liquid movement across the

slide as well as to avoid trapping air bubbles in the chamber during filling. In

addition, because a huge number of probe features are packed into a small area,

great efforts are needed for data processing, normalization, and interpretation in

high-density microarray hybridizations [97]. In gene expression analysis, many

thousands of genes must be simultaneously monitored to create a global picture

of cellular function, and so high-density microarrays are needed. On the other

hand, in many gene diagnostic applications, once a relatively small number of

genes are identified, low-density microarrays could be designed to screen these

genes [98]. This approach of low-density microarray has been proved to be

reliable, cost-effective, and much faster in data analysis and interpretation [99-

105].

1.2.2.1 Low-density microfluidic microarray method with pin-spotted probes

The microfluidic method is suitable for incorporation into low-density DNA

microarray analysis. The use of microchannels instead of large hybridization

chambers alleviated the need for complicated chip design to achieve

homogeneous hybridization across the chamber area. The pin-spotting method

that has been an often used technique in the fabrication of DNA microarrays [43],

24

has also been used in microfluidic microarray experiments. Both straight

microchannels [84-87, 106-108] and serpentine microchannels [66, 88-90, 109]

have been designed. For example, Chung et al. constructed a low-density

microarray with 5 probe spots and they found a signal increase of up to 6-fold in

a straight microchannel with flow hybridization, as compared to static

hybridization (Figure 1-11) [87]. The authors also found that the hybridization

efficiency was further improved by introducing extensional strain from the change

of flow velocities inside the microchannel [87, 110]. Wei et al. proposed a

microchannel hybridization method with microtrenched serpentine channels, as

shown in Figure 1-12 [90]. Sample consumption was reduced to 1 µL and

hybridization time was as low as 500 s for oligonucleotide targets. The signal-to-

noise ratio was improved by 30-fold in the discrimination of 2 oligonucleotides

with one base difference (Figure 1-12b). In terms of probe spotting, either

commercialized small arrayers [88, 89, 106, 109] or even micropipettors [85, 86,

107] have been used.

25

Figure 1-11 (a) Schematic drawing of a straight hybridization channel and probe spots (unit: mm). The flow direction was from top to bottom. (b) Fluorescent images after 30 min hybridization. Left: hybridization results at u = 1 cm·s-1; right: hybridization results at u = 0 cm·s-1 (control). (c) Comparison of the averaged fluorescence intensities after hybridization. Reproduced from [87] by permission of The Royal Society of Chemistry.

Figure 1-12 (a) An illustration showing scrambled discrete plugs being swept over probe rows in the serpentine channel. The black arrow denotes the shuttle flow inside the microchannel. (b) The fluorescence images are from 500-s microfluidic hybridization (top row), 500-s static hybridization (center row) or 2-h static hybridization (bottom row). The left column represents the results from 20-mer PM targets, and the right column represents the results from 20-mer MM targets with single base-pair mismatch in sequence center. Reprinted with permission from [90], Copyright © 2005 Oxford University Press.

(a) (b) PM MM

(b) (a)

Unit: mm

(c)

26

1.2.2.2 Low-density microfluidic microarray method using microfluidic printing and the intersection approach

In low-density DNA microarray analysis, the conventional pin-spotting

method has been used to create dot-like probe arrays [43]. The performance of

the hybridization assay is thus heavily influenced by the quality of printed probe

spots. In practice, because the spotting solutions are open to air and the spotting

pins are very close to each other, the pin-spotting method may suffer from

problems such as splashing, uneven evaporation and cross-contamination [111].

Clearly, the ability to create probe spots of high homogeneity would be beneficial,

because this would simplify image analysis and considerably enhance the

accuracy of signal detection. When a spotting buffer such as saline sodium

citrate (SSC) solution is used, spot uniformity is often poor due to the

hydrophobic properties of the microarray chip surface. Supplementary chemicals

such as dimethyl sulfoxide (DMSO) can reduce solution surface tension, and

hence improve spot uniformity. But DMSO could also introduce new problems,

such as spot size increase, as shown in Figure 1-13a [112]. Moreover, during the

blocking and cleaning procedures in the chemical post-processing steps, the

remaining unreacted probe molecules could diffuse away and smear the slide,

giving rise to “comet tail” artefacts [19, 113, 114]. Furthermore, when

microchannel hybridization is to be used later with the spotted microarray,

additional devices such as bolts and clamps must be used to ensure that the

entire hybridization microchannel is well aligned to probe rows without liquid

leakage [90].

27

A simple and effective method is to use a microfluidic network to print

probe line arrays on a surface. In this technique, an elastomeric plate with

microchannel features is bound to another solid substrate to form the microfluidic

network. Upon flowing sample solutions through the microchannels, different

probe lines can be printed. For instance, species such as small molecules [115],

lipids [116-119], proteins [120-127], DNA [91-93, 96, 128, 129], and

nanostructures [95, 130-132], have been patterned on a variety of solid surfaces

(gold, glass, or polymer). Because the immobilization reaction between the

biomolecules and the surface is confined in microchannels, uneven evaporation

and splashing are prevented. In practice, less than 1 µL of reagent is needed to

fill through each microchannel and the solution can be kept at room temperature

for hours without drying out [94].

Figure 1-13b to d show examples of patterned biomolecules (protein and

DNA) or nanoparticles on solid surfaces using the elastomeric microchannel

plate. The surface was typically chemically-modified to anchor different types of

molecules. The bonding between the polymer channel plate and the glass slide

was reversible, and a homogeneous distribution of the probe molecules along the

region enclosed by microchannels was achieved. The resolution of microfluidic

line patterning can be very small down to the submicrometer scale [126].

Moreover, because the flow operation in each microchannel can be manipulated

individually, the probe immobilization step and subsequent processes such as

washing and blocking can be independently conducted, and the individual

channels can thus be used for condition optimization tests (Figure 1-13c). Cross-

28

contamination resulting from the diffusion of unreacted agents is avoided during

the washing and blocking steps. The microfluidic approach is simple and flexible

for the printing of low-density probe arrays.

Figure 1-13 (a) Fluorescent image of the hybridization of DNA probes (5 and 10 µM) spotted on activated glass slides with increasing concentration of DMSO. Reprinted with permission from [112], Copyright © 2003 Oxford University Press. (b) Fluorescence images acquired in Cy3 (green) and Cy5 (red) channels show the parallel microfluidic patterning of multiple DNA probes. Reprinted with permission from [128], Copyright © 2009 American Chemical Society. (c) A 16-channel microfluidic network was used to pattern 20-µm lines of decreasing antigen coverage onto a PDMS substrate (left line to right line) using solutions of different concentrations. The antigen was labelled with rhodamine. Reprinted with permission from [120], Copyright © 2000 American Chemical Society. (d) Cross-polarized optical image of [Mo3Se3

-]∞ nanowire patterns on glass substrate made with microchannel networks. Reprinted with permission from [131], Copyright © 2000 American Chemical Society.

Low-density probe line arrays created from microfluidic methods have

been used successfully for nucleic acid analysis [65, 91-94, 96, 129, 133, 134]. In

these applications, the intersection approach was used. This approach was first

(a) (b)

(c) (d)

29

proposed for immunoassays [121-124] and patterning nanostructures [130]. It

was soon expanded to DNA analysis [65, 91-93, 96] and lipid-protein interaction

studies [119, 135]. As shown in Figure 1-14a, probe lines were printed onto a

chemically-modified glass slide, which was reversibly sealed with the first

microchannel plate. After peeling off the first plate, a second channel plate used

for target recognition was then assembled with the same glass slide. Here,

microfluidic channels were arranged vertically and were orthogonal to the

preprinted probe lines. Sample molecules flowing through the microchannels

intersected with the probe line arrays. If specific or complementary probe

molecules were encountered, the target molecules were retained and rectangular

patches were thus formed. These patches could then be detected by surface

plasma resonance imaging (Figure 1-14b) [96], or more often, by fluorescent

flatbed scanner [65, 91-94, 129, 133, 134]. Compared with the pin-spotted

microarray (in Figure 1-13a), the final hybridization patches obtained from the

microfluidic microarray were more regular and well organized which facilitated

the subsequent image analysis.

30

Figure 1-14 (a) Intersection approach for performing an immunoassay on a surface with microfluidic network. (i) Patterning different antigen molecules along the horizontal lines on a solid substrate. (ii) The area of the substrate left unpatterned during step (i) is blocked with bovine serum albumin (BSA) to prevent nonspecific binding in subsequent steps. (iii) Antibodies flowing through the channels of a second microfluidic network locally bind to the patterned antigens. (iv) Reading the rectangular binding patches reveals the amount of antibodies present in the samples. Reprinted with permission from [124], Copyright © 2001 American Chemical Society. (b) SPR difference images showing hybridization of a RNA sample onto an array of surface-bound probe DNA. Hybridization of the target RNA onto the probe DNA array was indicated by a change in the reflectivity. Reprinted with permission from [96], Copyright © 2001 American Chemical Society.

The intersection approach on the microfluidic microarray is well suited for

parallel sample hybridizations. Unlike the pin-spotted low-density DNA

microarray, the use of probe line arrays alleviates the need for precise alignment

between the hybridization channels and the probe dots [66]. The method can be

used for the investigation or comparison of optimal assay conditions, and for high

throughput multi-sample analysis. Benn et al. compared the hybridization rates

of 60-mer oligonucleotides in 8 concentrations from 10 pM to 10 nM on one chip

(a)

(b)

(i)

(ii)

(iii)

(iv)

31

[91]. Situma et al. used a 16 ×16 array to detect two different low-abundant DNA

point mutations in oncogenes [93]. We have also used a 16 ×16 microfluidic

microarray device to test the effect of probe coverage, ionic strength and

incubation time on DNA hybridization all conducted on the same chip. With

optimized conditions, we successfully discriminated PCR product samples from

two fungal pathogen genomic DNA [94]. More details can be found in Chapter 3.

1.2.3 Microfluidic DNA microarray using centrifugal pumping

Microfluidic DNA microarray analysis with high sample throughput

demands parallel sample hybridizations in multiple microchannels, which raises

the question of finding an effective way to deliver liquid simultaneously. For the

conventional pressure-driven method, Huang and colleagues recently presented

a microfluidic device integrated with pneumatically-controlled microvalves and

micropumps for parallel DNA hybridizations [64]. This device has been applied to

analyse 48 different DNA targets (18-mer oligonucleotides derived from the

Dengue virus genes) simultaneously (Figure 1-15). However, the pressure-driven

method could be complicated because each microchannel has to be connected

to pump tubing and synchronization has to be considered to ensure parallel flows

[66]. Moreover, high pressure is required for liquid delivery in long and narrow

microchannels, and this in turn requires very tight sealing between the

microfluidic channel plate and the substrate. For example, a steel clamp was

used to hold and tighten a microfluidic microarray assembly using syringe

pumping [90].

32

Figure 1-15 (a) Photograph of the microfluidic chip containing shuttle-flow channels, microvalves and micropumps. The entire chip consists of three layers, the top and middle layer were made of PDMS containing microvalves, micropumps, and shuttle flow channels; the bottom layer was a glass substrate with gel pads on which DNA probes were immobilized. The shuttle flow hybridization was realized by controlling the gas ports 1, 2 and 3 automatically. (b) Hybridization specificity assay using four serotypes of the Dengue virus under shuttle flow conditions (frequency 2 Hz). Channels 1 to 4 indicated target DNAs (200 nM) of serotypes I to IV, with repetitions in channels 5 to 8, respectively. The duration of hybridization process was 90 s and washing time was 30 s. Reproduced from [64] by permission of The Royal Society of Chemistry.

Electroosmotic flow (EOF) is another microflow method that has been

used for parallel pumping of multiple channels [136, 137], albeit with multiple

electrode connections. The EOF flow control method depends not only on the

applied voltage across the microchannel, but also on the surface properties of

the microchannel as well as the low ionic strength of the buffered solutions [138].

The high salt concentration typically found in DNA hybridization buffers [19, 139,

140] may result in excessive Joule heating and electrolysis [85, 141-143]. These

effects will lead to drastic changes of the solution temperature and pH [144]

causing instability of hybridization and SNP discrimination performance.

Therefore, only a few reports were published applying EOF flow to microfluidic

DNA microarray analysis [85].

(a) (b)

33

An alternative liquid pumping method is to utilize the body force of the

liquid column itself, and such a force can be created under a centrifugal force

field. As compared to other technologies, centrifugal pumping is easy to

implement and is not sensitive to the physiochemical properties of the liquid. It

can move fluids in a parallel manner in multiple channels of a wide range of

sizes. Moreover, the compact disk (CD) and related industries including disk

materials, disk fabrication, signal reading and error correction, as well as rotor

driven/control systems have been well-developed over the past decades. With

the MEMS technology, the centrifugal microfluidic platform could be combined

with microarray technology to develop portable and point-of-care analysis

devices [145-148].

Centrifugal microfluidic platforms have been developed in many

applications including nucleic acid analysis, immunoassay in biomedical

diagnosis, cell lysis, separation and extraction, and environmental assays [39,

60, 106-108, 149-168]. Flow dynamics studies as well as microfluidic operation of

the liquid in rotating radial microchannels were also investigated [136, 169-178].

More works regarding the application of centrifugal pumping to microfluidics can

be found in several reviews [147, 148, 179, 180]. Reports on the microfluidic

microarray DNA hybridization achieved using the centrifugal pumping method are

listed in Table 1-2. The table summarizes the sample properties (sample types,

sample number, sample volume and concentrations), chip properties (chip

materials and chip bonding methods), and hybridization conditions (time and

temperature) of various experiments including those developed in Dr. Li’s group.

34

Table 1-2 Microfluidic DNA microarray hybridization conducted using centrifugal pumping.

Ref. Number of probe features

Sample volume

(μL)

Sample concentr

ation Sample type

Number of

samples

Hybridization time and

temperature

Chip materials

Substrate materials

Chip bonding Methods and applications

[60] 22000 60 N/A cDNA 1 17 h at 65°C Glass Glass Rubber and clamp Planetary centrifugal flow in the chamber.

[107] 6 3 0.1-100 nM oligonucleotides 1 3min PDMS glass Reversible

bonding Intergraded washing and block buffer chambers.

[106] 150 2 10nM 368-bp PCR amplicons 1 5 min at RT PDMS glass Reversible

bonding Integrated washing and block buffer chambers, differentiated 4 different Staphylococcus species.

[108] 32 0.35 25nM to 200 nM oligonucleotides 12 90 to 300s at

RT PDMS glass Reversible bonding

Reciprocating flow induced from rotation-pause operation of the CD device enhanced hybridization.

[129] ǂ 100 1 20nM oligonucleotides or PCR amplicons 100 10 min at

45°C PDMS glass Reversible bonding

Hybridization using intersection approach, detection of pathogen PCR amplicons. Up to 100 samples could be analyzed in parallel.

[65] ǂ 384 1 0.1nM for oligos

oligonucleotides or PCR amplicons 384 1 min at RT PDMS glass Reversible

bonding Hybridization using intersection approach, 384 × 384 hybridization assays have been performed.

[134] ǂ 100 1 N/A oligonucleotides or PCR amplicons 100 3 min at RT PDMS glass Reversible

bonding

3-min hybridization using intersection approach, discrimination of two pathogen PCR amplicons. Up to 100 samples could be analyzed in parallel.

[133] ǂ 96 1 0.5 nM oligonucleotides 96 3 min at RT PDMS glass Reversible bonding

Intersection approach hybridization in 3 min. Up to 96 samples could be analyzed in parallel.

* N/A: Data not available; RT: room temperature. ǂ: The work was developed in Dr. Li’s group, in which ref. [129] and ref. [134] are done by me as given in Chapter 5 and Chapter 6, respectively.

35

Although centrifugal pumping has been successfully exploited in many

microfluidic assays, its application to DNA microarray analysis is still rare [60,

106-108]. Bynum and Gordon used several rotors on a centrifugal platform to

induce reciprocal flows inside the DNA hybridization chambers [60]. After 17 h of

reaction, the measured signals were 10 times higher than those obtained from

the conventional static method [60]. Madou and co-workers developed a

chamber hybridization system, where the assembly of a glass slide and a PDMS

microfluidic chamber was placed on a CD support that could hold up to 5 slides

[106, 107]. Here, low-density spotted probe arrays were used and the

hybridization reagents were positioned to be pumped sequentially. The authors

demonstrated the discrimination of 4 clinically relevant Staphylococcus species

by the automated hybridization process in 15 min performed at room temperature

[106]. Li et al. presented a CD-like microfluidic device capable of generating a

reciprocating flow of DNA samples within the microchannels and demonstrated

its application in rapid DNA hybridization assays [108]. Here, centrifugal force

was used to drive the sample solution to flow through the hybridization channel

into a temporary collection reservoir while capillary force pulled the solution back

into the hybridization channel during the stopping period (Figure 1-16). The

sample hybridization time was reduced to 90s and the sample volume was as

low as 350 nL [108]. A theoretical model of this reciprocal flow by the interplay of

capillary and centrifugal forces has also been proposed [181].

36

Figure 1-16 (a) Schematic representation of a CD microfluidic device sealed with a glass substrate for DNA hybridization. It consists of a PDMS CD slab containing twelve DNA hybridization assay units with immobilized DNA probe arrays. (b) Schematic diagram of a single DNA hybridization assay unit. (c) Hybridization specificity tests with the CD microfluidic device. Top: Dengue virus serotype 1 targets. Bottom: Dengue virus serotype 2 targets. The CD device was rotated at 22 Hz for 3 s and then stopped for 3 s during the reciprocating process, with a duration of 90 s. Reprinted with permission from [108], Copyright © 2009 Elsevier Ltd.

The limited applications of centrifugal pumping for high throughput nucleic

acid microarray analysis could be attributed to the use of the radial-direction

structure design. Currently, almost all of the fluidic patterns in centrifugal

microfluidic platforms are fabricated in the radial orientation. Due to the

mechanical strength of materials and the demand of miniaturization, the length of

the microfluidic structure in the radial direction is limited. For example, if a

centrifugal platform is built on a 120-mm regular CD with a 15-mm centre spindle

hole, the length of a microfluidic structure cannot be more than 53 mm.

Therefore, the applicable space is very limited and the capillary effect may

(c)

(a) (b)

37

dominate the liquid flow process in microchannels. In addition, it is well known

that the centrifugal force increases with the increase in radius, and the average

velocity of the liquid is proportional to the radial extent of the fluid in a

microchannel [147, 156]. Such an increasing flow velocity in the microchannel

demonstrates the challenges and difficulties in the design of fluidic structures

[136, 169, 172]. Furthermore, the radial-like microfluidic structure cannot be

easily adapted to the intersection approach because centrifugal pumping

appears to be only exploited in one direction. Therefore, spotted probe arrays

have to be used for microchannel hybridization, which limits sample throughput,

and a lot of disk space is wasted.

Recently, our group developed a new microfluidic DNA microarray

method, in which the centrifugal advantages were exploited twice, first in the

radial microchannels, and then in the spiral channels [133]. The microarray

assembly consists of two CD-like PDMS chips as microchannel plates and one

92-mm glass wafer as substrate [129, 133, 134]. In the first step, the radial

channel plate is assembled with the glass wafer for printing the radial probe line

arrays; during the second step, the spiral channel plate is sealed against the

wafer with the printed probe lines after removing the radial channel plate. In both

steps, liquid flows in the channels are driven by centrifugal pumping obtained

from spinning the assembly. Because the spiral channels are nearly orthogonal

to the radial direction, sample flows thus intersect with the probe lines printed on

the disk and result in the formation of hybridization patches after complementary

target molecules are applied. The hybridization patches are in a rectangular

38

shape and are easy to be analysed. More details on the microfluidic DNA

microarray using centrifugal-pumping technology can be found in Chapters 5-7.

1.3 Research outlines

The proposed work is mainly focused on developing microfluidic

microarray methods for fast, low-volume, and high throughput nucleic acid

detection and discrimination.

Microfluidic microarray hybridization based on the intersection method

was first studied on glass slides using pressure-driven microflows. Optimal

conditions of probe printing as well as sample hybridization were examined for

the detection of both oligonucleotide and PCR product samples. A novel method

based on the interaction between gold nanoparticles and DNA molecules was

constructed for the discrimination between DNA targets with single base-pair

difference. Moreover, to get a better understanding of the mass transport and

surface hybridization in microfluidic systems, mathematical models were

constructed to predict the microfluidic hybridization behaviour with numerical

simulations.

Another research objective was to realize fast and high throughput DNA

analysis using centrifugal pumping technology. Compared with the stop-flow

method on the slide format, the hybridization time was shorten to <3min and up

to 100 samples can be analysed simultaneously. The microflow driven by

centrifugal pumping was further studied with fluid dynamics. The study describes

39

how device geometry, i.e. channel height and width, as well as rotation speed

influence liquid transport phenomena.

1.4 Dissertation structure

The main content of this dissertation is organised as per the structure

shown in Figure 1-17. Specifically:

In Chapter 1, an overview on the recent advances in microfluidic DNA

microarray hybridization is given. Part of the work has been

published in Analytica Chimica Acta (L. Wang and P.C.H. Li,

“Microfluidic DNA Microarray Analysis: A Review”, Anal. Chim.

Acta, 2011, 687, 12–27).

In Chapter 2, methods and technologies involved in the DNA microchip

fabrication and analysis are illustrated. The related instrumentations

are also introduced.

In Chapter 3, microfluidic probe spotting and DNA target hybridization

were conducted on glass slides using pressure-driven flow. The

intersection approach was used in the microfluidic hybridization. Up

to 16 samples from fungal DNA were detected. The results have

been published in the Journal of Agricultural and Food Chemistry

(L. Wang, and P.C.H. Li, “Flexible microarray construction and fast

DNA hybridization conducted on a microfluidic chip for greenhouse

plant fungal pathogen detection”, J. Agri. Food Chem., 2007, 55,

10509-10516).

40

In Chapter 4, room temperature discrimination of two fungal DNA

fragments with one base-pair difference was achieved with the

assistance of gold nanoparticles. The work has been published on

Biomicrofluidics (L. Wang, and P.C.H. Li, “Gold nanoparticle-

assisted single base-pair mismatch discrimination on a microfluidic

microarray device”, Biomicrofluidics, 2010, 4, 032209, page 1-9).

In Chapter 5, a circular glass disk was used as a microarray substrate

to employ centrifugal-pumping technology. Two polymer channel

plate, in radial and spiral direction, respectively, were designed for

microchip hybridization through the intersection approach. Up to

100 oligonucleotide samples can be detected in 3 min. The work

has been published in Analytica Chimica Acta (L. Wang, P.C.H. Li,

H.-Z. Yu, and A.M. Parameswaran. “Fungal pathogenic nucleic acid

detection achieved with a microfluidic microarray device”, Anal.

Chim. Acta, 2008, 610, 97–104).

In Chapter 6, with an improved design in channel geometry and target

labels, high-throughput discrimination of fungal DNA samples was

achieved on disk platform. The work has been published in

Analytical Biochemistry (L. Wang, and P.C.H. Li, “Optimization of a

microfluidic microarray device for the fast discrimination of fungal

pathogenic DNA”, Anal. Biochem., 2010, 400, 282-288).

In Chapter 7, experimental measurement and mathematical modelling

of centrifugal-pumping flow in spiral microchannels are described.

41

Section 7.3 was written based on the work of Dr. Kropinski from the

Department of Mathematics with the aid of my experimental data.

Part of the chapter has been published in Lab on a Chip (L. Wang,

M.-C. Kropinski, and P. C. H. Li, “Analysis and modeling of flow in

rotating spiral microchannels: towards math-aided design of

microfluidic systems using centrifugal pumping”, Lab Chip, 2011,

11, 2097–2108).

In Chapter 8, theoretical modelling and simulation of microfluidic DNA

microarray hybridization were studied. The manuscript is in

preparation (L. Wang, and P.C.H. Li, “Modeling of microfluidic DNA

microarray hybridization and its application to microchip design and

optimization”).

In Chapter 9, different aspects of this research work are summarized,

and an outlook on future directions is discussed as well.

In Chapter 10, detailed procedures on the use of COMSOL program

are listed. The procedures have been applied to the simulation of

the microfluidic DNA microarray hybridization kinetics as depicted

in Chapter 8.

42

Figure 1-17 The organization structure of this dissertation.

Applications to DNA analysis (4 chapters)

Simple slide format and pressure-driven flow

Circular disk format with centrifugal-pumping flow

Chapter 3: Concept of intersection

approach and optimization

Chapter 4: Further

improvement using

nanoparticles

Chapter 5: Application to

both oligos and PCR amplicons

analysis

Chapter 6: Improvement

and application to the fast SNP discrimination

Microfluidic Microarray DNA hybridization

Chapter 7: Modeling and simulation of centrifugal-

pumping microflow

Chapter 8: Modeling and simulation of

microfluidic DNA hybridization

Modeling and simulation of flow and hybridization (2 chapters)

For faster and higher throughput

assay

43

2: MICROCHIP FABRICATION AND DETECTION INSTRUMENTATION

2.1 Introduction

Microfluidics is a rapidly developing technology evolved from the well-

established microelectronics industry. The fabrication of microfluidic devices is

usually based on bonding two chip plates together where microfluidic structures

are created on one or both of the chip plates. Because of the great success in

semiconductor-related technologies, silicon and glass materials were first used

as substrates in making the microstructures by photolithography [138, 182, 183].

These materials such as glass are easily obtained and some of the properties,

such as high chemical inertness and/or excellent optical transparency, have been

very attractive in many applications [184, 185]. However, microstructure etching,

hole drilling, and chip bonding processes with these materials are costly and time

consuming [55, 138]. In addition, if microfluidic structures are used for patterning

surface-immobilized molecules on support chips, reversible bonding is preferable

for later applications of these chips with patterned molecules. Microfluidic devices

made in plastics or polymers offer unique advantages in addition to functionality

such as bio-compatibility, disposability and ease of mass production at low cost

[184, 186, 187]. A wide range of materials such as polycarbonates (PC),

poly(ethylene terephthalate) (PET), polypropylene (PP), polystyrene (PS),

44

poly(methyl methacrylate) (PMMA), and polydimethylsiloxane (PDMS) have been

structured with microfluidic channels. During the last decades, numerous

researches have been done with polymer microfluidic chips and some works

related to DNA microarray hybridization are listed in Table 1-1.

The fabrication of microstructures on polymer substrates can be achieved

using various methods including laser ablation, wire imprinting, hot embossing,

injection molding, and casting. Using the laser ablation technology,

microstructures can be imprinted directly onto the polymer substrate by moving a

focused laser beam across the surface and fluidic prototypes can be rapidly

produced with high accuracy and repeatability [63, 83, 90, 188]. However, the

requirement for a high-power laser instrumentation has limited the applications of

this process [189]. The wire imprinting method is a low cost method used to

create microfluidic channels on plastics but it suffers from low accuracy and the

inability to create complicated patterns [190]. Hot embossing and injection

molding have been widely used in structuring thermoplastic materials [72, 86, 89,

191, 192]. These methods involve the use of molding masters that contain the

desired microstructures to stamp (reverse-duplicate) patterns onto the substrate,

which are easy to be applied to the mass production of microchips.

Casting is a micro-replication method using silicone rubber to form

microfluidic structures. This technology, which is also named as soft lithography,

was developed by Whitesides and co-workers and has received much interest in

academic research due to its simplicity and low cost [193-195]. PDMS is the

most widely used silicone-based elastomer in casting microfluidic structures [70,

45

195, 196]. It is optically clear, non-toxic, and, in general, is considered chemically

inert. During soft lithography processes, a mixture of the elastomer precursor and

its curing agent is poured over a molding master. After curing at room

temperature or higher, the soft elastomer copy or replica is peeled off the

molding master. By simply sealing the PDMS plate with microstructures against a

planar substrate, a microfluidic network is formed without the use of thermal

bonding or high pressure. This PDMS casting technology was the method used

for fabricating microfluidic devices throughout my experiments.

2.2 PDMS-based Microfabrication Processes

The microfluidic devices in this thesis are assemblies from reversibly

bonding a flat glass chip to a PDMS plate with microstructures. The PDMS

channel plate is a replica from casting PDMS prepolymer solutions to a SU-8

coated Si master or an etched Si master. The fabrication of the master with

positive relief microstructures is based on the popular photolithography

techniques where a series of processes are involved (Figure 2-1).

At first, the surface of silicon wafers are cleaned thoroughly and a layer of

negative photoresist is deposited on the surface. The thickness of the photoresist

film determines the channel depth of the moulded chip. Upon UV exposure,

geometric shapes printed on a photomask are transferred to the photoresist

layer. The following development and hard-baking procedures leave polymerized

photoresist with the microchannel patterns on the silicon wafer. Figure 2-1

depicts the procedures of making a PDMS-glass microfluidic device with a

negative photoresist (e.g. SU-8). More details are illustrated in the following

46

sections. Furthermore, because dust could contaminate surface and reduce

product quality in most of the procedures, the concentration of airborne particles

has to be controlled to a low level using clean room facilities (Clean Air Products,

577 Series, Minneapolis, MN) as shown in Figure 2-2.

Figure 2-1 Sequence for fabrication of the microfluidic assembly using negative photoresist SU-8. (a) a plain silicon wafer (b) Si wafer coated with photoresist film; (c) wafer exposed to UV light through a photomask; (d) photoresist developed; (e) PDMS prepolymer solution casted on the SU-8 master; (f) PDMS channel plate peeled off; (g) PDMS channel plate bound to a glass chip to produce assembly.

Silicon substrate

Negative photoresist

PDMS channel plate

Photomask

UV light

PDMS-glass chip assembly

Layer deposition by spin coating

UV exposure

PDMS casting

Bonding with glass chip

Developing

(a)

(b)

(d)

(c)

(e)

(f)

(g)

Glass chip

Weir structure

Microchannels

47

Figure 2-2 Photograph of the softwall clean room facilities used for making microchip and a bionocular microscope for surface examination.

2.2.1 Surface preparation and layer deposition processes

The fabrication of microfluidic structures starts with the deposition of a

photoresist film on a silicon wafer. The wafer is a thin, round slice of silicon

crystal that has a diameter of 100 mm or more and it has a very flat surface.

Under exposure to oxygen, the silicon surface oxidizes to form silicon dioxide

Air filtration system

Reflective microscope

Nitrogen gas cylinder

Spin coater

UV exposure system

Vacuum pump

Hot plate

48

(SiO2). The adhesion of photoresist to the wafer surface is critical to the faithful

replication of the micropatterns [197]. In our practice, dust particles and films of

grease were washed away with acetone and isopropanol. Because water has a

higher affinity to SiO2 than organic photoresist, contamination from humidity

absorption is another major concern and the silicon wafers were further baked at

>100°C to get rid of surface moisture, which would otherwise weaken the

subsequent photoresist adhesion.

Once the wafer is prepared, a thin-layer of photoresist is applied. Spin-on

coating is the most common deposition techniques for producing a photoresist

layer on the surface of a Si wafer. Figure 2-3 illustrates the spin-coating

processes as well as the spin coater used in my experiments (WS-400, Laurell

Technologies Corporation, North Wales, PA). A controlled volume of photoresist

is dispensed onto a wafer and the wafer is then spun to produce a uniform

photoresist film. In our work, SU8 (SU8-50, MicroChem Corporation, Newton,

MA), an epoxy resin, has been used as the negative photoresist [198-200].

Because the SU-8 film will act as both a photoresist for pattern transferring and a

master layer for later PDMS casting, the thickness of the film has to be set to

match the final microchannel height. The control on film thickness depends on

the properties of the photoresist as well as rotation speeds as shown in Figure

2-3(f). For example, to achieve a layer of SU8-50 with 75-µm thickness, the spin

coater has to be rotated at a speed of 1600 rpm for 30 s. After the spin coating

process, the wafer with photoresist layer is baked at a temperature (95 °C) for

few minutes to remove most of solvent from the resin for later UV exposure.

49

Figure 2-3 (a)-(d) Illustration of the spin coating processes for the creation of the SU8 photoresist layer on a silicon wafer. (e) a spin coater with controller (f) rotation speed versus the thickness of the SU8 resin layer. The figure is adapted from MicroChem® product datasheet. SU8-50 and SU8-100 are different from each other in viscosity for different applications. In this thesis, SU8-50 has been used throughout.

2.2.2 Photolithography processes

Photolithography is the process of transferring geometric shapes from a

photomask to the photoresist layer on the surface of a silicon wafer. The process

involves, pattern drawing, pattern transfer by UV exposure, development; and

hard-baking.

As shown in Figure 2-1, micropatterns are transferred to the photoresist

layer on the silicon wafer by UV irradiation. The exposure is performed on a

(a) Dispense a controlled amount of photoresist

(c) Rapidly ramp - up the spin speed throwing off excess photoresist

(b) Allow the photoresist to spread across the wafer

(d) Spin at a higher speed (e.g. 2000rpm) to form a thin film of photoresist

Spin speed (×1000 rpm)

Film

thic

knes

s µm

)

1000 1500 2000 2500 3000

50

100

150

200

(e) (f)

50

standard UV aligner (Model LS-150-3, Bachur & Associates, San Jose, CA)

where a dose of 300–400 mJ·cm−2 measured at 365 nm is necessary for the

cross-linking of the photoresist (Figure 2-4 (a)). The photomask is a transparent

film with the micropatterns printed on it at high resolution. Because a negative

photoresist, SU8, has been used, the photomask contains an inverse copy of the

pattern which is to be created on the wafer (Figure 2-1(c)). Upon the exposure to

the UV light, the SU8 solution becomes polymerized, and more difficult to

dissolve than the unexposed SU8 in the later developing process. Then, a post-

exposure bake at 95ºC for 15min is required to accelerate the cross linking of the

epoxy resin to form a rigid structure with a high cross-linking density [198]. The

unexposed material is then removed with an organic solvent (SU8 Developer,

MicroChem Corp., Newton, MA) in the later developing process (Figure 2-4 (b)).

Figure 2-4 (a) Silicon wafer aligned with photomask under UV irradiation; (b) SU8 layer with microstructures on a 5-inch silicon wafer after the developing process.

Considering that the photomask is used to generate the microstructures, it

is obvious that its print quality determines the final microstructure quality on the

photoresist layer. Figure 2-5 compares the PDMS channel plate made from two

(a) (b)

51

photomasks. Two resolutions, 3368 dpi and 20000 dpi, respectively, were used

in printing the photomasks. As shown in the figure, lower resolution (3368 dpi)

printing results in serrated-shape channel sidewall (see Figure 2-5(a) and (b)),

while the higher resolution printing gives a much smoother microchannel

sidewalls (see Figure 2-5(c) and (d)).

Figure 2-5 Comparison of the quality of PDMS microchannels fabricated from two photomasks of different printing resolutions. (a) Photomask printed with 3368 dpi. (b) The microscopic image of the PDMS channel plate made from (a). (c) Photomask printed with 20,000 dpi. (d) The microscopic image of the PDMS channel plate made from (c).

2.2.3 Casting and bonding processes

Throughout this work, PDMS prepolymer (Sylgard 184, Dow Corning) was

used throughout for casting the microchannel plate. The use of PDMS offers

several benefits due to its characteristics. It is optically transparent down to 300

100µm

(a)

(c)

(b)

(d)

100µm

100µm

100µm

52

nm, chemically inert, and nontoxic after being cured and has highly thermal

stability (up to 300 °C) [201]. During the casting process, a solid weir is first built

around the edge of the Si wafer with the SU8 master layer (Figure 2-6 (a)). The

PDMS prepolymer solution is mixed with its curing agent in a particular ratio,

such as 10:1. Then the mixture is poured into the well to obtain the replica of the

pattern. The volume of the solution is calculated to obtain a polymer thickness

around 1~2 mm. After curing at 60°C for 4h or at room temperature for 24h, the

PDMS replica is peeled off from the master. To facilitate the peeling off process,

the Si master is treated with a small amount of releasing agents, such as

dimethyldichlorosilane (PlusOne™ Repel-Silane ES, GE Healthcare,

Buckinghamshire, UK).

During the bonding process, the cured PDMS plate could be overlaid

directly onto a glass substrate at room temperature for reversible bonding (Figure

2-6 (b)). Microchannels thus formed from the PDMS grooves and the glass

surface (Figure 2-1(g)). Solution reservoirs (1 mm in diameter) at both ends of

channels were made by punching holes on the PDMS plate with a flat-end

syringe needle. Although plasma oxidation could render the PDMS channels

hydrophilic property for easy filling of aqueous solutions in subsequent

experiments [202], the method was not used in our work because the sealing

between glass surface and the plasma-oxidized PDMS plate is non-reversible

[203, 204].

53

Figure 2-6 (a) SU8 master with weir structure around the wafer for PDMS molding. (b) Assembly of the PDMS microchannel plate with a glass disk.

2.2.4 Comparison of molding master made from SU8 or wet etching

In addition to the SU8 molding master depicted in the previous sections, Si

wafer itself could be fabricated into a master through wet chemical etching. With

the help of a positive photoresist, channel patterns are transferred onto the pre-

coated mask layer of the Si master. Upon using chemical etchant solutions, such

as HF, positive relief structures are created on the silicon wafer. The procedures

are shown in Figure 2-7.

(a) (b)

54

Figure 2-7 Sequence for fabrication of the microfluidic assembly using a positive photoresist and subsequent HF etching. (a) A clean Si or glass wafer. (b) Cr and Au masked Si wafer (c) The wafer in (b) is coated with positive photoresist. (d) The wafer in (c) is exposed to UV light through a photomask; (e) photoresist is removed; (f) The exposed metal mask is etched; (g) The exposed Si or glass etched; (h) The photoresist and metal is stripped; (i) PDMS prepolymer solution casted on the SU8 master; (j) The cured PDMS channel plate is peeled off; (k) PDMS channel plate is sealed to a glass chip to produce assembly.

The advantage of the Si master is its high mechanical strength and

durability. It has been reported for ~100 casting procedures without significant

degradation [205], while a SU8 molding master can only allow up to 30 PDMS

replicas to be made [193]. Nevertheless, SU8 is a popular structural material for

lab-on-chip applications [206]. When compared with the silicon or glass master,

PDMS-glass chip assembly

Silicon substrate

PDMS channel plate

PDMS casting

Positive photoresist coating

Metal layer sputtering

Photomask

UV light

UV exposure

Photoresist developing

Metal mask etching

Wafer etching

Removal of metal and photoresist

Bonding with glass chip

Cr & Au

photoresist

(a)

(b)

(c)

(d)

(e)

(f)

(i)

(g)

(h)

(j)

(k)

55

SU8 master shows superiority due to many advantages that are outlined as

follows.

First, the fabrication procedures of SU8 master are simpler and less

tedious. It takes only three steps in total to fabricate a SU8 master from a Si

wafer (Figure 2-1). On the contrary, the number of steps is doubled in making a

silicon master. Furthermore, because no metal layer sputtering instrument or

highly toxic HF solution is involved in creating the SU8 master, an in-lab modular

cleanroom could be used for rapid prototyping of micropatterns (Figure 2-2).

Furthermore, the use of SU8 resin layer gives better channel shape and

the PDMS replica shows better sealing with glass surface. The microchannel

structures on the silicon master are created from wet etching process. As shown

in Figure 2-8 (b), the growth direction of the microstructure is not completely

vertical. The lateral etching of the channel structure can be a significant

disadvantage if a microchannel with high aspect ratio is desired. Therefore, the

channel width is larger than the value on the photomask and the spacing

between channels has to be carefully calculated. Meanwhile, the SU8 molding

master has been known for making microstructures with high aspect ratios and

channel walls are almost vertical to the chip plate (Figure 2-8 (d)) [200, 207].

Furthermore, the bottom of the microstructure in the silicon master is near round

making the corresponding top part of the microchannel in the PDMS plate near

round (Figure 2-8 (b)). This feature affects the integrity of the sealing of the

PDMS channel plate to the glass surface. On the other hand, with proper

development, the bottom of the microstructure on the SU8 master can be made

56

flat (Figure 2-8 (d)) and the resulting PDMS channel plate will seal tightly with the

glass surface. No side leakage across spiral channels has been observed even

when the disk assembly was spun at 3600 rpm for 10 min.

Figure 2-8 Comparison of the Si masters and microchannels fabricated using positive and negative photoresist. (a) The reflective microscope image of the Si master fabricated from positive photoresist followed with KOH etching. (b) Surface profile of the microchannels in (a); (c) The reflective microscope image of the Si master fabricated from SU8 photoresist. (d) Surface profile of the microchannels in (c).

2.3 Surface modification to generate aldehyde-functionalized glass chips.

The glass substrates were chemically modified to produce aldehyde-

functionalized surfaces using an established procedure (See Figure 2-9) [208].

Briefly, plain glass slides were cleaned with 10% NaOH solution for 10 minutes at

(a)

(c)

(b)

(d)

57

~80°C. After being rinsed with distilled water, the slides were treated with a

piranha solution (70:30 v/v, sulphuric acid to 30% hydrogen peroxide) for 1h at

~80°C. The slides were then rinsed with water and dried under a stream of

nitrogen.

Figure 2-9 Surface modification to generate aldehyde-functionalized glass surface. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

The cleaned slides were treated with a mixture of ethanol: H2O:

aminopropyltriethoxyl silane (95:3:2 by volume) for 2h under stirring; rinsed with

95% ethanol and deionized H2O; dried under nitrogen and baked at ~120°C for

1h. The aminated glass slides were then immersed in 5% glutaraldehyde in a 10x

PBS (phosphate buffered saline) solution overnight and washed with acetone

and deionized H2O. After being dried in a nitrogen gas stream, the aldehyde-

modified glass slides were stored in a dark place at 4°C before probe printing.

OHOHOH

EtO

SiEtO

HO

NH2

OSiO

ON O

H

OSiO

ONH2

O O

H H

Clean glass surface

- H2O

- 3 EtOH

APTES

Glutaraldehyde

58

2.4 Immobilization of probe DNA.

The PDMS channel plate was sealed against the aldehyde glass slide to

form microchannels as shown in Figure 2-1(g). Then, 0.8 µL of probe DNA

prepared in the spotting solution (1.0M NaCl + 0.15M NaHCO3) was added into

the inlet reservoirs using a micropipette. The probe solution was filled through the

channels by applying vacuum pumping at the outlets or by centrifugal pumping.

With incubation at room temperature for 30 min, covalent Schiff linkage was

formed between the amine ends of the probe oligonucleotides and the aldehyde

groups on the glass surface (Figure 2-10) [209].

Figure 2-10 Covalent attachment of aminated probe DNA to aldehyde glass surface. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

After washing the microchannels with 1 µL of washing solution (0.15%

Triton-X 100, 1.0M NaCl and 0.15M NaHCO3), the PDMS channel plate was then

peeled off and the glass slide was chemically reduced with a NaBH4 solution

(100 mg of NaBH4 was dissolved in 30 mL 1x PBS and 10 mL 95% ethanol) for

15 min to reduce the Schiff linkage to the more stable C-N single bond (Figure

+ Probe AB

+ NaBH4

Aldehyde modified glass surface Covalent Schiff linkage between probe DNA and glass surface

Covalent C-N linkage between probe ssDNA and glass surface

3

OSiO

ON O

H

3

OSiO

ON N

33

CGCCAGAGATACCAAAACTC

6

OSiO

O

HN

HN

3

CGCCAGAGATACCAAAACTC

65

For further

hybridization

59

2-10). The glass chip was then rinsed with deionized water for 2 min and dried by

nitrogen gas.

2.5 Quantification of fluorescent image and data analysis

Following the hybridization and washing procedures, the glass substrate

with hybridization patches was scanned on a confocal laser fluorescent scanner

(Typhoon 9410, Amersham Biosystems, now GE healthcare) as shown in Figure

2-11. The excitation wavelength was 488 nm (for fluorescein-labelled targets) or

633nm (for Cy5-labelled targets) and the PMT (Photomultiplier tube) voltage was

set to 600V. The scanned image was collected at a resolution of 10 or 25 µm and

was analysed by the ImageQuant 5.2 program. In the data quantification

procedure, lines or boxes were manually drawn to cover the rectangular

hybridization patches on the image, and the average fluorescent intensity of the

patch was recorded in the relative fluorescent unit (RFU).

Figure 2-11 Typhoon 9410 confocal laser fluorescent scanner.

60

3: FLEXIBLE MICROARRAY CONSTRUCTION AND FAST DNA HYBRIDIZATION CONDUCTED ON A MICROFLUIDIC CHIP FOR GREENHOUSE PLANT FUNGAL PATHOGEN DETECTION

3.1 Introduction

Plant diseases from fungal, bacterial and viral organisms have caused

serious economic losses in greenhouse vegetable industry annually [210, 211].

Effective disease control requires rapid identification of disease microorganisms.

Currently, DNA microarray chips have been widely used in various applications,

such as expression profiling, genotyping and species characterization [212]. For

example, Xu et al. and Bordoni et al. have identified genetically modified soybean

and maize using oligonucleotide microarrays [213, 214]. Warsen et al. and

Ronning et al. discriminated between fish pathogens or closely related crops,

respectively, with microarray technology [215, 216]. The conventional DNA

microarrays are generally constructed either by on-chip synthesis of

oligonucleotide probes or by spotting of pre-synthesized probes on activated

substrates [111, 212]. For assays, samples containing the labelled target were

applied by manually spreading 10-50 µL of solutions on the microarray area for

hybridization, and the process usually requires long-time incubation up to 16

hours [19, 105].

