D M Devicesfor Drug Deliveryand Cellular Biophysics · Recent advances in micro technologies have...

Development ofMicrofluidic Devices for Drug Delivery and CellularBiophysics

by

Jian Chen

A thesis submitted in conformity with the requirementsfor the degree of Doctor of Philosophy

Graduate Department of Electrical and Computer EngineeringUniversity of Toronto

Copyright© 2011 by Jian Chen

Abstract

Development of Microfluidic Devices for Drug Delivery and Cellular Biophysics

Jian Chen

Doctor of Philosophy

Graduate Department of Electrical and Computer Engineering

University of Toronto

2011

Recent advances in micro technologies have equipped researches with novel tools for inter-

acting with biological molecules and cells. This thesis focuses on the design, fabrication and

application of microfluidic platforms for stimuli-responsive drug delivery and the electrome-

chanical characterization of single cells.

Stimuli-responsive hydrogels are promising materials forcontrolled drug delivery due to

their ability to respond to changes in local environmental conditions. In particular, nano-

hydrogel particles have been a topic of considerable interest due to their rapid response times

compared to micro and macro-scale counterparts. Owing to their small size and thus low

drug-loading capacity, these materials are unsuitable forprolonged drug delivery. To address

this issue, stimuli-responsive implantable drug deliverymicro devices by integrating microfab-

ricated drug reservoirs with smart nano-hydrogel particles embedded composite membranes

have been proposed.

In one proposed glucose-responsive micro device, crosslinked glucose oxidase enables the

oxidation of glucose into gluconic acid, producing a microenvironment with lower pH values

to modulate the pH-responsive nanoparticles.In vitro glucose-responsive drug release profiles

were characterized and normoglycemic glucose levels in diabetic rats with device implantation

were also recorded.

The biophysical properties of single cells have recently been demonstrated as an important

indicator of disease diagnosis. Existing technologies arecapable of characterizing single pa-

ii

rameter either electrical or mechanical rapidly, but not both, which could only collect limited

information for cell status evaluation. To address this issue, two microfluidic platforms ca-

pable of simultaneously characterizing both the electrical and mechanical properties of single

cells based on electrodeformation and integrated impedance spectroscopy with micropipette

aspiration have been proposed.

In one proposed microfluidic device, a negative pressure wasused to suck cells contin-

uously through the aspiration channel with impedance profiles measured. By interpreting

impedance profiles, transit time and impedance amplitude ratio can be quantified as cellular

mechanical and electrical property indicators. Neural network based cell classification was

conducted, demonstrating that two biophysical parameterscould provide a higher cell classifi-

cation success rate than using electrical or mechanical parameter alone.

iii

Acknowledgements

First of all, I would like to bring my sincere appreciation tomy advisor, Professor Yu Sun

for his enduring support, guidance, and encouragement throughout my studies at the University

of Toronto. His enthusiasm for research has been a constant source of inspiration to me in my

current research and future career. He is a great writer, andI have learned a lot from his revision

of my writings.

I thank Professor Shirley Wu, Michael Chu and Claudia Gordijo (Department of Pharma-

ceutical Sciences), Professor Adrea Giacca, Khajag Koulajian and Simon Chiang (Department

of Physiology), Professor William Geddie, Professor Michael Jewett, Professor David Hedley

and May Cheung (Ontario Cancer Institute) for their collaboration and valuable discussions;

Professor Glenn Gulak, Professor Shirley Wu, Professor Kevin Truong, Professor Amr Helmy

and Professor Carolyn Ren for serving on my Ph.D. supervisory and defense committees.

I wish to thank all past and present members in the Advanced Micro and Nanosystems

Laboratory: Dr. Jianhua Tong, Dr. Wenhui Wang, Dr. Zhe Lu, Dr. Mohamed Abdelgawad,

Dr. Liming Yu, Dr. Changhai Ru, Dr. Xing Chen, Dr. Xuping Zhang, Dr. Yanliang Zhang,

Dr. Luke Macqueen, Dr. Keeyoung Kim, Dr. Xinyu Liu, Dr. Christopher Moraes, Yong

Zhang, Brandon Chen, Weiyin Chien, Yi Zheng, Pablo Neiman, Eric Wu, Jason Li, Steve To,

Clement Leung, Bogdan Beca, Qingyuan Tan, Ehsan Shojaei-Baghini, Haijiao Liu, Xutao Ye,

Nika Shakiba, Mario Moscovici, Lyndia Wu, David Xue, and others for the assistance and

collaboration.

Finally, I want to express my deep appreciation to my parentsand parents-in-law for their

support; and to my wonderful wife, Lijuan, and my lovely daughter, Jiashuo, for everything.

This research was supported by the Natural Sciences and Engineering Research Council of

Canada and the Canadian Institutes of Health Research.

iv

Contents

1 Introduction 1

1.1 Diabetic Management: Closed-Loop Insulin Delivery . . .. . . . . . . . . . . 1

1.1.1 Closed-Loop Insulin Delivery Devices with Discrete Components . . . 2

1.1.2 Glucose-Responsive Hydrogels in Closed-Loop Insulin Delivery . . . . 3

1.1.3 Implantable Drug Delivery Micro Devices . . . . . . . . . . .. . . . . 4

1.2 Single-Cell Electromechanical Property Characterization for Cancer Detection . 6

1.2.1 Microfluidic Devices for Single-Cell Electrical Property Characterization 9

1.2.2 Microfluidic Devices for Single-Cell Mechanical Property Characteri-

zation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.3 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 13

1.4 Dissertation Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . 13

2 pH-Responsive Drug Delivery Micro Device 14

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14

2.2 Working Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.3 Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 16

2.3.1 pH-Responsive Nanoparticle Synthesis and Characterization . . . . . . 16

2.3.2 Composite Membrane Synthesis and Characterization .. . . . . . . . 16

2.3.3 Micro Device Fabrication and Characterization . . . . .. . . . . . . . 17

2.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 19

v

2.4.1 Nanoparticle and Membrane Characterization . . . . . . .. . . . . . . 19

2.4.2 Micro Device Characterization . . . . . . . . . . . . . . . . . . .. . . 21

2.4.3 Biocompatibility Testing . . . . . . . . . . . . . . . . . . . . . . .. . 22

2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Glucose-Responsive Drug Delivery Micro Device (Type I) 24

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24



3.3.1 Synthesis and Characterization of Glucose-Responsive Membranes . . 26

3.3.2 Fabrication and Characterization of Glucose-Responsive Micro Devices 28


3.4.1 Parameter Optimization . . . . . . . . . . . . . . . . . . . . . . . . .29

3.4.2 Permeability Testing of Glucose-Responsive Membranes and Micro

Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4 Glucose-Responsive Drug Delivery Micro Device (Type II) 33

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33



4.3.1 Preparation and Characterization of Glucose-Responsive Membranes

with PDMS Grid Backbone . . . . . . . . . . . . . . . . . . . . . . . 35

4.3.2 Fabrication and Characterization of Glucose-Responsive Micro De-

vices with Enhanced Mechanical Strength . . . . . . . . . . . . . . . .36


4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

vi

5 Electrodeformation 45

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45



5.3.1 Experimental Procedures . . . . . . . . . . . . . . . . . . . . . . . .. 48

5.3.2 Numerical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51


5.4.1 Cell Elongation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5.4.2 Effect of Cells’ Electrical Properties . . . . . . . . . . . . . . . . . . . 58

5.4.3 Calculation of Young’s Modulus . . . . . . . . . . . . . . . . . . .. . 61

5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6 Impedance Spectroscopy and Micropipette Aspiration 66

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .66



6.3.1 Device Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.3.2 Device Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.3.3 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71


6.4.1 Impedance Measurement Results and Data Analysis . . . .. . . . . . 74

6.4.2 Cell Aspiration Results and Young’s Modulus Calculation . . . . . . . 78

6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

7 Impedance Spectroscopy and Constriction Channel 80

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .80



vii

7.3.1 Device Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

7.3.2 Cell Preparation and Device Operation . . . . . . . . . . . . .. . . . 83

7.3.3 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84


7.4.1 Osteoblast vs. Osteocyte . . . . . . . . . . . . . . . . . . . . . . . .. 88

7.4.2 EMT6 vs. EMT6/AR1.0 . . . . . . . . . . . . . . . . . . . . . . . . . 91

7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

8 Conclusions 98

8.1 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .99

8.2 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 100

viii

List of Tables

1.1 Summary of the electrical and mechanical properties of cancer cells with some

insights into underlining mechanisms . . . . . . . . . . . . . . . . . .. . . . 7

4.1 Permeability comparison of three cycles of micro devices using insulin con-

centration of 10 mg/ml at glucose concentration of 100 mg/dl and 400 mg/dl. . 39

4.2 Permeability comparison of two cycles of micro devices using insulin concen-

tration of 50 mg/ml at glucose concentration of 100 mg/dl and 400 mg/dl. . . . 40

5.1 Electrode dimensions and relevant parameters used in numerical simulation.

Reproduced with permission from [149]. . . . . . . . . . . . . . . . . .. . . . 52

5.2 Simulation results of electroydnamic forces as a function of cell electrical prop-

erties. Reproduced with permission from [149]. . . . . . . . . . .. . . . . . . 61

7.1 Electrode dimensions and relevant parameters used in numerical simulation. . . 86

7.2 Cell classification success rates. . . . . . . . . . . . . . . . . . .. . . . . . . 91

ix

List of Figures

1.1 Closed-loop insulin delivery devices with discrete components. (a) Key ele-

ments required for closed-loop blood glucose control. Reproduced with per-

mission from [6]. (b) Schematic representation of the s.c.-s.c. approach: sub-

cutaneous glucose monitoring and subcutaneous insulin delivery. Reproduced

with permission from [9]. (c) Schematic representation of the i.v.-i.p. ap-

proach: intravenous glucose monitoring and intraperitoneal insulin delivery.

Reproduced with permission from [7]. (d) Delays associatedwith the i.v.-i.v.,

i.v.-i.p., and s.c.-s.c. closed-loop control routes. Reproduced with permission

from [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 A closed-loop insulin delivery system utilizing a smallportion of glucose-

responsive micro-hydrogel. Reproduced with permission from [16]. . . . . . . 4

1.3 Schematic illustration of the permeation model for a composite membrane con-

taining temperature and pH-responsive hydrogel nanoparticles. Reproduced

with permission from [17]. . . . . . . . . . . . . . . . . . . . . . . . . . . . .5

x

1.4 Implantable drug delivery micro devices. (a) Schematicillustration of the

first implantable drug delivery micro device via electrochemical dissolution

approach. Reproduced with permission from [21]. (b) A representative section

of a reservoir-based micro device via electrothermal dissolution approach. Re-

produced with permission from [22]. (c) Electronic components on a printed

circuit board in package and the assembled implantable device under remote

control. Reproduced with permission from [23]. (d) Schematic illustration

of the first implantable drug delivery micro device with biodegradable mem-

branes. Reproduced with permission from [24]. . . . . . . . . . . .. . . . . . 6

1.5 Schematic illustrations of microfluidic devices for single-cell electrical prop-

erty characterization. (a) Flow cytometry based single-cell electrical property

characterization. Reproduced with permission from [78]. (b) Hydrodynamic

trapping based differential impedance characterization. Reproduced with per-

mission from [93]. (c) Aspiration based single-cell impedance spectroscopy.

Reproduced with permission from [47]. (d) Microhole based single-cell elec-

trical property characterization. Reproduced with permission from [97]. . . . . 10

1.6 Schematic illustrations of microfluidic devices for single-cell mechanical prop-

erty characterization using mechanisms of (a) micropipette aspiration (repro-

duced with permission from [100]), (b) electrodeformation(reproduced with

permission from [104]), (c) optical stretching (reproduced with permission

from [35]), (d) the constriction channel (reproduced with permission from

[126]) and (e) fluid stress (reproduced with permission from[50]). . . . . . . . 12

1.7 Schematic of electromechanical characterization of single cells using impedance

spectroscopy and deformable cantilevers. Due to the designcomplexity, the

device suffers from the absence of analytical electrical and mechanical mod-

els suitable for interpreting their measured raw data into cellular electrical and

mechanical properties. Reproduced with permission from [88]. . . . . . . . . . 12

xi

2.1 Illustration of the mechanism for pH-responsive drug release out of the mi-

cro device. Left: nanoparticles are in the swollen state when the surrounding

pH value is higher than pKa (acid dissociation constant) of the nanoparticles.

Right: Nanoparticles are in the shrunk state when the surrounding pH value is

lower than pKa. Resulting volumetric swelling and shrinking of the nanoparti-

cles control drug-release rates. Reproduced with permission from [131]. . . . . 15

2.2 Microfabrication steps for the integration of nano-hydrogel composite mem-

branes with patterned PDMS structures. Reproduced with permission from

[131]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.3 (a) Size transition of poly(NIPAm-MAA) nanoparticles resulting from varied

pH values at 37C. (b) SEM picture of a cross section of 40% nanoparticle-

loaded ethylcellulose membrane. Circled are nanoparticles into membrane

channels in ethylcellulose matrix. Reproduced with permission from [131]. . . 19

2.4 pH-dependent permeation of VB12 through ethylcellulose membranes (n=3)

that contain 40% w/w nanoparticles. Dilute acetic acid was added at 80-minute

to decrease pH from 7.4 to 4. Error bars represent standard deviations. Repro-

duced with permission from [131]. . . . . . . . . . . . . . . . . . . . . . .. . 20

2.5 Prototype micro devices for pH-responsive drug release. Reproduced with per-

mission from [131]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.6 Testing results of pH-responsive release. Data show twocycles of in vitro

testing of VB12 permeation on two micro devices. Dilute acetic acid was added

at 2-hour. Reproduced with permission from [131]. . . . . . . . .. . . . . . . 22

2.7 Total white blood cell (WBC) numbers for the control group (no device im-

plantation, n=6) and the device implantation group (n=6). Reproduced with

permission from [131]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

xii

3.1 Mechanism of hydrogel nanoparticle-embedded glucose-responsive micro de-

vices. With an increase in glucose concentration, pH-responsive nanoparticles

shrink in response to enzymatic oxidation of glucose to gluconic acid, which

leads to a higher drug release rate. . . . . . . . . . . . . . . . . . . . . .. . . 26

3.2 (top) Chitosan microparticles are used as anchors with free primary amine

groups available to crosslink with glutaraldehyde. (bottom) Schematic view

of the two-step glucose oxidase crosslinking methodology.. . . . . . . . . . . 27

3.3 Fabrication steps for the glucose-responsive micro devices. (a) A thin layer of

PDMS was spun on a glass slide. (b) The spun adhesive layer wastransferred

onto a PDMS drug reservoir through contact printing. (c) ThePDMS drug

reservoir was glued together with a chitosan embedded composite membrane

to form a drug-delivery micro device. (d) The device was soaked in solutions

of glutaraldehyde and glucose oxidase sequentially to crosslink with glucose

oxidase. (e) A fabricated prototype device. . . . . . . . . . . . . .. . . . . . . 28

3.4 Glucose oxidase activity of glucose-responsive membranes as a function of

the percentage of chitosan microparticles (anchor of glucose oxidase in com-

posite membranes). pH measurements as a function of time were taken in a

saline solution with a glucose concentration of 400 mg/ml, in which a glucose-

responsive membrane was soaked. . . . . . . . . . . . . . . . . . . . . . . .. 29

3.5 Effects of DBS 5%, 10% and 15% w/w on pH-dependent permeation of com-

posite membranes (5% chitosan w/w and 40% nanoparticle w/w). The ex-

periments were conducted in a side-by-side diffusion cell, with an effective

permeation area of 0.71 cm2. VB12 was used as the model drug. Dilute acetic

acid was added at 60-minute to decrease pH from 7.4 to 4. The calculated

permeability of composite membranes was 3.9±0.2×10−4 cm/min (5% DBS),

9.3±0.3×10−4 cm/min (10% DBS) and 4.7±0.3×10−4 cm/min (15% DBS) re-

spectively at pH 4. Error bars represent standard deviation. . . . . . . . . . . . 30

xiii

3.6 Diffusion testing of glucose-responsive composite membranes (40% w/w nanopar-

ticles, 5% chitosan and 10% DBS). Error bars represent standard deviation. . . 31

3.7 Glucose-responsive permeation of VB12 through glucose-responsive devices

(n=3) (the area for permeation is 0.78 cm2) with 5% w/w chitosan molecules,

10% w/w DBS, and 40% w/w nanoparticles. Error bars represent standard

deviation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.1 Cross-sectional schematic for PDMS grid-gel micro device embedded with

glucose oxidase and pH-responsive nanoparticles for glucose-responsive in-

sulin delivery. Reproduced with permission from [134]. . . .. . . . . . . . . . 35

4.2 (a) and (b) Standard soft lithography to form PDMS grids.(c) Pre-cured mem-

brane solution is transferred onto the PDMS grids through dip coating. (d)

Fully cured glucose-responsive membrane with PDMS grids. (e) Prototypes of

glucose-responsive composite membranes. Reproduced withpermission from

[134]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.3 (a-d) Microfabrication steps for forming glucose-responsive micro devices with

enhanced mechanical strength. (e) Prototypes of fabricated glucose-responsive

micro devices. Reproduced with permission from [134]. . . . .. . . . . . . . 37

4.4 Glucose-responsive permeation of insulin through membranes (n=3) with inte-

grated PDMS backbone. (a) Profile of insulin permeated across the membrane

in response to changes of step-wise glucose concentration variations (100, 200,

and 400 mg/dl, respectively). (b) Permeability of insulin as a function of glu-

cose concentration calculated from the slopes of the curves. Error bars repre-

sent standard deviation. Reproduced with permission from [134]. . . . . . . . . 39

xiv

4.5 (a) Three cycle permeation of insulin micro devices (10 mg/ml). Glucose lev-

els changed from 100 mg/dl (0-2.5h) to 400 mg/dl (2.5h-5h) in PBS 7.4 buffer.

Error bars represent standard deviation (n=7). (b) Permeability of insulin mi-

cro devices from normoglycemic (100 mg/dl) to hyperglycemic (400 mg/dl)

environment. Three cycles displayed with error bars representing standard de-

viation (n=7). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.6 (a) Two cycle permeation of insulin micro devices (50 mg/ml). Glucose levels

changed from 100 mg/dl (0-2.5h) to 400 mg/dl (2.5h-5h) in PBS 7.4 buffer.

Error bars represent standard deviation (n=3). (b) Permeability of insulin mi-

cro devices from normoglycemic (100 mg/dl) to hyperglycemic (400 mg/dl)

environment. Error bars represent standard deviation (n=3). . . . . . . . . . . . 41

4.7 Comparison of insulin permeated between 50 and 10 mg/ml insulin reservoir

concentrations with increase in glucose at 2.5 hour (100 to 400 mg/dl). Error

bars represent standard deviation (n=3,7). . . . . . . . . . . . . . . . . . . . . 42

4.8 Blood glucose measurements in STZ-Diabetic rats (SZT injection on Day 0

and micro device implantation on Day 2) as a function of number of devices

implanted. All data were collected from micro devices with insulin concentra-

tion of 50 mg/ml. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.9 Blood glucose measurements in STZ-Diabetic rats (SZT injection on Day 0

and device implantation on Day 2) using 25 mg/ml insulin solution. Error bars

represent standard deviation (n=3). . . . . . . . . . . . . . . . . . . . . . . . . 43

4.10 Short-term blood glucose profile of STZ-diabetic rats with implanted micro

devices during glucose challenge. Error bars represent standard deviation (n=6). 43

5.1 Schematic of positive dielectrophoresis (pDEP) and electrodeformation. Re-

produced with permission from [149]. . . . . . . . . . . . . . . . . . . .. . . 47

5.2 Fabrication steps for ITO based microelectrodes for electrodeformation. Re-

produced with permission from [149]. . . . . . . . . . . . . . . . . . . .. . . 49

xv

5.3 The experimental setup for cell electrodeformation testing. A cell is directly

placed on top of the electrodes. Rectangular AC signals are applied, and cell

deformation images are captured and processed. Reproducedwith permission

from [149]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.4 (a) Schematic of the numerical model used in the simulations. Half geometry

was simulated to reduce mesh size. All variables are defined with specific val-

ues listed in Table 5.1. (b-e) Electrodynamic forces (integration of the Maxwell

stress tensor around cell membrane in the z direction) as a function of length

of the simulation model ls, width of the simulation model lw, height of the

simulation model lh and the number of elements at the following condition:

electric field of 1 MHz at 1 Vp−p and the following electrical properties were

used:σmedium=1 mS/m,σcytoplasm=0.2 S/m, εmedium=εcytoplasm=80,εmembrane=20.

A mesh independent solution was achieved at 40,000 elements. (f) A picture

of meshing with 40 000 elements. Reproduced with permissionfrom [149]. . . 53

5.5 (a) Experimental electrodeformation of SiHa cells as a function of electric field

strength with an applied electric field of frequency 500 kHz,1 MHz, and 5

MHz. Sample size is 5 cells at 500 kHz (blue), 7 cells at 1 MHz (red) and 5

cells at 5 MHz (green). The deformation ratio is defined as theratio between

the elongation of the cell parallel to the applied electric field direction and the

original diameter of the cell before electrodeformation. (b) Numerical simu-

lations based electrodynamic forces as a function of frequency at 19 Vp−p un-

der surrounding medium conductivity of 1 mS/m with the following electrical

properties:εmembrane=10,σcytoplasm=0.1 S/m, εcytoplasm=40 (red);εmembrane=20,

σcytoplasm=0.4 S/m, εcytoplasm=80 (green) andεmembrane=30,σcytoplasm=0.7 S/m,

εcytoplasm=120 (blue). Reproduced with permission from [149]. . . . . . . .. . 59

xvi

5.6 Top: Images of electrodeformation of SiHa (a) and ME180 (b) cells as a func-

tion of electric field strength using a cell suspension of sucrose with 0.01%

BSA, electric field frequency of 1 MHz, and electrode gap of 20µm. Applied

electric field strength is indicated in brackets. Bottom: electrodeformation of

SiHa and ME180 cells at 19 Vp−p at 1 MHz with electrode gap of 20µm.

Sample size is 7 cells per cell line. Reproduced with permission from [149]. . . 60

5.7 Simulation results of electroydnamic forces for cell deformation as a function

of cell electrical properties. Reproduced with permissionfrom [149]. . . . . . . 62

5.8 Young’s modulus calculation as a function of the deformation ratio from nu-

merical simulations. For a given deformation ratio, 27 Young’s modulus values

were obtained based on simulation results, which reflected 27 cases of electri-

cal property variations shown in Table 5.2. The standard deviations (within

15% of the average value) represented the range of Young’s modulus values

due to cell electrical property variations. Simulations were conducted with the

electric field of 1 MHz and the surrounding medium conductivity of 1 mS/m.


5.9 Young’s modulus values of individual SiHa and ME180 cells determined from

electrodeformation by means of curve fitting experimental measurement re-

sults with simulation results. Standard deviation bars represent the effects of

cell electrical property variations on single-cell Young’s modulus calculation.

Individual SiHa and ME180 cells showed different Young’s modulus, which

represent cell heterogeneity. Reproduced with permissionfrom [149]. . . . . . 64

xvii

5.10 Comparison between the Young’s modulus values of SiHa and ME180 cells de-

termined from electrodeformation and micropipette aspiration. Sample size is

7 cells per cell line. Standard deviation bars of electrodeformation are mainly

due to the effect of cell electrical property variations from numerical simu-

lations and cell stiffness variations among individual cells from experiments.


5.11 Conventional micropipette aspiration experiments. Images of aspirated (a)

SiHa and (b) ME180 cells for different vacuum pressures (the aspirated lengths

and the suction pressure are indicated). Reproduced with permission from [149]. 65

6.1 Simultaneous electrical and mechanical characterization of single cells. A neg-

ative pressure is applied to trap a single cell at the entrance of the aspiration

channel. Cell deformations are recorded by imaging. Impedance measure-

ment is conducted via two Ag/AgCl electrodes connected with an impedance

analyzer. The electrical model of the aspiration channel isrepresented by

Rpipette and Cpipette in parallel. Cellular electrical components are represented

by Rcytoplasmand Cmembranein series. Rleak indicates sealing during cell aspira-

tion. Reproduced with permission from [161]. . . . . . . . . . . . .. . . . . . 68

6.2 Fabrication steps for forming the two-layer PDMS device. Reproduced with

permission from [161]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.3 Circuit models proposed to assess impedance data. Reproduced with permis-

sion from [161]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

6.4 Measured impedance amplitude (a) and phase (b) at ‘no cell trapping’, ‘cell

trapping at 50 Pa’ and ‘cell trapping at 100 Pa’ as a function of frequency

(n=18). Error bars represent standard deviation. Reproduced with permission

from [161]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

xviii

6.5 Fitting measured impedance data with cell trapping to electrical Model 2 and

Model 3. Error bars represent standard deviation. Reproduced with permission

from [161]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6.6 Calculated cell membrane capacitance (a) and cytoplasmresistance (b) of MC-

3T3 cells (n=18) by fitting experimental results with Model 3 under aspiration

pressure of 50 Pa and 100 Pa, respectively. Reproduced with permission from

[161]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6.7 (a) Images of cell aspiration at 0 Pa and 100 Pa. (b) Aspiration lengths of cells

at 50 Pa (0.813±0.351µm) and 100 Pa (1.771±0.623µm) (n=18). Error bars

represent standard deviation. Reproduced with permissionfrom [161]. . . . . . 78

7.1 Schematic of the microfluidic system for electrical and mechanical characteri-

zation of single cells using impedance spectroscopy and constriction channel.

