IN SITU SENSING AND MODELING OF MOLECULAR EVENTS AT THE CELLULAR
LEVEL
By
Ruiguo Yang

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Electrical Engineering - Doctor of Philosophy
2014

ABSTRACT
IN SITU SENSING AND MODELING OF MOLECULAR EVENTS AT THE CELLULAR
LEVEL
By
Ruiguo Yang
We developed the Atomic Force Microscopy (AFM) based nanorobot in combination with
other nanomechanical sensors for the investigation of cell signaling pathways. The AFM nanorobotics
hinge on the superior spatial resolution of AFM in imaging and extends it into the measurement of
biological processes and manipulation of biological matters. A multiple input single output control system was designed and implemented to solve the issues of nanomanipulation of biological
materials, feedback, response frequency and nonlinearity. The AFM nanorobotic system therefore
provide the human-directed position, velocity and force control with high frequency feedback, and
more importantly it can feed the operator with the real-time imaging of manipulation result from
the fast-imaging based local scanning. The use of the system has taken the study of cellular process
at the molecular scale into a new level.
The cellular response to the physiological conditions can be significantly manifested in cellular
mechanics. Dynamic mechanical property has been regarded as biomarkers, sometimes even regulators of the signaling and physiological processes, thus the name mechanobiology. We sought to
characterize the relationship between the structural dynamics and the molecular dynamics and the
role of them in the regulation of cell behavior. We used the AFM nanorobotics to investigate the
mechanical properties in realtime of cells that are stimulated by different chemical species. These
reagents could result in similar ion channel responses but distinctive mechanical behaviors. We
applied these measurement results to establish a model that describes the cellular stimulation and
the mechanical property change, a “two-hit” model that comprises the loss of cell adhesion and

the initiation of cell apoptosis. The first hit was verified by functional experiments: depletion of
Calcium and nanosurgery to disrupt the cellular adhesion. The second hit was tested by a labeling of apoptotic markers that were revealed by flow cytometry. The model would then be able to
decipher qualitatively the molecular dynamics infolded in the regulation of cell behavior.
To decipher the signaling pathway quantitatively, we employed a nanomechanical sensor at the
bottom of the cell, quartz crystal microbalance with energy dissipation monitoring (QCM-D) to
monitor the change at the basal area of the cell. This would provide the real time focal adhesion
information and would be used in accordance with the AFM measurement data on the top of the
cell to build a more complete mechanical profile during the antibody induced signaling process.
We developed a model from a systematic control perspective that considers the signaling cascade
at certain stimulation as the controller and the mechanical and structural interaction of the cell as
the plant. We firstly derived the plant model based on QCM-D and AFM measurement processes.
A signaling pathway model was built on a grey box approach where part of the pathway map was
delineated in detail while others were condensed into a single reaction. The model parameters
were obtained by extracting the mechanical response from the experiment. The model refinements
were conducted by testing a series of inhibition mechanisms and comparing the simulation data
with the experimental data. The model was then used to predict the existence of certain reactions
that are qualitatively reported in the literature.

ACKNOWLEDGMENTS

I would like to express my most sincere appreciation to my advisor, Dr. Ning Xi, for his expert
guidance, generous encouragement and support for my research.
In addition, I would like to thank all my PhD committee members: Dr. Lixin Dong, Dr. Robert
J. McGough, Dr. Donna H. Wang and Dr. Animesh A. Sinha. They offered me timely help and
unfailing support that improve the technical soundness and the presentation of this dissertation.
Furthermore, I would like to express my gratitude to all my colleagues, especially to Dr. King
Wai Chiu Lai, Dr. Carmen Kar Man Fung, Dr. Hongzhi Chen, Bo Song, Zhiyong Sun, Liangliang
Chen, Jianguo Zhao, and Yunyi Jia, for their support in the experiments and discussion.
Last but not the least, I want to thank my parents and my wife, Jie Guo. This dissertation would
not have been possible without their years of encouragement and continuous support. With love
and gratitude, I dedicate this dissertation to them.

iv

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

x

Chapter 1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Sensing techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1 Structural sensing and force sensing . . . . . . . . . . . . . . . . .
1.2.2 Structural sensing at the cell level . . . . . . . . . . . . . . . . . .
1.2.2.1 High resolution imaging by fluorescence imaging . . . . .
1.2.2.2 Scanning Electron Microscopy (SEM) imaging . . . . . .
1.2.2.3 Stochastic optical reconstruction microscopy (STORM) .
1.2.2.4 AFM with Raman spectroscopy . . . . . . . . . . . . . .
1.2.3 Nanomechanical sensors for biological detection . . . . . . . . . .
1.2.3.1 Cantilever arrays and nano-resonators . . . . . . . . . . .
1.2.3.2 Traction force microscopy (TFM) . . . . . . . . . . . . .
1.2.3.3 QCM-D detecting energy dissipation in cell adhesion . . .
1.2.4 Force sensing at the cell level . . . . . . . . . . . . . . . . . . . . .
1.2.5 Force sensing at the molecular level . . . . . . . . . . . . . . . . .
1.2.5.1 Tip functionalization based on self-assembly . . . . . . .
1.2.5.2 Single molecule force microscopy (SMFM) . . . . . . . .
1.3 Cell signaling pathway . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Models of molecular kinetics . . . . . . . . . . . . . . . . . . . . .
1.3.2 Michaelis-Menten reaction . . . . . . . . . . . . . . . . . . . . . .
1.3.3 Pathway regulation of cell adhesion . . . . . . . . . . . . . . . . .
1.4 Objectives and challenges for AFM based nanorobotics integrated platform
1.5 Organization of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
1
3
3
5
5
6
8
8
9
9
10
10
11
11
11
12
14
15
15
16
20
22

Chapter 2
Nanobiomechanics in the identification of pathophysiological conditions
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Cell Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Membrane, Channels and Cytoskeleton . . . . . . . . . . . . . . . . . . .
2.2.2 Mechanical model of cells . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2.1 Continuum models . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2.2 Structural models . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3 Importance of cell adhesion in the establishment of cell mechanical states .
2.2.4 Different model could apply at different scenarios . . . . . . . . . . . . . .
2.3 Manipulation and measurements at the molecular level . . . . . . . . . . . . . . .
2.3.1 DNA manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

23
23
24
24
26
27
32
33
34
34
35

v

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

2.3.2 Single cell manipulation . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.3 Filament manipulation and characterization . . . . . . . . . . . . . . . .
2.3.4 Nanodissection and nanosurgery . . . . . . . . . . . . . . . . . . . . . .
Cellular nanosurgery mimics the effect of antibody treatment in cell-cell adhesion

.
.
.
.

35
36
37
38

Chapter 3
In situ sensing of ion channel activities after glucose stimulation . . . .
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 TRPV1 immunofluorescence . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Loading of Fur-2 AM to RINm5f cells . . . . . . . . . . . . . . . . . . .
3.2.3 Glucose and capsaicin stimulation and dosage-dependent study . . . . . .
3.2.4 Simultaneous fluorescence imaging of [Ca2+ ]i and AFM measurements on
RINm5f cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.5 Actin filament staining and imaging . . . . . . . . . . . . . . . . . . . .
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1 Immunofluorescence of TRPV1 on RINm5F cells . . . . . . . . . . . . .
3.3.2 AFM stiffness observation of dosage responses to glucose and capsaicin
stimulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.3 [Ca2+ ]i and stiffness observation of glucose stimulation . . . . . . . . . .
3.3.4 [Ca2+ ]i and stiffness observation of capsaicin stimulation . . . . . . . . .
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

39
39
40
42
42
43

.
.
.
.

44
44
45
45

.
.
.
.
.

45
47
49
51
55

2.4

Chapter 4
In situ biomarker detection during signaling events . . . . . . . . . . .
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Visualization of keratinocyte junctions under physiologic and disease conditions
by AFM methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Mechanical characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 AFM based quantification of cell surface stiffness reveals autoantibodyassociated nanostructural changes . . . . . . . . . . . . . . . . . . . . .
4.3.2 Desmosome disruption by other means . . . . . . . . . . . . . . . . . . .
4.4 Two-hit hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 An integrated model for antibody-mediated damage in the skin . . . . . . . . . .
4.6 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.

63
64
67
72
73

Chapter 5
Modeling and experimental verifications of cell tensegrity .
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Desmosome disruption leads to decrease in cell stiffness . . . . . .
5.2.1 Desmosome disassembly results in stiffness decrease . . . .
5.2.2 AFM based nanosurgery resulting in the decrease of stiffness
5.3 Quantitative modeling based on six struts tensegrity structure . . . .
5.3.1 Without intermediate filaments . . . . . . . . . . . . . . . .
5.3.2 With intermediate filaments . . . . . . . . . . . . . . . . . .
5.4 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

75
75
77
78
80
85
88
89
92

vi

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

. 56
. 56
. 59
. 63

Chapter 6
Dynamic characterization of cell signaling pathways mechanics
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1.1 AFM energy dissipation and hysteresivity measurements . . . . .
6.1.2 QCM-D energy dissipation measurement . . . . . . . . . . . . .
6.2 Model development . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.1 AFM viscoelastic characterization . . . . . . . . . . . . . . . . .
6.2.2 QCM-D based cell membrane peeling model . . . . . . . . . . .
6.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

95
95
97
99
101
101
102
105
111

Chapter 7
Plant model derivation . . . . . . . . . . . . . . . . . . . .
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Model development . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1 The dynamics model for the sensor disk . . . . . . . . . . .
7.2.2 The mechanical model of cell adhesion . . . . . . . . . . .
7.2.3 Gaussian compensation of oscillation amplitude . . . . . . .
7.2.4 Numerical simulation . . . . . . . . . . . . . . . . . . . . .
7.3 Result and discussion . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.1 The dynamics of the sensor disk and the kinetics of the bond
7.3.2 Contributions from different energy dissipation source . . .
7.3.3 Relationship of bond number and energy dissipation . . . .
7.3.4 Experimental observations . . . . . . . . . . . . . . . . . .
7.4 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

113
113
115
116
117
120
122
122
122
125
126
128
129

Chapter 8
Modeling of signaling pathway dynamics
8.1 Introduction . . . . . . . . . . . . . . . . . . . .
8.2 Model development . . . . . . . . . . . . . . . .
8.2.1 Plant model . . . . . . . . . . . . . . . .
8.2.2 The controller model . . . . . . . . . . .
8.3 Result . . . . . . . . . . . . . . . . . . . . . . .
8.4 Chapter summary . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

131
131
135
136
138
142
145

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

Chapter 9
Conclusion and future work . . . . . . . . . . . . . . . . . . . . . . . .
9.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.1.1 Development of AFM nanorobotics for biomedical applications . . . . .
9.1.2 Dynamical monitoring of cellular structural and mechanical changes under
patho/physiological conditions . . . . . . . . . . . . . . . . . . . . . . .
9.1.3 Nanomechanical characterization and nanorobotic surgery delineate cell
signaling pathway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.1.4 Cell adhesion model with nanomechanical oscillator . . . . . . . . . . .
9.1.5 A systematic control model for cell signaling pathway . . . . . . . . . .
9.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2.1 Signal transduction and propagation study . . . . . . . . . . . . . . . . .
9.2.2 Functional model of the pathway based on mechanical profile . . . . . .

vii

. 147
. 147
. 147
. 148
.
.
.
.
.
.

149
150
151
151
151
152

BIBLIOGRAPHY . . . . . . . . . . . . . .

viii

. . . . . . . . . . . . . . . . . . . 154

LIST OF TABLES

Table 5.1

Parameter list for the numerical evaluation . . . . . . . . . . . . . . . . . 92

Table 7.1

The parameters in the model with their physical meanings used for the
simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

Table 8.1

State variables and ODEs . . . . . . . . . . . . . . . . . . . . . . . . . . 140

Table 8.2

Rate constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

ix

LIST OF FIGURES

Figure 1.1

Schematic drawing of AFM working principle . . . . . . . . . . . . . . .

4

Figure 1.2

High resolution images showing the cytoskeleton structure under the applied peak force of 5 nN, scan size: 30 µm. . . . . . . . . . . . . . . . .

6

The combination configuration of fluorescence microscopy with AFM.
Simultaneous imaging of B lymphoma cells labeled with anti-CD20 (FITC
labeled) surface receptor and the AFM is shown. . . . . . . . . . . . . . .

7

Figure 1.3

Figure 1.4

Tip functionalization with the intermediate linker molecules. . . . . . . . 12

Figure 1.5

SAM formation onto goad coated AFM tip. . . . . . . . . . . . . . . . . 13

Figure 1.6

The binding force measuring process. . . . . . . . . . . . . . . . . . . . 13

Figure 1.7

A: The diagram of cell adhesion regulation pathways. B: Image of focal
adhesion by florescence. C: AFM image of the living cells with focal
adhesion complex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Figure 2.1

A: The conversion of force-displacement curve to force-indentation curve;
B: The fitting of the force indentation curve by Hertz model to generate
the Youngs modulus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Figure 2.2

A: The cell structure is based on cytoskeleton tensegrity. B: Tensegrity
structures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Figure 3.1

A: The stimulus-secretion coupling model of β -cells. Increased plasma
glucose levels will increase the intracellular ATP/ADP ratio after glucose being uptaked by cells, which leads to the closure of ATP-dependent
potassium channels, membrane depolarization, and the increase in intracellular Ca2+ concentrations through the activation of voltage dependent
calcium channels (VDCC) and ultimately insulin secretion. Capsaicin
binds to the intracellular domain of the TRPV1 receptor and causes the
influx of Ca2+ , which triggers the release of Ca2+ into the cytoplasm from
the endoplasmic reticulum (ER) and increases [Ca2+ ]i concentrations. B:
The AFM based nanorobotic manipulator uses a joystick as position input
to facilitate the movement of the AFM tip in the planar area during the
force measurement process. . . . . . . . . . . . . . . . . . . . . . . . . . 41

x

Figure 3.2

Immunofluorescence imaging of TRPV1 receptors in RINm5f insulinoma
cells. A: Cell plasma membrane localization of TRPV1 receptors in
RINm5f cells, positive staining indicated by arrows. B: Control RINm5f
cells in which primary antibody was replaced with 1% BSA in PBS show
negative staining. Scale bar: 20 µ m. C: The stiffness changes before and
after administration of glucose at three concentrations, 16 mM, 20 mM,
and 40 mM, plus the control. * p<0.05 vs. before stimulation; p<0.05
vs. control; p<0.05 vs. glucose 16 mM; p<0.05 vs. glucose 16 mM and
20 mM (n=20). D: The stiffness changes before and after administration
of Cap (capsaicin) at four concentrations (1.67 nM, 16.7 nM, 167 nM
and 1.67 µM), with *p<0.05 vs. before stimulation; p<0.05 vs. control;
p<0.05 vs. Cap 1.67 nM; p<0.05 vs. Cap 1.67 nM and 16.7 nM. Each
datum is expressed as mean SEM (n=100). . . . . . . . . . . . . . . . . 46

Figure 3.3

Dynamic characterization of cellular stiffness simultaneously with [Ca2+ ]i
fluorescence imaging after glucose (16 mM) stimulation. A: The ion
channel activities overlaps with the overall cellular stiffness change in
the first phase when the onset of stimulation induced rapid structural rearrangement. A second sustained phase follows to maintain the structural
configuration and stiffness at a relatively higher level than the baseline.
B: A close-look at the lagging of stiffness response compared with that of
intracellular Ca2+ concentrations, a 2 minute-delay was observed. C, D,
E: The fluorescence imaging of [Ca2+ ]i before stimulation (C), 12 minutes after stimulation when the light intensity reaches the peak (D), and
20 minutes after stimulation (E) when the light intensity falls to the baseline. 48

