micro- and nano-scale approaches for disease detection and ... · ii micro- and nano-scale...
TRANSCRIPT
Micro- and Nano-Scale Approaches for Disease Detection and Characterization
by
Adam Hunter Mepham
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Institute of Biomaterials and Biomedical Engineering University of Toronto
© Copyright by Adam Hunter Mepham 2019
ii
Micro- and Nano-Scale Approaches for Disease Detection and
Characterization
Adam Hunter Mepham
Doctor of Philosophy
Institute of Biomaterials and Biomedical Engineering
University of Toronto
2019
Abstract
Devices that diagnose and characterize disease have the potential to greatly improve healthcare
worldwide. This thesis explores a number of different elements pertaining to device design and
application. A microfluidic device for the capture and profiling of circulating tumor cells (CTCs)
is tested against a rabbit model of cancer. This device demonstrates both an increase in CTC load
and aggressiveness which correlates with traditional computed tomography measurements. CTC
biology is also shown to differ markedly from tumour precursor cells. Next, a study of gold
microelectrode architecture is performed with the aim of improving performance towards
biosensing. A unique regime of gold ion concentration, applied voltage, and electrolyte viscosity
is determined which drives the assembly of a highly structured morphology. Further studies
illustrate growth mechanisms and the sensitivity of the electrode towards biomolecule detection.
Additionally, a microfluidic device for instrument-free manipulation of microscopic fluid
quantities is developed. This design allows the metering and dispensing of reagents in an intuitive
manner by combining a series of capillary valves and a simple push-button. This “Digit Chip” is
applied to the detection of antibacterial susceptibility alongside a simple smart-phone based
fluorimeter. Together these studies explore the application of electrochemical and microfluidic
modalities to the realm of disease monitoring.
iii
Acknowledgments
I would like to begin by thanking Professor Shana Kelley for her support throughout the tenure of
my PhD. Her unique blend of creativity and practicality helped me to pursue interesting projects
which nonetheless had clear and important applications in the world of biotechnology. Moreover,
her ability to assemble a group of friendly and intelligent individuals helped to create a warm and
interesting environment in which to practice research.
I would further like to thank the members of my advisory committee, Professor Ted Sargent,
Professor Aaron Wheeler, and Professor Axel Guenther. They all helped to keep me motivated
and see the larger picture in my work. Their guidance served me well in planning out my projects
and keeping my destination clear.
Next I would like to thank the innumerable labmates who helped to make my rough days tolerable
and my good days even better. My appreciation goes out to Dr. Jagotamoy Das, Dr. Mahmoud
Labib, Dr. Brenda Green, Dr. Libing Zhang, Surath Gomis, Fan Xia, Bill Duong, Zongjie Wang,
Dr. Ivaylo Ivanov, Laili Mahmoudian, Dr. David Tulumello, Dr. Brian Lam, Dr. Mahla Poudineh,
Dr. Andrew Sage, Dr. Sae Rin Jean, Dr. Simon Wisnovsky, Dr. Leyla Kermanshah, Dr. Yige Zhou,
Dr. Ying Wan, Dr. Sam Chang, Dr. Julie Shi, Alexandre Zaragoza, Thy Vu, and Thaddeus “Thad”
Gibbs. Special thanks to those who provided invaluable assistance in my publications, Dr. Reza
Mohamadi, Dr. Sharif Ahmed, Dr. Justin Besant, Dr. Ian Burgess, Dr. Sahar Mahshid and Dr. Sara
Mahshid. Further thanks to those members who keep the Kelley lab running smoothing;
specifically, Dr. Mark Pereira, Bob Christensen and especially Barbara Alexander. Finally, a shout
out to those who helped keep my days fun and interesting, Dr. Wendi Zhou, Peter Aldridge,
Wenhan Liu, Carine Nemr, Tanja Sack, Jenise Chen, David Philpott, Xiaolong Yang, Cindy Ma,
Hanie Yousefi, Dr. Eric Lei, , Dr. Tina Saberi Safaei, and Dr. Sarah Smith.
Last but not least, I would like to thank my family for all of their support over the years. I would
not have survived without the steady supply of peanut butter and almonds which fueled my body
and mind. I want to thank my parents, Peter and Linda, who have both encouraged me to pursue
what I enjoy. My brother, who called me a nerd. And my entire extended family, who pretend that
they understand when I ramble on about my work.
iv
Table of Contents
Acknowledgments.......................................................................................................................... iii
Table of Contents ........................................................................................................................... iv
List of Tables ............................................................................................................................... viii
List of Figures ................................................................................................................................ ix
List of Abbreviations .................................................................................................................... xii
Introduction .....................................................................................................................1
1.1 Point-of-Care Diagnostics ....................................................................................................1
1.2 Electrochemistry ..................................................................................................................3
1.3 Electrode Properties and Design ..........................................................................................6
1.4 Microfluidics ........................................................................................................................7
1.5 Scope of Thesis ....................................................................................................................9
Sorting of Circulating Tumor Cells during Disease Progression in an Animal Model
...................................................................................................................................................11
2.1 Abstract ..............................................................................................................................12
2.2 Introduction ........................................................................................................................12
2.3 Results and Discussion ......................................................................................................14
2.3.1 Animal study and microfluidic chip design ...........................................................14
2.3.2 Characterization of device function .......................................................................16
2.3.3 Measurement of tumor volumes and analysis of metastatic lymph nodes.............17
2.3.4 Correlation of CTC levels with disease progression..............................................18
2.3.5 Analysis of CTC sorting patterns during tumor growth ........................................20
2.3.6 Discussion ..............................................................................................................22
2.4 Conclusion .........................................................................................................................24
2.5 Methods..............................................................................................................................25
v
2.5.1 Animal model.........................................................................................................25
2.5.2 Tumor cell line propagation ...................................................................................25
2.5.3 CT imaging and image analysis .............................................................................25
2.5.4 Histopathological evaluation .................................................................................26
2.5.5 Capture of cell lines ...............................................................................................26
2.5.6 Immunocytochemistry of cell lines ........................................................................26
2.5.7 Isolation and fluorescent staining of CTCs ............................................................26
2.5.8 Capture efficiency of white blood cells .................................................................27
2.5.9 Image scanning and analysis ..................................................................................27
2.5.10 Flow cytometry ......................................................................................................27
2.5.11 Microchip fabrication.............................................................................................27
Mechanistic Control of the Growth of Three-Dimensional Gold Sensors ....................29
3.1 Abstract ..............................................................................................................................30
3.2 Introduction ........................................................................................................................30
3.3 Results and Discussion ......................................................................................................32
3.3.1 Outline of experiments ...........................................................................................32
3.3.2 Effects of solution conditions of 3D nanostructured microelectrode growth ........33
3.3.3 Nucleation and growth mechanism of 3D gold structures .....................................39
3.3.4 Study of current-time transients .............................................................................39
3.3.5 Analysis of i2/im2 vs t/tm .........................................................................................41
3.3.6 Surface area of nanostructured 3D microelectrodes ..............................................45
3.3.7 Study of DNA hybridization efficiency on the surface of 3D microelectrodes .....46
3.4 Conclusion .........................................................................................................................48
3.5 Methods..............................................................................................................................48
3.5.1 Chip fabrication .....................................................................................................48
vi
3.5.2 Electrodeposition ...................................................................................................48
3.5.3 Surface characterization .........................................................................................49
3.5.4 Nucleation and growth model ................................................................................49
3.5.5 Electrode functionalization ....................................................................................50
3.5.6 Sensor measurements .............................................................................................50
Power-free, digital and programmable dispensing of picoliter droplets using a Digit
Chip ...........................................................................................................................................51
4.1 Abstract ..............................................................................................................................52
4.2 Introduction ........................................................................................................................52
4.3 Results and discussion .......................................................................................................54
4.3.1 Overview of the digit chip .....................................................................................54
4.3.2 Bursting pressure model and design principles .....................................................56
4.3.3 Optimization of device geometry...........................................................................57
4.3.4 Designing a user-friendly interface ........................................................................59
4.3.5 Generation of a discrete concentration gradient ....................................................62
4.3.6 A low-cost platform for rapid determination of bacterial antibiotic susceptibility
................................................................................................................................63
4.3.7 Discussion ..............................................................................................................65
4.4 Conclusion .........................................................................................................................67
4.5 Methods..............................................................................................................................67
4.5.1 Digit chip fabrication .............................................................................................67
4.5.2 Fabrication of user-friendly interface ....................................................................67
4.5.3 Contact angle measurements..................................................................................67
4.5.4 Bursting pressure measurements............................................................................68
4.5.5 Measurements of applied pressure using the elastomeric button ...........................68
4.5.6 Chamber filling percentage measurements ............................................................68
vii
4.5.7 Generation of a discretized concentration gradient ...............................................68
4.5.8 Antibiotic susceptibility testing .............................................................................69
4.5.9 Fluorescence image acquisition and analysis ........................................................69
Conclusions and Future Outlook ...................................................................................70
5.1 Thesis Findings ..................................................................................................................70
5.2 Future Outlook ...................................................................................................................71
References ......................................................................................................................................74
Single Cell Capture Device .......................................................................................83
A.1 Background ........................................................................................................................83
A.2 Device Design ....................................................................................................................84
Supporting Information ...........................................................................................104
Supporting Information for Chapter 2 .............................................................................104
Supporting Information for Chapter 4 .............................................................................107
viii
List of Tables
Table A-1 Truth table for the multiplexer. .................................................................................... 98
ix
List of Figures
Figure 2-1 Overview of the study monitoring CTC heterogeneity in a rabbit cancer model. ...... 14
Figure 2-2 Characterization of capture. ........................................................................................ 16
Figure 2-3 VX2 tumor growth in New Zealand rabbits. ............................................................... 17
Figure 2-4 CTC sorting and analysis. ........................................................................................... 19
Figure 2-5 Comparison of sorting profile ..................................................................................... 22
Figure 3-1 Schematic of gold microelectrode experiemnts .......................................................... 32
Figure 3-2 Effects of concentration, viscosity, and voltage on electrodeposition of gold
microsensors are studied using scanning electron microscopy (SEM). ........................................ 35
Figure 3-3 Analysis of the interior structures of electrodeposited gold using FIB. ...................... 38
Figure 3-4 I-t curves during electrodeposition.............................................................................. 40
Figure 3-5 Nucleation during electrodeposition ........................................................................... 42
Figure 3-6 Two-dimensional time-varying simulation results for deposition of Au .................... 44
Figure 3-7 Effect of deposition overpotential on surface nanostructuring. .................................. 45
Figure 3-8 DNA detection assay based on 3D gold microsensors................................................ 47
Figure 4-1 Overview of the Digit Chip. ........................................................................................ 55
Figure 4-2 Experimental investigation of the device geometry and its optimization. .................. 58
Figure 4-3 The Digit Chip interface for controlled dispensing of droplets. ................................. 60
Figure 4-4 Generation of a discretized concentration gradient. .................................................... 62
Figure 4-5 Testing of antibiotic susceptibility. ............................................................................. 63
x
Figure A-1 Structure of a microfluidic weir. ................................................................................ 85
Figure A-2 A microfluidic weir with a parallel shunt channel. .................................................... 86
Figure A-3 Equivalent circuit diagram for a sequence of weir-shunt pairs. ................................. 87
Figure A-4 Encouraging flow to enter the weir. ........................................................................... 87
Figure A-5 Reducing shunt resistance by translating cells across streamlines. ............................ 88
Figure A-6 Design of the gravity driven cell alignment and capture device, which is operated in
the vertical orientation. ................................................................................................................. 90
Figure A-7 Function of the bleeding flow motif. .......................................................................... 91
Figure A-8 Two stage cell aligner. ............................................................................................... 92
Figure A-9 First “bleeding flow” element. ................................................................................... 93
Figure A-10 End of the first “bleeding flow” element. ................................................................ 93
Figure A-11 Reintroduction of removed fluid. ............................................................................. 94
Figure A-12 Second “bleeding flow” element. ............................................................................. 95
Figure A-13 The cell alignment module interfaced with the weir and shunt capture module. ..... 96
Figure A-14 Single cell captured in weir trap. .............................................................................. 96
Figure A-15 Photograph of integrated devce including membrane deflection valves. ................. 97
Figure A-16 Complete design of integrated devce. ...................................................................... 97
Figure A-17 Pressure manifold with three 4-way valves (left) and two 3-way valves (right). .... 99
Figure A-18 Pressure manifold feeding pressurized water reservoirs. ......................................... 99
Figure A-19 Cell capture region of the devce including all 8 valves.. ....................................... 100
xi
Figure A-20 A single trap in capture mode. ............................................................................... 101
Figure A-21 A single trap in incubation mode. .......................................................................... 102
Figure A-22 A single trap in release mode. ................................................................................ 103
Figure B-1 Nanoparticle characterization. .................................................................................. 105
Figure B-2 Cell line flow cytometry. .......................................................................................... 106
Figure B-3 VX2 EpCAM flow cytometry. ................................................................................. 106
Figure B-4 The critical advancing contact angle of PBS on PDMS as a function of oxygen plasma
exposure time. ............................................................................................................................. 107
Figure B-5 Accuracy of filling as a function of chamber size. ................................................... 107
Figure B-6 A replication of Figure 4-2 F-H using buttom pressess rather than using a hydraulic
pump, .......................................................................................................................................... 108
xii
List of Abbreviations
BSA: Bovine serum albumin
CT: Computed tomography
CTC: Circulating tumor cell
DI: Deionized
EMT: Epithelial-to-mesenchymal transition
EpCAM: Epithelial cell adhesion molecule
FDA: Food and drug administration
FIB: Focused ion beam
H&E: Hematoxylin and eosin
HBSS: Hank’s balanced salt solution
HCG: Human chorionic gonadotropin
LED: Light-emitting diode
MACS: Magnetic-activated cell sorting
MCH: Mercaptohexanol
PBS: Phosphate-buffered saline
PDMS: Polydimethylsiloxane
POC: Point-of-care
RGB: Red, green, blue
SEM: Scanning electron microscope
SWV: Square-wave voltammetry
TCEP: Tris(2-carboxyethyl)phosphine
VOI: Volume of interest
WBC: White blood cell
1
Introduction
1.1 Point-of-Care Diagnostics
There is currently a shift occurring in the realm of disease diagnostics. This transformation is
harnessing the power of innovative, more precise technologies to provide deeper and more
personalized insight into the disease state and to facilitate the development of treatments more
closely tailored to the unique needs of individual patients. Moreover, these approaches
significantly expedite testing and allow for more direct communication between clinical physicians
and patients. This represents a marked shift from previously recognized models of medical care.
The gold standard for medical testing has historically been laboratory testing.1 This approach
served the needs of patients well throughout the 20th century, as the application of medicine
became more systematic. Traditional laboratory tests were performed by trained professionals in
a controlled setting, and results were later communicated to clinicians and, in turn, patients. This
approach allowed the powerful techniques developed in the laboratory, such as bacterial culture
and genetic testing, to be brought to bear on the problems of identifying illnesses and determining
prognosis and appropriate treatments. This paradigm allowed for batch testing of large numbers of
samples, and provided accurate and actionable information.
Despite its utility, the traditional laboratory-based model suffers from a number of critical flaws.
Firstly, these tests often take a considerable amount of time. For example, traditional approaches
to identifying bacterial infections require a number of steps which can take multiple days to
complete.2 These typically include a culture stage, which serves to grow bacteria to a suitable
number, as well as multiple additional tests such as disc-diffusion assays and mass-spectrometric
testing. In critical situations, such as bacterial sepsis, such a delay can have a severe impact on
patient mortality and morbidity.
2
Secondly, testing is performed at a central laboratory, physically removed from the patient and
clinician. This has two major drawbacks. First, additional time is wasted as the samples must be
transferred to the central location prior to testing, and, in turn, the results must be communicated
to the patient. Additionally, due to the requirement of a large, centralized laboratory, this approach
is not feasible in regions where existing infrastructure is unavailable, such as low-resource
settings.3
Finally, laboratory testing is devised to have utility for the mass population, and as such is typically
not designed to address the idiosyncrasies of particular patients. As the importance of subtle
differences in diseases such as cancer have become more apparent, the need for tests that can give
deeper insight into each patient and individualize patient care has become more concrete.4 These
weaknesses in the traditional approach to probing disease have motivated the development of novel
diagnostic approaches.
Flaws in the existing paradigm for diagnostic testing are being addressed by the development of
point-of-care (POC) devices. POC devices are defined to be diagnostic devices which are used at
or near the location where patient care is administered. This approach has a number of benefits
when compared to centralized laboratory testing. POC testing is typically much faster than
laboratory assays, reducing the time required for diagnosis by as much as an order of magnitude.1
Such devices are also designed to require minimal external equipment, allowing for their use in
low-resource settings. Furthermore, a number of POC devices are designed to detect rare genetic
or protein markers, which allows treatment to be customized on a patient-by-patient basis.5
Most POC devices are examples of biosensors. A biosensor is defined as an apparatus that couples
a biological sensing element to a transducer in order to detect or quantify chemical species,
typically in a sample of biological origin.6 There are a wide variety of useful sensing elements and
transducers, allowing for a large number of possible device designs. The most important property
of a sensing element is the ability to recognize the chemical species of interest, especially in the
presence of a complex sample matrix. As such, two prominent classes of sensing elements are
proteins and nucleic acids. Among proteins, antibodies and enzymes are the most used, due to their
tendency to bind/catalyze a specific target or family of targets.7,8 For nucleic acids, DNA is often
chosen due to its ability to form a duplex with complementary strands of interest.9 DNA has a
number of excellent properties for a biosensor. It can detect both other DNA strands (e.g. genomic
3
DNA) and RNA (e.g. mRNA). It is also highly sensitive to nucleotide sequence, allowing for the
discrimination of mutant strands. Finally, it is stable and can be easily modified with chemical
handles, allowing its incorporation into diverse device architectures.10 In addition, a multitude of
transducers exist which lend themselves to varied applications. Each different transducer type is
tied to a different physicochemical phenomenon and provides a different method of signal readout.
Optical biosensors rely on changes in absorbance, fluorescence or luminescence, for detection
either by a photodetector or the human eye.11 Mechanical biosensors exploit changes of properties
such as resonance frequencies or exerted forces upon the binding of target molecules.12
Electrochemical biosensors measure changes in voltage, impedance or current produced by the
presence of the target.13
One POC biosensor that illustrates the utility of this modality is the home pregnancy test.14 This
test detects the presence in urine of human chorionic gonadotropin (HCG), a peptide hormone
produced by the placenta. Devices employ an antibody specific against HCG. In a traditional
pregnancy test, this antibody is tethered to an enzyme which catalyzes the reaction of a
chromogenic substrate. The presence of HCG localizes the antibodies to a particular strip on the
device, in turn localizing the chromogenic reaction and allowing the visual readout of a colored
band. The pregnancy test is a lateral flow assay, with the flow of urine through the device driven
by capillary action through a porous flow pad. This approach allows for the passive movement of
fluid along a device without the need for an active pumping mechanism.
Another POC biosensor is the handheld glucose monitor.15 This device quantifies the levels of
glucose in the blood, primarily for use by diabetic patients. The sensing element employed is an
enzyme, typically either glucose oxidase or glucose dehydrogenase. These enzymes catalyze the
oxidation of glucose and the concomitant reduction of an electrochemically-active mediator. This
mediator is in turn detected at an electrode, with the current produced ultimately proportional to
the glucose concentration. This electrochemical device is one of a burgeoning number of
biosensors that utilize electrochemical events as output.
1.2 Electrochemistry
Electrochemistry is a field that exists at the intersection of electricity and chemistry. Specifically,
it involves the study of the influence of electrical phenomena on chemical reactions, and,
conversely, the production of electrical current and potentials as a result of chemical reactions.
4
The underlying reason for this interrelation is the fact that all matter is, at the atomic level,
comprised of charged particles. Consequently, electric fields can influence the energy of different
chemical states, and therefore the application of electric fields can cause molecular composition
and molecular distribution to change accordingly.
The fundamental reaction underlying electrochemistry is the redox reaction. In this reaction, one
of the components of a system gains electrons (and is thus reduced), while the other loses electrons
(and is thus oxidized). This reaction can occur homogenously- for example, between two different
solute molecules within a solution- or heterogeneously, between molecules and an electrode. In
either case, electrons will flow in the direction that allows them to reach a lower energy state.
Most electrochemistry is centered on the idea of the electrochemical cell. An electrochemical cell
is comprised of at least two electrodes and, typically, a conductive electrolyte solution. Such cells
can behave in either a galvanic or electrolytic mode. A galvanic cell is one in which a spontaneous
chemical reaction produces electrical energy, which can in turn be measured or employed to
perform work. An example of a galvanic cell is a traditional battery. By contrast, an electrolytic
cell is one in which an externally generated potential is applied in order to promote chemical
reactions that would otherwise not be spontaneous. The electrolysis of water to produce hydrogen
and oxygen gas is a classic example of an electrolytic cell.
Electrochemical cells must contain at least two electrodes due to the paired nature of
electrochemical reactions: if electrons are flowing out of one electrode to reduce one chemical
species, an equal number of electrons must be flowing into a second electrode as a second chemical
species is oxidized. The electrode through which electrons are entering the electrolyte is known as
the cathode, whereas the electrode via which electrons are leaving the electrolyte is known as the
anode.
The majority of activity in electrochemical cells occurs at the interface between phases, such as
the interface between the electrode and the electrolyte. This interface marks the division between
two difference conductors, one electronic and one ionic. Due to the markedly different chemical
environments, the energy of an electron in each of the phases will tend to be different. Hence, if
the interface is to be at equilibrium, a potential difference must arise which offsets this chemical
difference in electronic energy and allows the total electrochemical energy (electrical + chemical
energy) to be equal. This potential difference is the driving force for the flow of electrons. In the
5
two-electrode cell, the potential between the two electrodes will be comprised of two parts: the
potential between the first electrode and the electrolyte, and the potential between the electrolyte
and the second electrode. This potential difference is produced by an electrical double layer which
exists at the electrode/liquid interface. At the electrode surface, at excess or dearth of electrons
produces a net charge. This charge is balanced by a diffuse layer of oppositely charged ions in
solution. Specifically adsorbed ions may also contribute charge to the interface. Changes in
potential modify both the charge density of the interface and the quantity/type of ions adsorbed,
leading to a transient capacitive current. Since such currents do not involve the transfer of charge
across the interface, they are known as non-faradaic currents, in contrast with the faradaic current
of typical redox reactions.
The paired nature of electrodes introduces a difficult problem for the study of electrochemistry.
Most of the time, it is only the chemical reaction that is occurring at one of the electrodes that is
the subject of interest (the “working” electrode). However, the potentials and currents involved are
typically dependent on nature of both of the electrodes present. In fact, the potential of an electrode
can only be defined in relation to a second electrode. This has necessitated the development of
reference electrodes. Reference electrodes are electrodes with a carefully controlled chemical
composition. Since the potential difference between the electrolyte and the electrode is a function
of the presence and arrangement of molecules, this serves to keep this potential difference fixed.
Consequently, any potential differences between the reference electrode and the electrode of
interest can be ascribed to changes at the second electrode/electrolyte interface.
When measurements are being performed at equilibrium, such a simple two electrode system is
functional. However, complications arrive in non-equilibrium states when current is flowing. Since
charge conservation necessitates that current flow through the reference electrode as well as the
working, the potential of the reference electrode must shift from its equilibrium value. In general,
reference electrodes are designed to be non-polarizable. This means that they are capable of
supplying large quantities of current in either direction without a significant change in voltage.