61

Since the microfluidic method is capable of reducing the sample volume

and of accelerating diffusion and reaction kinetics, DNA hybridization has been

conducted in microfluidic channels [89, 90, 106, 217]. Moreover, probe

immobilization can also be achieved using microchannels. The resulting probe

line arrays can be used in subsequent microfluidic hybridization analysis with an

intersection method. For instance, Liu et al. designed a glass chip containing four

line arrays of oligonucleotide probes, and parallel hybridizations of human

genomic DNA targets were completed by using a second chip consisting of 4

sample microfluidic channels [218]. Similarly, Lee et al. created a 3×3 array on a

gold substrate and conducted microfluidic hybridizations to detect RNA

fragments derived from a transgenic plant [96]. Benn et al. studied the mass

transfer efficiency and hybridization kinetics inside microfluidic channels using

60-mer oligonucleotide samples on an 8 × 8 array [91]. Situma et al. has

achieved the detection of two different low-abundant DNA point mutations in

KRAS2 oncogenes using the intersection method with 16-channel microfluidic

chips [93]. In these pathogen detection or mutation studies, a small set of probes

is sufficient, and so a low-density DNA microarray can be constructed, as

opposed to the high-density microarrays needed for large-scale gene expression

profiling. Moreover, parallel sample analysis was achieved with microfluidic

hybridization as well as the intersection method.

To date, the application of the microfluidic microarray method to

agricultural problems has been limited. Here, we dubbed the method as

microfluidic microarray assembly (MMA) and a 16×16 DNA array was designed

62

to aim at fast assays. In this work, we use the MMA technique successfully to

identify three PCR products prepared from three fungal pathogens: Didymella

bryoniae, which causes gummy stem blight of greenhouse cucumbers; and,

Botrytis cinerea and Botrytis squamosa, which cause downy mildew and are

recognised as pathogens of overwintered salad crops [105]. The two Botrytis

amplicons are different from each other with only one base-pair difference in the

centre. We employed two 21-mer oligonucleotide probes and managed to

distinguish the PCR products with the microfluidic microarray method. It has

been demonstrated that 1.4 ng PCR products (~260bp, 1.4ng/µL, 1µL) were

detected at 50°C and one-base-pair discrimination was achieved in five minutes.

The results obtained from the printed glass slide in this work showed an

improvement over our previous work [105], in which targets were detected only

when the probes were spotted on an agarose-coated glass slide and were

hybridized overnight.

Here, the MMA method consists of 2 steps of assembly process, see

Figure 3-1. In the first step, channel plate 1 is assembled with the glass chip via

reversible bonding. Aminated DNA probes are introduced into the microchannels

and are immobilized on the glass chip. A line microarray of probes is thus

created. After peeling off plate 1, channel plate 2 is then assembled with the

same glass chip. The sample solution that flows through the microchannels will

intersect the probe line arrays and hybridization is accomplished in a few

minutes. Both the creation of probe microarray and hybridization process in

microfluidic channels are capable of reducing the sample volume (< 1 µL) and of

63

preventing from evaporation and cross-contamination. In this work, it was found

that the MMA method not only provides a flexible probe creation method, but also

enhances the detection sensitivity and achieves differentiation of various

greenhouse pathogens.

Figure 3-1 The microfluidic microarray method using straight microchannels. (a) The creation of a DNA probe line array on an aldehyde-modified glass slide via straight microchannels. (b) The hybridization of DNA samples in straight channels orthogonal to the straight probe lines printed on the glass slide. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

3.2 Experimental

3.2.1 Materials.

All chemicals and solvents were purchased from BDH Tech Inc (Toronto,

Canada), unless stated otherwise, and used without further purification. 3-

Aminopropyltriethoxyl silane (APTES), 50% glutaraldehyde, sodium dodecyl

sulphate (SDS) and Triton X-100 were obtained from Sigma-Aldrich. Ultra-pure

water (18 MΩ·cm-1) was obtained from an Easypure RF purification system

(Dubuque, IA). Negative photoresist (SU8-50) and its developer were purchased

from MicroChem Corporation. An elastomer base, Sylgard® 184 silicone and its

1) Sealed against an aldehyde glass chip

2) Flow of aminated DNA probe solution

Channel plate 1 horizontal channels

1) Peel off channel plate 1.

Line arrays of DNA probes

Channel plate 2 vertical channels

1) Sealed against the glass chip with line microarray

2) Flow of DNA sample solution

Sample Hybridization

1) Peel off PDMS channel plate 2

Detection results

2) Scan with confocal fluorescence scanner

2) Reduction of Schiff base linkage Probe

immobilization

(a)

(b)

64

curing agent used to make polydimethylsiloxane (PDMS) were obtained from

Dow Corning Corporation (Midland, MI). The plain 3"×2" glass microscope slides

were purchased from Fisher Scientific Company (Ottawa, ON, Canada). All other

chemicals and solvents were purchased from BDH Tech Inc (Toronto, ON,

Canada) and used without further purification.

All oligonucleotides were synthesized and modified by Sigma-Genosys

(Oakville, ON, Canada). The sequences of probe oligonucelotides were designed

to detect two greenhouse plant pathogens, Botrytis cinerea (with probe AB and

ALB) and Didymella bryoniae (with probe AD) [105, 219]. The probes were

modified with an amine group at the 5’-end. In AB and AD molecules, this amine

group was spaced from their DNA sequences with a C6 linker, while a C12 linker

was used in ALB. Dual labelled probe ADF, which has the same DNA sequence

as that of probe AD, has its 3'-end also labelled with fluorescein. The probe was

used as a marker during probe immobilization and sample hybridization.

Oligonucleotide samples, B’F and FD’ which are complementary to the

sequences of probe AB and AD, respectively, were labelled with fluorescein at

the 5’-end. CD’ has the same sequence with FD’ but was labelled with Cy5 dye

at the 5’-end. The sequences of oligonucleotides used and their acronyms were

listed in Table 3-1.

65

Table 3-1 Oligonucleotides and PCR products used in this study

Acronym Length Sequence (5'-3')

AB 21-mer NH2-(CH2)6 –CGC CAG AGA ATA CCA AAA CTC

ALB 21-mer NH2-(CH2)12 –CGC CAG AGA ATA CCA AAA CTC

AD 22-mer NH2-(CH2)6 –CGC CGA TTG GAC AAA ACT TAA A

ADF 22-mer NH2-(CH2)6 –CGC CGA TTG GAC AAA ACT TAA A-Fluorescein

FB’ 21-mer Fluorescein or Cy5 –GAG TTT TGG TAT TCT CTG GCG

FD’ 22-mer Fluorescein or Cy5 –T TTA AGT TTT GTC CAA TCG GCG

CD’ 22-mer Cy5–T TTA AGT TTT GTC CAA TCG GCG

FB’P 264-bp

FBN’P 264-bp

FD’P 259-bp

Genomic DNA samples were extracted from cucumber–dextrose broth of

Botrytis cinerea, Didymella bryoniae and Botrytis squamosa at Agriculture and

Agri-Food Canada [105]. Three PCR products (FB’P, 264 bp; FD’P, 259 bp;

FBN’P, 264bp) were amplified and labelled with fluorescein from the 3 species,

respectively. The concentration of FB’P and FD’P was ~40 ng/μL (~250 nM),

while that of FBN’P was lower at about 6 ng/μl (~40 nM). The central sequences

of the sense strand of FB’P and FD’P are complementary to the sequences of

probe AB and AD, respectively, while FBN’P has one base-pair difference

(TTT:ATA instead of TAT:ATA in the center) from that of FB’P [220].

… –A AAT TCA AAA CAG GTT CGC CGC–…

Fluorescein –…–T TTA AGT TTT GTC CAA TCG GCG–…

Fluorescein –…–GAG TTT TGG TTT TCT CAG GCG–

… –CTC AAA ACC AAA AGA GTC CGC–

Fluorescein –…–GAG TTT TGG TAT TCT CTG GCG–

… –CTC AAA ACC ATA AGA GAC CGC–…

66

3.2.2 Fabrication of PDMS channel plates.

A 2"×2" PDMS channel plate was fabricated using the established

photolithographic method, as described in Section 2.2.2 [193]. The channel

pattern was designed using Visual Basic (Microsoft) and was printed on a

transparency to create the photomask at a resolution of 3368 dpi. The PDMS

channel plate consists of 16 parallel microchannels. The width of the straight

channels was 300 µm and the channel height was 30 μm. The length of the

straight section of each channel was 30 mm. Solution reservoirs (1 mm in

diameter) at both ends of channels were punched on the PDMS channel plate

using a flat-end syringe needle. The images of the assembly of a 2"×2" PDMS

channel plate on a 3"×2" glass slide are shown in Figure 3-2.

Figure 3-2 The images of the assembly of a 2"×2" PDMS channel plate on a 3"×2" glass

slide. (a) Image of16 channels filled with blue-dye solutions in horizontal direction. (b) Image of 16 channels filled with green-dye solutions in vertical direction. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

(a) (b)

67

3.2.3 Probe line printing, sample hybridization and result read-out by fluorescent scanning.

The glass substrates were chemically modified to produce aldehyde-

functionalized surfaces and aminated probe DNA was spotted in line array with

the PDMS channel plate. The detail procedures have been shown in Chapter 2.

In DNA target hybridization, the glass chip with probe line arrays was

covered with a PDMS channel plate. The straight channels were orthogonal to

the printed probe lines on the slide (as shown in Figure 3-1b). The DNA samples

(oligonucleotides or PCR products) were prepared in the hybridization buffer (4x

SSC + 0.2% SDS, unless stated otherwise). The PCR products were denatured

at 95°C for 4 min and then quickly cooled in an ice-water bath just before

hybridization. 1.0-µL DNA targets were added to the inlet reservoirs using an

automatic pipettor. Sample solutions in different reservoirs were then pumped

into channels by vacuum suction simultaneously applying at the 16 outlets. The

flow of DNA targets and their binding on the immobilized probes in a

microchannel was shown in Figure 3-3. Two methods were used to control the

hybridization temperature inside the microchannels. In the continuous-flow mode,

a Peltier device (CP1-12715, Thermal Enterprises, NJ) was placed under the

glass slide assembly and the hybridization temperature was adjusted by tuning

the voltage applied to the Peltier device. In the stop-flow incubation mode, the

assembly was incubated in a humidity box placed in an oven at a specified

temperature. Hybridization was achieved between complementary DNA targets

in solution and probe lines at the intersections, showing the hybridization patches

68

of 300 µm × 300 µm. The microchannels were rinsed immediately with 2-µL

hybridization buffer following hybridization.

Figure 3-3 Microfluidic hybridization of target DNA strands (in red) in the parabolic liquid

front to the probe DNA strands (in blue) immobilized on the glass chip surface. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

Following the hybridization and washing procedures, the glass slide was

scanned on a confocal laser fluorescent scanner (Typhoon 9410, Molecular

Dynamics, Amersham Biosystems). The resolution is 25 µm. The excitation

wavelength was 488 nm or 633 nm for fluorescein-labelled or Cy5-labelled

samples, respectively. The photomultiplier tube (PMT) voltage was set to 600 V.

The scanned image was analyzed by ImageQuant 5.2 software. In the data

quantification procedure, square frames (13 pixels ×13 pixels) was overlaid on

the square hybridization patches in the image. The average fluorescent signals of

the 169 pixels were measured in relative fluorescent unit (RFU).

69

3.3 Results and discussion

3.3.1 Flexible probe immobilization without spotting.

In conventional microarrays using the robotic spotting method, it usually

took 8-16 h for probe immobilization to complete after spotting [19]. To reduce

solvent evaporation during this process, a humidity chamber was used and

DMSO might have to be added in the probe solution. The latter may either

increase the spot size on the slide or reduce the actual amount of DNA fixed on

the solid support [112]. On the contrary, in the microfluidic printing method, the

probe solution, which was confined to the PDMS microchannels, would not dry

out even in several hours. Furthermore, shrinking of the channel dimensions to

micro-scale decreased the diffusion time and hence the incubation time needed

to complete probe immobilization [59, 63]. To study the effect of immobilization

time on the signal intensity, dual-labelled oligonucleotide probe (ADF) was flowed

through and incubated in the microchannels for different durations. It was found

that the fluorescent intensity was higher with immobilization time of 30 min than

with 15 min, as shown in Figure 3-4a. When immobilization time was more than

30 min, the signal intensity from immobilized probes did not increase by more

than 3%. Thus, 30 min was considered enough to achieve effective probe

immobilization.

70

Figure 3-4 Probe immobilization. (a) Effect of ionic strength of spotting solutions and immobilization time on the ADF signal. The probe line arrays were made by incubating 0.8 µL of 25-µM ADF prepared in 1.0M NaCl (grey bar) or 0.1M NaCl (white bar) in microchannels at different durations. The slide was chemically reduced and then washed with distilled water. The fluorescent signals were measured by scanning the slide at 488nm. (b) The fluorescent image of ADF probe lines (vertical green stripes) printed on the glass slide using different immobilization times. The brighter the probe lines, the stronger the fluorescent signals. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

The effect of solution ionic strength on probe immobilization was also

studied with the same chip. In the spotting buffer containing 1.0 M NaCl, we

observed a greater ADF signal intensity, as compared to that obtained in 0.1 M

NaCl (Figure 3-5a). Our observation is consistent with the previously reported

findings by Peterson et al. that single-stranded DNA is less adsorbed using a

buffer of a low salt concentration [221]. This is because the charged DNA strands

are better electrostatically shielded under the high ionic strength condition [222],

and so it could be easier for aminated oligonucleotides to come into contact and

to react with the aldehyde groups on the glass surface.

Flow direction

Probe ADF in 1.0M NaCl

20 mm

Probe ADF in 0.1M NaCl

(b)

0

6000

12000

18000

24000

30000

15 30 45 60 90 120

Immobilization time of Probe ADF (min)

Imm

obili

zati

on s

ign

al (

RFU

) in 1.0M NaClin 0.1M NaCl

(a)

71

Figure 3-5 Sample hybridization. (a) The immobilization signal of various ADF probe solutions (0.8-µL) at different concentrations (10 to 400 µM) that were incubated in microchannels for 2 h. After washing, the slide was scanned at 488nm. (b) The hybridization signals resulted from the above probe lines. Complementary oligonucleotides (CD’) labelled with Cy5 (100 nM, prepared in 1xSSC + 0.2%SDS) was hybridized to the ADF probe lines for 10 min. After washing, the slide was scanned at 633nm. (c) Overlaid dual-channel image of the same glass slide showing both printed probe lines (vertical green lines) and square hybridization patches (red) at intersections. ADF probe immobilization was achieved in duplicate at each concentration. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

The effect of the probe concentration on probe coverage was also studied.

As shown in Figure 3-5a, the fluorescent intensity is much higher using 25 µM

than 10 µM of ADF, but the rate of increase is reduced when the probe DNA

concentration goes from 25 to 400 µM. It was found that the signal obtained at

25-µM DNA probes has reached around 60% of the value obtained from 400-µM

probes. Apparently, the amount of immobilized oligonucleotides kept increasing

with the higher probe concentration, and the results did not generate an

optimized probe concentration for microfluidic probe printing. Further

(c) 17 mm

10 25 50 100 200 400

100-

nM D

’C

Probe ADF at different concentrations (μM)

Probe flow direction

0

5000

10000

15000

20000

25000

30000

0 50 100 150 200 250 300 350 400ADF probe concentration (µM)

Imm

ob

iliz

atio

n s

ign

als

(RFU

)

0

20000

40000

60000

80000

10 25 50 100 200 400

ADF probe concentration (µM)

Hyb

rid

izat

ion

sig

nal

s (R

FU)(a) (b)

72

experiments were carried out by the measuring the hybridization signals obtained

from these probe lines. Here, complementary oligonucleotides (CD’) labelled with

a second fluorescent tag, Cy5, were used to distinguish the hybridization signals

(at 633 nm) from the immobilization signals (at 488 nm). It is clear from Figure

3-5b that the hybridization signal has reached its highest value with probe lines

prepared with 50-µM ADF, and the amount of hybridized samples decreased

when higher probe concentrations were used. Our observations were consistent

with the findings from Le Berre et al. and Peterson et al. that higher probe density

on the glass surface could reduce the efficiency of duplex formation and the

kinetics of target capture procedure [112, 221]. The decreased hybridization

signals from probe lines at high concentrations might be due to steric constraints

resulted from the high probe density on the glass substrate. Accordingly, 25-µM

probe was chosen as a compromise for both signal sensitivity and reagent

savings. The surface coverage of the DNA probes on the surface was also

estimated. This was achieved by establishing a calibration graph correlating the

fluorescence signals to the concentrations of ADF solutions when they were

completely filled inside channels. It was found that the fluorescent intensity of the

probe lines created with 25-µM ADF was comparable to that of a 0.5-µM ADF

solution, leading to the surface density of 5 × 1011 strands/cm2, which was

comparable to the value of ~3 × 1012 strands/cm2 obtained in a recent study

[221].

73

3.3.2 Fast hybridization of multiple DNA samples.

Two sets of fluorescein-labelled oligonucleotide samples, FB’ and FD’, at

3 different concentrations, were hybridized to pre-printed oligonucleotide probe

lines of AB and AD at room temperature for 10 min. As shown in Figure 3-6a,

fluorescent patches showing successful hybridization occur at the intersections

between the vertical printed probe lines and horizontal microchannels filled with 3

samples of different concentrations (1, 10, 100 nM). It was found that 1μL of 1-

nM FB’ and FD’ could be detected. The signal-to-noise ratio of 1 nM of FB’ and

FD’ were 5.6 ± 0.4 and 4.1 ± 1.1, respectively. Good specificity of hybridization

was obtained, as shown in Figure 3-6b. The non-specific binding signals

remained low even at a high sample concentration of 100 nM. The hybridization

kinetics was found to be faster in microchannels than that in bulk solutions [89].

This experiment has demonstrated the detection of 2 different oligonucleotide

samples (1μL, at three concentration of 1, 10 and 100 nM) after hybridization at

room temperature for 10 min. Moreover, the patches formed in microfluidic

hybridizations are regular in shape, which is an important feature for subsequent

image analysis.

74

Figure 3-6 Hybridization of oligonucleotide samples to printed probe lines. (a)

Fluorescent images of the hybridizations of oligonulceotides (FB’ and FD’ prepared in 1X SSC + 0.2% SDS) with probe line arrays for 10 min at room temperature. (b) Histogram showing the fluorescent intensities of hybridization versus non-specific binding at various sample concentrations. The grey bars represent the signals of samples hybridized with their complementary probe sequences, i.e., FB’ with AB, or FD’ with AD; the white bars represent the non-specific binding. The error bars describe the standard deviations of the signals from 5 or 7 hybridization patches. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

After experiments with oligonucleotide samples, two PCR products (FB’P,

264 bp; FD’P, 259 bp) which were amplified from plant fungal cultures were

tested. Usually, the hybridization for PCR products requires higher temperature

and longer duration as compared to oligonucleotide samples. Therefore, in

conventional DNA microarray experiments, overnight incubation of samples in a

thermostated chamber is needed [223]. Accordingly, denatured PCR products

were hybridized to probe lines at 50 °C, which was at a temperature lower than

their melting temperature (~66 °C) [17]. Moreover, the results obtained from

different hybridization conditions were compared: the first set was carried out by

continuous flowing of samples for 5 min and the second and the third sets were

conducted by stop-flow incubation for 30 min or overnight. As shown in Figure

0

500

1000

1500

2000

2500

3000

3500

1nM-B'F 1nM-D'F 10nM-B'F

10nM-D'F

100nM-B'F

100nM-D'F

Hyb

rid

izat

ion

sig

nal

s (R

FU) With complementary probes

With non-complementary probes

AD AB AD AB AD

1-nM D’F

1-nM B’F

10-nM D’F

10-nM B’F

100-nM D’F

100-nM B’F

3 mm (a) (b)

75

3-7a, it was found that longer hybridization time gave stronger signals, but the

30-min hybridization signal for FB’P has reached ~75% of that obtained from

overnight hybridization; it was ~82% for FD’P. However, the two targets were

already distinguished clearly by using 5-min hybridization at continuous flow of

the PCR product samples, which suggested that the hybridization rate was

enhanced greatly by the microfluidic flow. Figure 3-7b depicts the fluorescent

images of the results obtained for hybridizations at 3 different flow/incubation

conditions. In the case of long-time hybridization, a humidified box had to be

used and the condensation of water vapour on the chemically modified glass

slide might create variations on the background signals shown in the fluorescent

images. However, in the case of 5-min flow, a low background was observed

because a short hybridization time was used and so no humidified box was used.

As compared to our previous work [105], where FB’P and FD’P were detected

with pin-spotted DNA microarrays, the new MMA method simplified the probe

creation step (no agarose, no spotting), reduced the sample volume (from 50 μL

to 1 μL) as well as shorten the hybridization time (from overnight to 5 min.). In

addition, multi-sample hybridizations were achieved all on the same chip.

76

Figure 3-7 Hybridization of PCR products to printed probe lines. (a) Fluorescent signals from the hybridization of 2.6-ng pre-denatured PCR products (FB’P and FD’P) at 50 ºC for 5-min flow, 30 min incubation and overnight incubation, respectively. (b) The fluorescent images corresponding to the left histogram of hybridizations of 2 samples to 7 probe lines. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

3.3.3 Effect of probe tether length and one-base-pair-difference discrimination.

In this work, we also studied the effect of different probe tether length on

the hybridization efficiency of PCR products with printed probe lines within

microfluidic channels. In a study reported by Shchepinov et al., they found that

the use of a longer tether in the immobilized probe led to a higher hybridization

signal [224]. In our hands, two aminated probes with 2 different tethers were

used: probe AB (with a C6 tether) and probe ALB (with a C12 tether), see Figure

3-8.

0

200

400

600

800

1000

B'PF D'PF B'PF D'PF B'PF D'PF

Hyb

ridi

zati

on s

ign

als

(RFU

)With complementary probesnon-specific binding

(b)

AB AD AB AD AB AD AB

Overnight incubation

30-min incubation

5-min flow

D’PF

B’PF

D’PF

B’PF

D’PF

B’PF

(a)

30-min incubation

Overnight incubation

5-min flow

77

Figure 3-8 The immobilized oligonucleotide probe with C6 or C12 tether as spaced from the glass substrate. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

After immobilizing these two probes (AB and ALB) under the same

conditions, the PCR products were hybridized to them at 50 °C for 5 min. As

shown in Figure 3-9 (a) and (b), probe ALB gave a ~3-fold higher hybridization

signals than probe AB when the PCR products (1.4 ng and 2.6 ng) were applied.

This finding is consistent with Shchepinov’s result, which is attributed to a less

steric hindrance where the longer tether probe is spaced farther away from the

glass substrate. However, a longer tether led to a decrease in the hybridization

specificity, and more non-specific binding (signals from FD’P) was observed,

shown as small black bars in Figure 3-9(a).

OSiO

O

HN

HN

Surface loaded C6 tethered probe DNA

3 5DNA

OSiO

O

HN

HN

3 5 DNA

Surface loaded C12 tethered probe DNA

78

Figure 3-9 Hybridization of PCR products to probes having different tether lengths. (a) Fluorescent signals from hybridization of PCR products (1.4 ng and 2.6 ng) to probes with 2 different tether lengths. (b) The fluorescent image corresponding to the left histogram of hybridizations of samples to 5 probe lines. (c) Effect of probe tether length on discrimination of PCR products with one-base-pair-difference. Fluorescent signals come from the fluorescent patches of three PCR products (FB’P, FBN’P and FD’P, 1.4ng for each). The small black bars come from the non-specific binding signals of FD’P and they are too low to be seen on probe AB. The error bars describe the standard deviations of the signals from 5 duplicated tests. (d) The fluorescent image corresponding to the left histogram of hybridizations of 3 samples to 5 probe lines. In all cases, the DNA targets were hybridized by continuous-flow method for 5 min at 50 ºC. Reprinted with permission from [94], Copyright © 2007 American Chemical Society.

The lower specificity obtained when a longer tether probe is used has an

impact on our study of the discrimination between two PCR products with one-

base-pair-difference. These two PCR products (FB’P, 264 bp; FBN’P 264bp)

were from two closely related subspecies, Botrytis cinerea and Botrytis

squamosa, respectively. They differ in only one base-pair at the centre of the

0

50

100

150

200

250

300

350

With probe AB With probe ALB

Hyb

ridi

zati

on s

ign

als

(RFU

) B'PFB'NPFD'PF

0

200

400

600

800

1000

1200

Hyb

ridi

zati

on s

ign

als

(RFU

)

(b)

AD AB ALB

2.6 ng/μL D’PF

1.4 ng/μL B’PF

2.6 ng/μL B’PF

1.4 ng/μL D’PF

AB ALB

(d)

AD AB AB ALB ALB

B’PF

BN’PF

D’PF

(c)

(a)

2.6 ng/μL

B'PF

D'PF

2.6 ng/μL

1.4 ng/μL 1.4 ng/μL

79

264-bp long sequences, see Table 3-1. The duplex formed between denatured

FBN’P and probe AB had a one base-pair mismatch (TTT:ATA instead of

TAT:ATA) in the centre of the PCR product. The melting temperature of the

mismatch duplex was calculated to be 5°C lower than that of the matched

duplex. The hybridization results were shown in Figure 3-9c and Figure 3-9d.

Although probe ALB gave higher hybridization signals for FB’P, the discrimination

between FB’P and FBN’P was lower. In terms of differentiation between FBN’P

and FB’P, it was found that the longer tethered probes captured more

mismatched DNA targets; the shorter tethered probe gave both lower non-

specific binding and a better discrimination between the two PCR products. It can

be seen from Figure 3-7c that the discrimination ratio of FBN’P over FB’P is 15%

for the probe AB, while the ratio becomes worse to ~60% for the longer tethered

probe ALB. Therefore, by using the short tethered probes, we have

demonstrated the capability of the MMA method to discriminate between low

amounts of PCR products (1.4ng) with one base-pair differentiation.

80

3.4 Conclusion

In this work, the microfluidic microarray assembly (MMA) method was

employed in which flexible probe array creation and fast DNA sample

hybridization were conducted in microchannels. The hybridization could be

fulfilled in minutes at the intersections between the sample channels and printed

probe lines. The process conducted in microfluidic channels was capable of

reducing the sample volume (< 1 µL), and of preventing the liquids from

evaporation and cross-contamination. It was demonstrated that one femtomole

DNA (1nM, 1µL) of oligonucleotide samples were detected in 10 minutes at room

temperature. The microfluidic method was also applied for greenhouse plant

fungal pathogen detection of two ~260-bp PCR products, Botrytis cinerea and

Didymella bryoniae, (1.4 ng/µL, 1µL, at 50 °C for 5 min). For the first time, fast

discrimination (in 5 min) between two 260-bp PCR products with one-base-pair-

difference was achieved with a discrimination ratio of 15%. It is demonstrated

that the MMA method provides the advantages of flexible probe creation, low

sample volume, good spot homogeneity, and fast hybridization rate, as applied to

agricultural problems.

81

4: GOLD NANOPARTICLE-ASSISTED SINGLE BASE-PAIR MISMATCH DISCRIMINATION ON A MICROFLUIDIC MICROARRAY DEVICE

4.1 Introduction

For many years, gold nanoparticles (GNPs) have been used in the fields

of bio-diagnostics based on different formats [225]. In the first format, GNPs are

labelled to thiol-modified oligonucleotide probes which are used for the specific

recognition of target DNA. Signal read-out is achieved based on the unique

properties of GNPs. For example, scanometrical methods were developed using

the catalytic effect of GNPs on silver staining [226-228]; a series of molecular-

beacon methods were constructed based on the strong quenching effect of

GNPs on fluorescence [229-232]; colorimetric methods were also developed

based on the hybridization-induced inter-linkage of GNP-labelled probes [233-

236].

Recently, the non-covalent binding of ssDNA to GNPs was discovered,

and this was applied to DNA analysis [237]. GNPs are usually stabilized via a

layer of adsorbed negative ions (such as citrate ions) on their surface [238], while

they are synthesized by citrate reduction of HAuCl4 in aqueous solutions [239,

240]. It was found that uncoiled ssDNA could bind to GNP surface through

attractive van der Waals forces between the bases and the negatively charged

82

GNPs [237]. This binding is tight and GNPs could be released only upon the

hybridization with complementary target DNA. Therefore, in the second format,

GNPs acted as competitors to discriminate between complementary and non-

complementary DNA targets. Subsequent duplex detection can be achieved

either by colorimetric method from salt-induced GNP aggregation or by

fluorescence-resonance-energy-transfer method [237, 241-243]. Since no thiol

modification is involved in this second format, the method is simple and it has

been used in oligonucleotide detection, single-base mismatch discrimination, as

well as aptamer-based applications in bulk solutions [237, 243-246].

Microfluidic technology offers the advantages of less sample consumption

and fast detection processes. Nevertheless, its applications to GNP-assisted

bioanalysis are mostly focused on immunoassay with GNP catalyzed silver

staining [247]. In terms of DNA analysis, a polydimethylsiloxane (PDMS)

microfluidic device was developed for DNA discrimination by hybridization-

induced GNP aggregation-related colour change [235]. However, only 15-mer

oligonucleotides were analyzed and concentration was tested at 6 µM.

In this work, two GNP-assisted methods for DNA analysis were evaluated.

In the first method, the glass surface was modified with a submonolayer of

GNPs, and probe DNA molecules were then applied and allowed to bind to

GNPs. The nanoscale spacing between GNPs on the surface resembles the

distance created from dendrimer coating, which allows better hybridization

efficiency and higher signals [112, 248]. In the second method, we found that

even when the probe DNAs were immobilized on a glass surface, they are also

83

able to hybridize with the GNP-bound target DNAs in solutions. GNP will then be

released and washed away. We thus combined the use of GNPs with microfluidic

technology and applied the method to the detection of PCR amplicons.

Moreover, it was found that the mismatched target DNAs showed less

hybridization with immobilized probes, as compared to the matched ones.

Therefore, the mismatched DNAs remained bound with GNPs in solution and

were washed away later. We have successfully achieved the room-temperature

discrimination of two 260-bp PCR products with one base-pair difference using

this second method.

4.2 Experimental

4.2.1 Fabrication of the PDMS-glass microchip.

PDMS channel plates consisting of 16 parallel microchannels were

fabricated, as described in Section 2.2.2.

4.2.2 DNA samples

Oligonucleotides were synthesized and modified by Sigma-Genosys

(Oakville, ON, Canada) or International DNA Technologies (Coralville, IA). The

21-mer DNA probes were modified with an amine group at the 5’-end. Target

oligonucleotides are either 21-mer or 50-mer with Cy5 dye at the 5’-ends. The

central 21 bases are complementary (perfect match) or one-base-pair mismatch

to the sequences of the probe molecules.

84

Two 264-bp PCR products (or amplicons) were amplified from genomic

DNA samples, and were labelled with Cy5 dyes as previously described [134].

The central sequences of the sense strand of perfect matched PCR amplicons

are complementary to the sequences of probe molecules, while the mismatched

amplicons have one base-pair difference from that of the perfect matched ones.

4.2.3 Depositing GNP layers on glass surface using microfluidic method.

In the first method of our work, a GNP sub-monolayer was coated on the

glass surface before probe immobilization. Here, the microchip with aminated

glass substrate was used. GNP solutions (20 nm in average diameter, Sigma Life

Science) at different volumes were flowed through the microchannels. GNPs thus

deposited onto the aminated glass surface and formed strip-like layers along the

microchannels. The uncovered amine groups on the glass surface were capped

by acetyl groups by filling through the microchannels with acetic anhydride [249].

After washing the microchannels with 95% ethanol followed with citrate buffer

solutions, the microchip was dried out for subsequent probe immobilization.

4.2.4 Preparation of DNA-GNPs conjugates.

In the second method of our work, sample DNA molecules were bound to

GNPs to form DNA-GNP conjugates before hybridizations. GNP solutions (5 nm

in average diameter, Sigma life science) were added into the DNA samples

(oligonucleotides or PCR products) in water. In the case of PCR products, the

mixtures were incubated at 95ºC to denature and uncoil the DNA chains and so

ssDNA molecules were produced for binding to GNPs non-covalently. The DNA-

85

GNP conjugates were snap-cooled in an ice-water bath before the hybridization

experiments. All the conjugate solutions were diluted to the desired concentration

before hybridization.

4.2.5 Probe line printing, sample hybridization and result read-out by fluorescent scanning.

The procedures of probe DNA printing, target hybridization, and signal

read-out are similar to the previous work depicted in Section 2.2.3.

4.3 Results and discussion

4.3.1 GNP-modified surface and its application to microarray DNA hybridization

In microarray technology, the effect of surface coverage on probe

orientation and sample hybridization signals has been studied by many groups

[221, 224, 250, 251]. It was found that probe density is a governing factor for the

amount of hybridization as well as for the kinetics of the target hybridization. In

order to maximize hybridization signals, the probe density should be high

enough. However, at a very high probe density, the steric hindrance by the

adjacent DNA probe molecules reduces the target capture efficiency and

hybridization kinetics. On the other hand, at a lower probe density, probes are

more accessible to targets and hybridization kinetics is faster [221]. Based on

this finding, different approaches have been proposed to achieve a controllable

probe density by adjusting the spacing of surface functional groups (such as –

CHO or –NH2) used for probe immobilization. For instance, nanoscale distance

86

was created by either dendrimer coating or self-assembled dendrimers on glass

surface [112, 248]. By covalent binding between modified DNA probes and the

pendant functional groups on the dendrimer molecules, probe density was

reduced and higher detection sensitivity was achieved in hybridizations.

In this work, we made use of a sub-monolayer of GNPs on glass surface

to achieve a nanoscale-controlled spacing for subsequent probe immobilization.

As depicted in Figure 4-1, GNPs deposited on the aminated glass surface are

bound electrostatically between the positive amine surface and negative GNP

surface. Due to the inter-particle repulsion from the surface negative charges of

GNPs, a self-assembled sub-monolayer forms on the glass surface [252, 253].

The “saturated” coverage was measured at ~30% by different groups and the

spacing has been calculated at around 12.5 nm between GNPs [254, 255]. This

spacing is comparable to the length of 21-mer oligonucleotides (~12nm) [256].

Since probe DNA molecules are to be immobilized onto the GNPs, the DNA

strands will also be separated in nanoscale spacing.

87

Figure 4-1 Schematic diagram of the formation of the nanoscale-controlled spacing between oligonucleotide probes using GNPs. (a) Protonated amine groups on the APTES-treated glass surface. (b) GNPs, with adsorbed citrate ions, are then deposited to the aminated glass surface and form a submonolayer. (c) The remaining non-reacted amine groups are deprotonated using a pH 11 buffer and masked by acetic anhydride (Ac2O). (d) When oligonucleotides probe DNA solutions flow through the submonolayer, the amine groups at the 5’-end of the molecules will bind only to GNPs which are well spaced out on the glass surface. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics.

We used a microfluidic method to deposit GNP sub-monolayer on glass

surface on a 16-microchannel PDMS chip. The confinement of solutions in

microchannels resulted in strip-like GNP layers on the glass surface. The inset in

Figure 4-2 shows pink strips after flowing through 10-µL GNP solutions. The pink

colour of the layer reflects an unaggregated status of gold nanoparticles on the

surface. It is because GNPs will induce a shift of the surface plasmon band and a

visual colour change (from red to blue) when they are in close contact [252].

Grabar et al. achieved different particle coverage by incubating with GNP

solutions for different times [254]. Unlike the bulk solution method, microfluidic

1) pH11 buffer

2) Ac2O 3) Citrate

buffer Amine-labelled probes

(b)

(c)

(a)

(d)

88

method offered flexibility to immobilize different amount of GNPs on the same

glass slide and deposition was very fast due to the high surface-to-volume ratio

in microchannels. The more solutions flowed through the channels, the more

nanoparticles were deposited on the surface and showed deeper colour. From

Figure 4-2b, it can be seen that the strips from applying 2-µL solutions is barely

visible. It was found that the GNP binding on glass withstood the later acylation

and washing steps.

Figure 4-2 Different modified surfaces for hybridization experiments. (a) Image of an APTES modified glass slide with immobilized GNP strips (b) Magnified image of the glass slide with insets showing (i) Flow through 2-µL GNP solutions, (ii) Flow through 10-µL GNP solutions. The glass slide was first treated with APTES solution to create an aminated surface. Then a PDMS channel plate was sealed against the glass slide. GNP solutions in different amounts were filled through 8 microchannels. The deposition of GNPs onto the aminated surface resulted in the pink strips. The rest of the microchannels were left empty to keep the glass surface aminated. The images were taken after washing the microchannels and peeling off the PDMS plate. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics.

The effectiveness of GNP submonolayer coated surface was evaluated

with the hybridization of oligonucleotides samples. With the PDMS-glass

microchip, probe DNA solutions were flowed through microchannels (vertically

oriented) and incubated with the GNP strips on the glass slide. DNA probe

molecules that are conjugated with amine groups at 5’-end were introduced for

(a)

(b) (i)

(ii)

89

immobilization to GNPs [252]. After probe immobilization, the PDMS channel

plate was peeled off and the second PDMS plate with horizontally oriented

microchannel was sealed against the glass slide for sample hybridization at the

intersection between the horizontal microchannel and vertical probe strips. This

orthogonal or intersection microfluidic method has been used in our previous

work and many other groups [91, 93, 94, 96, 129, 134]. The resulting

hybridization patches from oligonucleotides complementary perfect match

(PM) and one-base mismatch (MM) are shown in Figure 4-3. For comparison,

sample hybridization with probes immobilized on aminated glass surface without

GNPs was also shown. It can be seen that the hybridization signals on GNP-

modified surface are higher than those on aminated surface, and more GNP

coverage obtained by using more GNP solutions has resulted in even higher

signals.

Figure 4-3 Fluorescent images of the hybridization results from (a) 21-mer and (b) 50-mer target oligonucleotides at different modified surfaces. The targets are either complementary perfect- match (PM) or one-base mismatch (MM) in the sequence center with the probe DNA molecules. (a) Hybridization of Cy5 labelled 21-mer targets. The probe DNA lines were printed microfluidically on: aminated glass surface, on surface modified by flowing through 2-µL GNP solutions, and on surface modified by flowing through 10-µL GNP solutions (b) Hybridization of Cy5 labelled 50-mer targets. The probe DNA lines were printed microfluidically on: aminated glass surface, on GNP modified surface with Ac2O treatment to cap remaining amine groups, and on GNP modified surface without AC2O treatment. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics.

Probe DNA lines PM MM

Aminated surface

GNP-2uL

GNP-10uL

10 nM 50-mer oligonucleotide

targets

10 nM 21-mer oligonucleotide

targets

(a) (b)

Aminated f

GNP modified, but no AC2O treatment

GNP modified with AC2O treatment

Probe DNA lines PM PM

90

GNP modified surface offers advantages of nanoscale spacing in DNA

probe immobilization, and resulted in higher hybridization signals, and our results

have confirmed this. However, the method currently can only be applied to

oligonucleotide targets that have a similar sequence length with the probes, but

not to longer DNA targets. As shown in Figure 4-3b, when 50-mer targets were

applied, the hybridization signals (the middle image) are much lower than those

from aminated surface only (the top image). This is because the dangling end of

DNA strand with the fluorescent labels could drop on the GNPs and the

fluorescence signals are thus be quenched [229]. The bottom image in Figure

4-3b also shows the hybridization results on GNP modified surface but without

acetic anhydride treatment. Because GNPs formed a sub-monolayer and cannot

cover all the amine groups on the glass surface, the probe DNA molecules can

still bind to the aminated surface non-covalently and hybridize with the targets

later [257, 258]. These duplexes were not quenched by GNPs and can thus be

detected, though with a weaker signal than obtained from the purely-aminated

surface (top image).

4.3.2 GNP-DNA conjugates and its application to single-base-pair discrimination

The use of nanoscale spacing surface for DNA probe immobilization

enhanced hybridization signals. However, the discrimination ratio between

perfect-matched and mismatched DNA could not be improved, and longer

oligonucleotide samples could not be detected efficiently through the use of the

91

GNP modified surface. To solve this problem, a different strategy of using GNPs

was employed in the second part of our work.

Figure 4-4 (a) GNP-DNA conjugates from the incubation of target DNAs with gold nanoparticles. (b) Perfectly matched target DNAs desorbed from GNPs and hybridized to the surface immobilized probes. (c) Mismatched DNAs remained bound to GNPs and were washed away through microfluidic method. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics.

The principle of GNP-assisted DNA discrimination was illustrated in Figure

4-4. This is based on non-covalent binding between GNPs and targets, in

competition with the specific interaction between the targets and the immobilized

probes. This non-covalent binding is thought to act between GNPs and nitrogen

bases on the DNA strands, and is resulted from both hydrophobic interaction and

electrostatic adsorption [259]. Here, target DNA labelled with fluorescent

molecules were first incubated with GNPs solutions at high temperature. The

solution of GNP-DNA conjugates was then applied to pre-printed DNA probe

lines on the glass slide through microfluidic method. Because the base-pair

interaction between matched DNA chains is strong, sample DNA molecules

could desorb from GNPs and hybridize with the immobilized probes. The read-

GNP Labelled

target DNA Glass slide with hybridized DNA

DNA bound to

GNP

(a) (b)

Washing away GNPs

(c)

Glass slide with immobilized probes

92

out of the hybridization signals were achieved read-out through the fluorescent

labels on the sample DNA molecules. On the contrary, mismatched DNA showed

much less binding energy with the probes and thus still bound with GNPs. The

conjugates were washed away in the microfluidic flow and discrimination was

made.

Figure 4-5 (a) Images of hybridized patches of perfect-matched (PM) and mismatched (MM) target oligonucleotides in triplicate. Here, the oligonucleotides were pre-incubated with GNPs (5 nm) at different ratios. (b) Discrimination ratios between PM and MM duplexes. The discrimination ratios were calculated by dividing the signal of PM DNAs with that of MM DNAs (The higher ratio, the better). (c) The fluorescent hybridization signals from the images in (a), and the results at Oligo/GNPs = 1:1 are expanded and shown in the right inset. Reprinted with permission from [95], Copyright © 2010 American Institute of Physics.

The proposed GNP-assisted discrimination method was first verified with

two 50-mer oligonucleotides with one-base difference in centre. The samples

hybridized with the same probe molecules to produce two types of duplexes,

(a) Hybridization of oligonucleotides

(c)

PM

No GNPs added

MM PM MM PM MM

Oligo:GNPs = 2:1

(b)

0

400

800

1200

1600

Oligo:GNPs = 1:1

Fluo

resc

ent

inte

nsity PM

MM

Oligo:GNPs = 1:1

0

4000

8000

12000

16000

20000

No GNPs added Oligo:GNPs = 2:1

Oligo:GNPs = 1:1

Fluo

rese

cent

inte

nsity

(R

FU)

PM

MM

Ratio of PM/MM oligos

0

2

4

6

8

10

No GNPsadded

Oligo:GNPs= 2:1

Oligo:GNPs= 1:1

Dis

crim

inat

ion

ratio

s

93

namely the perfectly matched (PM) duplex and the mismatched (MM) duplex. As

shown in the images in Figure 4-5a, without pre-incubation with GNPs, the

hybridization signals from mismatched targets are very close to those obtained

from complementary targets. The discrimination ratio, which is the hybridization

signal ratio of PM duplexes over MM duplexes, is around 1.4. With the use of

GNP conjugates instead of free DNA molecules, the discrimination ratio was

raised up to 6.8. A clear discrimination between two oligonucleotides was

observed from the images in Figure 4-5a.

The effect of the molar ratio between GNPs and target oligonucleotides

on hybridization signals and discrimination ratios was also investigated. The

molar concentration of GNPs can be calculated from total gold concentration as

well as the size of the GNPs [260]. For 5-nm diameter GNPs in our study, the

particle molar concentration is around 86 nM. Conjugates of different GNP/DNA

ratios were thus prepared in this manner. Figure 4-5c compares different

hybridization intensities from conjugates of different Oligo/GNP ratios. It was

found that the more GNPs were incubated with DNA samples, the weaker were

the hybridization signals. The ratio at Oligo/GNP = 1:2 even resulted in non-

detectable hybridizations. This observation could be explained by the relatively

strong binding between GNPs and DNA as well as the slow kinetics of

desorption. Despite the reduction in the fluorescent intensities, GNPs does

enhance the discrimination of single base-pair mismatch, and the hybridization

signals are still adequate as shown in both the images and the inset graph in

Figure 4-5c.

94

Figure 4-6 (a) Images of hybridized patches of perfect-matched (PM) and mismatched (MM) PCR products in triplicate. Here, the amplicons were pre-incubated with GNPs (5 nm) at different ratios. (b) Discrimination ratios between PM and MM amplicons. The discrimination ratios were calculated by dividing the signal of PM DNAs with that of MM DNAs (The higher the ratio, the better). Reprinted with permission from [95], Copyright © 2010 American Institute of Physics.

This nanoparticle-assisted microfluidic method was applied to the room-

temperature discrimination of two related Botrytis subspecies, B. cinerea and B.

squamosa [105]. The two PCR amplicons differ in only one base pair in the

middle of the 264-bp long sequence [134]. Although these two targets were

already discriminated in our previous work, thermal stringency up to 62˚C has to

be applied to the microchip for washing away mismatched duplexes [134]. In this

work, amplicons were first incubated at 95˚C with GNPs. This incubation serves

for two reasons: one is to denature double-strand amplicons as the usual

procedures and another is to promote the subsequent binding of ssDNA to

GNPs. The later snap chilling procedures (at 4 ˚C) prevented the renaturation of

ssDNA molecules. Although both of the two complementary strands were bound

to GNPs and co-existed in the same solutions, they can be still used as samples

for later microarray hybridization because the renaturation between two long

ssDNA strands with high complexity is much slower than that between long

ssDNA and short oligonucleotide probes [256]. The discrimination ratio without

Ratio of PM/MM PCR amplicons

0

5

10

15

20

25

30

35

No GNPsadded

DNA:GNPs= 4:1

DNA:GNPs= 2:1

Dis

crim

inat

ion

ratio

s

(a) Hybridization of PCR amplicons PM MM PM MM PM MM

(b)

No GNPs added

DNA:GNPs = 4:1

DNA:GNPs = 2:1

95

the use of GNPs is ~3.6 while it goes up to ~27.7 with the assistance of

nanoparticles (Figure 4-6b). Compared with the discrimination ratio of ~6.7 using

temperature stringency at 50°C (results not shown), the GNP-assisted method has

not only improved discrimination, but also alleviated the need of high temperature

and related heating devices in the microfluidic chip applications.