Figure 3.4

Dynamic characterization of cellular stiffness simultaneously with [Ca2+ ]i
fluorescence imaging after capsaicin (16.7 nM) stimulation. A: The ion
channel shows similar pulse-shape response pattern as that of glucose
albeit with smaller peak value. B: The cells were also pretreated with
Capz 30 minutes before capsaicin stimulation to block TRPV1 receptors.
No increase in light intensity was observed for TRPV1-blockade samples
shown in the normalized gray scale before and after capsaicin stimulation. *p<0.05 vs. before stimulation; p <0.05 vs. Cap 16.7 nM. C, D, E:
The fluorescence imaging of [Ca2+ ]i before (C), 12 minutes after stimulation (D) when the light intensity reaches the plateau, and 20 minutes
after stimulation (E) when the light intensity falls to the base line. The
normalized gray scale value was obtained the same way as the Fig. 3.3. . 50

Figure 3.5

The overall trend of the stiffness responses. A: Glucose stimulation resulted in a two-phase response: a rapid increase phase in stiffness followed by a sustained elevation phase; B: Capsaicin stimulation induced a
slow and sustained increase in stiffness. . . . . . . . . . . . . . . . . . . 52

xi

Figure 3.6

The actin filament reorganization after treatment of low glucose based
medium as control, 16 mM glucose and 16.7 nM capsaicin for 0 min, 1
min, 5 min and 10 min. F-actin was labeled with CF488A-conjugated
phalloidin. The treatment of glucose and capsaicin causes time-lapse
change of cytoskeleton especially actin filament structures with distinct
patterns. A delayed disorganization of actin around 10 minutes after stimulation was observed in capsaicin treatment, whereas glucose triggered
the disorganization 1 minute after treatment, similar with the biophysical
data obtained by AFM nanomechanical analysis. No distinctive change
was observed for the control sample. . . . . . . . . . . . . . . . . . . . . 53

Figure 4.1

Integrated SEM and AFM imaging of intercellular adhesion structures.
A: Correlation of SEM imaging with AFM imaging. Identically cultured
plates of confluent HaCaT cells were fixed and imaged by SEM and Multimode AFM in increasing magnifications. The lower magnification images (SEM: A1, A2; AFM: B1, B2) show the cells with clear boundaries
between neighboring cells where cell-cell adhesion occurs. The higher
magnification images (SEM: A3, A4; AFM: B3, B4) show details of the
adhesion junction with strand-shaped structures in parallel distribution
between two cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

Figure 4.2

Integrated immunofluorescence (IF) and AFM imaging of intercellular
adhesion structures. The same area on a confluent slide of HaCaT cells
was captured simultaneously by IF and AFM after fixation of the cells.
For IF imaging, HaCaT cells were labeled with anti-cytokeratin antibodies (red) and anti-desmoplakin antibodies (green). AFM images were
captured by Bioscope AFM at increasing resolution with scan sizes of
100 µm (B2), 50 µm (B3) and 20 µm (B4). C: Correlation of IF imaging (C1) with AFM imaging [scan sizes of 100 µm (C2), 50 µm (C3)
and 20 µm (C4) after treatment with 10 g/ml of the pathogenic anti-Dsg3
antibody Px4-3 for 24h . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Figure 4.3

AFM imaging of autoantibody mediated desmosomal disruption in fixed
cells. HaCaT cells were grown to confluence, and left with no antibody
treatment, or were treated for 24h with (A) the pathogenic anti-Dsg3 antibody Px4-3, (B) non-pathogenic antibody Px4-4, (C) non-binding irrelevant control antibody goat anti-mouse Ig, or (D) irrelevant control
antibody . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

Figure 4.4

AFM imaging of autoantibody mediated desmosomal disruption in live
cellsconfluent HaCaT cells were imaged live before antibody treatment
or treated with the antibodies as in Fig. 4.3 for 30 min, 1 h, 2 h, 3 h, and 4 h 62

xii

Figure 4.5

AFM-guided elasticity measurements on HaCaT cells before and after
treatment with anti-Dsg3 and control antibodies. At each time point, between 20 and 40 HaCaT cells were tested, and the force curves obtained
from these measurements were processed to generate the corresponding
Young’s modulus as a measure of cellular elasticity. In parallel experiments, measurement were obtained before antibody treatment and after
treatment with 10 g/ml of the pathogenic anti-Dsg3 antibody Px4-3 (B),
the non-pathogenic antibody Px4-4 (C), the non-binding irrelevant control antibody goat anti-mouse Ig (D).*p<0.05. . . . . . . . . . . . . . . . 63

Figure 4.6

Cell desmosome disruption by calcium depletion. . . . . . . . . . . . . . 65

Figure 4.7

Apoptotic processes are induced by nonpathogenic and pathogenic antiDsg3 antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Figure 4.8

Apoptotic processes are induced by pathogenic anti-Dsg3 antibodies detected by flow cytometry and AFM nanomechanical analysis. A: No treatment; B: pathogenic antibody treatment and C: pathogenic antibody treatment blocked by FasL inhibitor . . . . . . . . . . . . . . . . . . . . . . . 70

Figure 4.9

The pathogenic anti-Dsg3 antibody-induced increase in keratinocyte stiffness can be blocked by caspase inhibitor, but not by Fas ligand neutralizer

71

Figure 5.1

Desmosome structure with different desmosomal proteins connecting each
other all the way to the intermediate filaments. AFM image and zoom-in
(B) correlates with fluorescence image (C) . . . . . . . . . . . . . . . . . 77

Figure 5.2

Young’s modulus comparison for normal and calcium depleted cells . . . 79

Figure 5.3

A: Illustration of composite cell structure during cell indentation and
dissection; B: Force-displacement curve without penetration of the cell
membrane; C: Force displacement curve with penetration of the cell membrane with the kink zoomed-in in the inset. . . . . . . . . . . . . . . . . . 81

Figure 5.4

A:The lateral force was the laser position shift in the horizontal direction
on the detector; B: The tangent direction of the trajectory defines the angle (θ ) between the cutting force F and the lateral force Fl ; C: The trajectory of the nanosurgery on the cell peripheral between neighboring cells
with the black triangle indicating the tip position; D: the XY position,
cutting force and the Z piezo position in the process of the nanosurgery
along the path in C with a cutting depth of 100 nm. . . . . . . . . . . . . 82

Figure 5.5

Nanodissection of intermediate filaments by AFM (before and after) with
cross section to indicate the cut-off bundle . . . . . . . . . . . . . . . . . 84

xiii

Figure 5.6

The stiffness response before and after AFM based nanodissection. . . . . 84

Figure 5.7

The configuration of the tensegrity structure without intermediate filaments, the black lines indicate the microtubule as the compressive elements and the blue line indicate the actin filament as the tensional elements 86

Figure 5.8

Young Modulus comparison with and without intermediate filaments from
simulation based on the structure model . . . . . . . . . . . . . . . . . . 93

Figure 6.1

AFM energy dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

Figure 6.2

Schematic of QCM-D principle . . . . . . . . . . . . . . . . . . . . . . . 100

Figure 6.3

The modeling of viscoelastic material under AFM based force displacement measurement during the retract process . . . . . . . . . . . . . . . . 101

Figure 6.4

Cell membrane peeling model at the basal area (A, B); and the cell model
as a whole (C) in which the cytoplasm was modeled as a Voigt element. . 103

Figure 6.5

AFM energy dissipation measurement on a single cell with and without
EGF treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Figure 6.6

AFM energy dissipation measurement on 100 different cells before and
after EGF treatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Figure 6.7

Dynamic mechanical responses (Energy dissipation) of A431 cells to
stimulations with four different concentrations of EGF: 0, 10 nM, 20 nM,
and 40 nM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

Figure 6.8

Dynamic mechanical responses (Hysteresivity) of A431 cells to stimulations with four different concentrations of EGF: 0, 10 nM, 20 nM, and 40
nM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Figure 6.9

QCM-D measurement of energy dissipation over time after EGF treatment with concentration of 0 nM, 10 nM, 20 nM and 40 nM . . . . . . . 108

Figure 6.10

AFM energy dissipation simulation result shows that it takes longer time
for the deformation to recover for a cell with higher viscosity η as indicated with η = 0.45 showing larger deformation overall. . . . . . . . . . 109

Figure 6.11

Cell peeling model simulation . . . . . . . . . . . . . . . . . . . . . . . 110

Figure 7.1

The QCM-D/cell focal adhesion energy dissipation model . . . . . . . . . 116

xiv

Figure 7.2

A: The initial spring displacement Ls0 − Lsr will be varied according to a
Gaussian profile of the oscillation amplitude A = Ls0 − Lsr with the maximum amplitude A0/2π of 1 nm and a variance of 3.16; B: Illustration
of the sensor disk with manually divided areas with equal distribution of
bonds, each was assigned with one amplitude. . . . . . . . . . . . . . . . 121

Figure 7.3

The simulation result at a given initial displacement of 1 nm for a period
of 5 µs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

Figure 7.4

Dynamic energy dissipation for the three channels . . . . . . . . . . . . . 125

Figure 7.5

The energy dissipation channel contribution with respect to the bond
number. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Figure 7.6

The energy dissipation increases linearly with the bond number. . . . . . 128

Figure 7.7

The simulation data comparison with the experimental observation . . . . 129

Figure 8.1

Diagram of control model of pathway dynamics . . . . . . . . . . . . . . 135

Figure 8.2

The exponential relationship of the input and output for the plant model. . 138

Figure 8.3

The system of molecules that ODEs were built upon. . . . . . . . . . . . 141

Figure 8.4

The fitting result for EGF treatment of A431 cells based on the control
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

Figure 8.5

Expression levels of MEK1/2, Rho and PLCγ . . . . . . . . . . . . . . . 144

Figure 9.1

EGF stimulated intra and intercellular signaling propagation: A1, no activation; A2, global activation; A3, Local activation. B1, Local stimulation; B2, Intercellular propagation . . . . . . . . . . . . . . . . . . . . . 152

xv

Chapter 1
Introduction

1.1

Overview

Nanobiotechnology, the application of nanotechnology to the characterization and analysis of biological systems is a fast evolving field that represents the intersection of nanoscale physics, biology
and engineering [1]. The centerpieces of this technology are the instrumentation and techniques
that allow for the sensing and manipulation of nanoscale structures and molecules. One of the
key instruments is the Scanning Probe Microscopy (SPM) which measures the near-field physical
interactions, mainly van der Waals forces, between the scanning probe and the materials underneath [2]. Atomic force microscopy (AFM), a member of the SPM family, has taken a front seat
in driving new developments of nanobiotechnology, and there is now ever-increasing emphasis on
applying nanoscale forces to investigate biological events.
AFM was developed by Binnig and Quate in 1986 [3], with the first commercialized AFM
available in 1989. AFM remains a high resolution imaging tool mainly used for surface science
research, helping researchers visualize materials, such as semiconductors, at nanometer scale [4,5]
or even sub-nano meter level to resolve atomic structures. Even today, new material characterizations still rely on the high spatial resolution capability of AFM, such as the identification of single
layer graphene [6, 7]. However, the application of AFM in semiconductor and material sciences
are being overwhelmed by new breakthroughs in AFM guided biological investigations [8–13].
The extension of AFM into nanobiotechnology relies on its unique working features that are par1

ticularly well suited for biological studies [14–16]. First, AFM can be operated in several environments, including vacuum, air and liquid, the most important of which is the liquid environment
where normal physiological conditions for biological matters can be maintained. This is a key factor for AFM studies as compared with other high resolution imaging techniques such as Scanning
Electron Microscopy (SEM). Second, due to the nature of its measurement properties, AFM does
not require complicated sample preparation where sample contamination and damage can possibly be introduced. For example, electron microscopy requires sample fixation, dehydration and
coating processes before each observation. Third, measurement using a sharp probe, as is done
in AFM, is less destructive compared with many other commonly used imaging techniques. Furthermore, AFM measurement in stable conditions delivers a relatively high signal to noise ratio
(SNR), allowing high image resolution at the nano and even sub-nano meter scale. With this level
of resolution the membrane structure of single cells, single proteins and DNA molecules can be
visualized in unparalleled detail.
The AFM tip can be controlled to mechanically pull, push and manipulate nanoscale objects,
working as a nanorobotic arm under precise computer control. Therefore, traditional concepts
in robotics such as feedback, stability, and frequency response can all be integrated into these
miniatured nanorobotic systems. AFM based nanorobotics has been a tremendously useful tool in
nanomanufacturing, as demonstrated by its ability to perform nanolithography, nanoassembly and
manipulating nano particles [17–19]. In biology, nanomanipulation using the AFM based robotics
system has substantially extended the reach of AFM in various biological fields. One such application is to functionalize the AFM tip to conduct so called single-molecule manipulations. Single
strand intermediate filaments have been stretched by a controlled AFM tip to measure their tensile
strength [20]. Nanosurgery has been performed on cell associated molecules with a resolution of
less than 100 nm [21]. Individual DNA and protein molecules have been manipulated. The term
2

single molecular force spectroscopy (SMFS) has been applied when using the technology to probe
forces at the molecular level on surface membranes [22, 23]. The cellular response to varying
physiological conditions is often manifested as alterations in cellular mechanics. A recent AFM
study [24] has shown that during apoptosis cytoskeleton elements such as actin filaments will form
a thick layer wrapping around the collapsed nucleus, thus resulting in a time lapse stiffness change;
first a decrease and then an increase. Dynamic mechanical properties such as elasticity and viscoelasticity can be revealed by AFM and may serve as biomarkers, or even regulators of signaling
events and physiological processes. Furthermore, AFM allows these changes in biological structures to be visualized through high resolution images. Thus, AFM based nanorobotic operations
are proving to be an exciting and novel technology that may be uniquely suited for the capture of
biologically relevant information linking mechanostructural and functional events.