However, in order to remove this effect entirely, a three-electrode system can be employed. This
setup introduces a counter electrode, in addition to the working and reference electrodes. The
voltage is controlled between the working electrode and the reference electrode; however, no
current flows through the reference electrode. Instead, the counter electrode passes an equal but
opposite current to the working electrode, completing the circuit. In general, the reaction occurring
6
at the counter electrode is not controlled, and as such may involve any of the chemical species
present.
Three-electrode systems can be used for a wide variety of different experiments. In potentiometry,
the potential between electrodes is measured. In voltammetry, the potential is controlled (either
held constant or varied) and the resultant current is measured. When the potential is fixed, the
technique is known as amperometry. By contrast, potential can be varied in a large number of
ways, both cyclic and acyclic. One such technique is square-wave voltammetry (SWV).16 The
potential in SWV is a superposition of a staircase voltage and a square wave. In each period two
measurements of current are made; one towards the end of the positive voltage step and one
towards the end of the negative voltage step. Delaying the measurement serves to separate the two
types of current, faradaic and non-faradaic. Since non-faradaic processes are generally transient,
they decay much more rapidly than faradaic ones. As such, waiting for non-faradaic currents to
decay to negligible values (using on the order of milliseconds) allows faradaic currents to
dominate. Generally, forward and reverse measurements are subtracted from one another to give
the final trace. This subtraction serves two purposes: it increases the magnitude of the peak due to
the difference in sign of the forward and reverse currents, and it helps to remove the interference
of faradaic currents from contaminants in the solution.
1.3 Electrode Properties and Design
Metal electrodes are widely employed as working electrodes, due to their excellent conductivity
and ease of manufacture. A number of properties of metal electrodes can strongly influence their
behavior, including electrode size/shape,17 material,18 and crystallinity19. The size of an electrode
dictates a number of important properties. In recent years, there has been a push towards producing
smaller electrodes, so-called ultramicroelectrodes.20 Such electrodes deliver enhanced
performance by reducing double-layer charging time, minimizing ohmic loss of potential, and
allowing for a hemispherical diffusion pattern to improve mass delivery. They are ideal for POC
devices, where a small footprint is important and the ability to concentrate analytes of interest into
a small region serves to improve sensitivity. Structure at a finer scale can also serve important
function, and three-dimensional electrodes can interact with a greater amount of solution than flat,
two-dimensional electrodes. Dendritic forms on the nano- or micro-scale serve to enhance surface
area without required enlarged footprint, and may also dictate the geometry of molecular binding.21
7
The material used strongly influences the nature of the interface. The kinetics of redox reactions
are a strong function of electrode material, and the use of certain metals over others can control
the relative rate of reactions in a manner largely independent of thermodynamic favourability. For
example, the reduction of carbon monoxide can proceed along a large number of different routes,
producing a variety of carbon-based structures. Which molecules are formed and in what
proportions is dictated primarily by the nature of the electrode, with judicious selection of electrode
material allowing for enrichment in molecules of interest.22 Similarly, the metal employed
influences which surface modifications can be made, such as the tethering of biological
molecules.23 This phenomenon further extends to the crystal facet of the metal.24 Since it is the
outer surface of the metal which interacts with the solution, it is the properties of this facet rather
than the bulk material which dictate behaviour. Different crystal planes have different surface
energies and different affinities for adsorbents. In the case of polycrystalline electrodes, the
observed effect is the superposition of all of the composite facets.
1.4 Microfluidics
One technology that has been applied extensively in diagnostic devices is microfluidics.
Microfluidic devices are those that serve to manipulate fluids and have a least one critical
dimension on the micrometre scale. Fluid manipulation in this regime is marked by a number of
deviations from macroscopic behaviour. In general, microfluidic devices have a very low Reynolds
number.25 The Reynolds number is an indicator of the ratio of inertial forces to viscous forces. Due
to the relatively low inertia of the flow, the fluid moves as a series of layers or lamina (laminar
flow). Consequently, convective mixing is largely absent in these systems and diffusion is the
primary mechanism by which mixing occurs.26 This allows for the very precise manipulation of
fluid lamina and molecules/particles present therein, facilitating controlled movement of
substances.27 Furthermore, due to the small volumes of fluid being manipulated, forces which scale
with surface area have proportionally greater impact. Surface tension is one such force.28 Surface
tension arises as a consequence of the increase in free energy at the interface between difference
phases. This tension acts within the plane of the interface (tangentially) and acts to minimize the
area of contact. When multiple phases exist in concert, a complicated series of forces arises which
seeks to reduce the interphase between low affinity phases and increase the interface between high
affinity phases. Although this phenomenon can produce unwanted consequences, careful design
can take advantage of these naturally arising forces to drive desired behaviors. These approaches
8
can often allow for more passively actuated devices, whose behaviours are guided by the chemical
nature of the components.29
A large number of techniques exist to produce microfluidic devices. The preferred technique
depends on the requisite spatial resolution as well as the number of devices required and concurrent
cost constraints. One of the most common techniques for microfluidic fabrication is
photolithography. Photolithography depends at its core on the ability of certain chemicals
(photoresists) to undergo a chemical change upon exposure to particular wavelengths of light.
Positive photoresists are those whose chemical bonds are weakened by light exposure, whereas
negative resists are strengthened. By passing light through a patterned photomask, the pattern can
be transferred to the photoresist and revealed by chemical development. Devices with very fine
features may require e-beam lithography, a technique which boasts resolutions as fine as 10 nm.30
Looser precision can be achieved by performing optical photolithography on silicon or glass
substrates.31 If resolutions >5 μm are adequate, soft lithography is a very useful technique.32 In
soft lithography, a mold is made using traditional lithographic techniques, typically using a
negative photoresist such as SU-8. A silicone elastomer, such as PDMS, is then poured onto the
mold and allowed to cure. This elastomer device can then be peeled off the mold and adhered to a
substrate, allowing for repeated use of the mold. PDMS has been widely employed in microfluidic
devices for biological samples due to its biocompatibility.33
The scale of microfluidic devices makes them ideally suited for the manipulation of cells (~10 μm
in diameter). This size compatibility, along with the convenience of laminar flow for deterministic
particle movement, has encouraged the development of a wide range of devices designed for cell
samples. Two such categories are devices for cell separation and devices for single cell isolation.
Cell separation devices serve to remove cells of interest from a sample matrix that contains
extraneous cells or other unwanted contaminants. These depend on exploiting characteristics of
the target population which are not shared by confounding populations, such as physical properties
(size, shape, deformability, etc.) or biochemical properties (surface markers, mRNA, etc.).34 In the
case of discrimination by surface markers, cells are often labelled by antibodies or aptamers which
are tethered to a convenient tag. These tags may be fluorescent, magnetic, dielectric, or actuated
by another means.35 By either active or passive separation the cells bearing this marker can then
be spatially separated from those without. This approach has wide applicability across biology,
allowing for the purification of stem cells and circulating tumor cells, among other rare cell
9
populations.36 Microfluidics also enables the fine positioning of cells, allowing for controlled
treatment and culture of cells isolated on an individual basis.37 Such devices have gained increased
interest given recent paradigm shifts in cell biology, driven by the recognition of subtle
heterogeneities within cell populations.38–40 These differences are especially pertinent in tumour
biology, where the genotypic/phenotypic differences between cells can have drastic effects on their
pathogenic potential.41,42
1.5 Scope of Thesis
This thesis seeks to investigate a number of topics pertaining to the development of devices as
disease diagnostics and as tools for the study of disease biology. It spans microfluidic design and
application, the fine tuning of electrodes towards the detection of biomolecules, and the first steps
towards a POC devices for use in low-resource settings.
In chapter 2, a microfluidic device is described that monitors the properties of circulating tumor
cells in an animal model of cancer. This device captures and sorts CTCs, profiling them on the
basis of the EpCAM surface marker as a surrogate for cell invasiveness. The results show a marked
increase in the number and aggressiveness of CTCs, in good agreement with CT scans of the
primary tumor and secondary metastases. We also illustrate the temporary reprieve following the
resection of the main tumor, alongside the return to an aggressive state as secondary tumors re-
establish the metastatic state.
In chapter 3, we explore the mechanisms driving the morphogenesis of gold microelectrodes
during electroplating and the requisite parameters to fine-tune electrode shape and performance.
The effects of gold ion concentration, applied voltage, and electrolyte viscosity are tested and the
regime for a high surface area structure is established. Further investigations into crystal structure
and theoretical modelling help to elucidate the underlying mechanisms. Finally, the electrodes are
challenged with the detection of DNA strands and performance is tested as a function of
morphology.
In chapter 4, our attention shifts to the design of a device for microfluidic manipulation without
the need for external instrumentation. This device, termed the “Digit Chip”, is based on the
underlying technology of the capillary valve. First we delve into the parameters that control the
reliability and bursting pressure of the valves, determining an optimal geometry. We then devise
10
a simple “push button” control which allows the user to actuate the valves in an intuitive manner
using only their finger. This completed platform is first applied to a simple task of producing a
concentration gradient. Following this, a more advanced design is used in conjunction with a cheap
cell phone-based fluorescence detector to perform antibiotic susceptibility testing of bacteria.
Finally, in chapter 5, we discuss the possible future directions of the research. The potential for
the combination of the various platforms is examined and the position of this work in the field as
a whole is outlined.
11
Sorting of Circulating Tumor Cells during Disease Progression in an Animal Model
In this chapter we explore the application of a microfluidic device previously developed in our lab
to monitoring the progression of cancer in a rabbit model. This microfluidic device extracts CTCs
from rabbit blood samples by labelling them with magnetic nanoparticles and applying a strong
external magnetic field. By varying the drag force present along the length of the device, it is tuned
so that cells with a large amount of the epithelial marker EpCAM are captured in earlier zones,
and cells with a lower amount of EpCAM are captured in later zones. Since reduced epithelial
expression has been correlated with more aggressive cancer, we are able to observe changes in
cancer biology. Over the course of the experiment the tumor growth is monitored and eventually
the tumour is excised, allowing us to monitor changes in CTC number and distribution across the
natural course of disease development as it might occur in a patient.
Reprinted with permission from Muhanna N.*, Mepham A.*, Mohamadi R. M., Chan H., Khan
T., Akens M., Besant J. D., Irish J., Kelley S. O., “Sorting of Circulating Tumor Cells during
Disease Progression in an Animal Model” Nanomedicine: Nanotechnology, Biology and Medicine,
2015, 11(7) 1613-20.
Link to publication online: https://doi.org/10.1016/j.nano.2015.04.017
Disclosure of work within this manuscript: A.M., N.M., and R.M.M. designed the experiments.
A.M performed microfluidic experiments and NM performed traditional hospital testing with the
assistance of T.K, H.C. and M.A. Data analysis and manuscript writing were performed by A.M.,
N.M., and T.K. with assistance from R.M.M., J.D.B., J.I. and S.O.K.
12
2.1 Abstract
Circulating tumor cells (CTCs) can be used as markers for the detection, characterization, and
targeted therapeutic management of cancer. We recently developed a nanoparticle-mediated
approach for capture and sorting of CTCs based on their specific epithelial phenotype. In the
current study, we investigate the phenotypic transition of tumor cells in an animal model and show
the correlation of this transition with tumor progression. VX2 tumor cells were injected into
rabbits, and CTCs were evaluated during tumor progression and correlated with computerized
tomography (CT) measurements of tumor volume. The results showed a dramatic increase of
CTCs during the four weeks of tumor growth. Following resection, CTC levels dropped but then
rebounded, likely due to lymph node metastases. Additionally, CTCs showed a marked loss of the
epithelial cell adhesion molecule (EpCAM) relative to precursor cells. In conclusion, the device
accurately traces disease progression and CTC phenotypic shift in an animal model.
2.2 Introduction
Current cancer staging methods inadequately predict tumor prognosis and response to therapy,
thus underscoring the need for new tumor characterization approaches.43,44 The heterogeneous
nature of tumors and difficulty in identifying metastases during early stages of cancer confound
the ability to predict tumor prognosis and determine appropriate therapy.45,46 A promising
approach, which may serve to address these problems, is the analysis of circulating tumor cells
(CTCs).
CTCs are putative precursors of metastases.47 Specifically, they are cells which are released from
the perimeter of the tumor and intravasate into the blood stream.48 These cells then circulate until
they encounter an appropriate niche, at which time they may extravasate into the surrounding
tissue. Rapid division then establishes a secondary tumor, which can ultimately produce its own
CTCs and continue this process.
Since CTCs are derived from primary tumors and appear to be the cells that establish metastatic
sites, they can provide a wealth of information regarding specific tumor biology and the driving
factors behind invasive disease.49,50 Furthermore, numerous studies have shown that CTCs in
blood samples may be used as a marker to predict survival and prognosis in metastatic cancer
13
patients.51,52 CTC levels appear to be correlated with disease spread,51–53 and elevated CTCs are
associated with poor prognosis and increased probability of metastasis.54
Importantly, CTCs represent the biological state of the evolving disease. A significant aspect of
CTC biology, which has gained considerable traction in recent years, is the loss of epithelial
character accompanying cancer progression during a process known as epithelial-to-mesenchymal
transition (EMT).55 This process involves a cellular reprogramming event that causes CTCs to
assume a more invasive phenotype relative to the epithelial phenotype prominent in the primary
tumor.48 As a consequence, epithelial surface markers such as EpCAM are down-regulated.56
Although the consequences of this transition are not well understood, it has been implicated in
increased metastatic potential, possibly stemming from increased cell mobility due to loss of
adhesion molecules.57 Furthermore, the population of CTCs with low levels of epithelial marker
expression, including EpCAM, has been shown to peak during times of disease progression as
compared to treatment response.55 Thus monitoring changing levels of epithelial markers, the most
prominent of which is EpCAM, should provide valuable insight into cancer progression and
metastasis.
Numerous studies have successfully captured CTCs from diverse tumor origins including lung,58
prostate,58 head and neck,59 melanoma,60 gastric and pancreatic cancers.58,61 However, despite
numerous advances in CTC capture techniques, the analysis of CTCs is still not part of routine
tumor staging in clinical practice. In fact, the CellSearch system is the only existing Food and Drug
Administration (FDA) cleared platform.62 Furthermore, the vast majority of CTC detection
methods simply count the absolute number of CTCs, without distinguishing between varieties of
CTC subpopulations. Consequentially, potentially valuable information may be overlooked.
Recently, we developed a new approach that provides a means to capture and classify of CTCs
with high sensitivity and selectivity using immunomagnetic nanoparticles captured within a novel
microfluidic device.63,64 Our CTC isolation technique depends on antibodies against EpCAM
attached to magnetic nanoparticles allowing for capture using a magnetic field. This device
spatially sorts CTCs on the basis of EpCAM expression, thus providing insight into differential
expression of epithelial markers. EpCAM is expressed by a wide variety of epithelial tumors and
is a generally accepted marker of CTCs,58 and is one of the markers known to show a significant
decrease during EMT, thus establishing it as a surrogate marker of this process.55 Sorting cells
14
according to EpCAM expression therefore provides a means to monitor phenotypic changes in
CTCs. This device was shown to allow the profiling of CTC subpopulations with differing
epithelial character in samples collected from prostate cancer patients. Here, we report the
application of this device to an animal model of cancer. Using a rabbit host and the VX2 tumor
model, we monitored the epithelial character of CTCs during tumor growth and following
resection of the tumor. A significant change in CTC profile is observed from more epithelial to
less epithelial as tumors progress. This is the first study to monitor these dynamics in an animal
model of cancer.
2.3 Results and Discussion
2.3.1 Animal study and microfluidic chip design
The overall timeline of the study is shown in Figure 2-1 A. Following tumor induction, CT scans
and CTC analysis were performed bi-weekly until the 4th week post-tumor induction. At this point,
the primary tumor was resected and CT scan/CTC analysis continued until terminal surgery during
the 6th week.
Figure 2-1 Overview of the study monitoring CTC heterogeneity in a rabbit cancer model.
(A) Timeline showing tumor induction, survival surgery and terminal surgery. (B) Microfluidic
CTC capture and sorting strategy. Cells are labeled with paramagnetic nanoparticles conjugated
to anti-EpCAM antibodies. Labeled cells are introduced into the microfluidic device, where they
are captured in low velocity regions (velocity valleys, shown in blue) adjacent to X structures.
15
(C) A device capable of capturing and sorting CTCs based on expression levels consists of 4
zones with increasing cross-sectional area allowing for capture of cells with high EpCAM
expression in the earlier zones and low EpCAM expression in the later zones. Cells are captured
and specifically identified as CTC by immunostaining. (D) The CTC sorting approach is
hypothesized to produce a profile of CTC subpopulations which can be monitored as a tumor
progresses, enabling evaluation of CTC heterogeneity during tumor growth.
The microfluidic chip employed for CTC analysis demonstrates a number of key design features
that allow for capture as a function of EpCAM expression. Cells are first tagged with magnetic
nanoparticles modified with anti-EpCAM antibodies (Figure 2-1 B). The average hydrodynamic
radius of the magnetic nanoparticles used in this study was 70-80 nm. Further characterization of
the nanoparticles was also performed (Appendix Figure B-1). There are numerous advantages to
the use of nanoparticles in this application. Firstly, due to their small size they are able to diffuse
much more quickly than larger particles. This reduces the incubation time considerably and allows
for more reproducible binding coverage. Secondly, the nanoparticles have limited steric hindrance,
which means that the number of particles binding to a cell is representative of the number of
surface markers. Finally, unbound particles are less likely to coalesce and thus are unlikely to
obstruct fluid flow. Following nanoparticle binding and cell introduction, the chip is sandwiched
between an array of neodymium magnets and flow is induced using a syringe pump. Cruciform
structures within the device serve as pockets of minimum flow (velocity valleys), which allow for
localized capture of CTCs.
Cells within the device experience two main forces (neglecting gravity at the microscopic scale),
namely, a drag force and the magnetophoretic force. The magnetophoretic force is proportional to
the number of nanoparticles on the surface of the cell as well as the intensity of the magnetic field,
which is held approximately constant by the particular arrangement of magnets. The drag force is
proportional to the fluidic flow rate. This is the basis of the geometric design of the chip. Notably,
the number of parallel chambers increases across the length of the device, in a stepwise manner
from 1, to 2, to 4, to 8, which effectively divides the chip into 4 different zones (Figure 2-1 C).
Cells are captured in a region where the magnetophoretic force and the drag force are equal in
magnitude but opposite in direction. For cells with a large number of bound nanoparticles, this
will occur in the first zone. For cells with a small number of bound nanoparticles, this will occur
in the later zones. Since the number of nanoparticles bound is a function of EpCAM expression,
16
cells with high nanoparticle binding (and thus high levels of EpCAM) are captured in the earlier
zones, whereas cells with low nanoparticle binding (and thus low levels of EpCAM) are captured
in the later zones. It was our hypothesis that a change in sorting pattern will be revealed as the
tumor develops and becomes more metastatic (Figure 2-1 D).
2.3.2 Characterization of device function
In order to test the prediction that the microfluidic device facilitates sorting by relegating low
EpCAM cells to later zones, a characterization of the device was performed. SKBR3 and MDA-
MB-231, two well-established cell lines with high and low EpCAM expression respectively, were
chosen. EpCAM expression levels were first characterized using flow cytometry, which confirmed
the large difference in EpCAM expression between the two cell lines (Appendix Figure B-2). Cells
were then subjected to capture in the microfluidic device at a flow rate of 600 μL/h, the same flow
rate used for all capture experiments.
Figure 2-2 Characterization of capture. (A) Capture distribution across the four zones for a
cell line with high EpCAM expression (SKBR3, black) and low EpCAM expression (MDA-MB-
231, gray). The lower EpCAM cells (see inset of cells stained for EpCAM, scale bar = 10 μm)
are captured in the later zones. (B) Capture sensitivity and specificity demonstrated by the
excellent capture of SKBR3 cells and the very low capture of non-specific white blood cells.
*0.04%
High EpCAM Low EpCAM
17
As shown in Figure 2-2 A, SKBR3 cells, which express high levels of EpCAM, were captured
predominantly in zones 1 and 2, whereas MDA-MB-231 cells, which express low levels of
EpCAM, were captured in zones 3 and 4. Furthermore, immunocytochemistry shows the typical
EpCAM expression pattern on cells captured in the early and later chambers of the device, which
further illustrates the capacity of the device to sort based on EpCAM (Figure 2-2 A, inset). Finally,
Figure 2-2 B shows the capture efficiency of the device towards target SKBR3 cells and
nonspecific WBCs. The device demonstrates both high sensitivity and specificity, highlighting its
applicability to in vivo testing.
2.3.3 Measurement of tumor volumes and analysis of metastatic lymph nodes
VX2 cells were injected into the thigh muscles of the rabbit cohort in order to seed the growth of
a tumor. Following this injection, tumor size and morphology were monitored by digital palpation
and CT scan biweekly.
Figure 2-3 VX2 tumor growth in New Zealand rabbits. (A) Representative CT images
demonstrating the progression of primary tumor and (B) lymph node metastasis at 5, 13 and 22
days post induction. The red boxes highlight the region of tumor growth. (C) Tumor volume for
18
all rabbits included in the study. (D) Lymph node metastasis for Rabbit 2. (E) Tumor histology
visualized by staining with (i) H&E and (ii) pan-CK. Lymph nodes stained with (iii) H&E and
(iv) pan-CK, confirming its metastatic status.
All 6 injected rabbits exhibited tumors that increased volume over the course of the study (Figure
2-3 A). On average, tumors typically became detectable within 7-10 days post-injection and over
time, increased exponentially in volume (Figure 2-3 C). Two weeks after injection, the tumors
spread to loco-regional lymph nodes in all 6 rabbits (the volume of the lymph node tumor is shown
for one of the rabbits in Figure 2-3 B, D). At four weeks post injection, survival surgery was
performed to resect the tumor and malignant lymph nodes. At four weeks post-resection, terminal
surgery was performed. Following terminal surgery, residual disease and metastatic lymph nodes
were histologically analyzed. Staining with H&E and pan-CK confirmed the tumors and lymph
nodes to be malignant in all rabbits (Figure 2-3 E).
2.3.4 Correlation of CTC levels with disease progression
In order to determine if the induced tumors would express sufficient EpCAM for on-chip capture,
flow cytometry experiments were conducted to analyze EpCAM levels on VX2 cells. These trials
revealed considerable expression of EpCAM by the VX2 cells, indicating that this surface antigen
was a suitable capture agent (Appendix Figure B-3).
19
Figure 2-4 CTC sorting and analysis. (A) Fluorescence micrographs showing immunostaining-
based identification of white blood cells and captured cancer cells at 50× magnification, scale bar
= 10 μm. (B) Number of CTCs (DAPI+, CK+, CD45−) identified in each zone of the chip over
the 6 week study, averaged over 6 rabbits. Gray bars represent average tumor volume. Notably,
the number of CTCs in zone 3 and 4 shows a more pronounced increase than those in zones 1
and 2. Following surgery, CTC numbers decrease rapidly before rebounding, likely due to
release of CTCs from secondary tumors.