4.4 Conclusion

In this work, two GNP-based DNA analysis methods using a microfluidic

device are presented. In the first method, fast coating of a submonolayer of

GNPs with nanoscale spacing on glass surface was achieved using a microfluidic

chip. Probe DNA molecules were then immobilized onto the GNP submonolayer.

The hybridization efficiency of the target oligonucleotides was improved due to

the nanoscale spacing among probe molecules. In the second method, target

DNA molecules, oligonucleotides or PCR amplicons, are first bound to GNPs and

then hybridized to the immobilized probe DNA on a glass slide. With the aid of

GNPs, we have successfully discriminated, at room temperature, between two

PCR amplicons (derived from closely related fungal pathogens, Botrytis cinerea

and Botrytis squamosa) with one-base-pair difference. DNA analysis on the

microfluidic chip avoids the use of large sample volumes, and only a small

amount of oligonucleotides (8-fmol) or PCR products (3-ng) was needed in the

experiment. The entire procedure is accomplished at room temperature in an

hour, and apparatus for high temperature stringency is not required.

96

5: FUNGAL PATHOGENIC NUCLEIC ACID DETECTION ACHIEVED WITH A CD-LIKE MICROFLUIDIC MICROARRAY DEVICE

5.1 Introduction

DNA microarrays are typically constructed either by on-chip synthesis of

oligonucleotide probes or by spotting of pre-synthesized probes on activated

substrates [111]. The applications of microfluidics to the fabrication of

microarrays are relatively recent and it has been reported that microchannels

provide a means to increase the hybridization rates over passive hybridization

using cover slips [85, 87, 89].The applications in which the DNA hybridizations

are conducted in microfluidic channels include the detection of oligonucleotides

[72, 84, 87, 261], bacterial DNA [109], clinical diagnostics of colorectal cancers

[86, 217], and single-nucleotide polymorphism studies [89, 106].

For the immobilization of DNA probes, microchannels have been used to

prepare line arrays, which were subsequently hybridized with DNA samples

delivered via a second set of microchannels (that orthogonally intersect with the

probe line arrays). Such a 2-D microarray method has been applied to the

analysis of oligonucleotide and PCR product samples [91, 93, 96, 262]. In these

reports, either capillary effect or pressure pumping was used to generate

microflows. We have also created a 16 × 16 microarray on rectangular glass

97

slides for fungal DNA analysis with pressure-pumping flows, see Chapter 3 and

Chapter 4.

Based on the unique design of spiral microchannel from our group, we

have been working on DNA microarrays fabricated on CD-like glass chips. With

the probe lines first created using radial channels, DNA hybridizations occurred

in the spiral channels that orthogonally intersected with the radial probe lines

[133, 263]. Liquid delivery in these channels (96 in total) can conveniently be

achieved by centrifugal pumping, in both the radial and spiral directions; this

method of liquid pumping has previously been applied only in the radial fashion

[106]. In this work, we apply this approach for the detection of two plant fungal

pathogens, Botrytis cinerea and Didymella bryoniae, which have caused serious

economic losses in greenhouse vegetable industry annually [210]. Traditional

methods of identification disease organisms can be slow and inconclusive, thus

preventing the timely implementation of the appropriate control measures.

Molecular diagnostic methods for plant pathogen include antibody-based and

nucleic acid-based approaches. The latter have increasingly been used in recent

years due to rapidity, simplicity and sensitivity [105, 264]. In this paper, coupled

with the proposed all-microfluidic assembly for probe immobilization and

hybridization, this DNA-based method has demonstrated the advantages of

flexible construction of oligonucleotide probe arrangements and multiple sample

capability in the subsequent target hybridizations. Moreover, we have achieved

accurate and faster hybridizations using smaller amount of samples (1 µL) than

in the conventional microarray method (30-50 µL).

98

5.2 Experimental

5.2.1 Materials

CD-like glass disks were purchased from Precision Glass & Optics (Santa

Ana, California, USA). They are 4” in diameter and were drilled with a 0.59”

centre hole. All other chemicals and materials are similar to those depicted in

Section 2.2.1. The sequence of the probes, marker, and targets are shown in

Table 3-1.

5.2.2 Surface modification of glass chips and the fabrication of PDMS channel plates

The glass substrates were chemically modified according to the procedure

shown in Section 2.3. The fabrication of PDMS channel plates has been depicted

in Chapter 2. As shown in Figure 5-1, the size of each PDMS plate is 92 mm in

diameter and ~2mm in thickness. The depths of the radial and spiral

microchannels were 25 or 50 μm, respectively. The solution reservoirs (1 mm in

diameter) were created on the channel plates by punching the PDMS chip using

a flat-tip syringe needle. All reservoirs were numbered for easy tracking during

solution introduction.

99

Figure 5-1 Assemblies of the PDMS channel plates with glass disks. (a) The circular channel plate with radial microchannels. All the channels (100 in total) were filled with dye solutions; (b) The image of an equi-force spiral channel plate assembled with a glass disk. Here, 30 out of 100 spiral microchannels, three in a group, were filled with dye solutions. The scale bar represents 10 mm.

5.2.3 Probe line array creation

As shown in Figure 5-2(a), the radial PDMS channel plate was sealed

against the aldehyde glass disk. 0.8-µL probe DNA solution was added into the

inlet reservoirs by a micropipette. The probe solution was filled through the

channels by spinning the assembly at 300 rpm for 10 min. With incubation at

room temperature for a certain time, covalent Schiff linkage was formed between

the amine ends of probe oligonucleotides and the aldehyde groups on the glass

surface [209]. After immobilization, the residual solutions were spun off the radial

channels at 3600 rpm for 10 min. The PDMS channel plate was then peeled off

and the glass disk was reduced with a NaBH4 solution (100 mg of NaBH4

dissolved in 40 mL of 0.75x PBS + 25% ethanol + 0.15% Triton X-100) for 15 min

10 mm

(a) (b)

100

to reduce the Schiff linkage to the C-N single bond. The glass disk was then

rinsed with deionized water for 2 min and dried by nitrogen gas stream.

Figure 5-2 Schematic diagram of probe immobilization and sample hybridization using CD-like microfluidic microarray assemblies. (a) Probe printing procedure: an array of radial probe lines was created on an aldehyde-modified glass disk using a radial channel plate. (b) Hybridization procedure: the hybridization occurring at the intersections between the spiral channels and radial probe lines, shown as colored patches in the last circle. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd.

5.2.4 Sample hybridization with spiral channel plate assembly

In DNA target hybridization, the glass disk printed with radial probe lines

was covered with a spiral PDMS channel plate (as shown in Figure 5-2(b)). The

direction of spiral channels was just orthogonal to the printed probe lines on the

disk (as shown in Figure 5-2(b)). 1.0 µL of fluorescein–labelled target DNA

solutions were added to the inlet reservoirs by a pipette. The spiral PDMS

channel plate assembly was spun at 700 rpm for 3 min on a rotating platform for

continuous liquid delivery. If the hybridization time was more than 10 min, the

channels were first filled by spinning at 700 rpm for 2 min, followed by stop-flow

incubation for a definite duration, unless stated otherwise. In this case, the whole

1) Sealed against a radial channel plate

2) Flow through DNA probe solutions.

Aldehyde glass disk with a central hole

Peel off PDMS radial channel plate

Radial probe lines printed on glass disk

Spiral Channel plate with microchannels

1) Sealed against the glass disk with probe lines

2) Flow through DNA sample solution

Hybridization at probe-line-channel intersections

1) Peel off spiral channel plate

Hybridization results on the disk

2) Scanned by fluorescent scanner

Spun the assembly to drive flows inside channels

(a)

(b)

101

assembly was removed from the rotating platform, put in a humidified container,

and was incubated in a temperature-controlled oven for duration up to 150 min.

Following hybridization, the CD-like assembly was spun at 3600 rpm for 10 min

to dry the channels and no further channel washing was needed. After removal of

the PDMS channel plate, the glass disk was rinsed with 2× SSC + 0.2% SDS for

1 min, then with 2× SSC for another 1 min, and was then dried in a nitrogen

stream.

5.2.5 Quantification of fluorescent image and data analysis

The glass disk with hybridization patches was scanned and analysed

following the procedures depicted in Section 2.5. The image was collected at a

resolution of 10 µm.

5.3 Results and discussion

5.3.1 Improvements in the design and fabrication of channel plates

In this work, radial and spiral channel width was designed to be 180 μm

and 80 μm, respectively. The spacing between spiral channels was 120 μm.

Since the samples in the spiral channels could only be hybridized at their

intersections with the probe lines, the dimension of the hybridization patches was

defined by the width of the radial and spiral channels, i.e. 180 μm × 80 μm. In the

previous reports, both radial and spiral channel widths were 60 μm, leading to the

dimension of the hybridization patches to be 60 μm × 60 μm [133, 263]. Though

the channel width in the new design has been increased, the dimension of the

disk was kept to be 92 mm in diameter by moving the inlet reservoirs closer to

102

the centre of the disk than in the case of the old design. With the wider channels

in the new design, the liquid transport from the inlet reservoir could be initiated at

a lower spinning speed (300 rpm for radial channels and 700 rpm for spiral

channels). On the contrary, with the 60-μm wide radial channels in the old

design, a spin rate of 500 rpm was used for liquid transport, and care must be

taken to avoid the liquid being spun out radially over the disk surface from the

inlets, rather than inside the radial channels. In the case of the 60-μm wide spiral

channels, the spinning speed must be at least 1800 rpm in order to obtain a

steady liquid flow inside the microchannels.

5.3.2 Hybridization of oligonucleotide samples on the CD-like MMA

After the radial probe lines were printed on the disk, the spiral channel plate was

sealed against the disk for sample delivery. As shown in Figure 5-3(a), a section

of the PDMS spiral channel plate that was sealed against the disk was shown, in

which 6 sample channels orthogonally intersected with 3 printed probe lines on

the disk. After hybridization, a series of rectangular patches were formed at the

intersections (see Figure 5-3(a) for schematic, and Figure 5-3(b) for actual

results, respectively). In one experiment shown in Figure 5-3(b), the probes AB

and AD were distributed alternatively and radially on the glass disk during the

probe immobilization step. With the assembly of a spiral channel plate, a group of

FB’ and FD’ solutions (2, 5 and 10 nM, each in quadruplicate) were tested with 3-

min spinning at room temperature. Successful hybridizations have been achieved

between the oligonucleotide samples and their complementary probe lines. As

depicted in Figure 5-3(b) inset, high specificity resulted in the alternative patterns

103

of fluorescent patches. For instance, along the radial AD probe lines, fluorescent

patches appeared only from the group of FD’ oligonucleotide samples, but not

from the oligonucleotides FB’. In each group of samples, increased

concentrations led to increased fluorescent intensities of the patches.

Figure 5-3 DNA hybridization on a CD-like glass chip. (a) The schematic diagram of hybridization patches formed at the intersections between 6 sample channels and 3 probe lines. (b) The fluorescent image of the entire glass disk. The 3 straight traces are resulted from the radial flow of the fluorescent marker during probe immobilization. The 3 grey spiral traces are resulted from the spiral flows of marker during sample hybridization. Hybridization results obtained in the rectangular region are expanded to give the middle inset. The right inset shows the groups of rectangular patches formed near the disk center, which are resulted from the hybridization of different concentrations of oligonucleotide samples with their complementary probe lines. The probe lines (AD or AB) were created by 40 min-incubation of 25-µM aminated oligonucleotide probes in the radial channels. Oligonucleotides hybridizations were achieved at room temperature in 3-min spinning at 700 rpm, and then dried out at 3600 rpm. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd.

AD

AB

AB

B’F

D’F AD

AB

AB

Rectangular hybridization patches after washing

Glass disk immobilized with probe line arrays

Samples in spiral channels flow over probe lines

(a)

(b)

10 nM B’F

The fluorescent marker line

5 nM B’F

2 nM B’F

10 nM D’F

Radial AD probe line

Radial AB probe line

5 nM D’F

104

On the same chip with the same conditions, calibration curves of both

oligonucleotide samples are shown in Figure 5-4(a). It was found that as low as

0.5 nM of FB’ or FD’ could be detected. As compared to the conventional

microarrays, in which ~ 50-µL sample is usually incubated at one time, the

microfluidic microarray method only requires as low as 1 µL of samples.

Moreover, this method does allow for the use of a flexible sampling volume to

improve sensitivity. For instance, with a 1-µL sample, the signals from 0.1 nM FB’

could not be detected effectively under the current conditions (Figure 5-4(b)).

However, by spinning the MMA five times each with 1 µL of the sample solutions,

an enough signal could be “accumulated”. It was found that the hybridization

intensity of 5 µL of 0.1-nM FB’ was comparable to that of 1 µL of 0.5 nM FB’

(Figure 5-4(b)). Therefore, the MMA method offers another way to detect the

samples with lower concentration of targets with an in-channel concentrating

step.

105

Figure 5-4 (a) The calibration graph of the hybridization signals from oligonucleotide FB’ (top curve) and FD’ (bottom curve). (b) The hybridization signals of FB’ samples using different solution concentrations and volumes. In all cases, the samples were hybridized to 25-μM probe lines at room temperature for 10 min and the error bars describe the standard deviations of the signals from 9 hybridization patches. The fluorescent intensities expressed have already been subtracted from the background. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd.

5.3.3 Hybridization of PCR products on CD-like MMA

In comparison with the oligonucleotide samples (~20 mer), the PCR

products have much longer strands (~260 bp). In this case, only a short section

(~20 bases) of the long strand can anneal to the oligonucleotide probes

immobilized on the glass surface, and the strand must align itself to be in register

with the complementary sequence of the probe. This re-annealing process

required longer time for PCR products than for the complementary

oligonucleotide samples, and the process should be promoted by higher

temperature [265]. It was found that 5-min continuous flow is not enough for the

detection of the PCR product samples (FD’P). As depicted in Figure 5-5(a), the

signal intensity from the hybridization of 20-nM FD’P with 10-min continuous flow

was less than half of the intensity from that of 5-nM FD’ conducted by 5-min

0

30

60

90

1µL of 0.5nM 1µL of 0.1nM 5µL of 0.1nM

Microfluidic operation

Flu

ores

cen

t in

ten

sity

(R

FU)

y = 53.518x + 30.584R2 = 0.9974

y = 29.52x + 41.408R2 = 0.9965

0

100

200

300

400

500

600

700

0 2 4 6 8 10Oligo sample concentration (nM)

Flu

ore

scen

ce i

nte

nsi

ty (

RFU

)

(b) (a)

106

continuous flow at 45 ºC, even though the PCR product had a 4-fold higher

concentration and took 2-fold longer hybridization time than the oligonucleotide

sample. Stop-flow method was thus applied in PCR product detections to

achieve stronger signals. The hybridization signals increased when the FD’P

samples were incubated inside the spiral channels for up to 2 hours under the

stop-flow conditions (Figure 5-5(a)).

Figure 5-5 (a) Comparison of different flow conditions for the hybridization of oligonucleotides and PCR products at 45 ºC. From left to right the hybridization conditions were: oligonucleotide samples (FD’, solid bar) by continuous flow for 5 min, PCR products (FD’P, hatched bar) by continuous flow for 10 min, or by stop-flow incubation for 10, 30, 60 and 120 min. For continuous flow, the spun rate was 500 rpm; whereas for stop-flow incubation, the samples were first introduced at a spun rate of 700 rpm for 2 min. For oligonucleotide sample, 5-nM FD’ was used because it already gave sufficiently high intensity; for PCR products, 20-nM FD’P was used, but the signal for 10-min stop-flow hybridization of FD’P was not detected and the variation of background was shown instead. The error bars describe the standard deviations of the signals from 9 hybridization patches. (b) The effects of time and temperature on the hybridization of PCR products. FD’P (20 nM) was introduced into spiral channels by spinning at 700rpm for 2min followed by incubation in an oven for different times at different temperatures. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd.

This long-time incubation also helps when the hybridizations are carried

out at higher temperatures. As shown in Figure 5-5(b), the fluorescence intensity

0

100

200

300

Flu

ores

cen

t in

ten

sity

(R

FU)

0

100

200

300

35ºC 45ºC 55ºC

Hybridization temperature

Flu

ore

scen

t in

ten

sity

(R

FU) 30min

60min 120min

Stop-flow

(a) (b)

5 min

10 min

10 min

30 min

60 min

120 min

Continuous-flow

107

increased with hybridization time and 2-hour incubation gave the highest signals

at each temperature. However, water vapour could diffuse out slowly through

PDMS [266], which may lead to unwanted sample drying inside channels during

the incubation process when the time was longer than 2 h. Therefore, the 2-hour

hybridization time was used in the subsequent work.

Figure 5-5(b) also depicted the effect of temperature on the signals. The

signal intensity at 45 ºC is higher than that obtained at 35 ºC, but it drops at 55

ºC. This result was consistent with that of the previous work [105], where FD’P

solution was incubated with its complementary probes on agarose-coated slides

at 42 ºC overnight.

5.3.4 Hybridization specificity on CD-like MMA

The hybridization specificity of both PCR products and oligonucleotide

samples to radial probe line arrays was further examined. In the conventional

DNA microarray method, in order to remove non-specific binding after overnight

incubation, a lengthy washing step with buffer solutions at different

concentrations is always needed during the post-hybridization procedure [223].

Moreover, the efficiency of washing is not always satisfactory because the

diffusion rate of the bulk solution to the glass surface is low, which might easily

cause comet-like dots to appear in the final results [19]. In contrast, the

microfluidic washing method is fast and efficient. By spinning the assembly at a

high speed for a few minutes, the sample flow continually removed any un-

hybridized DNA molecules and no further channel washing was needed. After 1-

108

min of brief cleaning of the dust on the glass surface with a buffer solution, the

glass disk was ready for fluorescent scanning.

(c) D N N N N N N N B N N N N N N N D

40nM D6'PF PCR

20nM D6'PF PCR

10nM D6'F oligo

5nM D6'F oligo

Figure 5-6 The hybridization specificity shown by the plots of fluorescent intensities versus locations in the spiral channels. The samples are fluorescein-labelled oligonucleotide (FD’) and PCR products (FD’P), each at 2 concentrations. (a) The samples flow from left to right intersecting the AB and AD probe lines alternatively. (b) The samples flow from left to right intersecting various probe lines in this sequence: AD, 7 non-complementary probe lines, AB, 7 non-complementary probe lines, and AD, as depicted in the box as D, N, B, N and D, respectively. (c) The fluorescent images correspond to the hybridizations and backgrounds shown in (b). All the probe lines were created by 40 min-incubation of 25 μM aminated oligonucleotide probe. Oligonucleotide sample hybridizations were achieved by 3-min spinning and dry-out at room temperature. For PCR products, they were firstly introduced into channels by spinning and then incubated in oven at 45°C. After hybridization for 2 hours, the DNA sample solutions inside channels were spun out at 3600rpm. Reprinted with permission from [129], Copyright © 2008 Elsevier Ltd.

Figure 5-6 showed the hybridization specificity of different samples on the

glass disk. The fluorescent intensities were extracted from the spiral track of

sample flows. We have explored 2 different formats to print the probe lines. In

(a)

500 1000 1500 2000 2500 3000 3500

20

40

60

80

100

120

140

Distance (pixels)

Flu

ores

cent

cou

nts

40-nM FD'P 20-nM FD'P

10-nM FD' 5-nM FD'

(b)

500 1000 1500 2000 2500

20

40

60

80

100

120

140

Distance (pixels)Fluo

resc

ent c

ount

s

109

Figure 5-6(a), the probes AB and AD were created in an alternative manner. The

known intersection locations between the radial probe lines and the spiral

channels help to identify the non-binding locations in order to measure the

background level. In reality, the probe creation is not conducted in this alternative

fashion. In other words, the probes do not repeat in close proximity. So in Figure

5-6b, every two AD probe lines were spaced by 15 oligonucleotide probes, which

are non-complementary to both FD’ and FD’P. The plots show high signal-to-

background ratios for the hybridizations to the two complementary probe lines

(AD) immobilized on both ends, as compared with the signals on the non-

complementary probes (AN and AB) in between. The fluorescent patches of the

data in Figure 5-6b were also shown in Figure 5-6(c), indicating the high

specificity of hybridizations achieved in the spiral microchannels.

110

5.4 Conclusions

Detection of fungal pathogenic DNA has been achieved at the intersection

between the spiral sample channels and the radial probe lines created with two

independent microfluidic channels plates. Hybridizations using oligonucleotides

and PCR products were tested at different conditions such as probe

concentrations, hybridization times and temperatures. It was found that the PCR

products were detected at the level of 1 µL of 20 nM (or 3 ng) at 45 °C in 2 h;

whereas the oligonucleotides could be detected as low as 0.5 fmol (in 1 µL) at

room temperature in 3 min. The advantages offered by the intersection approach,

such as flexibility in probe line construction, low sample consumption, fast

hybridization rate and multi-sample capability, have been demonstrated on a 100

× 100 microarray because of the ease in liquid delivery offered by centrifugal

pumping. This method will facilitate the fast diagnosis of greenhouse crop

diseases because there are usually many plant samples to handle.

111

6: OPTIMIZATION OF A CD-LIKE MICROFLUIDIC MICROARRAY DEVICE FOR THE FAST DISCRIMINATION OF FUNGAL PATHOGENIC DNA

6.1 Introduction

DNA microarray analysis is generally performed by passive hybridization

of samples on glass slides microspotted with probes. To reduce reagent

consumption and facilitate target diffusion to the substrate surface, microfluidics

has been integrated into the hybridization-based DNA microarray technology

[267]. One approach is to deliver one DNA target sample to the spotted probe

region enclosed within a microchamber [60, 72-74, 76, 86, 87, 106, 108], or a

serpentine microchannel [89, 90, 109]. However, these methods do not deal with

many samples simultaneously. To achieve this goal, another approach is to

deliver DNA targets via many microchannels which intersect with lines of

preprinted probes [84, 91, 93, 94, 96]. Such an approach has previously been

applied to immunoassays by Delamarche et al. and a few other groups [123,

125], and it is also very useful in parallel DNA hybridization, especially when

multiple samples are involved and a high probe density is not needed, such as in

single nucleotide polymorphism (SNP) studies [93, 94, 106]. In these reports, the

sample volume has been reduced down to 1μL, and the hybridization time was

shortened to be < 10 min, and so far up to 16 microchannels were ever used.

112

Parallel hybridizations in multiple sample microchannels greatly demand

an effective way in simultaneous liquid delivery. A common method used in

microfluidics is pressure-driven pumping. However, a high pressure is required

for liquid delivery in long and narrow microchannels, and this in turn requires a

very tight sealing between the microfluidic channel plate and the substrate [90].

An alternative method is to utilize the body force of the liquid column itself, and

such a force can be created under a centrifugal force field. As compared with

other methods, centrifugal pumping is easy to implement and it can move fluids

in a parallel manner in many channels of a wide range of sizes. Even with this

advantage, however, most of centrifugal pumping applications are still limited to

the radial direction [106, 108, 129, 133, 149, 152, 153, 156, 157].

To fully utilize the centrifugal advantage, we have already developed the

microfluidic microarray method for DNA hybridizations with a microchip

consisting of 96 channels [129, 133]. Here, hybridizations occurred in the spiral

microchannels that orthogonally intersected with the radial probe lines pre-

printed on a glass disk and liquid delivery in these channels has been achieved

by centrifugal pumping. The previous work applied mainly to the fast analysis of

oligonucleotide DNA. In this work, we extend the method for the fast detection of

~260-bp PCR products. With the improvements in flow control, microchannel

design and the use of Cy5 dye labels, the detection of less than 0.2-ng PCR

product (in 1 µL) was achieved within 3-min hybridization. This is an

improvement to our previous work using fluorescein as the fluorescent labels, as

113

described in Chapter 5. Moreover, fast single-base-pair discrimination of two

Botrytis species has been achieved with the improved method.

6.2 Experimental

6.2.1 Fabrication of microfluidic devices.

The fabrication of the radial and spiral PDMS channel plates has been

described in Chapter 2 and Chapter 5.

6.2.2 Oligonucelotides and the preparation of fungal pathogenic DNA samples.

The sequences of oligonucelotide probes, previously designed to detect

two greenhouse plant pathogens, Botrytis cinerea (with probe AB) and Didymella

bryoniae (with probe AD), have been listed in Table 3-1 [105]. The sequences of

the primers (Cy5-ITS2 and ITS1F) used for PCR are also shown in Table 6-1. All

oligonucleotides were synthesized and modified by Sigma-Genosys (Oakville,

ON, Canada) or International DNA Technologies (Coralville, IA).

Genomic DNAs of B. cinerea, D. bryoniae and B. squamosa were

provided by Carol Koch from Agriculture and Agri-Food Canada. The amplicons

were generated by PCR amplification of their respective genomic DNA using the

primers (ITS2 and ITS1F) specific to septate fungi, as described previously [105].

Fluorescent dyes were incorporated in PCR products through the Cy5 or

Fluorescein-labelled forward primer (Cy5-ITS2 or Fl-ITS2) during amplification.

After PCR, samples were purified with a DNA purification kit (QIAquick, Qiagen,

Mississauga, ON, Canada) to remove excess primers and nucleotides. The DNA

114

concentrations of the amplicons were measured using a UV-Vis

spectrophotometer (ND-1000, NanoDrop Technologies, Montchanin, DE). The

purified amplicons were named as CB’P (264 bp), CD’P (259 bp), and CBN’P

(264bp), related to the genomic DNA of B. cinerea, D. bryoniae and B.

squamosa. The central sequences of the sense strand of CB’P and CD’P are

complementary to the sequences of the probes AB and AD, respectively. But the

binding between CBN’P and probe AB has one base-pair mismatch as compared

to that between CB’P and probe AB (i.e. TTT:ATA instead of TAT:ATA in the

center of the DNA strand).

Table 6-1 Oligonucleotide primers used in this study

Acronym Length Sequence (5'-3')

Cy5-ITS2 20-mer Cy5-GCT GCG TTC TTC ATC GAT GC

ITS1F 22-mer CTT GGT CAT TTA GAG GAA GTA A

6.2.3 Sample hybridization and fluorescent detection.

As depicted in Chapter 5, the surface of the plain glass disk was

chemically modified and preprinted with radial probe lines [129]. The PDMS

spiral channel plate was then assembled with the disk (Figure 5-2). Afterwards,

the DNA target solutions (0.5-1.0 µL, prepared in 2.5x SSC plus 0.2% SDS) were

added to the inlet reservoirs using a micropipette. In centrifugal-flow method, the

disk assembly was mounted on a rotation platform which was enclosed in a

plastic box with temperature control. By spinning, the sample solutions

115

continuously flowed through the spiral channels and hybridized with the

preprinted radial probe lines. The control of the hybridization time was achieved

through adjusting residence time by using different rotation speeds (1400~3000

rpm). The flow visualization and residence-time measurement during disk

spinning were achieved with the assistance of a stroboscope light (Nova-Strobe

DA Plus, Monarch Instrument, Amherst, NH) [65, 129, 133]. In suction-flow

method, the sample solutions were filled through the microchannels by applying

a negative pressure at the outlet reservoirs. The flow rate in this method was

controlled by adjusting the pressure. As for the stop-flow method, sample

solutions stayed in the microchannels and the disk assembly was kept in an oven

at 42 ºC for 2 h.

In all cases, after hybridizations, there was no further channel washing,

but only whole-disk washing was used. This simplified the operation procedures.

First, the disk assembly was spun at 3600 rpm to dry the channels, and then the

PDMS chip was peeled off. Afterwards, the glass disk was washed briefly with 2x

SSC + 0.2% SDS (1 min.), and with 2x SSC (1 min.), and then dried in a nitrogen

stream. The glass disk was scanned on a laser fluorescent scanner to obtain the

hybridization results as previously described [129]. The excitation wavelength

was 488 nm or 633 nm for fluorescein-labelled or Cy5-labelled samples,

respectively. The photomultiplier tube (PMT) voltage was set to 600V.

116

6.3 Results and discussion

6.3.1 Fast and parallel hybridization using continuous centrifugal flow.

Microfluidic DNA hybridization has fast kinetics and sensitive detection

because continuous replenishment in addition to diffusion is involved in DNA

target transport to the surface-immobilized probes. The dynamic DNA

hybridization behaviours in microchannels have been characterized in numerous

reports [84, 87, 91]. In the previous chapter, 3-ng PCR products were detected

by microfluidic hybridization only after 2-h stop-flow and at 45°C [129]. In this

work, a lower amount of PCR products (CB’P and CD’P) were detected under

two flow conditions, continuous-flow as well as stop-flow. Figure 6-1(a) shows the

hybridization results obtained from the fluorescent image of the test disk. There,

18 samples in seven groups flowed through adjacent spiral microchannels and

intersected orthogonally with 3 preprinted radial AB probe lines. Since CB’P is

complementary to the probe sequence, hybridization occurred and rectangular

patches formed at the intersection areas. On the other hand, the non-

complementary CD’P samples were not retained by immobilized probe AB and

they were washed away immediately in the continuous-flow method. No patch

was seen at the intersection areas on the image. However, in the stop-flow

method, when PCR product samples were incubated in microchannels for a

longer time (2h), non-specific binding of DNA was observed along the spiral

channels shown as strips in the image. We believe that, in the continuous-flow

method, the dynamic sample flow prevented the dye-labelled DNA targets from

accumulating and binding non-specifically onto the glass surface. For a good

117

comparison of signal-to-noise ratio, Figure 6-1(b) depicted various fluorescent

intensities resulted from hybridizations in adjacent spiral channels on the bottom

radial AB probe line shown on Figure 6-1(a). For 3.2-ng PCR products, the

continuous-flow method gave fluorescent signals similar to those from the stop-

flow method, but this was achieved at the benefit of 40-fold decrease in

hybridization time. From Figure 6-1(b), it can also be seen that the background

signals from the continuous-flow method (the baseline of the first six peaks) are

much lower and more homogeneous than those from the stop-flow method

(shown as a series of small bumps). Therefore, the continuous centrifugal flow

method provided short reaction time as well as good signal-to-noise ratios for the

hybridization experiments.

118

Figure 6-1 Comparison between the continuous-flow and stop-flow methods of hybridization. (a) The fluorescent images show the PCR product hybridizations conducted with two different flow methods at 42 ºC. The dark rectangular patches represent the specific binding of the complementary targets. 1.0 μL of PCR products (CB’P or CD’P) were applied to the spiral microchannels. Probe AB was used to print radial probe lines (b) Line graph showing the average fluorescent intensities along the bottom radial probe line intersected with 20 spiral sample channels depicted in (a). (c) Hybridization signal comparison of two continuous-flow methods on same microchannel plate. Here, six groups of oligo probe AB were preprinted and distributed evenly on a glass disk as shown in the inset. Sample solutions (0.2ng CB’P in 2.5xSSC+0.2%SDS), driven by either centrifugal force or vacuum suction, flowed through and intersected with these probe lines. Each bar represents signals from the average of nine hybridization patches. The bar number in the graph matches to the number as shown in the inset, indicating the position of hybridization along the spiral channels. The error bars represented the standard deviation of hybridization signals from the nine patches. The sample volume is 0.8 μL and the rotation speed or vacuum was controlled to ensure an average 3 min of hybridization time. (d) Picture of the broken liquid column (in red) inside a 24-μm depth microchannel during vacuum suction. The inset shows the magnified picture of the area. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd.

The centrifugal-driven method was also compared to the vacuum suction

method for microfluidic DNA hybridization using sample CB’P and the

0

2000

4000

6000

8000

10000

12000

RFU

3-min flow

2h stop- flow(a) (b)

0

50

100

150

200

250

300

1 2 3 4 5 6Different groups of probe lines

RFU

By suctionBy spinning

(d)(c) 1

23

4

5 6

AB

AB

AB

3-min flow 2h stop- flow

6.4ng

CB’P 6.4ng

CB’P 3.2ng

CB’P

3.2 ng

CB’P 6.4ng

CD’P 6.4ng

CD’P

3.2ng

CD’P

119

complementary probe AB. Figure 6-1(c) depicted the comparison of hybridization

signals with two flow methods. Here, it was found in vacuum suction method,

there were more signal variations among adjacent spiral microchannels and

among different positions on each spiral channel. These two variations could be

attributed to two reasons. First, due to the tiny volume of microchannels, a small

variation in the pressure applied at the outlet might cause a considerable change

in liquid flows. Second, fragmentation of sample flows at the end of the channel

was found in the vacuum suction method (Figure 6-1(d)), which resulted in the

variation of hybridization intensities at different positions (1~6 as shown in the

inset of Figure 6-1(c)) along each spiral channel. On the contrary, continuous

centrifugal flow method gave higher signals together with less signal variations.

Considering the relative small errors among repeated experiments, parallel

hybridization in multiple channels could be readily attained on the microfluidic

disk by employing the centrifugal force field.

6.3.2 Enhanced sensitivity by the control of flow, channel depth and temperature.

After confirming that the continuous centrifugal flow produced good

hybridization results, the effect of the rotation speed on hybridization signals was

carefully examined. In centrifugal force driven microflow, the disk rotation speed

affects the flow rate of the liquid inside microchannels. This determines the

residence time of sample solutions which is defined as the net time from the start

of liquid filling in the microchannel to the complete dry-out of the liquid plug. We

predicted that a lower rotation speed (or a higher residence time) would result in

120

stronger hybridization signals. As shown in Figure 6-2(a), the residence time of 1-

μL sample solution decreased from 4.3 min to 1.3 min with the increase of the

rotation speed from 1400 rpm to 2400 rpm (1400 rpm is the lowest rotation speed

below which centrifugal pumping could not be achieved in this experiment). This

corresponded to the decrease in the hybridization signals as the rotation speed

increased. Therefore, it is concluded that, in the continuous centrifugal flow state,

the lowest achievable rotation speed gave the strongest sample hybridization

signal.

Hybridization efficiency obtained in microchannels of different channel

depths was investigated. Here, we compared two depths of microchannels, 24

μm and 75 μm, in two PDMS chips. To prevent any effect resulted from the

difference in volumetric flow rate, the rotation speeds were adjusted to ensure an

equal residence time of 3 min in each case. The resulting hybridization signal-to-

noise ratios for 0.2 and 3.2-ng PCR products were shown in Figure 6-2(b). It was

found that for both high and low concentrations of DNA samples, the shallower

microchannel gave a better sensitivity. The signal-to-noise ratios from the 24-μm

microchannel chip are twice as high as those from the 75-μm microchannel chip.

The reduction of microchannel height enhanced mass transport of target DNA to

the capture probes and thus generated higher hybridization signals. Moreover,

DNA molecules could be stretched in microflows, which offered further

advantages for hybridization efficiency [84, 87, 110].

121

Figure 6-2 Enhanced sensitivity due to smaller channel depth and higher temperature. (a) The bar graph showing the measured fluorescence intensities from the hybridization of 1.6-ng CB’P at different rotation speeds at room temperature. The line graph shows the corresponding residence times. The error bars were calculated from three measurements. (b) The effect of channel depth on hybridization signal intensity. Different amount of PCR products (CD’P) were hybridized at 42°C for 3 min. (c) Fluorescent intensities along an AB probe line. Ten samples (CB’P) were detected on the test disk at different concentrations and each sample was conducted in duplicate. From left to right, five samples (6.4, 3.2, 1.6, 0.8, and 0.2 ng) were first hybridized at room temperature (23°C) for 3 min. After dry-out of these channels, another group of 5 samples was hybridized at 42°C for 3 min. In all cases, the volume of sample solutions was 1.0 μL. The insets showed the magnified graphs of the hybridization signals of 0.2 ng CB’P. (d) Calibration curves of hybridization signals of PCR products (0.2~6.4 ng). The error bars represented the standard deviation of hybridization signals from six patches. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd.

The calibration curve for PCR product detection with the proposed method

was examined at room temperature and at 42°C on the same disk. Figure 6-2(c)

shows the signal variations resulted from various samples hybridized to one

radial probe line on the disk. It was found that the signals obtained at 42°C were

0

30

60

90

120

150

75um-Channel 24um-Channel

Sig

nal

-to-

noi

se r

atio

0.2ng CD'P 3.2 ng CD'P

0

1000

2000

3000

4000

5000

1.4 1.5 1.6 1.8 2 2.4

Spinning speed (*1000rpm)

Fluo

resc

ent i

nten

sity

(RFU

)

0

50

100

150

200

250

300

Res

iden

ce ti

me

(sec

)

RFU

Residence time

0

4000

8000

12000

16000

20000

0 1 2 3 4 5 6 7

Target DNA applied (ng)

RFU

0

4000

8000

12000

16000

20000

0 5 10 15

Distance (mm)

RFU

(a) (b)

(d)

At 23°C: y = 880x + 489

At 42°C: y = 2616x - 26

200

400

600

800

1000

200

400

600

800

1000(c)

Hybridized at 23 °C Hybridized at 42 °C

122

more than twice as high as the signals from the room temperature experiment.

This observation matches with the conventional microarray method, which is

usually performed at higher temperature for better sensitivity [65]. Nevertheless,

with flow hybridization, room temperature test also gave both enough specificity

and low background in a 3-min period, and 0.2-ng PCR products can be

detected. The corresponding calibration graphs from 5 different sample amounts

ranged from 0.2-6.4 ng were also depicted in Figure 6-3(d), showing good

linearity for both room-temperature and 42°C flow hybridization. It is concluded

that the best hybridization is obtained with the 24-μm deep spiral channel plate,

spun at 1400rpm and maintained at 42°C.

6.3.3 Enhanced detection sensitivity due to the use of Cy5 dye labels.

In Chapter 5, we reported that 3.2-ng fluorescein labelled PCR products

(FD’P) were barely detected at 45°C after 10-min hybridization [129]. With the

improvements achieved in this work, 0.2-ng Cy5 labelled PCR products gave a

signal-to-noise ratio of 41 in 3-min flow hybridization at 42°C (Figure 6-2(c)).

Although different hybridization behaviours related to other dye labels of DNA

were discussed in many reports [268-271], no comparison has been made

between the microchip hybridization signals from DNA labelled with Cy5 and

fluorescein dye molecules. In this experiment, the relative hybridization

intensities of DNA with the above mentioned labels were examined. Here single-

strand oligonucleotides complementary to the immobilized probes were used as

targets to rule out any possible effect from the overhanging strand and possible

secondary structure of longer PCR product samples [272]. From the hybridization

123

results in Figure 6-3a, it can be found that the signal-to-noise ratio of 0.1-μM FD’

is around 20 while this number for Cy5 dye labelled samples (CD’) at same

concentration is close to 1000. This phenomenon cannot be explained by

different fluorescent quantum efficiencies because this value of Cy5 dye label in

aqueous buffer is reported as ~0.2 while that of the fluorescein dye label is ~0.6

[273, 274].

To further examine if the presence of fluorescent labels may affect

hybridization efficiency, the Cy5-labelled and fluorescein-labelled samples was

mixed and applied to the microchip to study competitive hybridization. The results

were compared with the signals from solutions containing FD’ or CD’ alone. It

was found that the presence of different labelled targets showed a competitive

effect. In terms of the results scanned at the fluorescein wavelength (532nm), it

can be seen that CD’ in the mixture strongly inhibited the hybridization of FD’, to

a degree up to ~90% in comparison with pure FD’ samples. But the hybridization

signals from the mixture scanned at Cy5 wavelength (670nm) reduced only

~10% from those of pure CD’ samples (Figure 6-3(b)). The former observation

should not be caused by fluorescence resonance energy transfer (FRET)

because, while there was a large drop in the FD’ signal (scanned at 532nm),

there was no corresponding increase in the CD’ signals. The results clearly

showed that the hybridization efficiency of Cy5 dye labelled DNA is much higher

than that of fluorescein labelled DNA, and this could be explained by the strong

stabilizing effects of Cy5 on the duplex DNA [275].

124

Figure 6-3 Enhanced hybridization signals of Cy5-labelled oligonucleotides due to higher hybridization efficiency and less photobleaching. (a) signal-to-noise ratio from the hybridization of different labelled oligonucleotides with the same probe lines. The results of fluorescein-labelled DNA are expanded to give the right inset picture. Here, the concentration of both oligonucleotide samples is 0.1 μM. (b) Comparison of hybridization efficiency between Cy5-labelled and fluorescein-labelled oligonucleotides. The left two bars show the normalized fluorescent intensities from the hybridization of CD’ or CD’ + FD’ (scanned at 670nm). The right two bars show the normalized fluorescent intensities from the hybridization of FD’ or CD’ + FD’ (scanned at 532 nm). In all the cases, the final concentrations of different labelled oligonucleotides are 1 µM. (c) Photobleaching effect on different dye labelled oligonucleotides. Normalized fluorescent intensity from the chip with hybridized oligos at the specific wavelength of fluorescein dye. In all the cases, 1 µL of 0.1-µM oligo samples labelled with Cy5 or fluorescein dyes were applied and hybridized at room temperature for 45min in dark place. After removal of the channel plate, the glass chip was scanned consecutively for three times and it was then exposed to room light for 4 h before it was scanned for the fourth time. The error bars here represent the standard deviations of the hybridization signals of six patches. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd.

To examine if the photobleaching effect of the fluorescein label plays a

role, repeated scans were performed on the hybridization patches, and the mean

signal intensities were plotted as percentage of the signal intensity of the first

0

200

400

600

800

1000

1200

Sign

al-to

-noi

se ra

tio

Oligo-Cy5 Oligo-FL

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1st scan 2nd scan 3rd scan After 4hexposure

Nor

mal

ized

fluo

rens

cenc

e in

tens

ity

Oligo-Cy5, scanned at 670 nm

Oligo-Fl, scanned at 532 nm

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Scan at 670 nm Scan at 532 nm

Nor

mal

ized

fluo

rens

cenc

e in

tens

ity

(a

(b

(c)

Oligo-Cy5 only Oligo-FL only Oligo-Cy5 + oligo-FL (1:1)

0

10

20

800

1000

1200

S/N

ratio

125

scan. As illustrated in Figure 6-3(c), three consecutively repeated scans showed

no noticeable bleaching effect for both Cy5 and fluorescein labelled DNA since

fluorescent hybridization patches were subjected to laser exposure for less than

15 s. However, after exposing the hybridized chips under room light for 4h, there

was differential photobleaching effect: only 5% loss in the signals of Cy5-labelled

DNA, but 31% loss of the signals of fluorescein-labelled DNA. The results

demonstrated the superior signal stability of Cy5 dye over fluorescein dye labels.

In summary, an enhanced detection sensitivity with greater hybridization signals

were obtained due to higher hybridization efficiency and less photobleaching

effect of the Cy5 dye label.

6.3.4 Discrimination of PCR products with single base-pair difference.

With the optimization in flow and channel design, the specificity of the

microfluidic method was exploited further for the discrimination between two

related Botrytis species, B. cinerea and B. squamosa. The two PCR amplicons

differ in only one base pair in the middle of the 264-bp long sequence. In this

manner, they hybridized with the probe AB to produce two types of duplexes,

namely the perfectly matched (PM) duplex from CB’P and the mismatched (MM)

duplex from CBN’P. Because the latter duplex will have a lower melting

temperature, discrimination can be achieved by applying thermal stringency to

denature MM, but not PM. This is usually performed in the classical microarray

method for single-base-pair discrimination [90, 276]. Here, the hybridization

efficiency at four hybridization temperatures was consecutively tested on the

same disk. Figure 6-4 depicted the images and signal intensities of the

126

fluorescent patches resulted from the hybridization of samples in 3 spiral

channels to 3 preprinted radial probe lines. Although hybridization signals were

quite strong at 23°C, discrimination was poor. The PM signal reached a high

value at 42°C, and it was almost 2 times higher than that at 23°C (room

temperature). Thereafter, the PM signal was reduced at higher temperatures. As

for the mismatched duplexes (MM), the signal intensity kept decreasing with the

increase of hybridization temperature. The discrimination ratios between the PM

and MM duplexes, which were calculated by dividing the PM signals over the MM

signals, were plotted in Figure 6-4(b). It was found that hybridization at 62°C

gave the best differentiation with a ratio of ~14. As shown in the rightmost image

of Figure 6-4(a), the MM signals at 62°C are the lowest, and are comparable with

the background, indicating the best discrimination. Therefore, higher temperature

improves discrimination at the expense of signal intensity. In view of sufficient

hybridization intensity and good discrimination ratio, 52°C was chosen as

optimum.