1.2
1.2.1

Sensing techniques
Structural sensing and force sensing

The imaging principle of AFM is relatively simple. A cantilever with a sharp tip is operated to scan
across the sample surface to probe its topography. The main functioning unit is the piezoelectric
actuator which drives the cantilever both horizontally, the XY scanning direction, and vertically,
the Z motion direction. During the XY scanning, the interaction force (mostly van der Waals
forces and sometimes electrostatic forces) between the scanning AFM tip and the sample surface
will cause the bending of the cantilever. A laser beam which is reflected from the back of the
cantilever is able to record this bending with a position sensitive device (PSD). The Z piezo will
move accordingly to keep the tip at a constant oscillation (tapping mode) or in contact with the
sample (contact mode). A schematic of the working principle of AFM is shown in Fig. 1.1. By
3

X, Y, Z
piezoelectric
actuation

Laser beam
PSD
Cantilever

Sample

Figure 1.1 Schematic drawing of AFM working principle
Fig. 1. A schematic drawing of the AFM imaging principle.

this simple mechanism, complex biological structures can be visualized with nanoscale detail.
When imaged in liquid, biological samples are often rendered soft and delicate. Therefore, the
contact force must be well controlled to avoid damage to the sample surface. Typically for contact
mode AFM imaging, the applied force should be adjusted to less than 100 pN for live cell imaging
(this value varies for different cell types). At this interaction force, the deformation of the cell
membrane can be recovered [25]; any force larger than this value would pose potential irreversible
damage to the cell membrane and cytoskeleton. However, the applied force can also be used
to highlight the stiffer structures underneath the soft cell membrane. In this case, the AFM tip
will press against the soft cell membrane into the cytoplasm until the membrane is supported by
the underlying filaments structures, allowing the underlying structures to be highlighted in both
the topography and deflection images. The most important aspect for this imaging mode is to
accurately control the applied force.
Not only is the normal force applied by contact mode a potential damaging factor for biological
samples, but the lateral or frictional force that “scratches” the surface must also be considered. The
development of tapping mode AFM imaging has proved a good solution to this problem [26]. By
oscillating the cantilever at its resonance frequency in liquid with a pre-defined amplitude, the
sharp tip at the end of the cantilever is free of contact with the sample. The interaction force
between the sample and the tip reduces the oscillation amplitude and also induces a phase shift.
4

By controlling the vertical piezo to maintain the desired amplitude, topography information can be
obtained. In fact, the phase shift provides indirect, quasi-quantitative data for mechanical property
analysis.
The ever-developing techniques for AFM imaging of biological samples have led to a number
of breakthroughs in the biological sciences. The surface structure of bacteria cells was visualized with lateral resolution around 2 nm with an applied force around 80 pN [27]. Membrane
protein structures have been revealed with tetramer configurations at sub-angstrom resolution in
the vertical direction. Animal cells have been imaged live to obtain structural information on cell
membranes or underlying cytoskeleton structures [28]. At an even larger scale, cartilage tissues
were imaged to determine their fibril composite [29]. These experiments truly demonstrate the spatial resolving capability of AFM for biological materials under physiological conditions; in some
cases this has been used to confirm crystallography data for defined structural conformations.
The live cell images in Fig. 1.2 show the topography and the surface detail of keratinocytes in
phosphate buffered saline (PBS). The cytoskeleton beneath the cell membrane was highlighted by
the applied force, revealing actin filaments at the center of the cell and intermediate filaments at
the peripheral connecting neighboring cells.

1.2.2

Structural sensing at the cell level

1.2.2.1

High resolution imaging by fluorescence imaging

Several studies have utilized a combination of AFM with optical microscopy techniques to localize
desired targets (schematically shown in Fig. 1.3). For the majority of these studies, AFM is used
to image the topography of a sample while a fluorescence image is collected simultaneously. The
fluorescence image is used to guide the AFM tip to certain locations where a specific agent is

5

Fig. 2.
High
resolution
images
showing
the cytoskeleton
structure
under theunder
applied
forcepeak
of 5 nN,
Figure
1.2
High
resolution
images
showing
the cytoskeleton
structure
thepeak
applied
force
of 5 nN, scan size: 30 µm.
scan size: 30 μm.

labeled, allowing detailed structural images to be obtained around the labeled molecule. In other
cases, the AFM tip is employed as a mechanical device to manipulate the sample while at the same
time conducting fluorescence imaging. A few papers have examined the mechanical properties of
single molecules by using AFM to stretch them and simultaneously visualize the stretching process
by fluorescence imaging [30]. However, a concern with this combination of techniques is the
interference between the excitation light and the emission light from the fluorescence microscopy
and the laser for AFM deflection recording. The laser used for cantilever deflection detection
has a wavelength in near infrared (NIR) ranging from 700 to 1200 nm [31], while in most cases
fluorescence excitation and emission wavelength is less than 700 nm, in the visible light range
space; for FITC the excitation and emission wavelengths are 480 and 560 nm, respectively shown
in Fig. 1.3.

1.2.2.2

Scanning Electron Microscopy (SEM) imaging

If we consider the combination of fluorescence imaging with AFM to be more of an instrumental
combination, then AFM combined with SEM could be considered as a functional combination. A

6

XY Scanning
Piezo
PSD
700-1200 nm
Tip
AFM image from
Z piezo motion
Objective
lens
Dichroic
mirror
Emission
filter
560 nm

Z Direction
Piezo

Laser
Cantilever

Sample

Excitation
filter
480 nm Mercury
lamp

Figure
Thecombination
combination
configuration
of fluorescence
microscopy
with
AFM. Simultaneous
Fig.1.3
3. The
configuration
of fluorescence
microscopy
with AFM.
Simultaneous
imaging of
imaging of B lymphoma cells labeled with anti-CD20 (FITC labeled) surface receptor and the
AFM Bislymphoma
shown. cells labeled with anti-CD20 (FITC labeled) surface receptor and the AFM is shown.
few investigators have used SEM to confirm structures observed in AFM. SEM has much larger
scanning area, and thereby can be used to complement the AFM measurement. In one study, SEM
was operated in low acceleration voltage to observe vesicle structures on the cell membranes [32].
Vesicle pits normally contain a substantial amount of water. Thus, when imaged by AFM under
low loading force, the pit structure may not be elevated. Only when a certain contact force is
applied on the surface can the pit be visualized with higher contrast between the water-filled area
and the unfilled area. However, this pit structure can be directly observed under SEM. Before
SEM imaging, samples need to be dehydrated. Therefore, the water containing vesicles will be
highlighted after shrinkage due to the drying process. The inhomogeneity of the membrane surface
can then be confirmed.

7

1.2.2.3

Stochastic optical reconstruction microscopy (STORM)

A recent advancement of fluorescence microscopy has enabled, in principle, the optical imaging
that reaches molecular scale [33], which was prevented by light diffraction in traditional optical
imaging. The technique, STORM, short for stochastic optical reconstruction microscopy employs
fluorophores that can be switched on and off at different measurement cycles to enable the precise
measurement of positions of subset of fluorophores. When a subset of fluorophores was switched
on, it effectively eliminates the light overlapping and allows the positions of fluorophores to be
determined precisely.

1.2.2.4

AFM with Raman spectroscopy

The combination of AFM with Raman spectroscopy was introduced approximately ten years ago
termed as tip enhanced Raman spectroscopy (TERS) [34]. This method takes advantage of the high
spatial resolution of scanning probe microscopy to obtain topography images while performing
surface chemical sensing at the nanoscale. The AFM tip is coated with metal (normally silver) to
further enhance the Raman signal. When the laser is irradiated onto the metal surface of the AFM
tip closely engaged with the sample surface, field- enhancement occurs to excite molecules around
the tip that allows Raman spectroscopy with a higher lateral resolution as compared with surface
enhanced Raman spectroscopy (SERS). The most common application of TERS in bio-imaging is
to chemically characterize biomaterials [35]. Further development and refinement of this method
is required before it can be routinely applied to more complex biological samples and processes.

8

1.2.3

Nanomechanical sensors for biological detection

The capability of AFM was further enhanced by the introduction of other modes of detection,
e.g. capacitance, resistance and magnetisms. These detection modes leverage on the scanning
capabilities of AFM while provided new dimensions of information as sensors. Recently, the
development of micro- and nanoscale cantilevers utilized the same physical principle introduced
new flexibility as well as accuracy to the sensing of physical interactions at the molecular level.
Meanwhile, other type of sensors are trending towards the label free detection of force interactions
during biological processes, such as optical and acoustic sensors.

1.2.3.1

Cantilever arrays and nano-resonators

With the similar concept as in an AFM, microfabricated cantilevers have been contributing in the
detection of various biological phenomenons [36–38]. The versatility of cantilevers have been
demonstrated with mechanical, electrostatic or piezoelectric drives and detections. The nature of
these devices as nano-resonators enables them to sense extremely small mass changes or force
interactions, thus a high sensitivity is guaranteed in the detection of gases, chemical compounds,
biological receptors or DNA hybridizations. With the introduction of multi-cantilever arrays [39],
the limitation of low throughput has been effectively overcome. There have been cantilever arrays
coated with sugar contents for the detection of certain mutations in cancer cells [40], and nanoresonators monitoring the response of bacterial cells to specific antibiotics [41]. However, most of
these applications demonstrate exceptional accuracy in air, once working in liquid environments,
the viscous drag of the fluid will significantly reduce the quality factor of the sensor, rendering it
ineffective in numerous cases [42].

9

1.2.3.2

Traction force microscopy (TFM)

The label free detection of physical forces at the cell-cell and cell-extracellular matrix (ECM) interaction poses another set of applications for nanomechanical sensors. The force interactions at
the cell-ECM have been reported to regulate the normal growth of the cell, migration, metastasis
and so forth through the focal adhesion-orchestrated cell signaling pathways [43]. Thus to quantify
and map out the distribution of cell-ECM interactions is crucial to the understanding of the molecular mechanisms underlying the regulation processes. TFM utilizes the optical property to monitor
the substrate displacement caused by the cellular traction forces. The displacement normally is
revealed by the fluorescence beads that are embedded into the substrates and are coated with ECM
binding agents. It could provide the three dimensional force profile of the cell-ECM interface [44].
A special group of force measurement at the cell-ECM adhesion sites uses microfabricated
pillars [45] with same geometrical structure but different length, thus different elasticity, to indicate
the displacement caused by traction. The variations in matrix elasticity have been reported capable
of directing the differentiation of stem cells [46]. These revealing techniques, though effective,
will pose a different growth environment for cells from their natural inhabitant, or even from the
long established two dimensional petri-dish based culturing system. Thus, the effect may very well
be the

1.2.3.3

QCM-D detecting energy dissipation in cell adhesion

Quartz Crystal Microbalance with Dissipation (QCM-D) is another nanomechanical senor that
could measure interactions at the molecular level [47]. It has been used for molecular recognitions,
such as membrane formations [48], protein adsorption [49], and cell spreading [50]. It operates at
thickness-shear mode by oscillating a quartz crystal disk with high frequency (normally 5 MHz)

10

and low amplitude (less than 1 nm in lateral direction). It collects the frequency and amplitude
changes of the quartz crystal and reveals the mechanical/structural status of the sample attached
on top of the sensor disk: the shift in resonance frequency indicates the mass absorption or loss;
while the energy dissipation alteration could be used to derive the viscoelasticity property of the
adsorbed film on the disk [51, 52].

1.2.4

Force sensing at the cell level

Mechanical properties of structures can be characterized by recording force displacement curves
via AFM nano indentation. Following Hertzian model analysis of force curves, a quantitative
Young’s modulus can be generated. For this purpose, the AFM tip is driven by piezoelectric actuator along its Z direction towards the sample and then retracted back over a pre-defined distance.
The movement of the piezoelectric actuator is then used to drive the AFM cantilever vertically and
the deflection signal from the cantilever is recorded as a force-displacement curve.

1.2.5

Force sensing at the molecular level

1.2.5.1

Tip functionalization based on self-assembly

Functionalization of AFM tips by coating them with antibodies or ligands has extended the application of AFM for studying specific interactions at a molecular level, e.g. binding force for molecular interaction or molecular recognition. The functionalization process is mainly based on the
self-assembly ability of chemical agents at molecular level. There are three ways to functionalize
the AFM probes with functional agents. The simplest method is by direct coating of a functional
protein to the AFM tip. While this is not technically difficult, the binding strength here may not
be strong enough. A common strategy to improve the binding is to employ certain crosslinking

11

Cleaned
AFM tip

Amination

Protein
molecule

NH2

Linker
molecule
Thiolation

PEG
SH

PEG
SH

Fig. 13. Tip functionalization with the intermediate linker molecules.
Figure 1.4
Tip functionalization with the intermediate linker molecules.

reagents to link the functional agents onto the AFM probe. The intermediate linking method is a
much more complicated procedure, but it results in a larger binding force between the AFM tip
and the functional agents. It also provides a higher degree of spatial specificity. This functionalization method usually employs a linker to covalently bind the proteins in order to expose specific
site(s) within selected molecules. Polyethylene glycol (PEG) is a common crosslinking reagent.
A terminal thiol group can be first attached to the antibody or protein and this thiol group will
be able to covalently bind to PEG [53]. The AFM tip should be cleaned and processed to add
an amine group layer, which can then be cross-linked to PEG. This functionalization process is
schematically shown in Fig. 1.4. One of the most common ways to functionalize the AFM tip is
by the formation of Self-Assembled Monolayers (SAMs) [54]. The formation of SAM is achieved
by thiol chemistry, similar to the coating of polymer PEG. After the initial binding of thiol with
gold, the large space between molecules enables the thiolyated alkyls to be horizontally oriented.
As the density of the horizontally aligned alkyls increases, the lateral interaction through van der
Waals forces drives the alkyls to lift off and finally align upright as depicted in Fig. 1.5.

1.2.5.2

Single molecule force microscopy (SMFM)

The AFM tip was first brought into contact with the cell sample (Process A in the Fig. 1.6). When
there is affinity binding activity between the ligand coated AFM tip and the surface receptor, the
12

Thiolyated
alkyls
Gold coated
AFM tip

Figure 1.5Fig.
SAM
onto onto
goadgoad
coated
AFM
tip.tip.
14.formation
SAM formation
coated
AFM
A

D

Approach
Retract

Force

B

C

B
A
C

D

Z Displacement

PEG Ligand Receptor Cell

Figure 1.6 The binding force measuring process.

Fig. 15. The binding force measuring process.

adhesive forces dominate the interaction between the AFM tip and the sample below (Fig. 1.6B).
When the cantilever is pulled back from the sample, the adhesive interaction drag back the cantilever and this is detected by the high sensitivity cantilever bending (Fig. 1.6C). The bending of
the cantilever through the whole process A to D is recorded as a function of the vertical displacement of the piezo drive. From the force distance curve, this dragging effect can be calculated as
the force difference between the baseline and the minimum. Normally, drag effect always exists
when the AFM cantilever pulls back from the sample due to the van der Waals force, but the force
is quite small. To obtain the binding force for single bonds, more detailed analyses would need to
be performed.

13

1.3

Cell signaling pathway

Cell signaling is a system of communications that governs various cell activities. The ability of
cells to perceive the environment and respond accordingly determines the development and the
fate of the related cellular system, the malfunction of which would cause serious consequences,
diseases such as diabetes and cancer. Cellular signaling pathways have been studied extensively
in biological context using traditional methods such as protein pooling. Epidermal growth factor
(EGF) is a ligand for a cell surface receptor epidermal growth factor receptor (EGFR). Its activation by EGF can trigger a cascade of downstream signaling events. The A431 human epidermoid
carcinoma cell line has been widely accepted as a model for this type of studies. The EGF stimulated A431 cells have been reported to exhibit significant rounding and stiffening, a phenomenon
caused by cytoskeleton reorganization governed by cell signaling. We previously reported that
these changes could be detected by AFM based mechanical measurements, in which both cellular
stiffness and energy dissipation increased. Thus we conclude that a mechanical biomarker could
be successfully established to signify the EGF stimulation and activation. Systems biology studies normally try to map out the underlying structure of cell signaling network and the flow of
signal cascade. These networks are quite complex in their organizations. By far the traditional
reductionist approach has failed to elucidate the decisive signaling reactions and molecules that
are responsible for the cellular functions, which are often integrated outputs of multiple signaling
pathways. A quantitative understanding of regulation of cellular functions is also seriously lacking. Thus, it becomes necessary to introduce computational modeling approach at the system level
to examine cellular functions in order to determine the responsible properties and interactions of
the cellular components and assess how these components work together to generate an integrated
downstream signal output in response to various extracellular cues.