Cells are aspirated continuously through the small constriction channel with

impedance data, cell transit time, and cell elongation length measured simulta-

neously. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

7.2 Fabrication steps for forming the PDMS based micro device for single-cell

electromechanical property characterization using impedance spectroscopy and

constriction channel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83

7.3 (a) Impedance measurement of single cells (amplitude vs. time). Transit time

indicates cellular mechanical properties and impedance amplitude ratio indi-

cates cellular electrical properties. (b) A cell aspiratedin the constriction chan-

nel. As an indicator of the cell size, the elongation length was measured from

image processing approaches. . . . . . . . . . . . . . . . . . . . . . . . . .. 85

7.4 Scatter plot of impedance amplitude ratio vs. transit time (osteoblast) as a func-

tion of testing parameters (constriction channel cross-sectional area: 6µm×6

µm and 8µm×8µm; impedance measurement frequency: 10 kHz and 100 kHz;

aspiration pressure: 10 kPa). . . . . . . . . . . . . . . . . . . . . . . . . .. . 85

xix

7.5 (a) Schematic of the 2-D numerical model used in simulation. Geometric pa-

rameters are listed in Table 7.1. (b) A section of meshing with 380,000 el-

ements. (c) Simulation results of current density as a frequency of 100 kHz

(left) and 10 kHz (right). Arrow: total current density. . . .. . . . . . . . . . . 88

7.6 Scatter plot of transit time vs. cell elongation (a) and impedance amplitude

ratio vs. cell elongation length (b). Osteoblast (n=206), osteocyte (n=217),

impedance measurement frequency: 100 kHz, aspiration pressure: 10 kPa, and

constriction channel cross-sectional area: 6µm6µm. . . . . . . . . . . . . . . 89

7.7 Pattern recognition using neural network for classifying osteoblasts (n=206)

and osteocytes (n=217). (a) Confusion matrix with the input of cell elongation

length. Success rates: 91.2% (training group), 85.7% (validation group), and

90.5% (test group). (b) Confusion matrix with the input of transittime. Suc-

cess rates: 69.7% (training group), 69.8% (validation group), and69.8% (test

group). (c) Confusion matrix with the input of impedance amplitude ratio. Suc-

cess rates: 85.5% (training group), 83.8% (validation group), and85.3% (test

group). (d) Confusion matrix with the input of transit time and impedance am-

plitude ratio. Success rates: 87.5% (training group), 81.0% (validation group)

and93.7% (test group). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

7.8 Scatter plot of transit time vs. cell elongation length (a) and impedance am-

plitude ratio vs. cell elongation length (b). EMT6 (n=747), EMT6/AR1.0

(n=770), impedance measurement frequency: 100 kHz, aspiration pressure:

10 kPa, and constriction channel cross-sectional area: 8µm×8 µm. . . . . . . . 92

7.9 (a) Transit time and (b) impedance amplitude ratio as a function of cell elon-

gation length (EMT6 vs. EMT6/AR1.0). . . . . . . . . . . . . . . . . . . . . . 92

xx

7.10 The distribution of transit time of EMT6 and EMT6/AR1.0 cells within grouped

cell elongation lengths (40-45µm (EMT6 (n=112) and EMT6/AR1.0 (n=144)

(a), 45-50µm (EMT6 (n=192) and EMT6/AR1.0 (n=199) (b) and 50-55µm

(EMT6 (n=140) and EMT6/AR1.0 (n=134) (c). . . . . . . . . . . . . . . . . . 94

7.11 Pattern recognition using neural network for classifying EMT6 (n=747) and

EMT6/AR1.0 (n=770). (a) Confusion matrix with the input of cell elonga-

tion length. Success rates: 53.0% (training group), 49.6% (validation group),

and51.3% (test group). (b) Confusion matrix with the input of transittime.

Success rates: 55.3% (training group), 58.3% (validation group), and57.5%

(test group). (c) Confusion matrix with the input of impedance amplitude ratio.

Success rates: 57.0% (training group), 53.5% (validation group), and59.6%

(test group). (d) Confusion matrix with the input of transittime and impedance

amplitude ratio. Success rates of 63.2% (training group), 63.2% (validation

group), and70.2% (test group). . . . . . . . . . . . . . . . . . . . . . . . . . 95

xxi

Chapter 1

Introduction

1.1 Diabetic Management: Closed-Loop Insulin Delivery

Diabetes mellitus, often simply referred to as diabetes, isa group of metabolic disorders char-

acterized by hyperglycemia, which results from defects in insulin secretion, insulin action or

both. The three main forms of diabetes are type 1 and type 2 diabetes, and gestational diabetes.

Type 1 diabetes is a result of the body’s failure to produce insulin due to the loss of insulin-

producing beta cells in the pancreas. Type 2 diabetes is caused by the development and onset

of insulin resistance in the body, a condition in which cellsare unable to use insulin properly.

Gestational diabetes is a result of high blood glucose levels during pregnancy [1].

The prevalence of diabetes around the world is increasing rapidly. It is estimated that

diabetes affected 194 million adults in 2003. By 2007, the number of people affected had

increased to 246 million. The number of people with diabetesworldwide projected for 2025 is

more than 300 million. In addition, diabetes accounted directly for about 3.8 million deaths in

2007, approximately 6% of the total world mortality [2].

Diabetes is a chronic disease which cannot be cured except invery specific situations,

although considerable work is under way to better utilize islet cells and stem cells for treat-

ment [3, 4]. Management of this disease, however, is possible and centers on maintaining

1

Chapter 1. Introduction 2

normoglycemia via insulin therapy without causing hypoglycemia. For type 1 diabetes, in-

sulin therapy consists of administration of exogenous insulin to the body. In type 2 diabetes, a

combination of oral medications as well as insulin administration is employed [5].

1.1.1 Closed-Loop Insulin Delivery Devices with Discrete Components

The most effective approach to realize ideal control of insulin delivery is based upon the con-

cept of artificial beta-cell (closed-loop insulin deliverydevices) [6-9]. From a technological

point of view, the development of the discrete closed-loop insulin delivery device requires

the availability of three key elements: a safe and reliable insulin delivery device, an accurate

glucose-sensing unit, and a control system that modulates insulin delivery according to blood

glucose levels and variation trends. These three elements of the artificial beta-cell should be

functionally connected, work in a continuous, reactive andphysiological mode to attain sus-

tained normoglycemia (See Figure 1.1(a)) [6].

In general, there are two approaches for closed-loop insulin delivery, which are s.c.-s.c.

(subcutaneous-subcutaneous) approach (See Figure 1.1(b)) [9] and i.v.-i.p. (intravenous-intraperitoneal)

approach (See Figure 1.1(c)) [7]. The advantage of the s.c.-s.c. approach lies in the easy cali-

bration and replacement of the glucose sensor. However, thepresence of delays associated with

insulin delivery and the unpredictable insulin absorptionfrom subcutaneous delivery disallows

effective compensation of large disturbances of blood glucoselevels such as meals (See Figure

1.1(d)) [8].

In the i.v.-i.p. approach, implantable pumps currently provide the most effective and phys-

iological insulin delivery. Despite the advent of insulin pumps and the development of control

algorithms, research efforts in pursuing closed-loop insulin delivery by integrating separated

glucose sensors, control units and insulin pumps have not yet resulted in viable automated

closed-loop insulin delivery systems due to the lack of continuous glucose sensors that are ca-

pable of performing fast, accurate measurements and havingrelatively long-term stability in

the in vivo condition [7].


Figure 1.1: Closed-loop insulin delivery devices with discrete components. (a) Key elementsrequired for closed-loop blood glucose control. Reproduced with permission from [6]. (b)Schematic representation of the s.c.-s.c. approach: subcutaneous glucose monitoring and sub-cutaneous insulin delivery. Reproduced with permission from [9]. (c) Schematic representationof the i.v.-i.p. approach: intravenous glucose monitoringand intraperitoneal insulin delivery.Reproduced with permission from [7]. (d) Delays associatedwith the i.v.-i.v., i.v.-i.p., ands.c.-s.c. closed-loop control routes. Reproduced with permission from [8].

1.1.2 Glucose-Responsive Hydrogels in Closed-Loop Insulin Delivery

Glucose-responsive hydrogels offer unique opportunities for active insulin control in response

to surrounding glucose concentration variations [10-15].These tangled networks of crosslinked

polymer chains, when immersed in a solvent, manifest a reversible and abrupt swelling and

deswelling phase transition in response to the changes in environmental glucose concentra-

tions.

The transient swelling-deswelling behavior of the glucose-responsive hydrogel is limited

by the diffusion of glucose into the hydrogel and by the absorption and expulsion of the sol-

vent (usually water). In certain microscale applications,the path-lengths become short enough

to permit the hydrogels to be used as mechanical actuators with a reasonable response time.

Professor Ziaia and Siegel’s groups at Minnesota pioneeredthe development of a closed-loop

insulin delivery system (Figure 1.2) utilizing micro-hydrogels via MEMS fabrication [16]. In

this design, a thin hydrogel (p(MPBA-co-AAm) was confined in a cavity that allows an ex-

ternal glucose solution to diffuse through a stiff porous membrane. The hydrogel’s volume


change produces a deflection of a diaphragm that opens and closes an intake orifice of a valve

to deliver insulin.

Figure 1.2: A closed-loop insulin delivery system utilizing a small portion of glucose-responsive micro-hydrogel. Reproduced with permission from [16].

To drive a mechanical load, hydrogels need to have a reasonably large size (hundreds of

micrometers in thickness), which causes slow responses (20minutes to hours) and make these

micro devices physiologically less relevant. In order to obtain fast response and enhanced me-

chanical strength of stimuli-responsive hydrogels, Professor Wu’s group at Toronto developed

prototypes of nano-hydrogel membranes with imbedded nanoparticles (p(NIPAAm-co-MAA))

(See Figure 1.3) as intelligent nano-valves that respond totemperature and pH changes [17].

The much smaller sizes of hydrogel particles (hundreds of nanometers) lead to faster responses

and the nano-hydrogel composite membranes demonstrated significantly enhanced mechanical

strength.

1.1.3 Implantable Drug Delivery Micro Devices

Advances in the MEMS technologies enable the precise fabrication of arrays of miniature drug

delivery devices for implantation [18]. Figure 1.4(a) shows the schematic view of the first

implantable drug delivery micro device developed in Prof. Langer’s group at MIT [19-21]. In


Figure 1.3: Schematic illustration of the permeation modelfor a composite membrane contain-ing temperature and pH-responsive hydrogel nanoparticles. Reproduced with permission from[17].

this device, electrical pulses were employed to dissolve a gold membrane via electrochemical

dissolution approach, allowing the diffusion of drugs out of silicon-made reservoirs.

In contrast to the electrochemical dissolution approach described above, electrothermal

activation to open reservoirs for drug release has also beenproposed [22, 23]. The passage of

an electric current through the membrane causes it to disintegrate, thereby exposing the drugs

of the reservoir to the surrounding environment (Figure 1.4(b)). This design enabled specific

reservoirs to be addressed and opened remotely by integrating electronic components within

the micro devices (Figure 1.4(c)).

Implantable drug delivery micro devices with biodegradable membranes have also been

proposed for multi-dose drug delivery [24-26]. Figure 1.4(d) shows the first biodegradable

polymeric micro device formed by the compression-molding technique. Individual membrane

recipes were prepared using various ratios of lactic acid/glycolic acid and different molecular

weight polymers to control the release of reservoir contents from the devices [24]. An ad-

vantage of biodegradable polymeric micro devices comparedto conventional implantable drug

delivery microsystems is the elimination of a requirement for a second surgery to remove the

device. In addition, the lack of electronics reduces any size restrictions in terms of device


Figure 1.4: Implantable drug delivery micro devices. (a) Schematic illustration of the firstimplantable drug delivery micro device via electrochemical dissolution approach. Reproducedwith permission from [21]. (b) A representative section of areservoir-based micro device viaelectrothermal dissolution approach. Reproduced with permission from [22]. (c) Electroniccomponents on a printed circuit board in package and the assembled implantable device un-der remote control. Reproduced with permission from [23]. (d) Schematic illustration of thefirst implantable drug delivery micro device with biodegradable membranes. Reproduced withpermission from [24].

manufacture.

In these implantable drug delivery micro devices, althoughdrugs can be released on de-

mand through remote control or by the degradation of the sealing membranes, they are not

capable of regulating drug delivery rates and volumes according to local micro environmental

signals such as the pH value and the glucose concentration.

1.2 Single-Cell Electromechanical Property Characterization

for Cancer Detection

Cancer is an abnormal growth of cells caused by multiple changes in gene expression leading

to dysregulated balance of cell proliferation and cell death. This may ultimately evolve into a

population of cells that can invade tissues and metastasizeto distant sites, causing significant

morbidity and, if untreated, death of the host [27].


Cancer is an important cause of ill health worldwide. In 2008, estimated 12.4 million

people were diagnosed with cancer. The most common cancers,primarily breast, lung, stomach

and prostate cancer, accounted for 50% diagnoses. There are6.7 million reported deaths from

cancer annually, making up 12% of deaths worldwide. This is more than HIV/AIDS, malaria

and tuberculosis combined [28].

Diagnostic procedures for cancer include imaging, laboratory tests (including tests for tu-

mor markers), tumor biopsy, endoscopic examination, surgery, and genetic testing. A number

of types of tumor markers have been identified and these include nucleic acid-based markers

such as mutations, loss of genetic heterozygosity, microsatellite instability, and gene expression

microarrays, as well as protein markers, protein pattern recognition profiles and circulating tu-

mor cells. Some of these markers can be detected in the circulation or in body fluids, and some

request tumor tissue [27].

Besides the development of biochemical markers, cellular biophysical markers (e.g., elec-

trical properties of the cell membrane and cytoplasm [29-31] and mechanical properties of the

cytoskeleton [32, 33]) have been correlated with pathophysiological states in cancer. Table 1.1

summarizes the reported observations of the abnormal electrical and mechanical properties of

cancer cells as well as available information on the underline mechanisms that lead to these

variations [34-50].

Table 1.1: Summary of the electrical and mechanical properties of cancer cells with

some insights into underlining mechanisms

Cell lines Experimental methods Key observations

Breast cancer cells(MCF-7 and MDRderivatives)

Dielectrophoresis Cytoplasmic conductivity differenceswere observed due to different molecularmovement activities in cytoplasm [34].

Head and neck cancercells (686 LN vs. 686LN-M4e)

Micro electricalimpedance spec-troscopy

The poorly metastatic HNC cell line had ahigher impedance phase compared to thehighly metastatic HNC cell line [47].


Breast cancer cells(MCF-7, MDA-MB-231, MDA-MB-435,and MCF-10A)

Micro electricalimpedance spec-troscopy

In both magnitude and phase, significantdifferences were observed between thenormal cells and the cancer cells [39].

Bladder cancer cells(Hu609, HCV29,Hu456, T24, andBC3726)

Atomic force mi-croscopy (AFM)

Malignant cells were an more deformablethan benign counterparts [45].

Breast cancer cells(MCF-10A and MCF-7)

Microfluidic opticalstretcher

MCF-7 cells deformed more than MCF-10A due to a reduction in F-actin [35].

Epithelial pancreaticcancer cells (Panc-1)

Mechanical microplatestretcher

Drug treatment of pancreatic cells re-sulted in a reduction in stiffness as aresult of re-organization of the keratinmolecules [48].

TRAIL-expressingleukemic cells (Jurkat)

Micropipette aspiration Apoptosis-inducing TRAIL causedan in-crease in elastic stiffness of the cell [44].

Myeloid (HL60) andlymphoid leukemiccells (Jurkat)

Indentation with AFM HL60 myeloid cells were 18 times stifferthan Jurkat and six times stiffer than hu-man neutrophils [37].

Leukemic cells (ALLand AML)

Indentation with AFM Leukemic cell stiffness was increasedwhen exposed to chemotherapeuticagents [40].

Colon carcinoma cells(HT-29)

Flow experiments in aflow chamber

Changes in cell elasticity were locateddue to cytoskeletal reorganization [46].

Fibroblasts Micropipette aspiration An increase in deformability correlatingwith the transformed phenotype was ob-served [38].

Lung carcinoma cells(3LL)

Micropore filtration Nitric oxide caused a reduction in de-formability of the 3LL cells [42].

Melanoma cells (F1B16)

Micropore filtration Damage to cytoskeletal components wasnoticed to increase the deformability ofcells [41].

Erythroleukemic cells Micropipette aspiration MEL-W exhibited less deformability thanmodified ones due to compositionalchanges in the cytoskeleton proteins [43].

Fibroblasts AFM Cancer fibroblasts were significantlymore deformable than normal counter-parts [36].

Fibroblasts Microfluidic opticalstretcher

Deformability of SV-T2 cells was higherthan that of the benign cells, possibly as aresult of a reduction in F-actin [35].

Breast cancer cells(MCF-7 and MCF-10A)

Constriction channel MCF-7 cells deformed more than MCF-10A cells [49].


Breast cancer cells(MCF-10A, MCF-7and Mod MCF-7)

Microfluidics basedfluid stress

MCF-7 cells deformed more than MCF-10A cells. Modified MCF-7 cells de-formed more than MCF-7 cells [50].

1.2.1 Microfluidic Devices for Single-Cell Electrical Property Character-

ization

Advances in microfluidic technologies have led to the development of micro devices for single-

cell electrical property characterization. The three types of micro devices for electrical char-

acterization are patch clamp [51-64], electro-rotation [65-70] and micro-electrical impedance

spectroscopy (Micro-EIS) [29-31], respectively. Patch-clamp micro devices characterize cellu-

lar ion channel activities by sucking a cell membrane patch into a micropipette to form a high

electrical resistance seal. This technique is invasive since the cell membrane is intentionally

disrupted, and thus, cannot be used for long-term study or multiple types of measurements.

In electro-rotation, a rotating electric field is exerted ona suspended cell causing the cell to

rotate as a result of the Maxwell-Wanger polarization. Although electro-rotation is a powerful

technique for measuring cell membrane permittivity and cytoplasm conductivity, a proper cell

manipulation and position in the rotating electrical field is requested, which is labor intensive

and highly dependent on the operator experience.

Micro-EIS is a non-invasive technique in which a frequency-dependent excitation signal is

applied across a single cell to measure the corresponding current response [29-31]. Several flow

cytometry based microfluidic devices have recently been reported for single-cell impedance

measurement with rather high throughput [71-87] (see Figure 1.5(a)). Hydrodynamic trap-

ping methods have also been employed in conjunction with Micro-EIS for measuring cellular

impedance properties over extended periods of time [88-95](see Figure 1.5(b)). Furthermore,

two mechanisms utilizing vacuum aspiration [39, 47] and electrode surface modification [96]


were proposed to form tight cell adhesion on the measurementelectrodes (see Figure 1.5(c)).

In addition, micro-hole based chips modified from patch-clamp devices have also been pro-

posed to address the leakage current issue by forming propersealing between the aspirated

cell and the aspiration channel [97, 98] (see Figure 1.5(d)). The main concern of the Micro-

EIS technique is that differences in impedance profiles between measurements taken with and

without a cell’s presence are usually small and sometimes unobservable [96].

Figure 1.5: Schematic illustrations of microfluidic devices for single-cell electrical propertycharacterization. (a) Flow cytometry based single-cell electrical property characterization. Re-produced with permission from [78]. (b) Hydrodynamic trapping based differential impedancecharacterization. Reproduced with permission from [93]. (c) Aspiration based single-cellimpedance spectroscopy. Reproduced with permission from [47]. (d) Microhole based single-cell electrical property characterization. Reproduced with permission from [97].

1.2.2 Microfluidic Devices for Single-Cell Mechanical Property Charac-

terization

In the meanwhile, several microfluidic devices have also been developed to measure the me-

chanical properties of single cells based on various mechanisms including micropipette aspi-

ration [99, 100], electrodeformation [101-107], optical stretcher [35, 108-117], constriction

channel [49, 118-127] and fluid stress [50, 128-130].


In micropipette aspiration, a cell is deformed by applying anegative pressure through an

aspiration channel. By recording resulting geometrical changes of the cell and using well-

established aspiration mechanics models, raw data can be used to extract the cell’s Young’s

modulus (see Figure 1.6(a)). In electrodeformation, a cellplaced in an applied electric field be-

comes polarized due to the surface charge build-up and therefore is deformed electrically. By

interpreting the relationship between the deformation ratio and the applied electric field, cell

mechanical properties can be obtained (see Figure 1.6(b)).In an optical stretcher, a two-beam

laser trap is optimized to serially deform single suspendedcells by optically induced surface

forces to measure mechanical properties (see Figure 1.6(c)). In devices with constriction chan-

nels, the cells are squeezed through a channel with a smallercross-section area by hydraulic

pressure differences and the transit time is recorded as a mechanical property indicator (see

Figure 1.6(d)). By elaborately designing the configurationof micro channels, cells can be

exposed to various fluid stresses and the corresponding deformations can be used to quantify

cellular mechanical properties (see Figure 1.6(e)).

Discussions above reveal that a number of microfluidic devices have been demonstrated

for cell biophysical characterization. However, the majority of these devices are only capable

of characterizing either electrical properties (i.e., ionchannel activities, membrane capacitance

and cytoplasm resistance) or mechanical properties (i.e.,Young’s modulus) of a targeted cell.

For a more complete understanding of a cell’s properties, itis desirable to perform both elec-

trical and mechanical measurements on the same cell using the same device. The only device,

which was reported to perform both impedance and mechanicalcharacterization of single cells

[88], is complex in both design and the microfabrication process (see Figure 1.7). Due to the

design complexity, this reported device suffers from the absence of analytical electrical and

mechanical models suitable for interpreting their measured raw data into cellular electrical and

mechanical properties.


Figure 1.6: Schematic illustrations of microfluidic devices for single-cell mechanical propertycharacterization using mechanisms of (a) micropipette aspiration (reproduced with permis-sion from [100]), (b) electrodeformation (reproduced withpermission from [104]), (c) opticalstretching (reproduced with permission from [35]), (d) theconstriction channel (reproducedwith permission from [126]) and (e) fluid stress (reproducedwith permission from [50]).

Figure 1.7: Schematic of electromechanical characterization of single cells using impedancespectroscopy and deformable cantilevers. Due to the designcomplexity, the device suffers fromthe absence of analytical electrical and mechanical modelssuitable for interpreting their mea-sured raw data into cellular electrical and mechanical properties. Reproduced with permissionfrom [88].


1.3 Research Objectives

The overarching goal of this thesis is to develop new microfabricated tools for controlled drug

delivery and cellular biophysics studies. The objectives of this thesis are:

• To design, fabricate, and characterize (bothin vitro andin vivo) implantable

stimuli-responsive (pH and glucose-responsive) micro devices for controlled drug

delivery.

• To design, fabricate, and characterize new microfluidic platforms capable of

high-throughput electromechanical characterization of single cells.

1.4 Dissertation Outline

An overview of the ensuing chapters is as follows: Chapter 2 describes the design, fabrication

and characterization of a pH-responsive drug delivery micro device. Chapter 3 presents the de-

sign, fabrication, optimization and characterization of type I glucose-responsive drug delivery

micro devices. Chapter 4 proposes the design, fabrication,in vitro and in vivo characteriza-

tion of type II glucose-responsive drug delivery micro devices. Chapter 5 describes the design,

simulation, fabrication and characterization of a microfluidic device for single-cell electrome-

chanical property measurement using electrodeformation.Chapter 6 presents the design, fabri-

cation and characterization of a microfluidic device for single-cell electromechanical property

measurement using electrical impedance spectroscopy and micropipette aspiration. Chapter 7

proposes the design, fabrication and high-throughput characterization of a microfluidic device

for the measurement of single-cell electromechanical properties using electrical impedance

spectroscopy combined with a constriction channel. The thesis is concluded in Chapter 8, with

a summary and contributions of this research and feasible future research directions.

Chapter 2

pH-Responsive Drug Delivery Micro

Device

2.1 Introduction

This chapter presents a drug-delivery micro device integrating pH-responsive nano-hydrogel

particles functioning as intelligent nano valves. The polymeric micro device is monolithic

without requiring peripheral control hardware or additional components for controlling drug-

release rates. pH-responsive nanoparticles were synthesized and embedded into a composite

membrane. The resulting pH-responsive composite membranewas integrated with a PDMS

based micro reservoir via a room-temperature transfer bonding technique to form the proof-

of-concept micro device.In vitro release characterization of the micro devices was conducted

in which the release rate of Vitamin B12 (VB12) as a model drug increased dramatically when

the local pH value was decreased from 7.4 to 4. This device concept can serve as a platform

technology for intelligent drug delivery in response to various in vivo environmental signals.

The project entitled “pH-responsive drug delivery micro devices” is a collaborative project

among three research laboratories at the University of Toronto: Prof. Y. Sun (Department of

Mechanical and Industrial Engineering), Prof. S. Wu (Department of Pharmaceutical Sciences)

14

Chapter 2. pH-Responsive Drug Delivery Micro Device 15

and Prof. A. Giacca (Department of Physiology). M. Chu (Professor Wu’s group) and J. Chen

(Prof. Sun’s group) synthesized and characterized pH-responsive membranes. J. Chen and M.

Chu fabricated and characterized pH-responsive drug delivery micro devicesin vitro. K. Koula-

jian (Prof. Giacca’s group), J. Chen and M. Chu characterized the device biocompatibilityin

vivo.

2.2 Working Mechanism

As shown in Figure 2.1, the patterned PDMS structure forms a drug reservoir and provides

physical support for the thin nano-hydrogel embedded composite membrane. The embedded

hydrogel nanoparticles in the composite membrane detect environmental pH changes as intel-

ligent nano valves. Corresponding volumetric swelling andshrinking response of the nanopar-

ticles controls drug-release rates.

Figure 2.1: Illustration of the mechanism for pH-responsive drug release out of the microdevice. Left: nanoparticles are in the swollen state when the surrounding pH value is higherthan pKa (acid dissociation constant) of the nanoparticles. Right: Nanoparticles are in theshrunk state when the surrounding pH value is lower than pKa.Resulting volumetric swellingand shrinking of the nanoparticles control drug-release rates. Reproduced with permissionfrom [131].


2.3 Experimental Methods

2.3.1 pH-Responsive Nanoparticle Synthesis and Characterization

For nanoparticle synthesis, N-isopropylacrylamide (NIPAm), methacrylic acid (MAA), N,N-

bisacrylamide (BIS), and potassium persulfate (KPS) were from Sigma. Sodium lauryl sulfate

(SDS) was from Fisher. Ethylcellulose powder (viscosity 45) for composite membrane fabri-

cation was from Dow Chemical Company.