127

Figure 6-4 Differentiation of Botrytis species with single-base-pair difference at various hybridization temperatures. (a) 4 groups of fluorescent images showing 9 patches of matched duplex and 9 patches of mismatched duplex. Each image was obtained from the hybridization of sample solutions in 3 spiral channels intersecting with 3 probe lines at the specified temperature. Two PCR products, perfectly matched CB’P (PM) and mismatched CBN’P (MM), were tested in the experiment. (b) Bar graphs showing the fluorescent intensities from the hybridization patches in (a). The error bars represent the standard deviations of the hybridization signals of nine patches. The connected line shows the discrimination ratio between the PM and MM species. In these discrimination experiments, 1.0-μL PCR products (1.6-ng CB’P or CBN’P) were applied in triplicate to the inlet reservoirs of 6 spiral channels Then the sample solutions flowed into the spiral channels by spun at 1500 rpm for 3 min first at 23 ºC (room temperature). After this operation, the channels were dried out at 3600 rpm and the temperature of the assembly was raised to 42 ºC. Then another batch of sample solutions was applied to another 6 spiral channels on the same chip and the process was repeated. The procedure was repeated similarly for hybridization at 52 ºC and 62 ºC. Reprinted with permission from [134], Copyright © 2010 Elsevier Ltd.

0

1000

2000

3000

4000

5000

6000

23ºC 42ºC 52ºC 62ºC

Hybridization temperature

Sig

nal

in

ten

isty

(R

FU

)

0

3

6

9

12

15

Dis

crim

inat

ion

rat

io

PM

MM

Ratio

(a) (b)

23ºC 42ºC 52ºC 62ºC

PM MM PM MM PM MM PM MM

128

6.4 Conclusion

A microfluidic CD-like chip with spiral microchannels was further studied

for the detection of pathogen PCR samples. The centrifugal pumping method

used in this paper showed advantages over vacuum suction flows in parallel

solution delivery and signal variations between duplicated samples. The sample

residence time in the microchannel was predictable at a certain rotation speed.

The effect of microchannel depth was studied and enhanced sensitivity was

achieved with the use of shallower spiral channel plate (24-μm depth). Cy5 dye

labels were proved to show much higher hybridization efficiency as well as less

photobleaching effect, as compared with the fluorescein dye labels used in the

previous work. The calibration curves of PCR product detection at two

temperatures were constructed with a sample load from 0.2 to 6.4ng (in 1 μL).

With optimized conditions, the discrimination between two 260-bp PCR products

with one base-pair difference was achieved in duration as short as 3 min. The

microfluidic chip has the capacity to analyze up to 100 samples per chip

simultaneously.

129

7: ANALYSIS AND MODELING OF FLOW IN ROTATING SPIRAL MICROCHANNELS: TOWARDS MATH-AIDED DESIGN OF MICROFLUIDIC SYSTEMS USING CENTRIFUGAL PUMPING

* Section 7.3 and Appendix A were written based on the work of Dr.

Kropinski from the Department of Mathematics with the aid of my

experimental data.

7.1 Introduction

Centrifugal pumping has been demonstrated to be a convenient liquid-

driven method in microfluidic applications [180]. Unlike pressure-driven or

electrokinetic techniques, the method avoids the use of tubing or electrical

connections, and parallel sample transport can be easily achieved. Centrifugal

pumping has been successfully applied to immunoassays [156, 160, 277], multi-

ion analysis [149], protein crystallization screening [278], DNA hybridization [106,

108, 133], as well as blood separation and analysis [155, 159, 163]. By using

different microstructure designs, liquid flow and mixing can be controlled in

rotating microchannels. Recent microfluidic applications integrated with

centrifugal pumping method were summarized in many reviews [147, 148, 279,

280]. Although the approximate equations for flow rate estimation have been

given in these reports, more rigorous flow dynamics studies have been pursued

130

for the better understanding of liquid flows in rotating radial microchannels [136,

169, 171, 173-177, 281]. For example, analytical solutions were proposed for

centrifugal-force-driven filling flows in either circular or rectangular microchannels

[136, 173-175]. With commercial CFD tools, numerical solutions have also been

proposed and have been compared to experimental data [171, 176, 177].

Currently, almost all of the fluidic channel patterns used for centrifugal-

pumping microflow are designed in the radial orientation. This radial design has

several limitations: (i) the dimension of fluidic structure in radial direction is

restricted. The centrifugal microfluidic platform usually aims to fit the well-

developed compact disk (CD) system to utilize the signal reading and rotor

driven/control system [145-148, 282]. If a disk of standard CD size (120 mm in

diameter) is used, the maximal dimension in radial direction is ~5 cm. Therefore,

the applicable space is limited considering that no reversed flow design is

allowed in centrifugal pumping method. (ii) The centrifugal force increases with

the increase of radius. As characterized in the work of Duffy et al. [156], for a

liquid flowing into an empty radial channel from a reservoir, the average velocity

of the flow (U) is given by:

2 2

32Hd r rU Lρω

µ∆=

where the determinants include fluid properties (ρ: density; μ: viscosity),

the rotation rate (ω: angular velocity), configuration of the channels (dH: hydraulic

diameter; L: channel length), and the location of the liquid column ( r is the

average distance from the liquid column to the rotation center, and Δr is the

131

radial extent of the liquid column subject to centrifugal force). Therefore, the

average velocity of the liquid is not only related to the rotation speed, but is also

affected by the further extension of the liquid column in a microchannel. Such an

increasing flow velocity puts forward challenges and difficulties in the design of

fluidic structure especially near the rim of a disk. (iii) At last, when extra

pressures such as capillary pressure exists, it is hard to get a balanced flow for

both spatial and temporal control of flow with radially-distributed patterns. The

capillary effect is an important factor during the liquid flow process in short and

narrow channels [283, 284]. In many practical applications, microchannels are

made from polymeric materials such as polydimethylsiloxane (PDMS) or

polymethylmethacrylate (PMMA). The contact angle of water on these polymers

can be more than 90°, which generates a capillary barrier pressure (Pcb) at the

liquid front for aqueous based solutions. On the other hand, the centrifugal

pressure acting on the liquid in radial channels is given by [147]:

2mP r rρω= ∆

Liquid flow will not be initiated unless the pressure at the meniscus of

liquid front (Pm) is larger than the capillary barrier pressure (Pcb). Once the

rotation speed (ω) reaches the critical point, the liquid plug in microchannels

starts to move forward. However, with the advancing of the liquid front, Δr

increases and so does Pm. The balance between centrifugal driven force and

capillary barrier is then broken and the flow speed starts varying. A flow pumping

force which is not related to the length of the radial extent is thus needed.

132

In our previous work, we have designed an equiforce spiral microchannel

on a CD-like microchip and a constant centrifugal force component was

produced along the microchannel [133, 285]. Parallel transport and detection of

nucleic acid samples was performed for 3 min in spiral microchannels under the

centrifugal force [129, 133, 134]. From experimental observations, it was found

that the liquid flow driven by the centrifugal force in spiral microchannels was not

affected by the radial extent of the liquid front. The flow velocity is mainly

determined by the rotation speed of the disk assembly. In this work, we carried

out a detailed study of centrifugal-driven flows inside the equiforce spiral

microchannels. The experimental measurement of the microflow at different

rotation rates was performed by analyzing the digital-video recording of the liquid

flow. Then, the Navier-Stokes equations of incompressible flow were rewritten in

a new orthogonal curvilinear-coordinate system based on the spiral design. An

approximate analytical solution to the Navier-stokes equations was deduced and

fitted to the experimental data. In addition, we investigated the centrifugal flow

behaviours in spiral microchannels by carrying out three-dimensional numerical

simulations. The results were compared to the experimental observations and

the mathematical modeling solutions.

133

7.2 Chip design, microfabrication and flow measurement of the spiral microchannels

The layout of the spiral channel has been reported in previous chapters.

The fabrication of the PDMS channel plates was based on the soft

photolithography technique which has been reported in Chapter 2 [94, 129]. The

surface profile of the microchannels was obtained by using a profilometer

(Alphastep 500, Tencor Instruments, Mountain View, CA) and the measured

channel height and width are 33 μm and 90 μm, respectively.

The liquid flow in the spiral channels was visualized using aqueous

solutions colored with food dye. As shown in Figure 7-1a, a piece of white paper

preprinted with position lines was put underneath the glass disk, which was used

as a background for easy observation of the liquid front movement. The flow

visualization during disk spinning was achieved with the assistance of a digital

video recorder and a stroboscope light (Nova-Strobe DA Plus, Monarch

Instrument, Amherst, NH). The displacement of the front of the colored liquid

column can be measured by calculating the number of lines passed on the paper

background (Figure 7-1b). Time data were extracted from each frame of the

movie which was recorded at 30 frames per second.

Contact angles of solutions on the microchip surfaces were measured with

an OCA 15 Plus contact angle microscope (Dataphysics Instruments GmbH,

Filderstadt, Germany). The liquid drop images were taken with a charge coupled

device (CCD) camera. The liquid drop shape was snapped from the images to

134

infer the contact angles with suitable software (SCA 20 from Dataphysics

Instruments GmbH).

Figure 7-1 (a) The photograph of the equiforce spiral channel assembly before spinning. One inlet reservoir was filled with colored solutions and a piece of paper printed with grey position lines was placed underneath to facilitate the measurement of the liquid front position. (b) Magnified picture in (a) shows the image of inlet reservoirs and a microchannel partially-filled with red dye solution. Reproduced from [286] by permission of The Royal Society of Chemistry.

7.3 The mathematical model

The layout of the equiforce spiral channel has been reported in previous

work [285]. The spiral curve was designed so that the angle α between the radial

direction and the tangent (see Figure 7-2a) satisfies the following equation

cosr kα = (1)

where k is defined as the degree of spirality ( cosk r α= ) and it is a

constant. The value of k is determined by the inlet and outlet positions of the

spiral channel. In our case, the microchannel is confined within a 92-mm

diameter plate and the beginning and the end of the spiral curve are set to be 21

1cm

(a) (b)

135

mm and 42 mm from the center of the plate, respectively. Therefore, the value of

k equals to 3.32 mm for a 200-mm long spiral channel, which is calculated by the

initial r value of 21 mm and initial α value of 80.9001°. This k value is the same

for the final r of 42 mm and final α of 85.4644°. This geometry ensures that the

along-channel component of the centrifugal acceleration is constant.

The motion of an incompressible fluid in a centrifugally driven flow is

governed by the Navier-Stokes equations. Written in a frame of reference

rotating with the disk, these equations are:

21( ) ( ) 2 pt

νρ

∂+ ⋅∇ +Ω× Ω× + Ω× = − ∇ + ∇

∂u u u R u u (2)

together with the continuity equation

0∇⋅ =u (3)

for the fluid velocity u, time t and the pressure p. The constants ρ and ν

are the fluid density and kinematic viscosity, respectively. In this frame of

reference, the boundary conditions for the fluid velocity are the usual “no-slip”

boundary conditions. The term Ω × (Ω × R) is the centrifugal acceleration at

angular velocity Ω = ωk, R is the position within the channel, and 2Ω × u is the

Coriolis term.

In order to obtain a mathematically tractable physical model, we proceed

under the following assumptions:

1) The Coriolis force 2Ω×u is assumed negligible in comparison to

the centrifugal force.

136

2) Any surface tension effects are neglected.

3) The external pressure differences between the inlet and the outlet

boundaries are assumed to be zero.

4) The liquid front is advanced under a quasi-steady approximation;

i.e. / 0t∂ ∂ =u .

5) The flow is unidirectional and it is only along the s direction.

6) The microchannel has a short, radial “neck” which joins the inlet

reservoir to the entrance of the spiral channel. This neck region is

assumed to play a negligible role in the fluid motion within the spiral

portion of the channel. The role of this region will be investigated in

future studies.

Ducrée et al. carefully investigate assumptions 1 and 2, and determine

that these assumptions are valid for moderate rotation speeds [177]. Considering

that both the inlet and outlet reservoirs are open to air, assumption 3 is

reasonable, and this assumption has also been used in the work from other

groups [174, 177]. In terms of assumption 4, it is valid in the case of thin-film

flow, [287] and Kim et al. have further proved its validity for the flow in a rotating

rectangular microchannel [174]. The assumption 5 assumes that the transversal

channel flow, un and uz are negligible compared to the down-channel flow us and

they are thus equal to 0. The assumption 5 has been verified through numerical

simulation by CFD program as shown in Appendix B.

137

Before proceeding with the reduction of the Navier-Stokes equations

under the above assumptions, we introduce a right-handed orthogonal curvilinear

coordinate system (n, s, z), with unit vectors en, es, and k, respectively, that are

aligned with the spiral portion of the microchannel (Figure 7-2). The s-coordinate

measures the distance along the spiral, where s = 0 corresponds to the start of

the spiral, and s = l(t) is the position of the liquid front, n is the across-channel

distance, 0 ≤ n ≤ W, and z the channel height, 0 ≤ z ≤ H. Details of this

coordinate system and the reformulation of the Navier-Stokes equations for a

fluid velocity u = unen + uses +uzk and pressure p(n, s, z) are given in Appendix

A.

Figure 7-2 (a) The layout of the equiforce spiral channel and the orthogonal curvilinear coordinate system used for the mathematical modeling of the liquid flow inside the spiral microchannel. (b) 3-D magnified view of the spiral microchannel in the curvilinear coordinate system. H and W are channel height and width, respectively. Reproduced from [286] by permission of The Royal Society of Chemistry.

Under the assumption of a unidirectional flow only along the s direction, un

= uz = 0, which has been validated in narrow channels by the CFD simulation

(a) (b)

138

(Results shown in Appendix B), the the s-momentum Navier-Stokes equation

reduces to (Details can be found in Appendix A):

( )2 1,Ts

pk u n zs

ω νρ∂

− ∆ = −∂

. (4)

where the operator ΔT is,

2 2

2 2T

n z∂ ∂

∆ = +∂ ∂

,

Integrating equation (4) with respect to s gives

( ) [ ]20

1, ( , )Tsk u n z s p n s Pω ν

ρ − ∆ = − − ,

where P0 is the pressure at the inlet of the spiral channel. Neglecting any

external pressure difference between the inlet and the outlet implies

that ( )( ) 0, p l t n P= , and the only way this is satisfied is to require that

( ) 2, , 0 , 0Tsu n z k n W z Hν ω∆ = < < < < , (5)

which we solve together with the no-slip boundary conditions

( ) ( ) ( ) ( )0, , 0, , 0 , 0.s s s su z u W z u n u n H= = = = (6)

A global mass conservation argument determines that the velocity of the

liquid front must be equal to the average of us across the channel. Thus,

( )0 0

1 , H W

sdl u n z dndzdt WH

= ∫ ∫ (7)

139

Before solving equation (7), we seek the solution to (5) satisfying

boundary conditions (6) in the form of an eigenfunction expansion. This yields

( )2

21,3,5, 1,3,5,

16 1, sin sinsp q pq

k p qu n z n zpq W H

ω π ππ ν λ

∞ ∞

= =

=

∑ ∑

. (8)

where

2 2

pqp qW Hπ πλ = +

.

Substituting (8) into (7) and integrating yields

2

4 2 21,3,5, 1,3,5,

64 1 p q pq

dl kdt p q

ωπ ν λ

∞ ∞

= =

= ∑ ∑

. (9)

This equation provides the relation to construct the plot of the liquid

meniscus displacement (l) versus time (t) in order to fit the experimental data.

7.4 Numerical simulation of the microflow

Numerical simulations of the centrifugal filling flow in the spiral

microchannel were studied using an ESI-CFD package (Advanced CFD 2009,

ESI North America, Bloomfield Hills, MI). The dimensions and the geometry of

the 3-D models are shown in Figure 7-3. The geometry was generated from the

spiral pattern used in the microfabrication of the PDMS microchannels. The

channel cross-section is 90 μm (width) × 33 μm (depth), which were measured

from experimental observations. The 3-D geometry was computationally

discretized into structured quadratic elements as shown in Figure 7-3(b) and (c).

A grid with 35, 970 cells was used in the transient flow simulations. Here, the

140

model of the microchannel is comprised of three parts: an inlet reservoir of 1-mm

diameter, 2-mm long straight neck channel from the inlet, and the 200-mm long

spiral channel. The inlet starts at a distance of 19 mm from the center of rotation.

The simulations were run on a Dell T7400 workstation with 4 GB of RAM and

Intel Xeon processors; each simulation typically required ~1 week of computer

execution time. Details on the governing equations, parameter settings, and

simulation procedures are given in Appendix B.

Figure 7-3 (a) A bird’s-eye view of the 3D geometry of the equiforce spiral microchannel used in CFD simulations. (b) Magnified view of the inlet region with a tilted inlet reservoir and a neck channel. Point A denotes the starting position of the spiral channel where the neck channel connects. (c) Magnified view shows the structured grid of the microchannel near point A. The cell widths range from 15 µm to 75 µm. The grid lines are as close to 90° as possible and small cell aspect ratios were employed to ensure a good modeling quality. Reproduced from [286] by permission of The Royal Society of Chemistry.

Neck

Spiral

Inlet (a) (b)

(c)

Inlet

A

A

O

Disk center

141

7.5 Results and discussions

7.5.1 Characterization of microflows in equiforce spiral channels with experimental data

Liquid flows inside the rotating spiral channels were imaged by using a

stroboscope imaging system. Figure 7-4 depicts the images of the movement of

the liquid column inside a spiral microchannel. From Figure 7-4 (a) to (c), three

snapshot images of the disk in motion were captured, showing that the liquid

front moved from the 11th line then passing the 15th line to the 22nd line under

centrifugal force. From these images, the displacement of the liquid front was

measured, and the data was plotted versus time (Figure 7-4 (d) to (f)). The data

was obtained from three of the spiral channels in the same disk rotating at 1500

rpm. It was found that the filling flow is steady through the whole 200-mm

channel with a constant slope under a certain rotation speed. This is unlike the

centrifugal-pumping flow in radial microchannels, where the flow speed also

depends on the radial distance between the liquid front and the rotation center.

With our spiral microchannel design, the radial length of the liquid column has

negligible effect on the overall flow speed. The flow speed, which could be

deduced from the slope of the displacement curves, was found to be 5.10 ± 0.03

mm/s. For comparison, the average flow speed in a radial microchannel with

similar dimension and rotation speed was calculated to be ~44 mm/s [147].

The start length of the liquid front during measurement is not zero. This is

because it takes a couple of seconds for the disk to be accelerated from the

stationary state to reach the set rotation speed in our system and the liquid has

flowed into the microchannels for a few millimeters before a stable strobescopic

142

image can be obtained for analysis. Moreover, pipette pressure at inlets during

the sample loading procedure varies and the start length could be very different

for different measurements even at same rotation speed. In Figure 7-4 (d) to (f),

although the start lengths are 7.8 mm, 9.3 mm, and 12.6 mm, respectively, the

liquid front velocities are 5.1 mm/s in all cases. The coefficient of variation of the

experimental flow velocities measured from three microchannels is less than 1%.

For comparison, flow data from measuring the moving liquid rear was also

shown in Figure 7-4 (d) to (f). In our experiment, the total volume of a 90μm (W)

×33μm (H) ×200mm (L) microchannel is 0.5 μL. After a solution of this amount

fills through the microchannel, the inlet reservoir starts emptying out under the

centrifugal force. To test any change in flow speed during liquid filling and

emptying processes, the distance of the receding liquid rear was also measured

with the similar method and was depicted in the figures. From the linear fit of the

displacement-to-time graph it is concluded that the liquid receding flow is also

steady. Moreover, by comparing liquid filling and emptying curves, we see that

the two slopes are similar to each other and the coefficient of variation in

calculated flow rates is less than 10%. Therefore, the speed of liquid filling

process can represent steady flow in our spiral microchannels.

143

Figure 7-4 Liquid movement inside microchannels. (a), (b) and (c) show the 3 snapshot images of the disk spun at 1500 rpm, where 1 and 2 designate the 10th and 20th position lines, respectively. The black arrows denote the inlet reservoirs and the white arrows indicate the liquid front. The spiral trace extended as time increased, indicating the movement of the liquid front along the microchannels. (d), (e) and (f) represent the displacement-to-time plots showing the movement of both the liquid front and liquid rear meniscus in 3 adjacent spiral channels. The start lengths of the liquid front measurement are 7.8 mm, 9.3 mm, and 12.6 mm, respectively. Reproduced from [286] by permission of The Royal Society of Chemistry.

7.5.2 Comparison of solutions from the mathematical model with the experimental data.

The mathematical model was validated with the experimental data. By

using the measured parameters, equation (9) was plotted as a distance (l) vs.

time (t) curve. The comparison between the mathematical modeling solutions

and experimental data of liquid front movement at 6 different rotation speeds are

depicted in Figure 7-5(a). A fairly good agreement between approximate

theoretical model predictions and the experimental findings was observed at

rotation speeds ranging from 1400 to 2400 rpm. It is revealed that the deviations

0

40

80

120

160

200

0 20 40 60 80 100

Men

iscu

s di

spla

cem

ent

(mm

)

Time (seconds)

0

40

80

120

160

200

0 20 40 60 80 100

Men

iscu

s di

spla

cem

ent

(mm

)

Time (seconds)

0

40

80

120

160

200

0 20 40 60 80 100

Men

iscu

s di

spla

cem

ent

(mm

)

Time (seconds)

(a) (b) (c)

(d) (e) (f) Liquid front

Liquid rear

Liquid front

Liquid rear

Liquid front

Liquid rear

144

between the experimental and mathematical results appear to be very small at

lower rotation speeds. However, at higher rotation speeds, the velocity of liquid

front was slightly faster than that predicted by the mathematical model.

The mathematical model has also been used to evaluate the effect of the

variation in channel width and depths on the liquid flow behaviour. Currently,

microfluidic devices are usually fabricated using photolithograph techniques [193]

and the uniformity of microchannel width could be limited by many factors, such

as photomask resolution and the wavelength of UV light used in

photolithography. Figure 7-5(b) shows microscope images of two PDMS

microchannel plates. It can be seen that the channel shape varies and the width

deviation is as high as 10 % in the channel plate. In addition, depending on the

quality of the spin-coating and subsequent channel-developing processes, the

depth of microchannels also fluctuates. With the aid of the mathematical model,

the effect of the variation of channel dimensions on the centrifugal-driven flow

inside equiforce spiral channels is shown in Figure 7-5c. With a 10% variation in

channel width, the flow speed only changed by less than 3%. In contrast, the

variation of channel depth has much more influence on the flow. Up to a 16%

increase of flow speed was found with a 10% deeper channels.

The proposed model could help in the design of microchips with

equiforce microchannels. For instance, if a chip with a different spiral pattern is to

be designed, the mathematical model could offer a quick and accurate prediction

on the flow behaviour. We use this model to examine the flow behaviour of 2

different equiforce microchannel patterns used in our work (Figure 7-5(d)) [65].

145

Although a narrower microchannel (25 μm in height and 50 μm in width) was

used, the larger k value counteracted the effect of the narrow geometry, and the

liquid front velocity was in fact higher than that of wider and deeper channels.

Figure 7-5 (a) Liquid front displacement versus time at different rotation speeds. Data points represent experimental data and the solid lines are theoretical curves plotted using the mathematical model (The curves and data points were displaced horizontally for clarity). (b) Microscope images of 3 channels in a PDMS chip showing the width variation. The scale bar represents 100 μm. (c) Calculated variation of flow velocity as affected by the variations in channel width and depth using the proposed mathematical model. (d) Liquid front displacement versus time graph from two spiral-microchannel chips of different channel geometries. Data points represent experimental data and the solid lines are from the model prediction. Here, W, H, and U are channel width, channel height, and measured liquid-front velocity, respectively. Both microchips were rotated at 1500rpm. Reproduced from [286] by permission of The Royal Society of Chemistry.

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

-10% -5% 0% 5% 10%

Dimensional variations

Flow

spe

ed v

aria

tions

Channel widthChannel depth

020406080

100120140160180200

1400 rpm 1500 rpm 1600 rpm 1800 rpm 2000 rpm 2400 rpm

Liqu

id fr

ont (

mm

)

Time (Sec)

0

20

40

60

80

100

Time (Sec)

Liqu

id fr

ont (

mm

)

(a) (b)

(c)

10 sec 100 µm

5 sec

k = 3.3 mm k = 7.0 mm Math model

W: 50 μm H: 25 μm U: 6.8 mm/s

W: 90 μm H: 33 μm U: 5.1 mm/s

(d)

146

7.5.3 Validation of the assumptions in the mathematical modeling with the experimental data and simulation results

During the mathematical modeling, we employed 5 assumptions to

simplify the Navier-Stokes equations. Although the model shows reasonable

agreement with experimental data, further studies were performed to examine

the validity of these assumptions. Due to instrumental limitations, the resolution

of time could not be finer than 0.4 sec and the liquid front displacement could not

be measured accurately for a difference within 1.5mm. Moreover, the pressure

inside the microchannels could not be measured. Therefore, numerical studies

with commercial CFD program were used to visualize the transient flow filling

process during the initial stage. Glatzel et al. evaluated the performance of four

commercial programs for different microfluidic applications including centrifugal

microflows [171]. The authors recommended ESI-CFD for the simulation of free

surface flows involving capillary forces [171]. We adopted to use this program

which has also been used by many groups in the numerical studies of steady or

transient microflows with validation from experimental data [176, 177, 284].

7.5.3.1 The quasi-steady state approximation

In this paper we present a quasi-steady state approach for mathematical

modeling the flow-pattern. Here, it is assumed that u does not depend on time.

From the linear fit of the liquid front advancing curves in Figure 7-5(a), it is

already observed that the flow is steady during liquid filling through the 200-mm

spiral channel. Numerical studies were performed to find out how long it will take

for the transient filling flow to reach such a steady state. With the CFD program,

transient filling processes during the first 1.5 seconds were simulated. Here, the

147

inlet reservoir was prefilled with 0.5-μL liquid as what we usually did during the

experiments (Figure 7-6(a)). Because liquid in the reservoir usually crawled into

the spiral channel due to the pipette loading pressure, a small part of the

microchannel was also filled with liquid column as initial conditions in the

simulation (Figure 7-6(a)). This is the start length of the liquid column which

varies in different simulation studies to mimic the real experiments.

The simulation results of the speed of the liquid front at corresponding

times were collected and plotted in Figure 7-6b. It is found that the transient filling

speeds are high at the beginning. The shorter is the start length, the higher is the

initial flow speed. However, the liquid front movement quickly slows down and

reaches a quasi-steady state, shown as the flat part in Figure 7-6(b). This

situation can be considered as a steady flow because the flow velocity variation

is small. As shown in Figure 7-6(c), the flow velocities at 1.5 seconds differ from

each other by less than 5%. Moreover, these flow speed values are very close to

the experimental data run at the same rotation speed (Figure 7-6(c)). Therefore,

the quasi-steady state approximation in our mathematical modeling is reasonable

and valid except during initial stage, which is highly transient. The different start

length of liquid column in the spiral microchannels has a negligible effect on the

filling flow speed during most of the time.

148

Figure 7-6 (a) Image showing the initial conditions used for transient flow simulation. The blue color represents the part of the channel filled with air and the pink color denotes liquid solutions. (b) Simulated liquid front displacement versus time at 1500 rpm with different start lengths. (c) comparison of the liquid front velocities at 1500 rpm with experimental data, mathematical modeling, simulation of transient filling flow and simulation of steady flow. Reproduced from [286] by permission of The Royal Society of Chemistry.

7.5.3.2 The approximation of the negligible Coriolis force.

The Coriolis force is a fictitious force in a non-inertial rotation reference

frame. Since the force is perpendicular to the flow direction, it has been utilized to

induce secondary flows at high rotation speeds for applications in flow switching

or passive mixing [136, 172, 177]. In the mathematical model, the motion of an

incompressible fluid in a centrifugally driven flow is governed by the Navier-

Stokes equations under a rotation reference frame. The Coriolis term is

neglected to simplify analytical solutions, and this approximation has been used

0

2

4

6

8

10

12

14

0 0.3 0.6 0.9 1.2 1.5

Time (sec)

Liq

uid

fro

nt

velo

city

(m

m/

s)

0 mm5 mm7.8 mm9.3 mm12.6 mm

(a) (b)

Start length

Neck length

0

2

4

6

8

Experimentdata

Mathmodeling

Transientflow

simulation

Steady flowsimulation

Liqu

id fr

ont v

eloc

ity (m

m/s

)

(c)

149

by many mathematical studies on the flow velocity in radial microchannels at a

lower rotation speed [136, 173, 177].

To examine the effect of the Coriolis force on the flow in equiforce spiral

channels, both experimental data and simulation results were collected at two

rotation directions, anti-clockwise and clockwise directions. As the sign symbol of

the Coriolis force term in equation (1) depends on the direction of rotation, the

difference between the flow velocities under two different conditions should be

notable if the secondary flow induced by the Coriolis force is not negligible. Here,

a high rotation speed, 2400rpm, was used throughout the experiments and

simulations to demonstrate the possible effect of Coriolis force. The results were

shown in Figure 7-7. From the experimental data in Figure 7-7(a), it is found that

the rotation direction showed negligible effect on flow velocities; the liquid front

velocities at two rotation directions differ from each other by less than 2% (Figure

7-7(a) inset). The evaluation of the Coriolis effect with simulation results was

shown in Figure 7-7(b). Since the start length of the liquid column was set to be

same value as found in experimental observations, the unsteady part of the liquid

front velocity is very short and less than 0.1 s. Furthermore, the liquid front

velocities (in mm/s) at the quasi-steady state for both direction 11.3 ± 0.5, 11.2 ±

0.7, respectively, which are not significantly different. The results support the

assumption of the negligible Coriolis effect in our mathematical model.

150

Figure 7-7 (a) Liquid front displacement versus time from experimental data at 2400 rpm. The inset shows the liquid front velocities from fitting the experimental data in both cases. Because a longer acceleration time was required to achieve a higher rotation speed, the start length is much longer, which is around 50 mm. (b) Simulated results of the liquid front velocities at 2400 rpm from anti-clockwise and clockwise directions. Reproduced from [286] by permission of The Royal Society of Chemistry.

7.5.3.3 Scaling the effect of the surface tension force.

The effect of surface tension force on microflows has been studied

extensively [283, 284, 288-290]. In rectangular microchannels, the surface

tension force per unit mass can be deduced from the Young-Laplace equation

and is written as [177, 289]:

2 cos 1 1ss

PFl l W H

γ θ = = +

.

where Ps is the pressure from surface tension force at liquid front, l is the

length of the liquid column, γ is the surface tension of the liquid, θ is the contact

angle, W and H denote the width and depth of the channel, respectively. On the

other hand, because the centrifugal force in this work has been designed to be a

constant along the spiral microchannel, we thus obtain:

15 20 25 30 35 40 45 50 55 6040

80

120

160

200

Anti-clockwise Clockwise

Liqu

id fr

ont (

mm

)

Time (Sec)

(a) (b)

0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.000

0.004

0.008

0.012

0.016

0.020 Anti-clockwise Clockwise

Velo

city

(m/s

)

Time (Sec)

0

4

8

12

16

20

Anti-clockwise

Clockwise

Liqu

id fr

ont v

eloc

ity (m

m/s

)

151

2F kω ρω= .

where k is a constant and is defined in equation (1). The effect of the

surface tension force could be evaluated using a dimensionless ratio of the

centrifugal force (Fω) to the surface tension force (Fs):

( )2

2 coss

F lWHF W Hω ρω γ

γ θ=

+.

The ratio turns out to depend on ω, l and θ. Figure 7-8(a) shows the graph

of Fω and Fs versus different liquid column length and contact angles at a rotation

speed of 1500 rpm. The centrifugal force is constant along the spiral

microchannel and it is shown as a flat plane in the graph. However, the effect of

the surface tension force on the liquid flow is associated with both the length of

the liquid column and the contact angle. As seen in the plot, the surface tension

force is comparable to the centrifugal force only during the initial filling stage, but

not when the liquid column develops more into the spiral channel. A more

dominant factor is the contact angle of the liquid. When the solid surface is

hydrophobic and the contact angle is close to 90°, Fs can be completely

neglected. In our experiments, the solid substrate is either PDMS polymer or

chemically-treated glass with an organic layer on top, the contact angles of the

solutions on these substrates were usually found to range from 40° to 80°.

Therefore, surface tension force can only take effect during the very short initial

stage, which is highly transient. Figure 7-8(b) shows the CFD simulation results

predicting the difference in liquid front velocities at this stage for two different

solutions. The contact angles of solution A and B on the substrate were 60° and

152

46°, respectively. The liquid front velocity of solution A is 30% slower than

solution B during the first 1.5 s. On the other hand, we did not observe a

significant slower liquid front velocity from our experimental results using these 2

solutions (results not shown). The discrepancy could be attributed to the

phenomenon of the dynamic contact angles during liquid movement [173]. Figure

7-8(c) shows the measured contact angles of the two solutions on the PDMS

surface. The difference in static contact angles are around 15°. However, when

liquids are advancing, both contact angles increase and the difference in

dynamic contact angles is less than 5°. Therefore, if a solution with lower surface

tension coefficient and a hydrophobic substrate such as PDMS or modified glass

were used, which are common in the bio-application of microchips, neglecting the

surface tension force will not introduce a significant error. Moreover, because

surface tension is independent of rotation speed, this effect can also be

incorporated into the Navier-Stokes equations as a constant pressure term

added onto the liquid front in our future modified mathematical model [291].

153

Figure 7-8 (a) Comparison of the centrifugal force (Fω, shown as a gray plane.) and the surface tension force (Fs, shown as a color surface) along the microchannel at 1500 rpm (b) Simulated the liquid front velocities at 1500 rpm from two solutions (A and B) with different contact angles on the solid substrate (c) Measured dynamic contact angle change of the two solutions from experimental observations. Reproduced from [286] by permission of The Royal Society of Chemistry.

7.5.3.4 The effect of the neck channel

As seen in Figure 7-6(a), the microchannel has a short “neck” channel

which joins the inlet reservoir to the starting point of the spiral channel. The

average length of the neck channel is 2 mm. In our mathematical model, this

neck region is assumed to play a negligible role in the fluid motion within the

spiral portion of the channel. From the comparison of the results from modeling

and experiments (Figure 7-9(a)), the predicted liquid front velocity from the

analytical solution at 2400 rpm is around 30% less than that from experimental

0.000

0.002

0.004

0.006

0.008

0.010

0.012

0.014

0 0.3 0.6 0.9 1.2 1.5

Time (Seconds)

Liqu

id fr

ont v

eloc

ity (m

/s)

Contact angle = 60 degreeContact angle = 46 degree

40

45

50

55

60

65

70

75

80

0 50 100 150 200 250 300

Run number

Con

tact

ang

le (º

)

Solution 1 on PDMS

Solution 2 on PDMS

(a) (b)

050

100150

200 2040

6080

1000

2

4

6

8

10

12

x 104

Forc

e (P

a/m

)

Liquid length (mm) Contact angle ( )

(c)

Solution B

Solution A

154

observation. We speculate that the error can be attributed to the underestimation

of the pressure induced by the neck section. The external pressure differences

between the inlet and the outlet are zero because both boundaries are open to

air and the hydraulic pressure from the reservoir solution is negligible. However,

the mathematical modeling solution was solved starting from the beginning of the

spiral channel instead of the inlet reservoir. The rotation of the short liquid plug at

the neck channel under centrifugal field does create an extra pressure on the

flow inside the spiral channels. This pressure, which could be estimated using

the equation given in the work of Madou et al. [147], is proportional to the square

of the angular velocities (Figure 7-9(b)). At a high rotation speed such as 2400

rpm, this extra pressure is around 3 times of the pressure induced at 1400 rpm.

Therefore, a significant deviation between the analytical solution and

experimental data were observed at higher rotation speeds. Further improvement

in the mathematical model is expected to include this effect.

Figure 7-9 (a) Liquid front moving velocity under different rotation speeds. Both data from experimental data and from mathematical model prediction are shown. The error bars are from the measurement of 3 adjacent channels. (b) Pressure at the beginning of the spiral channels generated from the short liquid column in the neck section under different rotation speeds. Reproduced from [286] by permission of The Royal Society of Chemistry.

0

500

1000

1500

2000

2500

3000

3500

1200 1400 1600 1800 2000 2200 2400

Rotation speed (rpm)

Pres

sure

from

nec

k (P

a)

0

4

8

12

16

20

1400 1500 1600 1800 2000 2400

Rotation speed (rpm)

Liqu

id fr

ont v

eloc

ity (m

m/s

) Experimental measurmentsMathematical modeling

(a) (b)

155

7.6 Conclusions

This work describes the experimental measurement and mathematical

modeling of centrifugal-pumping flow in spiral microchannels. Quasi-steady flow

in rotating channels has been attained on a microfluidic glass disk with the

equiforce spiral channel design. The flow velocities and residence time of

solutions can be controlled by adjusting rotation speeds. In contrast to radial

channel design, liquid flow in the spiral microchannels is more manipulatable and

the residence time of the liquid is much extended. For example, at 1400rpm, the

centrifugally driven flow in the spiral microchannels is only at a speed of 4.3

mm/s. However, the average flow speed in a radial microchannel with similar

dimension could be up to ~45 mm/s, depending on the radial extent of the liquid

front from the inlet. Because many applications of microfluidics were focused on

reactions inside microchannels, a longer residence time is required and a much

slower flow speed is favourable. The spiral design has shown good spatial and

temporal control over the position of fluids in microscopic channels, which is an

advantage when longer reaction time is important in bio-analysis applications

done in a flow system. With the employment of centrifugal pumping force, parallel

analysis for multiple samples has been achieved in our previous work.

The mathematical modeling studies in this paper were based on solving

the Navier-Stokes equations in a customized orthogonal curvilinear-coordinate

system aligned with the channel geometry. Under proper assumptions, the liquid

flow speeds predicted by the mathematical model and by the CFD program

showed reasonable agreement with the experimental measurements at different

156

rotation rates. Moreover, the assumptions used in the mathematical model have

been examined with both experimental and computational simulation data. The

mathematical model gave a flexible and rather accurate analytical solution using

much less computing power. Compared with the commercial CFD program,

where results can only be obtained in weeks for one group of parameters, the

proposed mathematical studies gave an analytical solution with similar accuracy

in seconds. The proposed study demonstrated the effectiveness of the spiral

microchannel design in microfluidic applications using centrifugal force. With

modifications, this work could be adapted to the simulation and modeling of other

centrifugal-pumping microflow systems.

157

7.7 Appendix A: The spiral coordinate system

In order to obtain the coordinates of the equiforce spiral curve, we first

consider an infinitesimal segment dθ (Figure 7-10), from which we see that

tandrdrθ α= (10)

Figure 7-10 Section of the spiral curve and the infinitesimal segment used for the calculation. Here the black solid curve is the spiral curve and the blue dashed lines are radial lines from the disk center. r and θ are the usual polar coordinates. Reproduced from [286] by permission of The Royal Society of Chemistry.

Differentiating (1) with respect to α and substituting the result into (10)

yields

2tanddθ αα= ,

which after integrating gives θ = tan α − α. We can now view the polar

coordinates along the spiral curve as being parameterized by α,

( ) ( ) ( ) [ ]0 1 sec , tan , ,r kα α θ α α α α α α= = − ∈ . (11)

θ

dθ

dr

rdθ α

The spiral curve

O

158

where the interval [ ]0 1,α α satisfies ( ) ( )1 0 2θ α θ α π− = .

To obtain the distance along the spiral, s, we observe from Figure 7-10

that cosdr ds α= . Substituting this into (11) gives

ds rdr k

=

and integrating yields

2 20

1 ( )2

s r rk

= − . (12)

We now consider a three-dimensional spiral channel of width W and

height H and whose inner wall is described parametrically through (11). We

introduce an orthogonal curvilinear coordinate system, (n, s, z), that is aligned

with this geometry. Here, s is the distance along the spiral, 0 ≤ s ≤ L, n is the

cross-channel distance, 0 ≤ n ≤ W, and z is the height above the plate, 0 ≤ z ≤ H.

Let en, es, k be a right-handed system of unit vectors in the n, s, and z directions,

respectively. The unit vectors en and es are related to the unit vectors in polar

coordinates, er and eθ through the following:

cos sinα α= +s rθe e e , sin cosn α α= −rθe e e (13)

To rewrite the Navier-Stokes equations in spiral coordinates, we must first

know the position vector R to a point described by (n, s, z) in the channel. A

straightforward geometric argument gives

cos ( sin ) nr n r zα α= + + +sR e e k . (14)

159

In addition, we must determine the scale factors, or the rate of change of

arc length with respect to each coordinate direction. The scale factor in the s

direction is

1h ns

κ∂= = +

∂R

, 1

sinrκ

α= , (15)

where κ is the curvature. In both the n and z direction, the scale factors

have unit value. Following the methodology presented in Davis and Snider [292]

for general orthogonal curvilinear coordinate systems, we determine that the

gradient and the Laplacian operators are

1= nh s n z∂ ∂ ∂

∇ + +∂ ∂ ∂se e k ,

2

2

1 1 1( ) ( )hh s h s h n n z∂ ∂ ∂ ∂ ∂

∆ = + +∂ ∂ ∂ ∂ ∂

,

For a vector field s n n zu u u= + +su e e k , we reformulate the Navier-Stokes

equations (2) in terms of the spiral coordinate system as follows:

22

2 2

1 12 ( )s ns s n n s s n

s

u upu u u u k u u ut h h s h h s h s h

κ κ κ κω ω νρ

∂ ∂∂ ∂+ ⋅∇ − − + = − + ∆ − − − ∂ ∂ ∂ ∂

u , (16)

2 22 2

1 1 12 ( ) ( )n sn s s n n s

u upu u u n u u ut h n h s h h s h

κ κ κ κω ω νκ ρ

∂ ∂∂ ∂ + ⋅∇ + + − − = − + ∆ + − + ∂ ∂ ∂ ∂ u , (17)

1zz z

u pu ut z

νρ

∂ ∂+ ⋅∇ = − + ∆

∂ ∂u , (18)

1 ( ) 0s zn

u uh huh s n z∂ ∂∂

+ + =∂ ∂ ∂

, (19)

160

where

sn z

u u uh s n z

∂ ∂ ∂⋅∇ = + +

∂ ∂ ∂u .

We now proceed under the assumption of a unidirectional flow only along

the s direction, un = uz = 0. The continuity equation reduces to

0sus

∂=

∂,

which implies that us = us (n, z). Then only the s-momentum equation (16)

is used, and the equation reduces to

2 22

2 2

1 1( )s ss

u u pk h uh n n z h h s

κω νρ

∂ ∂∂ ∂− + − = − ∂ ∂ ∂ ∂ .

(20)

The scale factor in the s direction, h = 1 + nκ, is expected to have a value

close to 1 for since n (0 ≤ n ≤ W) is much smaller than 1. Furthermore, since we

expect the cross-channel gradients to be significant, given the confines of a small

cross-sectional area, we can further simplify the above by noting that

2 22

2 2s s

su u un z

κ∂ ∂+

∂ ∂ .

Introducing the operator ΔT,

2 2

2 2T

n z∂ ∂

∆ = +∂ ∂ ,

equation (20) becomes

( )2 1,Ts

pk u n zs

ω νρ∂

− ∆ = −∂

. (4)

161

7.8 Appendix B: Simulation set-up and the study on the transversal flow with the CFD program

7.8.1 Governing equations

The transient, three-dimensional numerical simulations of the centrifugal

flow in spiral microchannels were performed based on the flow module and free

surface module of the CFD program. In the flow module, we solve momentum

equation and continuity equation under a rotation reference frames. The flow is

considered to be laminar, incompressible, Newtonian and isothermal with velocity

field u governed by the Navier–Stokes and continuity equations, and surface

tension force (Fs) at the air/liquid interface is incorporated as an additional term in

the equations used in simulations.

In the free surface module, the volume of fluid (VOF) method was used to

model the location of the fluid front during filling of the microchannel [293]. The

system consists of two incompressible and immiscible fluids represented as the

liquid and gas phases, respectively. The VOF distribution has been determined

by solving a separate passive transport equation, given as

0f ft

∂+ ⋅∇ =

∂u

where f is the volume of fluids function (f = 1 in grids completely filled with

water, f = 0 in grids filled with air, and f has a fractional value corresponding to

the volume fraction of water in partially filled grids).

162

7.8.2 Volume, boundary, and initial conditions in the numerical simulations

Since both the inlet and outlet reservoirs were open to air without any

extra fluidic connections, the boundary conditions of inlet and outlet were set to a

fixed pressure of 0 Pa. This pressure condition is the same as that used in

literature reporting the simulation of the flow in a radial microchannel [171]. All

the walls were taken to be non-compliant (or rigid) and were defined as rotating

walls with the same angular velocity as the global frame of reference. The wall

boundary conditions used for the simulations involve no-slip condition with

contact angles at the front specified for each wall (to be consistent with the

experimental conditions, as obtained from the contact angle measurements of

the microchannel materials). The contact angle of the fluid with the bottom wall

(chemically-modified glass surface) is 52.8° ± 2.1°. The top and side walls of the

microchannel are PDMS, and the contact angle is 73.2° ± 2.5°.

During our experiments, the inlet reservoirs were filled with 0.5~1.0 μL of

solution before the chip assembly was spun. To match the experimental

conditions, we started the simulation with initial conditions that assumed 0.5 μL of

liquid present in the inlet reservoir. Moreover, because the liquid could crawl into

the microchannel under the pressure from the pipetting process, we also

assumed a certain start length of liquid column in the microchannel as initial

conditions.

The fluids under consideration are the buffer solution as the liquid, and air

as the gas. The thermo-physical properties of the solution were measured and

163

applied directly in the computational simulations. For air and the buffer solution

respectively (at room temperature), the densities are 1.614 kg·m-3 and 1020

kg·m-3, respectively, and the dynamic viscosity (µ) are 1.846 × 10-5 kg·m-1·s-1 and

1.25 × 10-3 kg·m-1·s-1, respectively. Because of the addition of surfactant in the

buffer solution, the surface tension of the air-liquid interface was measured at

0.03 N·m-1, a value smaller than that for pure water.

In the simulations, the uniform rotation speed was imposed at a steady-

state value in either clockwise or anti-clockwise direction to match the

experiments. The center of the rotation axis corresponded to the center of the CD

and was located at a distance 21 mm from the start of the spiral channel.