Furthermore, VX2 cells spiked into blood were used to confirm the efficacy of our
immunostaining approach (Figure 2-4 A). Specifically, anti-CK stain was used to identify cancer
cells, as cytokeratins are expressed by a wide variety of epithelial cells but are typically not
expressed by blood cells. CD45, a pan-leukocyte marker, was used as a counterstain to identify
20
contaminating white blood cells. This collection of stains enables the clear distinction of cancer
cells from white blood cells.
Over the course of the study, CTCs were captured and analyzed multiple times per week. Figure
2-4 B summarizes the CTC capture data averaged over the entire rabbit cohort (n = 6). The
utilization of nanoparticles allowed for an effective capture of CTCs even using a relatively small
volume of blood (1 mL). The CTCs in each zone are shown separately to elucidate the shifting
values of individual CTC subpopulations. Importantly, prior to tumor injection, all rabbits showed
very few or no CTCs, indicating a low rate of false positives. Levels of CTCs became significant
one week after tumor induction and continued to gradually increase for four weeks (Figure 2-4 B).
The increase in CTCs was correlated with growth of the primary tumor, which reached a maximum
at four weeks, just prior to survival surgery. Furthermore, CT imaging determined the development
of a secondary metastatic disease in most of the rabbits during the same timeline. These findings
suggested a positive correlation between the presence of CTCs and the appearance of metastatic
tumors, with a rise in CTCs preceding a detectable metastasis. At four weeks after tumor induction,
survival surgery for tumor resection was performed. Following this surgery, a drastic reduction in
CTC levels was observed, which suggests that the primary tumor was the major source of CTCs.
However, two weeks after surgery, a second increase in CTC levels was observed, indicating a
rebound in metastatic activity. This may be explained by the presence of metastatic lymph nodes
in the pelvis as revealed by CT scanning. This implies that CTCs might have originated not only
from the primary tumor but also from metastatic lymph nodes, and indicates a correlation between
CTCs and lymph node metastasis. Furthermore, this observation implies that surgical resection of
the primary tumor was insufficient in limiting disease spread. Taken together, these results suggest
that an increase in the number of CTCs is an indicative biomarker for metastatic tumors derived
from VX2 cancer cells.
2.3.5 Analysis of CTC sorting patterns during tumor growth
The sorting of CTCs on the basis of EpCAM revealed a notable capture trend over the progression
of the disease (Figure 2-4 B). Within our device, cells captured in Zone 1 represent those with high
levels of EpCAM, while those in Zone 2, 3, and 4 have decreasing levels of this epithelial marker.
In weeks 2, 3, and 4, the number of CTCs in zones 3 and 4 increased more rapidly than those in
zones 1 and 2. In fact, these cells account for most of the increase in the total number of captured
21
CTCs, which indicates a loss of EpCAM in the CTCs with tumor progression. This finding is
interesting, but reasonable considering that the reduction in epithelial cell adhesion molecules is a
well-established fact that is known to occur with tumor maturation.48,56 Notably, CTCs in zones 3
and 4 also showed the greatest reduction following surgery, which may reflect that these cells
originated from the primary tumor. However, CTCs in zones 3 and 4 also increase most drastically
in week 6, presumably as the lymph node metastases becomes increasingly dominant. The cells
with the lowest amount of EpCAM appear to be the most informative indicator to the state of the
disease. This is of particular interest as these are the most difficult to capture and as such are the
subpopulations most likely to be missed in conventional capture methods.
It is useful to benchmark the evolution of the surface expression profile reflected in the sorting
pattern we monitor against the sorting profile of the VX2 cells that are the precursor of the tumors
studied. In order to determine if the capture profile of the seeded VX2 cells differed from that of
the captured CTCs, experiments were performed where VX2 cells were spiked into healthy rabbit
blood. The majority of cells were captured, indicating that EpCAM expression was adequate for a
sufficient number of nanoparticles to be bound. Interestingly, a large number of VX2 cells were
captured in zone 1, with a steadily decreasing rate of capture in zones 2, 3 and 4 (Figure 2-5). This
is indicative of a high level of EpCAM expression, which is a reasonable result given the epithelial
origin of VX2 tumors.
22
Figure 2-5 Comparison of sorting profile for (i) VX2 cells and (ii) CTCs captured 4 weeks
after tumor induction, showing a prominent shift towards later zones as tumor growth progresses
with lower level of EpCAM expression.
Perhaps the most important revelation is the marked difference between the capture trends
exhibited by VX2 cells compared to that of CTCs collected after 4 weeks of tumor growth (Figure
2-5). Whereas the spiked VX2 cells were captured primarily in the earlier zones, the CTCs were
isolated primarily in the later zones. This indicates a dramatic reduction in the abundance of
EpCAM expression on the surface of CTCs. We speculate that the development of the tumor within
the host triggers the abrupt shift in the biology of the VX2 cells, possibly initiating a pronounced
loss of epithelial character.
2.3.6 Discussion
It is well recognized that CTCs are precursors to metastasis and can serve as an integral component
of tumor staging criteria. Furthermore, previous studies involving breast cancer patients have
shown that CTC data can be the best indicator of disease state and can provide improvement to the
staging process.65 Because blood collection is simple and minimally invasive, CTCs can
potentially be used as a real-time marker to monitor disease progression. They also possess the
23
potential to guide therapeutic management, indicating therapy effectiveness or necessity, even in
the absence of detectable metastases.66
Despite this utility, the clinical application of CTCs remains minimal. One reason for this is the
inability of commercially-available systems to detect CTCs with low levels of EpCAM.56 Other
limitations include negligible EpCAM expression levels in subpopulations of CTCs. This results
in missing significant information regarding the tumor cell differentiation level. In this study, we
demonstrated the ability of a microfluidic device to capture and sort CTCs with varying levels of
EpCAM, and infer a correlation with tumor progression.
Our investigation shows that elevated numbers of CTCs correlate with disease progression and
lymph node metastasis. Specifically, the number of CTCs peaked at the same time that the tumor
reached its maximum size. This is expected given that previous studies have demonstrated that
high CTC numbers correlate with aggressive disease, increased metastasis, and decreased time to
relapse.51,52,54 The dramatic reduction in CTCs post-surgery provided real-time information on
treatment effectiveness, which could potentially be useful in the assessment of surgery or
chemotherapeutic efficacy. Since chemotherapy regimens have highly variable efficacies for
different patients, rapid determination of the effectiveness of treatment using CTC markers could
allow for rapid treatment modulation.
The recurrence of CTCs after the tumor resection suggests that the CTCs might also originated
from metastases in lymph nodes, which further supports the theory that secondary tumors can
become a source for CTCs, and helps to explain the recurrence of metastatic disease in patients
that have undergone surgical resection. This is in agreement with the ability of many cancers to
recur despite complete surgical resection of the primary tumor.66,67
There is ongoing research regarding whether tumor cells undergo a reduction in epithelial character
during dissemination, resulting in a more mesenchymal or even more stem cell-like phenotype.
Our results corroborate this theory; over time, in addition to the increase in their numbers, we
inferred that CTCs likely exhibited a reduction in EpCAM. In this respect, the use of nanoparticles
is important due to the minimal influence of steric effects on bead coating efficiency. Since
EpCAM levels reflect the epithelial nature of a cell, they can serve as a surrogate marker for this
process, indicating an increase in invasive and migratory properties. This is consequential because
the loss of epithelial character is believed to be the gatekeeper under which tumor cells intravasate
24
into the blood.68 Furthermore, it has been suggested that CTCs with the highest plasticity tend to
be the most aggressive.69
An interesting insight provided by the present study was the difference in EpCAM expression
levels between the VX2 cell line used to produce the tumors and the CTCs seeded from these
tumors. This is an aspect that would typically be neglected by detection methods that capture all
CTCs in the same fashion, regardless of their biological state and their EpCAM expression level.
This reveals a phenotypic change occurring immediately upon insertion of cells into the biological
environment of the rabbit host, suggesting a phenotypic transition that can occur in cells on the
periphery of the tumor. Interestingly, this is similar to the result shown by Steinert et al, which
demonstrated a rapid reduction in EpCAM expression upon colorectal cancer-derived CTC
compared to the corresponding tumor tissue.70
As a future step a comprehensive study with large patient cohorts using different clinical subtypes
of cancer, with standardized detection and characterization of CTCs would be desirable to evaluate
the performance of the device for human blood samples. Further studies, incorporating a suite of
biomarkers indicative of cell phenotype changes would allow one to determine the prognostic
significance of different categories of CTCs in cancer patients and aid in the clinical management
of these patients. Finally, since the device allows for the sorting of CTCs, which remain viable,
this opens the door for genomic, proteomic and transcriptomic investigations of these cells in
various phenotypic states. CTC genomics is still in its infancy, which is primarily due to the lack
of technologies that are capable of capturing sufficient numbers of CTCs to analyze somatic
mutations.49,50 The next stage in the investigation of CTCs is their characterization using these
‘omics’ techniques, which can be used to identify numerous characteristics of tumors, allowing
for targeted therapeutic approaches.55,71,72 These techniques could find application with the current
device.
2.4 Conclusion
CTCs are an important class of circulating cancer markers that may enable noninvasive
determination of prognosis. This study demonstrated the successful capture of CTCs in an animal
model and the sorting of these CTCs on the basis of their EpCAM expression. The numbers of
isolated CTCs were positively correlated with tumor growth while CTC sorting profiles
concurrently indicated a shift towards reduced expression of EpCAM. This indicates that both
25
number and phenotypic distribution of CTCs may possess clinical relevance. Future studies
including a greater diversity of cancer subtypes, cell surface markers, and genomic approaches are
to be investigated.
2.5 Methods
2.5.1 Animal model
Experiments were performed using 6 New Zealand white rabbits weighing 2.5-3.0 kg. All animal
studies were performed in accordance with the University Health Network/University of Toronto
guidelines for the humane use of animals. Care, handling and maintenance of all animals used in
this study were conducted in a humane manner, as per the animal care experimental protocol
approved by the institutional Animal Care and Use Committee of University Health Network,
University of Toronto. Male rabbits (Charles River, Wilmington, Massachusetts) were injected
with 300 μL of a high-density (approximately 5 × 106/mL) cell suspension of VX2 squamous cell
carcinoma of the rabbit into the thigh muscles (quadriceps). Tumor development and lymph node
metastases were monitored using computed tomography (CT) images and clinical evaluation bi-
weekly, and were resected 4 weeks after tumor induction. Tumors and enlarged lymph nodes were
sent for pathology and immunostaining analysis. Blood samples for CTCs (2-3 mL), hematology
and comprehensive biochemistry analysis were collected biweekly, pre and post-operation.
2.5.2 Tumor cell line propagation
The VX2 tumor cell line is maintained in small tumor pieces that are frozen at −80 °C. Tumor cells
were propagated by injecting 500 μL of VX2 tumor into the quadriceps of propagating rabbits
(different from rabbits used for the CTC study) and were harvested after approximately 3 weeks.
The harvested tumor was placed in Hanks balanced salt solution (HBSS) in a sterile 100 mL
container. Prior to tumor induction in rabbits, the tumor pieces were thawed and cut into small
pieces using a sterile scalpel and subsequently placed on to a 70-μm-cell strainer sitting on a 50
mL tube (BD Falcon brand). A syringe plunger was used to mince the cells and ~500 μL HBSS
was used to suspend the cells in the strainer (repeated several times).
2.5.3 CT imaging and image analysis
CT imaging and image analysis CT imaging (Locus Ultra, GE Healthcare, Milwaukee, Wisconsin,
USA) was performed biweekly pre and post-surgical tumor resection (80 kVp, 50 mA). All CT-
26
based image analysis was performed using Microview (GE Healthcare, Milwaukee, Wisconsin,
USA) and custom in-house program written using MATLAB (MathWorks®, Natick,
Massachusetts, USA). The tumor volumes were contoured using a semi-automated threshold based
method. The mean and standard deviation of the voxel signal distribution within each VOI were
calculated.
2.5.4 Histopathological evaluation
Tumor and lymph node tissue samples were fixed in formalin after resection, embedded in paraffin
blocks, cut and stained with hematoxylin and eosin (H&E) and pan-cytokeratin (AE1/AE3), the
intermediate filaments of epithelial cells. All histopathology images were analyzed using
ImageScope (Leica Biosystems, Wetzlar, Germany) after scanning.
2.5.5 Capture of cell lines
100 SKBR3 or MDA-MB-231 cells in 100 μL were incubated with 10 μL of MACS anti-EpCAM
nanoparticles (130-061-101) for 10 min. Cells were then introduced into the chip at a rate of 600
μL/h for 10 min. A 200 μL PBS 1× rinse was then added followed by 100 μL of PBS-4%
formaldehyde and 100 μL PBS–0.2% Triton. Cells were stained with anti-pan-CK Alexa Fluor
488 from eBioscience (53-9003-82) and 10% DAPI.
2.5.6 Immunocytochemistry of cell lines
Cells were incubated in 1% BSA PBS for 10 min to prevent non-specific binding. Cells were then
stained with Alexa Fluor 647 anti-human CD326 (EpCAM) antibody (Biolegend, 324212) for 1
h. Cells were rinsed twice using 1× PBS before DAPI staining in 0.1%Triton-PBS.
2.5.7 Isolation and fluorescent staining of CTCs
In order to detect and characterize CTCs, we used an immunomagnetic system targeting
EpCAM.21 Captured cells then underwent immunocytological staining to confirm their identity
as tumor cells. CTCs were identified as EpCAM isolated cells that stained positive for DAPI and
pan-cytokeratin and negative for CD45. The number of detected CTCs obtained per mL of blood
was recorded for correlation with clinical parameters such as tumor size, lymph node metastasis
and duration from tumor inoculation. CTCs in each of the 4 zones were counted separately.
27
1 mL of rabbit blood was incubated with 10 μL of MACS anti-EpCAM nanoparticles (130-061-
101) for 10 min. This blood was then introduced into the chip, which had been pretreated with 1%
pluronic acid, at a rate of 600 μL/h for 100 min. 200 μL of PBS-EDTA was introduced through to
rinse out red blood cells. Subsequently, 100 μL of PBS-4% formaldehyde was added to fix the
cells. Next, 100 μL of 0.2% Triton in PBS was added to permeabilize cells. 100 μL of a CTC-
specific antibody (1 μL of anti-pan-CK Alexa Fluor 488 from eBioscience (53-9003-82) and 1 μL
of anti-CD45 APC from AbdSerotec (MCA1114F) in 98 μL of PBS-1% BSA) was added for 1 h
at 100 μL/h for immunostaining. Afterwards, 100 μL of PBS with 10%DAPI solution was added
at 600 μL/h for 10 min to stain nuclei. Finally, 200 μL of PBS was added to remove excess non-
specifically bound antibody.
2.5.8 Capture efficiency of white blood cells
Capture and staining were performed as described above, with white blood cells considered those
cells which were DAPI+/ CD45+/CK−. Capture efficiency was calculated assuming 106 WBC/ml
blood.
2.5.9 Image scanning and analysis
After immunostaining, chips were scanned using a 10× objective and a Nikon Eclipse Ti
microscope equipped with an automated stage controller and a cooled CCD (Hamamatsu,
Hamamatsu, Japan). Images were acquired with NIS Element software (Nikon, Tokyo, Japan).
Red, green and blue fluorescence images were recorded. The captured images were then analyzed
in NIS Elements and target cells were enumerated.
2.5.10 Flow cytometry
VX2 cells were prepared in suspension as for tumor propagation. SKBR3 and MDA-MB-231 cells
were collected from culture. Cells were incubated in 1% BSA blocking buffer for 30 min to prevent
non-specific binding. Anti-pan-CK Alexa Fluor 488 from eBioscience (53-9003-82) was added
and incubated for 1 h prior to cytometry. Cytometry was performed using the BD FACSCanto
(Becton Dickinson, Franklin Lakes, New Jersey, USA) flow cytometer with 488 nm laser
excitation and 530/30 nm detection.
2.5.11 Microchip fabrication
28
Microchips were fabricated using poly-dimethylsiloxane (PDMS) after production of an SU-8
master by lithography on silicon wafers, with a height of 80 nm (University Wafer, Massachusetts,
USA). A PDMS (Dow Chemical, Michigan, USA) copy of the master was produced and peeled
off the wafer. Holes were pierced at the inlet and outlet. PDMS and a glass slide were treated with
1 min of plasma and attached to form a permanent bond. Silicon tubing was inserted into the holes
at the outlet and inlet. Prior to use, chips were incubated with Pluronic F68 acid to reduce non-
specific binding. During cell capture, arrays of NdFeB N52 magnets (KJ Magnetics, Pennsylvania,
USA) were placed above and below the chip.
29
Mechanistic Control of the Growth of Three-Dimensional Gold Sensors
In the previous chapter, we explored the application of a microscale technology towards a
particular disease. In this chapter, we are continuing to examine the role of microscale technology
for disease probing, but are investigating a more fundamental topic; the development of novel
electrode architectures to improve biomarker sensing. Gold electrodeposition is a well-established
technique for electrode synthesis, due to the inert nature of gold electrodes and the proclivity of
gold to encourage the formation of monolayers of thiolated molecules. Here we test the effects of
a number of adjustable parameters (gold concentration, voltage, and electrolyte viscosity) and
observe the fundamentally different architectures that result. We further investigate the
mechanisms of the underlying growth patterns, and ultimately challenge the final electrodes with
biomolecular detection.
Reprinted with permission from Mahshid S.*, Mepham A.*, Mahshid S. S., Burgess I. B., Safaei
T.S., Sargent E.H., Kelley S.O., “Mechanistic Control of the Growth of Three-Dimensional Gold
Sensors” Journal of Physical Chemistry C, 2016, 120(37) 21123–21132. Copyright 2016
American Chemical Society.
Link to publication online: https://doi.org/10.1021/acs.jpcc.6b05158
Disclosure of work within this manuscript: A.M., S.M., S.S.M. and I.B.B. designed the
experiments. A.M. and S.M. performed experiments. Data analysis and manuscript writing were
performed by A.M., S.M., S.S.M. and I.B.B. with assistance from T.S.S., E.H.S., and S.O.K.
30
3.1 Abstract
Three-dimensional (3D) electrodes with large surface areas are highly effective biomolecular
sensors. These structures can be generated via the electrodeposition of gold inside microscale
apertures patterned on the surface of a microelectronic chip. Such electrodes enable the
ultrasensitive analysis of nucleic acids, proteins, and small molecules. Since the performance of
these electrodes is directly related to their surface area, the ability to control their microscale
morphology is critical. Here, we explore an electrochemical model based on the theory of
nucleation and growth to better understand how to control the morphology of these electrodes. The
insights gained from this model enabled us to create preferential conditions for the formation of
different morphological features. We demonstrate for the first time that electrodeposition of 3D
nanostructured microelectrodes inside a microscale aperture is governed by two stages of
nucleation and growth. The first stage involves the creation of primary nuclei at the bottom of the
aperture. The second stage features the generation of new nuclei upon exposure to the bulk
solution. Depending on the overpotential, the deposition is then continued by either rapid growth
of the original nuclei or fast growth of new nuclei. Faster electrodeposition at high overpotentials
promotes directional growth, generating spiky structures. More isotropic growth is observed with
low overpotentials, generating rounder features. Ultimately we determine the efficiency of DNA
hybridization on a variety of structures and identify the optimal morphologies for rapid
DNA−DNA duplex formation.
3.2 Introduction
Engineering three-dimensional (3D) structures on the micro and nanometer scales is of importance
for the production of high-performance materials for electronics and biological applications.73–77
Bottom-up fabrication via self-assembly lends itself to the creation of complex 3D architectures in
a variety of material systems.78–81 This approach relies on minimal direct control, relying instead
on pattern formation on the basis of thermodynamic equilibria or instabilities.82 In such a system,
the assembly process and the resultant material morphology can be manipulated via any parameters
that affect kinetic and thermodynamic properties.83
One of the materials that is used extensively for generation of 3D micro- and nanostructures is
gold.84 Gold structures exhibit a variety of morphologies depending on the method of synthesis.85
Furthermore, 3D gold structures are outstanding candidates for electrochemical biosensing
31
applications86 and catalyze a number of important chemical reactions.87,88 The specificity of
catalysis and the degree to which the reaction rates are amplified can be linked directly to the
morphological properties of 3D gold structures.89
Moreover, the crystal structure of gold is also important in determining its performance as a sensor
or catalyst. Different crystal facets of gold promote distinct interactions with molecular substrates
and targets.90 Thus, the capacity to modulate the chemical reactivity of different facets during the
assembly process is advantageous for the design of optimized structures. As such, methods that
control the number and orientation of grain boundaries are of significant interest.
A variety of methods have been used to synthesize gold structures, including sputtering, e-beam
lithography, and chemical and electrochemical approaches.91–94 Although each of these methods
has its advantages, none offers access to the suite of synthetically accessible parameters that are
available with electrochemical synthesis.94 We have previously developed electrodeposited, 3D
gold microelectrodes and found that they represent a promising platform for electrochemical
biomolecular detection.95–100 By creating sensors with large surface areas that protrude into
solution, collisional frequencies for biomolecular targets are enhanced.96 Also, the electrochemical
currents generated by these structures are amplified due to the efficient transport of redox-active
reporters via radial diffusion.97 We have further demonstrated that the introduction of
nanostructured roughness on the surface of these microelectrodes enhances their biosensing
performance.98 While this system has been applied successfully to the detection of cancer
biomarkers,99 infectious pathogen identification,95 and organ transplant assessment,100 we know
little about how the 3D features of these sensors influence their performance.
Here, we explore the growth mechanism of 3D gold microsensors and identify factors that enable
precise control of morphology and crystallinity. We demonstrate that by changing the gold ion
concentration, we can preferentially grow different structures and control directional growth of
spiky structures. By increasing the viscosity of the electrolyte solution, we can suppress the growth
and generate finer spiky structures. Additionally, by increasing the deposition overpotential, we
can also generate finer spiky structures through promoting multiple nucleation and rapid growth
of the nuclei clusters. The collection of these parameters enables a high degree of control over the
microscale morphology and crystallinity of gold assemblies. We further explore the effects of
voltage in more depth to visualize processes related to three dimensional nucleation and diffusion-
32
controlled growth.101,102 We ultimately investigate the efficiency of hybridization of DNA
duplexes on the surfaces of sensors with differing morphologies and identify the optimal
morphology for this type of application. The most effective sensors are generated with high
growth-to-nucleation ratios during electrodeposition.
3.3 Results and Discussion
3.3.1 Outline of experiments
Our approach to generating 3D microelectrodes using gold electrodeposition benefits from micro-
patterned apertures fabricated via photolithography on the surface of a glass chip. A schematic
outlining the system used in this study is shown in Figure 3-1; the application of this
microelectrode system for the analysis of specific DNA sequences is also depicted.
Figure 3-1 Schematic of gold microelectrode experiemnts (A) Gold 3D microelectrodes are
grown using electrodeposition on a gold substrate with 10 μm apertures as a template. In this
study, we explore parameters that could generate different sensor morphologies to determine the
optimal properties for biomolecular detectors. (B) The parameters of gold ion concentration,
Gold substratePhotoresist
Glass
target DNA
probe DNA
A B
C
33
solution viscosity, and applied voltage are varied to explore which regions encourage the growth
of particular morphologies. (C) 3D microsensors are tested for DNA hybridization. Blue strand:
capture probe that promotes sequence-specific binding of a target sequence. Red: Target DNA
strand complementary to probe. Target binding can then be read out using redox-active reporters.