The pH-responsive poly(N-isopropylacrylamide-co-methacrylic acid) (PNIPAm-MAA) nanopar-

ticles with 1:1 molar ratio of NIPAm to MAA were prepared by anaqueous dispersion poly-

merization process using BIS as the crosslinking agent, KPSas the initiator, and SDS as the

stabilizer. Specifically, monomer mixtures (763.83 mg of NIPAm and 0.57 ml of MAA) were

dissolved in 100 ml of distilled water. After the incorporation of 133.22 mg of BIS, 11.54 mg

of SDS was added to the reaction mixture. The reaction mixture was purged with nitrogen for

0.5 hour and then polymerization was initiated by the addition of 56.77 mg of KPS. The poly-

merization process was conducted under a nitrogen blanket at 70C for 4 hours at 200 rpm.

Typically, nanoparticles in aqueous solution with a polymer concentration of 1.5 wt% were

prepared. More concentrated samples were obtained throughcentrifugation.

Synthesized nanoparticles were dispersed in 10 mg/ml concentration with distilled water.

A volume of 50µl of solution was taken and placed into glass light scattering test tube with

700µl of phosphate buffer solution (PBS) at different pH values. Particle sizer measurements

were performed with a NICOMP 380 particle sizer.

2.3.2 Composite Membrane Synthesis and Characterization

To form a useful material for device construction, the nanoparticles were embedded into a

composite membrane. Casting of composite membranes started with dissolving 0.65 g of ethyl-

cellulose powder in 15 ml of anhydrous alcohol at room temperature (22C). Ethylcellulose

solution was mixed with dispersed nanoparticle solution for another 2 hours. Ethylcellulose/-


nanoparticle solution was then poured into a 10-cm Teflon plate and placed in a vacuum tank

for 15 minutes for degassing. Solution was allowed to dry for24 hours in a sealed desiccator to

slow evaporation and prevent contamination. SEM analysis of the membrane was performed

to obtain visual characterization of the nanoparticle and membrane.

Permeability testing of nano-hydrogel composite membranes was conducted in side-by-

side diffusion cells using VB12 as a model drug (Mw=1355). Multiple 2-cm diameter circular

samples were soaked in pH 7.4 phosphate buffer solution for 24 hours before testing. Side-

by-side water jacketed diffusion cells housed both the receptor and donor cells for permeabil-

ity testing. The volume of each cell was 3 ml and the area for permeation was 0.63 cm2.

Polyethylene tubes connected to a peristaltic pump (silicone tubes within the pump) allowed

for continuous flow from the receptor cell to a cuvette. The cuvette was placed in a UV spec-

trometer (Hewlett-Packard 8452A) for kinetic measurements of VB12 diffusion. The donor cell

was filled with 1 mg/ml VB12 solution. Data were collected over a 4-hour period at 10-minute

intervals at different pH values.

Solute permeability P=DK/h was determined from the relationship Mt=PSCd(t-tL) based

on Fick’s first law of diffusion where Mt is the mass of a drug permeated till time t (steady state

is reached in the membrane after a lag time, tL) with the solute concentration of Cd in the donor

cell; D and K are, respectively, the diffusion coefficient and partition coefficient of the drug; h

is the thickness of the membrane. Permeability (P) was determined from the slope of the curve

of M t vs. t at the steady state.

2.3.3 Micro Device Fabrication and Characterization

The integration of nano-hydrogel composite membranes and PDMS drug reservoirs must en-

sure the physical and functional integrity of the compositemembranes without exposing the

hydrogel nanoparticles to harsh processing conditions, such as UV, plasma, high temperatures

or wet chemical etchants.

The PDMS reservoir was formed via standard soft lithographyusing SU-8 as the mould


master. The thickness of the SU-8 master was approximately 1mm, obtained by spinning

two layers of SU-8-2100. After the PDMS drug reservoir was peeled off from the substrate,

room-temperature transfer bonding was used to integrate nano-hydrogel embedded composite

membranes. A thin layer of PDMS was spun on a substrate as a bonding adhesive layer (Figure

2.2(a)). Through micro-contact printing, the adhesive layer was transferred to the PDMS drug

reservoir (Figure 2.2(b) and (c)). The drug reservoir was then bonded to a nano-hydrogel

composite membrane and left to cure at room temperature (Figure 2.2(d)).

Figure 2.2: Microfabrication steps for the integration of nano-hydrogel composite membraneswith patterned PDMS structures. Reproduced with permission from [131].

In vitro testing was conducted on the micro devices filled with 5 mg/ml VB12 solution.

Device-containing chambers were soaked in a water bath at 37C and solutions around micro

devices were measured by a UV spectrometer. Data were collected over an 8-hour period with

a change in pH from 7.4 to 4 triggered by adding dilute acetic acid. After each change in pH,

the devices were washed several times with distilled water before re-testing.

To verify the suitability of the micro devices for short-term implantation use in animals,in

vivo biocompatibility testing was conducted in Sprague-Dawleyrats. The rats were randomly

divided into 2 groups and underwent implant surgery under isoflurane anesthesia. The sham


group (n=6) underwent surgery without the implantation of a device while rats in the test group

(n=6) were implanted subcutaneously in the interscapular tissue with ethylene oxide sterilized

empty micro devices for two weeks. Biocompatibility was assessed after implantation by quan-

tifying the number of white blood cells (WBC) present in whole blood. 1 mL of blood was

collected in an EDTA-vacutainer tube and the WBC count was determined based on size using

a coulter counter.

2.4 Results and Discussion

2.4.1 Nanoparticle and Membrane Characterization

Figure 2.3(a) shows that the diameters of the nanoparticlesdecrease dramatically from 490±41.6

nm (pH 7.4) to 327±29.3 nm (pH 4). Figure 2.3(b) shows a scanning electron microscopy pic-

ture of a composite membrane after freezing and cross-sectional fractioning. The membranes

have sufficient mechanical strength and suitable for physical handling.

Figure 2.3: (a) Size transition of poly(NIPAm-MAA) nanoparticles resulting from varied pHvalues at 37C. (b) SEM picture of a cross section of 40% nanoparticle-loaded ethylcellu-lose membrane. Circled are nanoparticles into membrane channels in ethylcellulose matrix.Reproduced with permission from [131].

Figure 2.4 shows a representative set of permeability testing data, which proves the pH

responsiveness of the nano-hydrogel composite membranes.After the addition of dilute acid to

decrease the pH value of the side-by-side diffusion cell from 7.4 to 4, VB12 amount diffused into


the receptor cell increased steadily with a calculated permeability of 1.17×10−5±1.46×10−6

cm/s.

Figure 2.4: pH-dependent permeation of VB12 through ethylcellulose membranes (n=3) thatcontain 40% w/w nanoparticles. Dilute acetic acid was added at 80-minute to decrease pHfrom 7.4 to 4. Error bars represent standard deviations. Reproduced with permission from[131].

Nanoparticle percentages ranging from 10% to 40% were embedded into the ethylcellu-

lose membranes. At 30% and lower, there was no obvious response to pH, possibly due to

the insufficient amount of nanoparticles to create nano-channels for drug diffusion. At 40%,

pH response was strong. Even higher percentages (>40%) resulted in uneven ethylcellulose

membranes.

Alcohol-based ethylcellulose was chosen as the base polymer since it showed highly satis-

factory film formation at room temperature, and membrane synthesis was relatively fast. Fur-

thermore, the membranes remained mechanically strong after the addition of nanoparticles.

Cellulose acetate, despite being a similar base polymer, required a much longer synthesis time

(24 hours vs.∼3.5 days), and the resulting membrane was poor in quality (e.g., surface uni-

formity). Aquacoat, an aqueous suspension of ethylcellulose, was also tested for membrane

synthesis. The resulting membranes were weak in mechanicalstrength and remained weak

with the addition of plasticizers, making physical handling of the membranes difficult.


2.4.2 Micro Device Characterization

Figure 2.5 shows prototypes of the pH-responsive drug-delivery micro devices. The drug reser-

voirs were constructed with PDMS due to its mechanical stability, the feasibility of precise

patterning using microfabrication, and its biocompatibility and insusceptibility to protein ad-

sorption and fouling. Other biocompatible materials such as poly(methyl methacrylate) and

poly(ethylene-co-vinyl acetate) can also be valid options. VB12 filling and refilling were con-

ducted through the backside of the PDMS reservoirs with a small-gauge syringe needle, made

possible by the self-sealing property of PDMS.

Figure 2.5: Prototype micro devices for pH-responsive drugrelease. Reproduced with permis-sion from [131].

Figure 2.6 shows VB12 release profiles of two-cycle tests of two micro devices. After the

addition of dilute acid to decrease the pH of the surroundingmedium from 7.4 to 4, VB12

amount in the surrounding medium increased steadily. The results proved the release concept

and the capability of the micro devices for pH-responsive drug delivery.

The slight lag time for diffusion could be due to the chain of chemical events. The hydro-

gen ions from acid need to come in contact with the embedded nanoparticles, which requires

diffusion into the membrane. Although the nanoparticle shrinkage time is short, drug diffusion

takes time to permeate the composite membrane. By adjustingnanoparticle percentages and

the thickness of the composite membranes, the response timecould be further shortened. The


Figure 2.6: Testing results of pH-responsive release. Datashow two cycles ofin vitro testingof VB12 permeation on two micro devices. Dilute acetic acid was added at 2-hour. Reproducedwith permission from [131].

slight drop of VB12 after the surrounding pH decreased from 7.4 to 4 could be attributed to

osmotic pressures and hydrogen ion gradients across the membrane which could have brought

a small amount of VB12 together with water into the micro devices, as nanoparticles began to

shrink.

2.4.3 Biocompatibility Testing

Micro devices after implantation were retrieved. They demonstrated satisfactory mechanical

strength for implantation. The total white blood cells (WBCs) at the end of the experiment

for the ”device implantation” (5.35×109/L±1.78×109) group were slightly but not significantly

increased compared to the ”sham” group (3.78×109/L±1.32×109) (Figure 2.7). This may in-

dicate the presence of some inflammation due to the foreign implant. The count of specific

inflammatory cells shows mostly an increase in the number of lymphocytes, which may be an

indication of chronic inflammation. This, however, was not significant.


Figure 2.7: Total white blood cell (WBC) numbers for the control group (no device implanta-tion, n=6) and the device implantation group (n=6). Reproduced with permission from [131].

2.5 Conclusion

In summary, this chapter demonstrated the first monolithic pH-responsive drug-delivery micro

device. The micro devices integrated pH-responsive nanoparticles as intelligent nano-valves

that were embedded into composite membranes.In vitro release characterization proved the

concept and the capability of the micro devices for pH-responsive drug delivery.In vivo bio-

compatibility testing verified the suitability of the devices for short-term implantation in rats.

Chapter 3

Glucose-Responsive Drug Delivery Micro

Device (Type I)

3.1 Introduction

Previously, we developed a pH-responsive drug delivery micro device by integrating a pH-

responsive hydrogel nanoparticle-embedded composite membrane with a PDMS drug reser-

voir. In this chapter, we present: (1) a glucose-responsivecomposite membrane with chem-

ically immobilized glucose oxidase on chitosan micro particles and pH-responsive hydrogel

nanoparticles, functioning as intelligent ‘nano-valves’in response to surrounding glucose con-

centration variations; and (2) a proof-of-concept glucose-responsive drug-delivery micro de-

vice by integrating the newly developed glucose-responsive composite membrane with a mi-

crofabricated drug reservoir. Membrane synthesis parameters were systematically investigated

and responsive release profiles were quantified by testing model drug (VB12) diffusion through

the glucose-responsive membranes. Release profiles of the glucose-responsive micro devices

were also measuredin vitro, demonstrating a marked increase in release rates of VB12 when

glucose levels in the surrounding media were increased from0 to 400 mg/dl.

In this study, VB12 was chosen as the model drug for the proof-of-concept permeability

24

Chapter 3. Glucose-Responsive Drug Delivery Micro Device (Type I) 25

testing of composite membranes and devices for the following reasons: (1) VB12 has a well-

characterized structure and a molecular weight (Mw=1355) similar to polypeptides of potential

interest (e.g., glucagon Mw=3485 and insulin Mw=5808). (2) VB12 has a higher stability com-

pared to glucagon and insulin, without the concern of protein denaturing by absorbance on

hydrophobic surfaces. (3) VB12 has a lower cost compared to polypeptide counterparts. (4)

VB12 shows a five-fold higher absorptivity than insulin in drug permeation using a UV spec-

trometer so that permeation changes can be more easily measured.

The project entitled “type I glucose-responsive drug delivery micro devices” is a collab-

orative project between two research laboratories at the University of Toronto: Prof. Y. Sun

(Department of Mechanical and Industrial Engineering) andProf. S. Wu (Department of Phar-

maceutical Sciences). M. Chu (Professor Wu’s group) and J. Chen (Prof. Sun’s group) syn-

thesized and characterized glucose-responsive membranes. J. Chen and M. Chu fabricated and

characterized glucose-responsive drug delivery micro devicesin vitro.


As shown in Figure 3.1, a patterned PDMS structure forms a drug reservoir and provides phys-

ical support for the glucose-responsive composite membrane. In the composite membrane,

chitosan microparticles work as anchors for chemically immobilizing glucose oxidase. The

embedded hydrogel nanoparticles detect local pH changes caused by gluconic acid resulting

from the oxidation of glucose, acting as intelligent ‘nano-valves’. Corresponding volumet-

ric swelling and shrinking of the hydrogel nanoparticles inresponse to environmental glucose

levels controls drug-release rates.


Figure 3.1: Mechanism of hydrogel nanoparticle-embedded glucose-responsive micro devices.With an increase in glucose concentration, pH-responsive nanoparticles shrink in response toenzymatic oxidation of glucose to gluconic acid, which leads to a higher drug release rate.


For nanoparticle synthesis, N-isopropylacrylamide (NIPAm, 99%), methacrylic acid (MAA),

N’N’-bisacrylamide (BIS), potassium persulfate (KPS), dibutyl sebacate (DBS) and glutaralde-

hyde (25%) were from Sigma-Aldrich. Sodium lauryl sulfate (SDS) was from Fisher Scien-

tific. Ethylcellulose powder (viscosity 45) for composite membrane fabrication was from Dow

Chemical Company. TM 2495 chitosan powder was from Chitoclear. Glucose oxidase (333

u/mg) was from Calzyme Laboratories.

3.3.1 Synthesis and Characterization of Glucose-Responsive Membranes

The synthesis of glucose-responsive membranes started with the formation of ethylcellulose-

based composite membranes containing pH-responsive nanoparticles (see section 2.3.1), chi-

tosan microparticles and DBS. Chitosan microparticles work as anchors for crosslinking glu-

cose oxidase. DBS was added to improve membrane quality (e.g., pH-responsive permeability)

as a plasticizer. Membrane synthesis parameters such as thepercentages of chitosan and DBS


were systematically varied and investigated. pH-responsive permeability testing of composite

membranes was conducted in the side-by-side diffusion cell using VB12 as the model drug,

described in section 2.3.2.

Immobilization of glucose oxidase was performed after the fabrication of the chitosan mi-

croparticle embedded membrane in a two-step process by soaking the membrane in a solution

of glutaraldehyde (crosslinker), and then in a glucose oxidase solution to complete chemical

immobilization (Figure 3.2).

Figure 3.2: (top) Chitosan microparticles are used as anchors with free primary amine groupsavailable to crosslink with glutaraldehyde. (bottom) Schematic view of the two-step glucoseoxidase crosslinking methodology.

For glucose oxidase activity measurements, each compositemembrane crosslinked with

glucose oxidase was initially washed and soaked in a saline solution. After glucose was added

into the saline solution to form a glucose concentration of 400 mg/dl, pH drop was recorded.

Permeability testing of the glucose-responsive membraneswas conducted in a similar man-

ner to pH-responsive permeability testing, except that concentrated glucose solutions (end glu-

cose concentration of 400 mg/dl) rather than dilute acetic acid were added as stimuli.



vices

To fabricate glucose-responsive drug-delivery micro devices (Figure 3.3(a-d)), standard soft

lithography was first used for forming a PDMS drug reservoir.Using a thin layer of uncured

PDMS as adhesive, a chitosan microparticle embedded pH-responsive composite membrane

was bonded to the PDMS drug reservoir and cured at room temperature. To make the devices

glucose responsive, the devices were soaked in a solution ofglutaraldehyde and then a solu-

tion of glucose oxidase for glucose oxidase immobilization. Figure 3.3(e) shows a fabricated

prototype device.

Figure 3.3: Fabrication steps for the glucose-responsive micro devices. (a) A thin layer ofPDMS was spun on a glass slide. (b) The spun adhesive layer wastransferred onto a PDMSdrug reservoir through contact printing. (c) The PDMS drug reservoir was glued together with achitosan embedded composite membrane to form a drug-delivery micro device. (d) The devicewas soaked in solutions of glutaraldehyde and glucose oxidase sequentially to crosslink withglucose oxidase. (e) A fabricated prototype device.

Glucose-responsive release testing was conducted on the micro devices filled with 5 mg/ml

VB12 solution. The micro devices were soaked in a water bath at 37C, and solutions around

micro devices were measured by a UV spectrometer. Data was collected over 4-hour periods

with an increase in glucose in the solution from 0 mg/dl to 400 mg/dl at 50-minute.



3.4.1 Parameter Optimization

The effect of the percentage of chitosan on glucose oxidase crosslinking was quantified. As

the percentage of chitosan content increased, the membranes showed lower pH-responsive per-

meability and poorer membrane quality, with no membrane formation above 10% w/w. By

measuring the pH drop in a glucose solution, in which a glucose-responsive membrane was

soaked, it was found that membranes with 4%-8% w/w chitosan particles showed compara-

ble glucose oxidase activity (Figure 3.4). After 20 minutes, the pH level dropped below 4.5,

demonstrating sufficient glucose oxidase activity.

Figure 3.4: Glucose oxidase activity of glucose-responsive membranes as a function of the per-centage of chitosan microparticles (anchor of glucose oxidase in composite membranes). pHmeasurements as a function of time were taken in a saline solution with a glucose concentrationof 400 mg/ml, in which a glucose-responsive membrane was soaked.

Figure 3.5 shows the effect of DBS percentages (5%, 10% and 15% w/w) on pH respon-

siveness of composite membranes (5% chitosan, 40% nanoparticles, before enzyme immobi-

lization). As the DBS percentage increased from 0% to 10% w/w, the membrane permeability

increased from 0 (0% w/w DBS) to 3.9±0.2×10−4 cm/min (5% w/w DBS) and to 8.4±0.3×10−4

cm/min (10% w/w DBS) at pH 4. When the percentage of DBS increased from 10% to15%


w/w, the membrane permeability decreased from 8.4±0.3×10−4 cm/min (10% w/w DBS) to

4.7±0.5×10−4 cm/min (15% w/w DBS). It was speculated that when the percentage of DBS

was too high (e.g.,>10%), as a small molecule, DBS filled the voids surrounding nanoparti-

cles in the membranes leading to decreased pH-responsive permeability changes.

Figure 3.5: Effects of DBS 5%, 10% and 15% w/w on pH-dependent permeation of compositemembranes (5% chitosan w/w and 40% nanoparticle w/w). The experiments were conductedin a side-by-side diffusion cell, with an effective permeation area of 0.71 cm2. VB12 was usedas the model drug. Dilute acetic acid was added at 60-minute to decrease pH from 7.4 to 4.The calculated permeability of composite membranes was 3.9±0.2×10−4 cm/min (5% DBS),9.3±0.3×10−4 cm/min (10% DBS) and 4.7±0.3×10−4 cm/min (15% DBS) respectively at pH4. Error bars represent standard deviation.

3.4.2 Permeability Testing of Glucose-Responsive Membranes and Micro

Devices

With an optimized membrane containing 5% w/w chitosan, 10% w/w DBS, and 40% w/w

nanoparticles, the pH-responsive membranes were soaked inseparate solutions of glutaralde-

hyde, then glucose oxidase for enzyme immobilization. Figure 3.6 shows VB12 release profiles


of glucose-responsive composite membranes in which the glucose concentration was increased

in solution from 0 mg/dl to 400 mg/dl at 60-minute. Tests showed an increase of VB12 perme-

ability in response to an increase of glucose levels from 0.7±0.5×10−5 cm/min to 6.2±0.5×10−5

cm/min.

Figure 3.6: Diffusion testing of glucose-responsive composite membranes (40% w/w nanopar-ticles, 5% chitosan and 10% DBS). Error bars represent standard deviation.

Figure 3.7 shows VB12 release profiles of glucose-responsive devices. Glucose concentra-

tion was increased from 0 mg/dl to 400 mg/dl at 60-minute in the surrounding medium. The

permeation tests showed a large increase of VB12 diffusion out of the devices, with a perme-

ability of 2.7±0.5×10−5 cm/min at a glucose concentration of 400 mg/dl. These results showed

the potential of the devices for glucose-responsive drug delivery.

3.5 Conclusion

This chapter demonstrated a new type of monolithic glucose-responsive drug-delivery micro

device. Diffusion testing demonstrated the capability of these devicesfor glucose-responsive

release, with an increase in release rates in response to increasing glucose levels. This device

concept may serve as a platform technology for intelligent drug delivery in response to environ-


Figure 3.7: Glucose-responsive permeation of VB12 through glucose-responsive devices (n=3)(the area for permeation is 0.78 cm2) with 5% w/w chitosan molecules, 10% w/w DBS, and40% w/w nanoparticles. Error bars represent standard deviation.

mental glucose variations. Further work will continue the optimization and fine-tuning of the

composite membrane parameters as well as the quantificationof insulin stability and delivery

kinetics through this glucose-responsive micro device.

Chapter 4

Glucose-Responsive Drug Delivery Micro

Device (Type II)

4.1 Introduction

Previously, in Professor Wu’s lab, a new glucose-responsive membrane by introducing bio-

inorganic nanohybrid composite and chemical crosslinkingof albumin with enzymes was de-

veloped [132]. This new nanohybrid material has been successfully applied as a glucose-

responsive plug in a small prototype implantable device as acontrolled insulin delivery system

[133]. However, owing to the soft nature of this membrane, the material would not be suitable

for use as a free membrane in relatively large areas.

In this chapter, we present: (1) a glucose-responsive composite membrane with enhanced

mechanical strength in which the albumin based composite membrane was supported with

PDMS micro grids; and (2) a monolithic glucose-responsive drug-delivery micro device by

integrating the composite membrane and a micro drug reservoir. Membrane responsive release

profiles were quantified by testing model drug diffusion (bovine insulin), with an increase in

release rates in response to increasing glucose levels.In vitro release profiles of the glucose-

responsive micro devices were measured with a marked increase in release rates of insulin

33

Chapter 4. Glucose-Responsive Drug Delivery Micro Device (Type II) 34

when glucose levels in the surrounding medium increased from 100 to 300 mg/dl. Blood

glucose measurements in STZ-diabetic rats were also conducted, recording normoglycemic

glucose levels in diabetic rats with micro device implantation, which further demonstrated the

feasibility of the micro devices for glucose-responsive insulin delivery.

The project entitled “type II glucose-responsive drug delivery micro devices” is a collab-

orative project among three research laboratories at the University of Toronto: Prof. Y. Sun

(Department of Mechanical and Industrial Engineering), Prof. S. Wu (Department of Phar-

maceutical Sciences) and Prof. A. Giacca (Department of Physiology). M. Chu (Professor

Wu’s group), C. Gordijo (Professor Wu’s group) and J. Chen (Prof. Sun’s group) synthesized

and characterized glucose-responsive membranes with PDMDgrids. J. Chen, M. Chu and C.

Gordijo fabricated and characterized glucose-responsivedrug delivery micro devicesin vitro.

S. Chiang (Prof. Giacca’s group), M. Chu, J. Chen, K. Koulajian (Prof. Giacca’s group) and

C. Gordijo characterized micro device performancein vivo.


As shown in Figure 4.1, a PDMS drug reservoir and a layer of PDMS grid provide physical

support for the composite membrane. In this composite membrane, enzymes (e.g., glucose

oxidase) are crosslinked with the albumin macromolecules,creating the base membrane. The

embedded hydrogel nanoparticles detect and respond to local pH changes caused by gluconic

acid resulting from the oxidation of glucose, acting as intelligent nanovalves. Corresponding

volumetric swelling and shrinking of the nanoparticles control the pore size of the membrane

and thus drug-release rate.


Figure 4.1: Cross-sectional schematic for PDMS grid-gel micro device embedded with glucoseoxidase and pH-responsive nanoparticles for glucose-responsive insulin delivery. Reproducedwith permission from [134].


4.3.1 Preparation and Characterization of Glucose-Responsive Membranes

with PDMS Grid Backbone

To fabricate glucose-responsive membranes, standard softlithography was first used for form-

ing a PDMS grid. SU-8 pillars were obtained via standard photolithography as a mold master

(Figure 4.2(a)). PDMS was spin coated on the SU-8 mold masterand the cured PDMS mem-

brane with a thickness of 150µm was peeled off from the substrate, producing a PDMS grid

(Figure 4.2(b)). PDMS grids were modified by oxygen plasma treatment and then soaked in 0.1

M aminopropyl trimethoxysilane solution for 24 hours at room temperature to produce primary

amine groups. A mixture of pre-cured membrane solution (albumin macromolecules, glucose

oxidase, pH-responsive hydrogel nanoparticles and glutaraldehyde molecules) was transferred

onto the modified PDMS grids through dip coating (Figure 4.2(c)) and fully cured (Figure

4.2(d)). Figure 4.2(e) shows prototypes of synthesized glucose-responsive composite mem-

branes.


Figure 4.2: (a) and (b) Standard soft lithography to form PDMS grids. (c) Pre-cured mem-brane solution is transferred onto the PDMS grids through dip coating. (d) Fully cured glucose-responsive membrane with PDMS grids. (e) Prototypes of glucose-responsive composite mem-branes. Reproduced with permission from [134].