14

1.3.1

Models of molecular kinetics

In post-genomic research, high-throughput gene and protein profiling techniques have enabled the
monitoring of biological systems in terms of expressed gens and proteins in real time at reasonable cost, and these information have helped people identify the signaling cascaded in various
biological process. The identification of pathway could take a different tone as in a sequence of
enzyme-catalyzed reactions of small molecules or signaling propagation though a set of protein
phosphorylation reactions and gene regulations. The investigation would thus involve the analysis
of protein-protein physical interaction, the kinetics simulation of small molecules or the steady
state pathway analysis.

1.3.2

Michaelis-Menten reaction

Enzyme kinetics is at the center of all the singling cascade events. Enzymes are highly specific
catalysts of biological systems and accelerate the chemical reaction of molecules to ensure the
proper propagation of signals, and biological reactions do not occur at perceptible rates in the
absence of enzymes.
When a signaling transduction reaction is catalyzed by an enzyme, the enzyme (kinase or
phosphatase) is not consumed or produced by this reaction, but it may form a temporary complex
with the substance in the reaction. For such a reaction, we can use Michaelis-Menten kinetics to
describe its reaction rate under the key assumption of quasi steady state approximation, which is
valid when the enzyme concentration is much lower than the substrate concentration and when
the enzyme is not allosteric. Michaelis-Menten kinetics is named for Leonor Michaelis and Maud

15

Menten and has the following formulation:

v=

vm S
km + S

(1.1)

in which v is the rate of reaction and S is the concentration. vm and km are maximum reaction rate
and Michaelis constant, and km equals the substrate concentration that leads to the reaction rate of
vm
2 .

1.3.3

Pathway regulation of cell adhesion

EGF is a compound that can bind to the cell surface receptor, EGFR [55]. The binding will trigger
a series of signaling cascades that have complicated effects and implications in the normal function
of cells. Specifically, the EGFR pathway could regulate cell migration, growth and proliferation
in normal physiological conditions; while in pathological conditions, it is reported to be closely
associated with the metastasis of certain cancers [56]. The EGFR pathway has been shown to be
an extremely complicated interconnected network recruiting and involving numerous intermediate
molecules [57]. One of the essential consequences of EGFR signaling is the regulation of cell
motility through the promotion and inhibition of assembly or disassembly of focal adhesion complexes [58], or the so called cell-extracellular matrix (ECM) adhesion. Overexpression of EGFR
has been associated with cancer metastasis which are featured by compromised cell adhesions.
Thus it is of paramount interest to dissect the signing pathways that regulate the cell adhesion after
EGFR signaling.
The focal adhesion complex comprises of discrete sites where physical contacts between the
cell cytoskeleton and the ECM, extracellular tissue that mainly composed of interlocked fibrous
proteins, are established. The foundation of the physical entanglement starts from the binding of
16

integrin molecules of their extracellular domain, a transmembrane receptor, to the RGD sequences
in the ECM (Fig. 1.7A). This binding will recruit a number of proteins in the intracellular domain
to tether the integrin molecule to the actin filament, one of the three cytoskeleton elements. Thus, a
change in the adhesion site, binding/unbinding of integrin molecules with RGD sequence, assembly/disassembly of focal adhesion complex, adhesion/deadhesion of cell motility, will essentially
tip the delicate balance established between the cell cytoskeleton and the ECM [59]. The focal adhesion was visualized as discrete sites of contact as shown in Fig. 1.7B and the cell structure was
captured by AFM live cell imaging in Fig. 1.7C. This biophysical process has been investigated
from the ligand-receptor binding dynamics perspective. Most of these models are derivations of the
thermodynamic framework pioneered by [60]. The biophysical property of cell adhesion emerges
from the delicate equilibrium of integrin-RGD binding, where the stochastic bond formation and
dissociation rates were expressed in the form of thermodynamics and the load bore by the integrinRGD bond. A number of models have extended the reach of the Bell model to more complicated
scenarios, such as the cell membrane peeling model when the blood cell in rolling motion through
the vein and these extended models have exert influence on the understanding of cell spreading,
migration, proliferation and differentiation [61].
The regulation of cell adhesion by EGFR signaling pathways has been reported to work through
three branches of cascades and their inter-branch crosstalks: the phosphoinositide 3-kinase (PI3K)
pathway, the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK)
pathway, and the phospholipase C (PLC) pathway [62]. Significant progress has been made since
the discovery of EGF and EGFR, resulting in the identification of key molecules in these pathways.
Nevertheless, the complexity of these pathways and their crosstalks prevents the full understanding of impacts of each pathway to the adhesion/deadehsion process, and ultimately cell motility
and migration in a bigger picture [63]. From the intricacy of the pathway dynamics it also raises
17

EGFR

A

Cell membrane
Nucleus
Cell-cell
adhesion

Pathway
molecules
Focal
Adhesion

Actin
filament

Integrin
ECM

B

C

Figure 1.7 A: The diagram of cell adhesion
regulation pathways.
5 nM Experiemental
data B: Image of focal adhesion by
florescence. C: AFM10
image of the living cells with focal adhesion complex.
5

-6

D (x10 )

0
the nonlinearity issues that are inherent to the various feedback and feedforward loops prevalent
-5
in these networks of-10
molecules, issues of paramount interest to the modeling and control commu-

nity [64].

-15

-20
To understand these exquisite
in the process of cell motility regulation,
Phase III
Phase I balances
Phase II of molecules
-25
0 20
40 computational
60 80 100 120
140 160
system biologists often refer
to the
modeling
techniques at the system level to eluTime (Min)
cidate the steps and molecules, and try to assess quantitatively the rise and fall of these molecules

in terms of expression level or concentrations at each step, as well as break down the contributions
from each component during the dynamic interactions [63]. The resulting product is computational
models written in the form of ordinary differential equations (ODE) derived from the reactions of
 

chemical species based on the kinetics of reaction with knowledge of sets of rate constants. This
approach has produced a number of well-defined pathway structures for various aspects of EGFR
signaling [65]. To analyze the modeling results and verify them by experiments, however, requires
18

significant amount of biochemical assays, which are often time consuming themselves.
To circumvent the problem, researchers have proposed a paradigm-shifting methodology to
integrate the biophysical model of cell structure and its dynamic remodeling and reconfiguration
process with the control and regulation from the signaling cascade upon stimulation by various extracellular cues. The only concrete model so far that has been proposed for mechanistic regulation
of cell adhesion by signaling pathways was developed by [66]. The Rho pathway was considered
in the model as the main regulator of cell adhesion and the level of enzymatic activities were converted to applied load similar as the Bell model. The lack of components from other key molecules
poses a challenge for the general acceptance of the model. The linkage between the expression
level of Rho and the applied load to the molecular bond is debatable.
The physical properties of the focal adhesion and its impact on the cell mechanical property as a
whole provide unique peek into the end product of the signaling pathway controlled cell adhesion.
The mechanical characterization of the strength of focal adhesion by an oscillating sensor and
its implication to cell mechanical state by AFM probing has been reported previously [67]. The
relationship between the size of focal adhesion and the readout from the QCM-D sensor has been
elaborated [68]. Thus to build an integrated EGFR signaling/cell adhesion model, we tried to shed
the light from the system control perspective, in which the cellular physical state and its interaction
with the nanomechanical sensor are deemed as the micro and nano mechanical system under the
control of the signaling pathway cascade. Leverage on the unique capability of QCM-D which
can provide real-time continues mechanical measurement, a mechanical model can be built as the
plant, and then various signaling pathway structures and molecular interactions of different rate
constants and parameters can be put on trial against the real-time data as to verify their legitimacy
and accuracy.

19

1.4

Objectives and challenges for AFM based nanorobotics integrated platform

AFM applications are providing increasing insights onto biological processes that are otherwise
not offered by other techniques. With its unique operating environment, AFM high resolution
structural imaging and nanomechanical measurement provide real-time structural and functional
information during the course of biological events. AFM imaging can be combined with other
imaging modalities such as florescence imaging and electron microscopy to boost its capability.
Furthermore, with the introduction of robotic concepts nanobiomanipulation can be performed
on single cells, DNA molecules and membrane proteins. However, there are still technological
challenges that need to be solved for even more precise monitoring of biological phenomenon.
AFM nanorobotics is not merely an extension on AFM, although some AFM could also perform
nanomanipulation, these systems lack the human interface and the precision control in position,
velocity and force, trademarks of traditional robotics.
Thus nanorobotics could be a replacement of AFM, since it integrates the basic elements together: normal imaging, area survey with compressive sensing based fast scan, and manipulation
with human interaction and position/force feedback. With these control capabilities, the nanorobot
can have a workspace of nanoscale, but the controllability of a traditional robot. Thus, the handling
of matters in the nanoscale would be achieved with stability, accuracy and efficiency. The AFM
nanorobotic system provide the human-directed position, velocity and force control with high frequency feedback, and more importantly it can feed the operator with the real-time imaging of manipulation result from the fast-imaging based local scanning. These technical advantages is crucial
to the success of nanobiomanipulations. Firstly, human directed control with a haptic device can
ensure that the survey fashion of measurement of binding force and local mechanical properties
20

can cover larger area and more precisely location wise. Secondly, three-dimensional force control
can provide the manipulation force tailored for different biological matters. Velocity control can
measure and extract the dynamics nature of biological materials than a static measurement model.
Last but not least, a fast imaging system to scan locally of the surrounding area could provide
the operator an updated image of manipulation result in real-time, a necessity in situations where
the temporal resolution of biological processes is in the minutes or even seconds. Biological events
are normally time sensitive. Most of these processes occur much faster than the time it takes to
capture a full frame AFM image. Therefore, a continuing goal is to improve temporal resolution
while maintaining spatial resolution. The use of high-speed AFM to track macromolecules may
be useful for this purpose [69]. Future exploration may include developing AFM imaging modalities correlated with a series of newer techniques to capture more precise biomarkers for various
biological processes.
AFM based nanobiomanipulation requires in situ sensing while the manipulation operation
is being performed. Thus, the AFM process needs to be able to switch between imaging mode
and manipulation mode periodically with high frequency during manipulation. In manipulation
mode the AFM tracks the operator input, while in imaging mode it performs in a zigzag motion
allowing the effect of the operations to be visualized simultaneously. Dual probe systems have also
been proposed in which one cantilever performs the manipulation while a second cantilever senses
the manipulation. This technique may be particularly useful for experiments involving local drug
delivery or site-specific nanosurgery [70]. Multiple probe sensing arrays have also been envisioned
to enhance the capacity of AFM readouts [71]. These exciting new developments in AFM suggest
that we are on the cusp of revolutionary technologies that will no doubt open unforeseen windows
into the mechanism of cellular and molecular process that underlie normal physiological as well
as disease states.
21

1.5

Organization of the Study

The first half of the study used AFM based nanorobotics and other imaging modalities such as
SEM, Fluorescence imaging and force measurement to characterize the role of cell-cell adhesion
(in the form of desmosome) plays in the cell model that we described as the tensegrity. The static
model will only provide the qualitative evaluation of the relationship between cell mechanical stimulation and the response. Thus we sought to build a dynamical model to address both the structural
dynamics and ultimately the molecular dynamics. The second half of the study mainly focused
on the development of AFM nanorobotics and QCM-D nanomechanical sensing in the identification of signaling pathways that mainly modifies the cell-ECM adhesion. The study verifies the
important role the cell-ECM plays in maintaining the cell force balance in the process of cell physiological conditions. In the mean time, the study would also provide the use of this model and
characterization as a screen tool used to replace the messy biochemical characterizations such as
protein blotting. Therefore, this technique would provide the real time dynamics of the molecules
in the regulation of cell-ECM adhesion. This would predict how the protein dynamics follows
certain biochemical stimuli in situ.

22

Chapter 2
Nanobiomechanics in the identification of
pathophysiological conditions

2.1

Introduction

Nanobiomechanics is an emerging field of study with the aim of understanding the mechanics
governing the various biological, especially cellular processes [72]. It started off with the study of
tissues and their mechanical behavior , thus so called biomechanics [73]. As the development of
nanotechnology enables the detection of forces at the nanometer scale, the field of biomechanics
has gradually derived a new frontier in nanobiomechanics, in which the study focuses on cells, or
even molecules [74].
This development has opened up the research in biological science with new possibilities. The
study of nanobiomechanics thus mainly evolves around the physical forces as small as a picoNewton -a trillions of a Newton. These physical forces govern, however undiscernibly, all the
biological process that happens in every instant in the cells. They work in every facet of the cell
life cycle, including the bindings of molecules, which are normally regarded as chemistry or biochemistry, and the growth and migration of cell body, in which the physical force could be the
regulator or the indicator, as well as the formation of cells into tissues, i.e. tissue morphogenesis.
The field of nanobiomechanics, setting in this context, aims to decipher of the underlying mechanism of forces in the regulation of cellular processes, and to mark the pathological state of cells
23

which would aid the diagnosis of diseases. For instance, the malaria infected red blood cells were
distorted and have difficulty passing through blood vesicles [75], and certain type of cancer cells
are 70 percent softer than their original phenotypes for the aid of metastasis [76].

2.2

Cell Mechanics

The study of cell mechanics is a subset of nanobiomechanics and they have taken on a new name,
mechanobiology [77].The understanding of the mechanical behavior of cells have been closely
associated with, or more precisely limited by the signatures of the techniques used for the characterizations.The engineering methodologies that have been driving the study of the cell mechanics
have evolved from micropipette aspirations to optical and magnetic twisting, and AFM probing.
The unique capabilities and shortcomings of each technique define what the research consider as
a proper structure the cell should be and which model could best depict these structures and behaviors. Thus, each technique would be limited by their can and cannot dos, which resulted in the
unfulfillment of understanding the whole picture. The models that followed the techniques have
also evolved from liquid drops to composite viscoelastic materials and finally to a interconnected
network of strings and struts encompassed by a membrane.

2.2.1

Membrane, Channels and Cytoskeleton

The cells, now we know, is a complex network of structures. The outmost layer is a semi-permeable
membrane, composed of drifting islands of lipid bilayer, hosting a range of receptors and ion
channels that can take the input of chemical signals and direct the influx and outflux of ions, as
well as manage the transportation of proteins in and out of the cell. The cytoplasm contains fluid
as media for the various biochemical reactions to take place as well as the cytoskeleton, and the
24

nucleus sits in the center of the cell and is composed of condensed chromosomes. The shape of
the cell is mainly determined by the intracellular cytoskeleton, since the lipid bilayer membrane
is extremely delicate and could not bear any load, thus it could not maintained a defined shape
without the underlying support of the fiber-intensive cytoskeleton structures.
The easiness of bending of membranes is reflected by the bending modulus, which is in the
range of 10−19 J. However, the low bending modulus dose not prevent the membranes to be a
material difficult to extend or compress in the lateral direction. The lateral compressibility of lipid
bilayer is in the range of 100-200 dyn/cm as determined by X-ray diffraction [78].
The ion channels are pore-forming proteins on the membrane. Their main functions include
maintaining a resting membrane potential and control the ions flows across the membrane, as well
as regulating cell volume. They can either controlled by voltage, i.e. voltage-gated, or by ligands,
i.e. ligand-gated. Or they can be classified based on the types of ions they are directing, such
as potassium or sodium channels. The gating of ion channels are closely related with the membrane mechanics as well as the cytoskeleton mechanics. The conformational changes undergone
by transmembrane ion channels in response to stimuli will pose a certain amount of free-energy
in the range of 10 kB T in the case of voltage gated channels [79], which would cause the subsequently deformation of the lipid bilayer. The endocrine secretion cells would also experience
major cytoskeleton changes, i.e. actin congregation at the site of granule secretion [80].
The cytoskeleton structure maintains the mechanical integrity of the cell structure. The microtubule, actin filament and intermediate filament are the three types of filaments that run through
the cytoplasm, connecting the membrane, the nucleus and holding the functioning unit in the cytoplasm together. These protein fibreous structures are with the diameter of a few to tens of nanometers. They form a inter-connected network which defines the mechanical response from external
load and the mechanical response in the normal and disease conditions when probed. They are
25

responsible for the establishment of cell-ECM contact through actin filament and they also coordinate the inter-cellular connection with neighboring cells through intermediate filaments.