The electrodeposition is carried out on a glass wafer coated with a thin layer of gold, which is
topped with a passivating layer of SU-8 photoresist (Figure 3-1 A). The small (10 μm) apertures
in this passivating layer expose the gold substrate to the electrolyte solution, allowing for localized
deposition of gold structures. This template-based approach allows for a hemispherical diffusion
pattern on top of the aperture that facilitates rapid growth and the faster emergence of structural
features compared to the growth achieved on larger planar surfaces.
The three major parameters explored in this study are the concentration of gold ions (AuCl4-), the
viscosity of the electrolyte solution, and the applied voltage. The concentration of gold ions plays
a critical role in the kinetics and thermodynamics of the transition from a dissolved state to a solid
state. Higher concentrations increase the capacity of the system to transfer sufficient quantities of
gold to the growing deposit and enhance electrodeposition kinetics. Viscosity affects the kinetics
of the deposition reaction by impeding replenishment of fresh gold ions. Applied voltage controls
the relative energies of the solid and dissolved states of gold. As the potential of the cathode is
made increasingly negative, the reduction of gold ions to metallic gold becomes more favorable.
By varying these three parameters, a phase space is generated featuring regions with different
growth regimes and different morphologies (Figure 3-1 B).
For applications where dilute diffusible molecular targets like DNA are being detected (Figure 3-1
C), it is critical to precisely control the morphology and surface area, as both factors influence
collisional frequencies as well as binding affinities. To control the formation of 3D structures
generated via gold electrodeposition, it is important to understand the influence of nucleation and
growth processes. We therefore not only investigate the influence of solution parameters on 3D
sensor electrodeposition but also investigate the mechanism of nucleation/growth and seek to link
it to the geometrical properties of the resultant structures.
3.3.2 Effects of solution conditions of 3D nanostructured microelectrode growth
34
The effects of varying the concentration and viscosity of the electrolyte solution are shown in
Figure 3-2 A and Figure 3-2 B. For each condition, a pair of SEM images are shown at a low (5
000×) and high (50 000×) magnification. This allows for the elucidation of changes in the structure
at two different scales; both the microscale, so that the overall shape of the structure can be
evaluated, and the nanoscale, which allows a detailed analysis of the morphological changes.
35
Figure 3-2 Effects of concentration, viscosity, and voltage on electrodeposition of gold
microsensors are studied using scanning electron microscopy (SEM). (A) SEM images for
structures generated with varied concentrations of gold ions. Varying the concentration of gold
reveals a number of different morphologies. A low (1 mM) concentration produces nanoscale
spike-like structures, whereas a 5 mM concentration produces leaf-like structures (inset shows
structure at 1000× to visualize entire structure) and higher concentrations promote the growth of
needles. (B) Increasing viscosity creates rounded structures without visible facets or needles. (C)
A
B
C
(1)(2) (3)
(4)(5) (6)
(7)(8)
(9)
(10)
(11)
(12)
36
Increasing deposition potential changes the structures from isotropic rough hemispheres to
highly anisotropic structures with a large population of needles. White scale bars are 20 μm,
yellow are 2 μm, and red are 40 μm. The potentials used for electrodeposition are displayed
above each SEM image and numbered for cross-referencing to Figure 3-4.
The first parameter explored was the concentration of the gold solution (Figure 3-2 A). A wide
range of concentrations were tested, ranging from 1 mM to 500 mM, with a roughly logarithmic
distribution (each sample approximately 3× more concentrated than the previous). The different
concentrations produced at least three qualitatively different growth modes. At the lowest
concentration tested (1 mM), the overall morphology features a rounded structure containing a
number of distinct lobes on the order of 20−30 μm in size. A study of the surfaces of these
structures reveals that the gold assumes a propensity toward nanoscale spike-like structures and is
highly porous.
At a 5-fold higher concentration (5 mM) of gold ions, the structure assumes a morphology that is
remarkably different from that at 1 mM. Instead of featuring rounded lobes, the structure is
dominated by a small number of large leaf-like structures. Upon closer inspection at a higher
magnification, these structures demonstrate remarkable geometric patterns. Each leaf has a central
ridge that runs lengthwise along the middle of the leaf from stem to tip. From this ridge, a number
of smaller veins extend. These veins in turn have smaller sub-veins. This fractal structure indicates
that each leaf is either a single crystal domain or a number of domains with fixed relative
orientations.
As the gold ion concentration is further increased to 16 mM, a new structural regime is
encountered. Leaf-like structures are reduced in number, and there appears a new structural
feature: needles. At this concentration, the needles are highly ridged, exhibiting a series of peaks
and valleys along their length. These needles do not have a consistent cross-sectional shape and
appear to be incompletely formed. As the concentration is further increased from 16 mM to 50 and
500 mM, these needles persist. However, at higher concentrations, the ridges are significantly
reduced and the needles develop a consistent pentagonal cross section. These needles are of
particular interest due to their unique shape, one that has been recognized before in gold
structures.103 Interestingly, higher gold ion concentrations do not necessarily create a greater
37
abundance of needles but rather promote thicker and smoother shapes. As such, the region for
creating the greatest number of needles lies in the concentration range between 50 and 500 mM.
In order to change the viscosity of the solution, varying amounts of glycerol were introduced into
the solution (Figure 3-2 B). Glycerol was chosen due to its high viscosity and its miscibility in
water. As the glycerol concentration is increased, the structures become more compact and dense
and the local features are distorted. The sharp edges and facets are replaced with much more
rounded features. The most profound change occurs at 75% glycerol. Here, the needles are not
present and are replaced with a largely amorphous arrangement that displays weaker directional
preference than the highly crystalline material. This experiment indicates that there is another
important ingredient for the promotion of needle morphology; namely, low viscosity. The
associated high mobility of ions seems integral to allow needles to dominate over largely
amorphous, many-grained structures.
The influence of applied potential was initially studied using linear voltammetry (Figure 3-2 C).
Variation of applied potential also creates a demonstrable effect on the final morphology (see SEM
insets of Figure 3-2 C). At very low overpotentials (+500mV) the growth is hemispherical overall,
with no noticeable anisotropy. At the microscale, the structures are disordered, with a large number
of grains of different sizes and orientations. As the magnitude of the potential is increased (i.e., the
cathode is made more negative), the structures gain a pronounced anisotropy, with regions of the
structure extending farther into the surrounding solution. At a potential of about +150 mV, the
needle morphology reemerges, although it is clearly not pentagonal at this voltage and is decorated
with a large number of offshoots. Approaching 0 mV, the needles display a full pentagonal
geometry. Therefore, to promote the formation of the needles and generate finer spiky structure, a
high overpotential (i.e., −250 mV) is needed in addition to a high gold concentration (SEM insets
of Figure 3-2 C). It is only in the combinatorial application of these parameters that this particularly
interesting morphology becomes the most stable manifestation of gold. Increasing overpotential
to more negative values (above −500 mV) causes the spiky structures to be replaced by a flaky
morphology with fine structuring.
Images of the exterior of the structures provide insight into the growth mechanism but do not
elucidate the extent of crystallinity of the structures. To investigate the internal structures of our
3D sensors more directly, FIB sectioning was performed and SEM images were collected of the
38
resulting cross sections. Using this approach, different crystal grains are revealed within the
structures.
Figure 3-3 Analysis of the interior structures of electrodeposited gold using FIB. (A, B) The
structures generated under high viscosity conditions shows grains with a wide variety of sizes
and shapes. (C, D) The leaf structure has a clear crystal twinning about the central plane,
indicated by two different shades of gray. (E, F) The needle-like structures have pentagonal cross
sections with a 5-fold twinning about the center axis. All scale bars are 2 μm.
The structures grown in 75% glycerol are shown in Figure 3-3 A and again in cross section in
Figure 3-3 B. In these amorphous structures there is no conserved orientation or direction to the
grains. Furthermore, the grains are of vastly different sizes and shapes, with no discernible pattern.
The leaf-like structures have a fundamentally different inner morphology than those generated
under high viscosity conditions. Figure 3-3 B shows an example of this type of structure displaying
a series of veins along its surface and a bilaterally symmetrical shape. A FIB cross section is shown
in Figure 3-3 E. Notably, this structure is again composed of multiple grains; however, it has a 2-
fold symmetry rather than the disordered arrangement seen previously. This symmetry explains
why the top and bottom of the leaf are similar in geometry. This grain boundary is likely a
consequence of crystal twinning about a common central plane.104
Figure 3-3 C shows a high-resolution SEM image of a well-formed needle. This structure has well-
defined facets that are modified by a series of small ridges perpendicular to the length of the needle.
A B C
D E F
39
FIB sectioning of this structure reveals a clear pentagonal symmetry (Figure 3-3 F). Moreover,
this symmetry is seen to stem from a 5-fold arrangement of unique crystal grains about a common
center point, with distinct grain boundaries existing between them.
3.3.3 Nucleation and growth mechanism of 3D gold structures
In most electrodeposition processes, the first phase of structural growth is characterized by rapid
nucleation.101 These nuclei then enter a growth phase, with additional nuclei forming at a reduced
rate. As such, crystalline structure is largely established at the time of nucleation. It is apparent
from the images presented in Figure 3-3 that at relatively low concentrations (around 5 mM), nuclei
are formed with a 2-fold twinning structure. By contrast, at higher concentrations, there is a shift
such that the 5-fold twinned structure is preferable. The appearance of some leaf-like structures
even at higher concentrations can also be attributed to local depletion of gold ions.
In the case of increased viscosity, the growth rate is severely attenuated relative to the rate of
nucleation. Additionally, the geometric entities that represent the lowest energy state under other
conditions are no longer preferred, and there is instead a largely random process of deposition.
This leads to a loss of directionality that manifests both as an increase in isotropy and a less faceted
surface at the local level. Evidently, in order to promote needle growth, it is critical to create
conditions where 5-fold twinned nuclei are the most favorable.
Although FIB imaging provides qualitative insights into the mechanisms of growth, a more
comprehensive understanding of this phenomenon was desired. Due to the wide array of different
structures that can be produced (bulky, spiky, flaky) within a small window of potentials, we
elected to explore the effects of electrodeposition voltage in further depth.
3.3.4 Study of current-time transients
The early stage of electrodeposition in our system is associated with a three-dimensional
nucleation process, where the number of nuclei and the rate of nucleus formation are strongly
governed by the deposition overpotential. We investigated the I−t current transients during
deposition to establish a relationship between deposition overpotential and the nucleation and
growth mechanisms in 3D microelectrodes (Figure 3-4).
40
Figure 3-4 I-t curves during electrodeposition (A) I−t curves for gold electrodeposition within
the potential range of +500 to −750 mV. Each trace is labeled with a number that corresponds to
a voltage specified in Figure 3-2. These traces deviate from linearity under certain conditions as
shown by the region “e” specified with a dotted line. (B) Electrodeposition data for low-to-
moderate overpotentials (500 to 50 mV). (C) Electrodeposition data for higher overpotentials (0
to −750 mV).
During deposition inside the microscale aperture, nucleation and growth are strongly affected by
the diffusion of ions around the edges. However, as the deposition proceeds outside the aperture,
the geometry of the 3D electrode increases and consequently provides a larger surface for electron
transfer over time.
From the I−t curves (Figure 3-4), we can distinguish several deposition phenomena that occur
within our microscale apertures. The different phases of growth can be visualized by examining
1st nucleation phase
2nd nucleation phase
A
B C
41
the I−t curves generated with low overpotentials, +500 and +400 mV (Figure 3-4 B, curves 1 and
2). The first rise is related to the double-layer formation (region a). A sharp drop in current then
occurs due to the depletion of ions following the nucleation process on the planar Au at the bottom
of the aperture (region b). Immediately thereafter, steady-state growth fills the aperture (region c).
The isotropic growth then proceeds with most of the electrodeposited gold growing outside the
aperture and the 3D electrodes expand (region d) to form a bulky structure (SEM images 1 and 2
in Figure 3-2 C). When the overpotential is increased to +300, +150, and +50 mV, a sharp rise in
the I−t curve is observed at the beginning of region d (Figure 3-4 B, curves 3−5). Here the
deposition front undergoes a second nucleation stage, with new nuclei clusters forming on top of
the deposit as it confronts the bulk solution outside the aperture. This behavior appears to be
responsible for the initiation of needle-like structures in the deposit as shown in the SEM images
(5 and 6) of Figure 3-2 C. At more negative potentials (0 to −450 mV), the second nucleation
process is gradually merged with the first (Figure 3-4 C, curves 6− 12).
Similar types of behavior have been previously reported for metallic structures electrodeposited
from microscale/nanoscale pores.103–106 However, extrapolation of the I−t responses in Figure 3-4
demonstrates a unique phenomenon in our system related to the deposition of 3D electrodes at
higher overpotentials (0 to −450 mV). Within region “e” highlighted in Figure 3-4, the current
traces exhibit a second rise as deposition proceeds beyond 20 s (see curves 6−10). This may stem
from the production of additional small spikes “budding” off the side of the larger ones, a
phenomenon which appears to occur in this same potential range (Figure 3-2, C).
3.3.5 Analysis of i2/im2 vs t/tm
The I−t response curves revealed that two stages of nucleation exist: first, the creation of nuclei on
the planar Au at the bottom of the aperture, and second, the generation of new clusters on top of
the deposition front caused by exposure to the solution outside of the aperture. The first stage of
deposition takes place on a planar electrode located at the bottom of a 10 μm aperture (as shown
in Figure 3-5 A). All the area around this planar electrode is made of nonconductive amorphous
glass forcing the deposition to initiate at the bottom of aperture.
Figure 3-5 illustrates the relationship between the deposition overpotential and the nucleation
process by comparing current responses in a dimensionless i2/im2 vs t/tm plot specified by Equation
3.1 (See Methods). Here, the instantaneous nucleation is compared to progressive nucleation.
42
Figure 3-5 Nucleation during electrodeposition (A) Schematic of the chip-based templates
used for electrodeposition, where a 10 μm aperture provides a confined area for the growth of a
3D gold structure. The dimensionless i2/im2 vs t/tm responses of microelectrodes electrodeposited
at (B) +500, (C) +300, (D) +150, (E) +50, (F1) 0, and (F2) −550 mV and the corresponding
B C D
E F
A
43
theoretical dimensionless models for instantaneous and progressive nucleation. White scale bars
are 20 μm, and red scale bars are 40 μm.
At low overpotentials of +500 mV (Figure 3-5 B) and +400 mV (not shown), nucleation follows
the progressive model. The low overpotential allows diffusion from the bulk solution into the
aperture to readily replenish the consumed ions, resulting in formation of random-size nuclei at all
the possible active sites of planar Au (schematic cross section). This results in the formation of the
bulky structure shown in the SEM image.
At higher overpotentials such as +300 mV (Figure 3-5 C), after the first stage of nucleation inside
the aperture (described by the progressive model), a second nucleation occurs as soon as the
deposit is exposed to the bulk solution. This nucleation follows the instantaneous model. The
schematic cross section represents the sudden creation of nuclei (orange clusters) on top of the
random-sized progressive nuclei (black clusters). This results in the formation of two different
microstructures merged together (SEM image of Figure 3-5 C).
Further increases in the overpotential result in a gradual change in the first nucleation stage from
progressive to instantaneous, while the second nucleation remains instantaneous. At +150 mV
(Figure 3-5 D), a mixture of both the instantaneous and progressive behavior occurs for the first
nucleation stage, with the dimensionless current located between the two extreme forms of
nucleation. The resulting cross section and the SEM image also show the presence of a bulkier
structure underneath, which turns into thick needles as a result of the second instantaneous
nucleation (orange clusters in schematic). However, at +50 mV, instantaneous behavior is the only
dominant regime in the first and second nucleation stages (Figure 3-5 E). This instantaneous
regime in the second nucleation stage promotes a number of smaller nuclei clusters and as a result
reduces the thickness of needles in the final morphology.
At negative potentials (0 to −700 mV), the nucleation and growth follows the instantaneous model
as shown in Figure 3-5 F. From the SEM images (parts E, F1, and F2 of Figure 3-5), the deposits
tend to have finer needles and the thick ones disappear completely. This is in principle related to
the faster nucleation rate on the deposition front, which results in the creation of smaller nuclei on
a large number of active sites (Figure 3-5 F1 at 0 mV). The size of nuclei clusters continues to
shrink with increasing overpotential. At overpotentials above −450 mV, the possible sites for
44
nucleation appear on the edges and along the wall of the aperture (schematic of Figure 3-5 F2).
This results in the formation of a flaky structure.
We used 2D finite-element numerical simulations in COMSOL to model the profile of the
deposition front in our system for both the minimum and maximum overpotentials.
Figure 3-6 Two-dimensional time-varying simulation results for deposition of Au at (Top)
+500 mV and (Bottom) −500 mV. The COMSOL geometry consists of an aperture (10 μm wide,
1 μm deep) surrounded by an electrolyte in a 100 μm × 40 μm rectangular space. The electrolyte
had a diffusion coefficient of D = 1 × 10−9 m2/s with initial concentration of c = 50 mM. The
model is based on a Nernst−Planck equation with electric potential boundary conditions. The
simulation results show the deposition front progression from the bottom of the aperture for the
minimum and maximum overpotential in our system. At the right, the profiles of deposited layers
on the vertical wall of the aperture are shown at +500 mV (top) and −500 mV (bottom). All scale
bars are 20 μm.
The time-series simulation results of Figure 3-6 A reveal the slow progression of the deposition
front from the bottom of the aperture and confirm the formation of bulky structures at very low
overpotentials (+500 mV). Likewise, Figure 3-6 B demonstrates the deposition profile at −500
mV, where the deposition front moves quickly along the wall of the aperture to the edge. This
results in the formation of flaky structures at extreme negative overpotentials. Figure 3-6 C and
Figure 3-6 D also demonstrate the profile of the deposit thickness along the vertical wall of the
aperture for the bulky and flaky structures, respectively.
45
3.3.6 Surface area of nanostructured 3D microelectrodes
We also investigated variations in surface area for structures generated with different deposition
overpotentials by monitoring cyclic voltammograms generated in sulfuric acid107 (Figure 3-7 A).
Figure 3-7 Effect of deposition overpotential on surface nanostructuring. (A) Surface area of
3D structures deposited at different overpotentials (from +500 mV to −750 mV) measured in 50
mM H2SO4. Insets: SEM images of structures generated with deposition potentials of (1) +150
mV, (2) 0 mV, (3) −150 mV, (4) −750 mV. (B) Schematic of the individual spikes with different
textures deposited in the range of (1) +150 to +50 mV, (2) 0 to −50 mV,( 3) −150 to −250 mV.
All scale bars are 2 μm.
We observe that (not surprisingly) the structures generated with low overpotential (+500 mV)
exhibit the smallest surface areas (∼1200 μm2). In contrast, spiky structures generated at higher
overpotentials boast a much larger surface area. Interestingly, a 3-fold increase is observed for the
spiky structures deposited at −150 mV relative to those deposited at +50 mV, despite a similar
overall footprint. High-resolution SEM images demonstrate that the needles within these structures
have finer nanoscale features that are produced only at higher overpotentials. This finding is also
46
in accordance with the model proposed in Figure 3-5, which predicts that increasing the
overpotential results in the generation of smaller clusters of nuclei with finer spiky morphology.
As the potential increases above −450 mV, the needles are replaced by fine flakes and the surface
area decreases to ∼2600 μm2 at −750 mV (SEM image, Figure 3-7 A, inset 4). These structures
have a different overall morphology, with nucleation occurring on the edge of the aperture
followed by planar growth (described in Figure 3-5 and Figure 3-6). As a result, fine flakes with
smooth texture are formed, limiting surface area.
3.3.7 Study of DNA hybridization efficiency on the surface of 3D microelectrodes
We compared the efficiency of DNA hybridization on a subset of the structures generated in our
mechanistic study. The efficiency and kinetics of DNA hybridization depend strongly on the
density and accessibility of DNA probes attached to a surface.108–110 In general, a higher density
of probe strands attached to the surface produces larger signals upon hybridization to redox-tagged
complementary strands. However, extremely high probe densities can diminish the rate of
hybridization and the efficiency of detection. This is due to the accumulation of more negative
charges and the steric hindrance of the populated capturing stands on the surface, which limits the
ability of target molecules to easily access and hybridize to their complements.108
To investigate the efficiency of DNA hybridization on our 3D sensors, we immobilized a closely
packed monolayer of thiolated single-stranded DNA probes on their surface. A complementary
DNA strand labeled with methylene blue is then used to monitor the kinetics of hybridization
(Figure 3-8 A).
47
Figure 3-8 DNA detection assay based on 3D gold microsensors. (A) Capture probe (blue) is
attached to the sensor. The complementary DNA strand (red) carries methylene blue redox tag
(MB) to the electrode surface, which generates an electrochemical signal upon hybridization of
DNA strands. (B) Kinetic responses corresponding to DNA−DNA hybridization at sensors
deposited with different overpotentials. (C) The hybridization rate and (D) the corresponding t1/2
values show the pronounced variation in the efficiency of hybridization induced by different
sensor morphologies.The kinetic responses (B) and the corresponding rate measurements (C)
reveal that the rate of DNA−DNA hybridization exhibits a strong dependence on sensor
morphology. Spiky structures with finer nanostructuring provide more accessibility for the target
hybridization and accelerate the rate of DNA−DNA hybridization. However, on the flaky
structures, the rate of hybridization is reduced, likely due to a blocking effect. The structure of
the flakes creates hidden sites that may trap probe strands during overnight immobilization but
have limited accessibility during the hybridization time. The calculated time t1/2 (the time needed
for half of the capturing strands to hybridize to signaling strands) is also reported in D.
target DNA
probe DNA
AMB
B
C D
48
3.4 Conclusion
We conducted a detailed study of the electrodeposition of three-dimensional gold microsensors
with varied morphologies. This study revealed the influence of a number of important parameters
on the assembly of gold both on a nano- and microscopic scale. This allowed for the customization
of parameters to encourage the growth of 5-fold twinned needles which are of particular promise
due to their proven utility in biosensing applications. At low overpotentials (+500 to +300 mV)
the progressive nucleation and low growth rate encourage the formation of bulky structures. By
increasing the overpotential, two discrete stages of nucleation resulted that promoted the creation
of spiky structures (+150 to −250 mV). The more instantaneous the nucleation processes
(encouraged by higher overpotential) are, the finer is the nanostructuring of the spiky structures.
At overpotentials above −450 mV, flaky structures are formed as a result of nucleation on the edge
of the aperture. In order to understand the importance of morphology in DNA-based biosensing
applications, we studied the rate of hybridization on the surface of these electrodes. Optimal
structures that promoted highly efficient hybridization kinetics were identified.
3.5 Methods
3.5.1 Chip fabrication
Gold-coated glass wafers (with chrome adhesion layer and positive photoresist coating) were
purchased from Telic Company (Valencia, CA). Gold was patterned to create a series of seven
leads using standard lithography of the photoresist followed by etching of the gold and chrome
layers. Negative photoresist (SU-8 2002) was applied by spin coating and developed using
photolithography to create single 10 μm apertures on each lead.