The permeation of bovine insulin (Mw = 5808) through the membrane was determined as

a function of glucose concentration using the side-by-sidediffusion cell system. Briefly, the

membrane was placed between two cells: a donor cell containing insulin in pH 7.4 PBS and

a receptor cell filled with the same solution but without the insulin (release medium). After

1.5-hour and 3-hour the glucose concentration in the cells was increased to 200 mg/dl and 400

mg/dl, respectively, by adding aliquots of highly concentrated glucose solution. During all the

course of the experiment the release medium was continuallypumped to a UV-flow cell and

the insulin permeation was determined by measuring insulinabsorbance automatically every

10 minutes atλ = 276 nm using an UV spectrometer. The rate of insulin permeation was

determined by the slope of the curves.


vices with Enhanced Mechanical Strength

To fabricate glucose-responsive micro devices, firstly a PDMS reservoir was made via standard

soft lithography using a metal mold master with a thickness of 2 mm (Figure 4.3(a)). After the

PDMS drug reservoir was peeled off from the mold master, the PDMS grids described earlier

were bonded with the drug reservoir via a thin layer of pre-cured PDMS as an adhesive layer


Figure 4.3: (a-d) Microfabrication steps for forming glucose-responsive micro devices withenhanced mechanical strength. (e) Prototypes of fabricated glucose-responsive micro devices.Reproduced with permission from [134].

(Figure 4.3(b)). The PDMS drug reservoir with PDMS grids wasmodified by oxygen plasma

treatment and then soaked in 0.1 M aminopropyl trimethoxysilane solution for 24 hours at

room temperature to produce primary amine groups. The mixture of the composite membrane

solution was cast on a Teflon plate and transferred to the activated PDMS grids via contact

printing (Figure 4.3(c)) and fully cured (Figure 4.3(d)). Figure 4.3(e) shows prototypes of

fabricated glucose-responsive micro devices.

In vitro insulin release from the micro devices was determined as a function of glucose

concentration in pH 7.4 PBS solution at 37C. Micro devices filled with bovine insulin (10

and 50 mg/ml) were individually placed in vials containing 5 ml of 100 mg/dl glucose in

the PBS medium. Vials were sealed and kept under constant mixing in a mini blot mixer.

After 2.5 hours glucose concentration was increased to 400 mg/dl. The same procedure was

repeated in a subsequent alternated cycle. The insulin release was determined by measuring

insulin absorbance manually every 30 minutes atλ = 276 nm. The rate of insulin release was

determined by the slope of the curves.


In order to further quantify the performance of the micro devices, animal testing was con-

ducted with the procedures as follows: Sprague-Dawley ratswere injected with 65 mg/ml

streptozotocin to induce diabetes before device implantation. Blood glucose was taken before

STZ-injection and after to ensure beta cell deficiency. Surgery was performed on the inner

belly of the rats, and insulin-filled micro devices (25 and 50mg/ml) were implanted in the

intraperitoneum. After device implantation blood glucoselevels were monitored to evaluate

the performance of glucose-responsive insulin delivery micro devices.

In vivo glucose challenges were conducted as follows: diabetic rats were implanted with

one micro device (25 mg/ml) and allowed to rest for 48 hours to confirm the decrease of blood

glucose to normal levels by the action of the implanted devices. Rats were then given an

injection of glucose (1 g/kg, 50% dextrose), and their blood glucose levels were continuously

monitored for 90 minutes.


Figure 4.4(a) and (b) show a representative set of permeability testing data, which demon-

strated the glucose responsiveness of the composite membranes. As shown in the figure, when

the glucose concentration in the medium was increased from 100 mg/dl to hyperglycemic lev-

els (200-400 mg/dl) the insulin permeation rate increased from 0.0015 mg/h/mm2 to 0.0022

mg/h/mm2 and 0.0046 mg/h/mm2, respectively (1.5-fold and 3.1-fold increase in insulin per-

meation, respectively).

Figure 4.5(a) and (b) showin vitro insulin release profiles of three-cycle tests of micro

devices with insulin concentration of 10 mg/ml (n=7). There was good glucose responsiveness,

with slight degradation over time among three cycles. Insulin release cycles from 1 to 3 had

P400/P100 ratios of 2.53, 2.28 and 2.01, respectively (see Table 4.1). Initial permeability

to insulin increased slightly from 4.56, 5.12 to 5.60×10−5 cm2/s from cycles 1 to 3. Final

permeability (400 mg/dl) reached 1.15, 1.17 and 1.13×10−4 cm2/s from cycles 1 to 3.


Figure 4.4: Glucose-responsive permeation of insulin through membranes (n=3) with inte-grated PDMS backbone. (a) Profile of insulin permeated across the membrane in response tochanges of step-wise glucose concentration variations (100, 200, and 400 mg/dl, respectively).(b) Permeability of insulin as a function of glucose concentration calculated from the slopes ofthe curves. Error bars represent standard deviation. Reproduced with permission from [134].

Table 4.1: Permeability comparison of three cycles of microdevices using insulin concentra-tion of 10 mg/ml at glucose concentration of 100 mg/dl and 400 mg/dl.

Cycle P400 P100 P400/P100

1 1.15× 10−4 cm2/s 4.56× 10−5 cm2/s 2.532 1.17× 10−4 cm2/s 5.12× 10−5 cm2/s 2.283 1.13× 10−4 cm2/s 5.60× 10−5 cm2/s 2.01

In terms of the total insulin permeated, it is roughly 1.1 units per cycle. This translates

to about 3.3 units over three cycles, which is reasonable considering the lethal dose that rats

can tolerate per day (<5 units). Also, insulin solutions retrieved from micro devices afterin

vitro testing showed negligible signs of aggregation (>90% transmittance) and yielded iden-

tical retention peaks to fresh insulin samples when analyzed via High-Performance Liquid

Chromatography (HPLC).

Althoughin vitro testing using 10 mg/ml insulin showed good cyclic repeatability of glucose-

responsive insulin release with a P400/P100 ratio of>2, the lower insulin reservoir concentra-

tion led to a lower net insulin permeation (∼15µg/hour, lower than the previous device reported


Figure 4.5: (a) Three cycle permeation of insulin micro devices (10 mg/ml). Glucose levelschanged from 100 mg/dl (0-2.5h) to 400 mg/dl (2.5h-5h) in PBS 7.4 buffer. Error bars representstandard deviation (n=7). (b) Permeability of insulin micro devices from normoglycemic (100mg/dl) to hyperglycemic (400 mg/dl) environment. Three cycles displayed with error barsrepresenting standard deviation (n=7).

in our group (55.2µg/hour) [133]). To address this issue, higher insulin concentration of 50

mg/ml was used for further testing. Figure 4.6(a) and (b) showin vitro insulin release profiles

of two-cycle tests for three micro devices with insulin concentration of 50 mg/ml. Insulin re-

lease cycles from 1 to 2 had P400/P100 ratios of 2.66 and 3.66, respectively (see Table 4.2).

Initial permeability to insulin was 3.83 and 2.52×10−5 cm2/s for cycle 1 and cycle 2, respec-

tively and final permeability (400 mg/dl) reached 1.02 and 0.92×10−4 cm2/s for cycle 1 and

cycle 2, respectively.

Table 4.2: Permeability comparison of two cycles of micro devices using insulin concentrationof 50 mg/ml at glucose concentration of 100 mg/dl and 400 mg/dl.

Cycle P400 P100 P400/P100

1 1.02× 10−4 cm2/s 3.83× 10−5 cm2/s 2.662 0.92× 10−4 cm2/s 2.52× 10−5 cm2/s 3.66

Figure 4.7 shows the comparison of insulin permeated between 50 (n=3) and 10 (n=7)


Figure 4.6: (a) Two cycle permeation of insulin micro devices (50 mg/ml). Glucose levelschanged from 100 mg/dl (0-2.5h) to 400 mg/dl (2.5h-5h) in PBS 7.4 buffer. Error bars representstandard deviation (n=3). (b) Permeability of insulin micro devices from normoglycemic (100mg/dl) to hyperglycemic (400 mg/dl) environment. Error bars represent standard deviation(n=3).

mg/ml insulin reservoir concentrations with increase in glucose at 2.5-hour (100 to 400 mg/dl).

According to the data collected, permeability ratios were higher for insulin concentration of 50

mg/ml, in which P400/P100 ratio reached 2.66 and 3.66, compared to 2.53, 2.28 and 2.01 from

insulin concentration of 10 mg/ml. Compared to the 10 mg/ml insulin group, the increase in

net insulin permeated was noted when 50 mg/ml insulin was used, reaching 36µg/hour at 400

mg/dl, which is closer to the previous device design (55.2µg/hour). One thing to note is the

consistency of the permeability, as both 10 and 50 mg/ml reach the same order of magnitude

at 400 mg/dl (1.15 vs 1.02×10−4 cm2/s, respectively), with slightly lower permeability values

at 100 mg/dl for 50 mg/ml insulin compared to 10 mg/ml insulin (4.56, 5.12, 5.60 vs 3.83,

2.52×10−5 cm2/s, respectively). This accounts for the higher ratio obtained for the 50 mg/ml

micro devices. The repeatability at 400 mg/dl is desirable, as the values are consistent for a

good insulin release profile. In addition, analysis of insulin solutions with UV transmittance

showed no aggregation from insulin denaturation (>95% for all samples).


Figure 4.7: Comparison of insulin permeated between 50 and 10 mg/ml insulin reservoir con-centrations with increase in glucose at 2.5 hour (100 to 400 mg/dl). Error bars represent stan-dard deviation (n=3,7).

Figure 4.8: Blood glucose measurements in STZ-Diabetic rats (SZT injection on Day 0 andmicro device implantation on Day 2) as a function of number ofdevices implanted. All datawere collected from micro devices with insulin concentration of 50 mg/ml.


Figure 4.9: Blood glucose measurements in STZ-Diabetic rats (SZT injection on Day 0 anddevice implantation on Day 2) using 25 mg/ml insulin solution. Error bars represent standarddeviation (n=3).

Figure 4.10: Short-term blood glucose profile of STZ-diabetic rats with implanted micro de-vices during glucose challenge. Error bars represent standard deviation (n=6).


Figure 4.8 shows the blood glucose measurements in STZ-diabetic rats (SZT injection on

Day 0 and device implantation on Day 2) as a function of numberof devices using insulin

concentration of 50 mg/ml. With the control STZ-rats, high blood glucose levels were ex-

pected after SZT injection, with most values reaching above20 mM. The blood glucose of

the one-device rat dropped significantly, reaching roughly5 mM glucose over 5 days, indi-

cating moderate hypoglycemia. The two-device and four-device rats showed a more severe

drop in blood glucose, with blood glucose levels below 4 mM for 5 days, indicating severe

hypoglycemia.

To deal with the issue of hypoglycemia, single device implantation with the insulin concen-

tration of 25 mg/ml was conducted in STZ-Diabetic rats (see Figure 4.9). After device implan-

tation, compared to the 50 mg/ml group, the blood glucose dropped from roughly 25 mM to 10

mM glucose, which gave a better blood glucose profile and maintained normoglycemia for 5

days. Figure 4.10 shows preliminary glucose challenge results. Once rats received the glucose

injection their blood glucose levels immediately increased to hyperglycemic levels (with the

peak of∼ 18 mM). Within one hour, blood glucose levels came back to thenormoglycemic

stage (lower than 10 mM), demonstrating the glucose-responsiveness of the micro devices.

4.5 Conclusion

In summary, this chapter demonstrated a new monolithic glucose-responsive drug-delivery mi-

cro device with enhanced mechanical strength by integrating PDMS microgrids as backbone.

In vitro release characterization proved the concept and the capability of the micro devices

for glucose-responsive drug delivery.In vivo release testing verified the feasibility of the de-

vices for short-term diabetic management. Further work will continue the optimization and

fine-tuning of the composite membrane parameters and conduct in vivo testing of the glucose-

responsive micro devices by optimizing implantation locations.

Chapter 5

Electrodeformation

5.1 Introduction

Differences in electromechanical properties of single cells have been correlated with patho-

physiological states such as cancer. The vast majority of existing micro devices are only ca-

pable of characterizing one biophysical marker: either electrical impedance or mechanical

stiffness. Simultaneous characterization of both impedance andelastic modulus in single cells

is expected to provide more accurate diagnostic outcomes. The objective of this study is to

develop a microfluidic platform for characterizing the electromechanical properties of single

cells.

This chapter presents the use of electrodeformation as a method for single-cell electrome-

chanical characterization in which mechanical propertiesof SiHa and ME180 cells (two cervi-

cal cancer cell lines) were quantified as a function of cell electrical properties. Electrodeforma-

tion was first demonstrated in 1984 when Engelhardt et al. reported the relationship between

applied voltages and corresponding deformation ratios of erythrocytes [101]. Later, they used

the same setup to calculate the elastic modulus of erythrocyte plasma membranes by treating

cells as conductive spheres (they neglected membrane polarization) [102]. Zimmermann et

al. reported the relationship between applied voltages andcorresponding electrodeformation

45

Chapter 5. Electrodeformation 46

ratios of cells via microfabricated electrodes, however, without interpreting raw data (voltage-

deformation) into cells’ Young’s modulus [107]. In a recentstudy, MacQueen et al. calculated

the elastic properties of cells using the Clausius-Mossotti factor with the effective dipole mo-

ment assumption [104], which is only valid when the scale of the electric field nonuniformity

is large compared to cell dimensions. In such case, the cell in the electric field is treated as an

infinitesimal charged particle whose presence has no disturbance on the electric field between

two microelectrodes. In the case of electrodeformation with a cell settled down on one elec-

trode tip (equilibrium location), it is under a highly non-uniform electric field and the effective

dipole moment approximation can lead to significant errors [135].

Although electrodeformation was demonstrated more than two decades ago, the technique

was only used in a limited number of studies because several critical questions remain unan-

swered. For example, the effect of cells’ electrical property variations on the electrical forces

experienced by the cells is still not well understood. Furthermore, due to the complexity of

the physical phenomena involved, there is no direct closed-form mathematical expression that

relates the applied electric field to the cellular mechanical stiffness.

In this study, single cells were deformed under an applied ACelectric field and correspond-

ing deformations were measured under certain experimentalconditions. Numerical simulations

were used to evaluate the applied electrodynamic forces based on the Maxwell stress tensor for-

mulation, which is suitable to treat a wide range of applied electric fields. In these simulations,

the relationship between applied voltages and deformationratios of cells with different Young’s

moduli was investigated. By comparing the measured morphological changes with those ob-

tained from numerical simulations, we were able to quantifyYoung’s modulus of SiHa cells

(601±183 Pa) and ME180 cells (1463±649 Pa). These values were consistent with Young’s

modulus values (SiHa: 400±290 Pa and ME180: 1070±580 Pa) obtained from conventional

micropipette aspiration.



When a cell suspended in a conductive medium is subjected to an electric field (Figure 5.1),

charges are trapped on cell surface and therefore an electrodynamic force distribution is applied

on the cell. If the electric field is non-uniform and the relative polarizability of the cell is higher

than that of the medium, this force distribution has a net resultant leading to cell translation

towards areas of higher electric fields (i.e., positive dielectrophoresis). In such case, the cell

moves and settles down on one electrode tip (i.e., the highest electric field region) where the

electrodynamic force on the cell is balanced out. Although the resultant force distribution

in the plane of motion is zero at this location, the distributed forces on the two cell halves

lead to elongation of the cell in a phenomenon called electrodeformation. The amount of cell

deformation induced depends on the magnitude of the electrodynamic forces generated (which

in turn depends on the applied electric field and the cell-medium electrical properties) and on

the cell stiffness.

Figure 5.1: Schematic of positive dielectrophoresis (pDEP) and electrodeformation. Repro-duced with permission from [149].



Unless otherwise indicated, all chemicals were obtained from Sigma-Aldrich (Oakville, ON,

Canada) and cell-culture reagents were from American Type Culture Collection (ATCC, Man-

assas, VA, USA). Materials required for device fabricationincluded indium-tin oxide (ITO)-

coated glass substrates (Delta Technologies Ltd, Stillwater, MN, USA), Shipley S1818 pho-

toresist and MF321 developer (Rohm and Haas, Marlborough, MA, USA). Bovine serum albu-

min (BSA) (Invitrogen Canada, Burlington, ON, Canada) was used in the electrodeformation

experiment.

5.3.1 Experimental Procedures

Device fabrication

ITO was chosen as the electrode material because it is transparent and facilitates inverted mi-

croscopy imaging. Microelectrodes were fabricated in the clean room facility of the Emerging

Communications Technology Institute at the University of Toronto. Glass slides coated with

200 nm ITO were cleaned in Acetone, Methanol and DI water, anddried on a hotplate (30 min

@ 150C). A 200 nm thick layer of silicon dioxide was deposited on ITO glass slides (5 min

@ 400C, deposition rate: 40 nm/min) using a plasma enhanced chemical vapor deposition

system (Oxford Instruments, UK) (see Figure 5.2).

Shipley S1818 photoresist was spun on the slide (4000 rpm for45 sec), soft-baked (1 min

@ 115C), and exposed to UV light (10 sec, 16 mW/cm2, 365 nm) through a chrome-on-glass

mask (University of Alberta Nanofabrication Facility, Alberta, Canada) using a Karl Suss MA6

mask aligner (Garching, Germany). Slides were then developed in MF321 developer for 60 sec,

and finally hard baked (1 hour @ 120C).

Silicon dioixde without protection from the patterned photoresist was etched away using

an inductive coupled plasma/reactive ion etching system (Trion Technology, FL, USA) with

CHF3 as the etchant gas (2 min, etch rate: 100 nm/min). The exposed ITO was then etched


Figure 5.2: Fabrication steps for ITO based microelectrodes for electrodeformation. Repro-duced with permission from [149].

away in a solution (HCl:HNO3:H2O 55: 7.5: 32.5 mL) to pattern the microelectrodes (2 min

@ 50C). The residual photoresist was removed in Acetone and the residual silicon dioixde

was removed by using CHF3 again as mentioned before.

Preparation of cell suspension

SiHa cells and ME180 cells (adherent human cervical carcinoma cell lines) were purchased

from American Type Culture Collection (Manassas, VA, USA) and maintained in Prof. D. Hed-

ley’s lab (Ontario Cancer Institute) with Eagle’s Minimum Essential Medium supplemented

with 10% fetal bovine serum and McCoy’s 5a Medium Modified supplemented with 10% fe-

tal bovine serum, respectively. Cells were cultured on tissue culture-treated polystyrene flasks

and immediately prior to an experiment, cells were trypsinized, centrifuged and resuspended in

isotonic sucrose solutions of 10.2% (weight to volume) plus0.01% BSA. Sucrose, extensively

used in experimental setups requiring positive dielectrophoresis (DEP) [136, 137], was used as

the cell suspension medium for its low conductivity. 0.01% BSA was added to decrease the

adhesion between cells and the substrate in the experiment [138].


Device operation

A droplet of the suspending solution was pipetted onto the electrode of the micro device and

single SiHa or ME180 cells were placed on the tip of one of the electrodes, using a home

developed automatic robotic manipulation platform. Robotic placement of single cells on the

electrode eliminated the need for single cell trapping techniques. Rectangular AC signals were

generated from a function generator (Model 4040 B and K Precision Corp, CA, USA) for

electrodeformation experiments (Figure 5.3). A minimal ACsignal (3 volts for the functional

generator we used) was applied to attract single cells to thetip of one electrode (equilibrium

location). If no cell lysis was noticed, the applied voltagewas increased 2 volts per step and

kept steady for 30 sec, with deformation pictures of cells taken. Then the voltage was increased

again in the same manner until cell lysis was noticed. Three different frequencies of 500 kHz,

1 MHz, and 5 MHz were used in this experiment.

Figure 5.3: The experimental setup for cell electrodeformation testing. A cell is directly placedon top of the electrodes. Rectangular AC signals are applied, and cell deformation images arecaptured and processed. Reproduced with permission from [149].

In order to quantify the geometric differences in electrodeformed cells, a sub-pixel ellipse

extraction algorithm was developed to process the capturedimages. The procedure consists of


a sequence of standard image processing steps adapted to thecontext of cell electrodeforma-

tion (such as smoothing, thresholding, edge detection, followed by a Hough transform) [139].

The direction and the lengths of deformation along the semimajor and semiminor axes were

obtained from the algorithm that calculates the deformation ratio.

Conventional micropipette aspiration for cell mechanicalcharacterization

To verify our electrodeformation technique, conventionalmicropipette aspiration experiments

were conducted on SiHa and ME180 cells. In the setup, a borosilicate glass micropipette tip

(5 µm diameter) was held by a micromanipulator (Sutter Instrument Company, CA, USA)

mounted on an inverted phase-contrast microscope. Attached to the pipette glass tube was an

in-house voltage-controlled vacuum source generator (a minimum pressure resolution of 8 Pa).

The experiment started with the submersion of the micropipette tip inside the cell-containing

medium and the positioning of the tip close to the surface of atarget cell. Then, a small nega-

tive pressure (usually 20-50 Pa) was applied to immobilize the cell with a complete seal. From

this reference state, subsequent larger suction pressureswere then applied and images of the

aspirated cell were captured. The Young’s modulus of the aspirated cell was estimated from

a common biomechanics model that approximates a cell as an elastic half-space solid (linear,

homogenous and incompressible) [140].

5.3.2 Numerical Analysis

Extensive simulations were conducted using the finite element analysis package COMSOL 3.4

(Burlington, MA, USA) to quantify the Young’s modulus from experimental data (voltage-

deformation). First, the electric field was calculated in the cell vicinity, and electrodynamic

forces exerted on the cell were computed by integrating the Maxwell stress tensor over the

cell surface. Second, a value of Young’s modulus of the cell was assumed, and the calculated

electrodynamic forces were used as a load to calculate cell deformation. Finally, the calcu-

lated deformations at different values of Young’s modulus were compared with experimental


results at the same conditions and an approximate value of the Young’s modulus of the cell was

extracted.

Geometrical parameters and physical properties

Figure 5.4(a) and Table 5.1 show the electrode and cell dimensions used in this study. The

optimum overall dimensions of the surrounding medium were determined by evaluating a se-

ries of cases with different lengths, widths, and heights. We started with a large model and

then reduced the model size gradually until a size is reached(see Figure 5.4(b)-(d)), which is

large enough to simulate infinite space with reasonable accuracy without unnecessary waste of

computational time. The optimum size had the following dimensions: length 100µm, width

60 µm and height 50µm. Since the model is symmetric, half the geometry was simulated to

minimize the number of elements.

Table 5.1: Electrode dimensions and relevant parameters used in numerical simulation. Repro-duced with permission from [149].

Parameter Value

Electrode length (le) 40µmElectrode width (we) 30µmElectrode tip angle (θ) 45

Electrode tip gap (ge) 20µmElectrode height (he) 0.2µmCell diameter (dc) 10µmCell membrane thickness (tc) 10 nmCell center height (hc) 5.3µmSimulation model length (ls) 100µmSimulation model height (hs) 50µmSimulation model width (ws) 60µm

Since exact electrical properties of SiHa and ME180 cells are not known, we simulated a

range of electrical properties of cells reported previously in the literature, to determine their

effects on generated electrodynamic forces. Ranges of cell electrical properties tested were

as follows [141-148]: membrane relative permittivityεmembraneof 10, 20 and 30, cytoplasm


Figure 5.4: (a) Schematic of the numerical model used in the simulations. Half geometry wassimulated to reduce mesh size. All variables are defined withspecific values listed in Table 5.1.(b-e) Electrodynamic forces (integration of the Maxwell stress tensor around cell membrane inthe z direction) as a function of length of the simulation model ls, width of the simulation modellw, height of the simulation model lh and the number of elements at the following condition:electric field of 1 MHz at 1 Vp−p and the following electrical properties were used:σmedium=1mS/m, σcytoplasm=0.2 S/m, εmedium=εcytoplasm=80, εmembrane=20. A mesh independent solutionwas achieved at 40,000 elements. (f) A picture of meshing with 40 000 elements. Reproducedwith permission from [149].


relative permittivityεcytoplasmof 40, 80 and 120, and cytoplasm conductivityσcytoplasmof 0.1,

0.4 and 0.7 S/m. In the electric field simulation, a quasi-static electricmodel (AC/DC module)

was used, with the governing equations and boundary conditions shown as follows.

Governing equations

In the case of a cell exposed to a non-uniform electric field, the electromechanics of the cell is

modeled as an electrodynamic force exerted upon a lossy dielectric spherical shell containing

a linear and isotropic conductive sphere, which is submersed in a lossy dielectric medium.

The electrostatic characteristics were obtained by solving the equation of continuity for the

conduction and displacement currents by explicitly showing its frequency dependence,

−∇ · ((σ + jωεrε0)∇φ) = 0 (5.1)

whereσ denotes the electrical conductivity of the cell,ω is the angular frequency of the driving

field, ε=εrε0 is the permittivity (εr is the relative permittivity of the medium andε0 is that of

vacuum) andφ is the electric potential. The electric fieldE and the displacementD can be

obtained from the gradient of the potentialφ,

E = −∇φ (5.2)

D = εrε0E (5.3)

The electrodynamic forceF assuming negligible magnetic contributions, upon the cellvolume

V, enclosed by a closed surfaceS, due to the applied external electric fieldE, at each point on

S, is given by

F =∫

v[ε(∇ · E)E + ε(E · ∇)E − 0.5∇(εE · E)]dV (5.4)

Equation (5.4) can be further simplified by using a tensor notation and transforming the volume


integral to a surface integral via the Gauss theorem. The resulting equation for the force per

unit area exerted on the surface of the cell becomes

F =∮

ST · ndS (5.5)

Ti j = ε(EiE j − 0.5δi j E2) (5.6)

with Ti j as the nine components of the Maxwell stress tensor (the indicesi andj refer to pairs

of x, y, and z axes andδi j is the Kronecker delta). The three diagonal elements ofTi j are known

to represent pressures while the off-diagonal elements represent shears. The unit vectorn is

normal to the surface.