2.2.2

Mechanical model of cells

Cells constantly receive mechanical stimuli from the external environment as well as from the internal signaling regulation [81]. Thus, mechanical cues can be crucial regulators of biochemical
processes that govern the cell fate. In the mean time, cell would also display certain mechanical
status at certain stages of their development or in pathological conditions [82]. From this perspective, mechanical states of the cell could also be used as biomarkers for the identification of
cell pathophysiological conditions. To utilize these information, however, requires the detection
mechanisms that have both high spatial and temporal resolution, such as an instrumentation based
on AFM and QCM-D. Besides, the interpretation of these mechanical information requires the
modeling technique that could take advantage of them.
The main cell mechanical models are divided into two different categories [83]: one is the
micro/nano structure approach for a cytoskeleton based cell structure model that mainly used for
adherent cells [84, 85] and only occasionally for suspended cells [86]. For cytoskeleton based cell
models, the cell would establish force balance through the interaction of the cytoskeletal network
and their interaction with the extra cellular matrix and/or the neighboring cells. Therefore the cell
is a discrete network of structures, bounded by the cell membrane that bears most of the force.
The cell membrane will not bear any forces that are significant enough to cause any deformation
or distortion of the cell body.

26

2.2.2.1

Continuum models

The early use of a micropipette aspiration in the study of cell mechanical property provoked a
series of models that consider the membrane as a incompressible layer and the material inside the
membrane as Newtonian fluid. This approach along with other that treat the cell as a composite
material all took the continuum mechanics methodology. The overall cell model was governed by
one constitutive equation.
The continuum mechanics [83] implores the material properties of the cells by considering
them as a composite material. It bears no condition for the cell biochemical processes since the
cells are consider as no different to a non-living composite material, either pure elastic or viscoelastic. Through experimental observations, proper constitutive equations can be built, and then
the related parameters can be identified. It is easy to derive and straightforward but the continuum
mechanics approach will not shine any light onto the detailed biochemical and molecular mechanisms underlying the cell mechanical states. Nevertheless, the parameters and characterizations of
distribution of stress and strain would provide more insights in building the micro/nano structure
models that would link the mechanical behavior and the cell biochemical processes. Continuum
mechanics models are normally divided into solid models [87], which consider the cell a single
material of either elastic or viscoelastic properties and cortical shell-liquid core models which
consider the cell as a viscoelastic material that enclosed inside a elastic membrane [88].
The viscoelastic models include the liquid drop models and the solid models. The liquid drop
model suits the early researches using micropipette as a manipulation and measurement tool. It
was developed by Yeung and Evans to describe the flow of the Leukocyte into the micropipette
with a slightly smaller diameter than the cell [88]. It considers the membrane or cortex as a fluid
layer with a constant tension and the cytoplasm as a Newtonian liquid droplet. The ensuing studies

27

have proposed different constitutive equations for the liquid drop inside the cortex, such as the
compound liquid drop with an additional layer of cortex for the nucleus, the shear thinning liquid
drop model, in which the liquid is modeled as a power law droplet, as well as the Maxwell liquid
drop model.
The solid models treat the whole cell as a homogeneous body without considering the membrane itself. The cell material has been perceived as either incompressible elastic solid or viscoelastic solid. For a pure elastic solid, a constitutive equation defines:

τi j = Gγi j

(2.1)

in which G is the shear modulus and is related to the Young’s modulus E by E = 2(1 + ν)G with
ν being the Poisson’s ratio. For a live cell in liquid environment, the pure elastic solid model is
inherently inadequate in describing the viscous behavior of cells. However, in most situations,
the approximation would not cause any defect in describing the mechanical condition of the cell.
Such practice has been constantly adopted by researchers with AFM nanoindentation measurement
which provides the force-displacement curve.
The force curve plots the relation between the cantilever displacement (z) and its vertical deflection (d). A hard sample surface is regarded as indefinitely stiff compared with the cantilever
and thus will not be indented at all. Therefore, the force curve has a slope of one, meaning the
deflection equals the cantilever displacement. The deflection sensitivity which correlates the photodiode voltage with the deflection of the cantilever can be calibrated by this hard surface force
curve. Moreover, it is a natural reference for soft surface force curves. The soft surface on which
indentation becomes dominant will have a slope less than one. The difference between the two in

28

B

500

glass
cell
Line with slope of 1

Deflection (nm)

400
300
200

(z-z0)-(d-d0)

100
0

0

100
200
300
Z displacement(nm)

Hertz model fitting
Force indentation curve

35
30
Force (nN)

A

400

25
20
15
10
5
0
-5

0

100

200

300

Indentation(nm)

400

500

 

Figure 2.1 A: The conversion of force-displacement curve to force-indentation curve; B: The fitting
of the force indentation curve by Hertz model to generate the Youngs modulus.
the deflection axis is the indentation depth δ .

δ = z−d

(2.2)

The point from where the force curve slope starts is called the contact point, where the initial
contact between the tip and the sample surface was established. Thus, the offset of both the displacement (z0 ) and the deflection (d0 ) before the contact point should be subtracted, as shown in
Fig. 2.1A.
δ = (z − z0 ) − (d − d0 )

(2.3)

The indentation force can be easily calculated by applying Hooke’s law to the cantilever, where
k is the spring constant of the cantilever.

F = k(d − d0 )

(2.4)

Therefore, the force displacement is converted to the force indentation curve which directly
29

depicts the stiffness of the cell surface (Fig. 2.1A). The force indentation curve can be further
fitted to the Hertzian model to generate the quantitative data on the Young’s modulus (E) of the cell
surface (Fig. 2.1B). For a conical tip the relationship between the applied force and the indentation
can be expressed as:
F=

2
E
δ2
πtan(α) 1 − υ 2

(2.5)

where α is the half opening angle of a conical shaped indenter and ε is the Poisson ratio.
A number of studies have used a spherical tip at the end of the cantilever instead of a conical
shape for indentation measurement. The fabrication process involves gluing a spherical particle,
normally polymeric and with 1 µm to 5 µm diameter, to one end of the cantilever. In this case, the
relationship between the indentation and the applied load will be slightly different:

F=

1 3
4 E
R2 δ 2
2
3 1−υ

(2.6)

where R is the radius of the sphere and the other quantities remain.
The main advantage of using a spherical AFM tip for nanoindentation is less damage to a
soft sample compared with the sharp conical tip, normally 10 to 40 nm in diameter at the apex.
The spherical tip can induce deformation of a large area; therefore, the global stiffness can be
measured [89]. On the other hand, a sharp tip indents a much smaller area, and it resolves local
stiffness very well. This is especially useful when the inhomogeneity of the biological samples
matters most. With cells, for example, the underlying cytoskeleton that supports the soft membrane
is not evenly distributed and organized. In this case, the dynamic distribution and organization of
cytoskeleton may be best monitored by measuring the local stiffness using the conical tip. To
identify the mechanical property of a cell population which is not homogenous, statistical analysis
of a large number of cells must be undertaken.
30

Force modulation imaging is an AFM technique that maps differences in surface stiffness [90].
It is based on contact mode AFM imaging. The basic principle for force modulation is to add
a small vertical oscillation of the tip during the contact scanning. That means the contact force
is modulated while the average is still the same as in contact mode. When the tip is modulated
with a vertical displacement around 10 nm with a frequency ranging from 0.5 Hz to 500 Hz [91],
there will be a difference in the resistance of this oscillation among areas with different mechanical
properties. This resistance can be measured as the viscoelasticity of the sample.
For a spherical shaped AFM tip, the applied normal force, a complex with different modulation
frequency can be denoted as (derived from Eq. 2.5) [90]:
1

1
3
3
R2
2 + E ∗δ 2 δ ∗)
(E
δ
F =
0
1 − υ2
2 1 0 1

∗

(2.7)

where υ is the Poisson ratio, R is the radius of the probe and E0 is the elastic modulus at zero
frequency. E1∗ is the frequency-dependent elastic modulus. δ1∗ is the indentation displacement
caused by the oscillation force, and δ0 is the main indentation caused by the applied force by
contact AFM.
Two components constitute the applied force: the contact force and the oscillation force. The
original Hertz model force is the same as Eq. 2.6. And the oscillation force is:
1

∗
Fosc

2R 2 4 ∗ 21 ∗
=
E δ δ
1 − υ2 3 1 0 1

(2.8)

The complex modulus can therefore be expressed as the relation between the applied complex

31

force and the indentation as well as the indentation displacement:

G∗ = G + iG =

∗
fosc
1

2δ1∗ (Rδ0 ) 2

(2.9)

From Eq. 2.9, we can calculate the elastic moduli of the viscoelastic biological samples with
the real part G as the storage modulus and the imaginary part G as loss modulus.
For viscoelastic model, the standard linear solid model has a configuration of a serial connection of spring and dashpot in parallel with a spring of different spring constant. The constitutive
equation defines:
τi j +

k1
µ
τ˙i j = k1 γi j + µ(1 + )γ˙i j
k2
k2

(2.10)

in which τi j is the stress component, and γi j is the strain component. k1 and k2 are the spring
constants and µ is the viscous damping of the dashpot.

2.2.2.2

Structural models

One of the structural models that caused a large impact is the tensegrity model that was introduced
by Ingber, which assigns the roles of three basic cytoskeletal elements [92](microtubule, actin
filament, also called microfilament, and intermediate filament) play in the establishment of force
balance of the cell structure based on the different intrinsic property of these cytoskeletal element.
The microtubule is the stiffest of the three, and would bear the compressional force and the actin
and intermediate filaments would bear the tensional force, thus the structure itself composed of
these three type of elements would be able to balance itself with the help of anchoring point through
cell adhesion [93](Shown in Fig. 2.2A). The change of structure, or so called reconfiguration of
the cell would go through the dynamic biochemical processes: assembly and disassembly of these
microtubule and filaments, crosslink of the filaments. Of these processes, one of the most important
32

 
A
3
3
2.5

6

3

2

11

9

8

6

1

2
7

1

1
3

2
3

1

0

3

8

13

5
16
8

15

2

3

2

8

6
5
4

1
3

10

9

8

18

16

-1

6
20
10

0
-0.5

20
17

3

6
4

19
5
18

1
4

8

7
14
6
13
5
12

9
22 8
10 23
28
24
14
11
27
25
13
12 26

-1
1

1

0

0

0
-1

-1

y-axis

x-axis

7

9
10 32
11

1

1
-1

y-axis

19

9

0

0
-1

1 21
15
2
16

0.5

7

17

-1
1

1

0

x-axis

4
13

1.5

7

1

0

-1
1

-1

y-axis

14
12

-0.5

0
-1

5

15
11

1.5

5
4

1

-0.5

-1
1

1

0.5

6
14
7

0

2

3
2.5

2

0.5

0.5

3
2.5
2

z-axis

1

10
12

z-axis

z-axis

7

-0.5

11
10

1.5

1.5

0

9

4

4

5

12

1

2

6

5

2

4

z-axis

2.5

y-axis

x-axis

-1
x-axis

B

 

Figure 2.2 A: The cell structure is based on cytoskeleton tensegrity. B: Tensegrity structures.
3

2.5

4

3

3

2.5

2.5

5
12
7
11
6
15
21 de20
is the 2regulation
of the
cell
adhesions,1 since
the14cell4 adhesions are1 the
facto
1
3
19 external support that
5
2

10

2

1

2
7

3

z-axis

z-axis

5
cell use to balance
itself
[66].
6
1
4
8

0.5

11

2

13

3

12

2
1.5
7

1

8

6
5
4

1
3

2

10

9

0.5

6
14
7

0

13

5
16
8

0

15
2

16

3

9
10 32
11

4

17

4

18

1.5
z-axis

9

1.5

1

1

8

7
14
6
13
5
12

0.5

8

17

7

16

6
20
10

0

9
22 8
10 23
28
24
14
11
27
25
13
12 26

15
2.2.3
Importance
of cell adhesion in the establishment of cell mechanical
-0.5
-0.5

-1
1

states
-1

1

0

y-axis

18

19

9

-0.5

-1
1

1

0

0
-1

-1

x-axis

y-axis

-1
1

1

0

0

0
-1

-1
x-axis

y-axis

-1
x-axis

Cell adhesion comprises two main categories: one is the cell-ECM adhesions, that’s normally
called the focal adhesions [92] as shown in Fig. 2.2B. It is a anchoring point of the cell to the
ECM. It recruits a cluster of proteins to form the mechanical stabilization mechanism; in the mean
time, it serves as the signaling harbor for the various downstream signaling pathways that regulate
the fundamental cell development and fate: cell migration, inflammation response, apoptosis, and
cancer metastasis [59]. Another type of cell adhesion is the cell-cell adhesion. They are mainly

33

 

cadherin based cell adhesions. The desmosome is one type of the cadherin based cell adhesion,
which is divalent-cation-dependent, proteolytic enzyme such as trypsin as well as removal of extracellular Ca2+ and Mg2+ with a divalent cation chelator such as EDTA would cause dissociation
of these adhesions [92].

2.2.4

Different model could apply at different scenarios

The fact that the cell structure and the dynamics are so complicated that no single model is detailed
or suitable enough for all applications and cell related process. Thus you would see through out
the whole study that different models have been utilized in order to parse the parameters of certain
properties, or to extract certain properties. In these cases, different assumptions were made to
facilitate the building of these models and to use these models for predictions. And also based
on this fact, we can deduce that no single measurement of imaging modality would be enough to
address these complicated and complex matters and related processes, thus you would also see the
use of a combination of imaging modalities and measurement techniques in order to build a more
complete profile of the cell dynamic responses during the extracellular stimulation process.

2.3

Manipulation and measurements at the molecular level

The identification and measurement of biomechanical information on live cells requires the use
of micro/nano manipulation and measurement instruments. Among them, the most popular is the
AFM as we introduced in the previous sections. The earliest measurement and manipulation often
employs the micropipette aspiration technique [94], in which a suction pressure is applied to a static
cell body until a hemispherical projection is formed inside the pipette. The relationship between
the applied suction pressure and the property of the cell body, in this case modeled as a liquid drop
34

enclosed by a cortical shell can be derived. Similar techniques of this sort is the optical tweezers
which also consider the cell as a shell-liquid model. Laser traps the microbeads which attached at
two distal end of the cell would cause significant deformation, normally stretching of the cell [95].
Magnetic twisting cytometry [96] was used to assess the rheological properties of smooth muscle
cells; a magnetic field of varying frequency was applied to the microbeads, initially magnetized in
a predefined direction, that were trapped on the membrane of the cell. An angle of rotation would
then be measured by a magnetic detector and used to interpret the complex modulus of the cell.