3.5.2 Electrodeposition
Electrodeposition of 3D microelectrodes was performed with a BASI Epsilon potentiostat in DC
voltammetry mode. A three-electrode setup was employed with voltage measured relative to an
Ag/AgCl reference electrode, and a platinum wire was used as the counter electrode. The
deposition time was adjusted for each particular structure in order to keep the quantity of deposited
gold roughly constant (i.e., the total coulombs of charge transferred). For the concentration series,
the solution was between 1 mM and 500 mM HAuCl4 in a 0.5 M HCl supporting electrolyte. The
voltage applied was 0 mV relative to Ag/AgCl. At 50 mM, the time of deposition was 100 s and
49
other deposition times were adjusted accordingly. For the voltage series, the concentration of
HAuCl4 was 50 mM and the voltage was varied from +500 mV to −700 mV. For the viscosity
series glycerol was added to between 0% and 75% v/v (HCl final concentration was maintained at
0.5 M), with the voltage set to 0 mV and the HAuCl4 concentration maintained at 50 mM. The
surface area of 3D microelectrodes was calculated by measuring the area of the Au reduction peak
(∼0.80 V vs Ag/ AgCl) in 10 mM H2SO4 solution. The cyclic voltammetry studies were carried
out in the range of 0−1.5 V, and the reduction charge was obtained by integrating the reduction
peak. The result was then divided by 500 μC/geometric cm2 to obtain a geometric estimation of
the surface area.107
3.5.3 Surface characterization
Scanning electron microscopy (SEM) images were acquired on a Quanta FEG 250 ESEM. The
instrument was used in the high vacuum mode, with typical parameters being 10 kV bias and a 2.5
nm spot size. Magnifications between 1000× and 50 000× were employed. Focused ion beam (FIB)
imaging was performed using the Hitachi nanoDUE’T NB5000. Samples were first grounded
using a thin layer of silver paste. A thin layer of protective tungsten was then applied to the area
of interest using the machine’s built-in deposition capacity. Sectioning was performed using high-
energy ions to reveal a cross section of the structure. SEM imaging was then done using the built-
in SEM module.
3.5.4 Nucleation and growth model
The study of electrochemical three-dimensional nucleation processes involved correlating the
current to the number of clusters on the electrode surface.101–106 As the nucleation rate per active
site on the surface, A, and the number density of active sites for nucleation, N0, vary with
overpotential of deposition, the potentiostatic current transients establish the relationship between
deposition overpotential and the kinetics of nucleation:101,102
𝑖 = (
𝑧𝐹𝐷1/2𝑐
𝜋1/2𝑡1/2) (1 − 𝑒𝑥𝑝{−𝑁0𝜋𝑘𝐷[𝑡 − (1 − 𝑒−𝐴𝑡)/𝐴]}), (3.1)
where i is the current density; z, distance normal to the plane; F, the Faraday constant; D, diffusion
coefficient; c, bulk concentration; t, time; N0, number density of active sites; k, dimensionless
constant for growth rate of diffusion zones. Considering the im as the maximum current taking
50
place at tm, Equation 3.1 can be presented in a dimensionless form by simply plotting i2/im2 vs t/tm
for different values of the dimensionless parameter α = N0πkD/A. The two extreme forms of
nucleation are defined for small α or fast nucleation on a limited number of active sites,
“instantaneous” nucleation, and large α or slow nucleation on a large number of active sites,
“progressive” nucleation.
3.5.5 Electrode functionalization
Thiolated-DNA strands (0.1 mM) were incubated with TCEP (10 mM) for 1 h to reduce the
disulfide bonds. The solution was then diluted to 100 nM PBS 1×, pH 7.0. Glass chips having a
set of seven microelectrodes were incubated in 100 μL of 100 nM solutions of thiolated strands
overnight. Chips were then rinsed with DI water and incubated in 100 μL of 3 mM MCH in buffer
for another 3 h to displace nonspecifically adsorbed DNA and passivate the remaining electrode
area. After thoroughly rinsing with DI water, chips were stored in buffer. The thiolated strand
surface density (i.e., the number of DNA molecules per unit area of the surface) was determined
to be ∼6 × 1012 strands/ cm2 by measuring peak current.108
3.5.6 Sensor measurements
Electrochemical measurements were performed at room temperature using an EmStatMUX
potentiostat multiplexer (PalmSens Instruments, The Netherlands) and a standard three-electrode
configuration containing a platinum counter electrode wire, Sigma-Aldrich and an Ag/AgCl (3 M
NaCl) reference electrode (CHI). Experimental data were collected using square wave
voltammetry from −0.05 to −0.45 V in increments of 0.001 V vs Ag/AgCl, with an amplitude of
50 mV and a frequency of 60 Hz. Peak currents were fitted using the manual fit mode in the
PSTrace software (PalmSens). All measurements were taken immediately after adding the reagents
to the solution (100 nM of signaling strands) to measure the kinetics of hybridization.
51
Power-free, digital and programmable dispensing of picoliter droplets using a Digit Chip
So far this thesis has illustrated a number of elements which have application in a POC setting, but
which still rely on external instrumentation. In this chapter, we explore a device architecture
designed to be used with a minimum of external equipment and to perform a number of simple
assays in an intuitive and low-cost manner. The primary motivation for this device is the ability to
move microscopic quantities of fluid in a controlled manner without requiring external pumping
infrastructure, since the manipulating of fluid is the most fundamental operation of a microfluidic
device. Fluid movement is actuated by the press of a button with a single finger (digit), allowing
intuitive operation. This macroscopic operation is converted into discrete microscopic flow
through the integration of a series of capillary valves which break upon each instance of applied
pressure. This button-valve pairing is applied in the production of a concentration gradient and, in
conjunction with a cheap smartphone-based fluorimeter, the measurement of bacterial antibiotic
susceptibility. This assay requires no additional components and can be performed with minimal
training.
Reprinted with permission from Mepham A., Besant J.D., Weinstein A. W., Burgess I.B., Sargent
E.S., Kelley S.O., “Power-free, digital and programmable dispensing of picoliter droplets using a
Digit Chip” Lab on a Chip, 2017, 17 1505-1514. Reproduced by permission of The Royal Society
of Chemistry.
Link to publication online: https://doi.org/10.1039/C7LC00199A
Disclosure of work within this manuscript: A.M., J.D.B. A.W.W. and I.B.B. designed the
experiments. A.M, J.D.B. and A.W.W. performed experiments. Data analysis and manuscript
writing were performed by A.M. and J.D.B. with assistance from A.W.W, I.B.B., E.H.S., and
S.O.K.
52
4.1 Abstract
There is a growing need for power-free methods to manipulate small volumes of liquids and
thereby enable use of diagnostic assays in resource-limited settings. Most existing self-powered
devices provide analog manipulation of fluids using paper, capillary or pressure-driven pumps.
These strategies are well-suited to manipulating larger micro- and milliliter-scale volumes at
constant flow rates; however, they fail to enable the manipulation of nanoliter and picoliter
volumes required in assays using droplets, capillary sampling (e.g. finger prick), or expensive
reagents. Here we report a device, termed the Digit Chip, which provides programmable and
power-free digital manipulation of sub-nanoliter volumes. The device consists of a user-friendly
button interface and a series of chambers connected by capillary valves that serve as digitization
elements. Via a button press, the user dispenses and actuates ultra-small, quantitatively-
programmed volumes. The device geometry is optimized using design models and experiments
and precisely dispenses volumes as low as 21 pL with 97% accuracy. The volume dispensed can
be tuned in 10 discrete steps across one order-of-magnitude with 98% accuracy. As a proof-of-
principle that nanoliter-scale reagents can be precisely actuated and combined on-chip, we deploy
the device to construct a precise concentration gradient with 10 discrete concentrations.
Additionally, we apply this device alongside an inexpensive smartphone-based fluorescence
imaging platform to perform a titration of E. coli with ampicillin. We observe the onset of bacterial
death at a concentration of 5 μg mL-1, increasing to a maximum at 50 μg mL-1. These results
establish the utility of the Digit Chip for diagnostic applications in low-resource environments.
4.2 Introduction
Despite recent advances in microfluidics for diagnostics, many of the most sensitive tests remain
unavailable in resource limited settings due to the prohibitive cost and power requirements of the
necessary instrumentation.111 Most microfluidic systems require costly electrically-powered
pumps for fluidic actuation, and this limits the deployment of these technologies in the challenging
field conditions of developing world environments, where in many cases the requirements of
portability and remote location may limit access to reliable sources of electric power.
53
To increase access to point-of-care diagnostic technologies, low-cost and low-power fluidic
actuation systems are needed.111,112 Recently, compelling strategies for passive fluid control have
been reported that have included lateral flow paper microfluidics,113–115 vacuum pumps,116,117
pressure pumps,118,119 fibres,120 and capillary pumps.29,121 Sophisticated sample manipulations are
possible using these power-free actuation systems, and multi-step assays have been
demonstrated.116,122–126 However, these analog approaches to fluid manipulation are typically
optimized to regulate the flow rate of bulk fluids, and have not yet been demonstrated to be well-
suited for complex manipulations of very small reagent volumes.
Many diagnostic assays require the use of small samples or reagent volumes, including those that
sample blood from a finger prick and those that require expensive reagents. As a result, several
chip-based techniques for digital manipulation of small volumes have been developed including
droplet127–129 and digital microfluidics.130–132 In both cases, droplet actuation requires a power
source such as an electric pump (droplet microfluidics) or a high voltage source (digital
microfluidics). These requirements limit the feasibility of these methods in resource-limited
environments.
Developing power-free platforms for digital manipulation of small sample volumes poses two
distinct challenges not present in their analog counterparts: 1) dispensing mechanisms need to be
very precise and optimized for small volumes;133,134 2) sample controls need to be modular and
compatible with a user-friendly interface. Innovative architectures for self-powered,113 or
manually-powered135 fluid manipulation, including the SlipChip136–139 have recently overcome a
number of key challenges; however, thus far these self-powered techniques only allow a limited
number of sequential manipulations and lack the modularity and programmability of powered
techniques such as digital microfluidics. Devices driven by finger pressing have also been
developed,140,141 however, these devices typically make use of one directional flow valves and as
such are considerably more difficult to manufacture. Moreover, these devices cannot manipulate
picoliter volumes of fluid. Finally, these devices typically metre only a single, fixed volume of
fluid.
Here, we develop a user-friendly interface for power-free and digital manipulation of small
volumes using capillary valve- dispensers as modular elements, which are linked to pressure-
regulated buttons. The device, termed the Digit Chip, uses the pressure applied from pressing
54
buttons manually to break individual capillary valves and move liquids in fixed-volume
increments. Using theory and experiments, we optimize the geometry and surface properties of the
valves and pressure-regulated buttons, and show that droplets can be dispensed and manipulated
accurately across a wide range of volumes down to the picolitre scale. Dispensing, actuating, and
mixing reagents are essential components of many important biological and chemical assays. To
demonstrate that small volumes can be precisely actuated and mixed, we use the Digit Chip to
create a precise, discretized concentration gradient with nanoliter volumes. We also illustrate how
this architecture enables a convenient and low-cost technique to measure the susceptibility of
bacteria to antibiotics. This is an important capability that could allow small collections of bacteria
to be assessed for drug resistance via phenotypic testing without the need for any type of traditional
lab infrastructure. The Digit Chip could also be applied in a variety of biological and chemical
assays which depend on reagent dilutions and chemical gradients including generation of standard
curves, optimization of reaction conditions,142 and chemotaxis.143
4.3 Results and discussion
4.3.1 Overview of the digit chip
The Digit Chip consists of a series of chambers connected by capillary valves that serve as
digitization elements (Figure 4-1 A).
55
Figure 4-1 Overview of the Digit Chip. (A) Schematic illustrating precise and user-
programmable dispensing of ultra-low volumes using a Digit Chip Chambers, which serve as the
digital elements, are connected in series by capillary valves. After each button press, an
additional chamber is filled with fluid. (B) A schematic illustrating the spontaneous filling of a
chamber after pressure-induced bursting of a capillary valve. (C) Theoretical bursting pressure as
a function of capillary valve width with a 90° valve expansion angle and 50 μm channel height.
(D) Theoretical bursting pressure as a function of valve expansion angle for various contact
angles assuming a 50 μm valve width and a 50 μm channel height.
Upon manual application of a pressure via a button, the fluid bursts through the first capillary valve
and enters the adjacent chamber (Figure 4-1 B). Through capillary pressure, the chamber fills
spontaneously until the fluid reaches the subsequent capillary valve and the flow is arrested. The
user can opt to fill the next chamber by re-pressing the button. The solution volume dispensed is
programmed by the number of times the user applies pressure to the button.
The principle underlying the design of the device relies on the function of capillary valves created
within the fluidic structure.144 At the interface between the narrow and wide regions, the liquid–
air interface becomes pinned and requires a certain threshold pressure to resume movement. A
valve exists between the narrow linear channel and the larger circular chamber, and the pressure
is supplied by the press of a button. Once the valve bursts, the fluid will spontaneously fill the
56
circular chamber. As long as the user only presses the button for a short period of time, the pressure
is relieved before the fluid reaches the subsequent valve. As such, the pressure is below the critical
pressure upon arrival at the next chamber, and motion ceases.
4.3.2 Bursting pressure model and design principles
Each valve must have a bursting pressure low enough that it can be readily applied by a human
finger, but large enough that the valve does not burst spontaneously. To calculate the capillary
valve bursting pressure, we use the following previously described equation for the maximum
sustainable pressure across the meniscus in a rectangular configuration of the capillary burst
valve:144
𝛥𝑃 = −2𝜎 (
cos(𝜃𝐼)
𝑤+
𝑐𝑜𝑠(𝜃𝐴)
ℎ), (4.2)
where ΔP is the pressure difference across the liquid–air interface, w is the valve width, h is the
channel height, σ is the surface tension of the liquid–air interface (72.9 mN m-1), θI is the contact
angle with the initial side-wall prior to the expansion, and θA is the critical advancing contact angle.
Upon the application of pressure, the meniscus will bulge until the contact angle with the new
sidewall after the valve reaches the critical advancing contact angle, θA. This occurs when θI = θA
+ β, where β is the valve expansion angle, or when θI is greater than 180°, the maximum sustainable
contact angle.144 Thus the valve will burst when θI equals the lower of θA + β or 180°. These
equations assume that the fluid is dispensed from an infinite reservoir and do not consider the
receding interface.
Using this equation, we modeled the bursting pressure as a function of valve width (Figure 4-1 C)
and valve expansion angle (Figure 4-1 D) for a variety of contact angles. As expected, the bursting
pressure increases with narrower valves and larger valve expansion angles. The calculations
indicate that a valve expansion angle of approximately 90° or higher is optimal as the bursting
pressure rapidly drops off for angles less than 90°. At low contact angles, the bursting pressure
approaches zero for valve widths around 25 μm, thus valves narrower than 25 μm are ideal.
It is critical that the device surface be hydrophilic so that the chambers fill spontaneously via the
capillary pressure after the valves burst. Thus, the contact angle of the fluid should be less than
57
90°. On the other hand, Figure 4-1 C and Figure 4-1 D suggest that at low contact angles, the
valves will burst spontaneously for most valve widths and valve expansion angles. Thus, the
surface should be engineered to be only slightly hydrophilic. We measured the critical advancing
contact angle of PBS on PDMS for a variety of oxygen plasma treatment times (Appendix Figure
B-4). We found that with a 30 s oxygen plasma exposure, the contact angle of PBS on PDMS is
70° which is compatible with the Digit Chip. PDMS is a useful material for this device because of
its compatibility with rapid prototyping. While its elastomeric properties are useful in the user
interface, they are not essential for the functioning of the wells. The stability of PDMS surface
chemistry was sufficient for our device to function properly within the same day of plasma treating.
However, using different materials for the wells (e.g. glass) that have a more stable surface
chemistry may be more advantageous if the device were to be mass-produced and stored for longer
periods before use.
4.3.3 Optimization of device geometry
Motivated by these calculations, we fabricated an array of devices with various capillary valve
widths and expansion angles to further refine the design. Devices were fabricated by pouring
PDMS on a 50 μm tall SU-8 master mold patterned using standard photolithography. After curing,
the PDMS was plasma treated and bonded to a glass substrate.
Figure 4-2 A shows the measured bursting pressure as a function of valve width and Figure 4-2 B
shows the bursting pressure as a function of valve expansion angle.
58
Figure 4-2 Experimental investigation of the device geometry and its optimization. (A)
Experimentally measured bursting pressure as a function of valve width. The dotted line
represents the calculated bursting pressure assuming a 70° contact angle and 90° valve expansion
angle. (B) Experimentally measured bursting pressure as a function of valve expansion angle.
The dotted line represents the calculated bursting pressure assuming a 70° contact angle and 50
μm valve width. (C) The measured bursting pressure as sequential chambers are filled. (D) The
measured volume dispensed as a function of chamber volume. (E) Images illustrating the
spontaneous filling of a 12 nL chamber after the valve is burst by a user-applied pressure.
Sequential dispensing of volumes in approximately (F) 12 nL, (G) 580 pL and (H) 140 pL
increments. Errors bars represent standard error.
We compared our measured values to the theoretical predictions and in both cases, we find good
agreement with the theory. In all of our designs, the channels connecting the wells were sufficiently
short to ensure that there was not enough buildup of flow momentum between wells to break a
capillary valve. Figure 4-2 C shows the bursting pressure as a function of the number of valves
filled. On average, we observe only a 5% increase in bursting pressure after each sequential valve
is filled.
59
We studied the accuracy of filling as a function of chamber size (Figure 4-2 D). A series of devices
were designed with chamber diameters ranging from 55 μm to 800 μm and channel heights ranging
from 5 μm to 50 μm tall. Since the capillary valve bursting pressure depends on the width of the
valve and the expansion angle, the sizes of the circular chambers can be freely adjusted to allow
for different volumes while performing in the same manner. The volume of these chambers ranged
from approximately 21 pL to 24 nL. The valve widths scale with the chamber size and range from
2.5 to 20 μm wide. We found that all chambers and wells, including those as small as 21 pL, could
be filled with at least 97% accuracy. The remaining chamber sizes filled with high accuracy
(Appendix Figure B-5). A small amount of error is caused by the incomplete filling of some
chambers due to the occasional formation of small bubbles. However, the valves were stable, not
spontaneously breaking even after many minutes. A small amount of evaporation does occur if the
device is left unattended for a manner of minutes, but this does not destabilize the air–liquid
interface.
This smallest dispensing volume achieved approaches the minimum achievable limit for our
design, which we estimate to be 15–20 pL. The fabrication tolerance of our photolithography
process creates minimum lateral sizes of the channel and chamber (to maintain the near 90°
spreading angle). The valve design places an additional fundamental constraint on the valve aspect
ratio, which limits how small the channel height can be. Since the free-energy barrier encountered
at the side walls in the chamber opening must always exceed the favourable free-energy change
associated with wetting the top and bottom walls, the width:height aspect ratio in the channels
connecting the chambers must stay below a maximum value. This maximum aspect ratio is
derivable from Equation 4.2 and given by:
[𝑤/ℎ]𝑚𝑎𝑥 = −
𝑐𝑜𝑠(𝜃𝐴 + 𝛽)
cos(𝜃𝐴). (4.3)
A contact angle of 70° and a spreading angle (β) of 90°, gives a maximum aspect ratio is
2.75, limiting how small the channel height can be as a function of the width. This limit was in
agreement with experiments showing that further reduction of the channel height for our smallest
chamber size led to spontaneous breaking of the valves.
4.3.4 Designing a user-friendly interface
60
We sought to design an interface that allows the user to easily apply the appropriate pressure to
break one capillary valve. The interface consists of two hollow PDMS chambers connected in
series to the sample loaded in the Digit Chip (Figure 4-3 A and Figure 4-3 B).
Figure 4-3 The Digit Chip interface for controlled dispensing of droplets. (A) Schematic
illustrating the user-friendly interface for precise application of pressure. The interface consists a
button and a pressure regulation chamber patterned in PDMS. When the user depresses the
button with a thumbpress, the applied pressure is controlled by the size of the pressure regulation
chamber. Pressure is vented through a small hole when the user releases the button. (B) Image of
the device with a nickel shown for scale. (C) The applied pressure as a function of the ratio of the
size of the button and pressure regulation chambers. (D) Measurement of the volume displaced
when the button is depressed. (E) The number of chambers filled as a function of the size of the
pressure regulation chamber. Single chambers could be accurately filled when the button
chamber was 15% the volume of the pressure regulator. (F) Images of solution dispensed in 24
nL increments using the PDMS button. (G) Accuracy of filling in 24 nL increments after each
sequential button press using the user-interface. Solution was dispensed more than 10 times
sequentially. Errors bars represent standard error.
When the user fully compresses the first chamber, which serves as the button, the gas within the
chamber is displaced and a pressure is applied to the sample in the Digit Chip. The applied pressure
61
is regulated by tuning the size of the second pressure-regulation chamber. The pressure is vented
through a small hole when the user releases the button.
The applied pressure difference generated by pressing the button can be approximated using
Boyle's law:
𝛥𝑃𝑎𝑝𝑝𝑙𝑖𝑒𝑑 = 𝑃𝑎𝑡𝑚 (
𝑉𝑏 + 𝑉𝑠
𝑉𝑠− 1), (4.4)
where ΔPapplied is the applied pressure difference, Patm is the atmospheric pressure of 101 325 Pa,
Vb is the volume of the button chamber and Vs is the volume of the regulation chamber and the
connective tubing. This equation assumes the button is fully compressed when pressed. This two-
chamber design limits the maximum pressure a user can apply to the valve (achieved when the
button is fully compressed) and therefore ensures that the user cannot press too hard.
We measured the applied pressure as a function of the ratio of the size of the button and pressure
regulation chambers (Figure 4-3 C). The measured pressures are lower than the theoretical
predictions (Figure 4-3 D), suggesting that the button chamber retains about 20% of its volume
when fully depressed.
To test the efficacy of this interface, we connected various designs in series to a Digit Chip with
20 μm wide valves and 24 nL chambers. We measured the number of chambers filled per button
press (Figure 4-3 E). We found that when using an interface with a ratio of the two chambers of
15%, we could accurately fill a single chamber per button press. The applied pressure from this
interface design is 11 kPa, which is higher than the approximately 6 kPa capillary valve bursting
pressure predicted by Equation 4.2. This is expected as the pressure generated by the user must be
greater than the valve bursting pressure due to the pressure drop along the tubing, fluid reservoir,
and channel and the fact that the flexible PDMS chambers expand under pressure. Figure 4-3 F
shows images of dispensing fluid in 24 nL increments using this interface. Using the interface, we
found that we could accurately dispense liquid in 24 nL increments over 10 times sequentially.
The volume dispensed ranged from 24 nL for 1 button press to over 300 nL for 13 button-presses
(Figure 4-3 G) and the chambers filled with over 98% filling accuracy and less than 3% standard
deviation (Figure 4-3 G). The volumes dispensed in Figure 4-2 F–H were further replicated using
the button to show its applicability to a variety of volumes (Appendix Figure B-6).
62
4.3.5 Generation of a discrete concentration gradient
Concentration gradients and reagent dilutions are important in a variety of biological and chemical
assays. As a demonstration that solutions can be accurately dispensed and combined, we used the
Digit Chip to mix small solution volumes in precise ratios (Figure 4-4 A).
Figure 4-4 Generation of a discretized concentration gradient. (A) Schematic illustrating on-
chip dispensing and mixing of reagents using the Digit Chip. After the sample is loaded, both
solutions are dispensed in various ratios in 24 nL increments. The two solutions were sent to a
central mixing chamber by manually injecting air using a syringe. (B) Air channels were valved
using a screw to depress the PDMS and block the channel. (C) On-chip generation of a
discretized concentration gradient using the Digit Chip. Errors bars represent standard error. (D)
Images acquired with an optical microscope after mixing the blue and yellow dye in various
ratios with the Digit Chip.