It is important to note that by employing the Maxwell stress tensor, there is no underlying

assumptions on the non-uniformity of the electric field as isneeded for the effective dipole

moment method frequently used in dielectrophoretic force calculation [135]. As a result, our

approach is more general and can more accurately predict electrodynamic forces on the cell

in regions of high field nonuniformity as is the case when the cell is at the tip of one of the

electrodes.

Boundary conditions

The driving potential was applied to the left electrode while ground potential was applied to

the right one. The other external boundaries were electrically insulated (n·J=0) to meet the

requirement of charge conservation, Equation (5.1), whereJ is the current density. Boundary

conditions on the plane of symmetry were set to satisfy Equation (5.1). At interfaces between

the cell surface and the internal/external medium, continuity of the electric fieldE, electric

displacementD, and current densityJ were applied according to

n · (D1 − D2) = ρs (5.7)


n × (D1 − D2) = 0 (5.8)

n · (J1 − J2) = 0 (5.9)

whereρs is the surface charge density.

Numerical method

We used a Lagrange-quadratic element type and the PARDISO direct solver for electric simu-

lations and the GMRES iterative stationary solver with geometric multigrid preconditioner for

mechanical simulations. The relative tolerance used as a convergence criterion was

ρ|M−1(b− Ax)| < tol.|M−1b| (5.10)

whereρ is the factor in error estimation (ρ=400 in the current study),M is the preconditioner

matrix,Ax=b is the system of equations to be solved, andtol.=10−6 is the relative tolerance.

Mesh independence

In initial tests, different meshes were employed to optimize the mesh size that yields a solution

independent of discretization. Figure 5.4(e) shows the effect of number of elements on the

electrodynamic forces acting in the z-direction, as an integration of the Maxwell stress tensor

along the cell surface. As shown, convergence was reached atabout 40 000 elements (Figure

5.4(f)).


Electrodeformation can only be observed under positive dielectrophoresis, where the cells an-

chor on one of the electrodes under the electrodynamic forces in the negative z-direction (i.e.,

downward forces). To achieve positive dielectrophoresis with the highest electrodynamic force

possible, proper choice of medium properties and applied frequency is crucial. A low medium


conductivity is required to make the cell more polarizable to induce positive dielectrophoresis

[135]. The lower the medium conductivity relative to that ofthe cell cytoplasm, the higher

the electrodynamic force, which is due to the larger difference in the electric field inside and

outside the cell. The low medium conductivity also results in a larger voltage drop across the

medium rather than across the cell membrane, which decreases the possibility of electrolysis.

Choice of the frequency of the applied electric field is of utmost importance. In DC or low

frequency fields, the dielectric cell membrane acts like an insulator and bears the most of the

voltage drop resulting in cell lysis at low applied potentials. Whereas at very high frequencies

when the effect of permittivity dominates over that of conductivity, the cell membrane becomes

electrically transparent, making the cell behave more likea homogeneous cytoplasm, with the

same permittivity as the surrounding medium resulting in smaller electrodynamic forces. Thus

we used a frequency range of 100 kHz to 10 MHz which generates high electrodynamic forces

and results in a shorter time duration per cycle for charge build-up on the cell surface, and thus

reduces the electrolysis effect.

5.4.1 Cell Elongation

Three frequencies: 500 kHz, 1 MHz, and 5 MHz were chosen to deform cells electrically with

the surrounding medium conductivity of 1 mS/m (see Figure 5.5(a)). The deformation ratio

is defined as the ratio between the elongation of the cell parallel to the applied electric field

direction and the original diameter of the cell before electrodeformation. In the experiment,

the applied voltage was increased in steps of 2 volts and keptsteady for 30 sec per step, with

cell deformation pictures recorded until electrolysis occurred.

As shown in Figure 5.5(a), the cell lysis voltage increased from 19 volts to 25 volts as the

applied frequency was increased from 500 kHz to 5 MHz, which agreed well with the theoreti-

cal analysis on cell electrolysis discussed previously. Under the same voltage, the deformation

ratios of cells at 500 kHz and 1 MHz were comparable while the deformation ratios of cells at

5 MHz were significantly lower, suggesting that 5 MHz is beyond the upper frequency limit


to generate highest electrodynamic forces possible. This was confirmed by simulation results

which show a decline in the electrodynamic force at frequencies higher than 1 MHz regardless

of cell electrical properties (see Figure 5.5(b)).

When subjected to electric fields, both SiHa and ME180 cells showed elongation parallel to

the applied electric field lines. Deformation ratios of SiHaand ME180 cells were respectively

1.066±0.0254 and 1.031±0.0257 at 19 Vp−p indicating a lower stiffness for SiHa cells (see

Figure 5.6).

5.4.2 Effect of Cells’ Electrical Properties

The value of the electrodynamic forces generated on cells cannot be exactly predicted unless

the electrical properties of the cell (i.e., cytoplasm conductivity and membrane permittivity) are

known. Since electrical properties of SiHa and ME180 cells are not known, the electrodynamic

forces were calculated at a range of cell electrical properties of twenty-seven different types of

cells. Twenty-seven independent simulations were performed to include all permutations of the

electrical parametersεmembrane=10,20,30, εcytoplasm= 40,80,120, andσcytoplasm=0.1,0.4,0.7

S/m.

As shown in Table 5.2, for a cell with unknown electrical properties, the simulated electro-

dynamic force fall into the range of 11.54±1.55 nN, by calculating the average and the standard

deviation of the electrical simulation results of 27 cases mentioned above. The simulated maxi-

mum electrodynamic force is 13.45 nN (16.5% higher than the average value) and the minimal

electrodynamic force is 9.17 nN (20.5% lower than the average value). In addition, among

these three studied parameters, membrane permittivity andcytoplasm conductivity have more

significant effects on generated electrodynamic forces than cytoplasm permittivity whose effect

on electrodynamic forces is negligible (see Figure 5.7).

As shown in Figure 5.7, membrane permittivity has the largest effect on the generated elec-

trodynamic force with an increase of 31% when membrane relative permittivity was increased

from 10 to 30. Cytoplasm conductivity has less significant effect on generated forces with an


Figure 5.5: (a) Experimental electrodeformation of SiHa cells as a function of electric fieldstrength with an applied electric field of frequency 500 kHz,1 MHz, and 5 MHz. Sample sizeis 5 cells at 500 kHz (blue), 7 cells at 1 MHz (red) and 5 cells at5 MHz (green). The deforma-tion ratio is defined as the ratio between the elongation of the cell parallel to the applied electricfield direction and the original diameter of the cell before electrodeformation. (b) Numericalsimulations based electrodynamic forces as a function of frequency at 19 Vp−p under surround-ing medium conductivity of 1 mS/m with the following electrical properties:εmembrane=10,σcytoplasm=0.1 S/m, εcytoplasm=40 (red);εmembrane=20, σcytoplasm=0.4 S/m, εcytoplasm=80 (green)andεmembrane=30,σcytoplasm=0.7 S/m, εcytoplasm=120 (blue). Reproduced with permission from[149].


Figure 5.6: Top: Images of electrodeformation of SiHa (a) and ME180 (b) cells as a functionof electric field strength using a cell suspension of sucrosewith 0.01% BSA, electric fieldfrequency of 1 MHz, and electrode gap of 20µm. Applied electric field strength is indicatedin brackets. Bottom: electrodeformation of SiHa and ME180 cells at 19 Vp−p at 1 MHz withelectrode gap of 20µm. Sample size is 7 cells per cell line. Reproduced with permission from[149].


Table 5.2: Simulation results of electroydnamic forces as afunction of cell electrical properties.Reproduced with permission from [149].

εmembrane εcytoplasm σcytoplasm Electrodynamic force (nN)

10 40 0.1 9.3210 40 0.4 9.7010 40 0.7 9.7210 80 0.1 9.2410 80 0.4 9.6910 80 0.7 9.7210 120 0.1 9.1710 120 0.4 9.6910 120 0.7 9.7120 40 0.1 11.6120 40 0.4 12.2320 40 0.7 12.2220 80 0.1 11.5020 80 0.4 12.2120 80 0.7 12.2820 120 0.1 11.1420 120 0.4 12.2220 120 0.7 12.2230 40 0.1 12.2630 40 0.4 13.4530 40 0.7 13.4430 80 0.1 12.5130 80 0.4 13.4230 80 0.7 13.4130 120 0.1 12.2430 120 0.4 13.3430 120 0.7 13.34

increase of 9% when cytoplasm conductivity was increased from 0.1 S/m to 0.7 S/m. The elec-

trodynamic force in this context is the integration of the Maxwell stress tensor over one half of

the cell in the x-direction (i.e., cell elongation direction) at the equilibrium location. A cell’s

equilibrium location was defined as the location on top of theelectrode where the net x-forces

vanish. Since the equilibrium location changed with different electrical properties tested, new

equilibrium points were relocated for each new set of parameters.

5.4.3 Calculation of Young’s Modulus

By comparing the calculated deformations at different values of Young’s modulus with exper-

imental results, the Young’s modulus of the cell was determined. Since a range of electrody-


Figure 5.7: Simulation results of electroydnamic forces for cell deformation as a function ofcell electrical properties. Reproduced with permission from [149].

namic forces was calculated for each case due to the uncertainty in cell electrical properties,

Young’s modulus was calculated as lying between a minimum and maximum value for each

deformation ratio measured.

As shown in Figure 5.8, for a deformed cell with electrical properties unknown, Young’s

modulus values from simulations fall into the following ranges: 2289±299 Pa (deformation

ratio: 1.02), 1115±149 Pa (deformation ratio: 1.04), 743±99 Pa (deformation ratio: 1.06),

557±75 Pa (deformation ratio: 1.08) and 446±60 Pa (deformation ratio: 1.10) respectively,

by calculating 27 values of Young’s modulus (correspondingto 27 cases of different electrical

properties) for a given deformation ratio. Overall, the standard deviations are within 15% of

the averages.

The deformation ratios of individual SiHa and ME180 cells collected from experiments

were used to fit the simulation results as mentioned above forYoung’s modulus calculation.

For each cell with a measured deformation ratio, 27 values ofYoung’s modulus were calculated

and represented by the average and the standard deviation inwhich individual SiHa and ME180

cells showed different Young’s modulus values due to cell heterogeneity (seeFigure 5.9).

Figure 5.10 shows quantified Young’s modulus from electrodeformation (601±183 Pa for


Figure 5.8: Young’s modulus calculation as a function of thedeformation ratio from numericalsimulations. For a given deformation ratio, 27 Young’s modulus values were obtained based onsimulation results, which reflected 27 cases of electrical property variations shown in Table 5.2.The standard deviations (within 15% of the average value) represented the range of Young’smodulus values due to cell electrical property variations.Simulations were conducted with theelectric field of 1 MHz and the surrounding medium conductivity of 1 mS/m. Reproduced withpermission from [149].

SiHa cells and 1463±649 Pa for ME180 cells). The average and the standard deviation of

Young’s modulus were calculated from 189 values per cell line corresponding to 7 experimen-

tally deformed cells in which for each deformed cell, there were 27 Young’s modulus values

due to different electrical properties. Conventional micropipette aspiration was used to verify

Young’s modulus values calculated using electrodeformation with values of Young’s modulus

400±290 Pa for SiHa cells and 1070±580 Pa for ME180 cells (see Figure 5.11).

5.5 Conclusion

This chapter demonstrated the use of electrodeformation ofbiological cells as a method to

quantify single-cell electromechanical properties. Electrodeformation experiments were con-

ducted to deform SiHa and ME180 cells under applied electricfields, in which they were

distinguished based on different deformation ratios. Simulation results demonstrated the effect


Figure 5.9: Young’s modulus values of individual SiHa and ME180 cells determined fromelectrodeformation by means of curve fitting experimental measurement results with simula-tion results. Standard deviation bars represent the effects of cell electrical property variationson single-cell Young’s modulus calculation. Individual SiHa and ME180 cells showed differ-ent Young’s modulus, which represent cell heterogeneity. Reproduced with permission from[149].

Figure 5.10: Comparison between the Young’s modulus valuesof SiHa and ME180 cells de-termined from electrodeformation and micropipette aspiration. Sample size is 7 cells per cellline. Standard deviation bars of electrodeformation are mainly due to the effect of cell electricalproperty variations from numerical simulations and cell stiffness variations among individualcells from experiments. Reproduced with permission from [149].


Figure 5.11: Conventional micropipette aspiration experiments. Images of aspirated (a) SiHaand (b) ME180 cells for different vacuum pressures (the aspirated lengths and the suctionpressure are indicated). Reproduced with permission from [149].

of cell electrical property variations on the relationshipbetween applied voltages and deforma-

tions of cells with different Young’s modulus. By comparing the experimentally measured de-

formations with those obtained from numerical simulations, we were able to quantify Young’s

modulus of SiHa (601±183 Pa) and ME180 cells (1463±649 Pa), which were consistent with

Young’s modulus values (SiHa: 400±290 Pa and ME180: 1070±580 Pa) obtained from con-

ventional micropipette aspiration. Further work will focus on the characterization of single

cells’ electrical and mechanical properties simultaneously by integrating electrodeformation

with impedance measurements to further decouple the combined effect of cells’ electrical and

mechanical properties on their electrodeformed behaviors.

Chapter 6

Impedance Spectroscopy and

Micropipette Aspiration

6.1 Introduction

Differences in electromechanical properties of single cells have been correlated with cancer

[35, 39, 47, 49, 50] and malaria [88, 150-154]. The vast majority of existing micro devices are

only capable of characterizing one biophysical marker: either electrical impedance [29-31] or

mechanical stiffness [155-157]. Simultaneous characterization of both impedance and elastic

modulus in single cells is expected to provide more accuratediagnostic outcomes.

The only device, which was reported to perform both impedance and mechanical charac-

terization of single cells, is complex in both design and microfabrication processes [88]. In this

device, cantilever deformation ratios were used to indicate cell mechanical properties, which

are difficult to be translated into cell’s Young’s modulus. The design electrically suffered from

the problem of leakage currents, and no electrical model is available for the interpretation of

the raw data into cell’s electrical parameters. In terms of microfabrication, four photolithogra-

phy steps, dry and wet etching of silicon, and metal and dielectric deposition and lift-off were

required for device construction.

66

Chapter 6. Impedance Spectroscopy andMicropipette Aspiration 67

To address this issue, this chapter presents a simple microfluidic device capable of simul-

taneous mechanical and electrical characterization of single cells. The device performs two

types of cellular characterization (impedance spectroscopy and micropipette aspiration) on a

single chip to enable cell electrical and mechanical characterization. To investigate the perfor-

mance of the device design, electrical and mechanical properties of MC-3T3 osteoblast cells

were measured. Based on electrical models, membrane capacitance of MC-3T3 cells was de-

termined to be 3.39±1.23 pF and 2.99±0.82 pF at the aspiration pressure of 50 Pa and 100

Pa, respectively. Cytoplasm resistance values were 110.1±37.7 kΩ (50 Pa) and 145.2±44.3 kΩ

(100 Pa). Aspiration length of cells was found to be 0.813±0.351µm at 50 Pa and 1.771±0.623

µm at 100 Pa. Quantified Young’s modulus values were 377±189 Pa at 50 Pa and 344±156

Pa at 100 Pa. Experimental results demonstrate the device’scapability for characterizing both

electrical and mechanical properties of single cells.


The device, consisting of two layers of PDMS, is simple in both design and fabrication (see

Figure 6.1). The application of a low negative pressure traps a single cell at the entrance of

the aspiration channel while the magnitude of this pressurecontrols the cell aspiration length.

Cellular deformation is recorded as a function of increasing pressure while cellular impedance

is measured via two Ag/AgCl electrodes inserted into culture medium. The design ofthis de-

vice overcomes several technological limitations: (1) theleakage current is minimized through

proper sealing between the aspirated cell and the aspiration channel; (2) The electrode polar-

ization problem (impedance profile distortion in the low frequency domain due to the effect

of the electrical double layer) is minimized by using Ag/AgCl nonpolarizable electrodes; (3)

equivalent circuit models are straightforward to establish for determining cellular components

(e.g., membrane capacitance and cytoplasm resistance); and (4) existing micropipette aspira-

tion models enable the quantitative extraction of the Young’s modulus values of cells.


Figure 6.1: Simultaneous electrical and mechanical characterization of single cells. A negativepressure is applied to trap a single cell at the entrance of the aspiration channel. Cell deforma-tions are recorded by imaging. Impedance measurement is conducted via two Ag/AgCl elec-trodes connected with an impedance analyzer. The electrical model of the aspiration channelis represented by Rpipette and Cpipette in parallel. Cellular electrical components are representedby Rcytoplasmand Cmembranein series. Rleak indicates sealing during cell aspiration. Reproducedwith permission from [161].





assas, VA, USA). Materials required for device fabricationincluded SU-8 photoresist (Mi-

croChem Corp., Newton, MA, USA) and 184 silicone elastomer (Ellsworth Adhesives Canada,

Burlington, ON, Canada).

6.3.1 Device Fabrication

The two-layer channel masters (see Figure 6.2) were fabricated in the clean room facility of the

Emerging Communications Technology Institute (ECTI) at the University of Toronto. Glass

slides were cleaned in acetone, methanol and DI water, and dried on a hot plate (30 min @

150C). The first layer of SU-8 (5µm thick) is made of SU-8-5 to form cell aspiration channels,

which was spun on the slide (500 rpm for 5 sec+ 1500 rpm for 30 sec), soft-baked (2 min @

65C+ 5 min @ 95C), and exposed to UV light (7 sec, 16 mW/cm2, 365 nm) through the first

chrome-on-glass mask (University of Alberta Nanofabrication Facility, Edmonton, Alberta,

Canada) using a Karl Suss MA6 mask aligner (Garching, Germany). Slides were then baked

on a hot plate (1 min @ 65C + 2 min @ 95C) to cross-link the exposed SU-8-5 without

development (Figure 6.2(a)).

The second layer of SU-8 (25µm thick) is made of SU-8-25 to form cell loading channels.

SU-8-25 was spin coated on the glass slide covered with the first layer of SU-8-5 (without

development) (500 rpm for 5 sec+ 2000 rpm for 30 sec), soft-baked (3 min @ 65C + 7 min

@ 95C), aligned and exposed to UV light (12 sec, 16 mW/cm2, 365 nm) through the second

chrome-on-glass mask (University of Alberta Nanofabrication Facility, Edmonton, Alberta,

Canada) (see Figure 6.2(b)). Slides were then baked on a hot plate (1 min @ 65C + 3 min @

95C) to cross-link the exposed SU-8-25, developed in SU-8 developer for 60 sec, and finally

hard baked (2 hours @ 175C).


Figure 6.2: Fabrication steps for forming the two-layer PDMS device. Reproduced with per-mission from [161].

PDMS prepolymer and curing agent were mixed at a ratio of 10 : 1, degassed in a vacuum

desiccator, poured on a channel master placed in an aluminumfoil plate, and finally baked in

a convection oven (15 min @125C) (see Figure 6.2(c)). PDMS channels were then peeled

from the SU-8 master and reservoir holes were punched through. Bonding to glass slides was

conducted using a portable corona treater where PDMS piecesand glass slides were treated

with a BD20-AC corona treater (Electro-Technic Products Inc., Chicago, IL, USA) for 30

seconds per piece, pressed together, and baked on a hot plate(1 hour @ 100C) (see Figure

6.2(d)).

6.3.2 Device Operation

MC-3T3 cells (osteoblast) were purchased from American Type Culture Collection (Manassas,

VA, USA) and cultured in GIBCOT M Minimum Essential Medium Alpha (1X) supplemented

with 10% fetal bovine serum. Cells were cultured in tissue culture-treated polystyrene flasks.

Immediately prior to an experiment, cells were trypsinized, centrifuged and resuspended in

GIBCOT M Minimum Essential Medium Alpha (1X) with a concentration of1 million cells per

milliliter. Cell passage generations between p3 and p10 were used in this experiment.

The device was first filled with culture medium. Before cell trapping, an impedance profile

was recorded for reference (Agilent-4294A Impedance Analyzer, Agilent Technologies, Inc.,


Santa Clara, CA, USA). A droplet of cell suspension (MC-3T3 cells) was then pipetted to the

entrance of the cell loading channel. A negative pressure from 50 Pa to 100 Pa generated

from a custom developed pump was applied to trap and aspiratea cell at the entrance of the

cell aspiration channel. Cell images were taken by an inverted microscope (Olympus IX801,

Olympus Canada Inc., Markham, ON, Canada). Impedance data with the frequency range of

100 Hz-1 MHz (excitation voltage: 100 mV) were recorded by the impedance analyzer. After

characterization, a high negative pressure (∼2,000 Pa) was used to completely remove the cell

from the aspiration channel. Thus, the device was ready to perform measurements on the next

cell.

6.3.3 Data Processing

Impedance profile analysis

To interpret the measured impedance data, three electricalmodels are proposed (see Figure

6.3). Model 1 is used to fit the impedance data without cell trapping, in which Rpipette and

Cpipetterepresent the equivalent resistance and capacitance of theaspiration channel (see Figure

6.1). Model 2 and Model 3 are used to model the situation with cell trapping. Rleak in Model 2

represents the cell blockage of electrical field (see Figure6.1), and this model does not include

cell’s electrical components. In Model 3, cellular electrical components are considered, which

are represented as Cmembrane(capacitance of the cell membrane) and Rcytoplasm(resistance of the

cytoplasm) connected in series. In these equivalent circuit models, the electrical double layer

issue was not considered since Ag/AgCl non-polarizable electrodes were used and therefore,

no electrical double layer was produced.

MATLAB programs (MathWork, Natick, MA, USA) were developedto fit impedance pro-

files to the aforementioned electrical models. Since the mathematic expressions of these mod-

els are highly nonlinear, nonlinear least-squares fitting was employed for optimization [158].

In order to address the potential concern of convergence to alocal solution due to the inap-


Figure 6.3: Circuit models proposed to assess impedance data. Reproduced with permissionfrom [161].

propriate choice of a group of initial values, a loop function was used to enumerate the initial

values of interest, subsequent to which nonlinear least-squares curve fitting was conducted for

each case. Optimization results from each group of initial values were then compared to locate

the best curve fitting case and the corresponding optimized parameters.

Image processing of cell aspiration and Young’s modulus calculation

In order to quantify the aspiration length, a sub-pixel contour extraction algorithm was devel-

oped to process the captured images. The procedure consistsof a sequence of image processing

steps adapted to the context of cell elongation (such as smoothing, thresholding, edge detec-

tion, followed by a Hough transform) [139]. The Young’s modulus values of the aspirated cells

were calculated based on the elastic half-space solid model[140].


Differences in impedance profiles between measurements with andwithout single cells are

usually small and sometimes unobservable [96]. For example, the differential single-cell

impedance analysis using hydrodynamic cell trapping, reported in [93] , had only a 20% to

30% impedance difference with and without cell trapping. This small difference can be due to

poor contact between the cell and electrodes, which led to current leakage [47].


To improve cell-electrode adhesion, a microfluidic device was developed where a negative

pressure was used to suck the measured cell into close contact with the electrodes [47]. Dif-

ferences were observed in impedance phase profiles of two cell lines. However, the impedance

amplitude was not found statistically different, which proves the existence of leakage current

although tight cell-electrode adhesion was formed. Another approach to reduce leakage cur-

rent involves surface modification of the electrode and minimization of microelectrode contact

with the electrolyte [96], which resulted in an impedance amplitude increase of approximately

40% in the low frequency domain (several kHz). However, at these frequencies the amplitude

difference is dominated by the presence of an electric double layer suggesting that the observed

increase in impedance amplitude may result from the disturbance of the electrical double layer

(vs. from the cell). It was speculated in [96] that the nanometer gap between a cell and elec-

trodes is the current leakage source and the effect of the electrical double layer due to electrode

polarization can further distort impedance profiles in the low frequency domain.

In order to further reduce the leakage current and remove theelectrical double layer ef-

fect, a micro hole-based chip modified from the vertical patch-clamp technique utilizing the

four-electrode arrangement was proposed [97]. This approach resulted in a greater than 100%

impedance amplitude increase after cell trapping. It can befurther improved by proper control

of the aspiration pressure to form a tight seal between the cell and the microhole, mimicking

the situation used in conventional patch-clamping (i.e., giga-ohm seal).

The microfluidic device developed in this study is based on lateral patch clamping and

micropipette aspiration. Compared to the microhole-baseddevice [97], this design has a few

advantages: (1) accurate control of negative pressure enables a proper seal of the cell with

the aspiration channel to minimize the leakage current issue; (2) the device performs lateral

aspiration rather than vertical aspiration (as in [97]), enabling microscopy measurement of cell

elongation and thus, mechanical characterization while electrical impedance profiles are mea-

sured; and (3) the geometry of lateral aspiration dramatically reduces the capacitive coupling

between the cell loading channel and the aspiration channel, which is important for low-noise


impedance recording.

6.4.1 Impedance Measurement Results and Data Analysis

Figure 6.4 shows the impedance amplitude profiles (Figure 6.4(a)) and phase profiles (Figure

6.4(b)) as a function of frequency, both with and without cell trapping at an aspiration pressure

of 50 Pa and 100 Pa, respectively. Error bars represent standard deviations (n=18). The values

of measured impedance amplitude and phase at ‘no cell trapping’ are steady in the low and

medium frequency domain (433.7±50.1 kΩ and -0.1±0.015 degree at 1 kHz), modeled as a

pure resistor Rpipette. The average values are consistent with those reported by Cho et al. [97].

Since Ag/AgCl electrodes were used in this design, the electrical double layer effect was not

observed. In the high frequency domain, a decrease in both amplitude and phase occurred

(374.7±35.5 kΩ and -25.5±4.8 degree at 1 MHz), indicating the presence of a capacitance,

Cpipette.

Circuit Model 1 was used to fit the impedance profiles at ‘no cell trapping’. The calculated

Rpipette and Cpipette are 434.4±50.3 kΩ and 0.19±0.03 pF. The transition frequency at which

impedance amplitude starts to decrease is roughly 100 kHz, which is much higher than the

transition frequency (∼ 5 kHz) in Cho et al. [97], indicating a lower parasitic capacitance

Cpipette present in this design because of the lateral aspiration structure [53, 63].