The computational model included the flow module to solve the fluid

dynamics field as well as the free surface module to model the advancing of the

liquid front during microchannel filling. This transient process was solved using an

unsteady solver with implicit first-order Euler time integration under rotation

reference frames (i.e., centrifugal microfluidics). Simulations were performed in

automatic time steps with a scale around 1 × 10-5 s. The air/liquid interface was

reconstructed using a second order piecewise linear interface calculation. Explicit

formulations were used and each time step was calculated via the Courant-

Friedrichs-Lewy parameter (set to 0.1), which restricts the size of the time step to

ensure the free surface did not cross an entire element within that time step.

164

7.8.3 The effect of mesh grid density on the computation time and accuracy during the numerical simulations

Two different resolutions were used in the CFD simulation of the system.

A structured grid with 35, 970 cells was used in the transient flow simulations and

the cell numbers at the cross section are 5×2 cells. In steady-flow simulation, the

whole microchannel was filled with solution and a grid of much higher density

was used. The total computational cells are 1,834,560 with a symmetry plane

and the cell numbers at the cross section are 20×12 cells. Because the channel

was initially filled with air, the free-surface module in CFD-ACE program was

used to solve the surface-reconstruction procedures. To keep the run times

within reasonable limits, the number of cells was simplified resulting in a much

lower grid density.

To examine if the mesh density in transient-flow simulation is sufficient for

the calculation of the liquid front velocity, we compared the results from steady-

flow simulations using two resolutions. With low-density mesh, the cross-channel

resolution is 5 × 2 cells (Figure 7-11a(i)). With high-density mesh, the cross-

channel resolution of the computational grid is 20×12 cells (Figure 7-11b(i)). It

can be seen in Figure 7-11a(ii) and b(ii) that the predicted down-channel flow

velocities (us) are similar from these two meshes. Table 7-1 also lists the

maximum down-channel flow velocities (us_max) calculated from both grid

settings. The us_max calculated under 5 × 2 grids is only 0.4% in difference from

that calculated under 20×12 high density grids. The results indicated that the

low-density mesh is sufficient in predicting the liquid front position, which is

determined by the down-channel flow velocities. On the other hand, the 5 × 2

165

grids did not give accurate distribution on transversal cross-channel flows (un and

uz, as shown in Figure 7-11a(iii) and b(iii)) and the maximum values (un_max and

uz_max) calculated from both mesh settings are very different from each other

(Table 7-1). Because the purpose of the transient filling-flow simulation study is

to accurately predict the position of the advancing liquid column in a rotating

microchannel, accurate calculation of down-channel flow velocity (us) is more

important than the calculation of cross-channel transversal flows (un and uz).

Table 7-1 Comparison of the simulation results from two different mesh densities.*

us_max (mm/s)

un_max (mm/s)

uz_max (mm/s)

CPU time of each cycle

(Sec)

Low-density mesh 24.5 0.030 3.5×10-5 178

High-density mesh 24.6 0.043 0.012 2928

* Because the purpose of the transient filling-flow simulation study is to accurately predict the position of the advancing liquid column in a rotating microchannel, accurate calculation of down-channel flow velocity (us) is more important than the calculation of cross-channel transversal flows (un and uz).

The low-density mesh saved a lot of computation time. As seen in Table

7-1, for steady-flow simulation, the total computation time of one cycle for the

model with low-resolution grids is less than 3 min, compared with almost 1h for

the simulation of the high-resolution grids. Considering that it usually took ~5,000

to 10, 000 cycles during our simulation practice, the use of the low-resolution

grids reflects a compromise between overall accuracy and computational effort in

the simulation of multi-phase problems.

166

Figure 7-11 Comparison of the simulation results using two meshes with different resolutions. (a) low-density mesh, cross-channel resolution is same as that used in the manuscript (5 × 2 cells). (b) high-density mode, the cross-channel resolution of the computational grid is 20×12 cells. (i) computational grids showing the cross-section of the microchannel. (ii) the color map of the down-channel flow velocities (us) at the cross-section of the microchannel.(iii) the color map of the transversal flow velocities (un) at the cross-section of the microchannel. The center black lines in high-resolution grid results are the symmetry plane.

7.8.4 Validation of the assumption of negligible cross-channel flows.

The cross-channel secondary flows are also estimated by simulation.

Figure 7-12(b) shows the flow vector map combining un and and uz in a cross-

section of the spiral microchannel as shown in Figure 7-12(a). Figure 7-12(c) to

(e) show the color maps of flow speeds in 3 directions. The maximal us, and un

are found at 24, and 0.045 mm/s, respectively. Therefore, the values of

secondary flow velocity un are less than 0.2% of that in down-channel direction.

In terms of another cross-section flow uz, it is even smaller and is calculated at

0.012 mm/s. therefore, the computed un and uz are “small” in some significant

(a) (b) (i)

(ii)

(iii)

(i)

(ii)

(iii)

mm/s

167

sense relative to the down-channel flow us. We believe that the assumption 5 in

the body text of un = uz = 0 is valid.

Figure 7-12 (a) Schematic diagram showing the position of the cross-section part of the microchannels used for secondary flow analysis. (b) The flow vector map combining un and uz. (c) The color map of us on the cross section. (d) The color map of un on the cross section. (e) The color map of uz on the cross section. The center black lines in (b) to (e) are the symmetry plane used during simulation.

(a) (c)

(d)

(e) (b)

mm/s

168

8: MODELING OF MICROFLUIDIC DNA MICROARRAY HYBRIDIZATION AND ITS APPLICATIONS TO EXPERIMENTAL OPTIMIZATION

8.1 Introduction

Like other biosensor technologies, DNA Microarray is based on the

fundamental property of nucleic acid strands specifically binding (hybridizing) to

their complementary strands [294]. Generally, probe DNA molecules are

immobilized on a solid surface and form an array of microspots. Then, labeled

target DNA strands in solution phase are applied to the surface. Because the

immobilized probe DNA binds to its complementary strand, detection is achieved

through the read-out of the tagged markers on the retained target molecules.

The conventional microarray method occurs in a static mode, where

sample DNA in bulk solutions are transported to the reactive surface solely

through molecular diffusion. This slow process could limit the rate of the

hybridization assay, leading to long-term incubation (12-18 h) and poor detection

efficiency [19]. The recent microfluidic technology enhances the mass transport

by introducing convection to the microarray system [84, 91]. Moreover, the

microfluidic method shows the advantages of less sample usage (down to

submicrolitre) and reduced assay time (in minutes) compared to the bulk solution

method [129].

169

The kinetics of DNA microarray hybridization, which is heterogeneous, is

much more complex than that observed in homogeneous liquid-phase DNA

hybridization [10, 295]. Additional parameters from mass transport mechanism,

such as bulk and surface diffusivity, flow velocity, channel geometries, etc. are

involved. A sound and detailed mathematical modeling of the ongoing diffusion

and reaction processes is thus needed to properly design both convection- and

diffusion-driven DNA microarray systems. Because the system shares many

common features with other biosensors, the theoretical studies on these

heterogeneous interactions have provided considerable insight into the process

of DNA surface hybridization. For example, a simplified two-compartment model

was originally proposed for the antigen-antibody reaction in protein biosensors

[296-301] and it was then used to analyze the on-chip DNA hybridization [302,

303]. Another diffusion-reaction model proposed by Chan et al. was developed

from a cell receptor-ligand model [304] and it has been regarded as a pioneering

work on static microarray hybridization [305, 306]. In his work, hybridizations of

targets from bulk phase and from initial non-specific surface adsorption were

considered and the model has successfully predicted the enhanced kinetics from

sparse probe coverage. Other models have also been reported to further

describe the effect of different physico-chemical parameters on kinetics during

static hybridization [57, 307, 308]. Regarding the recent microfluidic technology,

Erickson et al. coupled Chan’s work with convection transport relations and they

examined their model with experimental results from oligonucleotide targets

[309]. Similar models have been reported for other microfluidic-based surface

170

biosensors [310-312]. Because the mathematical description of these models

involves the use of nonlinear partial differential equations (PDE), numerical

solutions have been achieved with the increasing computing power of modern

computers [303, 309, 311-313].

The present work aims at providing a comprehensive analysis of the

combined effect that the many different process variables have on DNA

microarray hybridization in diffusion- and convection-driven systems. The use of

dimensionless parameters and variables in our work brings in a universal

character to the model, transforming a specific situation into a generic case, as it

has been applied in a few studies on static hybridization [314, 315].The proposed

model was solved numerically and the results have been used to predict the

calibration range of oligonucleotide targets, and to examine the effect of flow

rate, channel depth and probe coverage on hybridization kinetics. Furthermore,

the problems about hybridization spot morphology and signal variation from spots

to spots are also explained for the first time with the present model. With the

validation from experimental data, our work could provide a quick route for the

experimentalist to estimate the importance of different parameters in a

microfluidic microarray chip and thus convenient criteria which can be used in the

design process.

171

8.2 Experimental

The fabrications of the microchip used and the subsequent microfluidic

hybridization analysis have been depicted in the previous chapters.

8.3 Model construction for microfluidic DNA microarray hybridization

8.3.1 Navier-Stokes and continuity equations

The configuration of a PDMS plate with a rectangular microchannel is

shown in Figure 8-1(a) and the liquid inside the channel is assumed to flow at a

steady state in the absence of body forces like gravity. Incompressible Navier-

Stokes equations (2) are used along with the continuity equation (1) to find the

velocity profile throughout the domain.

0∇⋅ =u (1)

21 p νρ

⋅∇ = − ∇ + ∇u u u (2)

In this equation, u is the velocity vector, p is the pressure, ρ is the density,

and ν is the kinematic viscosity of the solution. Equation (2) can be further

simplified based on the thin-film flow theory, which studies the steady flow of

viscous liquid between two rigid boundaries with small distance apart [316-318].

For a typical microfluidic flow in our experiments, the height scale of the flow (H ~

30 µm) is much smaller than the length scale of the flow (L > 30 mm). Moreover,

considering that the typical flow speed (U) of the microfluidic flow is around 1

172

mm/s, the inertial term ( ⋅∇u u ) in equation (2) can be ignored because

2

Re 1HL

<<

. Therefore, the Navier-Stokes equation is reduced to:

20 p µ= −∇ + ∇ u (3)

where µ is the dynamic viscosity of the solution.

The pressure-driven, steady-state laminar flow in long and rigid channels

is also known as Hagen-Poiseuille flow [319]. The transversal flow velocities in y

and z direction are small and negligible. In microfluidics, the aspect ratio of a

rectangular channel can often be large and the channel is well approximated by

an infinite parallel-plate configuration. We thus could drop out the y coordinate

and use 2-D geometry as shown in Figure 8-1(b).

173

Figure 8-1 (a) The configuration of a PDMS plate with a rectangular microchannel filled with solutions. The bottom wall of the microchannel is immobilized with 3 probe strips. (b) The schematic diagram showing the simplified 2-D geometry of the above microchannel used for later numerical simulation. The two diagrams are not drawn to scale.

A simplified format of the equation (3) for the 2-D channel is then

expected:

2

2

1u dpz dxµ∂

= −∂

(4)

With no-slip boundary condition, the solution of equation (4) is a simple

parabola

1 ( )2

dpu H z zdxµ

= − − (5)

(a)

Outlet

Bulk solution of target DNA

Inlet

Probe regions Spacing

Bottom wall

Upper wall

H

(b)

PDMS channel plate

Microchannel filled with solutions

3 immobilized probe strips on the

glass substrate

z

x

174

If the average flow velocity (U) is defined as volumetric flow rate divided

by the cross-section area [319], then

2

12H dpU

dxµ= − (6)

and the 2-D flow velocity equation (5) becomes:

6 1z zu UH H = −

(7)

8.3.2 Mass transport and hybridization kinetics

In DNA microarray hybridization, target DNA molecules are dissolved in

bulk solution and then applied to the chip immobilized with probe arrays. As

shown in Figure 8-2, the target DNA molecule moves from bulk solution to glass

surface and hybridizes with the immobilized probe DNA. Therefore, many studies

have been based on modelling target mass transport from the bulk solution to the

reactive surfaces [311, 312, 320]. In addition to this bulk-to-surface interaction,

Chan et al. also proposed another mechanism regarding the heterogeneous

hybridization of solution-phase DNA [305, 309]. Here, target DNA molecules

adsorb non-specifically to the glass substrate first [321, 322]. Then the absorbed

target molecules could move along the surface and hybridize with the probe DNA

in a two-dimensional manner. Our model incorporates the transport of target

molecules in solution phase, the direct bulk-to-surface hybridization kinetics, non-

specific binding of target to bottom wall, as well as the subsequent surface-to-

surface hybridization kinetics.

175

Figure 8-2 (a) Schematic representation of proposed heterogeneous DNA hybridization model. The probe DNA molecules are shown as blue wave-like ribbons in the center region of the glass substrate. The target DNA molecule is a red twisted ribbon with a red pellet at one end which represents the fluorescence label. The hybridized DNA are shown as duplexes in the probe region. (b) The three kinetic processes used in the later model construction regarding the heterogeneous hybridization of target DNA: k3 and k-3 represent bulk-to-surface hybridization and dessociation of targets to the bulk phase; k2 and k-2 represent surface-to-surface hybridization and dissociation of the nonspecifically adsorbed targets; and ka and kd represent the reversible nonspecific adsorption and desorption of the targets to the surface.

Region of immobilized probe arrays

Labelled target DNA in bulk solution

Glass substrate

Hybridized DNA duplex on probe site

Non-specifically adsorbed target DNA

Bulk solution

(a)

(b) 3

3

a

d

k

k

k

k

−

Target DNA (in bulk phase) + Probe DNA (immbolized) Hybridized Duplex

Target DNA (in bulk phase) Target DNA (adsorbed on the surface)

Target DNA (adsorbed on the surface) + Prob 2

2

k

k−

e DNA (immbolized) Hybridized Duplex

176

8.3.2.1 Mass transport through diffusion and convection in bulk solution

The target DNA molecules are carried by the microflow and transported to

the reactive surfaces through convection and diffusion. Therefore, the process

can be described by the convection–diffusion equation [323]:

( )2C D C Ct

∂= ∇ − ⋅∇

∂u (8)

Where C is the target concentration in the bulk solution, and D is the

diffusion coefficient of the target DNA molecules in bulk phase. In the 2-D

microchannel model shown in Figure 8-1(b), lateral flow velocity in z direction is

small and negligible. The transient two-dimensional mass transport equation can

be rewritten as follows:

2 2

2 2

C C C CD ut x z x

∂ ∂ ∂ ∂= + − ∂ ∂ ∂ ∂

(9)

8.3.2.2 Surface hybridization and adsorption kinetics

Considering the two mechanisms of the hybridization of target DNA

molecules with the surface-immobilized probes, the formation rate of the

hybridized duplexes at the probe region is the sum of two hybridization rates as

shown in follows:

bhy shyR Rt

∂Θ= +

∂ (10)

where Θ is the surface concentration of the hybridized duplexes, Rbhy is

the hybridization rate of the target DNA from bulk phase, and Rshy is the

hybridization rate of the target DNA non-specifically adsorbed on the glass

177

surface. We assumed that the upper PDMS wall is inert and does not adsorb

DNA molecules. Because the hybridization capacity at a probe site is limited by

the total number of immobilized probes, the maximal surface concentration of

hybridized duplexes is equal to the probe density ( 0Θ ) and ( )0Θ −Θ is the

concentration of the available probe molecules on the surface. Therefore, the

bulk-to-surface hybridization rate Rbhy can be analyzed according to Langmuir

kinetics [324]:

( )3 0 3bhy wR k C k−= Θ −Θ − Θ (11)

where Cw is the concentration of the target DNA molecules adjacent to the

wall, k3 and k-3 are the hybridization and denaturation rate constant of the bulk-to-

surface hybridization process, respectively.

Similarly, the hybridization rate of the adsorbed target DNA molecules is

given as:

( )2 0 2shyR k kη −= Θ −Θ − Θ (12)

where η is the surface concentration of non-specifically adsorbed target

DNA molecules, k2 and k-2 are the hybridization and denaturation rate constant of

the surface-to-surface hybridization process, respectively. Substitute equation

(11) and (12) to equation (10), we get:

( ) ( )3 0 3 2 0 2wk C k k kt

η− −

∂Θ= Θ −Θ − Θ + Θ −Θ − Θ ∂

(13)

In order to compute the rate of change of the surface concentration of

non-specifically adsorbed target DNA ( tη∂ ∂ ), we consider the processes

178

including non-specific adsorption, diffusion along surface, and subsequent

hybridization. Therefore, the reaction-diffusion equation for the adsorbed

molecules is given as [325]:

2s ads shyD R R

tη η∂= ∇ + −

∂ (14)

where Ds is the surface diffusion coefficient of the target DNA on the glass

surface and Rads is the rate of adsorption. If we define ηmax as the maximal

adsorbable number of the target DNA at a unit area of the glass surface, Rads can

be written as

( )maxads a w dR k C kη η η= − − (15)

where ka and kd are the adsorption and desorption rate constant of the

target DNA molecules, respectively.

Substitute equation (12) and (15) to equation (14), we get:

( ) ( )2max 2 0 2s a w dD k C k k k

tη η η η η η −

∂= ∇ + − − − Θ −Θ − Θ ∂

(16)

The equation (9), (13), and (16) are coupled by a flux balance at the wall

( 0zJ = ):

( ) ( )

0 0

max 3 0 3

( , )( , , )

wz z ads bhy

a w d w

C x tC x z tJ D R Rz t

k C k k C kη η η

= =

−

∂∂= − = − = +

∂ ∂

= − − + Θ −Θ − Θ

(17)

179

8.3.3 Non-dimensionalization of the reaction-diffusion and kinetic equations.

Non-dimensionalization is a useful tool in simplifying physical systems

[315, 317]. It rescales the parameters and variables so that all computed

quantities are of relatively similar magnitude. The units in the system are also

wiped out. In the governing equations shown in the last section, the different

variables such as time (t), channel length (x), channel height (z), down-channel

flow velocity (u) and bulk concentration (C) are divided by their respective

characteristic values (measured experimentally). The resulted dimensionless

forms are given as follows:

*2

ttH D

= , *2

xxUH D

= , * zzH

= , * uuU

= , *

0

CCC

= (18)

where 2H D is the diffusion timescale for a DNA molecule to travel from

channel top to channel bottom and it has been used as the characteristic time

scale, U is the average liquid flow velocity in microchannels and it can be

measured experimentally, and C0 is the initial bulk concentration of the target

DNA molecules.

With the characteristic values of flow velocity and channel height, the

dimensionless form of the 2-D flow velocity in equation (7) becomes:

* * *6 (1 )u z z= − (19)

and the transient 2-D mass transport equation (9) is rewritten as:

* 2 * 2 * **

* 2 * 2 * 2 *

1( ) ( )

C C C Cut Pe x z x

∂ ∂ ∂ ∂= + −

∂ ∂ ∂ ∂ (20)

180

where Pe is the Péclet number of the flow and it is defined as UHPe D= .

Similarly, for the reaction-diffusion equation (16) of the surface

adsorption/hybridization procedures, the dimensionless variables are

*

max

ηηη

= , *

0

ΘΘ =

Θ, *

0

ww

CCC

= (21)

Then the non-dimensionalization form of equation (16) becomes:

( )

( )

* 2 ** * *0 max

* 2 * 2max 0

* * *0 2 max 2

max 2 max

1 1( )

1

s a dw

a

s

s

D C H k H kCt D Pe x D k C

D k H kD D H k

ηη η η ηη

η ηη η

−

∂ ∂= + ⋅ − − ∂ ∂

Θ− ⋅ ⋅ −Θ − Θ

(22)

Introduce surface/bulk diffusivity ratio (φ ), relative surface capacities (ε1,

ε2, and ε3), two Damköhler numbers (Da1 and Da2) and two dimensionless

equilibrium constants ( 'D

K and '2K ), as shown

sDD

φ = , 01

max

C Hεη

= , 02

max

εηΘ

= , max1

ak HDaDη

= , 2 max2

s

k HDaD Hη

= , '

0D

d

a

kKk C

= ,

' 22

2 max

kKk η

−= (23)

Thus the equation (16) becomes

( ) ( )* 2 *

* * ' * * * ' *1 1 2 2 2* 2 * 2 1 1

( ) w DDa C K Da Kt Pe xη φ η ε η η φε η∂ ∂ = + − − − −Θ − Θ ∂ ∂

(24)

Similarly, for the bulk-to-surface hybridization kinetics shown in the

equation (13):

181

( ) ( )*

* * * * * *0 3 0 3 2 max 2*

0 3 0 2 max

1 1sw

s

C H k H k D k H kCt D k C D D H k

η ηη

− − Θ∂Θ= ⋅ −Θ − Θ + ⋅ −Θ − Θ ∂ Θ

(25)

Introduce relative hybridization capacity (ε3), Damkohler number (Da3) and

the dimensionless equilibrium constant ( '3K ),

03

0

C Hε =Θ

, 3 03

k HDaDΘ

= , ' 3 33

0 3 0

K kKC k C

−= = (26)

Then the dimensionless form of equation (13) is shown as:

( ) ( )*

* * ' * * * ' *3 3 3 2 2* 1 1wDa C K Da K

tε φ η∂Θ = −Θ − Θ + −Θ − Θ ∂

(27)

For the flux balance, the dimensionless form of equation (17) is shown as:

( ) ( )* * *

* * ' * * * ' *0 1 3 3*

( , , ) 1 1z w D wC x z Da C K Da C K

zτ η η=

∂ − = − − + −Θ − Θ ∂ (28)

The equation (19), (20), (24), (27), and (28) are used in the later numerical

simulation works.

8.3.4 Physical interpretation of the dimensionless numbers:

Five main dimensionless parameters have been introduced in the non-

dimensionalization process and they are Pe, φ , ε, Da, and K'.

The Péclet number (Pe) is a fundamental dimensionless number in

convective transport in both macroscale and microscale flow [314]. If we rewrite

its definition as follows

2 / Characteristic diffusion time/ Characteristic convection time

UH H DPeD H U

= = = (29)

182

it can be seen clearly that the Péclet number is the ratio between the

characteristic diffusion time ( 2 /H D ) and the characteristic convection time

( /H U ). Typically, flow-through and microfluidics-based assays are characterized

by high Péclet numbers (Pe >> 1), such that the convective transport dominates

over diffusion and localized depletion around probe areas is eliminated.

Considering that the flow speed is typically around 1mm/s and channel height is

30 µm in our microfluidic chips, Pe number is calculated to be ~ 200 for 21-mer

oligonucleotide targets. Therefore, microfluidics-based DNA hybridization

alleviates the limitation from diffusion, and hybridization is controlled mostly by

the reaction rate of the surface hybridization.

Biomolecules can be adsorbed to a variety of solid surfaces and these

molecules could diffuse along the surfaces [306]. During DNA microarray

hybridization, the adsorbed target molecules in the no-probe-covered regions can

diffuse to probe regions and contribute to the surface hybridization process. We

use the dimensionless number φ to calculate the ratio between the surface

diffusion coefficient (Ds) and the bulk diffusion coefficient (D) of the DNA. The

higher is the value of φ , the faster is the movement of the non-specifically

adsorbed target molecules.

The relative surface capacities consist of 3 dimensionless numbers (ε1, ε2,

and ε3). For any unit area of glass surface inside a microchannel, ε1 corresponds

to the ratio of the maximal amount of target DNA in the bulk solution ( 0C H ) to the

surface adsorption capacity of the targets ( maxη ) at that area. ε2 is a measure of

183

surface hybridization capacity relative to the surface adsorption capacity. Similar

to ε1, ε3 is the ratio of the maximal amount of target DNA in the bulk solution

( 0C H ) to the surface hybridization capacity of the targets at the unit area ( 0Θ ).

Small values of ε1 and ε3 indicate a high relative surface-capturing capacity

leading to strong depletion and a longer saturation time. On the other hand, small

ε2 indicate that the solid surface will consume too much analytes by non-specific

adsorption and pre-hybridization coatings such as using salmon sperm DNA or

bovine serum albumin is needed to prevent target depletion from non-specific

binding [19].

Damköhler number (Da) is a dimensionless number and it has been used

by many groups to relate reaction timescale to mass transport timescale of a

microfluidic system [307, 311, 312, 315, 324, 326]. In this work, three Damköhler

numbers (Da1, Da2, and Da3) are defined: Da1 is the ratio of surface adsorption

rate to the bulk diffusive mass transport rate of target molecules; Da2 is the ratio

of surface-to-surface hybridization rate to the surface diffusive mass transport

rate of target molecules; Da3 is the ratio of bulk-to-surface hybridization rate to

the bulk diffusive mass transport rate of target molecules. In a static

hybridization, when Da >> 1, the transport to the surface is strongly diffusion-

limited, while at Da << 1, the transport becomes reaction-limited at the surface

and a flat (constant) concentration profile can be expected across the channel.

Three dimensionless equilibrium dissociation constants (KD', K2', and K3'.)

are also introduced during the non-dimensionalization procedures. They have

been used as measures of the thermodynamic stability of the adsorbed targets

184

(KD') or hybridized duplexes (K2', and K3') [326]. In terms of the microfluidic DNA

microarray analysis, the smaller values of these parameters indicate higher

stability of the adsorbed targets or the hybridized duplexes.

8.4 Simulation of microfluidic DNA microarray hybridization

8.4.1 Numerical simulation with COMSOL program

Numerical simulations and visualization were performed with COMSOL

Multiphysics 3.5a. The considered domain is assumed to be infinitely wide and to

have a height (H) equal to the thickness of the microchannel (Figure 8-1(b)). The

no-slip boundary conditions are applied to both upper and bottom walls during

flow simulation. Constant flow velocity (e.g. 5mm/s) is specified at the inlet. The

outlet is assumed to be at atmospheric pressure. Moreover, we assume that the

upper wall is inert and does not adsorb DNA molecules. As seen in Figure 8-1(b),

the bottom wall of the microchannel is comprised of two parts: probe regions

where probe DNA molecules are immobilized at a concentration of 0Θ and

spacing areas which are bare glass surface. Both non-specific adsorption and

target hybridization will occur at probe regions while the spacing areas only

adsorb target molecules.

Figure 8-3 An example of meshed 2-D geometry of the microchannel. The grid closed to bottom wall was meshed with much higher density because it is close to the reactive surface with probe spots.

185

The 2-D geometry in Figure 8-1(b) was computationally discretized into

unstructured triangular cells as shown in Figure 8-3. The total number of mesh

elements is ~ 6000 for a typical model. The simulations were run on a HP

M8407C computer with 6 GB of RAM and an Intel® Core™2 Quad processor;

each simulation typically required ~5 min of computation time.

8.4.2 Parameter values used during simulation

During the non-dimensionalization processes, a series of parameters have

been used to scale the dimensionless variables. To verify and compare the

effectiveness of the model, the values of these parameters are extracted from

our experimental results and literature reports. For example, the characteristic

value of the microflow velocity (U) is picked from the experimentally-measured

average flow velocity and it ranges from 0.1 to 5mm/s. The channel height (H) is

set to 30 µm, which is a typical value used in our work [94]. The concentration of

the target DNA molecules varies from 0.1 nM to 1µM, as reported in our previous

works [94, 134]. We have estimated that the probe density ( 0Θ ) on the aldehyde-

modified glass surface is ~ 5 × 1011 strands/cm2 or 8.3 × 10-9 mol/m2 [94]. This

probe coverage is comparable to the values obtained elsewhere [139, 221, 250,

251, 327, 328]. The average radius occupied per probe molecule (Rp) is 7 nm.

In our model, the maximal surface concentration of non-specifically

adsorbed target molecules (ηmax) has been used in the non-dimensionalization

process. Because the adsorption of DNA is non-specific, a full monolayer of

targets could be considered as maximal at spacing region [309]. On the other

hand, probe regions are not fully packed with probe molecules, as calculated

186

from the probe density, and target DNA could adsorb to these regions as well.

Therefore, ηmax is given by [309]:

20

max 2

1 A p

A t

N RN Rπ

ηπ

− Θ= (30)

where NA is the Avogadro’s Number, Rp is the radius of a probe site, and

Rt are the gyration radius of an adsorbed target molecule. For 21-mer

oligonucleotides and longer single-stranded DNA (260 bases from denatured

PCR products), the calculated gyration radius are 2 nm and 7 nm, respectively

[329].

The diffusion of DNA molecules in bulk aqueous solution has been studied

extensively [59, 329-332]. According to the Stokes-Einstein correlation, the

diffusion coefficients of macromolecules (D) are proportional to temperature, and

they are inversely proportional to the bulk phase viscosity and to the

hydrodynamic radius, which is highly related to the molecular size [330, 333].

Lukacs et al. has suggested an empirical equation on the diffusion coefficients of

duplex DNA in water: D = 4.9 × 10-6 cm2/s · [base pair number] -0.72. The equation

is applicable to double-stranded DNA fragments with sizes of 21–6, 000 bp [59].

As for single-stranded DNA, the diffusion coefficients have been reported to

follow an empirical equation as: D = 7.38 ×10-6 cm2/s · [base number] -0.539 from

approximately 10 to104 bases at room temperature in water [331]. Moreover, it

has been reported that the diffusion coefficients are influenced by experimental

conditions. For instance, low ionic strength and the presence of fluorescent

187

labels on DNA molecules could reduce the diffusion coefficient, but by less than

one order of magnitude of difference [329, 334].

Surface diffusion can lead to the enhancement of the hybridization

efficiency of DNA microarray [305, 306]. This two-dimensional diffusion of DNA

molecules is usually slower than that in bulk solutions and it could be very

different on various surfaces. Chan et al. measured the surface diffusion

coefficients (Ds) of 21-base oligonucleotides on two types of silanized glasses:

APTES-modified and N-methyl- APTMS-modified, and they found that the

values ranged from ~1 ×10-9 to 7 ×10-9 cm2/s [306]. These values are applied to

the simulation of DNA hybridization in our work.

The adsorption and desorption parameters of oligonucleotide on various

types of glass substrate have also been measured by Chan et al. [306, 321].

From their results the value of adsorption equilibrium constant (Ka) can be

estimated at ~1×104 M−1 and the desorption rate constant (kd) between 0.15 and

0.45 s−1, depending on the substrate type. Because no other data is available for

predicting the adsorption behaviour of different DNA sequences, we assume that

these values are constant for the adsorption/desorption processes of any DNA

sequence on the glass surface.

The kinetics of the heterogeneous hybridization between bulk-phase

target DNA and the surface-immobilized probes was thought to be slower than

that of homogeneous solution hybridization due to steric interference [15].

Okahata et al. measured the kinetic parameters of oligonucleotide microarray

hybridization [16]. The effects of probe spacers, sequence length, ionic strengths,

188

temperature, and mismatch location have also been discussed. From their

results the value of hybridization rate constant (k3) can be estimated at ~1×105

M−1·s−1 and the denaturation rate constant (k-3) is ~2×10-4 s−1, for room

temperature hybridization of 20-mer oligonucleotides with 20-mer immobilized

probes. These values are of the same order with the experimental results from

other groups’ work using different measurement methods [15, 335-337]. In terms

of the rate constant of the 2-D hybridization (k2) from surface-adsorbed target

DNA, it can be estimated from the following equation [309]:

2

3

163

s

p

Dkk DRπ

= (31)

We assume that the denaturation rate constants of both bulk-to-surface (k-

3) and surface-to-surface (k-2) processes are equal.

189

8.5 Results and discussion

8.5.1 Bulk concentration profiles of static hybridization and microfluidic hybridization

The model was used to visualize the profile of bulk concentration (C)

around probe zones under static or flow hybridization. In conventional microarray

hybridization, if ~100µL of target solution was applied to an area of 20 mm × 20

mm on a glass slide printed with probe arrays [19], the thickness of the liquid film

is calculated to be 0.25 mm. On the other hand, in microfluidic hybridization,

~1µL of target solution is confined to a microchannel and flows over the probe

arrays. The depth of the microchannel is usually less than 100 µm. In our study,

microchannel plates with average height of 30 µm have been used [94]. By

solving the proposed model numerically, the bulk concentration (C) profiles after

3-min static and microfluidic hybridization of oligonucleotide targets are shown in

Figure 8-4(a) and (b), respectively. From the color map, it can be seen that the

target DNA concentration are consumed to a different degree along the bottom

wall during static hybridization (Figure 8-4(a)). The calculated sample

concentration at bottom wall (Cw) adjacent to the no-probe regions drops from 10

nM to 8.2 nM. This decrease in bulk concentration is due to the non-specific

adsorption of target molecules on the glass surface. At the probe regions, the

hybridization process significantly depletes the Cw down to 1.2 nM. On the

contrary, microfluidic flows prevent localized target depletion around probes by

continuous replenishment with fresh sample solutions. The lowest concentration

of bulk solution at bottom wall (Cw) is 9.6 nM after 3-min hybridization. Higher

190

local target concentration results in faster hybridization rate. Figure 8-4(c) shows

the simulated signal-to-time curves for 15-min hybridization. The average signal

is almost 3 times as high as that obtained from the static method at 15 min.

Figure 8-4 (a) Concentration profile of static hybridization. (b) Concentration profile of

microflow hybridization. The average flow velocity is 1 mm/s and the flow direction is shown as a blue arrow. In both (a) and (b), an array of 3 probe spots has been incorporated in the model. The probe region width and spacing are at 200 µm. The sample concentrations at inlets are set to 10 nM and the hybridization time is 3 min.The bottom rainbow bars denote the bulk concentration scale. (c) Kinetic curves showing the change of the hybridization fraction vs. time during the static or microflow hybridization in 15 min.

8.5.2 Prediction of the detectable range of oligonucleotide targets

The detection range of DNA microarray analysis depends upon the

performance of both the scanner and the microarray sensor. Although modern

scanners usually have a 16-bit digitization in signal outputs (216 = 65,536), the

actual dynamic range is less than 5 orders [338]. It is because the detection

range is confined by the probe coverage (upper limit) as well as the fluorescent

background (lower limit) of the solid substrate. Since the background signals

could take 2 orders in the detection of Cy5-labelled targets, the actual range of

1.2 nM

Concentration map of static hybridization

Flow

10 nM

9.6 nM 10 nM

0.25 mm

0.03 mm

(a)

(b)

(c)

Concentration map of microfluidic hybridization

0 200 400 600 800

0.0

0.2

0.4

0.6

0.8

1.0

Hybr

idiza

tion

fract

ion

Time (Sec)

Microfluidic Static

191

fluorescence signal detection in the microarray scanner thus spans only 3 orders

of magnitude.

To examine the detectable range of the microarray, oligonucleotide targets

at 100, 10, 1, 0.1, and 0.01nM were prepared and hybridized to the probe arrays

through microchannels. The resulted fluorescence images are shown in Figure

8-5(a). It was found that the hybridization signals obtained from 1µL of 100-nM

target solution was almost saturated. Higher concentration did not give significant

increase in fluorescence intensities. On the other hand, 1 µL of 0.01-nM targets

could not be detected at the experimental conditions. The signal-to-noise ratio

from the hybridization of 1µL of 0.1-nM target solution is ~6, which is close to the

detection limit. Therefore, the dynamic range of the oligonucleotide analysis

spans 3 order of magnitudes from 0.1~100 nM.

The experimental data can be compared with the simulation results from

the proposed model. As shown in Figure 8-5(b), the calculated relative

hybridization fraction is close to saturation ( * 1Θ = ) for 1µL of 100-nM oligo

targets after 12-min hybridization. This fraction decreases at lower target

concentrations. For instance, the number of hybridized DNA from 1µL of 0.1-nM

and 0.01-nM targets is around 1/260 and 1/3200, respectively, of that from 100-

nM targets (saturated hybridization). Since the scanner cannot read signals less

than 1/1000 of the saturated value (3 orders), it is impossible to detect 1µL of

0.01-nM DNA samples under the current experimental conditions. To achieve a

lower detection limit in concentration, further improvement on scanner

performance, probe capacity or different detection labels might be used.

192

Figure 8-5 (a) The image of hybridization patches by flowing through 1-µL oligonucleotide samples at different concentrations in 12 min. The hybridization patches from two vertical repeated probe lines are shown. (b) The normalized hybridization signal intensities from the experimental results in (a). The error bars come from the standard deviation of 5 hybridization patches. The simulated hybridization signals are also shown for comparison.

8.5.3 Effects of flow rates and channel dimensions on hybridization signals

In solution-phase hybridization, the time needed to reach thermodynamic

equilibrium depends only on reaction kinetics. On the contrary, in microarray

analysis, since probe molecules are immobilized on a solid surface, target

molecules in bulk phase have to be transported to the surface probe region by

diffusion before hybridization reactions. This diffusion procedure has been

proven to be a limiting factor due to the slow diffusion of DNA molecules [72, 81,

84, 126, 303, 310, 317]. Microfluidic method introduces additional mass transport

through convection and thus enhances the hybridization signal. This advantage

could be evaluated by the dimensionless Péclet number (Pe), which is the ratio

between the characteristic convection rate and diffusion rate. When Pe>>1, the

convective transport dominates over diffusion and localized depletion is

0.0001

0.001

0.01

0.1

1

100 nM 10 nM 1 nM 0.1 nM 0.01 nM

Rel

ativ

e h

ybri

diz

atio

n r

atio

Target concentrations

Simulation

Experimental100 nM

10 nM

1 nM

0.1 nM

Oligo targets (a) (b)

193

eliminated, resulting in a fast mass transport. A high Pe number can be simply

achieved by flushing the target solution through the hybridization channel at a

high flow speed. However, at a fixed hybridization time, a faster flow will result in

a larger sample consumption compared to a slower flow, unless some

recirculation facilities are used to reduce the sample consumption [72-74]. On the

other hand, when the sample volume is fixed, the total residence time decreases

at high flow rates and the hybridization signal could be reduced. It is thus

important to find optimum flow rate for the quantification of microfluidic DNA

hybridization within small sample volumes.

We performed experiments at different flow rates to identify the effects of

mass transport on the hybridization events. The same amount of oligonucleotide

target solutions (1 µL) was introduced to the probe region by either stop flow or

dynamic flow. In the former method, hybridization time was set to 30 min, while

the actual hybridization time in the latter method depends on the flow rate used.

For instance, hybridization time of 1 µL at 0.1 µL/min was 10 min. The resulted

fluorescence images and signals are shown in Figure 8-6(a). As expected, it is

found that diffusion-limited stop-flow hybridization did not give a strong signal

even though the hybridization time is the longest. In terms of the dynamic

hybridization, although flow with the highest flow rate (2 µL/min) is least likely to

be diffusion-limited, it resulted in lowest hybridization signals. On the contrary,

flow with the lowest flow rate (0.1µL/min) in the experiment gave the highest

hybridization signals.

194

Figure 8-6 (a) The fluorescent image of microchannel hybridization results at different flow rates. Six flow rates, from left to right, 2.0, 1.0, 0.5, 0.3, 0.2, and 0.1 µL/min have been applied. The hybridization image in the right most is from 30-min stop-flow hybridization. 1 µL of 10-nM oligonucleotide samples were applied to each microchannel. (b) The relative hybridization intensities from the left image. The error bars are from the standard deviation of 5 hybridization patches.

These experimental results can be interpreted by the proposed model.

Four flow rates ranging from 0.01 to 10 µL/min were simulated and the

corresponding Pe values are from 10 to 10000. For 1 µL of sample solutions, the

hybridization time is calculated to be from 6 s to 6000 s, depending on the flow

rates. The stop-flow hybridization was also studied for comparison. Figure 8-7(a)

shows the average bulk concentration of targets adjacent to the probe regions

(Cw*) at the end of hybridization. It was found that the stop-flow method did

cause a significant local depletion of target DNA at the probe region. The relative

bulk sample concentration (Cw*) decreased to ~0.003 after 30-min hybridization.

On the other hand, even at the lowest flow speed (Pe = 10) it is sufficient to

eliminate the depletion, and the relative target concentration above the probe

0.0

0.2

0.4

0.6

0.8

1.0

2.0 1.0 0.5 0.3 0.2 0.1 Stop

Flow rate (µL/min)

No

rma

lize

d h

ybri

diz

ati

on

inte

nsi

ty

(a) (b)

Sam

ple

flow

dire

ctio

n

Flow rate (µL/min) 2.0 1.0 0.5 0.3 0.2 0.1 Stop

195

region is close to 0.75. In the other cases, although the flow speed is increased

up to 1, 000 times, this local concentration does not vary much and it goes up to

only 0.88 at Pe = 10, 000. Therefore, diffusion is not a limiting factor anymore

when Pe >10 and the hybridization signals depend on reaction kinetics instead.

Figure 8-7(b) shows the simulated hybridization fractions ( *Θ ) at different

flow rates. Because the sample volume is fixed to 1 µL, the reaction time is

inversely proportional to the delivery speed. Similar to the experimental data,

longer time results in higher signals and the highest hybridization fraction is

found at the lowest flow speed (Pe = 10). Nevertheless, the assay time is too

long in this case (~6000 s) and saturation occurs at ~2000 s; this approach could

be impractical for large sample volumes [339]. It is also found that the benefit on

hybridization fraction from slower flow velocity shows marginal improvement. For

example, a 10-fold decrease in the delivery speed from Pe = 10, 000 to Pe = 1,

000 results in an almost 10-fold increase in the hybridization signal. However, if

the flow speed decreases at the same ratio from Pe = 100 to Pe = 10, the

increase in the hybridization signal is less than 2 folds and this is at the expense

of 10-fold reaction time. Therefore, optimized flow speed with the consideration

of both the reaction time and signal intensities can be found around Pe = 100,

where the average flow speed is between 0.1~1 mm/s and the hybridization time

is at the level of 10 min with the 30-µm deep microchannel used in the

experiment .

196

Figure 8-7 (a) The simulated results of the relative bulk target concentrations adjacent to the probe regions. (b) The simulated hybridization kinetics at different flow conditions. The hybridization time for the stop-flow method is 30-min. In the dynamic flow method, four flow speeds, from left to right, 0.01 to 1 mm/s have been simulated and the corresponding Pe values are from 10 to 10000. 1 µL of 10-nM oligonucleotide samples were applied to each microchannel.

Reduction of channel height is another strategy to efficiently deliver target

molecules to the surface probes [134, 326, 340, 341]. At same flow speeds,

shallower microchannel results in a longer residence time as compared to deeper

channels. The increase of hybridization signals is thus expected due to longer

reaction time in the shallower microchannels. On the other hand, if the total

residence time are controlled to be the same for both types of microchannels, the

flow speed in shallower microchannels is thus higher, which results in an

increase in the convective flux (from higher flow speed). Moreover, the

characteristic diffusion time of target molecules in shallow channels is less than

that in deep channels, which will also benefit the surface hybridization. As seen

from the simulation results in Figure 8-8(a), the simulated hybridization signal

increases 17% by decreasing the channel depth from 75 µm to 24 µm.

(a) (b)

0.1 1 10 100 10001E-4

1E-3

0.01

0.1

1

Rel

ativ

e hy

brid

izat

ion

fract

ion

Time (Sec)

Pe = 10 Pe = 100 Pe = 1000 Pe = 10000

0.001

0.01

0.1

1

Stopflow

Pe =10

Pe =100

Pe =1000

Pe =10000

Bu

lk c

on

cen

tra

tio

n a

t w

all

(Cw

*)

197

Compared to the experimental findings in Figure 8-8(b), the increase of

hybridization signals from shallower microchannels is even higher at ~50%. The

difference could be due to the further improvement from the stabilization effect on

hybridized DNA induced by microflow, i.e. the increase of flow speed might

benefit duplex formation [87, 342].

On the other hand, the reduction of channel or chamber height shows

adverse effect on static microarray hybridization. From the simulation results

shown in Figure 8-8(a), the hybridization signal decreases ~75% with the shallow

microchannel when using the stop-flow method. This is understandable because

less sample solutions are accommodated at the probe region and the local

depletion around the sensing area is even worse for shallower channels.

Figure 8-8 (a) The simulated hybridization intensities in microchannels of 2 different

depths. Here, 2 types of microchannels, 75 µm and 24 µm, are simulated and the hybridization time is 3 min in all cases. The Pe number of both types of microchannels is 1000. In stop-flow method, the hybridization time is set to 30 min. (b) Experimental results from flow hybridization of 1-µL samples in the 2 types of microchannels.

(a) (b)

0

1000

2000

3000

4000

75-um 24-um

Rel

ativ

e Fl

uo

resc

ence

Un

it

0

0.1

0.2

0.3

0.4

0.5

75-µm 24-µm

Rel

ativ

e h

ybri

diz

atio

n f

ract

ion Dynamic

Stop-flow

198

8.5.4 Effect of probe coverage on hybridization signals

In DNA microarray analysis, the probe DNA molecules are immobilized on

a glass surface, and then heterogeneous hybridization occurs when targets in

bulk phase hybridize with surface-immobilized probes. The hybridization signals

depend on the probe coverage in two ways. First, the maximal amount of

captured targets cannot exceed the amount of probe molecules at the sensing

region. Therefore, the higher is the probe density, the higher is the hybridization

capacity as well as the maximum hybridization signals. Secondly, the kinetics of

target capture process is affected by probe coverage. The current view of duplex

formation for short oligonucleotides is nucleation first, followed by helix zipping

[265]. For the target molecule to hybridize at the surface, it has to first penetrate

into the probe DNA film. Therefore, it is possible that lateral interactions with

nearby probe DNA molecules affect the kinetics due to steric constraints. In fact,

surface hybridization of tightly-packed probe molecules has been reported with

less efficiency than loosely-packed probes [112, 221, 335].