In this device, two separate solutions can be dispensed up to 10 times in increments of 24 nL. After
dispensing the liquid, the fluid is actuated towards the central mixing chamber by manually
injecting air using a syringe. The air channels are closed during dispensing by screw valves,145
which depress the channel (Figure 4-4 B).
We dispensed 10 different ratios of blue and yellow dyes in 24 nL increments and mixed the dyes
in the central chamber on chip. For each concentration, the sum of the number of both droplets
63
dispensed was kept constant at 10, corresponding to a total volume after mixing of 240 nL. Figure
4-4 C shows the measured ratios of the two solutions mixed on chip. The r2 value of the fit to the
line representing the expected ratio of the dyes is 0.98. The deviation in our measured values is
due to the compounded error of filling 10 wells and the fact that some liquid remains trapped in
the chambers after injecting air. Figure 4-4 D shows images of the resulting colors generated from
mixing the two dyes.
4.3.6 A low-cost platform for rapid determination of bacterial antibiotic susceptibility
In order to demonstrate the usefulness of the Digit Chip for a practical application, a device was
developed which tests the effect of antibiotic concentration on bacterial growth.146
Figure 4-5 Testing of antibiotic susceptibility. (A) Schematic illustrating architecture of
device. Insets show Digit Chip motif (i) and capillary valve between bacterial metering channel
(yellow) and growth chamber (ii). (B) Simple, smartphone-base fluorescence platform allowing
for excitation with green light and imaging of resultant red fluorescence. (C) Time series
demonstrating the conversion of non-fluorescent resazurin to highly fluorescent resorufin by
bacterial metabolism. (D) Curve illustrating the effect of ampicillin titration on the viability of E.
coli.
64
The chip is comprised of a pair of channels leading into a central growth chamber, each with a
respective air inlet (Figure 4-5 A). A single outlet is present on the far side of the growth chamber.
The top channel is the standard Digit Chip motif, with 10 chambers separated by capillary valves
(Figure 4-5 A, i). By filling the desired number of chambers with antibiotic solution and expelling
the metered antibiotic into the growth chamber, between 1 and 10 equivalents of antibiotic are
introduced. The side channel is a single bacterial metering channel separated from the growth
chamber by a capillary valve (Figure 4-5 A, ii). This chamber is filled with a fixed volume of
bacterial solution, which can be transported by air into the growth chamber where it mixes
diffusively with the antibiotic solution. This device can then be submerged in a 37°C water bath
to allow for bacterial growth.
To observe the viability of the bacteria, resazurin dye was added to the bacterial medium. Only in
the presence of actively metabolizing bacteria will the resazurin be reduced to its fluorescently
active product, resorufin. In order to measure the fluorescence of this product, a simple smart
phone-based imaging platform was devised (Figure 4-5 B). The excitation system consists of a
high-powered green LED light source, along with a collimating lens to focus the excitation light
and a plastic green filter to remove extraneous wavelengths. This light is focused onto the growth
chamber of the chip from an oblique angle to minimize detection of the excitation light. The
emitted light is passed through a 589 nm bandpass filter and imaged using an LG G3 smartphone
camera. To magnify the image of the growth chamber, a PDMS lens was fabricated and affixed
directly onto the back of the smartphone, covering the camera lens. The resultant RGB image was
then split into its component red, green and blue channels, and the average intensity of the red
channel was measured. In keeping with the intended use of the Digit Chip in low resource settings,
the entire platform including all components can be purchased and constructed for under 60 USD.
In order to confirm the function of the platform, a high concentration (1.0 × 108) of bacteria was
introduced into the growth chamber in the absence of antibiotic and the fluorescence was imaged
every 45 minutes for 3 hours. The resultant images display monotonically increasing fluorescence
as the bacteria replicate and metabolize (Figure 4-5 C). Evidently, the system is more than adequate
for detecting the conversion of resazurin to resorufin and thereby measuring bacterial viability.
The susceptibility of E. coli to ampicillin was chosen as a suitable test case to confirm the efficacy
of the Digit Chip. A wide range of concentrations (0, 1, 2, 5, 10 and 50 μg mL-1) of antibiotic were
65
introduced using the discretized measurement chambers and two stock solutions (100 μg mL-1 and
1000 μg mL-1). These were mixed with bacterial solution (12.5 × 106 cfu mL-1) and incubated for
8 h prior to measurement of fluorescence. The results are illustrated graphically in Figure 4-5 D.
The graph follows a standard sigmoidal shape, with bacterial death beginning at a concentration
of about 5 μg mL-1 and increasing until 50 μg mL-1. These results are in good agreements with past
studies, which have shown a minimum inhibitory concentration of 2 μg mL-1147 and complete
inhibition of growth at 50 μg mL-1.148 Evidently, the ability of the Digit Chip to create titrations
with a wide range of concentrations is ideally suited to investigating bacterial antibiotic
susceptibility. Moreover, due to the extremely small volumes used (on the order of 100 nL) very
small quantities of bacterial sample and antibiotic solution are required. This is particularly
important in situations where the sample is limited by biological or financial constraints.
Furthermore, this device is amenable to any combination of bacterial species and antibiotic,
allowing for widespread implementation.
4.3.7 Discussion
This type of self-powered digital fluidic device may be useful for a wide class of assays in low-
resource settings that require low sample or reagent volumes (e.g. pin-prick assays), or that require
complex multi-step manipulations or precise timing that are difficult to automate using existing
passive analog fluidics. Using this device, the user controls the time at which various reagents are
introduced which eliminates the need for built-in timing mechanisms. This device is especially
useful for assays in which the volumes dispensed are systematically varied, such as titrations. This
device architecture is compatible with many common fabrication techniques currently used to
make channel-based microfluidics at low cost.149 As with other microfluidic technologies, the
replacement of PDMS with more cost-effective alternatives (e.g. plastics, glasses) might also
accompany the transition to larger scale manufacturing.149 For our device this would have the
added benefit of enabling us to choose materials whose surface chemistry is more stable and
suitable for long-term storage.
Along with these advantages, the Digit Chip does have some weaknesses. The nature of the
capillary valve requires that metered fluid be displaced and replaced with air before the valve is
re-established, meaning that the chambers need to be flushed with air between subsequent
66
dispensings. Furthermore, valve dimensions need to be precise to allow for correct behaviour.
However, these drawbacks are relatively minimal given the simplicity and accuracy of the device.
We applied the Digit Chip to evaluating antimicrobial resistance, which is an important target
application to enable in resource-limited settings. The features of the Digit Chip would make it
straightforward to carry out the type of drug titration that is typically used to assess whether a
frontline antibiotic will be efficacious. This approach is also applicable to any of the multitude of
tests in biology and chemistry that require titration. For example, the testing of dissociation
constants and the activity of enzymes could be tested using minute quantities of reagents, provided
that a measurable color change occurs.
The adoption of this technology in a wider variety of assays will also include its functioning with
complex fluids, such as whole blood, that have a higher viscosity and contain suspended particles.
Although the core principles of our device are independent of viscosity, whose only independent
effect on our chips was to change the speed with which the chambers filled, the presence of
suspended particles (e.g. blood cells) would place size restrictions on our chambers, with the
smaller geometries having the capacity to become easily clogged.
Increasing the complexity of manipulations and involving more reagents in the Digit Chip would
require more inlets, and a 2D configuration of these buttons and valves. Valving mechanisms must
be added at each node to control the directionality of pressure applied, just as was illustrated here
to make the concentration gradient. One natural extension of the existing device would be to have
two parallel Digit Chip motifs emptying into a common chamber, with one having chambers 10×
larger than the other. This “ones and tens” configuration would allow any integer value between 1
and 100 to be metered and dispensed with only a small increase in complexity.
Just as has been shown with digital microfluidics,130–132 a vast modular library of manipulations
and different assays could be built from the Digit Chip using a few stock 1D and 2D configurations
of wells, valves and buttons. These device configurations would need to be optimized for the
maximum number and type of fluid manipulations possible, but not with a need for any reagents
to be loaded beforehand. Therefore a single chip design could be used for many assays (e.g. any
titration involving a given number of reagents) or rapid experiments performed in the field.
However, this modularity comes at the cost of easy assay automation. A general limitation of user-
67
programmable digital fluidic platforms is that more sophisticated assays may require more buttons
and thus, greater user involvement and more complicated instructions
4.4 Conclusion
In summary, we introduce a device for power-free and user programmable manipulation of small
sample volumes. A sequence of microfluidic chambers connected by capillary valves connected
to pressure-regulated buttons allow for discretized dispensing of sub-nanoliter volumes. This
device architecture enabled power-free dispensing of volumes in increments as low as 21 pL and
near 97% accuracy. The volume dispensed could be fine-tuned in 10 increments across one order-
of-magnitude. The number of volume increments could be increased in future iterations of the
device. A series of devices is used to generate a concentration gradient with 10 discrete
concentrations in sub-microliter volumes, and to determine the susceptibility of bacteria to various
concentrations of antibiotic. This device could help bring a new class of assays, which require
sophisticated manipulations of small volumes, to low-resource settings.
4.5 Methods
4.5.1 Digit chip fabrication
Using standard photolithography, we patterned a 50 μm tall SU-8 3050 (Microchem, MA) layer
on a silicon wafer (University Wafer, MA). PDMS (Dow Chemical, MI) was dispensed onto the
wafer and cured for 1 hour at 67 °C. After curing, the PDMS was removed from the silicon wafer
and holes were punched to form the inlet and outlet.
4.5.2 Fabrication of user-friendly interface
The mold was printed using a μPrint 3D printer. PDMS was dispensed onto the mold and cured at
67 °C for 1 hour. The PDMS was removed and holes were punched to create outlets and a small
vent in the button chamber. The PDMS was plasma treated and bonded to a glass slide. Silicone
tubing was used to attach the button to the Digit Chip.
4.5.3 Contact angle measurements
PDMS was treated with an oxygen plasma for a variety of exposure times. Advancing contact
angles were measured using ImageJ to analyse images of the droplets acquired using a camera.
68
4.5.4 Bursting pressure measurements
Valve bursting pressures were measured using a syringe pump connected to the chip. After
introducing PBS (pH 7.4) into the Digit Chip, the syringe pump was connected to the chip with
9.5 cm of silicone tubing (0.76 mm inner diameter). The initial volume of gas in the 1 mL syringe
and tubing was measured. The syringe was slowly compressed at a rate of 20–50 μL min-1 while
monitoring the capillary valve under a microscope. The pump was stopped and the pressure was
relieved as soon as the valve burst. The volume change of gas contained in the syringe and tubing
was recorded. The induced pressure was calculated using the initial and final volumes using the
ideal gas law. The changes in pressure were measured at least 5 times per chip.
4.5.5 Measurements of applied pressure using the elastomeric button
The outlet of the user-interface was connected to silicone tubing (0.76 mm inner diameter) filled
with a plug of PBS buffer (pH 7.4). After compressing the button, we recorded the change in
displacement of the plug. This change in displacement of the plug was used to calculate the volume
of gas displaced while pressing the button. This volume change was converted into a pressure
change using Boyle's law.
4.5.6 Chamber filling percentage measurements
The filling percentage of chambers was measured by acquiring optical images under a microscope
(Nikon) and analyzing the images using ImageJ.
4.5.7 Generation of a discretized concentration gradient
The device was fabricated from PDMS as stated above. Auxiliary air outlets were connected to
each fluid inlet to enable fluid to be pumped in from both sides. Screw valves were fabricated on
each air outlet by 3D printing a chuck to suspend the screws above the channel. The valves were
closed or opened by turning the screw clockwise or counter-clockwise. Before dispensing the dyes,
the air channels were blocked by closing the valves to ensure that liquid did not enter the air
channels. Both samples were introduced and dispensed for the desired number of chambers. The
sum of the number chambers filled with blue and yellow dyes was held constant at 10 for all
concentrations. Samples used were ddH2O with food coloring at 2 drops per mL. Devices were not
plasma treated or bonded to allow them to be reusable so 20% EtOH was added to the solutions to
lower the contact angle. To prevent leakage, the channels were held against the substrate under
69
light pressure. After dispensing, the valve was removed. Using a 1 mL syringe, air was injected by
hand to move the droplets to the middle chamber. Mixing was aided by pressing on the middle
chamber 3 times. Using an optical microscope, a picture of the mixed sample was recorded. The
ratio of the dye was measured using ImageJ by recording the intensity of the dye and comparing
it to the intensity of a bulk solution of dye mixed using standard laboratory pipettes and injected
into the on-chip mixing chamber.
4.5.8 Antibiotic susceptibility testing
The device was fabricated in PDMS as stated above. Ampicillin in 1× PBS (pH 7.4), either 100 μg
mL-1 or 1000 μg mL-1, was introduced into the antibiotic inlet. 1, 2, 5 or 10 chambers were filled
in the standard manner. Air was then used to push the antibiotic into the growth chamber. The
bacteria metering chamber was then filled with the bacterial solution (12.5 × 106 cfu mL-1 E. coli
in LB broth with 50 mM TBS pH 8.5 and 1 mM resazurin). Air was used to push this plug of fluid
into the growth chamber where it diffusively mixed with the antibiotic. The chip was then
incubated in a 37 °C water bath (to prevent evaporation) for 8 h.
4.5.9 Fluorescence image acquisition and analysis
A Luxeon Rebel Color LED (Green, LEDSupply) was used as a light source. A collimating lens
was used to focus the light into a beam. This beam was passed through a plastic green filter
(Roscolux) to remove any extraneous wavelengths and directed onto the growth chamber of the
Digit Chip. Emitted light was passed through a 589 nm bandwidth filter (Edmund Optics) and
imaged using a LG G3 cell phone camera equipped with an adhered PDMS lens.150 Images were
captured using the free Open Camera app and image analysis was performed using ImageJ. The
red channel of the RGB image was extracted and the average intensity of the pixels in the growth
chamber was measured. Each concentration of antibiotic was performed in triplicate and values
were normalized to maximum and minimum pixel values, with error bars showing standard error.
70
Conclusions and Future Outlook
5.1 Thesis Findings
In this thesis we explored a variety of topics in device design and application towards the probing
of biology, with the ultimate purpose of shedding light on the disease state and informing potential
treatments. Multiple aspects of devices were investigated with an eye towards application at the
POC, including microfluidic innovations, improvements in electrode design towards enhanced
sensing, simplification of platforms for greater accessibility, and testing using genuine biological
samples
In chapter 2, we challenged a microfluidic device to reveal the evolution of cancer in an animal
model. This technique allowed us to probe the progression of the disease by using the blood as a
liquid biopsy, thereby identifying changes in circulating tumour cells over time. The unique nature
of the device allowed both the number of CTCs and their invasive potential to be measured
multiple times a week, giving a high-resolution glimpse into the disease course. It was revealed
that both the total number of CTCs and the fraction of low-epithelial character CTCs increased
monotonically, in step with the growth of the primary tumour. This approach also allowed us to
observe the decrease in CTC load upon removal of the primary tumor, as the source for the majority
of CTCs was resected. Perhaps most importantly, we saw a marked rebound in CTC levels at a
later time, implying the presence of metastatic disease. The distinctive capacity of the device to
discriminate between CTC subpopulations had further utility in highlighting the biological changes
enacted by tumor growth and seeding.
One of the most effective means of biosensing is the use of microelectrodes. In chapter 3, we
pursued a fundamental understanding of the mechanisms underlying gold electrodeposition,
towards the development of improved sensing performance. We discovered significant and
separate effects for each of applied voltage, concentration of gold precursor, and modulation of
71
solution viscosity. A specific combination of conditions was established which reliably produced
a high surface area electrode with an intriguing geometry. Deeper probing of the underlying
mechanisms showed how this structure was rooted in nucleation and crystal twinning events early
in electrode growth. Finely tuned electrodes showed improved performance in the electrochemical
detection of nucleic acids, indicating possible application in biosensing applications.
A further necessity in the widespread implementation of point-of-care devices is the ability to
manipulate microscopic quantities of fluid in a reliable manner while not requiring the use of an
external instrument. As a step towards this end we examined the potential of capillary valves for
the metering and dispensing of minute quantities of fluid. Following a parametric search to
elucidate the dimensions best suited for reproducible behaviour, we devised a repeated valve motif
for discrete volume measurement. The actuation was made intuitive and power-free by pairing this
motif with a small external button, each press of which triggered the bursting of a single valve.
This architecture was first used to create a concentration gradient, demonstrating a functionality
with broad applications across chemistry and biology. Next, an application for the determination
of bacterial resistance and requisite antibiotic concentration was illustrated.
5.2 Future Outlook
The desire to create new and more informative devices for probing human disease is pushing
research forward on a large number of fronts. The work performed here aims to help advance this
goal. The work in chapter 2 served as a good initial assessment of the technology for tracking
cancer progression and identifying changes in cancer biology. However, this investigation was
confined to a single animal model of disease. A natural next step is investigating a variety of cancer
types, ideally in human patients. In fact, some studies to this effect have been performed in our
lab, and reveal the utility of the technology across cancers of varied origin and invasiveness.
Moreover, there is great potential in further probing the biology of the captured cells, since the
device retains cellular viability. Together with improvements in the sensitivity and reliably of
single cell “omics” techniques, devices such as this should soon be able to give unprecedented
insight into the abnormal cellular processes at work in cancerous cells. With this information in
hand, the ability to create more targeted and effective therapies will be significantly enhanced,
paving the way for a host of new cancer drugs with reduced side effects and reduced chances of
recurrence.
72
Similarly, the development of new electrodes with new properties is important not only for its own
sake, but for the advances in POC performance that it enables. Improved understanding of the
processes dictating electrode growth during electrodeposition allows for the custom tuning of
electrode properties. By controlling surface area and crystallinity of electrodes, improved ability
to capture biomolecular targets can be achieved without requiring increases in electrode footprint.
This in turn permits the incorporation of electrodes into more compact devices, a necessity if we
are to produce surgically implantable sensors such as those required for an artificial pancreas.
Apart from POC applications, there is evidence that gold electrodes with controlled surface
morphology have the ability to efficiently catalyze the reduction of greenhouse gases into useful
fuel stocks. Further refining the existing electrode architecture towards this end would be an
exciting avenue to pursue.
The primary idea behind the Digit Chip is allowing laboratory-style operations to be performed
using small samples volumes and in an intuitive manner, so as to allow for use by individuals with
minimal training. This simplification of testing is paramount for widespread adoption in areas with
limited resources. One envisioned extension of this technology is a small handheld cartridge which
features a small number of buttons facilitating basic laboratory functions (mixing, metering,
moving, etc.). By incorporating a simple optical or electrochemical sensor into such a device, a
variety of typical biochemical tests could be performed on a single platform. Such a device should
remain self-contained and power-free, with the possible exception of a simple battery for powering
the sensor element.
In addition to the specific outlooks of each project individually, there is potential for combining
the elements of each to provide additional functionality. Microelectrodes have the potential to
probe cell metabolic activity, which could add an extra dimension of characterization to the CTC
capture device. By placing electrodes in each of the chambers it may be possible to measure the
concentrations of key metabolites and see if these correlate to the degree of epithelial character.
Similarly, incorporation of electrodes into the antibiotic susceptibility model of the Digit Chip
would allow for resazurin to be detected electrochemically. This alteration would increase
sensitivity while simultaneously obviating the need for an external smart phone. Finally,
modification of the Digit Chip would allow for the discrete movement of cells between
compartments. By releasing cells from the CTC capture device and introducing them into this cell
manipulation Digit Chip, each CTC could potentially be individually cultured and characterized.
73
Taken together, the results of this thesis represent a step forward in the production of advanced
devices for probing the disease state. Their future application will improve disease
characterization, allowing for more accurate prognoses and better informed treatment.
74
References
1. Yager, P. et al. Microfluidic diagnostic technologies for global public health. Nature 442,
412–418 (2006).
2. Kelley, S. O. New technologies for rapid bacterial identification and antibiotic resistance
profiling. SLAS Technol. Transl. Life Sci. Innov. 22, 113–121 (2017).
3. Kuhn, L. et al. Human papillomavirus DNA testing for cervical cancer screening in low-
resource settings. J. Natl. Cancer Inst. 92, 818–825 (2000).
4. McDermott, U. & Settleman, J. Personalized cancer therapy with selective kinase inhibitors:
an emerging paradigm in medical oncology. J. Clin. Oncol. 27, 5650–5659 (2009).
5. Bissonnette, L. & Bergeron, M. G. Infectious disease management through point-of-care
personalized medicine molecular diagnostic technologies. J. Pers. Med. 2, 50–70 (2012).
6. Kissinger, P. T. Biosensors—a perspective. Biosens. Bioelectron. 20, 2512–2516 (2005).
7. DeLisa, M. P. et al. Evanescent wave long-period fiber Bragg grating as an immobilized
antibody biosensor. Anal. Chem. 72, 2895–2900 (2000).
8. Yemini, M., Reches, M., Gazit, E. & Rishpon, J. Peptide nanotube-modified electrodes for
enzyme−biosensor applications. Anal. Chem. 77, 5155–5159 (2005).
9. Sassolas, A., Leca-Bouvier, B. D. & Blum, L. J. DNA biosensors and microarrays. Chem.
Rev. 108, 109–139 (2008).
10. El-Sagheer, A. H. & Brown, T. Click chemistry with DNA. Chem. Soc. Rev. 39, 1388
(2010).
11. Cunningham, B. et al. A plastic colorimetric resonant optical biosensor for multiparallel
detection of label-free biochemical interactions. Sensors Actuators B Chem. 85, 219–226
(2002).
12. Arlett, J. L., Myers, E. B. & Roukes, M. L. Comparative advantages of mechanical
biosensors. Nat. Nanotechnol. 6, 203–215 (2011).
13. Wang, J. et al. DNA electrochemical biosensor for the detection of short DNA sequences
related to the human immunodeficiency virus. Anal. Chem. 68, 2629–2634 (1996).
14. Ehrenkranz, J. R. L. Home and point-of-care pregnancy tests: a review of the technology.
Epidemiology 13, S15–S18 (2002).
15. Oliver, N. S., Toumazou, C., Cass, A. E. G. & Johnston, D. G. Glucose sensors: a review
of current and emerging technology. Diabet. Med. 26, 197–210 (2009).
16. Osteryoung, J. G. & Osteryoung, R. A. Square wave voltammetry. Anal. Chem. 57, 101A–
110A (1985).
17. Kunduraci, M., Al-Sharab, J. F. & Amatucci, G. G. High-power nanostructured LiMn2 -
xNixO4 high-voltage lithium-ion battery electrode materials: electrochemical impact of
electronic conductivity and morphology. Chem. Mater. 18, 3585–3592 (2006).
18. Pang, S.-C., Anderson, M. A. & Chapman, T. W. Novel electrode materials for thin-film
ultracapacitors: comparison of electrochemical properties of sol-gel-derived and
electrodeposited manganese dioxide. J. Electrochem. Soc. 147, 444 (2000).
75
19. Xu, S., Luo, Y. & Zhong, W. Investigation of catalytic activity of glassy carbon with
controlled crystallinity for counter electrode in dye-sensitized solar cells. Sol. Energy 85,
2826–2832 (2011).
20. Heinze, J. Ultramicroelectrodes in electrochemistry. Angew. Chemie Int. Ed. English 32,
1268–1288 (1993).