2.3.1

DNA manipulation

DNA molecules are mechanically flexible and physicochemically stable. The capability of AFM
to maneuver objects at the molecular level has allowed the manipulation of DNA. Based on the
behavior model of DNA, the operation can be displayed in real time in an augmented reality environment as studied by Li and coworkers [8].

2.3.2

Single cell manipulation

There are a number of different mechanisms for single cell manipulation, including micropipette
aspiration, optical trapping, magnetic twisting, and nanoindentation, etc. Each technique works
differently to fit different mechanical models for single cell structures. Regarding AFM based single cell manipulation, the main operation is AFM based nanoindentation. There are two different
types of AFM based nanomanipulation: unpenetrated nanoindentation and penetrated nanoindentation.
For unpenetrated nanoindentation, the sharp AFM tip only deforms the membrane surface of
the cell. Because the cell membrane is supported by and propagated through the cytoskeleton,

35

deformation at the surface can result in cytoskeletal re-configuration. Altered cytoskeleton organization results in an altered response to the applied load, and is eventually manifested in the form
of changes in the overall stiffness of the cell body. There is also work to suggest that the tension in the cell membrane is mainly responsible for the stiffness behavior of the cell body, like a
balloon filled with liquid [83]. However, recent findings indicate that the cell may behave more
like a globally interconnected structure covered by a soft membrane, like a tent that is nailed to
the ground [85]. This is also consistent with AFM imaging data where high applied contact force
elevates the intracellular actin network.
For penetrated nanoindentation, the sharp AFM tip is used as a nanoneedle to penetrate the
membrane and reach the actin filament network underneath. The applied force just before the penetration can be correlated with the indentation depth just before penetration. One important potential application of penetrated nanoindentation is for controlled drug delivery [97]. The nanoneedle
can be fabricated with a hollow inner structure whereby certain drugs can be pumped into the cell
upon penetration in a controlled fashion.

2.3.3

Filament manipulation and characterization

The cytoskeletal structure supports the cell, performs functions such as vesicle transportation and
protein anchoring, and regulates the cell shape and movement. It composed of three different filament structures: actin filaments, intermediate filaments and microtubules. Each group of proteins
is critical for cellular functioning. Dynein is a basic protein that is assembled together to make
microtubules. It functions as a motor protein that converts chemical energy conserved in ATP to
mechanical energy required for movement [92]. This process may provide a powerhouse for the
development of future self-assembly fabrication processes to construct complex devices from nano
building blocks.
36

Recently, mechanical properties of intermediate filaments have been studied [53]. For one
study, intermediate filaments absorbed on the mica surface were stretched by AFM tip. The manipulation operation itself in this case was routine, but the findings from the experiment were
striking. Intermediate filaments could be stretched more than two fold of original length. This
in vitro experiment also confirmed the role of intermediate filament in the maintenance of mechanical integrity of the cell body and in providing strength for the entire tissue. These roles are
performed particularly well when cells are undergoing a large extent of deformation, and can be
systematically studied by this AFM nanomanipulation technique.

2.3.4

Nanodissection and nanosurgery

A very special group of operations using AFM involves nanodissection, or nanosurgery. The main
purpose of this operation is to isolate, or partially isolate a group of cellular and/or molecular
structures from other structures. There is literature on the dissection of DNA molecules, proteins,
cell membranes, intermediate filaments, among other structures. Most of these experiments were
performed for the sole purpose of demonstrating the capability of AFM based nanobiomanipulation
systems in handling and changing biological matters. Our group has extended the application of
nanosurgery to demonstrate that it can be a novel tool to re-create and disrupt specific functional
biological activities with unparalleled precision.

37

2.4

Cellular nanosurgery mimics the effect of antibody treatment in cell-cell adhesion

The identification and measurement of biomechanical information on live cells require the use of
micro/nano manipulation and measurement instruments. Among them, the most popular is the
AFM as we introduced in the previous sections. The earliest measurement and manipulation often
employs the micropipette aspiration technique [94], in which a suction pressure is applied to a static
cell body until a hemispherical projection is formed inside the pipette. The relationship between
the applied suction pressure and the property of the cell body, in this case modeled as a liquid drop
enclosed by a cortical shell can be derived. Similar technique of this sort is the optical tweezers
which also consider the cell as a shell-liquid model. Laser traps the microbeads which attached at
two distal end of the cell would cause significant deformation, normally stretching of the cell [95].
Magnetic twisting cytometry [96] was used to assess the rheological properties of smooth muscle
cells; a magnetic field of varying frequency was applied to the microbeads, initially magnetized in
a predefined direction, that were trapped on the membrane of the cell. An angle of rotation would
then be measured by a magnetic detector and used to interpret the complex modulus of the cell.

38

Chapter 3
In situ sensing of ion channel activities after
glucose stimulation

3.1

Introduction

The molecular mechanism of insulin secretion from pancreatic β -cells in times of elevated glucose
levels has been well documented over the past decade. Insulin is produced by β -cells to maintain
glucose homeostasis. When the extracellular glucose level rises, glucose will be transported to
the cytoplasm through the glucose transporter GLUT2 [98]. The intracellular glucose level rise
will increase the pace of glucose metabolism resulting in increased ATP/ADP ratio inside the β cell [99] followed by the closure of the ATP-sensitive K+ channel (KATP) [100]. The closure
of KATP channels causes the increase in plasma membrane (PM) potential [101] and cellular
depolarization, a process leading to activation of voltage-dependent calcium channels (VDCC).
Activation of VDCC allows Ca2+ to flow into the cytoplasm and gives rise to the intracellular
Ca2+ levels ([Ca2+ ]i ) [102]. In response to the elevated [Ca2+ ]i , insulin-containing vesicles will
fuse with plasma membrane to release the insulin [103]. The whole process is referred to as the
stimulation-secretion coupling, a biphasic process in nature. The initial 5-10 minutes with the
rapid insulin exocytosis is referred to as the first phase [104], while the subsequent second phase is
characterized by a sustained insulin exocytosis, an ATP-dependent process which releases insulincontaining vesicles in reserve pools further away from the releasing site [105].
39

Capsaicin, the pungent component of chili peppers, is a specific agonist of the transient receptor
potential vanilloid type 1 (TRPV1) channel or the vanilloid receptor [106]. TRPV1 receptor protein
is expressed in nerve fibers innervating the pancreatic islets of rats and mice [107]. It has also been
reported that TRPV1 expression is detected in rat insulinoma cell lines, RIN and INS-1 [108,
109]. Capsaicin may modulate insulin secretion through activation of capsaicin-sensitive afferent
nerves [110]. Additionally, capsaicin may stimulate islet β -cells by binding to the intracellular
domain of TRPV1 and thereby activating TRPV1 channels [111, 112] as illustrated in Fig. 3.1A.
Thus the regulatory role of capsaicin on insulin secretion may be direct or indirect, a complex
effect which remains to be defined. In comparison with glucose stimulation, capsaicin stimulated
insulin secretion from islet β -cells is reported to be a sustained process [109] without the biphasic
characteristics.
With the unique capability of AFM based nanorobotic manipulator (Fig. 3.1B), the current
study went a step further and combines optical microscopy and AFM to unravel the ion channel
dynamics in insulin release upon glucose and capsaicin stimulation. The ion channel activity was
monitored by capturing fluorescence images of intracellular Ca2+ levels upon stimulation. Simultaneously, cellular stiffness was measured in real time with AFM nanoindentation to reveal the
reorganization of the cell membrane and cytoskeleton resulting from ion channel reconfiguration.

3.2

Materials and methods

The insulinoma β -cell line RINm5f cells (ATCC, Manassas, VA) were cultured in RPMI-1640
medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal calf serum and 1% penicillin
and streptomycin (Invitrogen, Carlsbad, CA) at 37◦ C in a humidified atmosphere containing 5%
CO2 . The RINm5f cells used were between passages 20 and 30.

40

ure 1
A

B

Figure 3.1 A: The stimulus-secretion coupling model of β -cells. Increased plasma glucose levels
will increase the intracellular ATP/ADP ratio after glucose being uptaked by cells, which leads
to the closure of ATP-dependent potassium channels, membrane depolarization, and the increase
in intracellular Ca2+ concentrations through the activation of voltage dependent calcium channels (VDCC) and ultimately insulin secretion. Capsaicin binds to the intracellular domain of the
TRPV1 receptor and causes the influx of Ca2+ , which triggers the release of Ca2+ into the cytoplasm from the endoplasmic reticulum (ER) and increases [Ca2+ ]i concentrations. B: The AFM
based nanorobotic manipulator uses a joystick as position input to facilitate the movement of the
AFM tip in the planar area during the force measurement process.

41

3.2.1

TRPV1 immunofluorescence

For immunoreaction, RINm5f cells were seeded on glass-bottom petri dishes (MatTek Co., Ashland, MA) coated with poly-L-ornithine and incubated under standard conditions until they were
90% confluent. Cultured cells were then processed for immunofluorescence. After three washes
with phosphate buffered saline (PBS), cells were fixed by incubation in 4% (v/v) paraformaldehyde in PBS for 15 min at room temperature, rinsed with PBS and incubated overnight at room
temperature with anti-TRPV1 antibody, goat polyclonal IgG diluted 1:250 with 1% (w/v) bovine
serum albumin (BSA) in PBS (Santa Cruz Biotechnology, Santa Cruz, CA). Control samples were
incubated with 1% BSA in PBS. After three washes in PBS, fluorescence was revealed by incubation for 2 h in AlexaFluor488 labeled secondary antibody, donkey anti-goat IgG diluted 1:100
(Invitrogen, Carlsbad, CA) in PBS. Imaging of immunofluorescence was performed with an epifluorescence microscope equipped with the Nikon B-2A filter (excitation wavelength 450-490 nm).
Images were acquired using a digital camera (Photometrics, Tucson, AZ) connected to the microscope and image was captured using software QCapture Pro 6.0.

3.2.2

Loading of Fur-2 AM to RINm5f cells

Fura-2 AM (Invitrogen, Carlsbad, CA) was diluted in the low glucose base medium (2 mM glucose
+ 1mM CaCl2 +140 mM NaCl + 2.8 mM KCl + 2mM MgCl2 + 10 mM Hepes-NaOH) to a final
concentration of 0.5 µ M. The RINm5f cells were rinsed with PBS three times. One ml of diluted
Fura-2 AM ester solution was added to the dish and incubated for 30 minutes at 37◦ C. The cells
were then washed with PBS three times and then incubated in the low glucose base medium for
additional 30 minutes to allow complete de-esterification of intracellular AM esters.

42

3.2.3

Glucose and capsaicin stimulation and dosage-dependent study

All the experiments were performed at room temperature. Before each glucose stimulation experiment, the cells in the glass-bottom petri dishes loaded with Fura-2 AM were incubated with
5 ml of low glucose base medium (2 mM glucose, 1mM CaCl2 +140 mM NaCl + 2.8 mM KCl
+ 2mM MgCl2 + 10 mM Hepes-NaOH) for additional 30 minutes until equilibrium. For glucose
stimulation, a volume of 40 µl of high glucose medium (2M) was added into the dish to make
the final glucose concentrations at 16 mM; a 40 µl of low glucose base medium was added to the
control plates. Similar with glucose stimulation, the cells in the glass-bottom petri dishes loaded
with Ca2+ binding agents were incubated with 5 ml of low glucose base medium for additional
30 minutes until equilibrium before capsaicin was applied. One ml of low glucose base medium
with capsaicin (Sigma-Aldrich, St. Louis, MO) concentrations of 100 nM was added to the dish to
obtain the final concentrations of 16.7 nM; One ml of low glucose base medium was added to the
control plates.
Dosage-dependent experiments for glucose and capsaicin stimulation were carried out in the
similar method described above. Dosages of 40 mM, 20 mM and 16 mM for glucose and 1.67
nM, 16.7 nM, 167 nM and 1.67 µM for capsaicin were examined with AFM analysis. One ml of
low glucose based medium was used as control. For glucose stimulation, the stiffness calculated
from the first 20 force curves before switching to high glucose were averaged as the baseline.
The moduli from the first 20 minutes of measurements after addition of high glucose medium
were averaged as the positive results. For capsaicin, 100 cells were randomly selected before the
stimulation for base line measurement and another 100 cells 20 minutes after the stimulation on the
same cell population, stiffness was measured and compared statistically. To test the specificity of
capsaicin stimulation, the TRPV1 was blocked with 1.0 µM capsazepine (Capz) 30 minutes prior

43

to the capsaicin stimulation. Each data represents a minimum of three independent experiments.

3.2.4

Simultaneous fluorescence imaging of [Ca2+ ]i and AFM measurements
on RINm5f cells

Before adding glucose or capsaicin, 10 force displacement curves were taken as the baseline on the
center of the same cell. After stimulation, force displacement curves were taken at a frequency of 1
Hz to dynamically monitor the stiffness change in real time for an hour. To obtain higher sensitivity
and to prevent potential damage of the sharp AFM tip to the delicate cell membrane, force applied
to indent the cell was kept minimal. The maximum force applied is controlled at below 10 nN
with the resulted deformation less than 100 nm, and all the force curves are taken at a loading rate
of 1 Hz, corresponding to a tip velocity of 5.98 µm/s. At the same time, the cells loaded with
Fura-2 AM was observed under the inverted microscope with excitation wavelength 340-380 nm
(UV-2E/C Nikon filter). Images were captured in a similar fashion as in the immunofluorescence
for TRPV1 except at a much shorter exposure time (0.3 sec). The time-lapse images taken at 1
frame/s were then processed with ImageJ software (NIH) to obtain the mean gray scale values.
The images were displayed in 8 bit gray scale (0-255) and each cell in each image was outlined
and from which the gray scale value was obtained, averaged and normalized. Each data represents
a minimum of three independent experiments.

3.2.5

Actin filament staining and imaging

Cells grown on glass coverslips were allowed to pre-incubate in 1 mL of the PBS at 37 C for
1 h. After pre-incubation, the buffer was removed and replaced with either 16 mM of glucose
or 16.7 nM of capsaicin; the low glucose base medium was used for the control samples. After

44

various reaction times (0 min, 1 min, 5 min and 10 min) at 37◦ C, cells on glass coverslips were
fixed/permeabilized in a solution containing 0.1% Triton X-100 and 3% paraformaldehyde in PBS
buffer for 20 min. They were then washed three times by PBS buffer with 1% BSA. The resulting
cells were stained with CF488A phalloidin conjugate (Biotium, Inc., Hayward, CA) in PBS with
1% BSA at room temperature for 20 min. Then the excess phalliodin solution was removed by
rinsing with 1% BSA in PBS. Coverslips with stained cells were imaged with an inverted fluorescence microscope (Photometrics, Tucson, AZ). Images were processed with the use of QCapture
Pro 6.0 software.

3.3
3.3.1

Results
Immunofluorescence of TRPV1 on RINm5F cells

Immunofluorescence staining of RINm5f cells with antibody selective for TRPV1 receptors revealed the positive fluorescence signals, primarily distributed peripherally of the cells (Fig. 3.2A),
consistent with the notion that TRPV1 receptors are localized in the cell membrane of RINm5f
cells [109]. The control samples incubated with 1% BSA in PBS show negative immunoreactivity (Fig. 3.2B). The result confirms that RINm5f cells express TRPV1 receptors on their cell
membrane.