The effect of probe density on hybridization signals was evaluated with the

proposed model. Here, an equation proposed by Chan et al. was used to

estimate the relationship between probe coverage and the kinetic constant of

target DNA in bulk solution hybridized to surface immobilized probe molecules

[305]. The equation is shown as follows:

23

33

A pDN Rk

π χσ

= (32)

199

where NA is Avogadro constant, χ3 is the reaction success probability and it is

chosen as 0.001 [305], σ3 is the persistence length of the target DNA, and Rp is

the radius each probe molecule occupied. If we assume that there is no overlap

of probe molecules in the sensing region, Rp can be calculated using the

following equation:

0

1 12p

A

RN

=Θ

(33)

The higher is the density of the packed surface probes; the lower is the Rp value.

For the same DNA targets, the kinetic constant of surface hybridization (k3) is

inversely proportional to the density of surface probes. This relationship has been

confirmed experimentally by Michael et al.’s work, where they found that with

probe coverage decreased from 5.2 × 1011 to 1.38 × 1011 strands/cm2, the

hybridization kinetic constant increases almost 4 times from 0.9 × 105 to 4.3 ×

105 M-1·s-1 [335]. On the other hand, the reaction rate of denaturation did not

change significantly with the change of the probe coverage from the experimental

findings in Michel et al.’s work [335]. Therefore, an assumption that the k-3

remains constant for different probe densities is reasonable and is used in our

subsequent simulations.

Figure 8-9(a) shows both simulated and experimental results of the probe

coverage effect. As depicted in our previous work, the probe coverage on a glass

slide was controlled through using different concentrations of probe solutions

during spotting procedures [94]. The resulted probe densities vary from 2.3 ×

1011 to 7.9 × 1011 strands/cm2, which are comparable to the values obtained in

200

the studies by other groups [139, 221, 250, 251, 327, 328]. As seen in Figure

8-9(a), although the probe coverage increased almost 4 times, the hybridization

intensities from experimental measurements increased only around 1.5 times.

Moreover, no significant increase of hybridization signals was observed when

probe density is higher than 5.4 × 1011 strands/cm2. The results from simulation

are comparable to the experimental findings. The hybridization intensity

increased around 2 times and it reached a plateau after the probe density is

higher than 4.7 × 1011 strands/cm2.

The effect of probe coverage on hybridization is also described by the

simulated kinetic data. Because the kinetic constant of surface hybridization (k3)

is inversely proportional to probe densities, the lowest probe coverage resulted in

the fastest hybridization. As shown in Figure 8-9(b), the hybridization ratio ( *Θ ) is

saturated in the given time when the probe coverage is the lowest at 2.3 × 1011

strands/cm2 and the *Θ decreases significantly with the increasing probe density.

On the other hand, this dimensionless hybridization ratio is just a relative fraction

and it reflects the ratio between the hybridized probe molecules with the initial

number of the probe molecules ( *0Θ = Θ Θ ). To compare the hybridization

signals at different probe coverage, one has to multiply the ratio with the

corresponding initial probe density. Therefore, higher probe coverage can

provide higher surface concentration of un-hybridized probe molecules, which

could counteract the effect from lower kinetic constant, and the plateau shape in

Figure 8-9(a) is thus formed. The signals could be further enhanced when longer

hybridization time is used, as shown in Figure 8-9(b).

201

From Figure 8-9(a), it is also found that the simulated results give a higher

hybridization ratio compared to the experimental findings. This could be

attributed to the setting of the reaction success probability, χ3, in equation (42).

Though it is chosen as 0.001, it could be lower for diluted target solutions [305].

Figure 8-9 (a) The experimental and simulation results of the effect of probe coverage on hybridization signals. Top curve with triangle labels shows the simulated hybridization signals at different probe coverage; Bottom curve with square labels shows the experimental hybridization signals at different probe coverage. (b) The simulated kinetic curves of hybridization at different probe densities.

8.5.5 Assay conditions on the improvement of signal intensity and specificity/selectivity

In conventional microarray analysis, target solutions are usually incubated

with the surface-bound probes for static hybridization in 12-18 hours [19]. The

assumption behind this protocol is that the hybridization time is sufficiently long

and a thermodynamic equilibrium can be considered to be attained [343, 344]. In

this case, when non-specific adsorption and diffusion limitation of target

0 30 60 90 120 150 180

0.0

0.2

0.4

0.6

0.8

1.0

Rel

ativ

e hy

brid

izat

ion

ratio

Time (Sec)

2.3E+11 4.7E+11 5.4E+11 5.8E+11 6.5E+11 7.9E+11

0.5

1.0

1.5

2.0

2.5

2.3 4.7 5.4 5.8 6.5 7.9Probe density (×1011 Strands/cm2)

No

rma

lize

d h

yb

rid

iza

tio

nra

tio

s

(a) (b) Probe density (strands/cm2)

Experimental hybridization signals

Simulated hybridization signals

202

sequences are neglected, the equilibrium fraction of hybridized duplexes ( *eqΘ )

can be described by the Langmuir model [345]:

*'

0 3 3 0 3

1 11 1

eqeq k k C K−

ΘΘ = = =

Θ + + (34)

where '3K is the dimensionless equilibrium dissociation constant define in

equation (26). It is obvious that the more stable is a DNA duplex (lower '3K ), the

more is the amount of target sequences hybridized on a probe spot ( eqΘ ). The

equilibrium constant is calculated from the Gibbs free energy of duplex formation

(ΔG) via ( )31 expK G RT= −∆ and ΔG has been quantitatively described by the

nearest-neighbour model [30, 346-348]. In DNA microarray applications, it is

common that multiple probes and/or multiple samples are involved on one chip.

Non-match and mismatch (MM) duplexes are usually less stable than perfect-

matched (PM) duplexes due to the less number of hydrogen bond interactions

(as seen in Figure 8-10). These duplexes can be discriminated from PM

duplexes by the comparison of the signal intensities, which are linked to the

hybridization fractions (Θ ) at the probe regions.

203

Figure 8-10 Sequence structures showing (a) a perfect-matched duplex (B-B’) and (b) a one-base-pair-mismatched duplex (B-NB’) from two 21-mer oligonucleotides.

Table 8-1 lists the thermodynamic data from 6 groups of DNA duplexes.

Each group includes a 21-bp perfect-matched duplex and a similar one with

single-base-pair-mismatch at the duplex center. The central three base pairs of

the mismatched duplexes are listed in the table to show the mismatch types. The

thermodynamic data are calculated based on the nearest-neighbour model [30,

349]. The equilibrium dissociation constants (K3’) of mismatched and perfect-

matched duplexes are compared by ratios. From the thermodynamic data it can

be seen that the most stable molecules among the 12 DNAs is the perfect-

matched J duplex (ΔG = -91.2 kJ/mol) while the mismatched L duplex shows the

biggest free-energy difference (The equilibrium constant ratio is 577) from its

perfect-matched fellows.

(b) (a)

204

Table 8-1 The calculated free energies of 6 groups of DNA duplexes.

Duplex name

Mismatched part for MM duplexes

ΔG of PM duplex

(kJ/mol)*

ΔΔG of PM and MM

duplexes (kJ/mol)**

'3'3

(MM)(PM)

KK

B ATA:TTT -71.5 10.04 32

J TCA:ACT -91.2 15.89 242

K GGG:CTC -76.9 7.11 12

L TAC:ACG -84.1 18.4 577

M GGG:CTC -79.5 7.11 12

N CCC:GAG -82.8 14.22 136

* Calculated at 50°C and in 0.4M NaCl using the DINAMelt Web Server [349]. ** ΔΔG = (ΔG of MM duplexes) – (ΔG of PM duplexes).

In the microfluidic method, the equilibrium time of surface hybridization

has been reduced since the diffusion limitation of target molecules is eliminated.

However, the time required to reach thermal equilibrium may still exceed the

assay time, e.g. 3 min. In this situation, the discrimination ratio between perfect-

matched (PM) and mismatched (MM) DNA may not be predicted accurately by

the comparison of ΔG values. For instance, the experimental results of the above

6 groups of DNA duplexes from 3-min microfluidic hybridization (shown in Figure

8-11) are not consistent with the free energy data. The most stable DNA (duplex

J) did not appear as the one with the highest hybridization intensity. Despite that

the perfect-matched B duplex has a similar free energy (ΔG = -71.5 kJ/mol) with

205

that of the mismatched M duplex (ΔG = -72.4 kJ/mol), the hybridization signal

from the target B DNA is almost 10 times of that from the target M DNA.

Furthermore, the discrimination ratio is not related to the K3’ ratio. For

example, the experimental findings do not support the expectation that the

PM/MM duplexes of N should have a better discrimination ratio, as compared

with the PM/MM duplexes of B and K. Moreover, the signal intensities for L

duplexes are considerably low and there is even a false discrimination for the

mismatched duplexes. The experimental results indicate that the hybridization

efficiency and discrimination ratio of DNA duplexes cannot be reliably predicted

with the calculated thermodynamic equilibrium data.

Figure 8-11 The experimental hybridization results from the microfluidic hybridization of 6 groups of DNA. Each group represents two targets with one-base difference at the sequence center and the probe molecules are complementary to one of them. The 12 DNA solutions were hybridized in separated microchannels but under the same conditions at ~50°C and in 2.5x SSC for 3 min.

0.0

0.2

0.4

0.6

0.8

1.0

B J K L M NDifferent duplexes

Nor

mal

ized

hyb

ridi

zati

onin

ten

sity

Mismatched duplexes

Perfect-matched duplexes

206

For fast DNA analysis on microfluidic chips, kinetic constants for both

forward bulk-to-surface hybridization (k3) and reverse denaturation (k-3) dominate

signal intensity and selectivity. With the proposed model, we evaluate the effect

of these two kinetic constants on the intensity and selectivity of microarray

hybridization. As given in equation (26), k3 has been nondimensionalized to a

Damköhler number (Da3), to compare the effect from the duplex dissociation, k-3

is also non-dimensionalized to a dimensionless Damköhler number, Dn, and it is

a ratio of characteristic diffusion time and duplex dissociation time.

3 0

0

Characteristic diffusion timeCharacteristic duplex dissociation time

k HDnC D− Θ= = (35)

The dimensionless dissociation equilibrium constant ( 3 'K ) is then obtained

by

' 33

3 0 3

k DnKk C Da

−= = (36)

The two Damköhler numbers (Da3 and Dn,) were arbitrarily assigned to 6

values ranging from 0.001 to 100 where most of the experimental values could

fall in this range. The hybridization time are set from 3 min to 60 min to mimic

typical microfluidic hybridization times. The results are shown in Figure 8-12.

207

Figure 8-12 The simulated effect of the 2 Damköhler numbers (Da3 and Dn) on relative hybridization fraction after (a) 3-min, (b) 15-min, and (c) 60-min microfluidic hybridization. The corresponding dimensionless dissociation equilibrium constants (K3') are also shown as dashed lines in the 3 graphs. The colorscale bar beside each graph denotes the level of the relative hybridization fraction.

As the value of K3’ (dimensionless equilibrium dissociation constant)

decreases from 1000 to 0.001 (as represented by the light-blue dash lines in the

figures), DNA duplex is more stable and the hybridization intensity increases in a

general trend (shown as colour maps). For a certain combination of Da3 and Dn

numbers, longer hybridization time usually results in higher signals, as seen from

Figure 8-12(a) to (c). Therefore, if only the signal intensity is considered, any

1E-3 0.01 0.1 1 10 1001E-3

0.01

0.1

1

10

100

DN n

umbe

r

Da number

0.00.10.20.30.40.50.60.70.80.91.0

1E-3 0.01 0.1 1 10 1001E-3

0.01

0.1

1

10

100DN

num

ber

Da number

0.00.10.20.30.40.50.60.70.80.91.0

1E-3 0.01 0.1 1 10 1001E-3

0.01

0.1

1

10

100

DN n

umbe

r

Da number

0.00.10.20.30.40.50.60.70.80.91.0

(a) (b)

t = 3 min t = 15 min

t = 60 min (c)

208

experimental condition that can promote duplex association (higher 3k ) and can

stabilize the DNA duplexes (lower 3k− ) will boost the hybridization fraction.

On the other hand, if multiple samples are to be hybridized to the same

probe arrays, it would be desirable if PM and MM DNA duplexes can be

discriminated from each other by the difference in signal intensities. For two

target sequences of completely different sequences, the difference in the free

energy of duplex formation with the same probe sequence is very large and it

has no problem in discrimination. However, for duplexes with only a couple of

mismatched base pairs, the difference of duplex formation free energy (ΔΔG)

from the perfect-matched ones could be as low as ~7 kJ/mol (as shown in Table

8-1) and the calculated equilibrium constants only differ by ~1 order of

magnitude. In this case, their thermodynamic behaviours alone are not sufficient

to produce the successful discrimination of these two sequences. For example, in

Figure 8-12(a), even when two duplexes with different dissociation equilibrium

constant (K3' = 0.001 and 0.01, respectively), the fraction of hybridization could

be similar at 0.2~0.3 after 3-min hybridization and it would be very hard to

discriminate between these two target sequences.

The selectivity between the PM and MM DNA duplexes could be improved

by including the hybridization kinetics during the study. During the short-term 3-

min hybridization, Da3 (duplex association kinetics) has more impact on

hybridization signals as compared to Dn (duplex dissociation kinetics). For

example in Figure 8-12(a), if the Da3 number is fixed at 1, the fraction of

hybridized molecules increases a little from 0.1 to 0.3 while the Dn number

209

decreases 6 order of magnitude from 100 to 0.001. On the contrary, if the Dn

number is fixed at 0.01, which is a common value for surface-bound DNA, the

fraction of hybridized molecules increases from 0.1 to saturated when the Da3

number increases for 2 orders from 1 to 100. The duplex dissociation kinetic

constant (Dn) starts to take effect in longer-term reactions such as 15-min and

60-min hybridization (Figure 8-12(b) and (c)). During DNA microarray analysis,

experimental conditions such as pH, ionic strength, additives, temperature, and

sequence compositions show different impacts on the two kinetic rate constants

(Da3 and Dn) [10, 16, 350]. By carefully choosing the combination of assay

conditions, discrimination between the perfect-matched and mismatched targets

could be achieved.

Another condition to achieve mismatch discrimination and to get rid of

non-specific adsorption before final image analysis is to rinse the hybridized

spots with buffers of lower ionic strength [19]. The proposed model confirms the

importance of this washing step in the discrimination of PM/MM duplexes and in

the alleviation of non-specific cross hybridization. Here, a different initial condition

was applied in which the probe spot was fully occupied by target molecules first

and blank buffer solutions (C0 = 0) were then introduced. Because the

microfluidic flow continuously replenishes the channel, all the dissociated

molecules will be removed immediately from the probe region. Figure 8-13

depicts the kinetics curves of this washing step, where the effect of different Dn

values is compared. The Dn number of the top curve (Dn = 0.1) is a typical value

of perfect-matched 20-mer duplexes. The simulation of the rest of the curves

210

represents the washing procedure on the less stable duplexes with higher Dn

values. These values were arbitrarily assigned with respect to the top one from

perfect-matched duplexes. It can be seen that washing with blank buffer did not

lead to substantial loss of the PM duplexes. After 3-min washing, the duplex

fraction is 95% of the initial value. However, the signal loss is more significant for

less stable targets. For the dissociation constant of 10-times higher than that of

PM duplexes (Dn = 1), 3-min washing results in 30% of signal loss. Since the

dissociation kinetic constant of the non-specific bound targets is usually 100-fold

of that of the perfect-matched ones [16, 306], the microfluidic method can

effectively wash away the non-specific bound DNA molecules in a short period.

Figure 8-13 The kinetic curves of the microchannel washing procedure at different dissociation constants. In all the cases the Da values were set to 1, which is a typical value for 20-mer duplex formation.

8.5.6 Spot morphology

The quantification of DNA microarray hybridization results relies on the

image analysis of the hybridization spots on microarrays. For automated

microarray image processing, an “ideal” microarray image is usually

0 200 400 600

0.0

0.2

0.4

0.6

0.8

1.0

Rela

tive

hybr

idiza

tion

fract

ion

Washing time (Sec)

Dn = 0.1 Dn = 0.2 Dn = 0.5 Dn = 1 Dn = 2 Dn = 5 Dn = 10

211

characterized by a low background intensity with minimal artifacts, pre-defined

spot shape (morphology), and distinct spot intensity with minimal uncertainty and

minimal deviation between repeated spots [351].

As shown in Figure 8-14(a), one of the irregularities often seen in pin-

spotted microarray images is donut-shaped spots with the signal intensity high at

the spot edge and low at the spot centre [352, 353]. On the other hand, when

microchannel probe spotting and intersection approach for hybridization are

used, this irregularity appears as sandwich shapes for rectangular spots (Figure

8-14(b)). The presence of a donut or sandwich-shaped intensity pattern leads to

high variation per spot and creates extra difficulties in subsequent image analysis

[354].

Figure 8-14 (a) Top: The fluorescence image of the hybridization results from pin-spotted DNA microarray. Bottom: The fluorescence intensity along the yellow dashed line in the top image. (b) Top: The fluorescence image of the stop-flow hybridization results from channel-spotted DNA microarray. Here, the sample channel was in horizontal and crossed two vertical probe lines. Bottom: The fluorescence intensity along the yellow dashed line in the top image.

0

2000

4000

6000

8000

10000

Distance

Rel

ativ

e fl

uor

esce

nt

un

it

0

500

1000

1500

2000

Distance

Rel

ativ

e fl

uor

esce

nt

un

it

(a) (b)

1 mm 1 mm

212

For pin-spotted DNA microarray, one possible reason of the donut-shaped

spots is the fast evaporation process and the fixation of the contact line during

probe spotting stage [355, 356]. Microfluidic spotting method effectively

circumvents the above problems because the probe solution is confined within

microchannels and it is washed away after immobilization without dry-out.

Nevertheless, we still observed sandwich-shaped spots in stop-flow hybridization

in microchannels as seen in Figure 8-14(b).

Another cause for the occurrence of donut patterns on DNA microarrays is

diffusion limitation [352]. During static hybridization in microchannels, target

molecules diffuse from bulk solution to the probe surface in both vertical and

horizontal directions. Because the ratio of channel length and height is much

higher than 1, the sample molecules in bulk solution right above the spot surface

will be depleted in a short time (Figure 8-4(a)) and most of the target molecules

approach the probe spot from both sides. Before moving over the spot surface

and reaching the center of the spot, these molecules will have more chances to

bind to the edges of the spot and hence the donut/sandwich effect. The dynamic

hybridization method promotes mass transport by introducing extra convection

flow. If target molecules can be replenished immediately by the convection flow

of bulk solution (As shown in Figure 8-4(b)), the non-homogeneous donut

patterns are less likely to be formed.

Dimensionless numbers can provide an estimation of the combined effect

from different process variables on hybridization in diffusion- and convection-

driven DNA microarray systems. The non-homogeneity pattern of hybridization

213

spots is due to the diffusion limitation of target molecules, and microfluidic

hybridization could improve the spot morphology through the convection effect.

To evaluate the minimal convection rate needed to achieve a homogeneous

hybridization spot, we combine the Da number and Pe number together to get a

dimensionless ratio of the characteristic convection rate to the hybridization rate,

i.e.,

3 0

Characteristic convection rateCharacteristic hybridization rate

Pe UDa k

= =Θ

(37)

With the proposed model, we simulated the profiles of bulk target

concentration at the bottom wall (Cw*) as well as the hybridization signal along

the probe region. As shown in Figure 8-15(a), the profiles of bulk concentration at

bottom wall for static hybridization is reversely bell-shaped. The target DNA

molecules are almost depleted near the center of probe region. Due to the non-

specific adsorption, the bulk concentration at the rest part of the bottom wall is

also less than the initial concentration. On the other hand, the target

concentration profile for dynamic hybridization is asymmetric. At the left of the

probe region, the bulk concentration is very close to the initial value because of

the replenishment from the convection flow. The target concentration at bottom

wall drops along the probe region due to the hybridization event and is recovered

later at the no-probe region. The lowest target bottom concentration occurs at the

right edge of the probe region and the degree of the concentration drop depends

on the ratio between the convection rate and hybridization rate. When the flow is

fast enough, Pe/Da>1000, which corresponds to a flow speed at ~1mm/s for

214

oligonucleotide sample in the 30-µm microchannel, the target concentration drop

is less than 10% from the initial bulk concentration.

Figure 8-15(b) shows the hybridization fractions ( *Θ ) along the probe

region from static and dynamic hybridizations with different Pe/Da ratios. The

relative standard deviation (RSD) of the hybridization signal across the probe

region in static hybridization is near 40% and a donut-shape spot is thus

expectable. For dynamic hybridization method, the asymmetric distribution of the

bulk concentration at the bottom wall results in asymmetric hybridization fractions

across the probe region. We have observed this type of hybridization spot during

experiments at slower flow rates. Under faster flow conditions, such as Pe/Da >

1000, this non-homogeneity can be considered as negligible in subsequent

image analysis because the RSD of the hybridization signal change across a

hybridization spot is less than 5%.

215

Figure 8-15 (a) The simulated relative bulk concentration profile of target DNA at the bottom wall with a probe spot. Here, both static and dynamic hybridizations are simulated. For the latter method, 5 different flow rates were tested resulted in 5 Pe/Da ratios. (b) The simulated hybridization fraction profiles at the bottom wall with different hybridization methods. In all cases, the hybridization time is set to 3 min.

The non-homogeneous patterns could be eliminated when using PCR

products or narrow probe lines. Compared to the short oligonucleotides, the

hybridization kinetics of longer PCR amplicons is slower and the Pe/Da ratio is

thus higher. Our experimental data have confirmed that the sandwich pattern has

not been found for PCR targets under stringent hybridization conditions even

using the stop-flow method. In addition, the narrower probe lines also reduce the

non-homogeneity of the hybridization spots. Figure 8-16 compared the simulated

hybridization results on 5 probe lines with widths ranging from 50 µm to 1mm.

For static hybridization, if the probe lines are wider than 100 µm, the non-

homogeneity of hybridization spot is significant (Figure 8-16(a)). However, when

the probe line is as narrow as 50 µm, the signals along the probe region become

uniform and the RSD of hybridization intensities is less than 5%. The simulation

(a) (b)

0.0

0.2

0.4

0.6

0.8

1.0R

elat

ive

bulk

con

cent

ratio

n a

t bot

tom

wal

l

Distance (mm)

Pe/Da = 10000 Pe/Da = 1000 Pe/Da = 100 Pe/Da = 10 Static hybridization 0.0

0.1

0.2

0.3

0.4

0.5

Rel

ativ

e hy

brid

izat

ion

fract

ion

Distance (mm)

Pe/Da = 10000 Pe/Da = 1000 Pe/Da = 100 Pe/Da = 10 Static hybridization

216

results have been partially validated by the experimental data, where 50-µm

probe lines gave homogeneous hybridization patch after stop-flow hybridization

(Figure 8-16(a) insets) and 1-mm probe lines showed sandwich-like patterns

(Figure 8-14(b)). By further calculation, it is found that the diffusion time for a

target molecule moving over the probe spot is comparable to its characteristic

hybridization time, leading to a more uniform hybridization.

For dynamic hybridization, although the signals are asymmetric along the

probe region, the variation is not as high as that of the static hybridization. In all

the cases, the RSD of the hybridization signal change across a hybridization spot

is less than 10% (Figure 8-16(b)).

Figure 8-16 The simulated hybridization fractions along the bottom wall with a probe region. The width of the probe region varies from 50 µm to 1mm. (a) Static hybridization. The insets show the experimental results from the hybridization of oligonucleotide samples with complementary probe lines with a width of 50 µm. The concentration of the oligonucleotide samples is 100, 10, and 1 nM, respectively, from left to right. (b) Dynamic hybridization.

Rel

ativ

e hy

brid

izat

ion

sign

al in

tens

ities

Cross distance (mm)

Rel

ativ

e hy

brid

izat

ion

sign

al in

tens

ities

Cross distance (mm)

(a) (b)

300 µm

200 µm

100 µm

50 µm

1 mm

300 µm

200 µm

100 µm

50 µm

1 mm

217

8.5.7 Signal variation from spot to spot

During DNA microarray fabrication and analysis, small defects on the chip

substrate such as surface scratches or particles may appear, which results in

fake signals and interferes with subsequent spot quantitation [19, 65]. Therefore,

it is necessary to use repeated probe spots/lines to correct for these possible

errors. In practice, these spots with the same type and density of probe

sequences are usually next to each other in geometry for the ease of image

analysis. However, if they compete for the target molecules in an unequal way,

hybridization performance on these repeated spots are no longer equal and the

resulted signals are thus biased. For example, during the static hybridization, a

depletion layer around the probe spot evolves as time elapses and the

concentration of target molecules in this layer is much smaller than that in bulk

solutions, as depicted in the previous section. If the repeated probe spots are

very closed to each other, the depletion layer could eventually override the

spacing between probe spots (as shown in Figure 8-17(a) and (b)). For the spots

at the center of the probe groups, the average hybridization signal will be lower

than those at the rim especially after long-term incubation (Figure 8-17(c)).

218

Figure 8-17 Simulated graphs showing the evolvement of bulk concentration profiles with a group of 3 probe spots. (a) After 1-min static hybridization. (b) After 15-min hybridization. (c) Simulation results showing the evolvement of the discrepancy of hybridization fractions on the 3 probe spots at different time.

The signal variation from repeated probe spots appears in a different way

for microfluidic dynamic hybridization because the localized depletion around the

probe groups is negligible as depicted previously. Here, an extreme condition

was tested where 10 repeated probe lines were hybridized with oligonucleotide

samples of different concentrations. From the experimental results shown in

Figure 8-18, it was found that the signal intensities did not vary considerably for

the sample prepared in low ionic strength or stringent buffer solutions (1xSSC

buffer). For example, the RSD of the signals from 10 hybridization spots is 11%

and 7%, for 100-nM and 10-nM targets prepared in 1xSSC, respectively.

However, in a less stringent buffer (4xSSC), if the target concentration is less

than 10 nM, a unidirectional decrease in signal intensity along the flow path was

0.0

0.1

0.2

0.3

0.4

0.5

1 2 3

Re

lati

ve h

ybri

diz

ati

on

fra

ctio

n

1 min3 min5 min10 min15 min

Static hybridization at 1 min

Static hybridization at 15 min

1.1 nM 6.6 nM

(a)

(b)

(c)

2.0 nM 10 nM

219

observed and the variation from spot to spot is as high as 20%. The experimental

data can be interpreted by the model.

It is well known that DNA-DNA hybridization is highly dependent on the

ionic strength of bulk solutions as measured by [Na+] [12, 16]. High ionic strength

buffer is usually considered as less stringent and results in higher hybridization

signals. Moreover, Studier et al. have found that the hybridization rate in low-salt

concentration (<1.0M) is proportional to the cube of [Na+] [12]. The dependence

of the hybridization rate becomes less steep in higher salt concentration.

Because the salt concentration in our work ranges from 0.15M NaCl to 0.6M

NaCl, this cubic relation of [Na+] has been used during simulation. As seen in the

simulated concentration profile in Figure 8-18(c), fast hybridization kinetics in

high ionic strength buffer significantly depletes the target concentration in the

liquid front and the bulk concentration decreases along the flow path. Because

the probe spots at the right keep reacting with targets of lower bulk

concentration, the hybridization signals are systematically lower than those at the

left. The position-related hybridization signal could have an adverse effect on the

comparison of signals from different samples. Therefore, low stringent or high

ionic strength condition should be avoided especially in the discrimination of

short oligonucleotides using the microfluidic approach.

220

Figure 8-18 The experimental results from the microfluidic hybridization of oligonucleotide samples at 4 concentrations. (a) Samples are prepared with 1x SSC. (b) Samples are prepared with 4xSSC. 0.1-nM targets were not detected in (a). The bottom numbers represent the probe row number. (c) and (d) show the simulated profile of bulk concentration during hybridization in 4x SSC and 1x SSC buffer, respectively.

Targets in 4xSSC

1.E+00

1.E+01

1.E+02

1.E+03

1.E+04

1.E+05

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10

Rel

ativ

e Fl

uore

senc

e U

nit

Targets in 1xSSC

1.E+00

1.E+01

1.E+02

1.E+03

1.E+04

1.E+05

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10

Rel

ativ

e Fl

uore

senc

e U

nit

(a) (b)

0 nM 10 nM Concentration map from microfluidic hybridization

Flow direction

Flow direction (c)

(d)

100 nM

10 nM

1 nM

0.1 nM

221

8.6 Conclusions

Our results highlight the importance of incorporating physiochemical

factors and microchip parameters in both the design and the data analysis of

microfluidic DNA microarrays. Through dimensional analysis and numerical

simulations, we demonstrate that, (i) Convective transport is an efficient way to

enhance the kinetics of diffusion-limited hybridization; (ii) The detection range of

a DNA microarray is predictable with the proposed model; (iii) Fast flow rate is

not necessary and a Pe number of 10 is high enough to eliminate the diffusion

limitation; (iv) Low probe coverage facilitates surface reaction but limits

hybridization capacity; (v) Thermodynamic data of DNA duplex formation cannot

be used directly for the prediction of hybridization fraction, especially for short

time flow hybridization; (vi) The hybridization spot morphology and the signal

variation from spot to spot depend on the microchannel geometry, flow

conditions, as well as hybridization kinetics. The proposed model could be

extended for the design and characterization of all relevant microfluidic device

prototypes involving surface reactions.

222

9: CONCLUSIONS AND PERSPECTIVES

9.1 Concluding remarks

Hybridization on DNA microarrays has been a versatile technique in

genomic research. Because of their flexibility and low production cost,

microarrays with low-density probes have been used in nucleic acid diagnostic

applications as well as single-nucleotide polymorphism detections. Recently,

microfluidic technology has been combined with DNA microarray method through

an intersection approach. Here, probe molecules are immobilized microfluidically

as probe lines on substrates and DNA samples hybridize at intersections with

orthogonally positioned microchannels. This intersection microfluidic microarray

method avoids the exact alignment of the hybridization microchannels and the

spot shape is more regular for later image analysis. The method shows the

advantages of less sample usage, fast reaction kinetics, as well as multiple

sample capability. Moreover, higher throughput analysis has been achieved by

the use of centrifugal pumping with a CD-like microchip. This method utilizes the

body force from liquid itself and it is free of additional solution interface contacts

such as from electrodes or syringes and tubing. Centrifugal force driven flow also

features the ease of parallel reactions. This thesis contributes to the application

of the microfluidic DNA microarray method under three different topics described

as follows.

223

9.1.1 Microfluidic surface hybridization

In Chapter 3, some fundamental aspects regarding microfluidic surface

hybridization were studied on rectangular chips. The probe densities on the

glass substrates at different spotting conditions were evaluated and the numbers

were applied to the later theoretical modeling. In Chapter 6, the hybridization

behaviours of DNA targets with two fluorescent labels, fluorescein and Cy5, were

studied and compared. It was found that sample DNA with Cy5 labels showed

less photo-bleaching effect as well as much lower detection limit. In Chapter 3,

integrated temperature control was realized with a Peltier chip and was used

successfully in mismatch DNA discrimination. In Chapter 4, the need of

temperature stringency during mismatch discrimination was further relieved by

pre-binding the target DNA with gold nanoparticles. This form of discrimination

has not been previously described in the microfluidics community.

9.1.2 Applications and modeling of the CD-like chip using centrifugal pumping flow

In previous work from our group, a CD-like chip format was invented

where centrifugal pumping was used in the intersecting microchannel

hybridization. Chapter 5 and Chapter 6 further developed the concept. The

quality of the CD-like chip was enhanced with improved manufacture processes,

where high-resolution photomasks were used and SU8 coated on a silicon wafer

directly acted as the molding master. The resulted polymer channels are

smoother with better sealing with the glass substrate. As a result, the flow rate

variation from channel to channel was improved and the liquid movement was

224

controlled at a slower spin rate. Moreover, it is also found the channel geometry

affects the hybridization and shallower channels gave higher hybridization

signals. With all these improvements, 3-min detection and discrimination of 1-

femtomol PCR amplicons were achieved on the microchip for the first time.

Chapter 7 describes the experimental measurement and mathematical

modelling of centrifugal-pumping flows in the CD-like microchip. It is

demonstrated that steady flow has been attained with the equiforce spiral

channel design. The flow velocities and residence time of solutions have been

predicted with both a mathematical model and computer simulation. In contrast to

radial channel designs, liquid flow in the spiral microchannels is more

manipulatable and the residence time of the liquid is much extended. With

modifications, this flow dynamics study could be adapted to the modeling and

simulation of other centrifugal-pumping microflow systems.

9.1.3 Modeling and simulation of microfluidic hybridization kinetics.

As a complement to the previous experimental investigations, Chapter 8

provides a theoretical modeling and simulation of microfluidic DNA microarray

hybridization kinetics. Four questions have been answered in this chapter. (1)

For a given amount of sample (usually at ~10 nM and with a volume as small as

1 µL) and chip configuration, what would be the optimal flow rate to achieve the

highest-possible hybridization fraction? (2) Lower probe coverage facilitates

hybridization kinetics but also decreases the hybridization capacity, what would

the balanced values be? (3) Can we use the thermodynamics data of solution

hybridization to predict microfluidic microarray hybridization behaviour? (4) What

225

are the reasons of the observed non-homogeneous hybridization spots and

signal variation from spot to spot? The simulation results provide useful device

design criteria for the experimentalist to predict and control the behaviour of a

microfluidic DNA microarray experiment given its geometry and operation

conditions. Moreover, the proposed model and simulations are transferable to the

analysis of other surface-based biosensors.

9.2 Outlook and research perspective

9.2.1 Sensitivity

The main aim of the project is to develop a robust biochip platform with

good sensitivity for fast DNA analysis. Signal enhancement has been achieved

by flow hybridization in microchannels. In addition, hybridization capacity and

kinetics are highly dependent on the surface chemistry parameters and target

accessibility to immobilized probes [112]. As compared to planar surfaces, we

have developed a submonolayer of gold nanoparticles on glass substrate for

probe immobilization and the hybridization of oligonucleotide targets was

enhanced [95]. Other methods to increase surface area are also available. For

instance, microbeads present a larger surface capacity for probe immobilization

as well as faster target hybridization kinetics [357]. The sensitivity of DNA

detection could be further improved by the introduction of microbeads to the

current system.

226

9.2.2 Selectivity

In terms of selectivity, the advantages of gold nanoparticles have been

demonstrated in the discrimination of perfect-matched and mismatched

duplexes. However, only central-mismatch duplexes have been examined so far.

Further experiments on the discrimination efficiency of gold nanoparticles on

different mismatch types are to be studied. Moreover, the effect of nanoparticle

sizes on the discrimination efficiency is also an interesting subject.

9.2.3 Integration and packaging.

Research in microfluidic technology often shares the profit from the

development of microelectronics industry. Without exception, the centrifugal

microfluidic platform aims to utilize the well-developed compact disk (CD) system

[147, 148, 280]. The rotor driven/control system of the optical drive could be

revised to adapt to the need of microflow pumping. The current bulky and

complex scanner could be replaced by the laser head, lens, and optical pick-ups,

although their movement have to be reprogrammed since the hybridization

patches are positioned along spiral or radial paths. Moreover, hybridization data

could be collected in a more efficient way. All these studies will satisfy the quest

for miniaturization, faster analysis and the highest possible degree of integration

and eventually lead to the advent of portable microfluidic devices for high

throughput DNA assays.

227

9.2.4 Modeling and simulation of centrifugal-pumping flows

During the mathematical modeling of the centrifugal pumping flows, the

analysis has been focused on the spiral part of the microchannels. Under a high

rotation speed, such as 2400 rpm, the difference between the predicted and

experimental values of flow velocity is not negligible. We speculate that this

difference can be attributed to the extra pressure from the neck channel outside

the spiral part. Future studies could thus be pursued on the problem, which

would be helpful in estimating the effect of the microchannel connection between

different geometries on microflow behaviours.

9.2.5 Modeling and simulation of hybridization kinetics

The proposed kinetics model in Chapter 8 can be further modified for the

microfluidic microarray hybridization of PCR products in which inter-conversion

among multiple components in bulk solutions is considered. Unlike single-

stranded short oligonucleotide samples, PCR products are double-stranded DNA

with a much longer length of 100~1000 bp. Before hybridization analysis, they

have to be denatured at 95°C to break most of the duplexes into two

complementary single-stranded DNA. During microarray hybridization, the PCR

products in bulk solutions could appear in two formats: linear or loose-coiled

single-stranded DNA (LC) and renatured double-stranded duplexes (DX). The

two formats are in equilibrium to each other, see equation (1). If there is any

secondary structure on the single-strand molecule, a supercoiled form (SC) will

possibly be produced, see equation (4) [358, 359]. Therefore, the bulk sample

solution of PCR products could be a mixture of three or more components rather

228

than a single component as in oligonucleotide samples. Obviously, only DNA in

the linear or loose-coiled single-stranded form could hybridize to the immobilized

probes. On the other hand, the complementary strands could compete with the

immobilized probe DNA for the labelled target strands. Moreover, the possible

secondary structure of the single-stranded target also decreases the effective

concentration of the target strands. Therefore, the surface DNA hybridization

model can be modified to reflect the above situations.

The kinetic model in the study assumes that these DNA forms interconvert

and hybridize according to the mechanisms given as below:

dx

dx

kcomp k

LC LC DX−

+

(1)

and the kinetic equation is

LCdx DX dx LC LC comp

dC k C k C Cdt − −= − (2)

where CLC and CLC-comp are the bulk concentration of the single-stranded

target DNA molecules and their complementary strands, respectively. kdx and k-dx

are the hybridization and denaturation rate constant of the self-hybridization

process in bulk phase, respectively. For symmetric PCR amplification, which has

been used in our experiments, CLC = CLC-comp = C, therefore,

2dx DX dx

dC k C k Cdt −= − (3)

For the formation of the secondary structure leading to the supercoiled

form,

229

sc

sc

k

kLC SC

−

(4)

the kinetic equation is

sc SC scdC k C k Cdt −= − (5)

where CSC is the bulk concentration of the super-coiled target DNA

molecules. ksc and k-sc are the formation and dissociation rate constant of the

supercoiled form in the bulk solution, respectively.

The overall diffusion-convection-reaction model in 2-D microchannel for

linear single-stranded target molecules is written as:

2 22

2 2 dx DX dx sc SC scC C C CD u k C k C k C k Ct x z x − −

∂ ∂ ∂ ∂= + − + − + − ∂ ∂ ∂ ∂

(6)

By coupling the above equation with the surface reaction models depicted

in Chapter 8, the hybridization fraction of PCR products from microfluidic

microarray method can be solved.

During the modeling and simulation of hybridization kinetics, literature

values of parameters have been used. However, many of them are measured

based on the solution reactions which are not necessary same as those of

surface-based reactions [16, 360]. Moreover, the kinetics constant of DNA

hybridization may vary with different sequences. It is thus necessary to pursue

experimental measurements to find out the values of these parameters under

microfluidic conditions. In addition, the effect of microflow on DNA conformation

230

and later hybridization has to be studied and to be incorporated in the future

model [87, 110].

231

10: APPENDIX

10.1 COMSOL simulation steps for the kinetics study of microfluidic microarray hybridization

In this section, we provide a step-by-step guide based on the simulation

work of Gervais et al. [312]. The simulations use the multiphysics capabilities

built in the COMSOL program to simulate time-dependent concentration profiles

in the presence of convection, diffusion and DNA surface reaction under arbitrary

conditions. Geometry coupling is used to link the reaction occurring at the

surface to the reaction in the bulk. The specific example was used to generate

the data used in the analysis performed in Chapter 8. This tutorial will work on

COMSOL versions 3.5a. It may not work on earlier versions and cannot be

guaranteed to work on more advanced versions of the program (COMSOL

version 4.0 and up).

The physical model solves the dimensionless equations system

summarized in Chapter 8. The convection-diffusion PDE in the bulk is couple to

the DNA surface reaction ODE though a flux boundary condition at the bottom

wall. The resulting system is nonlinear due to the multiplied surface and bulk

concentration in the surface reaction equation.

232

10.1.1 Channel geometry construction

To simulate transport and reaction at a channel surface one must define

two different geometries in COMSOL: bulk geometry and a surface one. Since in

all microfluidic channels studied in this thesis, the channel width is larger than its

height, we proceed to a simplified 2-D model assuming parallel plate geometry.

The detail steps are list as follows:

1. Open “Multiphysics\Model Navigator”;

2. Click on “Add geometry”. Add a 2-D geometry named “Bulk” with

independent variables Lx and Hy (in that order to yield a normalized

length variable x* and normalized height variable z*, respectively,

as depicted in Chapter 8);

3. Click on “Add geometry”. Add a 1D geometry named “Surface” with

independent variable Lx;

4. Click “OK” to close window;

5. Select the “Bulk” in the geometry tabs; Go to “Draw\Draw

Objects\Rectangle-Square”. Draw a rectangle of dimension 2 × 1

with lower left base point located at the origin (0, 0). The channel

we seek to model has normalized length / ( )dLx x U t= ⋅ from 0 to 2

and normalized height /Hy z H= from 0 to 1, where x and z are the

real device dimension coordinates; td is the characteristic time scale

( 2dt H D= ). We fix the aspect ratio of our simulations and adjust

233

the length scale with a scaling factor ( sf ) to avoid consuming too

much computing powers.

6. Go to “Draw\Specify Objects\Point”, draw six points with Lx

coordinate “0.6, 0.8, 1.0, 1.2, 1.4, 1.6”, respectively.

7. Similarly, select the 1-D “Surface” geometry tab and draw a 2 unit-

long line in 18 segments with Lx coordinate “0, 0.6, 0.8, 1.0, 1.2,

1.4, 1.6, 2.0”, respectively.

Figure 10-1 The 2-D geometry used in the COMSOL simulation.

10.1.2 Choosing the physics models:

After the geometries have been constructed, we choose the proper

physics models to include the physical principles.

8. In “Multiphysics\Model Navigator”, select for the 2-D geometry

“Chemical Engineering Module\ Mass Transport\ Convection and

Diffusion\Transient Analysis”. Rename the dependent variable

234

“Conc” for the normalized bulk concentration variable ( *C ) and click

“Add”;

9. In the 1-D geometry, create a transient diffusion module

(convection is not necessary at the surface) with dependent

variable “Theta” for the normalized surface concentration of

hybridized target DNA ( *Θ );

10. In the 1-D geometry, create another transient diffusion module with

dependent variable “Eta” for the normalized surface concentration

of non-specifically adsorbed target DNA ( *η );

11. Click “OK” and return to COMSOL's main window;

10.1.3 Domain Coupling:

The variable “Theta” and “Eta” in the 1-D geometry must be redefined in

the 2-D domain with a different name. Similarly the variable “Conc” must be

redefined in the 1-D domain to represent the concentration at the wall ( *wC ).

Because we assume that the PDMS walls (the upper and side walls) do not

adsorb or retain any DNA molecules, the coupling is only applied to the bottom

wall (the glass substrate). The procedure is the following:

12. While in the “Bulk” Geometry, select “Options\ Extrusion Coupling

Variables\ Boundary Variables” to bring up the extrusion variables

window;

235

13. On the boundaries corresponding to the bottom wall enter the

following:

On the lower wall, create the variable “ConcWall = Conc”,

Under the Destination tab, select the “surface” geometry as the

destination and the lower wall boundary as the origin,

Under the Source Vertices tab, select all the vertices in the

boundary,

Under the Destination Vertices tab, select all the vertices in the

domain to which it corresponds in the 1-D geometry.

14. Repeat step 12 in the “Surface” geometry and create the other

extrusion variables “ThetaSurf = Theta” and “EtaSurf = Eta” which

links all the surface concentrations in 1-D domain to the lower

boundary in the 2-D domain;

10.1.4 Constant and Variable Specifications

As depicted in Chapter 8, a series of parameters have been used to scale

the dimensionless variables and they are either from experimental

measurements or from reference values. During COMSOL simulation, these

parameters are entered as global constants. Moreover, the non-dimensionalized

parameters, such as Pe and Da numbers, are calculated from the constants and

are inputted as global expressions.

236

15. In “Options\constants” enter the problems experimental

parameters, one per line, within a consistent unit system. The

constants entered are global:

Name Expression Description

c0 1e-8[mol/L] Initial bulk concentration of analyte (10 nM)

cs0 8.3e-9[mol/L*mm] Total number of surface probes (on one probe spot)

D 5e-5[mm^2/s] Diffusion coefficient of 21-mer oligonucleotides

Ds 5e-13[m^2/s] Surface diffusion coefficient of 21-mer oligonucleotides

he 0.03[mm] Channel height

L 2[mm] Channel length

U 5[mm/s] Average fluidic velocity (from experiment)

Lgeom 2 Length of the geometry

k3f 7.5e4[L/(mol*s)] Hybridization rate constant of bulk to surface

k3r 2e-4[1/s] Dissociation rate constant

ka 3e3[L/(mol*s)] Adsorption rate constant

kd 0.3[1/s] Desorption rate constant

Rt 4[nm] Gyration radius of target oligonucleotides

Na 6.022e23[1/mol] Avogadro constant

237

16. Select “Options\Global Expressions “. Enter the variables that apply

to both bulk and the surface:

Name Expression Description

Pe U*he/D Péclet number

Da1 ka*camax*he/D Damköhler number of the target adsorption process.