The values of measured impedance amplitude and phase at ‘cell trapping at 50 Pa’ are also

steady in the low and medium frequency domain (601.9±54.6 kΩ and -0.19±0.055 degree at

1 kHz) (see Figure 6.5(a) and (b)), modeled as two resistors (Rpipette and Rleak(50Pa)) connected

in series. Rleak indicates cell blockage of the electrical filed at the entrance of the aspiration

channel. In the high frequency domain, a decrease in both amplitude and phase occurred

(414.5±31.3 kΩ and -31.5±3.4 degree at 1 MHz).

Circuit Model 2 and Model 3 were used to fit the impedance profiles at ‘cell trapping at

50 Pa’. Compared to Model 2, Model 3 better fits impedance data(see Figure 6.5(a) and (b)),

proving that the data reflects the properties of the cells. The values Rleak(50Pa), Cmembraneand


Figure 6.4: Measured impedance amplitude (a) and phase (b) at ‘no cell trapping’, ‘cell trap-ping at 50 Pa’ and ‘cell trapping at 100 Pa’ as a function of frequency (n=18). Error barsrepresent standard deviation. Reproduced with permissionfrom [161].


Figure 6.5: Fitting measured impedance data with cell trapping to electrical Model 2 and Model3. Error bars represent standard deviation. Reproduced with permission from [161].

Rcytoplasmusing Model 3 were determined to be 167.5±46.7 kΩ, 3.39±1.23 pF and 110.1±37.7

kΩ (see Figure 6.6).

Figure 6.6: Calculated cell membrane capacitance (a) and cytoplasm resistance (b) of MC-3T3cells (n=18) by fitting experimental results with Model 3 under aspiration pressure of 50 Paand 100 Pa, respectively. Reproduced with permission from [161].

Compared to the patch-clamp technique, Rleak obtained here is relative low. Since the

aspiration channel in this device is not exactly circular, forming effective seal between the

cell membrane and the aspiration channel is more difficult and leakage may have taken place

[99]. The calculated Cmembranevalues are consistent with those (roughly 1µF/cm2) obtained

from patch clamp [159].


The values of measured impedance amplitude and phase at ‘cell trapping at 100 Pa’ are

steady in the low and medium frequency domain (704.2±58.3 kΩ and -0.30±0.007 degree at 1

kHz) (see Figure 6.5(c) and (d)), modeled as two resistors (Rpipette and Rleak(100Pa)) in series. In

the high frequency domain, a decrease in both amplitude and phase occurred (426.6±29.8 kΩ

and -33.5±3.9 degree at 1 MHz).

Circuit Model 2 and Model 3 were used to fit the impedance profiles at ‘cell trapping at 100

Pa’. Similar to the case ‘cell trapping at 50 Pa’, Model 3 better fits impedance data (see Figure.

6.5(c) and (d)), demonstrating that the data reflects cell properties. Rleak(100Pa), Cmembraneand

Rcytoplasmusing Model 3 were determined to be 272.5±57.0 kΩ, 2.99±0.82 pF and 145.2±44.3

kΩ (see Figure 6.6). Rleak(100Pa) is 60% higher than Rleak(50Pa), which indicates a better seal

of the cell with the aspiration channel at 100 Pa and therefore, a further decrease in leakage

current. As the aspiration pressure was increased from 50 Pato 100 Pa, a larger portion of a cell

was aspirated into the aspiration channel, which causes a roughly 30% increase in Rcytoplasm.

As to the 10% Cmembranedecrease, it is also possibly caused by cell shape change. Cmembrane

is represented by two capacitors in series. The first capacitor corresponds to the portion of

the membrane sucked into the aspiration channel, and the second capacitor corresponds to

the portion of the membrane outside the aspiration channel.When the aspiration pressure

increases, there is no change for the effective area of the portion of the membrane sucked into

the aspiration, which is determined by the cross-section ofthe channel. At the same time,

the effective area of the cell membrane outside the channel decreases, causing Cmembraneto

decrease.

It is worth noting that in the impedance characterization micro devices using vacuum aspi-

ration for cell positioning, there is a potential concern that aspiration pressures may modulate

the on-off stages of mechanical-sensitive ion channels on cell membranes, and therefore, affect

impedance profiles. In this study, this concern was not takeninto consideration as in existing

literature [39, 47]. The reason is that the lowest current measured in our experiments was sev-

eralµAs (excitation voltage: 100 mV, the highest impedance amplitude collected<800 kΩ),


which is three orders higher than the typical current level in ion channel experiments (several

nAs only) [53, 57, 59, 63], proving that the effect of ion channels is negligible.

6.4.2 Cell Aspiration Results and Young’s Modulus Calculation

Figure 6.7(a) shows the morphology of a cell before aspiration and at a negative pressure of

100 Pa. Aspiration length of cells (n=18) was found to be 0.813±0.351µm at 50 Pa and

1.771±0.623µm at 100 Pa (see Figure 6.7(b)). Calculated Young’s modulus values of MC-

3T3 cells were 377±189 Pa at 50 Pa and 344±156 Pa at 100 Pa using the infinite half-space

model of conventional micropipette aspiration.

Figure 6.7: (a) Images of cell aspiration at 0 Pa and 100 Pa. (b) Aspiration lengths of cells at50 Pa (0.813±0.351µm) and 100 Pa (1.771±0.623µm) (n=18). Error bars represent standarddeviation. Reproduced with permission from [161].

Since the aspiration channel in this device is not exactly circular, forming perfect seal be-

tween the cell and the aspiration channel is difficult and leakage may take place at the corners

of the aspiration channel. This leakage may result in highervacuum pressure required to as-

pirate cells to a certain distance compared to conventionalmicropipettes, possibly resulting in

Young’s modulus values lower than the use of conventional glass micropipettes. Interestingly,

the Young’s modulus values of 377±189 Pa at 50 Pa and 344±156 Pa at 100 Pa determined


using this micro device are not much lower than those obtained from the use of conventional

micropipettes on the same type of cells (555±183 Pa) [160]. The insignificant difference can

be attributed to the fact that the 5µm microchannels exhibit a vaulted rather than a strictly

rectangular cross section, alleviating the leakage problem to an extent compared to perfect

rectangular aspiration channels [53, 63].

6.5 Conclusion

This chapter demonstrated the use of a micro device for both electrical and mechanical charac-

terization of single cells using impedance spectroscopy and micropipette aspiration on the same

device. Impedance profiles and cell aspiration lengths wererecorded by an impedance analyzer

and microscopy imaging. Membrane capacitance of MC-3T3 cells was found to be 3.39±1.23

pF and 2.99±0.82 pF at the aspiration pressure of 50 Pa and 100 Pa, respectively while cy-

toplasm resistance values were determined to be 110.1±37.7 kΩ (50 Pa) and 145.2±44.3 kΩ

(100 Pa). Quantified Young’s modulus was 377±189 Pa at 50 Pa and 344±156 Pa at 100 Pa.

Chapter 7

Impedance Spectroscopy and Constriction

Channel

7.1 Introduction

The device reported in the previous chapter is capable of performing two types of cellular

characterization (impedance spectroscopy and micropipette aspiration) on a single chip to en-

able cell electrical and mechanical characterization. However, this design suffers from low

throughput, which is incapable of collecting statistically significant data. To address this issue,

this chapter presents a microfluidic device enabling rapid electrical and mechanical character-

ization of single cells. Cells are aspirated continuously through a constriction channel with

cell elongations and impedance profiles measured simultaneously. Cell transit time through

the constriction channel and the impedance amplitude ratioare quantified as cell’s mechanical

and electrical property indicators.Using this method, statistically significant datasets can be

obtained for single cells.

80

Chapter 7. Impedance Spectroscopy and Constriction Channel 81


This study presents a microfluidic device for single-cell electrical and mechanical character-

ization using impedance spectroscopy and constriction channel (see Figure 7.1). Cells are

aspirated continuously through a constriction channel while cell elongations and impedance

profiles are measured simultaneously using microscopy imaging and an impedance analyzer.

Transit time and the impedance amplitude ratio are quantified as cell’s mechanical and electri-

cal property indicators while cell elongation length inside the channel is used as a measure of

cell size.

Figure 7.1: Schematic of the microfluidic system for electrical and mechanical characterizationof single cells using impedance spectroscopy and constriction channel. Cells are aspiratedcontinuously through the small constriction channel with impedance data, cell transit time, andcell elongation length measured simultaneously.

The proposed design represents several technical advancements: (1) compared to previ-

ously reported micro devices targeting single cell electromechanical property characteriza-

tion, a higher number of cells per cell type have been characterized (e.g., EMT6 (n=747)

and EMT6/AR1.0 (n=770)); (2) electrically, leakage current is minimized due to the proper


sealing between the aspirated cell and the sidewalls of the constriction channel, which en-

ables this technique to distinguish not only cell types withsignificant difference in cell size

distribution (osteoblasts vs. osteocytes) but also cell types with a comparable size distribution

(EMT6 and EMT6/AR1.0); (3) mechanically, the effect of cell size on transit time is investi-

gated by comparing the testing results on cell types with comparable size distribution (EMT6

and EMT6/AR1.0); (4) Neural network based pattern recognition is used for multi-parameter

cell type classification.




assas, VA, USA). Materials required for device fabricationincluded SU-8 photoresist (Mi-

croChem Corp., Newton, MA, USA) and 184 silicone elastomer (Ellsworth Adhesives Canada,

Burlington, ON, Canada).

7.3.1 Device Fabrication

The fabrication process for the two-layer PDMS device was described in detail in Chapter 6.

Briefly, the channel mold masters (see Figure 7.2) were fabricated using standard soft lithog-

raphy. The cell constriction channel was formed from the first layer of SU-8 (5µm, SU-8 5)

on a glass substrate (Figure 7.2(a)). A second layer of SU-8 (25 µm, SU-8 25) was then spin

coated on the glass substrate covered with the first layer of SU-8, soft-baked, and exposed to

UV light with alignment (Figure 7.2(b)), followed by post-exposure bake, development and

hardbake (Figure 7.2(c)) to form the cell loading channel. PDMS prepolymer and curing agent

were mixed, degassed, poured on channel masters and baked inan oven (Figure 7.2(d)). PDMS

channels were then peeled from the SU-8 masters (Figure 7.2(e)) with through holes punched

and bonded to a glass slide for cell experiment use (Figure 7.2(f)).


Figure 7.2: Fabrication steps for forming the PDMS based micro device for single-cell elec-tromechanical property characterization using impedancespectroscopy and constriction chan-nel.

7.3.2 Cell Preparation and Device Operation

MC-3T3 cells (osteoblast) were purchased from American Type Culture Collection (ATCC)

(Manassas, VA, USA) and cultured withα-MEM media supplemented with 10% fetal bovine

serum and 1% penicillin and streptomycin. MLO-Y4 osteocyte-like cells were obtained from

Prof. L. Bonewald (University of Missouri-Kansas City) andmaintained in Prof. L. You’s lab

(Institute of Biomaterials and Biomedical Engineering, University of Toronto) withα-MEM

media supplemented with 2.5% calf serum, 2.5% fetal bovine serum, and 1% penicillin and

streptomycin [162]. Murine breast cancer wild type EMT6 cells were obtained from Prof.

R. Tannock (the Hospital for Sick Children, University of Toronto) and maintained in Prof.

S. Wu’s lab (Department of Pharmaceutical Sciences, University of Toronto) withα-MEM

media supplemented with 10% fetal bovine serum and 1% penicillin and streptomycin. P-

glycoprotein overexpressing, drug resistant EMT6/AR1.0 cells were made multidrug resistant

by treating EMT6 cells with 1µg/ml doxorubicin (an anticancer drug) [163].

The microfluidic device was first filled with culture medium. Adroplet of cell suspen-

sion was pipetted to the entrance of the cell loading channel. A negative pressure of 10 kPa


generated from a custom developed pump aspirated cells continuously through the constric-

tion channel. Cell images were taken by an inverted microscope (Olympus IX801, Olympus

Canada Inc., Canada). Impedance data were recorded by an impedance analyzer (Agilent -

4294A, Agilent Technologies, Inc., USA) with two Ag/AgCl nonpolarizable electrodes in-

serted into culture medium.

7.3.3 Data Analysis

When a cell is aspirated through the constriction channel, it blocks electric fields and leads

to higher impedance amplitude values compared to the case without the presence of a cell

in the constriction channel. The time duration for this increased impedance amplitude is in-

terpreted astransit time(i.e., the time duration taken by a cell to squeeze through the chan-

nel), which reflects cell’s mechanical properties (Figure 7.3(a)). The ratio between the highest

impedance amplitude value captured during cell’s squeezing through the constriction channel

and the impedance amplitude value with no cell in the constriction channel is defined as the

impedance amplitude ratio, which is used as the cell’s electrical property indicator (Figure

7.3(a)).

Numerical simulation was conducted using the finite elementanalysis package COMSOL

3.5 (Burlington, MA, USA) to model a cell’s passing through the constriction channel. Table

7.1 shows the cell dimensions and electrical parameters used in simulation. For simplicity, a

rectangular shape was used to model the cell elongation in the constriction channel). The total

meshing element was approximately 380,000.

In order to measure the cell elongation length inside the constriction channel, a background

subtraction technique was developed to process the captured images by a CMOS camera (601f;

Basler; Ahrensburg, Germany) (Figure 7.3(b)). The background image was stationary and the

lighting conditions were kept unchanged during the experiments. The procedure consists of a

sequence of image processing steps adapted to the context ofcell elongation (frame differenc-

ing, thresholding, particle removal using erosion, and edge detection along the channel) [164].


Figure 7.3: (a) Impedance measurement of single cells (amplitude vs. time). Transit time indi-cates cellular mechanical properties and impedance amplitude ratio indicates cellular electricalproperties. (b) A cell aspirated in the constriction channel. As an indicator of the cell size, theelongation length was measured from image processing approaches.

Figure 7.4: Scatter plot of impedance amplitude ratio vs. transit time (osteoblast) as a functionof testing parameters (constriction channel cross-sectional area: 6µm×6 µm and 8µm×8 µm;impedance measurement frequency: 10 kHz and 100 kHz; aspiration pressure: 10 kPa).


Table 7.1: Electrode dimensions and relevant parameters used in numerical simulation.

Parameter Value

Constriction channel length 200µmConstriction channel width 8µmCell elongation length 50µmCell elongation width 7.9µmCell membrane thickness 10 nmCulture medium conductivity 1 S/mCulture medium relative permittivity 80PDMS conductivity 0 S/mPDMS relative permittivity 5Cell membrane conductivity 0 S/mCell membrane relative permittivity 20Cell cytoplasm conductivity 0.4 S/mCell cytoplasm relative permittivity 80

For measuring cell transit time, impedance profiles rather than caputred images were used to

interpret transit time since compared to the Basler camera (30 frames per second), the sampling

rate of the impedance analyzer is higher (80 points per second).

A two-layer back propagation neural network was used for pattern recognition (MATLAB

2010, MathWork, USA). The input data has three groups of parameters collected from experi-

mental results on cells (osteocytes, osteoblasts, EMT6 andEMT6/AR1.0), namely, transit time,

impedance amplitude ratio, and cell elongation length. Neural network was used to classify os-

teoblasts from osteocytes and to classify EMT6 from EMT6/AR1.0. Taking the classification

of EMT6 (n=747) and EMT6/AR1.0 (n=770) as an example, the complete dataset was divided

into training data (70%), validation data (15%), and testing data (15%) to quantify cell classi-

fication success rates. In order to avoid the inappropriate selection of the number of neurons,

a loop function was used to enumerate the neuron number from 5to 200 with the highest cell

classification success rate recorded.



When a cell is forced to squeeze through the constriction channel, it blocks electric fields and

leads to an increase in impedance amplitude values. Interpreting the impedance data, one

can extract transit time that a cell takes to squeeze throughthe constriction channel. In the

meanwhile, as there is a tight seal between the cell and the constriction channel walls, which

can effectively decrease current leakages. Hence, this experimental setup provides an ideal

situation for cellular electrical property characterization. In summary, both the mechanical

indicator (transit time) and the electrical indicator (impedance amplitude ratio) were obtained

from the interpretation of impedance data.

Proper selection of the applied electric field’s frequency is important. When the frequency

is low (e.g., 100 Hz to 1 kHz), electric field lines pass aroundthe cell membranes, and impedance

data can only reflect the sealing properties between the celland the constriction channel. When

the frequency is too high (e.g., 1 MHz), the impedance of the cell membrane is too low to block

electric field lines, making the difference with and without cell inside the constriction channel

trivial.

In this study, two frequencies (10 kHz and 100 kHz) were selected for testing. As the

frequency was increased from 10 kHz to 100 kHz, there was an obvious impedance amplitude

ratio decrease (see Figure 7.4). Numerical simulation confirms that due to the cell membrane,

more electric field lines pass around the cell at 10 kHz compared to 100 kHz. At 100 kHz,

cellular membrane impedance is lower and therefore, more electric field lines penetrate the cell

membrane (see Figure 7.5). Thus, as the frequency increasesfrom 10 kHz to 100 kHz, cellular

electrical properties rather than cell-channel sealing properties are reflected by impedance data.

All experiments used 100 kHz as the characterization frequency.

The effect from different cross-sectional areas of the constriction channel must also be

understood. When the cross-sectional area is too small (e.g., 4 µm×4 µm), it was noticed in

experiments that cells were elongated too much and often broken into several sections inside

the constriction channel. When the cross-sectional area istoo large (e.g., 10µm×10µm), many


Figure 7.5: (a) Schematic of the 2-D numerical model used in simulation. Geometric param-eters are listed in Table 7.1. (b) A section of meshing with 380,000 elements. (c) Simulationresults of current density as a frequency of 100 kHz (left) and 10 kHz (right). Arrow: totalcurrent density.

cells passed through the constriction channel without any resistance and hence, there was no

proper seal formed. Figure 7.4 shows the collected transit time and impedance amplitude ratio,

measured on osteoblasts, for channel cross-sectional areaof 6 µm×6 µm and 8µm×8 µm. It

can be seen that as the channel cross-sectional area was decreased from 8µm×8 µm to 6µm×6

µm, transit time increased correspondingly, demonstratingthat not only cell deformability but

also cell size and channel cross-sectional area have an effect on transit time.

7.4.1 Osteoblast vs. Osteocyte

Micro devices with a constriction channel of 6µm×6 µm were used to characterize osteoblasts

(n=206) and osteocytes (n=217) (impedance measurement frequency: 100 kHz, aspiration

pressure: 10 kPa, see Figure 7.6). Compared with osteocytes, osteoblasts have a larger cell

elongation length (64.51±14.98µm vs. 39.78±7.16µm), longer transit time (1.84±1.48 sec vs.

0.94±1.07 sec), and a higher impedance amplitude ratio (1.198±0.071 vs. 1.099±0.038).


Figure 7.6: Scatter plot of transit time vs. cell elongation(a) and impedance amplitude ratiovs. cell elongation length (b). Osteoblast (n=206), osteocyte (n=217), impedance measurementfrequency: 100 kHz, aspiration pressure: 10 kPa, and constriction channel cross-sectional area:6 µm6µm.

Neural network-based cell classification resulted in cell classification success rates of 69.8%

(transit time), 85.3% (impedance amplitude ratio), and 93.7% (both transit time and impedance

amplitude ratio), suggesting that biomechanical (transittime) and bioelectrical (impedance am-

plitude ratio) parameters, when used in combination, couldprovide a higher cell classification

success rate than using electrical or mechanical parameteralone (see Table 7.2 and Figure 7.7).

Interestingly, using cell elongation length data only, cell classification success rate was as high

as 90.5%. This is due to the fact that significant size differences exist between osteoblasts

and osteocytes. This size difference may also account for their differences in transit time and

impedance amplitude ratio.

Note that both impedance amplitude and impedance phase datawere collected in exper-

iments and used as input for cell classification. The successrates were 85.3% (impedance

amplitude ratio only), 72.1% (impedance phase difference only), and 86.8% (impedance am-

plitude and phase data used together). Compared to the use ofimpedance amplitude ratio

only, a combined use of impedance amplitude ratio and phase did not significantly improve

cell classification results. Therefore, only impedance amplitude ratio was used as the electrical

parameter for neural network based cell classification.


Figure 7.7: Pattern recognition using neural network for classifying osteoblasts (n=206) andosteocytes (n=217). (a) Confusion matrix with the input of cell elongationlength. Successrates: 91.2% (training group), 85.7% (validation group), and90.5%(test group). (b) Confusionmatrix with the input of transit time. Success rates: 69.7% (training group), 69.8% (validationgroup), and69.8% (test group). (c) Confusion matrix with the input of impedance amplituderatio. Success rates: 85.5% (training group), 83.8% (validation group), and85.3% (test group).(d) Confusion matrix with the input of transit time and impedance amplitude ratio. Successrates: 87.5% (training group), 81.0% (validation group) and 93.7% (test group).


7.4.2 EMT6 vs. EMT6/AR1.0

Table 7.2: Cell classification success rates.

Cell type Cell elongation Transit time Amplitude ratio Time+ Ratio

Osteoblast vs. Osteocyte 90.5% 69.8% 85.3% 93.7%EMT6 vs. EMT6/AR1.0 51.3% 57.5% 59.6% 70.2%

The micro device was also applied to test EMT6 (n=747) and EMT6/AR1.0 (n=770) cells

(impedance measurement frequency: 100 kHz, aspiration pressure: 10 kPa, constriction chan-

nel cross-section: 8µm×8 µm). EMT6/AR1.0 cells are from drug treated EMT6 cells, having

almost the same size distributions. Figure 7.8(a) shows a scatter plot of transit time vs. cell

elongation length, indicating that the number of EMT6/AR1.0 cells with transit time less than

0.1 sec is higher than that of EMT6 cells. Figure 7.8(b) reveals a linear trend between cell

elongation length and impedance amplitude ratio with different slopes (0.0022µm−1 vs. 0.0028

µm−1) and different y-axis intersections (0.990 vs. 0.967) for EMT6 and EMT6/AR1.0.

We further investigated the effect of cell size on transit time and impedance amplitude ratio

of EMT6 and EMT6/AR1.0 cells within the cell elongation range of 40-55µm (see Figure 7.9).

This range was chosen since the majority of cells under measurement fell into this range (see

Figure 7.8). As the cell elongation length increases, thereis an increase in transit time and

impedance amplitude ratio for both EMT6 and EMT6/AR1.0 cells. From the perspective of

mechanical property characterization, EMT6/AR1.0 cells have a lower transit time, compared

to EMT6 cells, which are 0.17±0.12 sec vs. 0.20±0.17 sec (cell elongation length: 40-45µm),

0.21±0.16 sec vs. 0.26±0.28 sec (cell elongation length: 45-50µm), and 0.33±0.28 sec vs.

0.35±0.27 sec (cell elongation length: 50-55µm), respectively.

Figure 7.10 shows in further detail that EMT6/AR1.0 cells have a lower transit time than

EMT6 with comparable cell sizes. For the cell elongation range of 40-45µm, a higher fraction

of EMT6/AR1.0 cells have transit time lower than 0.1 sec (0.12 vs. 0.29) while higher fractions

of EMT6 cells have transit time in the range of 0.1-0.175 sec (0.57 vs. 0.37 ) and higher than


Figure 7.8: Scatter plot of transit time vs. cell elongationlength (a) and impedance ampli-tude ratio vs. cell elongation length (b). EMT6 (n=747), EMT6/AR1.0 (n=770), impedancemeasurement frequency: 100 kHz, aspiration pressure: 10 kPa, and constriction channel cross-sectional area: 8µm×8 µm.

Figure 7.9: (a) Transit time and (b) impedance amplitude ratio as a function of cell elongationlength (EMT6 vs. EMT6/AR1.0).


0.375 sec (0.10 vs. 0.08) (see Figure 7.10(a)). For the rangeof 45-50µm, a higher fraction of

EMT6/AR1.0 cells have transit time lower than 0.1 sec (0.08 vs. 0.22) while higher fractions

of EMT6 cells have transit time in the range of 0.1-0.2 sec (0.52 vs. 0.40) and higher than

0.4 sec (0.15 vs. 0.10) (see Figure 7.10(b)). For the range of50-55µm, a higher fraction of

EMT6/AR1.0 cells have transit time lower than 0.175 sec (0.28 vs. 0.37) while higher fractions

of EMT6 cells have transit time in the range of 0.175-0.275 sec (0.24 vs. 0.18) and higher than

0.575 sec (0.16 vs. 0.13) (see Figure 7.10(c)). In summary, EMT6/AR1.0 cells have lower

transit time compared to EMT6 cells, which may indicate a lower stiffness resulting from the

treatment of doxorubicin [163, 165].

From the perspective of electrical property characterization, for cell elongation range of

40-45µm, EMT6 cells have a higher impedance amplitude ratio compared to EMT6/AR1.0

(see Figure 7.9(b)), which are 1.085±0.023 vs. 1.084±0.026 (EMT6 vs. EMT6/AR1.0). For

the ranges of 45-50µm and 50-55µm, EMT6/AR1.0 cells have higher impedance amplitude

ratios, which are 1.092±0.027 vs. 1.090±0.020 and 1.107±0.028 vs. 1.100±0.024.

The impedance amplitude increase during a cell’s passing through the constriction channel

is caused by the impedance of cell membrane and cytoplasm. At100 kHz, electric field lines

penetrate two portions of the cell membrane and cytoplasm that is connected in series with the

two portions of the cell membrane (see Figure 7.5). The cell membrane capacitance Cmembraneis

estimated as (cell membrane permittivity)×(constriction channel cross-section area)/(cell mem-

brane thickness), which is independent of cell elongation length. In the meanwhile, cytoplasm

resistance Rcytoplasmcan be estimated as (cell elongation length)/ (constriction channel cross-

section area×cytoplasm conductivity), which is a linear function of cellelongation length.

Figure 7.9(b) shows that there is a linear trend between impedance amplitude ratio and cell

elongation length, indicating the effect of Rcytoplasmon impedance amplitude ratio. The slope

difference between EMT6 and EMT6/AR1.0 (0.0018 vs. 0.0023µm−1) suggests differences in

cytoplasm conductivity. More specifically, since the slopeof EMT6/AR1.0 cells is higher than

the slope of EMT6 cells, EMT6/AR1.0 cells may have lower cytoplasm conductivity.