21. Li, F., Han, X. & Liu, S. Development of an electrochemical DNA biosensor with a high
sensitivity of fm by dendritic gold nanostructure modified electrode. Biosens. Bioelectron.
26, 2619–2625 (2011).
22. Li, C. W., Ciston, J. & Kanan, M. W. Electroreduction of carbon monoxide to liquid fuel
on oxide-derived nanocrystalline copper. Nature 508, 504–507 (2014).
23. Lucarelli, F., Marrazza, G., Turner, A. P. F. & Mascini, M. Carbon and gold electrodes as
electrochemical transducers for DNA hybridisation sensors. Biosens. Bioelectron. 19, 515–
530 (2004).
24. Schouten, K. J. P., Qin, Z., Pérez Gallent, E. & Koper, M. T. M. Two pathways for the
formation of ethylene in CO reduction on single-crystal copper electrodes. J. Am. Chem.
Soc. 134, 9864–9867 (2012).
25. Purcell, E. M. Life at low Reynolds number. Am. J. Phys. 45, 3–11 (1977).
26. Kamholz, A. E. & Yager, P. Theoretical analysis of molecular diffusion in pressure-driven
laminar flow in microfluidic channels. Biophys. J. 80, 155–160 (2001).
27. Inglis, D. W. Efficient microfluidic particle separation arrays. Appl. Phys. Lett. 94, 013510
(2009).
28. Lee, J. & Kim, C.-J. Surface-tension-driven microactuation based on continuous
electrowetting. J. Microelectromechanical Syst. 9, 171–180 (2000).
29. Safavieh, R. & Juncker, D. Capillarics: pre-programmed, self-powered microfluidic circuits
built from capillary elements. Lab Chip 13, 4180–4189 (2013).
30. Liu, K. et al. Simple fabrication scheme for sub-10 nm electrode gaps using electron-beam
lithography. Appl. Phys. Lett. 80, 865–867 (2002).
31. Vong, T. et al. Site-specific immobilization of DNA in glass microchannels via
photolithography. Langmuir 25, 13952–13958 (2009).
32. Xia, Y. & Whitesides, G. M. Soft lithography. Annu. Rev. Mater. Sci. 28, 153–184 (1998).
33. Bélanger, M.-C. & Marois, Y. Hemocompatibility, biocompatibility, inflammatory and in
vivo studies of primary reference materials low-density polyethylene and
polydimethylsiloxane: a review. J. Biomed. Mater. Res. 58, 467–477 (2001).
34. Bhagat, A. A. S. et al. Microfluidics for cell separation. Med. Biol. Eng. Comput. 48, 999–
1014 (2010).
35. Kang, J. H. et al. A combined micromagnetic-microfluidic device for rapid capture and
culture of rare circulating tumor cells. Lab Chip 12, 2175–2181 (2012).
36. Pratt, E. D., Huang, C., Hawkins, B. G., Gleghorn, J. P. & Kirby, B. J. Rare cell capture in
microfluidic devices. Chem. Eng. Sci. 66, 1508–1522 (2011).
37. El-Ali, J., Sorger, P. K. & Jensen, K. F. Cells on chips. Nature 442, 403–411 (2006).
76
38. Copley, M. R., Beer, P. A. & Eaves, C. J. Hematopoietic stem cell heterogeneity takes center
stage. Cell Stem Cell 10, 690–697 (2012).
39. Raj, A., Rifkin, S. A., Andersen, E. & van Oudenaarden, A. Variability in gene expression
underlies incomplete penetrance. Nature 463, 913–918 (2010).
40. Elowitz, M. B., Levine, A. J., Siggia, E. D. & Swain, P. S. Stochastic gene expression in a
single cell. Science 297, 1183–6 (2002).
41. Navin, N. et al. Tumour evolution inferred by single-cell sequencing. Nature 472, 90–94
(2011).
42. Jordan, N. V. et al. HER2 expression identifies dynamic functional states within circulating
breast cancer cells. Nature 537, 102–106 (2016).
43. Quirke, P. et al. The future of the TNM staging system in colorectal cancer: time for a
debate? Lancet Oncol. 8, 651–657 (2007).
44. Baxter, N. N. & Tuttle, T. M. Inadequacy of lymph node staging in gastric cancer patients:
a population-based study. Ann. Surg. Oncol. 12, 981–987 (2005).
45. Epstein, F. H. & Schnipper, L. E. Clinical implications of tumor-cell heterogeneity. N. Engl.
J. Med. 314, 1423–1431 (1986).
46. Junttila, M. R. & de Sauvage, F. J. Influence of tumour micro-environment heterogeneity
on therapeutic response. Nature 501, 346–354 (2013).
47. Aceto, N. et al. Circulating tumor cell clusters are oligoclonal precursors of breast cancer
metastasis. Cell 158, 1110–1122 (2014).
48. Thiery, J. P. Epithelial–mesenchymal transitions in tumour progression. Nat. Rev. Cancer
2, 442–454 (2002).
49. Heitzer, E. et al. Complex tumor genomes inferred from single circulating tumor cells by
array-CGH and next-generation sequencing. Cancer Res. 73, 2965–2975 (2013).
50. Zong, C., Lu, S., Chapman, A. R. & Xie, X. S. Genome-wide detection of single-nucleotide
and copy-number variations of a single human cell. Science 338, 1622–6 (2012).
51. Cohen, S. J. et al. Relationship of circulating tumor cells to tumor response, progression-
free survival, and overall survival in patients with metastatic colorectal cancer. J. Clin.
Oncol. 26, 3213–3221 (2008).
52. Stopeck, A. et al. Circulating tumor cells, disease progression, and survival in metastatic
breast cancer. 781–791 (2004).
53. Bidard, F.-C. et al. Single circulating tumor cell detection and overall survival in
nonmetastatic breast cancer. Ann. Oncol. 21, 729–733 (2010).
54. Chaffer, C. L. & Weinberg, R. A. A perspective on cancer cell metastasis. Science 331,
1559–64 (2011).
55. Yu, M. et al. Circulating breast tumor cells exhibit dynamic changes in epithelial and
mesenchymal composition. Science 339, 580–584 (2013).
56. Gorges, T. M. et al. Circulating tumour cells escape from epcam-based detection due to
epithelial-to-mesenchymal transition. BMC Cancer 12, 178 (2012).
77
57. Tsukamoto, H. et al. Irradiation-induced epithelial–mesenchymal transition (EMT) related
to invasive potential in endometrial carcinoma cells. Gynecol. Oncol. 107, 500–504 (2007).
58. Nagrath, S. et al. Isolation of rare circulating tumour cells in cancer patients by microchip
technology. Nature 450, 1235–1239 (2007).
59. Nichols, A. C. et al. Detection of circulating tumor cells in advanced head and neck cancer
using the Cellsearch system. Head Neck 34, 1440–1444 (2012).
60. Rao, C. et al. Circulating melanoma cells and survival in metastatic melanoma. Int. J. Oncol.
38, 755–760 (2011).
61. Bidard, F. C. et al. Disseminated and circulating tumor cells in gastrointestinal oncology.
Crit. Rev. Oncol. Hematol. 82, 103–115 (2012).
62. Naoe, M. et al. Detection of circulating urothelial cancer cells in the blood using the
Cellsearch system. Cancer 109, 1439–1445 (2007).
63. Mohamadi, R. M. et al. Nanoparticle-mediated binning and profiling of heterogeneous
circulating tumor cell subpopulations. Angew. Chemie Int. Ed. 54, 139–143 (2015).
64. Besant, J. D. et al. Velocity valleys enable efficient capture and spatial sorting of
nanoparticle-bound cancer cells. Nanoscale 7, 6278–6285 (2015).
65. Cristofanilli, M. et al. Circulating tumor cells in metastatic breast cancer: biologic staging
beyond tumor burden. Clin. Breast Cancer 7, 34–42 (2007).
66. Alix-Panabieres, C. & Pantel, K. Circulating tumor cells: liquid biopsy of cancer. Clin.
Chem. 59, 110–118 (2013).
67. EBCTCG (Early Breast Cancer Trialists’ Collaborative Group) et al. Effect of radiotherapy
after mastectomy and axillary surgery on 10-year recurrence and 20-year breast cancer
mortality: meta-analysis of individual patient data for 8135 women in 22 randomised trials.
Lancet (London, England) 383, 2127–35 (2014).
68. Baum, B., Settleman, J. & Quinlan, M. P. Transitions between epithelial and mesenchymal
states in development and disease. Semin. Cell Dev. Biol. 19, 294–308 (2008).
69. Hiltermann, T. J. N. et al. Circulating tumor cells in small-cell lung cancer: a predictive and
prognostic factor. Ann. Oncol. 23, 2937–2942 (2012).
70. Steinert, G. et al. Immune escape and survival mechanisms in circulating tumor cells of
colorectal cancer. Cancer Res. 74, 1694–1704 (2014).
71. Wang, D. & Bodovitz, S. Single cell analysis: the new frontier in ‘omics’. Trends
Biotechnol. 28, 281–290 (2010).
72. Joosse, S. A. & Pantel, K. Biologic challenges in the detection of circulating tumor cells.
Cancer Res. 73, 8–11 (2013).
73. Valiev, R. Materials science: nanomaterial advantage. Nature 419, 887–889 (2002).
74. Wang, J. Nanomaterial-based electrochemical biosensors. Analyst 130, 421 (2005).
75. Rose, M. et al. Hierarchical micro- and mesoporous carbide-derived carbon as a high-
performance electrode material in supercapacitors. Small 7, 1108–1117 (2011).
76. Gittens, R. A. et al. The roles of titanium surface micro/nanotopography and wettability on
78
the differential response of human osteoblast lineage cells. Acta Biomater. 9, 6268–6277
(2013).
77. Potyrailo, R. & Naik, R. R. Bionanomaterials and bioinspired nanostructures for selective
vapor sensing. Annu. Rev. Mater. Res. 43, 307–334 (2013).
78. Noorduin, W. L., Grinthal, A., Mahadevan, L. & Aizenberg, J. Rationally designed
complex, hierarchical microarchitectures. Science 340, 832–7 (2013).
79. Grinthal, A. et al. Steering nanofibers: an integrative approach to bio-inspired fiber
fabrication and assembly. Nano Today 7, 35–52 (2012).
80. Phillips, K. R. et al. Tunable anisotropy in inverse opals and emerging optical properties.
Chem. Mater. 26, 1622–1628 (2014).
81. Akdim, B., Pachter, R. & Naik, R. R. Self-assembled peptide nanotubes as electronic
materials: an evaluation from first-principles calculations. Appl. Phys. Lett. 106, 183707
(2015).
82. Pacholski, C., Kornowski, A. & Weller, H. Self-assembly of ZnO: from nanodots to
nanorods. Angew. Chemie - Int. Ed. 41, 1188–1191 (2002).
83. Whitesides, G. M. & Grzybowski, B. Self-assembly at all scales. Science 295, 2418–21
(2002).
84. Jasuja, K. & Berry, V. Implantation and growth of dendritic gold nanostructures on
graphene derivatives: electrical property tailoring and raman enhancement. ACS Nano 3,
2358–2366 (2009).
85. Grzelczak, M., Pérez-Juste, J., Mulvaney, P. & Liz-Marzán, L. M. Shape control in gold
nanoparticle synthesis. Chem. Soc. Rev. 37, 1783–1791 (2008).
86. Woo, S. M., Gabardo, C. M. & Soleymani, L. Prototyping of wrinkled nano-
/microstructured electrodes for electrochemical DNA detection. Anal. Chem. 86, 12341–
12347 (2014).
87. Hashmi, A. S. K. & Hutchings, G. J. Gold catalysis. Angew. Chemie Int. Ed. 45, 7896–7936
(2006).
88. Liu, M. et al. Enhanced electrocatalytic CO2 reduction via field-induced reagent
concentration. Nature 537, 382–386 (2016).
89. Campbell, C. T. The active site in nanoparticle gold catalysis. Science 306, 234–235 (2004).
90. Ming, T. et al. Growth of tetrahexahedral gold nanocrystals with high-index facets. J. Am.
Chem. Soc. 131, 16350–16351 (2009).
91. Golan, Y., Margulis, L. & Rubinstein, I. Vacuum-deposited gold films. Surf. Sci. 264, 312–
326 (1992).
92. Lin, Y., Zou, Y., Mo, Y., Guo, J. & Lindquist, R. G. E-beam patterned gold nanodot arrays
on optical fiber tips for localized surface plasmon resonance biochemical sensing. Sensors
10, 9397–9406 (2010).
93. Zhao, P., Li, N. & Astruc, D. State of the art in gold nanoparticle synthesis. Coord. Chem.
Rev. 257, 638–665 (2013).
94. Cherevko, S. & Chung, C.-H. Direct electrodeposition of nanoporous gold with controlled
79
multimodal pore size distribution. Electrochem. commun. 13, 16–19 (2011).
95. Soleymani, L., Fang, Z., Sargent, E. H. & Kelley, S. O. Programming the detection limits
of biosensors through controlled nanostructuring. Nat. Nanotechnol. 4, 844–848 (2009).
96. Soleymani, L. et al. Hierarchical nanotextured microelectrodes overcome the molecular
transport barrier to achieve rapid, direct bacterial detection. ACS Nano 5, 3360–3366
(2011).
97. Soleymani, L. et al. Nanostructuring of patterned microelectrodes to enhance the sensitivity
of electrochemical nucleic acids detection. Angew. Chemie Int. Ed. 48, 8457–8460 (2009).
98. Bin, X., Sargent, E. H. & Kelley, S. O. Nanostructuring of sensors determines the efficiency
of biomolecular capture. Anal. Chem. 82, 5928–5931 (2010).
99. Fang, Z. et al. Direct profiling of cancer biomarkers in tumor tissue using a multiplexed
nanostructured microelectrode integrated circuit. ACS Nano 3, 3207–3213 (2009).
100. Sage, A. T. et al. Fractal circuit sensors enable rapid quantification of biomarkers for donor
lung assessment for transplantation. Sci. Adv. 1, e1500417 (2015).
101. Scharifker, B. & Hills, G. Theoretical and experimental studies of multiple nucleation.
Electrochem. Acta 28, 879–889 (1983).
102. Scharifker, B. R. On the theory of the potentiostatic current transient for diffusion-
controlled three-dimensional electrocrystallization processes. J. Electrochem. Soc. 146,
1005–1012 (1999).
103. Hyde, M. E. & Compton, R. G. A review of the analysis of multiple nucleation with
diffusion controlled growth. J. Electroanal. Chem. 549, 1–12 (2003).
104. Johnson, C. J., Dujardin, E., Davis, S. A., Murphy, C. J. & Mann, S. Growth and form of
gold nanorods prepared by seed-mediated, surfactant-directed synthesis. J. Mater. Chem.
12, 1765–1770 (2002).
105. Ghahremaninezhad, A. & Dolati, A. Diffusion-controlled growth model for
electrodeposited cobalt nanowires in highly ordered aluminum oxide membrane. in ECS
Transactions 28, 13–25 (The Electrochemical Society, 2010).
106. Xia, Y., Xiong, Y., Lim, B. & Skrabalak, S. E. Shape-controlled synthesis of metal
nanocrystals: simple chemistry meets complex physics? Angew. Chemie Int. Ed. 48, 60–
103 (2009).
107. Das, J. & Kelley, S. O. Protein detection using arrayed microsensor chips: tuning sensor
footprint to achieve ultrasensitive readout of CA-125 in serum and whole blood. Anal.
Chem. 83, 1167–1172 (2011).
108. Mahshid, S. S., Camiré, S., Ricci, F. & Vallée-Bélisle, A. A highly selective electrochemical
DNA-based sensor that employs steric hindrance effects to detect proteins directly in whole
blood. J. Am. Chem. Soc. 137, 15596–15599 (2015).
109. Peterson, A. W. The effect of surface probe density on DNA hybridization. Nucleic Acids
Res. 29, 5163–5168 (2001).
110. Nakano, S., Miyoshi, D. & Sugimoto, N. Effects of molecular crowding on the structures,
interactions, and functions of nucleic acids. Chem. Rev. 114, 2733–2758 (2014).
80
111. Yager, P. et al. Microfluidic diagnostic technologies for global public health. Nature 442,
412–418 (2006).
112. Kelley, S. O. et al. Advancing the speed, sensitivity and accuracy of biomolecular detection
using multi-length-scale engineering. Nat. Nanotechnol. 9, 969–980 (2014).
113. Lutz, B. et al. Dissolvable fluidic time delays for programming multi-step assays in
instrument-free paper diagnostics. Lab Chip 13, 2840–2847 (2013).
114. Renault, C., Koehne, J., Ricco, A. J. & Crooks, R. M. Three-dimensional wax patterning of
paper fluidic devices. Langmuir 30, 7030–7036 (2014).
115. Liu, H. & Crooks, R. M. Three-dimensional paper microfluidic devices assembled using the
principles of origami. J. Am. Chem. Soc. 133, 17564–17566 (2011).
116. Laksanasopin, T. et al. A smartphone dongle for diagnosis of infectious diseases at the point
of care. Sci. Transl. Med. 7, 273re1 (2015).
117. Dimov, I. K. et al. Stand-alone self-powered integrated microfluidic blood analysis system
(simbas). Lab Chip 11, 845–850 (2011).
118. Gong, M. M., MacDonald, B. D., Vu Nguyen, T. & Sinton, D. Hand-powered microfluidics:
a membrane pump with a patient-to-chip syringe interface. Biomicrofluidics 6, 044102
(2012).
119. Begolo, S., Zhukov, D. V, Selck, D. a, Li, L. & Ismagilov, R. F. The pumping lid:
investigating multi-material 3d printing for equipment-free, programmable generation of
positive and negative pressures for microfluidic applications. Lab Chip 14, 4616–4628
(2014).
120. Safavieh, R., Zhou, G. Z. & Juncker, D. Microfluidics made of yarns and knots: from
fundamental properties to simple networks and operations. Lab Chip 11, 2618–2624 (2011).
121. Safavieh, R., Tamayol, A. & Juncker, D. Serpentine and leading-edge capillary pumps for
microfluidic capillary systems. Microfluid. Nanofluidics 18, 357–366 (2015).
122. Yetisen, A. K., Akram, M. S. & Lowe, C. R. Paper-based microfluidic point-of-care
diagnostic devices. Lab Chip 13, 2210–2251 (2013).
123. Thom, N. K., Lewis, G. G., Yeung, K. & Phillips, S. T. Quantitative fluorescence assays
using a self-powered paper-based microfluidic device and a camera-equipped cellular
phone. RSC Adv. 4, 1334–1340 (2014).
124. Funes-Huacca, M. et al. Portable self-contained cultures for phage and bacteria made of
paper and tape. Lab Chip 12, 4269–4278 (2012).
125. Deiss, F., Funes-Huacca, M. E., Bal, J., Tjhung, K. F. & Derda, R. Antimicrobial
susceptibility assays in paper-based portable culture devices. Lab Chip 14, 167–171 (2014).
126. Luo, L., Li, X. & Crooks, R. M. Low-voltage origami-paper-based electrophoretic device
for rapid protein separation. Anal. Chem. 86, 12390–12397 (2014).
127. Guo, M. T., Rotem, A., Heyman, J. A. & Weitz, D. A. Droplet microfluidics for high-
throughput biological assays. Lab Chip 12, 2146–2155 (2012).
128. Rosenfeld, L., Lin, T., Derda, R. & Tang, S. K. Y. Review and analysis of performance
metrics of droplet microfluidics systems. Microfluid. Nanofluidics 16, 921–939 (2014).
81
129. Kim, M. et al. Optofluidic ultrahigh-throughput detection of fluorescent drops. Lab Chip
15, 1417–1423 (2015).
130. Ng, A. H. C., Chamberlain, M. D., Situ, H., Lee, V. & Wheeler, A. R. Digital microfluidic
immunocytochemistry in single cells. Nat. Commun. 6, 7513 (2015).
131. Eydelnant, I. A., Betty Li, B. & Wheeler, A. R. Microgels on-demand. Nat. Commun. 5,
3355 (2014).
132. Witters, D., Knez, K., Ceyssens, F., Puers, R. & Lammertyn, J. Digital microfluidics-
enabled single-molecule detection by printing and sealing single magnetic beads in
femtoliter droplets. Lab Chip 13, 2047–2054 (2013).
133. Puntambekar, A., Choi, J.-W., Ahn, C. H., Kim, S. & Makhijani, V. Fixed-volume metering
microdispenser module. Lab Chip 2, 213–218 (2002).
134. Hitzbleck, M. et al. Capillary soft valves for microfluidics. Lab Chip 12, 1972–1978 (2012).
135. Iwai, K. et al. Finger-powered microfluidic systems using multilayer soft lithography and
injection molding processes. Lab Chip 14, 3790–3799 (2014).
136. Shen, F. et al. Multiplexed quantification of nucleic acids with large dynamic range using
multivolume digital RT-PCR on a rotational slipchip tested with hiv and hepatitis C viral
load. J. Am. Chem. Soc. 133, 17705–17712 (2011).
137. Shen, F. et al. Nanoliter multiplex pcr arrays on a slipchip. Anal. Chem. 82, 4606–4612
(2010).
138. Pompano, R. R., Platt, C. E., Karymov, M. A. & Ismagilov, R. F. Control of initiation, rate,
and routing of spontaneous capillary-driven flow of liquid droplets through microfluidic
channels on slipchip. Langmuir 28, 1931–1941 (2012).
139. Liu, H., Li, X. & Crooks, R. M. Paper-based slippad for high-throughput chemical sensing.
Anal. Chem. 85, 4263–4267 (2013).
140. Li, W. et al. Squeeze-chip: a finger-controlled microfluidic flow network device and its
application to biochemical assays. Lab Chip 12, 1587–1590 (2012).
141. Xu, K., Begley, M. R. & Landers, J. P. Simultaneous metering and dispensing of multiple
reagents on a passively controlled microdevice solely by finger pressing. Lab Chip 15, 867–
876 (2015).
142. Leng, J. & Salmon, J.-B. Microfluidic crystallization. Lab Chip 9, 24–34 (2009).
143. Li Jeon, N. et al. Neutrophil chemotaxis in linear and complex gradients of interleukin-8
formed in a microfabricated device. Nat. Biotechnol. 20, 826 (2002).
144. Cho, H., Kim, H.-Y., Kang, J. Y. & Kim, T. S. How the capillary burst microvalve works.
J. Colloid Interface Sci. 306, 379–385 (2007).
145. Elizabeth Hulme, S., Shevkoplyas, S. S. & Whitesides, G. M. Incorporation of prefabricated
screw, pneumatic, and solenoid valves into microfluidic devices. Lab Chip 9, 79–86 (2009).
146. Cira, N. J., Ho, J. Y., Dueck, M. E. & Weibel, D. B. A self-loading microfluidic device for
determining the minimum inhibitory concentration of antibiotics. Lab Chip 12, 1052–1059
(2012).
147. Greenwood, D. & O’Grady, F. A comparison of the effects of ampicillin on escherichia coli
82
and proteus mirabilis. J. Med. Microbiol. 2, 435–441 (1969).
148. Kitagawa, M. et al. Complete set of ORF clones of escherichia coli ASKA library (a
complete set of e. coli k-12 ORF archive): unique resources for biological research. DNA
Res. 12, 291–299 (2006).
149. Chin, C. D., Linder, V. & Sia, S. K. Commercialization of microfluidic point-of-care
diagnostic devices. Lab Chip 12, 2118–2134 (2012).