Da2 k2*camax*he/(Ds/he) Damköhler number of the surface-to-surface hybridization process.

Da3 k3*cs0*he/D Damköhler number of the bulk-to-surface hybridization process.

td he^2/D Diffusion time scale

sf Lgeom*td*U/L Scaling factor

k2 (k3*phi/Rp)*(16/3*pi) Kinetic association constant for nonspecific adsorption of target strands.

k2r k3r Kinetic dissociation constant for nonspecific adsorption of target strands.

epsilon1 c0*he/camax Ratio of bulk-phase targets to maximum adsorbable targets.

epsilon2 cs0/camax Ratio of surface adsorption capacity relative to hybridization capacity.

epsilon3 c0*he/cs0 Ratio of bulk-phase target numbers to maximum hybridizable target numbers.

KD kd/ka/c0 Dimensionless dissociation equilibrium constant for nonspecific adsorption of targets.

K2 k2r/k2/camax Dimensionless dissociation equilibrium constant for surface-to-surface hybridization of targets.

K3 k3r/k3/c0 Dimensionless dissociation equilibrium constant for bulk-to-surface hybridization of targets.

phi Ds/D Ratio between the surface diffusion coefficient (Ds) and the bulk diffusion coefficient (D).

Rp sqrt(1/(Na*cs0))/2 Radius of each probe molecule occupied on the glass surface.

camax (1-pi*Rp^2*Na*cs0) /(pi*Na*Rt^2)

ηmax at probe regions on the bottom wall.

camaxx 1/(pi*Na*Rt^2) ηmax at no-probe regions on the bottom wall.

Da1x ka*camaxx*he/D Damköhler number of the target adsorption process at no-probe regions.

238

17. In the “Bulk” geometry tab, select “Options\Expressions\Scalar

Expressions”. Enter the variables that apply to the bulk

Name Expression Description

V 6*Hy*(1-Hy) Normalized flow velocity profile

Rabs epsilon1*Da1*(Conc*(1-EtaSurf)-KD*EtaSurf) Dimensionless adsorption rate at probe regions

Rbhy epsilon3*Da3*(Conc*(1-ThetaSurf)-K3*ThetaSurf) Dimensionless hybridization rate at probe regions

Rabsx epsilon1*Da1x*(Conc*(1-EtaSurf)-KD*EtaSurf) Dimensionless adsorption rate at no-probe regions

18. In the “Surface” geometry tab, select

“Options\Expressions\Subdomain Expressions”, At the domains

where probe spots locate, enter the surface reaction rate as seen

from the surface:

Name Expression Description

Rabs epsilon1*Da1*(ConcWall*(1-Eta)-KD*Eta) Dimensionless adsorption rate at probe regions

Rshy phi*epsilon2*Da2*(Eta*(1-Theta)-K2*Theta) Dimensionless hybridization rate at probe regions

Rbhy epsilon3*Da3*(ConcWall*(1-Theta)-K3*Theta) Dimensionless adsorption rate at no-probe regions

239

10.1.5 Subdomain equation settings:

The subdomain settings describe the physics on a model’s main domain,

which is divided into subdomains. We can set different values for coefficients that

define the PDE on the subdomain. Moreover, material properties are also defined

in the subdomains.

19. In “Multiphysics\Model Navigator”, select the concentration model

with variable “Conc”. In “Physics\Subdomain settings” enter the

following parameter (linked to the variables previously defined) in

the convection-diffusion equation (“Conc” tab):

Time-scaling coefficient: δts = td

Diffusion coefficient: D anisotropic = (sf/Pe)^2 0 0 1 (Since the

normalization constants are different axially and vertically, the

effective diffusion in the geometry becomes anisotropic),

Reaction rate : R = 0 (there is no reaction in the subdomain)

Lx-velocity: u = sf *V (accounts for the normalized length)

“init” tab, initial condition can be 0 (start from empty) or any other

value.

leave the “element” tab unchanged,

20. In “Multiphysics\Model Navigator”, select the concentration model

with variable “Theta”. In “Physics\Subdomain settings” enter the

following:

Time-scaling coefficient: ts= td (since = t/td)

Diffusion coefficient: D isotropic = 0,

240

Reaction rate : for probe regions, R = Rbhy+Rshy/epsilon2; for no-

probe regions, R = 0

Leave the other tabs unchanged.

The model computes the time dependent hybridization rate at the surface.

This rate will vary with Hy depending on the concentration right above it in the

bulk (Cw).

21. In “Multiphysics\Model Navigator”, select the concentration model

with variable “Eta”. In “Physics\Subdomain settings” enter the

following:

Time-scaling coefficient: ts= td (since = t/td)

Diffusion coefficient: D isotropic = phi*(sf/Pe)^2,

Reaction rate : for probe regions, R = Rabs-Rshy; for no-probe

regions, R = Rabs

Leave the other tabs unchanged.

The model computes the time dependent adsorption rate at the surface.

This rate will vary with z depending on the concentration right above it in the bulk

(Cw), surface diffusion coefficient (Ds), and probe coverage ( 0Θ ).

241

10.1.6 Boundary conditions settings:

Boundary conditions define the interface between the model geometry and

its surroundings. Different boundary conditions can be specified on different

boundaries.

22. For each of the models (variables: Conc, Theta, and Eta) enter the

boundary conditions in “Physics\boundary settings”:

a. for the “Conc” variable:

i. boundary (inlet): concentration c0= 1 or

insulation/symmetry for static hybridization,

ii. boundary (upper wall): flux = 0,

iii. boundary (lower wall with probe spots):

flux= -Rabs/epsilon1-Rbhy/epsilon3,

iv. boundary (lower wall without probe spots):

flux= - Rabsx/epsilon1,

v. boundary (outlet): convective flux.

b. for the “ Theta “ variable:

i. boundary (left vertex): flux=0,

ii. boundary (right vertex): flux=0.

c. for the “ Eta “ variable:

i. boundary (left vertex): flux=0,

242

ii. boundary (right vertex): flux=0.

10.1.7 Meshing

A mesh is a partition of the geometry model into small units of simple

shapes. It must be fine enough to resolve the sharp spatial distribution of

analytes that may occur near the reacting walls. In the bulk, some memory can

be saved by specifying a coarser mesh since the gradients present there are

lower than at the surface. We therefore propose the following meshing method:

23. Select “Mesh\Mesh parameters” from the main menu to bring up

the customized mesh window. For the “Bulk” geometry, enter the

following:

In the subdomain tab, enter a maximum element size of 0.05. This

size limits ensures that, for a geometry of height 1 as in the present

case, there will always be at least 20 grid points across it.

In the boundary tab, enter a maximum element size of 0.005 along

the reactive walls to ensure a finer characterization of these

regions.

Click on remesh to recreate the mesh. The default mesh growth

rates are sufficient in our case.

24. Select the “Surface” geometry and select Mesh\Mesh parameters”

from the main menu;

In the subdomain tab, enter a maximum element size of 0.004 for

all domains. This number of points at the surface will be sufficient to

243

match those specified at the boundary in the “Bulk” geometry. Here

a different number from 0.005 was used to avoid computational

noise in the simulation results.

10.1.8 Solver Parameters and Solver Manager

COMSOL Multiphysics includes a number of different solvers for PDE-

based problems. However, we do not need to select a solver explicitly in most

cases. Once the “Analysis Types” are selected as “Transient” based on our

convection-diffusion model, COMSOL will automatically choose a solver

compatible to our model. We then have to set the time duration and stepping for

the simulation studies.

25. In “solver\solver parameters”, select the “general” tab and specify

the following: Time stepping: times = range(0,1,180), for a 3-min

hybridization. Other tabs can be left as they are. Click “Ok” to return

to the main window.

26. In the “solver\solver manager” window, select the “initial value” tab

and specify: Initial value: “Initial value expression”, the rest can be

left untouched. In the “solve for” and “output” tabs, values can be

left as default (i.e. solves for all variables). In some particular

cases, it may save time to solve only for certain variables.

However, all coupled variables must always be solved together or

the solution will yield a trivial solution, Click “OK” to return to the

main menu.

244

10.1.9 Post-processing and visualization

COMSOL Multiphysics provides numerous post-processing and

visualization tools for the analysis of simulation results from the solvers. For

example, “Domain Plot” has been used to render the colormap of the bulk

concentration inside microchannels; “Cross-section Plot” was used to draw the

change of a variable along a specified line; the hybridization kinetics curve at a

spot was plotted with “Subdomain Integration” under a specified time range.

245

REFERENCE LIST

[1] J. D. Watson, and F. H. C. Crick, “Molecular structure of nucleic acids: A structure for deoxyribose nucleic acid,” Nature, 1953, 171, 737-738.

[2] K. Vlahovicek, L. Kajan, and S. Pongor, “DNA analysis servers: Plot.it, bend.it, model.it and IS,” Nucleic Acids Research, 2003, 31, 3686-3687.

[3] E. F. Pettersen, T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt, E. C. Meng, and T. E. Ferrin, “UCSF chimera: A visualization system for exploratory research and analysis,” Journal of Computational Chemistry, 2004, 25, 1605-1612.

[4] P. Yakovchuk, E. Protozanova, and M. D. Frank-Kamenetskii, “Base-stacking and base-pairing contributions into thermal stability of the DNA double helix,” Nucleic Acids Research, 2006, 34, 564-574.

[5] D. Voet, and J. G. Voet, Biochemistry, 3rd ed., New York: J. Wiley & Sons, 2004.

[6] V. A. Bloomfield, D. M. Crothers, and I. Tinoco, Nucleic acids: structures, properties, and functions, Sausalito, CA.: University Science Books, 2000.

[7] Y. Yin, and X. S. Zhao, “Kinetics and dynamics of DNA hybridization,” Accounts of Chemical Research, 2011, 44, 1172-1181.

[8] D. Porschke, and M. Eigen, “Co-operative non-enzymatic base recognition III. Kinetics of the helix-coil transition of the oligoribouridylic oligoriboadenylic acid system and of oligoriboadenylic acid alone at acidic pH,” Journal of Molecular Biology, 1971, 62, 361-364, IN311-IN315, 365-381.

[9] M. E. Craig, D. M. Crothers, and P. Doty, “Relaxation kinetics of dimer formation by self complementary oligonucleotides,” Journal of Molecular Biology, 1971, 62, 383-392, IN317, 393-401.

[10] J. G. Wetmur, and N. Davidson, “Kinetics of renaturation of DNA,” Journal of Molecular Biology, 1968, 31, 349-370.

[11] J. L. Sikorav, H. Orland, and A. Braslau, “Mechanism of thermal renaturation and hybridization of nucleic acids: Kramers' process and universality in Watson-Crick base pairing,” Journal of Physical Chemistry B, 2009, 113, 3715-3725.

246

[12] F. W. Studier, “Effects of conformation of single-stranded DNA on renaturation and aggregation,” Journal of Molecular Biology, 1969, 41, 199-209.

[13] C. R. Cantor, C. L. Smith, and Human Genome Project., Genomics : the science and technology behind the Human Genome Project, New York: Wiley, 1999.

[14] T. Strachan, and A. P. Read, Human molecular genetics, 4th ed., New York: Garland Science, 2011.

[15] M. M. A. Sekar, W. Bloch, and P. M. St John, “Comparative study of sequence-dependent hybridization kinetics in solution and on microspheres,” Nucleic Acids Research, 2005, 33, 366-375.

[16] Y. Okahata, M. Kawase, K. Niikura, F. Ohtake, H. Furusawa, and Y. Ebara, “Kinetic measurements of DNA hybridisation an an oligonucleotide-immobilized 27-MHz quartz crystal microbalance,” Analytical Chemistry, 1998, 70, 1288-1296.

[17] J. G. Wetmur, “DNA probes: Applications of the principles of nucleic-acid hybridization,” Critical Reviews in Biochemistry and Molecular Biology, 1991, 26, 227-259.

[18] R. Owczarzy, Y. You, B. G. Moreira, J. A. Manthey, L. Y. Huang, M. A. Behlke, and J. A. Walder, “Effects of sodium ions on DNA duplex oligomers: Improved predictions of melting temperatures,” Biochemistry, 2004, 43, 3537-3554.

[19] D. Bowtell, and J. Sambrook, DNA microarrays : a molecular cloning manual, Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press, 2003.

[20] T. Strachan, and A. P. Read, Human molecular genetics, 2nd ed., New York: Wiley, 1999.

[21] T. Koressaar, and M. Remm, “Enhancements and modifications of primer design program Primer3,” Bioinformatics, 2007, 23, 1289-1291.

[22] J. G. Mulle, V. C. Patel, S. T. Warren, M. R. Hegde, D. J. Cutler, and M. E. Zwick, “Empirical evaluation of oligonucleotide probe selection for DNA microarrays,” PLoS ONE, 2010, 5, e9921.

[23] G. Bernardi, “Chromatography of nucleic acids on hydroxyapatite columns,” Methods in Enzymology, 1971, 21, 95-139.

[24] J. H. Lee, C. J. Park, J. S. Shin, T. Ikegami, H. Akutsu, and B. S. Choi, “NMR structure of the DNA decamer duplex containing double T center dot G mismatches of cis-syn cyclobutane pyrimidine dimer: Implications for DNA damage recognition by the XPC-hHR23B complex,” Nucleic Acids Research, 2004, 32, 2474-2481.

247

[25] S. Kim, and A. Misra, “SNP genotyping: Technologies and biomedical applications,” Annual Review of Biomedical Engineering, 2007, 9, 289-320.

[26] A. J. Brookes, “The essence of SNPs,” Gene, 1999, 234, 177-186.

[27] K. Nakatani, “Chemistry challenges in SNP typing,” Chembiochem, 2004, 5, 1623-1633.

[28] A. Vainrub, and B. M. Pettitt, “Theoretical aspects of genomic variation screening using DNA microarrays,” Biopolymers, 2004, 73, 614-620.

[29] R. H. Lipsky, C. M. Mazzanti, J. G. Rudolph, K. Xu, G. Vyas, D. Bozak, M. Q. Radel, and D. Goldman, “DNA melting analysis for detection of single nucleotide polymorphisms,” Clinical Chemistry, 2001, 47, 635-644.

[30] J. SantaLucia, and D. Hicks, “The thermodynamics of DNA structural motifs,” Annual Review of Biophysics and Biomolecular Structure, 2004, 33, 415-440.

[31] T. Naiser, J. Kayser, T. Mai, W. Michel, and A. Ott, “Stability of a surface-bound oligonucleotide duplex inferred from molecular dynamics: A study of single nucleotide defects using DNA microarrays,” Physical Review Letters, 2009, 102.

[32] M. Arcellana-Panlilio, and S. M. Robbins, “Cutting-edge technology: I. Global gene expression profiling using DNA microarrays,” American Journal of Physiology-Gastrointestinal and Liver Physiology, 2002, 282, G397-G402.

[33] U. Maskos, and E. M. Southern, “Oligonucleotide hybridizations on glass supports: A novel linker for oligonucleotide synthesis and hybridization properties of oligonucleotides synthesized in situ,” Nucleic Acids Research, 1992, 20, 1679-1684.

[34] R. M. Ruprecht, N. C. Goodman, and Spiegelm.S, “Determination of natural host taxonomy of RNA tumor-viruses by molecular hybridization - application to Rd-114, a candidate human virus - (origin-feline-dna.rna-oncornavirus),” Proceedings of the National Academy of Sciences of the United States of America, 1973, 70, 1437-1441.

[35] A. Sassolas, B. D. Leca-Bouvier, and L. J. Blum, “DNA biosensors and microarrays,” Chemical Reviews, 2008, 108, 109-139.

[36] S. Russell, L. A. Meadows, and R. R. Russell, Microarray technology in practice, Amsterdam ; Boston: Academic Press/Elsevier, 2009.

[37] M. Schena, Microarray analysis, New York: Wiley-Liss, 2003.

248

[38] M. J. Buck, and J. D. Lieb, “ChIP-chip: considerations for the design, analysis, and application of genome-wide chromatin immunoprecipitation experiments,” Genomics, 2004, 83, 349-360.

[39] J. D. Hoheisel, “Microarray technology: beyond transcript profiling and genotype analysis,” Nature Reviews Genetics, 2006, 7, 200-210.

[40] D. D. Dalma-Weiszhausz, J. Warrington, E. Y. Tanimoto, and G. Miyada, “The affymetrix GeneChip (R) platform: An overview,” DNA Microarrays Part A: Array Platforms and Wet-Bench Protocols, 2006, 410, 3-28.

[41] R. J. Lipshutz, S. P. A. Fodor, T. R. Gingeras, and D. J. Lockhart, “High density synthetic oligonucleotide arrays,” Nature Genetics, 1999, 21, 20-24.

[42] M. J. Okoniewski, and C. J. Miller, “Comprehensive analysis of Affymetrix exon arrays using bioconductor,” PLoS Computational Biology, 2008, 4.

[43] I. Barbulovic-Nad, M. Lucente, Y. Sun, M. J. Zhang, A. R. Wheeler, and M. Bussmann, “Bio-microarray fabrication techniques: A review,” Critical Reviews in Biotechnology, 2006, 26, 237-259.

[44] M. Schena, D. Shalon, R. W. Davis, and P. O. Brown, “Quantitative monitoring of gene-expression patterns with a complementary-DNA microarray,” Science, 1995, 270, 467-470.

[45] A. Relogio, C. Schwager, A. Richter, W. Ansorge, and J. Valcarcel, “Optimization of oligonucleotide-based DNA microarrays,” Nucleic Acids Research, 2002, 30.

[46] T. R. Hughes, M. Mao, A. R. Jones, J. Burchard, M. J. Marton, K. W. Shannon, S. M. Lefkowitz, M. Ziman, J. M. Schelter, M. R. Meyer, S. Kobayashi, C. Davis, H. Y. Dai, Y. D. D. He, S. B. Stephaniants, G. Cavet, W. L. Walker, A. West, E. Coffey, D. D. Shoemaker, R. Stoughton, A. P. Blanchard, S. H. Friend, and P. S. Linsley, “Expression profiling using microarrays fabricated by an ink-jet oligonucleotide synthesizer,” Nature Biotechnology, 2001, 19, 342-347.

[47] Y. H. Yang, M. J. Buckley, and T. P. Speed, “Analysis of cDNA microarray images,” Briefings in Bioinformatics, 2001, 2, 341-349.

[48] B. J. Cheek, A. B. Steel, M. P. Torres, Y. Y. Yu, and H. Yang, “Chemiluminescence detection for hybridization assays on the flow-thru chip, a three-dimensional microchannel biochip,” Analytical Chemistry, 2001, 73, 5777-5783.

[49] H. C. King, and A. A. Sinha, “Gene expression profile analysis by DNA microarrays - Promise and pitfalls,” Jama-Journal of the American Medical Association, 2001, 286, 2280-2288.

249

[50] J. H. Lee, E. Ko, Y. E. Kim, J. Y. Min, J. Liu, Y. Kim, M. Shin, M. Hong, and H. Bae, “Gene expression profile analysis of genes in rat hippocampus from antidepressant treated rats using DNA microarray,” BMC Neuroscience, 2010, 11, 152.

[51] R. B. Stoughton, “Applications of DNA microarrays in biology,” Annual Review of Biochemistry, 2005, 74, 53-82.

[52] M. L. Bulyk, “DNA microarray technologies for measuring protein-DNA interactions,” Current Opinion in Biotechnology, 2006, 17, 422-430.

[53] J. Shendure, R. D. Mitra, C. Varma, and G. M. Church, “Advanced sequencing technologies: Methods and goals,” Nature Reviews Genetics, 2004, 5, 335-344.

[54] A. Khademhosseini, “Chips to hits: microarray and microfluidic technologies for high-throughput analysis and drug discovery,” Expert Review of Molecular Diagnostics, 2005, 5, 843-846.

[55] C. Situma, M. Hashimoto, and S. A. Soper, “Merging microfluidics with microarray-based bioassays,” Biomolecular Engineering, 2006, 23, 213-231.

[56] R. H. Liu, K. Dill, H. S. Fuji, and A. McShea, “Integrated microfluidic biochips for DNA microarray analysis,” Expert Review of Molecular Diagnostics, 2006, 6, 253-261.

[57] K. Pappaert, P. Van Hummelen, J. Vanderhoeven, G. V. Baron, and G. Desmet, “Diffusion-reaction modelling of DNA hybridization kinetics on biochips,” Chemical Engineering Science, 2003, 58, 4921-4930.

[58] H. Y. Dai, M. Meyer, S. Stepaniants, M. Ziman, and R. Stoughton, “Use of hybridization kinetics for differentiating specific from non-specific binding to oligonucleotide microarrays,” Nucleic Acids Research, 2002, 30, e86.

[59] G. L. Lukacs, P. Haggie, O. Seksek, D. Lechardeur, N. Freedman, and A. S. Verkman, “Size-dependent DNA mobility in cytoplasm and nucleus,” Journal of Biological Chemistry, 2000, 275, 1625-1629.

[60] M. A. Bynum, and G. B. Gordon, “Hybridization enhancement using microfluidic planetary centrifugal mixing,” Analytical Chemistry, 2004, 76, 7039-7044.

[61] K. Pappaert, J. Vanderhoeven, P. Van Hummelen, B. Dutta, D. Clicq, G. V. Baron, and G. Desmet, “Enhancement of DNA micro-array analysis using a shear-driven micro-channel flow system,” Journal of Chromatography A, 2003, 1014, 1-9.

[62] J. Hong, J. B. Edel, and A. J. deMello, “Micro- and nanofluidic systems for high-throughput biological screening,” Drug Discovery Today, 2009, 14, 134-146.

250

[63] G. Keramas, G. Perozziello, O. Geschke, and C. B. V. Christensen, “Development of a multiplex microarray microsystem,” Lab on a Chip, 2004, 4, 152-158.

[64] S. Huang, C. Li, B. Lin, and J. Qin, “Microvalve and micropump controlled shuttle flow microfluidic device for rapid DNA hybridization,” Lab on a Chip, 2010, 10, 2925-2931.

[65] H. Chen, L. Wang, and P. C. H. Li, “Nucleic acid microarrays created in the double-spiral format on a circular microfluidic disk,” Lab on a Chip, 2008, 8, 826-829.

[66] M. Dufva, J. Petersen, and L. Poulsen, “Increasing the specificity and function of DNA microarrays by processing arrays at different stringencies,” Analytical and Bioanalytical Chemistry, 2009, 395, 669-677.

[67] J. Cheng, E. L. Sheldon, L. Wu, A. Uribe, L. O. Gerrue, J. Carrino, M. J. Heller, and J. P. O'Connell, “Preparation and hybridization analysis of DNA/RNA from E-coli on microfabricated bioelectronic chips,” Nature Biotechnology, 1998, 16, 541-546.

[68] R. G. Sosnowski, E. Tu, W. F. Butler, J. P. OConnell, and M. J. Heller, “Rapid determination of single base mismatch mutations in DNA hybrids by direct electric field control,” Proceedings of the National Academy of Sciences of the United States of America, 1997, 94, 1119-1123.

[69] M. Syrzycka, M. Sjoerdsma, P. C. H. Li, M. Parameswaran, M. Syrzycki, C. A. Koch, and R. S. Utkhede, “Electronic concentration of negatively-charged molecules on a microfabricated biochip,” Analytica Chimica Acta, 2003, 484, 1-14.

[70] J. C. McDonald, D. C. Duffy, J. R. Anderson, D. T. Chiu, H. K. Wu, O. J. A. Schueller, and G. M. Whitesides, “Fabrication of microfluidic systems in poly(dimethylsiloxane),” Electrophoresis, 2000, 21, 27-40.

[71] R. C. Anderson, X. Su, G. J. Bogdan, and J. Fenton, “A miniature integrated device for automated multistep genetic assays,” Nucleic Acids Research, 2000, 28, e60.

[72] M. Noerholm, H. Bruus, M. H. Jakobsen, P. Telleman, and N. B. Ramsing, “Polymer microfluidic chip for online monitoring of microarray hybridizations,” Lab on a Chip, 2004, 4, 28-37.

[73] H. H. Lee, J. Smoot, Z. McMurray, D. A. Stahl, and P. Yager, “Recirculating flow accelerates DNA microarray hybridization in a microfluidic device,” Lab on a Chip, 2006, 6, 1163-1170.

[74] J. Liu, B. A. Williams, R. M. Gwirtz, B. J. Wold, and S. Quake, “Enhanced signals and fast nucleic acid hybridization by microfluidic chaotic mixing,” Angewandte Chemie-International Edition, 2006, 45, 3618-3623.

251

[75] N. B. Adey, M. Lei, M. T. Howard, J. D. Jensen, D. A. Mayo, D. L. Butel, S. C. Coffin, T. C. Moyer, D. E. Slade, M. K. Spute, A. M. Hancock, G. T. Eisenhoffer, B. K. Dalley, and M. R. McNeely, “Gains in sensitivity with a device that mixes microarray hybridization solution in a 25-µm-thick chamber,” Analytical Chemistry, 2002, 74, 6413-6417.

[76] J. Vanderhoeven, K. Pappaert, B. Dutta, P. Van Hummelen, and G. Desmet, “Comparison of a pump-around a diffusion-driven, and a shear-driven system for the hybridization of mouse lung and testis total RNA on microarrays,” Electrophoresis, 2005, 26, 3773-3779.

[77] J. Vanderhoeven, K. Pappaert, B. Dutta, P. Van Hummelen, and G. Desmet, “DNA microarray enhancement using a continuously and discontinuously rotating microchamber,” Analytical Chemistry, 2005, 77, 4474-4480.

[78] P. K. Yuen, G. S. Li, Y. J. Bao, and U. R. Muller, “Microfluidic devices for fluidic circulation and mixing improve hybridization signal intensity on DNA arrays,” Lab on a Chip, 2003, 3, 46-50.

[79] R. H. Liu, R. Lenigk, R. L. Druyor-Sanchez, J. N. Yang, and P. Grodzinski, “Hybridization enhancement using cavitation microstreaming,” Analytical Chemistry, 2003, 75, 1911-1917.

[80] M. K. McQuain, K. Seale, J. Peek, T. S. Fisher, S. Levy, M. A. Stremler, and F. R. Haselton, “Chaotic mixer improves microarray hybridization,” Analytical Biochemistry, 2004, 325, 215-226.

[81] F. Raynal, F. Plaza, A. Beuf, P. Carriere, E. Souteyrand, J. R. Martin, J. P. Cloarec, and M. Cabrera, “Study of a chaotic mixing system for DNA chip hybridization chambers,” Physics of Fluids, 2004, 16, L63-L66.

[82] J. M. Hertzsch, R. Sturman, and S. Wiggins, “DNA microarrays: Design principles for maximizing ergodic, chaotic mixing,” Small, 2007, 3, 202-218.

[83] Y. J. Liu, and C. B. Rauch, “DNA probe attachment on plastic surfaces and microfluidic hybridization array channel devices with sample oscillation,” Analytical Biochemistry, 2003, 317, 76-84.

[84] J. H. S. Kim, A. Marafie, X. Y. Jia, J. V. Zoval, and M. J. Madou, “Characterization of DNA hybridization kinetics in a microfluidic flow channel,” Sensors and Actuators B-Chemical, 2006, 113, 281-289.

[85] D. Erickson, X. Z. Liu, U. Krull, and D. Q. Li, “Electrokinetically controlled DNA hybridization microfluidic chip enabling rapid target analysis,” Analytical Chemistry, 2004, 76, 7269-7277.

252

[86] Y. Wang, B. Vaidya, H. D. Farquar, W. Stryjewski, R. P. Hammer, R. L. McCarley, S. A. Soper, Y. W. Cheng, and F. Barany, “Microarrays assembled in microfluidic chips fabricated from poly(methyl methacrylate) for the detection of low-abundant DNA mutations,” Analytical Chemistry, 2003, 75, 1130-1140.

[87] Y. C. Chung, Y. C. Lin, M. Z. Shiu, and W. N. T. Chang, “Microfluidic chip for fast nucleic acid hybridization,” Lab on a Chip, 2003, 3, 228-233.

[88] Y. C. Chung, Y. C. Lin, C. D. Chueh, C. Y. Ye, L. W. Lai, and Q. Zhao, “Microfluidic chip of fast DNA hybridization using denaturing and motion of nucleic acids,” Electrophoresis, 2008, 29, 1859-1865.

[89] R. Lenigk, R. H. Liu, M. Athavale, Z. J. Chen, D. Ganser, J. N. Yang, C. Rauch, Y. J. Liu, B. Chan, H. N. Yu, M. Ray, R. Marrero, and P. Grodzinski, “Plastic biochannel hybridization devices: a new concept for microfluidic DNA arrays,” Analytical Biochemistry, 2002, 311, 40-49.

[90] C. W. Wei, J. Y. Cheng, C. T. Huang, M. H. Yen, and T. H. Young, “Using a microfluidic device for 1 μl DNA microarray hybridization in 500 s,” Nucleic Acids Research, 2005, 33, e78.

[91] J. A. Benn, J. Hu, B. J. Hogan, R. C. Fry, L. D. Samson, and T. Thorsen, “Comparative modeling and analysis of microfluidic and conventional DNA microarrays,” Analytical Biochemistry, 2006, 348, 284-293.

[92] Y. C. Li, Z. Wang, L. M. L. Ou, and H. Z. Yu, “DNA detection on plastic: Surface activation protocol to convert polycarbonate substrates to biochip platforms,” Analytical Chemistry, 2007, 79, 426-433.

[93] C. Situma, Y. Wang, M. Hupert, F. Barany, R. L. McCarley, and S. A. Soper, “Fabrication of DNA microarrays onto poly(methyl methacrylate) with ultraviolet patterning and microfluidics for the detection of low-abundant point mutations,” Analytical Biochemistry, 2005, 340, 123-135.

[94] L. Wang, and P. C. H. Li, “Flexible microarray construction and fast DNA hybridization conducted on a microfluidic chip for greenhouse plant fungal pathogen detection,” Journal of Agricultural and Food Chemistry, 2007, 55, 10509-10516.

[95] L. Wang, and P. C. H. Li, “Gold nanoparticle-assisted single base-pair mismatch discrimination on a microfluidic microarray device,” Biomicrofluidics, 2010, 4, 32209.

[96] H. J. Lee, T. T. Goodrich, and R. M. Corn, “SPR imaging measurements of 1-D and 2-D DNA microarrays created from microfluidic channels on gold thin films,” Analytical Chemistry, 2001, 73, 5525-5531.

[97] P. Stafford, Methods in microarray normalization, Boca Raton: CRC Press, 2008.

253

[98] A. Bouchie, “Shift anticipated in DNA microarray market,” Nature Biotechnology, 2002, 20, 8.

[99] J. P. Gillet, T. J. Molina, J. Jamart, P. Gaulard, K. Leroy, J. Briere, I. Theate, C. Thieblemont, A. Bosly, M. Herin, J. Hamels, and J. Remacle, “Evaluation of a low density DNA microarray for small B-cell non-Hodgkin lymphoma differential diagnosis,” Leukemia & Lymphoma, 2009, 50, 410-418.

[100] G. Meneses-Lorente, F. de Longueville, S. Dos Santos-Mendes, T. P. Bonnert, A. Jack, S. Evrard, V. Bertholet, A. Pike, P. Scott-Stevens, J. Remacle, and B. Sohal, “An evaluation of a low-density DNA microarray using cytochrome P450 inducers,” Chemical Research in Toxicology, 2003, 16, 1070-1077.

[101] F. de Longueville, D. Surry, G. Meneses-Lorente, V. Bertholet, V. Talbot, S. Evrard, N. Chandelier, A. Pike, P. Worboys, J. P. Rasson, B. Le Bourdelles, and J. Remacle, “Gene expression profiling of drug metabolism and toxicology markers using a low-density DNA microarray,” Biochemical Pharmacology, 2002, 64, 137-149.

[102] J. P. Gillet, T. Efferth, D. Steinbach, J. Hamels, F. de Longueville, V. Bertholet, and J. Remacle, “Microarray-based detection of multidrug resistance in human tumor cells by expression profiling of ATP-binding cassette transporter genes,” Cancer Research, 2004, 64, 8987-8993.

[103] D. Tejedor, S. Castillo, P. Mozas, E. Jimenez, M. Lopez, M. T. Tejedor, M. Artieda, R. Alonso, P. Mata, L. Simon, A. Martinez, M. Pocovi, and S. F. Grp, “Reliable low-density DNA array based on allele-specific probes for detection of 118 mutations causing familial hypercholesterolemia,” Clinical Chemistry, 2005, 51, 1137-1144.

[104] P. Alvarez, P. Saenz, D. Arteta, A. Martinez, M. Pocovi, L. Simon, and P. Giraldo, “Transcriptional profiling of hematologic malignancies with a low-density DNA microarray,” Clinical Chemistry, 2007, 53, 259-267.

[105] C. A. Koch, P. C. H. Li, and R. S. Utkhede, “Evaluation of thin films of agarose on glass for hybridization of DNA to identify plant pathogens with microarray technology,” Analytical Biochemistry, 2005, 342, 93-102.

[106] R. Peytavi, F. R. Raymond, D. Gagne, F. J. Picard, G. Jia, J. Zoval, M. Madou, K. Boissinot, M. Boissinot, L. Bissonnette, M. Ouellette, and M. G. Bergeron, “Microfluidic device for rapid (< 15 min) automated microarray hybridization,” Clinical Chemistry, 2005, 51, 1836-1844.

[107] G. Y. Jia, K. S. Ma, J. Kim, J. V. Zoval, R. Peytavi, M. G. Bergeron, and M. J. Madou, “Dynamic automated DNA hybridization on a CD (compact disc) fluidic platform,” Sensors and Actuators B-Chemical, 2006, 114, 173-181.

254

[108] C. Y. Li, X. L. Dong, J. H. Qin, and B. C. Lin, “Rapid nanoliter DNA hybridization based on reciprocating flow on a compact disk microfluidic device,” Analytica Chimica Acta, 2009, 640, 93-99.

[109] Y. J. Liu, C. B. Rauch, R. L. Stevens, R. Lenigk, J. N. Yang, D. B. Rhine, and P. Grodzinski, “DNA amplification and hybridization assays in integrated plastic monolithic devices,” Analytical Chemistry, 2002, 74, 3063-3070.

[110] J. W. Larson, G. R. Yantz, Q. Zhong, R. Charnas, C. M. D'Antoni, M. V. Gallo, K. A. Gillis, L. A. Neely, K. M. Phillips, G. G. Wong, S. R. Gullans, and R. Gilmanshin, “Single DNA molecule stretching in sudden mixed shear and elongational microflows,” Lab on a Chip, 2006, 6, 1187-1199.

[111] M. Campas, and I. Katakis, “DNA biochip arraying, detection and amplification strategies,” Trac-Trends in Analytical Chemistry, 2004, 23, 49-62.

[112] V. Le Berre, E. Trevisiol, A. Dagkessamanskaia, S. Sokol, A. M. Caminade, J. P. Majoral, B. Meunier, and J. Francois, “Dendrimeric coating of glass slides for sensitive DNA microarrays analysis,” Nucleic Acids Research, 2003, 31, e88.

[113] K. Iida, and I. Nishimura, “Gene expression profiling by DNA microarray technology,” Critical Reviews in Oral Biology & Medicine, 2002, 13, 35-50.

[114] S. Taylor, S. Smith, B. Windle, and A. Guiseppi-Elie, “Impact of surface chemistry and blocking strategies on DNA microarrays,” Nucleic Acids Research, 2003, 31, e87.

[115] S. J. Hinder, S. D. Connell, M. C. Davies, C. J. Roberts, S. J. B. Tendler, and P. M. Williams, “Compositional mapping of self-assembled monolayers derivatized within microfluidic networks,” Langmuir, 2002, 18, 3151-3158.

[116] K. A. Smith, B. K. Gale, and J. C. Conboy, “Micropatterned Fluid Lipid Bilayer Arrays Created Using a Continuous Flow Microspotter,” Analytical Chemistry, 2008, 80, 7980-7987.

[117] J. D. Taylor, K. S. Phillips, and Q. Cheng, “Microfluidic fabrication of addressable tethered lipid bilayer arrays and optimization using SPR with silane-derivatized nanoglassy substrates,” Lab on a Chip, 2007, 7, 927-930.

[118] M. Zagnoni, M. E. Sandison, and H. Morgan, “Microfluidic array platform for simultaneous lipid bilayer membrane formation,” Biosensors & Bioelectronics, 2009, 24, 1235-1240.

255

[119] E. D. Goluch, A. W. Shaw, S. G. Sligar, and C. Liu, “Microfluidic patterning of nanodisc lipid bilayers and multiplexed analysis of protein interaction,” Lab on a Chip, 2008, 8, 1723-1728.

[120] I. Caelen, A. Bernard, D. Juncker, B. Michel, H. Heinzelmann, and E. Delamarche, “Formation of gradients of proteins on surfaces with microfluidic networks,” Langmuir, 2000, 16, 9125-9130.

[121] C. A. Rowe, L. M. Tender, M. J. Feldstein, J. P. Golden, S. B. Scruggs, B. D. MacCraith, J. J. Cras, and F. S. Ligler, “Array biosensor for simultaneous identification of bacterial, viral, and protein analytes,” Analytical Chemistry, 1999, 71, 3846-3852.

[122] C. A. Rowe, S. B. Scruggs, M. J. Feldstein, J. P. Golden, and F. S. Ligler, “An array immunosensor for simultaneous detection of clinical analytes,” Analytical Chemistry, 1999, 71, 433-439.

[123] K. E. Sapsford, P. T. Charles, C. H. Patterson, and F. S. Ligler, “Demonstration of four immunoassay formats using the array biosensor,” Analytical Chemistry, 2002, 74, 1061-1068.

[124] A. Bernard, B. Michel, and E. Delamarche, “Micromosaic immunoassays,” Analytical Chemistry, 2001, 73, 8-12.

[125] E. Delamarche, A. Bernard, H. Schmid, B. Michel, and H. Biebuyck, “Patterned delivery of immunoglobulins to surfaces using microfluidic networks,” Science, 1997, 276, 779-781.

[126] E. Delamarche, D. Juncker, and H. Schmid, “Microfluidics for processing surfaces and miniaturizing biological assays,” Advanced Materials, 2005, 17, 2911-2933.

[127] E. Delamarche, A. Bernard, H. Schmid, A. Bietsch, B. Michel, and H. Biebuyck, “Microfluidic networks for chemical patterning of substrate: Design and application to bioassays,” Journal of the American Chemical Society, 1998, 120, 500-508.

[128] M. Geissler, E. Roy, G. A. Diaz-Quijada, J. C. Galas, and T. Veres, “Microfluidic patterning of miniaturized DNA arrays on plastic substrates,” ACS Applied Materials & Interfaces, 2009, 1, 1387-1395.

[129] L. Wang, P. C. H. Li, H. Z. Yu, and A. M. Parameswaran, “Fungal pathogenic nucleic acid detection achieved with a microfluidic microarray device,” Analytica Chimica Acta, 2008, 610, 97-104.

[130] Y. Huang, X. F. Duan, Q. Q. Wei, and C. M. Lieber, “Directed assembly of one-dimensional nanostructures into functional networks,” Science, 2001, 291, 630-633.

256

[131] B. Messer, J. H. Song, and P. D. Yang, “Microchannel networks for nanowire patterning,” Journal of the American Chemical Society, 2000, 122, 10232-10233.

[132] J. Y. Shiu, C. W. Kuo, and P. L. Chen, “Actively controlled self-assembly of colloidal crystals in microfluidic networks by electrocapillary forces,” Journal of the American Chemical Society, 2004, 126, 8096-8097.

[133] X. Y. Peng, P. C. H. Li, H. Z. Yu, M. Parameswaran, and W. L. Chou, “Spiral microchannels on a CD for DNA hybridizations,” Sensors and Actuators B-Chemical, 2007, 128, 64-69.

[134] L. Wang, and P. C. H. Li, “Optimization of a microfluidic microarray device for the fast discrimination of fungal pathogenic DNA,” Analytical Biochemistry, 2010, 400, 282-288.

[135] J. D. Taylor, M. J. Linman, T. Wilkop, and Q. Cheng, “Regenerable tethered bilayer lipid membrane arrays for multiplexed label-free analysis of lipid-protein interactions on poly(dimethylsiloxane) microchips using SPR imaging,” Analytical Chemistry, 2009, 81, 1146-1153.

[136] J. Kim, H. Kido, R. H. Rangel, and M. J. Madou, “Passive flow switching valves on a centrifugal microfluidic platform,” Sensors and Actuators B-Chemical, 2008, 128, 613-621.

[137] S. C. Jacobson, T. E. McKnight, and J. M. Ramsey, “Microfluidic devices for electrokinetically driven parallel and serial mixing,” Analytical Chemistry, 1999, 71, 4455-4459.

[138] P. C. H. Li, Microfluidic lab-on-a-chip for chemical and biological analysis and discovery, Boca Raton: Taylor & Francis/CRC Press, 2006.

[139] P. Gong, and R. Levicky, “DNA surface hybridization regimes,” Proceedings of the National Academy of Sciences of the United States of America, 2008, 105, 5301-5306.

[140] R. Levicky, and A. Horgan, “Physicochemical perspectives on DNA microarray and biosensor technologies,” Trends in Biotechnology, 2005, 23, 143-149.

[141] G. Y. Tang, D. G. Yan, C. Yang, H. Q. Gong, J. C. Chai, and Y. C. Lam, “Assessment of Joule heating and its effects on electroosmotic flow and electrophoretic transport of solutes in microfluidic channels,” Electrophoresis, 2006, 27, 628-639.

[142] X. C. Xuan, B. Xu, D. Sinton, and D. Q. Li, “Electroosmotic flow with Joule heating effects,” Lab on a Chip, 2004, 4, 230-236.

[143] I. Rodriguez, and N. Chandrasekhar, “Experimental study and numerical estimation of current changes in electroosmotically pumped microfluidic devices,” Electrophoresis, 2005, 26, 1114-1121.

257

[144] B. J. Kirby, and E. F. Hasselbrink, “Zeta potential of microfluidic substrates: 1. Theory, experimental techniques, and effects on separations,” Electrophoresis, 2004, 25, 187-202.

[145] R. A. Potyrailo, W. G. Morris, A. M. Leach, T. M. Sivavec, M. B. Wisnudel, and S. Boyette, “Analog signal acquisition from computer optical disk drives for quantitative chemical sensing,” Analytical Chemistry, 2006, 78, 5893-5899.

[146] S. A. Lange, G. Roth, S. Wittemann, T. Lacoste, A. Vetter, J. Grassle, S. Kopta, M. Kolleck, B. Breitinger, M. Wick, J. K. H. Horber, S. Dubel, and A. Bernard, “Measuring biomolecular binding events with a compact disc player device,” Angewandte Chemie-International Edition, 2006, 45, 270-273.

[147] M. Madou, J. Zoval, G. Y. Jia, H. Kido, J. Kim, and N. Kim, “Lab on a CD,” Annual Review of Biomedical Engineering, 2006, 8, 601-628.

[148] D. D. Nolte, “Invited Review Article: Review of centrifugal microfluidic and bio-optical disks,” Review of Scientific Instruments, 2009, 80, 101101

[149] A. S. Watts, A. A. Urbas, E. Moschou, V. G. Gavalas, J. V. Zoval, M. Madou, and L. G. Bachas, “Centrifugal microfluidics with integrated sensing microdome optodes for multiion detection,” Analytical Chemistry, 2007, 79, 8046-8054.

[150] R. D. Johnson, I. H. A. Badr, G. Barrett, S. Y. Lai, Y. M. Lu, M. J. Madou, and L. G. Bachas, “Development of a fully integrated analysis system for ions based on ion-selective optodes and centrifugal microfluidics,” Analytical Chemistry, 2001, 73, 3940-3946.

[151] I. H. A. Badr, R. D. Johnson, M. J. Madou, and L. G. Bachas, “Fluorescent ion-selective optode membranes incorporated onto a centrifugal microfluidics platform,” Analytical Chemistry, 2002, 74, 5569-5575.

[152] J. Steigert, M. Grumann, T. Brenner, L. Riegger, J. Harter, R. Zengerle, and J. Ducree, “Fully integrated whole blood testing by real-time absorption measurement on a centrifugal platform,” Lab on a Chip, 2006, 6, 1040-1044.

[153] M. Gustafsson, D. Hirschberg, C. Palmberg, H. Jornvall, and T. Bergman, “Integrated sample preparation and MALDI mass spectrometry on a microfluidic compact disk,” Analytical Chemistry, 2004, 76, 345-350.

[154] J. Steigert, T. Brenner, M. Grumann, L. Riegger, S. Lutz, R. Zengerle, and J. Ducree, “Integrated siphon-based metering and sedimentation of whole blood on a hydrophilic lab-on-a-disk,” Biomedical Microdevices, 2007, 9, 675-679.

258

[155] J. L. Zhang, Q. Q. Guo, M. Liu, and J. Yang, “A lab-on-CD prototype for high-speed blood separation,” Journal of Micromechanics and Microengineering, 2008, 18, 125025, p125021-125026.

[156] D. C. Duffy, H. L. Gillis, J. Lin, N. F. Sheppard, and G. J. Kellogg, “Microfabricated centrifugal microfluidic systems: Characterization and multiple enzymatic assays,” Analytical Chemistry, 1999, 71, 4669-4678.

[157] P. Andersson, G. Jesson, G. Kylberg, G. Ekstrand, and G. Thorsen, “Parallel nanoliter microfluidic analysis system,” Analytical Chemistry, 2007, 79, 4022-4030.