Figure 7.10: The distribution of transit time of EMT6 and EMT6/AR1.0 cells within groupedcell elongation lengths (40-45µm (EMT6 (n=112) and EMT6/AR1.0 (n=144) (a), 45-50µm (EMT6 (n=192) and EMT6/AR1.0 (n=199) (b) and 50-55µm (EMT6 (n=140) andEMT6/AR1.0 (n=134) (c).


Figure 7.11: Pattern recognition using neural network for classifying EMT6 (n=747) andEMT6/AR1.0 (n=770). (a) Confusion matrix with the input of cell elongationlength. Suc-cess rates: 53.0% (training group), 49.6% (validation group), and51.3% (test group). (b)Confusion matrix with the input of transit time. Success rates: 55.3% (training group), 58.3%(validation group), and57.5% (test group). (c) Confusion matrix with the input of impedanceamplitude ratio. Success rates: 57.0% (training group), 53.5% (validation group), and59.6%(test group). (d) Confusion matrix with the input of transittime and impedance amplitude ratio.Success rates of 63.2% (training group), 63.2% (validationgroup), and70.2% (test group).


Since membrane capacitance is cell elongation independent, the effect of membrane ca-

pacitance on impedance amplitude ratio should be reflected from the intersection of the linear

fitting, (1.0076 vs. 0.9847) for EMT6 and EMT6/AR1.0 shown in Figure 7.9(b). More specif-

ically, since the intersection of EMT6 cells is higher than EMT6/AR1.0, the impedance value

of membrane capacitance of EMT6 cells is higher than that of EMT6/AR1.0, which translates

into lower membrane capacitance values for EMT6 cells compared to EMT6/AR1.0 cells.

The success rate of classifying EMT6 vs. EMT6/AR1.0 cells using cell elongation length

alone is only 51.3% (EMT6 vs. EMT6/AR1.0, Table 7.2 and Figure 7.11), due to the insignif-

icant difference in cell size (cell elongation length: 51.47±11.33µm vs. 50.09±9.70µm) for

EMT6 and EMT6/AR1.0 cells. Cell classification success rates of 57.5% (transit time), 59.6%

(impedance amplitude ratio), and 70.2% (both transit time and impedance amplitude ratio)

suggest that biomechanical (transit time) and bioelectrical (impedance amplitude ratio) param-

eters, when used in combination, could provide a higher cellclassification success rate than

using electrical or mechanical parameter alone.

7.5 Conclusion

This chapter presented a microfluidic measurement system for mechanical and electrical char-

acterization of single cells using constriction channel and impedance spectroscopy. The device

was used to test osteoblasts and osteocytes, demonstratingthat osteoblasts, compared with

osteocytes, have a larger cell elongation length, longer transit time, and a higher impedance

amplitude ratio. The micro device was also used to distinguish EMT6 from EMT6/AR1.0 cells

with the comparable size distribution. Neural network based pattern recognition produced

the cell classification success rates of 51.3% (cell elongation), 57.5% (transit time), 59.6%

(impedance amplitude ratio), and 70.2% (both transit time and impedance amplitude ratio).

These preliminary cell classification results suggest thatbiomechanical and bioelectrical pa-

rameters, when used in combination, could provide a higher cell classification success rate


than using electrical or mechanical parameter alone. The system capable of collecting both

electrical and mechanical data can also be a useful tool for fundamental cellular biophysical

studies.

Chapter 8

Conclusions

In this study, three implantable stimuli-responsive drug delivery micro devices have been de-

veloped with pH- and glucose-responsiveness, respectively. For pH-responsive drug delivery

micro devices,in vitro release characterization proved the concept and the capability of the

micro devices for pH-responsive VB12 delivery. In vivo biocompatibility testing verified the

feasibility of the pH-responsive devices for short-term implantation applications.

For type I glucose-responsive drug delivery micro devices,the permeation tests recorded a

large increase of VB12 diffusion out of the devices at high glucose concentrations. Fortype II

devices,in vitro testing verified the glucose-responsiveness by recording a3-fold increase of

insulin permeability in response to an increase of glucose levels from 100 mg/dl to 400 mg/dl.

In vivo testing verified the feasibility of the micro device for short-term diabetic management.

In the meanwhile, three types of micro devices targeting single-cell electromechanical

property characterization were developed. The micro device using electrodeformation distin-

guished SiHa from ME180 cells based on different deformation ratios and quantified Young’s

modulus of SiHa (601±183 Pa) and ME180 cells (1463±649 Pa) by comparing the experimen-

tally measured deformations with those obtained from numerical simulations.

The micro device based on impedance spectroscopy and micropipette aspiration was de-

veloped to quantify membrane capacitance, cytoplasm resistance and Young’s modulus of os-

98

Chapter 8. Conclusions 99

teoblasts, demonstrating the capability of simultaneous single-cell electromechanical property

characterization. The micro device based on impedance measurement and constriction chan-

nel was proposed to distinguish cell types not only with quite different size distributions (os-

teoblasts vs. osteocytes) but also cell types with comparable sizes (EMT6 vs. EMT6/AR1.0).

Neural network based cell classification suggested that biomechanical (transit time) and bio-

electrical (impedance amplitude ratio) parameters, when used in combination, could provide a

higher cell classification success rate than using electrical or mechanical parameter alone.

8.1 Contributions

The contributions of this research include:

1. Design, fabrication,in vitro andin vivocharacterization of the first-of-its-kind implantable

pH-responsive drug delivery micro device integrating pH-responsive nanoparticles with

microfabrication.

2. Design, fabrication andin vitro characterization of the first-of-its-kind implantable glucose-

responsive drug delivery micro device (type I) integratingpH-responsive nanoparticles,

glucose oxidase and chitosan microparticles (anchor of glucose oxidase).

3. Design, fabrication,in vitro and in vivo characterization of glucose-responsive insulin

release micro devices (type II) with PDMS grids as backbonesembedded with pH-

responsive nanoparticles, glucose oxidase and bovine serum albumin (base membrane).

4. Design, fabrication and characterization of micro devices for single-cell electromechan-

ical property characterization using electrodeformation.

5. Design, fabrication and characterization of the first-of-its-kind single-cell electrome-

chanical property characterization micro devices integrating impedance spectroscopy

and micropipette aspiration on a single chip.

Chapter 8. Conclusions 100

6. Design, fabrication and characterization of the first-of-its-kind micro devices for single-

cell electromechanical property characterization using impedance measurement and con-

striction channel.

8.2 Future Directions

Many possibilities exist for further extending this research. Examples are:

1. To systematically conductin vivo testing of the type II glucose-responsive micro de-

vices. Potential parameters to be screened include implantation locations (subcutaneous

vs. intraperitoneal), formulation of insulin (combination of insulin molecules with sur-

factants) and device geometry parameters (e.g., drug reservoir sizes and dimensions of

PDMS grids).

2. To decouple the combined effect of cells’ electrical and mechanical properties on their

electrodeformed behaviors. To achieve this objective, a four-electrode setup can be used

with two inner electrodes for impedance measurement and twoouter electrodes for elec-

trodeformation. Electrical properties collected from impedance measurement can be

used as input parameters to quantify cellular electrodeformation ratios, which can fur-

ther lead to Young’s modulus calculation. This setup could enable the characterization

of cellular Young’s modulus, membrane permittivity and cytoplasm conductivity, simul-

taneously.

3. To interpret cell-size independent electromechanical properties from measured transit

time and impedance amplitude ratio. In order to quantify specific membrane capacitance

and cytoplasm conductivity, electrical modeling should bedeveloped with the help of

numerical simulations. For mechanical property interpretation, fluid-structural coupling

simulation with moving mesh may be conducted to model this phenomenon for further

data interpretation.

Bibliography

[1] I. N. Scobie, Atlas of diabetes mellitus, 3rd ed. London:Informa Healthcare 2007.

[2] I. N. Scobie, K. Samaras, and I. W. Campbell, Fast facts: diabetes mellitus, 3rd ed.

Abingdon: Health Press, 2009.

[3] L. Leoni and T. A. Desai, “Micromachined biocapsules forcell-based sensing and de-

livery,” Advanced Drug Delivery Reviews, vol. 56, pp. 211-29, Feb 10 2004.

[4] J. J. Meier, A. Bhushan, and P. C. Butler, “The potential for stem cell therapy in dia-

betes,”Pediatric Research, vol. 59, pp. 65R-73R, Apr 2006.

[5] D. Porte, R. S. Sherwin, A. Baron, M. Ellenberg, and H. Rifkin, Ellenberg and Rifkin’s

diabetes mellitus: theory and practice, 6th ed. New York: McGraw-Hill, Health Profes-

sions Division, 2003.

[6] E. Renard, “Implantable closed-loop glucose-sensing and insulin delivery: the future for

insulin pump therapy,”Current Opinion in Pharmacology, vol. 2, pp. 708-16, Dec 2002.

[7] E. Renard, G. Costalat, H. Chevassus, and J. Bringer, “Artificial beta-cell: clinical

experience toward an implantable closed-loop insulin delivery system,”Diabetes and

Metabolism, vol. 32, pp. 497-502, Dec 2006.

[8] R. Hovorka, “Continuous glucose monitoring and closed-loop systems,”Diabetic

Medicine, vol. 23, pp. 1-12, Jan 2006.

101

Bibliography 102

[9] R. Hovorka, ”Closed-loop insulin delivery: from bench to clinical practice,”Nature

Reviews Endocrinology, vol. 7, pp. 385-95, 2011.

[10] S. Chaterji, I. K. Kwon, and K. Park, “Smart polymeric gels: redefining the limits of

biomedical devices,”Progress in Polymer Science, vol. 32, pp. 1083-1122, Aug 2007.

[11] P. Gupta, K. Vermani, and S. Garg, “Hydrogels: from controlled release to pH-

responsive drug delivery,”Drug Discovery Today, vol. 7, pp. 569-579, May 15 2002.

[12] A. Kikuchi and T. Okano, “Pulsatile drug release control using hydrogels,”Advanced

Drug Delivery Reviews, vol. 54, pp. 53-77, Jan 17 2002.

[13] J. Kost and R. Langer, “Responsive polymeric delivery systems,”Advanced Drug De-

livery Reviews, vol. 46, pp. 125-48, Mar 1 2001.

[14] T. Miyata, T. Uragami, and K. Nakamae, “Biomolecule-sensitive hydrogels,”Advanced

Drug Delivery Reviews, vol. 54, pp. 79-98, Jan 17 2002.

[15] R. V. Ulijn, N. Bibi, V. Jayawarna, P. D. Thornton, S. J. Todd, R. J. Mart, A. M. Smith,

and J. E. Gough, “Bioresponsive hydrogels,”Materials Today, vol. 10, pp. 40-48, Apr

2007.

[16] A. Baldi, Y. D. Gu, P. E. Loftness, R. A. Siegel, and B. Ziaie, “A hydrogel-actuated en-

vironmentally sensitive microvalve for active flow control,” Journal of Microelectrome-

chanical Systems, vol. 12, pp. 613-621, Oct 2003.

[17] K. Zhang and X. Y. Wu, “Temperature and pH-responsive polymeric composite mem-

branes for controlled delivery of proteins and peptides,”Biomaterials, vol. 25, pp. 5281-

91, Oct 2004.

[18] M. Staples, K. Daniel, M. J. Cima, and R. Langer, “Application of micro- and nano-

electromechanical devices to drug delivery,”Pharmaceutical Research, vol. 23, pp. 847-

63, May 2006.

Bibliography 103

[19] Y. Li, H. L. Ho Duc, B. Tyler, T. Williams, M. Tupper, R. Langer, H. Brem, and M. J.

Cima, “In vivo delivery of BCNU from a MEMS device to a tumor model,”Journal of

Controlled Release, vol. 106, pp. 138-45, Aug 18 2005.

[20] Y. Li, R. S. Shawgo, B. Tyler, P. T. Henderson, J. S. Vogel, A. Rosenberg, P. B. Storm, R.

Langer, H. Brem, and M. J. Cima, “In vivo release from a drug delivery MEMS device,”

Journal of Controlled Release, vol. 100, pp. 211-9, Nov 24 2004.

[21] J. T. Santini, Jr., M. J. Cima, and R. Langer, “A controlled-release microchip,”Nature,

vol. 397, pp. 335-8, Jan 28 1999.

[22] J. M. Maloney, S. A. Uhland, B. F. Polito, N. F. Sheppard,Jr., C. M. Pelta, and J. T.

Santini, Jr., “Electrothermally activated microchips forimplantable drug delivery and

biosensing,”Journal of Controlled Release, vol. 109, pp. 244-55, Dec 5 2005.

[23] J. H. Prescott, S. Lipka, S. Baldwin, N. F. Sheppard, Jr., J. M. Maloney, J. Coppeta,

B. Yomtov, M. A. Staples, and J. T. Santini, Jr., “Chronic, programmed polypeptide

delivery from an implanted, multireservoir microchip device,” Nature Biotechnology,

vol. 24, pp. 437-8, Apr 2006.

[24] A. C. R. Grayson, I. S. Choi, B. M. Tyler, P. P. Wang, H. Brem, M. J. Cima, and R.

Langer, “Multi-pulse drug delivery from a resorbable polymeric microchip device,”Na-

ture Materials, vol. 2, pp. 767-772, Nov 2003.

[25] J. Intra, J. M. Glasgow, H. Q. Mai, and A. K. Salem, “Pulsatile release of biomolecules

from polydimethylsiloxane (PDMS) chips with hydrolytically degradable seals,”Jour-

nal of Controlled Release, vol. 127, pp. 280-7, May 8 2008.

[26] W. H. Ryu, M. Vyakarnam, R. S. Greco, F. B. Prinz, and R. J.Fasching, “Fabrication of

multi-layered biodegradable drug delivery device based onmicro-structuring of PLGA

polymers,”Biomedical Microdevices, vol. 9, pp. 845-53, Dec 2007.

Bibliography 104

[27] R. W. Ruddon and ebrary Inc., Cancer biology, 3rd ed. NewYork: Oxford University

Press, 2007.

[28] K. Bellenir, Cancer sourcebook, 6th ed. Detroit, MI: Omnigraphics, 2011.

[29] H. Morgan, T. Sun, D. Holmes, S. Gawad, and N. G. Green, “Single cell dielectric

spectroscopy,”Journal of Physics D-Applied Physics, vol. 40, pp. 61-70, Jan 7 2007.

[30] T. Sun and H. Morgan, “Single-cell microfluidic impedance cytometry: a review,”Mi-

crofluidics and Nanofluidics, vol. 8, pp. 423-443, Apr 2010.

[31] A. Valero, T. Braschler, and P. Renaud, “A unified approach to dielectric single cell

analysis: impedance and dielectrophoretic force spectroscopy,” Lab on a Chip, vol. 10,

pp. 2216-2225, 2010.

[32] D. H. Boal, Mechanics of the cell. Cambridge ; New York: Cambridge University Press,

2002.

[33] C. R. Ethier and C. A. Simmons, Introductory biomechanics: from cells to organisms.

Cambridge: Cambridge University Press, 2007.

[34] H. M. Coley, F. H. Labeed, H. Thomas, and M. P. Hughes, ”Biophysical characterization

of MDR breast cancer cell lines reveals the cytoplasm is critical in determining drug

sensitivity,”Biochim Biophys Acta, vol. 1770, pp. 601-8, Apr 2007.

[35] J. Guck, S. Schinkinger, B. Lincoln, F. Wottawah, S. Ebert, M. Romeyke, D. Lenz, H.

M. Erickson, R. Ananthakrishnan, D. Mitchell, J. Kas, S. Ulvick, and C. Bilby, “Op-

tical deformability as an inherent cell marker for testing malignant transformation and

metastatic competence,”Biophysical Journal, vol. 88, pp. 3689-3698, 2005.

[36] S. Park, D. Koch, R. Cardenas, J. Kas, and C. K. Shih, ”Cell motility and local vis-

coelasticity of fibroblasts,”Biophysical Journal, vol. 89, pp. 4330-42, Dec 2005.

Bibliography 105

[37] M. J. Rosenbluth, W. A. Lam, and D. A. Fletcher, ”Force microscopy of nonadherent

cells: a comparison of leukemia cell deformability,”Biophysical Journal, vol. 90, pp.

2994-3003, Apr 2006.

[38] K. A. Ward, W. I. Li, S. Zimmer, and T. Davis, “Viscoelastic properties of transformed-

cells - role in tumor-cell progression and metastasis formation,” Biorheology, vol. 28,

pp. 301-313, Aug 1991.

[39] K. H. Han, A. Han, and A. B. Frazier, “Microsystems for isolation and electrophysio-

logical analysis of breast cancer cells from blood,”Biosensors and Bioelectronics, vol.

21, pp. 1907-1914, Apr 15 2006.

[40] W. A. Lam, M. J. Rosenbluth, and D. A. Fletcher, ”Chemotherapy exposure increases

leukemia cell stiffness,”Blood, vol. 109, pp. 3505-3508, Apr 15 2007.

[41] T. Ochalek, F. J. Nordt, K. Tullberg, and M. M. Burger, ”Correlation between cell de-

formability and metastatic potential in B16-F1 melanoma cell variants,” Cancer Re-

search, vol. 48, pp. 5124-5128, Sep 15 1988.

[42] S. Igawa, I. Hayashi, N. Tanaka, H. Hiruma, M. Majima, T.Kawakami, M. Hirose, N.

Masuda, and H. Kobayashi, ”Nitric oxide generated by iNOS reduces deformability of

Lewis lung carcinoma cells,”Cancer Science, vol. 95, pp. 342-347, Apr 2004.

[43] W. J. Yao, L. Gu, D. G. Sun, W. B. Ka, Z. Y. Wen, and S. Chien,“Wild type p53 gene

causes reorganization of cytoskeleton and, therefore, theimpaired deformability and

difficult migration of murine erythroleukemia cells,”Cell Motility and the Cytoskeleton,

vol. 56, pp. 1-12, Sep 2003.

[44] K. Chen, D. Li, Y. H. Jiang, W. J. Yao, X. J. Wang, X. C. Wei,J. Gao, L. D. Xie, Z. Y.

Yan, Z. Y. Wen, and S. Chien, ”Influence of expressed TRAIL on biophysical properties

of the human leukemic cell line Jurkat,”Cell Research, vol. 14, pp. 161-168, Apr 2004.

Bibliography 106

[45] M. Lekka, P. Laidler, D. Gil, J. Lekki, Z. Stachura, and A. Z. Hrynkiewicz, ”Elasticity

of normal and cancerous human bladder cells studied by scanning force microscopy,”

European Biophysics Journal, vol. 28, pp. 312-6, 1999.

[46] T. Korb, K. Schluter, A. Enns, H. U. Spiegel, N. Senninger, G. L. Nicolson, and J. Haier,

“Integrity of actin fibers and microtubules influences metastatic tumor cell adhesion,”

Experimental Cell Research, vol. 299, pp. 236-247, Sep 10 2004.

[47] Y. Cho, H. S. Kim, A. B. Frazier, Z. G. Chen, D. M. Shin, andA. Han, “Whole-cell

impedance analysis for highly and poorly metastatic cancercells,” Journal of Micro-

electromechanical Systems, vol. 18, pp. 808-817, Aug 2009.

[48] M. Beil, A. Micoulet, G. von Wichert, S. Paschke, P. Walther, M. B. Omary, P. P. Van

Veldhoven, U. Gern, E. Wolff-Hieber, J. Eggermann, J. Waltenberger, G. Adler, J. Spatz,

and T. Seufferlein, ”Sphingosylphosphorylcholine regulates keratinnetwork architecture

and visco-elastic properties of human cancer cells,”Nature Cell Biology, vol. 5, pp. 803-

811, Sep 2003.

[49] H. W. Hou, Q. S. Li, G. Y. H. Lee, A. P. Kumar, C. N. Ong, and C. T. Lim, “Deformabil-

ity study of breast cancer cells using microfluidics,”Biomedical Microdevices, vol. 11,

pp. 557-564, Jun 2009.

[50] D. R. Gossett, H. T. K. Tse, S. Lee, A. T. Clark, and D. Di Carlo, “Deformability cy-

tometry: high-throughput, continuous measurement of cellmechanical properties in ex-

tesional flow,” inProc. of the Int. Conf. on Miniaturized Systems for Chemistry and Life

Sciences 2010, Groningen, The Netherlands, 3-7 October, 2010, pp. 1382-1384.

[51] C. C. Chen and A. Folch, “A high-performance elastomeric patch clamp chip,”Lab on

a Chip, vol. 6, pp. 1338-1345, Oct 2006.

[52] N. Fertig, R. H. Blick, and J. C. Behrends, “Whole cell patch clamp recording performed

on a planar glass chip,”Biophysical Journal, vol. 82, pp. 3056-3062, Jun 2002.

Bibliography 107

[53] C. Ionescu-Zanetti, R. M. Shaw, J. G. Seo, Y. N. Jan, L. Y.Jan, and L. P. Lee, “Mam-

malian electrophysiology on a microfluidic platform,”Proceedings of the National

Academy of Sciences of the United States of America, vol. 102, pp. 9112-9117, Jun

28 2005.

[54] K. G. Klemic, J. F. Klemic, M. A. Reed, and F. J. Sigworth,“Micromolded PDMS

planar electrode allows patch clamp electrical recordingsfrom cells,” Biosensors and

Bioelectronics, vol. 17, pp. 597-604, Jun 2002.

[55] T. Lehnert, M. A. M. Gijs, R. Netzer, and U. Bischoff, “Realization of hollow SiO2

micronozzles for electrical measurements on living cells,” Applied Physics Letters, vol.

81, pp. 5063-5065, Dec 23 2002.

[56] . Lehnert, D. M. T. Nguyen, L. Baldi, and M. A. M. Gijs, “Glass reflow on 3-

dimensional micro-apertures for electrophysiological measurements on-chip,”Microflu-

idics and Nanofluidics, vol. 3, pp. 109-117, Feb 2007.

[57] S. Li and L. W. Lin, “A single cell electrophysiologicalanalysis device with embedded

electrode,”Sensors and Actuators a-Physical, vol. 134, pp. 20-26, Feb 28 2007.

[58] B. Matthews and J. W. Judy, “Design and fabrication of a micromachined planar patch-

clamp substrate with integrated microfluidics for single-cell measurements,”Journal of

Microelectromechanical Systems, vol. 15, pp. 214-222, Feb 2006.

[59] W. L. Ong, J. S. Kee, A. Ajay, N. Ranganathan, K. C. Tang, and L. Yobas, “Buried

microfluidic channel for integrated patch-clamping assay,” Applied Physics Letters, vol.

89, p. 093902, Aug 28 2006.

[60] W. L. Ong, K. C. Tang, A. Agarwal, R. Nagarajan, L. W. Luo,and L. Yobas, “Mi-

crofluidic integration of substantially round glass capillaries for lateral patch clamping

on chip,”Lab on a Chip, vol. 7, pp. 1357-1366, 2007.

Bibliography 108

[61] S. Pandey, R. Mehrotra, S. Wykosky, and M. H. White, “Characterization of a MEMS

biochip for planar patch-clamp recording,”Solid-State Electronics, vol. 48, pp. 2061-

2066, Oct-Nov 2004.

[62] R. Pantoja, J. M. Nagarah, D. M. Starace, N. A. Melosh, R.Blunck, F. Bezanilla, and J.

R. Heath, “Silicon chip-based patch-clamp electrodes integrated with PDMS microflu-

idics,” Biosensors and Bioelectronics, vol. 20, pp. 509-517, Oct 15 2004.

[63] J. Seo, C. Ionescu-Zanetti, J. Diamond, R. Lal, and L. P.Lee, “Integrated multiple patch-

clamp array chip via lateral cell trapping junctions,”Applied Physics Letters, vol. 84, pp.

1973-1975, Mar 15 2004.

[64] A. Stett, V. Bucher, C. Burkhardt, U. Weber, and W. Nisch, “Patch-clamping of primary

cardiac cells with micro-openings in polyimide films,”Medical and Biological Engi-

neering and Computing, vol. 41, pp. 233-240, Mar 2003.

[65] C. Dalton, A. D. Goater, J. P. H. Burt, and H. V. Smith, “Analysis of parasites by elec-

trorotation,”Journal of Applied Microbiology, vol. 96, pp. 24-32, 2004.

[66] G. De Gasperis, X. B. Wang, J. Yang, F. F. Becker, and P. R.C. Gascoyne, “Automated

electrorotation: dielectric characterization of living cells by real-time motion estima-

tion,” Measurement Science and Technology, vol. 9, pp. 518-529, Mar 1998.

[67] R. Holzel, “Non-invasive determination of bacterial single cell properties by electroro-

tation,” Biochimica et Biophysica Acta-Molecular Cell Research, vol. 1450, pp. 53-60,

May 6 1999.

[68] M. Kriegmaier, M. Zimmermann, K. Wolf, U. Zimmermann, and V. L. Sukhorukov,

“Dielectric spectroscopy of Schizosaccharomyces pombe using electrorotation and elec-

troorientation,”Biochimica et Biophysica Acta-General Subjects, vol. 1568, pp. 135-

146, Dec 5 2001.

Bibliography 109

[69] J. Yang, Y. Huang, X. Wang, X. B. Wang, F. F. Becker, and P.R. C. Gascoyne, “Di-

electric properties of human leukocyte subpopulations determined by electrorotation as

a cell separation criterion,”Biophysical Journal, vol. 76, pp. 3307-3314, 1999.

[70] X. F. Zhou, J. P. H. Burt, and R. Pethig, “Automatic cell electrorotation measurements:

studies of the biological effects of low-frequency magnetic fields and of heat shock,”

Physics in Medicine and Biology, vol. 43, pp. 1075-1090, May 1998.

[71] H. E. Ayliffe, A. B. Frazier, and R. D. Rabbitt, “Electric impedance spectroscopy us-

ing microchannels with integrated metal electrodes,”Journal of Microelectromechanical

Systems, vol. 8, pp. 50-57, Mar 1999.