150. Sung, Y.-L., Jeang, J., Lee, C.-H. & Shih, W.-C. Fabricating optical lenses by inkjet printing
and heat-assisted in situ curing of polydimethylsiloxane for smartphone microscopy. 20,
47005–47006 (2015).
151. Papalexi, E. & Satija, R. Single-cell RNA sequencing to explore immune cell heterogeneity.
Nat. Rev. Immunol. 18, 35–45 (2017).
152. Narsinh, K. H. et al. Single cell transcriptional profiling reveals heterogeneity of human
induced pluripotent stem cells. J. Clin. Invest. 121, 1217–1221 (2011).
153. Chattopadhyay, P. K., Gierahn, T. M., Roederer, M. & Love, J. C. Single-cell technologies
for monitoring immune systems. Nat. Immunol. 15, 128–135 (2014).
154. Roeder, I., Loeffler, M. & Glauche, I. Towards a quantitative understanding of stem cell–
niche interaction: experiments, models, and technologies. Blood Cells, Mol. Dis. 46, 308–
317 (2011).
155. Jordan, N. V. et al. HER2 expression identifies dynamic functional states within circulating
breast cancer cells. Nature 537, 102–106 (2016).
156. Brouzes, E. et al. Droplet microfluidic technology for single-cell high-throughput
screening. Proc. Natl. Acad. Sci. 106, 14195–14200 (2009).
157. Lagus, T. P. & Edd, J. F. A review of the theory, methods and recent applications of high-
throughput single-cell droplet microfluidics. J. Phys. D. Appl. Phys. 46, 114005 (2013).
158. Chung, J., Kim, Y.-J. & Yoon, E. Highly-efficient single-cell capture in microfluidic array
chips using differential hydrodynamic guiding structures. Appl. Phys. Lett. 98, 123701
(2011).
159. Lin, C.-H. et al. A microfluidic dual-well device for high-throughput single-cell capture and
culture. Lab Chip 15, 2928–2938 (2015).
160. Miyanishi, K. et al. Effects of hydrostatic pressure and transforming growth factor- β 3 on
adult human mesenchymal stem cell chondrogenesis in vitro. Tissue Eng. 12, 1419–1428
(2006).
161. Pertoft, H. & Laurent, T. C. Isopycnic separation of cells and cell organelles by
centrifugation in modified colloidal silica gradients. in Methods of Cell Separation 25–65
(Springer US, 1977). doi:10.1007/978-1-4684-0820-1_2
162. Yamada, M. & Seki, M. Hydrodynamic filtration for on-chip particle concentration and
classification utilizing microfluidics. Lab Chip 5, 1233–1239 (2005).
163. Yamada, M. & Seki, M. Microfluidic particle sorter employing flow splitting and
recombining. Anal. Chem. 78, 1357–1362 (2006).
83
Single Cell Capture Device
A.1 Background
It was historically the practice in cell biology to treat all of the cells from a particular tissue or
culture as being essentially identical. Part of this paradigm was rooted in the poor sensitivity of
early techniques, which often required that large quantities of cells be pooled and lysed in order to
detect biomolecules of interest. In this approach, the differences in genotype or phenotype within
these populations were ignored or treated as noise obscuring the desired information. However,
more recent investigations have shed light on the fundamental importance of this heterogeneity in
diverse biological processes.151–153 These variations stem from a number of different sources. One
of the most fundamental is small changes in the genotypes of cells due to unique mutations.41 Other
differences are a consequence of the history of cell, including the particular niche it inhabits and
the stimuli it has been exposed to.154 More transitory discrepancies arise as a consequence of the
fundamentally stochastic nature of molecular interactions, especially in the case of regulatory
proteins whose binding prompt a cascade of effects.40 In general, these differences are of greatest
import in scenarios where a small number of cells can have a disproportionately large impact. This
is the case during embryonic development, stem cell development and tumor biology.39,155
The desire to more fully characterize this cellular heterogeneity has motivated the development of
assays which allow for the analysis of protein, mRNA, DNA, and various metabolites at a single
cell level. Consequently, there exists a need for devices that allow for the precise manipulation of
single cells: specifically, the ability to isolated, treat, incubate and release cells in an individually
addressable manner. These devices can roughly be divided into single-phase and multi-phase
types. Multi-phase devices employ droplets of one medium immersed in a second medium.156,157
By trapping individual cells in these droplets, cells can then be manipulated using the full suite of
droplet fluidic techniques. Single-phase devices don’t isolate cells in different media but instead
84
use physical obstacle or force fields to keep cells apart.158,159 An ideal device would be as simple
as possible while still supporting all of the desired functions; this suggest the use of single phase
device which employs primarily passive elements.
A.2 Device Design
The first stage of any cell manipulation device must accept a stream of cells and individually
position them into isolated traps. There are two methods by which this can be done: the stochastic
“shotgun” approach and the deterministic approach. In the stochastic approach, a large number of
traps are used and the design depends on the probability that a cell will encounter a trap at some
point in its path through the device. In this case, which trap the cell encounters is largely random.
By contrast, the deterministic approach involves a series of traps such that the first cell to enter the
device is directed towards the first trap, the next towards the subsequent trap, and so on. Each
method has its advantages. As compared with the deterministic approach, shotgun approach is
typically capable of considerably higher throughput and needs less stringent design. However, this
approach may allow cells to escape and does not preserve the identity of the cell. The identity of
the cell is important if the incoming cells are being delivered from an upstream measurement, e.g.
magnetic or fluorescent sorting; maintaining cell identity will allow the measured properties to
remain associated with the correct cell. Moreover, such designs typically require more traps than
the number of cells, which increases the difficulty of addressing particular captive cells. Both of
these issues are accentuated in situations where the number of cells being analyzed is small, such
as in rare cell applications.38,39
To preserve the ability of the device to work with diverse cell populations and circumvent the need
for cell tagging/external force fields, the decision was made to capture cells purely on the basis of
physical obstruction. The structure which allows for this physical capture is the microfluidic weir
(Figure A-1).
85
Figure A-1 Structure of a microfluidic weir. The channel abruptly narrows to a diameter too
small for cell passage.
A weir is a simply an abrupt narrowing of the microfluidic channel that does not allow the passage
of particles above a certain diameter. Once a cell encounters the weir it is lodged in place, held
trapped by the pressure difference between the regions upstream and downstream of the
obstruction. It is important that the pressure the cell is exposed to be kept to a minimum both in
magnitude and duration, since exposure to high pressures can modify cell biology or damage cell
integrity.160 In order to capture multiple cells, a series of weir structures is required. However, the
structures cannot simply be chained together, since the filling of the first weir arrests flow and
subsequent weirs would remain unfilled. Instead, a shunt structure is included in parallel to each
weir, which allows later cells to circumvent filled traps (Figure A-2).
86
Figure A-2 A microfluidic weir with a parallel shunt channel. The lower channel includes a
weir structure, whereas the upper shunt is a simple wide channel.
The inclusion of a shunt, however, introduces a further complication. Even when approaching an
empty weir, there is a possibility that the cell will enter the shunt channel rather than the weir
channel. The path that a cell follows is determined by the streamline that the cell occupies, as well
as the relative resistances of the two channels. Specifically, the fraction of the flow that enters each
fork is inversely proportional to the resistance of that fork (Figure A-3).
87
Figure A-3 Equivalent circuit diagram for a sequence of weir-shunt pairs. At each fork the
cell will be guided to either the weir or the shunt depending on flow resistance. The lower the
relative resistance of each channel, the greater portion of the flow that enters it.
If the centre of mass of a cell occupies one of the streamlines that enters a given channel, the cell
will do so as well. Thus, to maximize the chance that a cell enters an empty weir instead of
diverting around it, the resistance of the weir relative to the shunt must be minimized. Since the
weir must be small enough to occlude the cell, it must be smaller than the size of the cell to be
captured (on the order of 10 μm). Thus, the resistance of the weir is bounded by a minimum value
and is not freely adjustable. Consequently, the resistance of the shunt must be increased in order
to encourage cell flow into the weir (Figure A-4).
Figure A-4 Encouraging flow to enter the weir. A distribution of cells across the channel is
shown, with each cell in a different colour. The resistance of the shunt must be much greater than
RWEIR
RSHUNT
RWEIR
RSHUNT
RWEIR
RSHUNT
RWEIR
RSHUNT
RW
RS >> RW
88
that of weir to direct the majority of the cells into the capture area. RS is shunt resistance, RW is
weir resistance.
Specifically, the portion of flow that enters the weir channel, QW, is:
𝑄𝑊 =
𝑅𝑠
𝑅𝑠 + 𝑅𝑊 , (A.1)
Where RS is the shunt resistance and RW is the weir resistance. A shunt resistance 10-fold that of
the weir ensures that approximately 90% of fluid approaching an empty weir will enter it rather
than diverting around it. Assuming that cells are roughly evenly distributed within the flow, a
matching 90% will enter the trap. Increasing the resistance of the shunt is not without consequence,
however. Once a weir is occupied, the pressure difference that a cell is exposed to is the product
of the volumetric flow and the shunt resistance. In order to keep the pressure on the cell low, the
volumetric flow would thus need be kept minimal, which severely limits device throughput. The
only way to solve this problem is to adjust the cell distribution across the streamlines such that it
is no longer random. Specifically, by moving the cells as close to the wall proximal the weir as
possible, a large fraction of the cells can enter the weir despite only a small fraction of flow doing
so (Figure A-5).
Figure A-5 Reducing shunt resistance by translating cells across streamlines. By driving the
cells to one lateral edge, only a small amount of flow is required to enter the weir, allowing a
RW
RS << RW
89
dramatic reduction in shunt resistance. Shunt resistance is now much smaller than weir
resistance, instead of the converse.
This allows for greatly improved performance, and is limited only by the size of the cell.
Specifically, the cell cannot lie on any streamline which is closer to the wall than the cell radius.
By integrating all of the flow which lies between the wall and this streamline of nearest approach,
the fraction of flow which must enter the weir to ensure complete capture (assuming cells have
been fully aligned to the wall) can be determined. This quantity is further reduced by the parabolic
nature of flow in a channel. Since flow is slowest near the wall, the fraction of flow which lies
between the wall and a distance x from the wall is even less than the ratio of x to the entire width
of the channel. Employing a channel width of 200 μm and a cell diameter of 15 μm, this means
that only 1-2% of the flow is required to enter the weir to ensure complete capture. Consequently,
the resistance of the shunt can be decreased to less than 1/10 the resistance of the weir, a more than
two order of magnitude reduction in the pressure that captured cells experience. This modification
is crucial for a sequential series of traps to capture cells in a deterministic fashion while also
retaining high throughput.
Moving cells across streamlines is especially difficult in the microfluidic regime, where fluid
lamina do not mix and tend to maintain their relative positioning. Nonetheless, it is possible, and
can be performed either by an externally generated force field or by passive structures within the
device. Candidate external forces include electrophoretic, dielectrophoretic, magnetic, and
gravitational forces. Of these, gravity is the simplest, since it is omnipresent and does not need to
be generated manually. The displacement of cells due to gravity is known as cell settling. Cell
settling is a consequence of the mismatch in densities between a cell and a typical aqueous solution.
In general, a cell is a few percent more dense than water, due to the various biomolecules contained
within.161 Consequentially, the gravitational force on a cell is slightly greater than the buoyancy
force, and cells will tend to settle downwards. The velocity with which a cell settles is determined
by the interaction between this net gravitational force and the drag force, and is calculated here:
𝑣𝑠𝑒𝑡𝑡𝑙𝑖𝑛𝑔 =
𝑔𝑑2(𝜌𝑐 − 𝜌)
18𝜇, (A.2)
where g is the acceleration due to gravity, ρC is the density of the cell, ρ is the density of the fluid,
and μ is the dynamic viscosity of the fluid. For a typical cell, this velocity is on the order of 1-10
90
μm/s. Thus, for a cell to traverse a 200 μm wide channel, it must remain in this channel for a time
on the order of 1 minute. The time that a cell spends in this alignment channel is a function of three
parameters: the depth of the channel, the length of the channel, and the volumetric flow rate. More
specifically, the residence time is proportional to both the length and depth of the alignment
channel, and inversely proportional to the flow rate. In order to maintain as high a volumetric flow
rate as possible to enable good throughput, the depth and length of the channel should be
maximized. The height of the channel is constrained to some degree by the height and aspect ratio
of the SU8, with 200 μm a suitable dimension. In order to maintain a relatively rapid 10 μL/min
flowrate with a height and depth of 200 μm and a residence time of one minute, the length of the
channel needs to be 25 cm! This is prohibitively long for a single, straight channel. By instead
using a serpentine channel, the effective length of the alignment channel can be increased without
a significant change in device footprint. This produces an additional issue; the settling in each
return channel is equal and opposite that in each forward channel, and as a consequence there is
no net cell displacement. This can be fixed by reducing the depth of the return channels. By
reducing the height of these channels to 50 μm, the time spent therein is one quarter the time spent
in the forward channels, and the net cell displacement remains in the downward direction. Using
8 of these switchback channels, each 50 mm in length, was shown to be sufficient to move all cells
to the intended wall, and allow for capture using the sequential weir traps (Figure A-6).
Figure A-6 Design of the gravity driven cell alignment and capture device, which is
operated in the vertical orientation. The green channel has a depth of 200 μm and the red
channel has a depth of 50 μm to allow for net displacement due to gravity. Aligned cells then
encounter a series of 8 weir-shunt pairs.
This approach allows for very simple and robust fabrication of the alignment portion of the device.
However, it does present a number of drawbacks. Firstly, the device must be oriented in the vertical
direction during cell capture, which complicates imaging. Secondly, and more importantly, the
91
cells spend a large amount of time near the wall. This is especially true of cells which happened to
enter the device proximal to this wall. This can cause cells to slow down significantly and requires
that the wall be pretreated with surfactants to discourage sticking.
An alternative approach that uses passive fluidic design in place of gravity was implemented to
mitigate these issues. In general, channel design in the laminar regime is incapable of diverting
cells due to the tendency of streamlines to retain their relative position. Although this is true, it
omits an important fact which can be exploited to great effect; namely, the finite size of cells. Since
cells have a certain radius (on the order of 5 – 10 μm), they cannot approach any wall more closely
than this limiting distance. This is not true of the fluid flow as a whole, which consists of essentially
infinitesimal water molecules. One design motif that takes advantage of this is the “bleeding flow”
element (Figure A-7).162
Figure A-7 Function of the bleeding flow motif. (Top) The fluid located between the dashed
line and the wall exits through the side channel, with the remaining fluid remaining in the main
channel and migrating outwardly to replace the removed volume. As long as the centre of a
particle lies medial to the dashed line, it will remain in the main channel. (Bottom) Resistances
need to be determined such that the shaded portion of flow exits at each branch point. Adapted
with permission from 162. Copyright 2005 The Royal Society of Chemistry.
92
This consists of a large primary channel through which particles are flowing, and from which small
quantities of fluid are siphoned off by a series of smaller “bleeder” channels. By judicious selection
of the resistance of these channels, small volumes of fluid are diverted at each junction. As long
as these quantities are small enough, only streamlines which lie between the wall and a distance
of one cell radius will be diverted. This means that cells which are not yet touching the wall are
moved closer to it, whereas cell which are already touching the wall simply remain touching it.
With enough of these channels, most of the cells in a channel will eventually come in contact with
the wall. The asymptotic nature of this approach means, however, that cells which begin at the far
side of the channel will tend not to migrate the entire distance. This can be rectified by using two
of these motifs in series, with a slight modification (Figure A-8).163
Figure A-8 Two stage cell aligner. The device consists of two sequential elements each
employing the “bleeding flow” motif. Each coloured region is expanded in a separate figure.
93
Figure A-9 First “bleeding flow” element. Each side channel draws a small amount of fluid,
causing the cell to migrate closer to the wall.
Figure A-10 End of the first “bleeding flow” element. The cell has approached the right wall,
so it remains near the bottom of the wider channel.
94
Figure A-11 Reintroduction of removed fluid. The fluid siphoned off in the first element is
returned. As long as at least half of the total flow was siphoned, this creates a buffer (shown by
blue dotted line) which ensures the cell will be leftward of the centre line when entering the
second “bleeding flow” element.
95
Figure A-12 Second “bleeding flow” element. Since the cell begins on the left half of the
channel, it quickly migrates the remaining distance to the left wall.
In this case, the flow siphoned off during the first element is reintroduced between the first and
second elements. As long as at least half of the flow had been siphoned, this reintroduced fluid
pushes all cells to one side of the channel centerline.
From this point, a second motif can be used to sequester cells along one wall. The design of such
a device is made possible by modelling the channels as resistors and simulating flow in analogy
with an electric circuit. The device was designed using a custom Python script which determined
the requisite dimensions of the channels while keeping in accordance with manually defined
parameters for SU8 tolerances (max/min height, max/min height, max aspect ratio). Testing of this
device demonstrated that cells are regularly moved to the intended wall and remain there. Upon
interfacing of this alignment module with the shunt and weir capture module (Figure A-13), it was
determined that cells are indeed moved to the intended streamlines and enter the weirs rather than
bypassing them via the shunts (Figure A-14).
96
Figure A-13 The cell alignment module interfaced with the weir and shunt capture module.
Figure A-14 Single cell captured in weir trap.
The device as outlined is capable of accepting a stream of cells and directing them, in single file,
to a series of traps. Additional features are required in order to allow for targeted introduction of
chemicals, isolated incubation, and addressable release. These features all require the introduction
of non-passive elements into the design. To this end, membrane deflection valves were
incorporated (Figure A-15, Figure A-16)
97
Figure A-15 Photograph of integrated devce including membrane deflection valves. Scale
bar is 1 cm.
Figure A-16 Complete design of integrated devce. The lower region performs the cell
alignment feature while the upper region performs cell capture and addressing. White channels
are 25 μm tall with rectangular cross section for flow, green channels are 25 μm tall with
rounded cross section for valves, and blue channels are control lines for opening and closing
valves.
98
These valves add a second channel which lies parallel to, and vertically offset from, the flow
channels, a so-called “control” channel. The two channels are separated by a thin PDMS
membrane. Upon application of pressure to the control channel this membrane deflects vertically
and blocks passage through the flow channel. A binary-tree channel, hereafter called the
addressing channel, was introduced to allow individual access to specified traps. This addressing
channel is controlled by 6 valves that are arranged to allow binary addressing. Specifically, these
channels comprise a multiplexer which accepts 6 inputs and outputs one of the eight traps. These
inputs are not independent; rather, they are arranged in complementary pairs (A and ~A, B and ~B
and C and ~C), such that the pressure state of any valve should be opposite its complement. This
means that the states of A, B and C uniquely determine which trap is addressed, as shown in Table
A-1.
Table A-1 Truth table for the multiplexer.
A B C Selected Trap #
0 0 0 1
0 0 1 2
0 1 0 3
0 1 1 4
1 0 0 5
1 0 1 6
1 1 0 7
1 1 1 8
These control lines were actuated using a pressure manifold outfitted with three separate 4-way
valves. These valves each supplied a pair of complementary lines, and ensured that the pair
remained in opposite pressurization (Figure A-17). These lines pressurize water reservoirs which
in turn actuate the on-chip valves (Figure A-18).
99
Figure A-17 Pressure manifold with three 4-way valves (left) and two 3-way valves (right).
The 4-way valves control the addressing channel, the 3-way valves control the device mode.
Figure A-18 Pressure manifold feeding pressurized water reservoirs.
100
Two additional valve lines were required to allow for different device modes (capture, treatment,
incubation) to be accessed. One valve blocks vertical flow and must be in the closed position
during capture and incubation, and in the open position during addressable release. The other valve
blocks horizontal flow and must be closed during addressable release and incubation, and open
during capture. These valves increase the total number of valves to 8, and are also actuated by the
external pressure manifold (Figure A-19).
Figure A-19 Cell capture region of the devce including all 8 valves.. The addressable channel
at the bottom branches in a binary-tree manner into each of the eight weir capture sites above.
The lower six blue channels form the multiplexer, with the two upper blue channels controlling
the mode of operation.
The possible device modes are illustrated below, along with the states of the valves required to
enable each.
101
Figure A-20 A single trap in capture mode. Vertical flow is blocked (blue superposed on
green) whereas horizontal into the weir is unblocked (Green superposed on blue). Cells enter the
weir channel and are captured.
102
Figure A-21 A single trap in incubation mode. Flow in both the vertical and horizontal
directions is blocked (blue superposed on green). The cell is contained in a small, isolated
region.
103
Figure A-22 A single trap in release mode. Horizontal flow is blocked (blue superposed on
green) but vertical flow is unblocked. By addressing the multiplexer to a single trap, the cell
contained therein is selectively released.
Together, the alignment module, weir and shunt pairs, and membrane deflection valves comprise
a device which is able to deterministically allocate an incoming series of cells into sequential traps,
isolate them spatially for incubation, and release a selected cell from the device in an addressable
manner. By placing this device upstream of a single-cell analysis module, such as a device for
single cell sequencing, rare cells can be triaged for downstream analysis and released individually
for measurement.
104
Supporting Information
Supporting Information for Chapter 2
Characterization of magnetic nanoparticles: In order to determine if the nanoparticle were suitable
to this application a characterization of their size and morphology was conducted. This was
performed using two different techniques. First, the hydrodynamic particle diameter was measured
using dynamic light scattering (DLS). This revealed the average hydrodynamic radius to be 70-80
nm (Figure B-1, Top) This is considerably higher than the nominal value of 50 nm provided by
the manufacturer. However, since larger particles produce larger signals in DLS, this technique
often overestimates particle size. As such, the discrepancy may be quite minor. This small diameter
should facilitate close packing of particles and allow the number of beads to closely reflect the
surface expression of EpCAM. In order to further characterize the beads, SEM images were taken.
There images show the beads to have a consistent round morphology, which should again engender
close packing (Figure B-1, Bottom)
105
Figure B-1 Nanoparticle characterization. (Top) DLS measurement of nanoparticle size
distribution. (Bottom) TEM image showing nanoparticle size and morphology.
In order to confirm that SK-BR-3 and MDA-MB-231 both expressed EpCAM and that there was
a large difference in expression between the two, flow cytometry was performed. An anti- EpCAM
antibody conjugated directly to Alexa Fluor 488 was used (Figure B-2). Notably, both cells stain
positive for EpCAM. The majority of MDA-MB-231 cells showed an intensity below 103 whereas
the majority of SKBR3 cells appear between 103 and 104. However, there are still a considerable
number of SKBR3 cells below 103, which explains why some SK-BR-3 cells tend to be captured
in the later zones. In addition, In order to confirm that VX2 cells had sufficient EpCAM expression
for capture using the device, flow cytometry was performed. Cells were stained with APC anti-
EpCAM antibodies (Figure B-3). As shown below, VX2 cells showed a considerable expression
of EpCAM.
106
Figure B-2 Cell line flow cytometry.
Figure B-3 VX2 EpCAM flow cytometry.
107
Supporting Information for Chapter 4
Figure B-4 The critical advancing contact angle of PBS on PDMS as a function of oxygen
plasma exposure time. Error bars represent standard error.
Figure B-5 Accuracy of filling as a function of chamber size.
108
Figure B-6 A replication of Figure 4-2 F-H using buttom pressess rather than using a
hydraulic pump, showing the same ability to reliably meter discrete quantities of liquid.