[158] S. W. Lee, J. Y. Kang, I. H. Lee, S. S. Ryu, S. M. Kwak, K. S. Shin, C. Kim, H. I. Jung, and T. S. Kim, “Single-cell assay on CD-like lab chip using centrifugal massive single-cell trap,” Sensors and Actuators A-Physical, 2008, 143, 64-69.

[159] S. Haeberle, T. Brenner, R. Zengerle, and J. Ducree, “Centrifugal extraction of plasma from whole blood on a rotating disk,” Lab on a Chip, 2006, 6, 776-781.

[160] S. Lai, S. N. Wang, J. Luo, L. J. Lee, S. T. Yang, and M. J. Madou, “Design of a compact disk-like microfluidic platform for enzyme-linked immunosorbent assay,” Analytical Chemistry, 2004, 76, 1832-1837.

[161] A. Penrose, P. Myers, K. Bartle, and S. McCrossen, “Development and assessment of a miniaturised centrifugal chromatograph for reversed-phase separations in micro-channels,” Analyst, 2004, 129, 704-709.

[162] A. P. Wong, M. Gupta, S. S. Shevkoplyas, and G. M. Whitesides, “Egg beater as centrifuge: isolating human blood plasma from whole blood in resource-poor settings,” Lab on a Chip, 2008, 8, 2032-2037.

[163] B. S. Lee, J. N. Lee, J. M. Park, J. G. Lee, S. Kim, Y. K. Cho, and C. Ko, “A fully automated immunoassay from whole blood on a disc,” Lab on a Chip, 2009, 9, 1548-1555.

[164] L. G. Puckett, E. Dikici, S. Lai, M. Madou, L. G. Bachas, and S. Daunert, “Investigation into the applicability of the centrifugal microfluidics development of protein-platform for the ligand binding assays incorporating enhanced green fluorescent protein as a fluorescent reporter,” Analytical Chemistry, 2004, 76, 7263-7268.

[165] M. Grumann, J. Steigert, L. Riegger, I. Moser, B. Enderle, K. Riebeseel, G. Urban, R. Zengerle, and J. Ducree, “Sensitivity enhancement for colorimetric glucose assays on whole blood by on-chip beam-guidance,” Biomedical Microdevices, 2006, 8, 209-214.

259

[166] Y. K. Cho, J. G. Lee, J. M. Park, B. S. Lee, Y. Lee, and C. Ko, “One-step pathogen specific DNA extraction from whole blood on a centrifugal microfluidic device,” Lab on a Chip, 2007, 7, 565-573.

[167] D. A. Duford, D. D. Peng, and E. D. Salin, “Magnetically driven solid sample preparation for centrifugal microfluidic devices,” Analytical Chemistry, 2009, 81, 4581-4584.

[168] J. P. Lafleur, A. A. Rackov, S. McAuley, and E. D. Salin, “Miniaturised centrifugal solid phase extraction platforms for in-field sampling, pre-concentration and spectrometric detection of organic pollutants in aqueous samples,” Talanta, 2010, 81, 722-726.

[169] J. M. Chen, P. C. Huang, and M. G. Lin, “Analysis and experiment of capillary valves for microfluidics on a rotating disk,” Microfluidics and Nanofluidics, 2008, 4, 427-437.

[170] S. Haeberle, T. Brenner, H. P. Schlosser, R. Zengerle, and J. Ducree, “Centrifugal micromixer,” Chemical Engineering & Technology, 2005, 28, 613-616.

[171] T. Glatzel, C. Litterst, C. Cupelli, T. Lindemann, C. Moosmann, R. Niekrawletz, W. Streule, R. Zengerle, and P. Koltay, “Computational fluid dynamics (CFD) software tools for microfluidic applications: A case study,” Computers & Fluids, 2008, 37, 218-235.

[172] T. Brenner, T. Glatzel, R. Zengerle, and J. Ducree, “Frequency-dependent transversal flow control in centrifugal microfluidics,” Lab on a Chip, 2005, 5, 146-150.

[173] M. Liu, J. Zhang, Y. Liu, W. M. Lau, and J. Yang, “Modeling of flow burst, flow timing in Lab-on-a-CD systems and its application in digital chemical analysis,” Chemical Engineering & Technology, 2008, 31, 1328-1335.

[174] D. S. Kim, and T. H. Kwon, “Modeling, analysis and design of centrifugal force driven transient filling flow into rectangular microchannel,” Microsystem Technologies, 2006, 12, 822-838.

[175] D. S. Kim, and T. H. Kwon, “Modeling, analysis and design of centrifugal force-driven transient filling flow into a circular microchannel,” Microfluidics and Nanofluidics, 2006, 2, 125-140.

[176] J. Siegrist, M. Amasia, N. Singh, D. Banerjee, and M. Madou, “Numerical modeling and experimental validation of uniform microchamber filling in centrifugal microfluidics,” Lab on a Chip, 2010, 10, 876-886.

[177] J. Ducree, S. Haeberle, T. Brenner, T. Glatzel, and R. Zengerle, “Patterning of flow and mixing in rotating radial microchannels,” Microfluidics and Nanofluidics, 2006, 2, 97-105.

260

[178] M. S. Al-Qahtani, and M. N. Basha, “Prediction of flow and heat transfer in narrow rotating rectangular channels using Reynolds stress model,” Heat and Mass Transfer, 2008, 44, 505-516.

[179] J. V. Zoval, and M. J. Madou, “Centrifuge-based fluidic platforms,” Proceedings of the IEEE, 2004, 92, 140-153.

[180] D. Mark, S. Haeberle, G. Roth, F. von Stetten, and R. Zengerle, “Microfluidic lab-on-a-chip platforms: requirements, characteristics and applications,” Chemical Society Reviews, 2010, 39, 1153-1182.

[181] J. L. Garcia-Cordero, L. Basabe-Desmonts, J. Ducree, and A. J. Ricco, “Liquid recirculation in microfluidic channels by the interplay of capillary and centrifugal forces,” Microfluidics and Nanofluidics, 2010, 9, 695-703.

[182] G. M. Whitesides, “The origins and the future of microfluidics,” Nature, 2006, 442, 368-373.

[183] D. J. Harrison, K. Fluri, K. Seiler, Z. H. Fan, C. S. Effenhauser, and A. Manz, “Micromachining a miniaturized capillary electrophoresis-based chemical-analysis system on a chip,” Science, 1993, 261, 895-897.

[184] H. Becker, and C. Gartner, “Polymer microfabrication technologies for microfluidic systems,” Analytical and Bioanalytical Chemistry, 2008, 390, 89-111.

[185] H. Becker, and L. E. Locascio, “Polymer microfluidic devices,” Talanta, 2002, 56, 267-287.

[186] H. Becker, and C. Gartner, “Polymer microfabrication methods for microfluidic analytical applications,” Electrophoresis, 2000, 21, 12-26.

[187] J. H. Ng, and L. L. Ilag, “Biochips beyond DNA: technologies and applications,” Biotechnology Annual Review, 2003, 9, 1-149.

[188] H. Klank, J. P. Kutter, and O. Geschke, “CO2-laser micromachining and back-end processing for rapid production of PMMA-based microfluidic systems,” Lab on a Chip, 2002, 2, 242-246.

[189] J. P. Sercel, "UV laser micro-materials processing of MEMS, microfluidics, sensors, LEDs and other miniature devices," Nanotech-NSTI Nanotechnology Conference and Trade Show, Boston, Massachusetts, USA, 2004, pp. 518-520.

[190] Y. H. Chen, and S. H. Chen, “Analysis of DNA fragments by microchip electrophoresis fabricated on poly(methyl methacrylate) substrates using a wire-imprinting method,” Electrophoresis, 2000, 21, 165-170.

[191] C. W. Tsao, and D. L. DeVoe, “Bonding of thermoplastic polymer microfluidics,” Microfluidics and Nanofluidics, 2009, 6, 1-16.

261

[192] U. M. Attia, S. Marson, and J. R. Alcock, “Micro-injection moulding of polymer microfluidic devices,” Microfluidics and Nanofluidics, 2009, 7, 1-28.

[193] D. C. Duffy, J. C. McDonald, O. J. A. Schueller, and G. M. Whitesides, “Rapid prototyping of microfluidic systems in poly(dimethylsiloxane),” Analytical Chemistry, 1998, 70, 4974-4984.

[194] G. M. Whitesides, E. Ostuni, S. Takayama, X. Y. Jiang, and D. E. Ingber, “Soft lithography in biology and biochemistry,” Annual Review of Biomedical Engineering, 2001, 3, 335-373.

[195] P. Kim, K. W. Kwon, M. C. Park, S. H. Lee, S. M. Kim, and K. Y. Suh, “Soft lithography for microfluidics: a review,” Biochip Journal, 2008, 2, 1-11.

[196] J. C. McDonald, and G. M. Whitesides, “Poly(dimethylsiloxane) as a material for fabricating microfluidic devices,” Accounts of Chemical Research, 2002, 35, 491-499.

[197] K. A. Reinhardt, and W. Kern, Handbook of silicon wafer cleaning technology, 2nd ed., Norwich, NY: William Andrew, 2008.

[198] H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaud, and P. Vettiger, “SU-8: a low-cost negative resist for MEMS,” Journal of Micromechanics and Microengineering, 1997, 7, 121-124.

[199] J. M. Shaw, J. D. Gelorme, N. C. LaBianca, W. E. Conley, and S. J. Holmes, “Negative photoresists for optical lithography,” IBM Journal of Research and Development, 1997, 41, 81-94.

[200] A. del Campo, and C. Greiner, “SU-8: a photoresist for high-aspect-ratio and 3D submicron lithography,” Journal of Micromechanics and Microengineering, 2007, 17, R81-R95.

[201] D. K. Cai, and A. Neyer, “Cost-effective and reliable sealing method for PDMS (polydimethylsiloxane)-based microfluidic devices with various substrates,” Microfluidics and Nanofluidics, 2010, 9, 855-864.

[202] C. Khan-Malek, and D. Bodas, “Formation of more stable hydrophilic surfaces of PDMS by plasma and chemical treatments,” Microelectronic Engineering, 2006, 83, 1277-1279.

[203] Y. Berdichevsky, J. Khandurina, A. Guttman, and Y. H. Lo, “UV/ozone modification of poly(dimethylsiloxane) microfluidic channels,” Sensors and Actuators B-Chemical, 2004, 97, 402-408.

[204] G. S. Ferguson, M. K. Chaudhury, H. A. Biebuyck, and G. M. Whitesides, “Monolayers on disordered substrates: Self-assembly of alkyltrichlorosilanes on surface-modified polyethylene and poly(dimethylsiloxane),” Macromolecules, 1993, 26, 5870-5875.

262

[205] P. S. Dittrich, and P. Schwille, “An integrated microfluidic system for reaction, high-sensitivity detection, and sorting of fluorescent cells and particles,” Analytical Chemistry, 2003, 75, 5767-5774.

[206] P. Abgrall, V. Conedera, H. Camon, A. M. Gue, and N. T. Nguyen, “SU-8 as a structural material for labs-on-chips and microelectromechanical systems,” Electrophoresis, 2007, 28, 4539-4551.

[207] G. Liu, Y. Tian, and Y. Kan, “Fabrication of high-aspect-ratio microstructures using SU8 photoresist,” Microsystem Technologies, 2005, 11, 343-346.

[208] H. Wang, J. Li, H. P. Liu, Q. J. Liu, Q. Mei, Y. J. Wang, J. J. Zhu, N. Y. He, and Z. H. Lu, “Label-free hybridization detection of a single nucleotide mismatch by immobilization of molecular beacons on an agarose film,” Nucleic Acids Research, 2002, 30, e61.

[209] N. Zammatteo, L. Jeanmart, S. Hamels, S. Courtois, P. Louette, L. Hevesi, and J. Remacle, “Comparison between different strategies of covalent attachment of DNA to glass surfaces to build DNA microarrays,” Analytical Biochemistry, 2000, 280, 143-150.

[210] T. C. Paulitz, and R. R. Belanger, “Biological control in greenhouse systems,” Annual Review of Phytopathology, 2001, 39, 103-133.

[211] W. R. Jarvis, and V. W. Nuttall, Cucumber diseases, Harrow, Ontario: Research Station, 1979.

[212] E. De Rinaldis, and A. Lahm, DNA microarrays : current applications, Wymondham, U.K.: Horizon Bioscience, 2007.

[213] J. Xu, S. F. Zhu, H. Z. Miao, W. S. Huang, M. Y. Qiu, Y. Huang, X. P. Fu, and Y. Li, “Event-specific detection of seven genetically modified soybean and maizes using multiplex-PCR coupled with oligonucleotide microarray,” Journal of Agricultural and Food Chemistry, 2007, 55, 5575-5579.

[214] R. Bordoni, A. Germini, A. Mezzelani, R. Marchelli, and G. De Bellis, “A microarray platform for parallel detection of five transgenic events in foods: A combined polymerase chain reaction - ligation detection reaction - universal array method,” Journal of Agricultural and Food Chemistry, 2005, 53, 912-918.

[215] A. E. Warsen, M. J. Krug, S. LaFrentz, D. R. Stanek, F. J. Loge, and D. R. Call, “Simultaneous discrimination between 15 fish pathogens by using 16S ribosomal DNA PCR and DNA microarrays,” Applied and Environmental Microbiology, 2004, 70, 4216-4221.

263

[216] S. B. Ronning, K. Rudi, K. G. Berdal, and A. Holst-Jensen, “Differentiation of important and closely related cereal plant species (Poaceae) in food by hybridization to an oligonucleotide array,” Journal of Agricultural and Food Chemistry, 2005, 53, 8874-8880.

[217] M. Hashimoto, F. Barany, and S. A. Soper, “Polymerase chain reaction/ligase detection reaction/hybridization assays using flow-through microfluidic devices for the detection of low-abundant DNA point mutations,” Biosensors & Bioelectronics, 2006, 21, 1915-1923.

[218] R. H. Liu, H. X. Chen, K. R. Luehrsen, D. Ganser, D. Weston, J. Blackwell, and P. Grodzinski, “Highly parallel integrated microfluidic biochannel arrays,” 14th IEEE International Conference on Micro Electro Mechanical Systems, Technical Digest, 2001, 439-442.

[219] A. HolstJensen, L. M. Kohn, and T. Schumacher, “Nuclear rDNA phylogeny of the Sclerotiniaceae,” Mycologia, 1997, 89, 885-899.

[220] The sequence of FB'P and FBN'P could be found in GenBank at http://www.ncbi.nlm.nih.gov.

[221] A. W. Peterson, R. J. Heaton, and R. M. Georgiadis, “The effect of surface probe density on DNA hybridization,” Nucleic Acids Research, 2001, 29, 5163-5168.

[222] T. M. Herne, and M. J. Tarlov, “Characterization of DNA probes immobilized on gold surfaces,” Journal of the American Chemical Society, 1997, 119, 8916-8920.

[223] J. Hager, “Making and using spotted DNA microarrays in an academic core laboratory,” Methods in Enzymology, 2006, 410, 135-168.

[224] M. S. Shchepinov, S. C. CaseGreen, and E. M. Southern, “Steric factors influencing hybridisation of nucleic acids to oligonucleotide arrays,” Nucleic Acids Research, 1997, 25, 1155-1161.

[225] R. Wilson, “The use of gold nanoparticles in diagnostics and detection,” Chemical Society Reviews, 2008, 37, 2028-2045.

[226] Y. Weizmann, F. Patolsky, and I. Willner, “Amplified detection of DNA and analysis of single-base mismatches by the catalyzed deposition of gold on Au-nanoparticles,” Analyst, 2001, 126, 1502-1504.

[227] J. J. Storhoff, S. S. Marla, P. Bao, S. Hagenow, H. Mehta, A. Lucas, V. Garimella, T. Patno, W. Buckingham, W. Cork, and U. R. Muller, “Gold nanoparticle-based detection of genomic DNA targets on microarrays using a novel optical detection system,” Biosensors & Bioelectronics, 2004, 19, 875-883.

[228] T. A. Taton, C. A. Mirkin, and R. L. Letsinger, “Scanometric DNA array detection with nanoparticle probes,” Science, 2000, 289, 1757-1760.

264

[229] B. Dubertret, M. Calame, and A. J. Libchaber, “Single-mismatch detection using gold-quenched fluorescent oligonucleotides,” Nature Biotechnology, 2001, 19, 365-370.

[230] D. J. Maxwell, J. R. Taylor, and S. M. Nie, “Self-assembled nanoparticle probes for recognition and detection of biomolecules,” Journal of the American Chemical Society, 2002, 124, 9606-9612.

[231] S. P. Song, Z. Q. Liang, J. Zhang, L. H. Wang, G. X. Li, and C. H. Fan, “Gold-nanoparticle-based multicolor nanobeacons for sequence-specific DNA analysis,” Angewandte Chemie-International Edition, 2009, 48, 8670-8674.

[232] J. Zhang, L. H. Wang, H. Zhang, F. Boey, S. P. Song, and C. H. Fan, “Aptamer-Based multicolor fluorescent gold nanoprobes for multiplex detection in homogeneous solution,” Small, 2010, 6, 201-204.

[233] M. S. Han, A. K. R. Lytton-Jean, B. K. Oh, J. Heo, and C. A. Mirkin, “Colorimetric screening of DNA-binding molecules with gold nanoparticle probes,” Angewandte Chemie-International Edition, 2006, 45, 1807-1810.

[234] J. J. Storhoff, A. D. Lucas, V. Garimella, Y. P. Bao, and U. R. Muller, “Homogeneous detection of unamplified genomic DNA sequences based on colorimetric scatter of gold nanoparticle probes,” Nature Biotechnology, 2004, 22, 883-887.

[235] K. Hosokawa, K. Sato, N. Ichikawa, and M. Maeda, “Power-free poly(dimethylsiloxane) microfluidic devices for gold nanoparticle-based DNA analysis,” Lab on a Chip, 2004, 4, 181-185.

[236] W. Zhao, M. A. Brook, and Y. F. Li, “Design of gold nanoparticle-based colorimetric biosensing assays,” Chembiochem, 2008, 9, 2363-2371.

[237] H. X. Li, and L. Rothberg, “Colorimetric detection of DNA sequences based on electrostatic interactions with unmodified gold nanoparticles,” Proceedings of the National Academy of Sciences of the United States of America, 2004, 101, 14036-14039.

[238] R. J. Hunter, Foundations of colloid science, 2nd ed., Oxford ; New York: Oxford University Press, 2001.

[239] J. W. Slot, and H. J. Geuze, “A new method of preparing gold probes for multiple-labeling cyto-chemistry,” European Journal of Cell Biology, 1985, 38, 87-93.

[240] M. C. Daniel, and D. Astruc, “Gold nanoparticles: Assembly, supramolecular chemistry, quantum-size-related properties, and applications toward biology, catalysis, and nanotechnology,” Chemical Reviews, 2004, 104, 293-346.

265

[241] H. X. Li, and L. J. Rothberg, “DNA sequence detection using selective fluorescence quenching of tagged oligonucleotide probes by gold nanoparticles,” Analytical Chemistry, 2004, 76, 5414-5417.

[242] J. Griffin, A. K. Singh, D. Senapati, P. Rhodes, K. Mitchell, B. Robinson, E. Yu, and P. C. Ray, “Size- and distance-dependent nanoparticle surface-energy transfer (NSET) method for selective sensing of Hepatitis C virus RNA,” Chemistry-a European Journal, 2009, 15, 342-351.

[243] K. Cho, Y. Lee, C. H. Lee, K. Lee, Y. Kim, H. Choi, P. D. Ryu, S. Y. Lee, and S. W. Joo, “Selective aggregation mechanism of unmodified gold nanoparticles in detection of single nucleotide polymorphism,” Journal of Physical Chemistry C, 2008, 112, 8629-8633.

[244] A. P. Xin, Q. P. Dong, C. Xiong, and L. S. Ling, “Colorimetric recognition of DNA intercalators with unmodified gold nanoparticles,” Chemical Communications, 2009, 1658-1660.

[245] P. C. Ray, “Diagnostics of single base-mismatch DNA hybridization on gold nanoparticles by using the hyper-Rayleigh scattering technique,” Angewandte Chemie-International Edition, 2006, 45, 1151-1154.

[246] H. K. Li, J. H. Huang, J. H. Lv, H. J. An, X. D. Zhang, Z. Z. Zhang, C. H. Fan, and J. Hu, “Nanoparticle PCR: Nanogold-assisted PCR with enhanced specificity,” Angewandte Chemie-International Edition, 2005, 44, 5100-5103.

[247] E. D. Goluch, S. I. Stoeva, J. S. Lee, K. A. Shaikh, C. A. Mirkin, and C. Liu, “A microfluidic detection system based upon a surface immobilized biobarcode assay,” Biosensors & Bioelectronics, 2009, 24, 2397-2403.

[248] B. J. Hong, S. J. Oh, T. O. Youn, S. H. Kwon, and J. W. Park, “Nanoscale-controlled spacing provides DNA microarrays with the SNP discrimination efficiency in solution phase,” Langmuir, 2005, 21, 4257-4261.

[249] M. Fuentes, C. Mateo, L. Garcia, J. C. Tercero, J. M. Guisan, and R. Fernandez-Lafuente, “Directed covalent immobilization of aminated DNA probes on aminated plates,” Biomacromolecules, 2004, 5, 883-888.

[250] P. Gong, G. M. Harbers, and D. W. Grainger, “Multi-technique comparison of immobilized and hybridized oligonucleotide surface density on commercial amine-reactive microarray slides,” Analytical Chemistry, 2006, 78, 2342-2351.

[251] J. H. Watterson, P. A. E. Piunno, C. C. Wust, and U. J. Krull, “Effects of oligonucleotide immobilization density on selectivity of quantitative transduction of hybridization of immobilized DNA,” Langmuir, 2000, 16, 4984-4992.

266

[252] K. C. Grabar, R. G. Freeman, M. B. Hommer, and M. J. Natan, “Preparation and Characterization of Au Colloid Monolayers,” Analytical Chemistry, 1995, 67, 735-743.

[253] S. T. Liu, T. Zhu, R. S. Hu, and Z. F. Liu, “Evaporation-induced self-assembly of gold nanoparticles into a highly organized two-dimensional array,” Physical Chemistry Chemical Physics, 2002, 4, 6059-6062.

[254] K. C. Grabar, P. C. Smith, M. D. Musick, J. A. Davis, D. G. Walter, M. A. Jackson, A. P. Guthrie, and M. J. Natan, “Kinetic control of interparticle spacing in Au colloid-based surfaces: Rational nanometer-scale architecture,” Journal of the American Chemical Society, 1996, 118, 1148-1153.

[255] S.-T. Liu, R.-S. Hu, Z.-H. Zhu, T. Zhu, and Z.-F. Liu, “Reorganization of gold nanoprticles in self-assembled nanostructures,” Acta Physico-Chimica Sinica, 2000, 16, 294-298.

[256] D. Voet, and J. G. Voet, Biochemistry, 3rd ed., New York: J. Wiley & Sons, 2004.

[257] S. V. Lemeshko, T. Powdrill, Y. Y. Belosludtsev, and M. Hogan, “Oligonucleotides form a duplex with non-helical properties on a positively charged surface,” Nucleic Acids Research, 2001, 29, 3051-3058.

[258] Y. Belosludtsev, B. Iverson, S. Lemeshko, R. Eggers, R. Wiese, S. Lee, T. Powdrill, and M. Hogan, “DNA microarrays based on noncovalent oligonucleotide attachment and hybridization in two dimensions,” Analytical Biochemistry, 2001, 292, 250-256.

[259] H. X. Li, and L. J. Rothberg, “Label-free colorimetric detection of specific sequences in genomic DNA amplified by the polymerase chain reaction,” Journal of the American Chemical Society, 2004, 126, 10958-10961.

[260] M. A. Hayat, Colloidal gold : principles, methods, and applications, San Diego: Academic Press, 1989.

[261] J. A. A. West, K. W. Hukari, T. Shepodd, and A. A. Hux, "A high density microfluidic microarray platform for rapid genomic profiling," The 8th International Conference on Miniaturized Systems for Chemistry and Life Sciences, Malmö, Sweden, 2004, pp. 656-658.

[262] R. H. Liu, H. Chen, K. R. Luehrsen, D. Ganser, D. Weston, J. Blackwell, and P. Grodzinski, "Highly parallel integrated microfluidic biochannel arrays," The 14th IEEE International Conference on Micro Electro Mechanical Systems (MEMS 2001), Interlaken, Switzerland, 2001, pp. 439 - 442.

267

[263] X. Y. Peng, P. C. H. Li, L. Wang, H. Z. Yu, M. Parameswaran, and W. L. Chou, "Microfluidic microarray assembly and its applications to multi-sample hybridization," The 9th International Conference on Miniaturized Systems for Chemistry and Life Sciences, Boston, USA, 2005, pp. 823-825.

[264] E. Ward, S. J. Foster, B. A. Fraaije, and H. A. McCartney, “Plant pathogen diagnostics: immunological and nucleic acid-based approaches,” Annals of Applied Biology, 2004, 145, 1-16.

[265] R. Garrett, and C. M. Grisham, Biochemistry, 2nd ed., Fort Worth: Saunders College Pub., 1999.

[266] M. W. Toepke, and D. J. Beebe, “PDMS absorption of small molecules and consequences in microfluidic applications,” Lab on a Chip, 2006, 6, 1484-1486.

[267] J. Mairhofer, K. Roppert, and P. Ertl, “Microfluidic Systems for Pathogen Sensing: A Review,” Sensors, 2009, 9, 4804-4823.

[268] Y. C. M. Staal, M. H. M. van Herwijnen, F. J. van Schooten, and J. H. M. van Delft, “Application of four dyes in gene expression analyses by microarrays,” BMC Genomics, 2005, 6, 101.

[269] W. T. Liu, J. H. Wu, E. S. Y. Li, and E. S. Selamat, “Emission characteristics of fluorescent labels with respect to temperature changes and subsequent effects on DNA microchip studies,” Applied and Environmental Microbiology, 2005, 71, 6453-6457.

[270] J. L. Ballard, V. K. Peeva, C. J. S. deSilva, J. L. Lynch, and N. R. Swanson, “Comparison of Alexa Fluor (R) and CyDye (TM) for practical DNA microarray use,” Molecular Biotechnology, 2007, 36, 175-183.

[271] K. K. Dobbin, E. S. Kawasaki, D. W. Petersen, and R. M. Simon, “Characterizing dye bias in microarray experiments,” Bioinformatics, 2005, 21, 2430-2437.

[272] J. E. Noble, L. Wang, K. D. Cole, and A. K. Gaigalas, “The effect of overhanging nucleotides on fluorescence properties of hybridising oligonucleotides labelled with Alexa-488 and FAM fluorophores,” Biophysical Chemistry, 2005, 113, 255-263.

[273] R. B. Mujumdar, L. A. Ernst, S. R. Mujumdar, C. J. Lewis, and A. S. Waggoner, “Cyanine dye labeling reagents: Sulfoindocyanine succinimidyl esters,” Bioconjugate Chemistry, 1993, 4, 105-111.

[274] Z. D. Wang, J. H. Lee, and Y. Lu, “Highly sensitive "turn-on" fluorescent sensor for Hg2+ in aqueous solution based on structure-switching DNA,” Chemical Communications, 2008, 6005-6007.

268

[275] B. G. Moreira, Y. You, M. A. Behlke, and R. Owczarzy, “Effects of fluorescent dyes, quenchers, and dangling ends on DNA duplex stability,” Biochemical and Biophysical Research Communications, 2005, 327, 473-484.

[276] B. J. Hong, V. Sunkara, and J. W. Park, “DNA microarrays on nanoscale-controlled surface,” Nucleic Acids Research, 2005, 33, e106.

[277] N. Honda, U. Lindberg, P. Andersson, S. Hoffman, and H. Takei, “Simultaneous multiple immunoassays in a compact disc-shaped microfluidic device based on centrifugal force,” Clinical Chemistry, 2005, 51, 1955-1961.

[278] G. Li, Q. Chen, J. J. Li, X. J. Hu, and J. L. Zhao, “A compact disk-like centrifugal microfluidic system for high-throughput nanoliter-scale protein crystallization screening,” Analytical Chemistry, 2010, 82, 4362-4369.

[279] J. Ducree, S. Haeberle, S. Lutz, S. Pausch, F. von Stetten, and R. Zengerle, “The centrifugal microfluidic bio-disk platform,” Journal of Micromechanics and Microengineering, 2007, 17, S103-S115.

[280] R. Gorkin, J. Park, J. Siegrist, M. Amasia, B. S. Lee, J. M. Park, J. Kim, H. Kim, M. Madou, and Y. K. Cho, “Centrifugal microfluidics for biomedical applications,” Lab on a Chip, 2010, 10, 1758-1773.

[281] H. Cho, H. Y. Kim, J. Y. Kang, and T. S. Kim, “How the capillary burst microvalve works,” Journal of Colloid and Interface Science, 2007, 306, 379-385.

[282] F. S. Ligler, and J. S. Erickson, “Bioengineering: Diagnosis on disc,” Nature, 2006, 440, 159-160.

[283] L. J. Yang, T. J. Yao, and Y. C. Tai, “The marching velocity of the capillary meniscus in a microchannel,” Journal of Micromechanics and Microengineering, 2004, 14, 220-225.

[284] A. A. Saha, and S. K. Mitra, “Numerical study of capillary flow in microchannels with alternate hydrophilic-hydrophobic bottom wall,” Journal of Fluids Engineering-Transactions of the ASME, 2009, 131, 061202.

[285] X. Y. Peng, and P. C. H. Li, “Centrifugal pumping in the equiforce spiral microchannel,” Canadian Journal of Pure and Applied Sciences, 2008, 2, 551-556.

[286] L. Wang, M. C. Kropinski, and P. C. H. Li, “Analysis and modeling of flow in rotating spiral microchannels: towards math-aided design of microfluidic systems using centrifugal pumping,” Lab on a Chip, 2011, 11, 2097-2108.

[287] G. K. Batchelor, An introduction to fluid dynamics, Cambridge,: U.P., 1967.

269

[288] A. A. Saha, S. K. Mitra, M. Tweedie, S. Roy, and J. McLaughlin, “Experimental and numerical investigation of capillary flow in SU8 and PDMS microchannels with integrated pillars,” Microfluidics and Nanofluidics, 2009, 7, 451-465.

[289] N. Ichikawa, K. Hosokawa, and R. Maeda, “Interface motion of capillary-driven flow in rectangular microchannel,” Journal of Colloid and Interface Science, 2004, 280, 155-164.

[290] D. S. Kim, K. C. Lee, T. H. Kwon, and S. S. Lee, “Micro-channel filling flow considering surface tension effect,” Journal of Micromechanics and Microengineering, 2002, 12, 236-246.

[291] J. U. Brackbill, D. B. Kothe, and C. Zemach, “A Continuum Method for Modeling Surface-Tension,” Journal of Computational Physics, 1992, 100, 335-354.

[292] H. F. Davis, and A. D. Snider, Introduction to vector analysis, 7th ed., Dubuque, IA: Wm. C. Brown, 1995.

[293] C. W. Hirt, and B. D. Nichols, “Volume of fluid (VOF) method for the dynamics of free boundaries,” Journal of Computational Physics, 1981, 39, 201-225.

[294] E. Southern, K. Mir, and M. Shchepinov, “Molecular interactions on microarrays,” Nature Genetics, 1999, 21, 5-9.

[295] R. J. Britten, and E. H. Davidson, “Studies on nucleic-acid reassociation kinetics: Empirical equations describing DNA reassociation,” Proceedings of the National Academy of Sciences of the United States of America, 1976, 73, 415-419.

[296] P. Schuck, and A. P. Minton, “Analysis of mass transport-limited binding kinetics in evanescent wave biosensors,” Analytical Biochemistry, 1996, 240, 262-272.

[297] R. A. Vijayendran, F. S. Ligler, and D. E. Leckband, “A computational reaction-diffusion model for the analysis of transport-limited kinetics,” Analytical Chemistry, 1999, 71, 5405-5412.

[298] D. G. Myszka, X. He, M. Dembo, T. A. Morton, and B. Goldstein, “Extending the range of rate constants available from BIACORE: Interpreting mass transport-influenced binding data,” Biophysical Journal, 1998, 75, 583-594.

[299] W. Kusnezow, Y. V. Syagailo, S. Ruffer, K. Klenin, W. Sebald, J. D. Hoheisel, C. Gauer, and I. Goychuk, “Kinetics of antigen binding to antibody microspots: Strong limitation by mass transport to the surface,” Proteomics, 2006, 6, 794-803.

270

[300] D. E. Leckband, and R. A. Vijayendran, “A quantitative assessment of heterogeneity for surface-immobilized proteins,” Analytical Chemistry, 2001, 73, 471-480.

[301] P. Schuck, “Use of surface plasmon resonance to probe the equilibrium and dynamic aspects of interactions between biological macromolecules,” Annual Review of Biophysics and Biomolecular Structure, 1997, 26, 541-566.

[302] S. Blair, J. Bishop, and A. M. Chagovetz, “Kinetics of multiplex hybridization: Mechanisms and implications,” Biophysical Journal, 2008, 94, 1726-1734.

[303] J. Bishop, S. Blair, and A. Chagovetz, “Convective flow effects on DNA biosensors,” Biosensors & Bioelectronics, 2007, 22, 2192-2198.

[304] D. Axelrod, and M. D. Wang, “Reduction-of-dimensionality kinetics at reaction-limited cell-surface receptors,” Biophysical Journal, 1994, 66, 588-600.

[305] V. Chan, D. J. Graves, and S. E. McKenzie, “The biophysics of DNA hybridization with immobilized oligonucleotide probes,” Biophysical Journal, 1995, 69, 2243-2255.

[306] V. Chan, D. J. Graves, P. Fortina, and S. E. McKenzie, “Adsorption and surface diffusion of DNA oligonucleotides at liquid/solid interfaces,” Langmuir, 1997, 13, 320-329.

[307] R. Singh, J. Nitsche, and S. T. Andreadis, “An Integrated Reaction-Transport Model for DNA Surface Hybridization: Implications for DNA Microarrays,” Annals of Biomedical Engineering, 2009, 37, 255-269.

[308] C. Gadgil, A. Yeckel, J. J. Derby, and W. S. Hu, “A diffusion-reaction model for DNA microarray assays,” Journal of Biotechnology, 2004, 114, 31-45.

[309] D. Erickson, D. Q. Li, and U. J. Krull, “Modeling of DNA hybridization kinetics for spatially resolved biochips,” Analytical Biochemistry, 2003, 317, 186-200.

[310] M. Zimmermann, E. Delamarche, M. Wolf, and P. Hunziker, “Modeling and optimization of high-sensitivity, low-volume microfluidic-based surface immunoassays,” Biomedical Microdevices, 2005, 7, 99-110.

[311] T. M. Squires, R. J. Messinger, and S. R. Manalis, “Making it stick: convection, reaction and diffusion in surface-based biosensors,” Nature Biotechnology, 2008, 26, 417-426.

[312] T. Gervais, and K. F. Jensen, “Mass transport and surface reactions in microfluidic systems,” Chemical Engineering Science, 2006, 61, 1102-1121.

271

[313] J. R. Pascault, and H. S. Zhou, “Enhancement of DNA hybridization kinetics in microarrays by convective transport,” Chemical Engineering Communications, 2008, 195, 167-186.

[314] J. Berthier, and P. Silberzan, Microfluidics for biotechnology, 2nd ed., Boston: Artech House, 2010.

[315] K. Pappaert, and G. Desmet, “A dimensionless number analysis of the hybridization process in diffusion- and convection-driven DNA microarray systems,” Journal of Biotechnology, 2006, 123, 381-396.

[316] D. J. Acheson, Elementary fluid dynamics, New York: Oxford University Press, 1990.

[317] T. M. Squires, and S. R. Quake, “Microfluidics: Fluid physics at the nanoliter scale,” Reviews of Modern Physics, 2005, 77, 977-1026.

[318] A. K. Vuppu, A. A. Garcia, S. K. Saha, P. E. Phelan, M. A. Hayes, and R. Calhoun, “Modeling microflow and stirring around a microrotor in creeping flow using a quasi-steady-state analysis,” Lab on a Chip, 2004, 4, 201-208.

[319] H. Bruus, Theoretical microfluidics, Oxford ; New York: Oxford University Press, 2008.

[320] S. Jomeh, and M. Hoorfar, “Numerical modeling of mass transport in microfluidic biomolecule-capturing devices equipped with reactive surfaces,” Chemical Engineering Journal, 2010, 165, 668-677.

[321] V. Chan, S. E. McKenzie, S. Surrey, P. Fortina, and D. J. Graves, “Effect of hydrophobicity and electrostatics on adsorption and surface diffusion of DNA oligonucleotides at liquid/solid interfaces,” Journal of Colloid and Interface Science, 1998, 203, 197-207.

[322] D. Pastre, L. Hamon, I. Sorel, E. Le Cam, P. A. Curmi, and O. Pietrement, “Specific DNA-protein interactions on mica investigated by atomic force microscopy,” Langmuir, 2010, 26, 2618-2623.

[323] R. F. Probstein, Physicochemical hydrodynamics : an introduction, 2nd ed., Hoboken, N.J.: John Wiley & Sons, Inc., 2003.

[324] M. F. Hagan, and A. K. Chakraborty, “Hybridization dynamics of surface immobilized DNA,” Journal of Chemical Physics, 2004, 120, 4958-4968.

[325] B. A. Grzybowski, Chemistry in motion : reaction-diffusion systems for micro- and nanotechnology, Chichester, U.K.: Wiley, 2009.

[326] D. Mocanu, A. Kolesnychenko, S. Aarts, A. Troost-Dejong, A. Pierik, E. Vossenaar, and H. Stapert, “Mass transfer effects on DNA hybridization in a flow-through microarray,” Journal of Biotechnology, 2009, 139, 179-185.

272

[327] A. Jayaraman, C. K. Hall, and J. Genzer, “Computer simulation study of probe-target hybridization in model DNA microarrays: Effect of probe surface density and target concentration,” Journal of Chemical Physics, 2007, 127, -.

[328] D. S. Dandy, P. Wu, and D. W. Grainger, “Array feature size influences nucleic acid surface capture in DNA microarrays,” Proceedings of the National Academy of Sciences of the United States of America, 2007, 104, 8223-8228.

[329] B. Tinland, A. Pluen, J. Sturm, and G. Weill, “Persistence length of single-stranded DNA,” Macromolecules, 1997, 30, 5763-5765.

[330] M. K. Liu, and J. C. Giddings, “Separation and measurement of diffusion-coefficients of linear and circular DNAs by flow field-flow fractionation,” Macromolecules, 1993, 26, 3576-3588.

[331] E. Stellwagen, Y. J. Lu, and N. C. Stellwagen, “Unified description of electrophoresis and diffusion for DNA and other polyions,” Biochemistry, 2003, 42, 11745-11750.

[332] A. E. Nkodo, J. M. Garnier, B. Tinland, H. J. Ren, C. Desruisseaux, L. C. McCormick, G. Drouin, and G. W. Slater, “Diffusion coefficient of DNA molecules during free solution electrophoresis,” Electrophoresis, 2001, 22, 2424-2432.

[333] D. E. Smith, R. M. Robertson, and S. Laib, “Diffusion of isolated DNA molecules: Dependence on length and topology,” Proceedings of the National Academy of Sciences of the United States of America, 2006, 103, 7310-7314.

[334] J. R. Borden, C. J. Paredes, and E. T. Papoutsakis, “Diffusion, mixing, and associated dye effects in DNA-microarray hybridizations,” Biophysical Journal, 2005, 89, 3277-3284.

[335] W. Michel, T. Mai, T. Naiser, and A. Ott, “Optical study of DNA surface hybridization reveals DNA surface density as a key parameter for microarray hybridization kinetics,” Biophysical Journal, 2007, 92, 999-1004.

[336] H. P. Lehr, M. Reimann, A. Brandenburg, G. Sulz, and H. Klapproth, “Real-time detection of nucleic acid interactions by total internal reflection fluorescence,” Analytical Chemistry, 2003, 75, 2414-2420.

[337] M. R. Henry, P. W. Stevens, J. Sun, and D. M. Kelso, “Real-time measurements of DNA hybridization on microparticles with fluorescence resonance energy transfer,” Analytical Biochemistry, 1999, 276, 204-214.

273

[338] L. Q. Wang, and Z. K. Lu, “Analyzing the dynamic range of fluorescent detection in microarray scanner,” Proceedings of International Symposium on Biophotonics, Nanophotonics and Metamaterials, 2006, 128-130.

[339] H. Parsa, C. D. Chin, P. Mongkolwisetwara, B. W. Lee, J. J. Wang, and S. K. Sia, “Effect of volume- and time-based constraints on capture of analytes in microfluidic heterogeneous immunoassays,” Lab on a Chip, 2008, 8, 2062-2070.

[340] C. Phillips, M. Jakusch, H. Steiner, B. Mizaikoff, and A. G. Fedorov, “Model-based optimal design of polymer-coated chemical sensors,” Analytical Chemistry, 2003, 75, 1106-1115.

[341] O. Hofmann, G. Voirin, P. Niedermann, and A. Manz, “Three-dimensional microfluidic confinement for efficient sample delivery to biosensor surfaces. Application to immunoassays on planar optical waveguides,” Analytical Chemistry, 2002, 74, 5243-5250.

[342] K. Yamashita, M. Miyazaki, Y. Yamaguchi, H. Nakamura, and H. Maeda, “The change of activation energy in microchannel laminar flow as demonstrated by kinetic analysis of the DNA duplex-coil equilibrium,” Lab on a Chip, 2008, 8, 1171-1177.

[343] J. C. Walter, K. M. Kroll, J. Hooyberghs, and E. Carlon, “Nonequilibrium effects in dna microarrays: A multiplatform study,” Journal of Physical Chemistry B, 2011, 115, 6732-6739.

[344] J. Hooyberghs, M. Baiesi, A. Ferrantini, and E. Carlon, “Breakdown of thermodynamic equilibrium for DNA hybridization in microarrays,” Physical Review E, 2010, 81, 012901.

[345] A. Halperin, A. Buhot, and E. B. Zhulina, “On the hybridization isotherms of DNA microarrays: the Langmuir model and its extensions,” Journal of Physics-Condensed Matter, 2006, 18, S463-S490.

[346] K. J. Breslauer, R. Frank, H. Blocker, and L. A. Marky, “Predicting DNA Duplex Stability from the Base Sequence,” Proceedings of the National Academy of Sciences of the United States of America, 1986, 83, 3746-3750.

[347] D. Tulpan, M. Andronescu, and S. Leger, “Free energy estimation of short DNA duplex hybridizations,” BMC Bioinformatics, 2010, 11, 105.

[348] S. Weckx, E. Carlon, L. De Vuyst, and P. Van Hummelen, “Thermodynamic behavior of short oligonucleotides in microarray hybridizations can be described using Gibbs free energy in a nearest-neighbor model,” Journal of Physical Chemistry B, 2007, 111, 13583-13590.

274

[349] N. R. Markham, and M. Zuker, “DINAMelt web server for nucleic acid melting prediction,” Nucleic Acids Research, 2005, 33, W577-W581.

[350] J. B. Fiche, A. Buhot, R. Calemczuk, and T. Livache, “Temperature effects on DNA chip experiments from surface plasmon resonance imaging: Isotherms and melting curves,” Biophysical Journal, 2007, 92, 935-946.

[351] P. Bajcsy, "An overview of DNA microarray image requirements for automated processing," The IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, California, USA, 2005, pp. 147.

[352] K. Pappaert, H. Ottevaere, H. Thienpont, P. Van Hummelen, and G. Desmet, “Diffusion limitation: a possible source for the occurrence of doughnut patterns on DNA microarrays,” Biotechniques, 2006, 41, 609-616.

[353] J. M. Moran-Mirabal, C. P. Tan, R. N. Orth, E. O. Williams, H. G. Craighead, and D. M. Lin, “Controlling microarray spot morphology with polymer liftoff arrays,” Analytical Chemistry, 2007, 79, 1109-1114.

[354] Q. H. Li, C. Fraley, R. E. Bumgarner, K. Y. Yeung, and A. E. Raftery, “Donuts, scratches and blanks: robust model-based segmentation of microarray images,” Bioinformatics, 2005, 21, 2875-2882.

[355] R. D. Deegan, O. Bakajin, T. F. Dupont, G. Huber, S. R. Nagel, and T. A. Witten, “Capillary flow as the cause of ring stains from dried liquid drops,” Nature, 1997, 389, 827-829.

[356] R. Blossey, and A. Bosio, “Contact line deposits on cDNA microarrays: A "twin-spot effect",” Langmuir, 2002, 18, 2952-2954.

[357] W. C. W. Chan, T. L. Jennings, K. S. Rahman, and S. Fournier-Bidoz, “Effects of microbead surface chemistry on DNA loading and hybridization efficiency,” Analytical Chemistry, 2008, 80, 2849-2856.

[358] S. Lane, J. Evermann, F. Loge, and D. R. Call, “Amplicon secondary structure prevents target hybridization to oligonucleotide microarrays,” Biosensors & Bioelectronics, 2004, 20, 728-735.

[359] Y. Gao, L. K. Wolf, and R. M. Georgiadis, “Secondary structure effects on DNA hybridization kinetics: a solution versus surface comparison,” Nucleic Acids Research, 2006, 34, 3370-3377.

[360] R. B. Towery, N. C. Fawcett, P. Zhang, and J. A. Evans, “Genomic DNA hybridizes with the same rate constant on the QCM biosensor as in homogeneous solution,” Biosensors & Bioelectronics, 2001, 16, 1-8.