[72] G. Benazzi, D. Holmes, T. Sun, M. C. Mowlem, and H. Morgan, “Discrimination and

analysis of phytoplankton using a microfluidic cytometer,”IET Nanobiotechnology, vol.

1, pp. 94-101, Dec 2007.

[73] K. Cheung, S. Gawad, and P. Renaud, “Impedance spectroscopy flow cytometry: on-

chip label-free cell differentiation,”Cytometry Part A, vol. 65A, pp. 124-132, Jun 2005.

[74] G. A. Ferrier, S. F. Romanuik, D. J. Thomson, G. E. Bridges, and M. R. Freeman, “A

microwave interferometric system for simultaneous actuation and detection of single

biological cells,”Lab on a Chip, vol. 9, pp. 3406-12, Dec 7 2009.

[75] S. Gawad, K. Cheung, U. Seger, A. Bertsch, and P. Renaud,“Dielectric spectroscopy

in a micromachined flow cytometer: theoretical and practical considerations,”Lab on a

Chip, vol. 4, pp. 241-51, Jun 2004.

[76] S. Gawad, L. Schild, and P. Renaud, “Micromachined impedance spectroscopy flow

cytometer for cell analysis and particle sizing,”Lab on a Chip, vol. 1, pp. 76-82, 2001.

Bibliography 110

[77] D. Holmes and H. Morgan, “Single cell impedance cytometry for identification and

counting of CD4 T-cells in human blood using impedance labels,” Analytical Chemistry,

vol. 82, pp. 1455-1461, Feb 15 2010.

[78] D. Holmes, D. Pettigrew, C. H. Reccius, J. D. Gwyer, C. van Berkel, J. Holloway, D.

E. Davies, and H. Morgan, “Leukocyte analysis and differentiation using high speed

microfluidic single cell impedance cytometry,”Lab on a Chip, vol. 9, pp. 2881-2889,

2009.

[79] D. Holmes, J. K. She, P. L. Roach, and H. Morgan, “Bead-based immunoassays using a

micro-chip flow cytometer,”Lab on a Chip, vol. 7, pp. 1048-1056, 2007.

[80] C. Kuttel, E. Nascimento, N. Demierre, T. Silva, T. Braschler, P. Renaud, and A. G.

Oliva, “Label-free detection of Babesia bovis infected redblood cells using impedance

spectroscopy on a microfabricated flow cytometer,”Acta Tropica, vol. 102, pp. 63-68,

Apr 2007.

[81] H. Morgan, D. Holmes, and N. G. Green, “High speed simultaneous single particle

impedance and fluorescence analysis on a chip,”Current Applied Physics, vol. 6, pp.

367-370, Jun 2006.

[82] G. Schade-Kampmann, A. Huwiler, M. Hebeisen, T. Hessler, and M. Di Berardino,

“On-chip non-invasive and label-free cell discriminationby impedance spectroscopy,”

Cell Proliferation, vol. 41, pp. 830-40, Oct 2008.

[83] L. L. Sohn, O. A. Saleh, G. R. Facer, A. J. Beavis, R. S. Allan, and D. A. Notterman,

“Capacitance cytometry: Measuring biological cells one byone,” Proceedings of the

National Academy of Sciences of the United States of America, vol. 97, pp. 10687-

10690, Sep 26 2000.

[84] T. Sun, S. Gawad, C. Bernabini, N. G. Green, and H. Morgan, “Broadband single cell

impedance spectroscopy using maximum length sequences: theoretical analysis and

Bibliography 111

practical considerations,”Measurement Science and Technology, vol. 18, pp. 2859-2868,

Sep 2007.

[85] T. Sun, N. G. Green, S. Gawad, and H. Morgan, “Analyticalelectric field and sensitivity

analysis for two microfluidic impedance cytometer designs,” IET Nanobiotechnology,

vol. 1, pp. 69-79, Oct 2007.

[86] T. Sun, D. Holmes, S. Gawad, N. G. Green, and H. Morgan, “High speed multi-

frequency impedance analysis of single particles in a microfluidic cytometer using max-

imum length sequences,”Lab on a Chip, vol. 7, pp. 1034-1040, 2007.

[87] T. Sun, C. van Berkel, N. G. Green, and H. Morgan, “Digital signal processing methods

for impedance microfluidic cytometry,”Microfluidics and Nanofluidics, vol. 6, pp. 179-

187, Feb 2009.

[88] Y. H. Cho, T. Yamamoto, Y. Sakai, T. Fujii, and B. Kim, “Development of microfluidic

device for electrical/physical characterization of single cell,”Journal of Microelectrome-

chanical Systems, vol. 15, pp. 287-295, Apr 2006.

[89] S. Z. Hua and T. Pennell, “A microfluidic chip for real-time studies of the volume of

single cells,”Lab on a Chip, vol. 9, pp. 251-256, 2009.

[90] L. S. Jang, H. H. Li, J. Y. Jao, M. K. Chen, and C. F. Liu, “Design and fabrication of mi-

crofluidic devices integrated with an open-ended MEMS probefor single-cell impedance

measurement,”Microfluidics and Nanofluidics, vol. 8, pp. 509-519, Apr 2010.

[91] L. S. Jang and M. H. Wang, “Microfluidic device for cell capture and impedance mea-

surement,”Biomedical Microdevices, vol. 9, pp. 737-43, Oct 2007.

[92] L. S. Jang, C. C. Wu, and C. F. Liu, “Fabrication of microfluidic devices for packaging

CMOS MEMS impedance sensors,”Microfluidics and Nanofluidics, vol. 7, pp. 869-875,

Dec 2009.

Bibliography 112

[93] D. Malleo, J. T. Nevill, L. P. Lee, and H. Morgan, “Continuous differential impedance

spectroscopy of single cells,”Microfluidics and Nanofluidics, vol. 9, pp. 191-198, Aug

2010.

[94] V. Senez, E. Lennon, S. Ostrovidov, T. Yamamoto, H. Fujita, Y. Sakai, and T. Fujii,

“Integrated 3-D silicon electrodes for electrochemical sensing in microfluidic environ-

ments: application to single-cell characterization,”IEEE Sensors Journal, vol. 8, pp.

548-557, May-Jun 2008.

[95] M. H. Wang and L. S. Jang, “A systematic investigation into the electrical properties of

single HeLa cells via impedance measurements and COMSOL simulations,”Biosensors

and Bioelectronics, vol. 24, pp. 2830-2835, May 15 2009.

[96] M. Thein, F. Asphahani, A. Cheng, R. Buckmaster, M. Q. Zhang, and J. Xu, “Response

characteristics of single-cell impedance sensors employed with surface-modified micro-

electrodes,”Biosensors and Bioelectronics, vol. 25, pp. 1963-1969, Apr 15 2010.

[97] S. B. Cho and H. Thielecke, “Micro hole-based cell chip with impedance spectroscopy,”

Biosensors and Bioelectronics, vol. 22, pp. 1764-1768, Mar 15 2007.

[98] C. D. James, N. Reuel, E. S. Lee, R. V. Davalos, S. S. Mani,A. Carroll-Portillo, R.

Rebeil, A. Martino, and C. A. Apblett, “Impedimetric and optical interrogation of single

cells in a microfluidic device for real-time viability and chemical response assessment,”

Biosensors and Bioelectronics, vol. 23, pp. 845-851, Jan 18 2008.

[99] M. Abdelgawad, C. Wu, W. Y. Chien, W. B. Geddie, M. A. S. Jewett, and Y. Sun,

“A fast and simple method to fabricate circular microchannels in polydimethylsiloxane

(PDMS),” Lab on a Chip, vol. 11, pp. 545-551, 2011.

[100] W. Kim and A. Han, “A micro-aspirator chip using vacuumexpanded microchannels

for high-throughput mechanical characterization of biological cells,” inProc. of the Int.

Bibliography 113

Conf. on Miniaturized Systems for Chemistry and Life Sciences 2010, Groningen, The

Netherlands, 3-7 October, 2010, pp. 253-255.

[101] H. Engelhardt, H. Gaub, and E. Sackmann, “Viscoelastic properties of erythrocyte mem-

branes in high-frequency electric fields,”Nature, vol. 307, pp. 378-80, Jan 26-Feb 1

1984.

[102] H. Engelhardt and E. Sackmann, “On the measurement of shear elastic moduli and vis-

cosities of erythrocyte plasma membranes by transient deformation in high frequency

electric fields,”Biophysical Journal, vol. 54, pp. 495-508, Sep 1988.

[103] M. Krueger and F. Thom, “Deformability and stability of erythrocytes in high-frequency

electric fields down to subzero temperatures,”Biophysical Journal, vol. 73, pp. 2653-

2666, Nov 1997.

[104] L. A. MacQueen, M. D. Buschmann, and M. R. Wertheimer, “Mechanical properties of

mammalian cells in suspension measured by electro-deformation,” Journal of Microme-

chanics and Microengineering, vol. 20, pp. 5007, Jun 2010.

[105] V. L. Sukhorukov, H. Mussauer, and U. Zimmermann, “Theeffect of electrical deforma-

tion forces on the electropermeabilization of erythrocytemembranes in low- and high-

conductivity media,”Journal of Membrane Biology, vol. 163, pp. 235-45, Jun 1 1998.

[106] P. K. Wong, W. Tan, and C. M. Ho, “Cell relaxation after electrodeformation: effect of

latrunculin A on cytoskeletal actin,”Journal of Biomechanics, vol. 38, pp. 529-35, Mar

2005.

[107] U. Zimmermann, U. Friedrich, H. Mussauer, P. Gessner,K. Hamel, and V. Sukhoruhov,

“Electromanipulation of mammalian cells: fundamentals and application,”IEEE Trans-

actions on Plasma Science, vol. 28, pp. 72-82, Feb 2000.

Bibliography 114

[108] N. Bellini, K. C. Vishnubhatla, F. Bragheri, L. Ferrara, P. Minzioni, R. Ramponi, I.

Cristiani, and R. Osellame, “Femtosecond laser fabricatedmonolithic chip for optical

trapping and stretching of single cells,”Optics Express, vol. 18, pp. 4679-4688, Mar 1

2010.

[109] F. Bragheri, L. Ferrara, N. Bellini, K. C. Vishnubhatla, P. Minzioni, R. Ramponi, R.

Osellame, and I. Cristiani, “Optofluidic chip for single cell trapping and stretching fab-

ricated by a femtosecond laser,”Journal of Biophotonics, vol. 3, pp. 234-243, Apr 2010.

[110] A. E. Ekpenyong, C. L. Posey, J. L. Chaput, A. K. Burkart, M. M. Marquardt, T. J.

Smith, and M. G. Nichols, “Determination of cell elasticitythrough hybrid ray optics

and continuum mechanics modeling of cell deformation in theoptical stretcher,”Applied

Optics, vol. 48, pp. 6344-6354, Nov 10 2009.

[111] C. W. Lai, S. K. Hsiung, Y. Q. Chen, A. Chiou, and G. B. Lee, “An integrated 2-D active

optical fiber manipulator with microfluidic channel for optical trapping and manipula-

tion,” Journal of Microelectromechanical Systems, vol. 17, pp. 548-557, Jun 2008.

[112] C. W. Lai, S. K. Hsiung, C. L. Yeh, A. Chiou, and G. B. Lee,“A cell delivery and

pre-positioning system utilizing microfluidic devices fordual-beam optical trap-and-

stretch,”Sensors and Actuators B-Chemical, vol. 135, pp. 388-397, Dec 10 2008.

[113] F. Lautenschlager, S. Paschke, S. Schinkinger, A. Bruel, M. Beil, and J. Guck, “The

regulatory role of cell mechanics for migration of differentiating myeloid cells,”Pro-

ceedings of the National Academy of Sciences of the United States of America, vol. 106,

pp. 15696-15701, Sep 15 2009.

[114] B. Lincoln, H. M. Erickson, S. Schinkinger, F. Wottawah, D. Mitchell, S. Ulvick, C.

Bilby, and J. Guck, “Deformability-based flow cytometry,”Cytometry Part A, vol. 59A,

pp. 203-209, Jun 2004.

Bibliography 115

[115] B. Lincoln, F. Wottawah, S. Schinkinger, S. Ebert, andJ. Guck, “High-throughput rhe-

ological measurements with an optical stretcher,”Cell Mechanics, vol. 83, pp. 397-423,

2007.

[116] T. W. Remmerbach, F. Wottawah, J. Dietrich, B. Lincoln, C. Wittekind, and J. Guck,

“Oral cancer diagnosis by mechanical phenotyping,”Cancer Research, vol. 69, pp.

1728-1732, Mar 1 2009.

[117] S. K. Teo, A. B. Goryachev, K. H. Parker, and K. H. Chiam,“Cellular deformation and

intracellular stress propagation during optical stretching,” Physical Review E, vol. 81, p.

1924, May 2010.

[118] M. Abkarian, M. Faivre, and H. A. Stone, “High-speed microfluidic differential

manometer for cellular-scale hydrodynamics,”Proceedings of the National Academy

of Sciences of the United States of America, vol. 103, pp. 538-542, Jan 17 2006.

[119] S. Gabriele, A. M. Benoliel, P. Bongrand, and O. Theodoly, “Microfluidic investigation

reveals distinct roles for actin cytoskeleton and myosin IIactivity in capillary leukocyte

trafficking,” Biophysical Journal, vol. 96, pp. 4308-4318, May 20 2009.

[120] S. Gabriele, M. Versaevel, P. Preira, and O. Theodoly,“A simple microfluidic method to

select, isolate, and manipulate single-cells in mechanical and biochemical assays,”Lab

on a Chip, vol. 10, pp. 1459-1467, 2010.

[121] S. C. Gifford, J. Derganc, S. S. Shevkoplyas, T. Yoshida, and M. W. Bitensky, “A detailed

study of time-dependent changes in human red blood cells: from reticulocyte maturation

to erythrocyte senescence,”British Journal of Haematology, vol. 135, pp. 395-404, Nov

2006.

[122] T. Herricks, M. Antia, and P. K. Rathod, “Deformability limits of Plasmodium

falciparum-infected red blood cells,”Cellular Microbiology, vol. 11, pp. 1340-1353,

Sep 2009.

Bibliography 116

[123] Y. Hirose, K. Tadakuma, M. Higashimori, T. Arai, M. Kaneko, R. Iitsuka, Y. Yamanishi,

and F. Arai, “A new stiffness evaluation toward high speed cell sorter,” inProc. of the

IEEE Int. Conf. on Robotics and Automation 2010, Anchorage, Alaska, USA, 3-8 May

2010, pp. 4113-4118.

[124] W. G. Lee, H. Bang, H. Yun, J. Lee, J. Park, J. K. Kim, S. Chung, K. Cho, C. Chung,

D. C. Han, and J. K. Chang, “On-chip erythrocyte deformability test under optical pres-

sure,”Lab on a Chip, vol. 7, pp. 516-519, 2007.

[125] F. Y. Leong, Q. Li, C. T. Lim, and K. H. Chiam, “Modeling cell entry into a micro-

channel,”Biomechanics and Modeling in Mechanobiology, pp. 1-12, 2010.

[126] M. J. Rosenbluth, W. A. Lam, and D. A. Fletcher, “Analyzing cell mechanics in hema-

tologic diseases with microfluidic biophysical flow cytometry,” Lab on a Chip, vol. 8,

pp. 1062-1070, 2008.

[127] J. P. Shelby, J. White, K. Ganesan, P. K. Rathod, and D. T. Chiu, “A microfluidic model

for single-cell capillary obstruction by Plasmodium falciparum infected erythrocytes,”

Proceedings of the National Academy of Sciences of the United States of America, vol.

100, pp. 14618-14622, Dec 9 2003.

[128] A. M. Forsyth, J. D. Wan, W. D. Ristenpart, and H. A. Stone, “The dynamic behavior of

chemically ”stiffened” red blood cells in microchannel flows,”Microvascular Research,

vol. 80, pp. 37-43, Jul 2010.

[129] Y. Katsumoto, K. Tatsumi, T. Doi, and K. Nakabe, “Electrical classification of single

red blood cell deformability in high-shear microchannel flows,” International Journal

of Heat and Fluid Flow, vol. 31, pp. 985-995, 2010.

[130] S. S. Lee, Y. Yim, K. H. Ahn, and S. J. Lee, “Extensional flow-based assessment of

red blood cell deformability using hyperbolic converging microchannel,”Biomedical

Microdevices, vol. 11, pp. 1021-1027, Oct 2009.

Bibliography 117

[131] J. Chen, M. Chu, K. Koulajian, X. Y. Wu, A. Giacca, and Y.Sun, “A monolithic poly-

meric microdevice for pH-responsive drug delivery,”Biomedical Microdevices, vol. 11,

pp. 1251-1257, 2009.

[132] C. R. Gordijo, A. J. Shuhendler, and X. Y. Wu, “Glucose-responsive bioinorganic

nanohybrid membrane for self-regulated insulin release,”Advanced Functional Mate-

rials, vol. 20, pp. 1404-1412, May 10 2010.

[133] C. R. Gordijo, K. Koulajian, A. J. Shuhendler, L. D. Bonifacio, H. Y. Huang, S. Chiang,

G. A. Ozin, A. Giacca, and X. Y. Wu, “Nanotechnology-enabledclosed loop insulin

delivery device:in vitro and in vivo evaluation of glucose-regulated insulin release for

diabetes control,”Advanced Functional Materials, vol. 21, pp. 73-82, Jan 7 2011.

[134] J. Chen, C. R. Gordijo, M. Chu, X. Y. Wu, and Y. Sun, ”A monolithic micro device

for glucose-responsive insulin delivery,”in Proc. of Solid-State Sensor, Actuator and

Microsystems Workshop (Hilton Head 2010), Hilton Head Island, SC, USA, 6- 10 June

2010, pp. 266-269.

[135] T. B. Jones, Electromechanics of particles. Cambridge ; New York: Cambridge Univer-

sity Press, 1995.

[136] D. S. Gray, J. L. Tan, J. Voldman, and C. S. Chen, “Dielectrophoretic registration of

living cells to a microelectrode array,”Biosensors and Bioelectronics, vol. 19, pp. 1765-

74, Jul 15 2004.

[137] B. M. Taff and J. Voldman, “A scalable addressable positive-dielectrophoretic cell-

sorting array,”Analytical Chemistry, vol. 77, pp. 7976-83, Dec 15 2005.

[138] A. R. Wheeler, W. R. Throndset, R. J. Whelan, A. M. Leach, R. N. Zare, Y. H. Liao,

K. Farrell, I. D. Manger, and A. Daridon, “Microfluidic device for single-cell analysis,”

Analytical Chemistry, vol. 75, pp. 3581-6, Jul 15 2003.

Bibliography 118

[139] D. Forsyth and J. Ponce, Computer vision : a modern approach. Upper Saddle River,

N.J.: Prentice Hall, 2003.

[140] R. M. Hochmuth, “Micropipette aspiration of living cells,” Journal of Biomechanics,

vol. 33, pp. 15-22, Jan 2000.

[141] L. M. Broche, N. Bhadal, M. P. Lewis, S. Porter, M. P. Hughes, and F. H. Labeed, “Early

detection of oral cancer - Is dielectrophoresis the answer?” Oral Oncology, vol. 43, pp.

199-203, Feb 2007.

[142] S. Chin, M. P. Hughes, H. M. Coley, and F. H. Labeed, “Rapid assessment of early

biophysical changes in K562 cells during apoptosis determined using dielectrophoresis,”

International Journal of Nanomedicine, vol. 1, pp. 333-7, 2006.

[143] H. M. Coley, F. H. Labeed, H. Thomas, and M. P. Hughes, “Biophysical characterization

of MDR breast cancer cell lines reveals the cytoplasm is critical in determining drug

sensitivity,”Biochimica et Biophysica Acta, vol. 1770, pp. 601-8, Apr 2007.

[144] L. Duncan, H. Shelmerdine, M. P. Hughes, H. M. Coley, Y.Hubner, and F. H. Labeed,

“Dielectrophoretic analysis of changes in cytoplasmic ionlevels due to ion channel

blocker action reveals underlying differences between drug-sensitive and multidrug-

resistant leukaemic cells,”Physics in Medicine and Biology, vol. 53, pp. N1-7, Jan 21

2008.

[145] P. R. C. Gascoyne, X. B. Wang, Y. Huang, and R. F. Becker,“Dielectrophoretic separa-

tion of cancer cells from blood,”IEEE Transactions on Industry Applications, vol. 33,

pp. 670-678, 1997.

[146] C. Huang, A. Chen, L. Wang, M. Guo, and J. Yu, “Electrokinetic measurements of

dielectric properties of membrane for apoptotic HL-60 cells on chip-based device,”

Biomedical Microdevices, vol. 9, pp. 335-43, Jun 2007.

Bibliography 119

[147] F. H. Labeed, H. M. Coley, and M. P. Hughes, “Differences in the biophysical prop-

erties of membrane and cytoplasm of apoptotic cells revealed using dielectrophoresis,”

Biochimica et Biophysica Acta, vol. 1760, pp. 922-9, Jun 2006.

[148] J. Yang, Y. Huang, X. Wang, X. B. Wang, F. F. Becker, and P. R. Gascoyne, “Dielectric

properties of human leukocyte subpopulations determined by electrorotation as a cell

separation criterion,”Biophysical Journal, vol. 76, pp. 3307-14, Jun 1999.

[149] J. Chen, M. Abdelgawad, L. M. Yu, N. Shakiba, W. Y. Chien, Z. Lu, W. R. Geddie, M. A.

S. Jewett, and Y. Sun, “Electrodeformation for single cell mechanical characterization,”

Journal of Micromechanics and Microengineering, vol. 21, pp. 054012, May 2011.

[150] F. K. Glenister, R. L. Coppel, A. F. Cowman, N. Mohandas, and B. M. Cooke, “Contri-

bution of parasite proteins to altered mechanical properties of malaria-infected red blood

cells,” Blood, vol. 99, pp. 1060-3, Feb 1 2002.

[151] J. P. Mills, L. Qie, M. Dao, C. T. Lim, and S. Suresh, “Nonlinear elastic and viscoelastic

deformation of the human red blood cell with optical tweezers,” Mechanics and Chem-

istry of Biosystems, vol. 1, pp. 169-80, Sep 2004.

[152] G. B. Nash, E. O’Brien, E. C. Gordon-Smith, and J. A. Dormandy, “Abnormalities in

the mechanical properties of red blood cells caused by Plasmodium falciparum,”Blood,

vol. 74, pp. 855-61, Aug 1 1989.

[153] M. Paulitschke and G. B. Nash, “Membrane rigidity of red blood cells parasitized by dif-

ferent strains of Plasmodium falciparum,”Journal of Laboratory and Clinical Medicine,

vol. 122, pp. 581-9, Nov 1993.

[154] S. Suresh, J. Spatz, J. P. Mills, A. Micoulet, M. Dao, C.T. Lim, M. Beil, and T. Seuf-

ferlein, “Connections between single-cell biomechanics and human disease states: gas-

trointestinal cancer and malaria,”Acta Biomaterialia, vol. 1, pp. 15-30, Jan 2005.

Bibliography 120

[155] D. H. Kim, P. K. Wong, J. Park, A. Levchenko, and Y. Sun, “Microengineered platforms

for cell mechanobiology,”Annual Review of Biomedical Engineering, vol. 11, pp. 203-

233, 2009.

[156] S. A. Vanapalli, M. H. G. Duits, and F. Mugele, “Microfluidics as a functional tool for

cell mechanics,”Biomicrofluidics, vol. 3, 2009.

[157] O. Loh, A. Vaziri, and H. Espinosa, “The potential of MEMS for advancing experiments

and modeling in cell mechanics,”Experimental Mechanics, vol. 49, pp. 105-124, 2009.

[158] T. F. Coleman and Y. Y. Li, “An interior trust region approach for nonlinear minimization

subject to bounds,”Siam Journal on Optimization, vol. 6, pp. 418-445, May 1996.

[159] H. Zhang, A. L. Qu, J. Luo, and J. Luo, “Error analysis ofC-m measurement under

the whole-cell patch-clamp recording,”Journal of Neuroscience Methods, vol. 185, pp.

307-314, Jan 15 2010.

[160] M. Moscovici, W.-Y. Chien, M. Abdelgawad, and Y. Sun, “Electrical power free, low

dead volume, pressure-driven pumping for microfluidic applications,”Biomicrofluidics,

vol. 4, pp. 046501-10, 2010.

[161] J. Chen, Y. Zheng, Q. Y. Tan, Y. L. Zhang, J. Li, W. R. Geddie, M. A. S. Jewett, and Y.

Sun, “A microfluidic device for simultaneous electrical andmechanical measurements

on single cells,”Biomicrofluidics, vol. 5, pp. 014113, Mar 2011.

[162] C. A. Simmons, W. Y. Cheung, C. Liu, R. M. L. Tonelli-Zasarsky, and L. D. You, “Os-

teocyte apoptosis is mechanically regulated and induces angiogenesisin vitro,” Journal

of Orthopaedic Research, vol. 29, pp. 523-530, Apr 2011.

[163] G. Bradley, P. F. Juranka, and V. Ling, “Mechanism of multidrug resistance,”Biochimica

et Biophysica Acta, vol. 948, pp. 87-128, Aug 3 1988.

Bibliography 121

[164] R. M. Haralick, S. R. Sternberg, and X. H. Zhuang, “Image-analysis using mathematical

morphology,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 9,

pp. 532-550, Jul 1987.

[165] P. R. Twentyman, J. G. Reeve, G. Koch, and K. A. Wright, “Chemosensitisation by

verapamil and cyclosporin A in mouse tumour cells expressing different levels of P-

glycoprotein and CP22 (sorcin),”British Journal of Cancer, vol. 62, pp. 89-95, Jul 1990.

D M Devicesfor Drug Deliveryand Cellular Biophysics · Recent advances in micro technologies have...

Documents

Transcript of D M Devicesfor Drug Deliveryand Cellular Biophysics · Recent advances in micro technologies have...