3.3.2

AFM stiffness observation of dosage responses to glucose and capsaicin stimulation

The normal Young’s modulus value for insulinoma RINm5f cells in the low glucose medium was
around 7.0 kPa, a similar value as previous reports by AFM based studies [113]. High glucose

45

Figure 2

A

B

Young's modulus (Pa)

20.0k

15.0k

Before stimulation
After stimulation

┼

Δ
┼

*

*

10.0k

5.0k

╪
Δ
┼

*

D
15.0k

Young's modulus (Pa)

C

10.0k

ol
M
M
M
Contr ose 16 m ose 20 m ose 40 m
Gluc
Gluc
Gluc

*

*

╪

╪

╪

5.0k

0.0

0.0

Before stimulation
After stimulation
Δ
┼
┼

ol
M
M
M
M
M
Contr ap 1.67 n ap 16.7 n ap 16.7 n ap 167 nap 1.67 u
C
C
C M+C
C
u
1
Capz

Figure 3.2 Immunofluorescence imaging of TRPV1 receptors in RINm5f insulinoma cells. A: Cell
plasma membrane localization of TRPV1 receptors in RINm5f cells, positive staining indicated by
arrows. B: Control RINm5f cells in which primary antibody was replaced with 1% BSA in PBS
show negative staining. Scale bar: 20 µ m. C: The stiffness changes before and after administration
of glucose at three concentrations, 16 mM, 20 mM, and 40 mM, plus the control. * p<0.05
vs. before stimulation; p<0.05 vs. control; p<0.05 vs. glucose 16 mM; p<0.05 vs. glucose
16 mM and 20 mM (n=20). D: The stiffness changes before and after administration of Cap
(capsaicin) at four concentrations (1.67 nM, 16.7 nM, 167 nM and 1.67 µM), with *p<0.05 vs.
before stimulation; p<0.05 vs. control; p<0.05 vs. Cap 1.67 nM; p<0.05 vs. Cap 1.67 nM and
16.7 nM. Each datum is expressed as mean SEM (n=100).

46

increased the stiffness, and the increase displayed a dose-dependent pattern (Fig. 3.2C). The modulus increase for 40 mM glucose stimulation was more than one fold, reaching 14.8 kPa. Stiffness
increased to 13.8 kPa and 12.0 kPa for 20 mM and 16 mM, respectively. The control experiment
with an equal volume of low glucose medium showed no significant changes in stiffness. Thus
cellular stiffness in response to various doses of glucose stimulation is dose-dependent. The dose
studies for capsaicin stimulation showed a different pattern compared with glucose. Statistical
analysis showed that stiffness increased for lower concentrations of capsaicin at 1.67 nM and 16.7
nM but not at higher concentrations of 167 nM and 1.67 µM (Fig. 3.2D). These data support the
finding [114] that the high dose of capsaicin had insignificant effects on insulin secretion, possibly
due to desensitization of TRPV1 induced by high concentrations of capsaicin [113]. The control samples and the samples treated with Capz, an antagonist of TRPV1, showed no significant
change of mechanical property (Fig. 3.2D). Taken together, these results indicate that increases in
cell stiffness are linked to insulin secretion, and the nanomechanical properties may well serve as
biomarkers for insulin secretion.

3.3.3

[Ca2+ ]i and stiffness observation of glucose stimulation

To look at the activities of ion channel and cellular mechanical property under stimulation, real
time observation was performed on a single RINm5f cell. The normal Young’s modulus value for
the individual RINm5f cell was around 7.0 kPa for the 10 minutes baseline measurement before
the addition of high glucose medium to make the final glucose concentration 16 mM in the dish.
Immediately following the glucose stimulation at 16 mM, altered ion channel activity was indicated
by the sharp increase by about 20% in light intensity of [Ca2+ ]i imaging (shown in Fig. 3.3A).
The light intensity remained at a relatively higher level for 3-4 minutes before it dropped to the
baseline. It is worth mentioning that the initial slight drop of light intensity was probably due
47

Figure 3

A
Stiffness

Light intensity

18.0k

0.68

16.0k

0.66

14.0k

0.64

12.0k

0.62

10.0k

0.60

8.0k

B

0.70

Normalized greyscale

Young's modulus (Pa)

20.0k

0.58

6.0k

Time (Min) 0.56

Glucose

4.0k

0 5 10 15 20 25 30 35 40 45 50 55 60 65
Min 10

C

Glucose

0

5

10

Min 12

D

15
Min 20

20
Time (Min)

E

Figure 3.3 Dynamic characterization of cellular stiffness simultaneously with [Ca2+ ]i fluorescence
imaging after glucose (16 mM) stimulation. A: The ion channel activities overlaps with the overall
cellular stiffness change in the first phase when the onset of stimulation induced rapid structural
rearrangement. A second sustained phase follows to maintain the structural configuration and
stiffness at a relatively higher level than the baseline. B: A close-look at the lagging of stiffness
response compared with that of intracellular Ca2+ concentrations, a 2 minute-delay was observed.
C, D, E: The fluorescence imaging of [Ca2+ ]i before stimulation (C), 12 minutes after stimulation
when the light intensity reaches the peak (D), and 20 minutes after stimulation (E) when the light
intensity falls to the baseline.

48

to the photo-bleaching effect, and for the same reason, the imaging capture stopped at minute 20
when the light intensity was bleached to the level below the baseline. The fluorescence images
of RINm5f cells at minute 10 (before stimulation), 12 (plateau was reached) and 20 (returns to
baseline) were shown in Figs. 3.3C, D and E. At the same time, the cellular stiffness increased
sharply as well, and it took about 5 minutes for the stiffness to reach a plateau at around 18.0 kPa.
This was followed by a rapid decline at a similar rate as the climbing. About 10 minutes post
stimulation, the stiffness dropped to a value around 10.5 kPa, about 1.5 times the baseline value
and leveled off through the rest of the observation. Note that in Fig. 3B, a close up look of the
stiffness and light intensity measurement 10 minutes prior and 10 minute post stimulation showed
a 2-minute lagging of mechanical property changes when compared with the ion channel activities.
The nanomechanical analysis results were echoed by the changes of cytoskeleton structures. The
staining of actin showed a disorganized filament structure one minute after glucose stimulation
while the control displayed no visible changes shown in Fig. 3.6.

3.3.4

[Ca2+ ]i and stiffness observation of capsaicin stimulation

Upon capsaicin stimulation at 16.7 nM, the light intensity increased immediately (Fig. 3.4A). Lack
of altered light intensity was observed for samples treated with Capz 30 minutes prior to capsaicin
stimulation (Fig. 3.4B). A pulse-like response lasted approximately 3-5 minutes before it dropped
to the baseline. The images corresponding to minute 10, 12 and 20 were shown in Fig. 3.4C, D
and E. The Ca2+ influx response after capsaicin treatment experienced a similar pattern as that
of glucose. The baseline value of stiffness for the cell population under capsaicin stimulation is
similar to the one in glucose experiment. However, no rapid increase of stiffness was observed following the administration of capsaicin. Instead, the stiffness increased gradually, reached the peak
at approximately 11.0 kPa about 20 minutes after the stimulation. This sustained stiffness response
49

Figure 4

20.0k

Stiffness

Light intensity

0.68

16.0k

0.66

14.0k

Peak time P

12.0k

0.64

10.0k

0.62

8.0k

0.60

6.0k

0.58

4.0k

Capsaicin

2.0k

Time (Min)

Normalized greyscale

Young's modulus (Pa)

18.0k

B

0.70

0.56

-5 0 5 10 15 20 25 30 35 40 45 50 55 60 65
Min 10

C

Before stimulation
At peak time P

0.7
Normalized greyscale

A

*

0.6

┼

0.5
0.4
0.0

Cap 16.7 nM

Capz 1um + Cap 16.7 nM

Min 20

Min 12

E

D

Figure 3.4 Dynamic characterization of cellular stiffness simultaneously with [Ca2+ ]i fluorescence
imaging after capsaicin (16.7 nM) stimulation. A: The ion channel shows similar pulse-shape
response pattern as that of glucose albeit with smaller peak value. B: The cells were also pretreated
with Capz 30 minutes before capsaicin stimulation to block TRPV1 receptors. No increase in light
intensity was observed for TRPV1-blockade samples shown in the normalized gray scale before
and after capsaicin stimulation. *p<0.05 vs. before stimulation; p <0.05 vs. Cap 16.7 nM. C, D,
E: The fluorescence imaging of [Ca2+ ]i before (C), 12 minutes after stimulation (D) when the light
intensity reaches the plateau, and 20 minutes after stimulation (E) when the light intensity falls to
the base line. The normalized gray scale value was obtained the same way as the Fig. 3.3.

50

stopped after the plateau was reached and remained at that value for the rest of the observation.
Similar change of actin filament structures was observed but with a distinctive temporal pattern;
the structural disorganization didn’t start to surface until after 10 minutes of capsaicin stimulation
as shown in Fig. 3.6.

3.4

Discussion

It is well documented that increased glucose concentration in the extracellular environment would
lead to increases in plasma membrane potential, elevated [Ca2+ ]i and ultimately insulin secretion, a
process commonly referred to as stimulus-secretion coupling [115–117]. Studies have also shown
that capsaicin is an effective compound stimulating insulin secretion from insulinoma β -cells including RINm5f and INS-1 cells [109, 114]. Although glucose and capsaicin have been shown to
induce distinctive patterns of insulin release, i.e., a biphasic secretion response upon glucose but
not capsaicin stimulation [118, 119], the biophysical characteristics of cells in response to these
stimuli have not been well defined. The present study aimed to study the temporal relation of
ion channel activity, cell membrane biophysical dynamics, and cytoskeleton reorganization of β cells in real time with AFM nanoindentation to reveal potential mechanisms underlying distinct
insulin-secretion processes in response to glucose and capsaicin.
The results from the present study using AFM based nanomechanical analysis show for the first
time that the β -cells biomechanically respond to glucose stimulation in a biphasic manner, consistent with the pattern of insulin release. The cellular stiffness increases rapidly and then decreases
sharply within 10 minutes of glucose stimulation, followed by a sustained second phase to maintain the cellular stiffness at a relative higher level than the baseline. The biphasic pattern of cell
stiffness in response to glucose stimulation is distinct from that induced by capsaicin (fitted curve

51

Figure 5

A

B

Young's modulus (Pa)

20.0k

Glucose

Capsaicin

15.0k

10.0k

5.0k
0

10

20
30
40
Time (Min)

50

60 0

10

20
30
40
Time (Min)

50

60

Figure 3.5 The overall trend of the stiffness responses. A: Glucose stimulation resulted in a twophase response: a rapid increase phase in stiffness followed by a sustained elevation phase; B:
Capsaicin stimulation induced a slow and sustained increase in stiffness.
showing the overall trend in Fig. 3.5). Capsaicin induces a sustained and slow increase of cellular stiffness without a rapid increase or decrease pattern, a change consistent with that of insulin
secretion in response to capsaicin. Although it is known that biophysical changes may result from
reorganization of cytoskeleton structure to facilitate the transport and fusion of insulin-containing
vesicles with the plasma membrane [80, 118, 119], the nature of fine changes in cellular structure
and mechanical property presents a daunting challenge for the detection and measurement techniques. The present study shows that biophysical markers captured by AFM based nanomechanical
analysis may be used as indicators for altered ion channel activities and thereby insulin secretion.
The initial phase of sharp increases in stiffness in response to glucose stimulation overlaps temporally with altered ion channel activities that result in elevated intracellular Ca2+ concentrations,
with a few minutes of delay in stiffness to reach the peak. These results indicate that structural
changes in the cell membrane or cytoskeleton follow the rapid responses of ion channels activated
by stimuli. In contrast, the increases in stiffness in response to capsaicin stimulation do not coincide temporally with increased intracellular Ca2+ concentrations, with a long delay in stiffness to
52

Figure 6
Control

0 Min

1 Min

5 Min

10 Min

0 Min

1 Min

5 Min

10 Min

0 Min

1 Min

5 Min

10 Min

Glucose

Capsaicin

Figure 3.6 The actin filament reorganization after treatment of low glucose based medium as control, 16 mM glucose and 16.7 nM capsaicin for 0 min, 1 min, 5 min and 10 min. F-actin was
labeled with CF488A-conjugated phalloidin. The treatment of glucose and capsaicin causes timelapse change of cytoskeleton especially actin filament structures with distinct patterns. A delayed
disorganization of actin around 10 minutes after stimulation was observed in capsaicin treatment,
whereas glucose triggered the disorganization 1 minute after treatment, similar with the biophysical data obtained by AFM nanomechanical analysis. No distinctive change was observed for the
control sample.
reach the peak. These results indicate that structural changes in the cell membrane or cytoskeleton
in response to TRPV1 activation induced by capsaicin or to glucose are dissimilar and agonistspecific albeit the changes in [Ca2+ ]i concentrations showed similar pulse-response patterns for
capsaicin and glucose. The results of immunocytochemistry study of actin filament concur with
the biophysical measurements obtained with AFM examination. As seem in Fig. 3.6, the treatment
of glucose and capsaicin causes time-lapse changes of cytoskeleton especially actin filament structures with distinct patterns. A delayed disorganization of actin around 10 minutes after stimulation
was observed in capsaicin treatment, whereas glucose triggered the disorganization 1 minute after
treatment, similarly with the biophysical data obtained by AFM nanomechanical analysis.
53

The biphasic model of insulin secretion from pancreatic β -cells is well established and can be
elicited by several compounds including glucose, arginine, and protein kinase C activators [120,
121]; under certain conditions, the first phase of insulin secretion may appear alone, without the
second phase [120]. Until recently, there was no compound or treatment condition that elicited the
second phase only. Our data shows that the initial calcium influx induced by glucose or capsaicin
is similar with a sharp increase in [Ca2+ ]i , but it diverges hereafter. While a signaling cascade
situated between altered [Ca2+ ]i and insulin secretion induced by capsaicin has yet to be fully
elucidated, our data from cell stiffness changes and actin reorganization support the notion that
capsaicin causes only a delayed-phase insulin secretion. It has been indicated that the delayed
phase of insulin secretion involves the Rho family of kinases that interacts with and causes the rearrangement of actin cytoskeleton in β -cells [118]. Interestingly, TRPV1 activation with capsaicin
may trigger the Rho kinase pathway [121]albeit this has yet to be shown in pancreatic β -cells.
Taken together, our data show that in contrast to well described glucose-stimulated dual phase secretion of insulin, capsaicin causes a delayed phase of insulin secretion through TRPV1 activation
resulting in actin reorganization and cell stiffness changes.
The islet β -cell insulin secretion process is highly regulated by electrical and ion channel activities [122]. The electrophysiological properties of ion channels play a crucial role in the secretion
process, the defect of which may lead to metabolic disorders [123]. These ion channel properties have been investigated to elucidate the underlying secretion mechanism through membrane
capacitance measurements. Patch-clamp technique has been employed to measure the capacitance
increase resulting from the membrane expansion due to insulin exocytosis [124]. While effective, patch-clamp studies are normally time consuming and demand highly skilled workers. More
importantly, electrically-regulated insulin secretion is coupled with biophysical changes of the
membrane [119], and the electromechanical coupling can be revealed fully by integrated studies
54

of electrical activity and biophysical monitoring [125].

3.5

Chapter summary

In this chapter, we have successfully established a biomechanical marker during the stimulated
ion channel responses. All the evidence indicates that islet β -cells exhibit a different physical
state during the insulin secretion process, on which the biophysically biphasic glucose stimulation
and single-phased capsaicin stimulation depend. These biophysical features may be addressed by
identifying and characterizing signatures of ion channel activities. Many have reported that β cells showed similar patterns of Ca2+ levels and insulin release in vitro and in vivo in responses
to glucose stimulation [126, 127]. Taking into consideration the fact that islet cells tend to cluster
in vivo [126], changes in biophysical property are expected with three dimensional environments
as compared with a cell monolayer. A continued study with cell encapsulation would be better
positioned to address these speculations. The current study aimed to develop and apply novel techniques such as this AFM nanorobotic system to examine physical properties of pancreatic insulin
secreting cells by characterizing nanomechanical properties of a single cell or a cell population so
that unique biomarker(s) in response to general or specific channel activators may be identified.
Among many biophysical characterization techniques used for living cells, AFM stands out for a
better fit for the task. In recent years, AFM has become an increasingly versatile tool in studies of
biological processes for in situ nanoscale measurements [14]. With the use of the AFM nanoindentation, nano- even pico- Newton forces may be applied for the biophysics dynamic measurement,
rendering the capture of tiny changes of cellular structure and property possible. The integrated
approach with AFM and optical microscopy, aimed at unraveling the dynamic responses to ion
channel stimulation, may provide a new perspective in the study of pancreatic β -cell stimulation.

55

Chapter 4
In situ biomarker detection during signaling
events

4.1

Introduction

Desmosomal junctions are specialized structures critical to cellular adhesion within epithelial
tissues. Disassembly of these junctions (acantholysis) can occur following autoimmune attack.
Pemphigus vulgaris (PV) is a prototypical organ-specific human autoimmune disease characterized clinically by flaccid blister formation in the skin and mucous membranes that can be lifethreatening. Histologically, there is an intraepidermal split resulting from acantholysis of suprabasilar keratinocytes that occurs in the presence of autoantibodies to specific desmosomal proteins, primarily desmoglein 3 (Dsg3) [128]. Anti-Dsg3 autoantibodies have been shown to induce acantholysis in cultured keratinocytes [129] and blister formation in vivo in neonatal mice [130]. Splenocytes from Dsg3-/- mice adoptively transferred into Rag2-/- recipient mice (that express Dsg3)
produce anti-Dsg3 antibodies that lead to the development of suprabasilar acantholysis in the oral
mucosa and esophagus [130].
Although lesion development in patients with PV is generally associated with high titers of
anti-Dsg3 autoantibodies, the precise molecular mechanisms by which autoantibodies direct the
loss of cell-cell adhesion is not known. In particular, it is unclear if acantholysis is the direct result
of structural changes at the keratinocyte cell surface that occur subsequent to autoantibody binding
56

and/or is dependent upon functional changes within the cell. Moreover, there is now clear evidence
that PV patients harbor both anti-Dsg3 antibodies that lead to blister formation (pathogenic) and
anti-Dsg3 antibodies that do not lead to blister formation (nonpathogenic) [129, 131]. While both
pathogenic and nonpathogenic autoantibodies are capable of binding to Dsg3 molecules expressed
within desmosomal bodies at the keratinocyte surface, it is not known why they lead to different
clinico-pathological outcomes.
To address these outstanding questions, novel experimental technologies and strategies are
required. To date, it has been extremely challenging to study the morphology of cell junctions
by standard light microscopy in living cells because of their small size (< 1 µm) and complex
structure [132, 133]. While electron microscopy has provided insight into the fine structure of
cell adhesion molecules [134], a model system for addressing dynamic changes due to (patho-)
physiological mechanisms has been lacking.
We utilized established, as well as novel roboticized AFM methods to visualize desmosomes
in physiologic and disease conditions at the nanoscale, and determine detailed nano-structural
correlates of the acantholythic process not previously attainable. Furthermore, we have linked
AFM data to functional alterations in cell behavior to develop a new paradigm for autoantibody
mediated tissue destruction in the skin. We reveal for the first time that pathogenic anti-Dsg 3
antibody attack is associated with changes in cellular stiffness that are distinct from the changes
induced by non-pathogenic antibody. Both pathogenic and non-pathogenic autoantibodies induce
an early, but incomplete, disruption of intercellular adhesion (“hit 1”), but pathogenic antibodies
alone lead to a later induction of apoptosis-related signaling (“hit 2”).

57

A

B

A1

1.5 μm

B1

20 μm

B1

20 μm
0

A2

1.5 μm

B2

2 μm

2 μm

A3

0
500 nm

B3

1 μm

1 μm
0

A4

B4

500 nm

1 μm

1 μm
0

SEM

AFM

Figure 4.1 Integrated SEM and AFM imaging of intercellular adhesion structures. A: Correlation
of SEM imaging with AFM imaging. Identically cultured plates of confluent HaCaT cells were
fixed and imaged by SEM and Multimode AFM in increasing magnifications. The lower magnification images (SEM: A1, A2; AFM: B1, B2) show the cells with clear boundaries between
neighboring cells where cell-cell adhesion occurs. The higher magnification images (SEM: A3,
A4; AFM: B3, B4) show details of the adhesion junction with strand-shaped structures in parallel
distribution between two cells.

58

4.2

Visualization of keratinocyte junctions under physiologic
and disease conditions by AFM methods
A

B

A1

IF

0

A2

B2

A3

B3

A4

B4

1.2 μm

AFM

0

B1

IF

1.0 μm

AFM

Figure 4.2 Integrated immunofluorescence (IF) and AFM imaging of intercellular adhesion structures. The same area on a confluent slide of HaCaT cells was captured simultaneously by IF and
AFM after fixation of the cells. For IF imaging, HaCaT cells were labeled with anti-cytokeratin
antibodies (red) and anti-desmoplakin antibodies (green). AFM images were captured by Bioscope
AFM at increasing resolution with scan sizes of 100 µm (B2), 50 µm (B3) and 20 µm (B4). C:
Correlation of IF imaging (C1) with AFM imaging [scan sizes of 100 µm (C2), 50 µm (C3) and
20 µm (C4) after treatment with 10 g/ml of the pathogenic anti-Dsg3 antibody Px4-3 for 24h

To validate the applicability of AFM for the investigation of cell-cell adhesion, we conducted
high magnification AFM imaging of fixed HaCaT cells, a well-established keratinocyte cell line.
59

We show that the AFM images correspond well with SEM imaging (considered the gold standard in
high resolution imaging of detailed intercellular adhesion structures in previous work [135])(Fig. 4.1A
and B).
Next, we demonstrate a close correlation of AFM images with IF images that visualize desmoplakin, an anchoring protein that links desmosomal proteins to intermediate filaments (Fig. 4.2A,
B). Autoantibody-treated cells (Fig. 4.2B) display a distinctive enlarged spacing between neighboring cells when compared to untreated cells (Fig. 4.2A), indicating cell dissociation that is clearly
observed both by IF and AFM in fixed cells likely based on a rearrangement of desmosomal and
cell-structural components [136].
In the higher resolution AFM images of untreated fixed keratinocytes (Fig. 4.3A), we observe structures in parallel organization between the cell membranes. However, in cells incubated with pathogenic anti-Dsg3 antibody or non-pathogenic anti-Dsg3 antibody for 24h, these
structures linking adjacent cells are no longer clearly detectable (Fig. 4.3A and B, respectively),
indicating that Dsg3-specific antibody binding leads to a perturbation of normal cell-cell adhesion. Conversely, the addition of two disease-irrelevant antibodies (goat anti-mouse Ig and mouse
anti-human HLA A, B, C) does not alter the intercellular structures observed without treatment
(Fig. 4.3C and D, respectively).
To visually follow disease processes in real-time, we obtained images of living keratinocytes
in culture before and after treatment with the same test and control antibodies used above over
a time-course of 30 min to 4 hs. Prior to antibody treatment (0h), multiple intact intercellular
structures can be observed. After treatment with the blister-inducing pathogenic anti-Dsg3 antibody, an initial perturbation of cell-cell adhesion can be visualized between 30m-1h, with all cells
completely disconnected after 2h (Fig. 4.4A), likely as a result of cytoskeletal reorganization as
shown previously for pathogenic Ab attack [137]. Treatment with the non-pathogenic antibody,
60

Fixed cells
24 h
with Ab

24 h
no Ab

A

B

C

D
0

500 nm

Figure 4.3 AFM imaging of autoantibody mediated desmosomal disruption in fixed cells. HaCaT
cells were grown to confluence, and left with no antibody treatment, or were treated for 24h with
(A) the pathogenic anti-Dsg3 antibody Px4-3, (B) non-pathogenic antibody Px4-4, (C) non-binding
irrelevant control antibody goat anti-mouse Ig, or (D) irrelevant control antibody
which does not lead to blister formation in vivo, results in a similar initial disruption of cell-cell
adhesion structures as seen with the pathogenic antibody. However, the non-pathogenic antibody
associated changes appear with a delayed onset and do not lead to a complete loss of cellular shape
(Fig. 4.4B). It is theoretically possible that the presumed “non”-pathogenic antibody does hold
pathogenic potential at much higher concentrations per target cell not tested here. Neither of the

61

irrelevant control antibodies alters the intercellular structures between adjacent cells at the corresponding time points (Fig. 4.4C and D). These data demonstrate the facility of AFM technology
not only to document autoantibody associated effects at static time points in fixed specimens, but
also to track autoantibody mediated effects on intracellular adhesion in real-time.
Living cells
0h
no Ab

30 min

1h

2h
with Ab

3h

4h

A

B

C

D

0

1.5 μm

Figure 4.4 AFM imaging of autoantibody mediated desmosomal disruption in live cellsconfluent
HaCaT cells were imaged live before antibody treatment or treated with the antibodies as in Fig. 4.3
for 30 min, 1 h, 2 h, 3 h, and 4 h

62

4.3
4.3.1

Mechanical characterization
AFM based quantification of cell surface stiffness reveals autoantibodyassociated nanostructural changes
*

100k

A

80k

Young's moduluss (Pa)

Young's modulus (Pa)

100k

60k
40k
20k
0

1-2 h

0-1 h

2-3 h

20k

100k

*
*

80k

*

60k

*

40k
20k
0
1-2 h
0-1 h
Add Antibody

Before

2-3 h

Time line

1-2 h
0-1 h
Add Antibody

Before

2-3 h

3-4 h

Time line

D

80k
60k
40k
20k
0

3-4 h

*

40k

3-4 h

Young's modulus (Pa
a)

You
ung's modulus (Pa
a)

C

*

60k

Time line
100k

B

80k

0
0h

*

1-2 h
0-1 h
Add Antibody

Before

2-3 h 3-4 h

Time line

Figure 4.5 AFM-guided elasticity measurements on HaCaT cells before and after treatment with
anti-Dsg3 and control antibodies. At each time point, between 20 and 40 HaCaT cells were tested,
and the force curves obtained from these measurements were processed to generate the corresponding Young’s modulus as a measure of cellular elasticity. In parallel experiments, measurement were
obtained before antibody treatment and after treatment with 10 g/ml of the pathogenic anti-Dsg3
antibody Px4-3 (B), the non-pathogenic antibody Px4-4 (C), the non-binding irrelevant control
antibody goat anti-mouse Ig (D).*p<0.05.

Dynamic mechanical properties of cells are increasingly being recognized as indicators and
63

regulators of physiological processes. For example, it has been shown that changes in cell elasticity/stiffness can effectively and reliably differentiate cellular conditions, such as non-cancerous vs.
metastatic cells [28, 76, 138], or the induction of apoptosis [24, 139]. To obtain dynamic measurements under physiological and pathophysiological conditions, we continuously measured elasticity/stiffness changes over a 4-hour time course following antibody treatment of HaCaT cells. In
untreated cells, or those treated with irrelevant control antibodies, the overall elasticity remained at
the same level over the measurement interval of 4 hours with means ranging from 32 kPa to 36 kPa
(Fig. 4.5A, B, and D, respectively). In the presence of the non-pathogenic antibody Px4-4, however, elasticity initially decreases within 2 hs from baseline levels of 32.5 ± 5.1 kPa to 17.6 ± 2.6
kPa, then returns to baseline after 2 hs, and remains between 35 kPa and 40 kPa thereafter, slightly
higher than the initial value (Fig. 4.5C). In cells treated with the pathogenic antibody Px4-3, stiffness first drops at 1 hour to 18.6 ± 4.6 kPa, similar to the results with non-pathogenic antibody.
However, in contrast to the non-pathogenic antibody condition, at 2 hs there is a consistent, sharp
increase (three-fold) in stiffness to levels around 70 kPa for the remaining period (Fig. 4.5B).

4.3.2

Desmosome disruption by other means

To define the (patho-) physiological correlates of the early decrease in cell stiffness induced by
both pathogenic and non-pathogenic antibody binding, we first investigated how the directed detachment of intercellular adhesion molecules via the manipulation of calcium levels in culture
media (calcium depletion) or by nanorobotic surgical methods (nanodissection) affects elasticity
measurements.
The formation of desmosomes requires the presence of calcium [140]. By depleting calcium
from the growth medium in vitro, the formation of desmosomes can be interrupted. Calciumdepleted cells grow separately from their neighboring cells, or, if they had previously reached
64

A1

A2

A3

B1

B2

B3

A

B

1.5 μm

1
0
-1
0

C1

C2

2

5 10 15 20 25 m

1
0
-1
0

5

1.5 μm

0

C3

2
Height (m)

C

Height (m)

2

3.5 μm

0

Height(m
m)

0

10 15 20 25m

1
0
-1
0

5 10 15 20 25m

Figure 4.6 Cell desmosome disruption by calcium depletion.
confluence, dissociate from adjacent cells. In our experiments, cells grown in calcium-containing
medium were shown to form stable cell attachments to their neighboring cells, have a somewhat
flat appearance (diameter 25µm, height 1 µm), and intercellular filame nts bridging neighboring
cells can clearly be observed (Fig. 4.6A1, B1, C1). Calcium depletion leads to a loss of cell-cell
adhesion that can be observed microscopically by an enlarged spacing between cells, and via AFM
by a disappearance of intercellular filaments (Fig. 4.6A2 and B2). Additionally, calcium depleted
cells become approximately three times smaller in diameter (10 µm), but almost three times higher
(3 µm) (Fig. 4.6C2). Once Ca2+ containing medium is reconstituted to cultures, cells recover to
normal pre-depletion shape (Fig. 4.6A3, B3, C3). AFM-measurements of cellular stiffness at the 3

65

states (pre-depletion, post-depletion, and recovery) show that the loss of intercellular adhesion, and
thus change in cell shape, is tightly correlated with a decrease in cellular stiffness (from 36.5 ±4.7
kPa to 19.1 ± 3.9 kPa). The re-formation of intercellular adhesion, however, restores stiffness back
to levels seen before the loss of cell-cell connections (35.2 ±6.7 kPa).
As an alternative method of disrupting intercellular adhesion, we performed novel, targeted
nanodissection at the intracellular adhesion junction of keratinocytes using an AFM based nanorobotic
surgery system. Corresponding AFM measurements of cellular elasticity indicate that the stiffness of an individual HaCaT cell decreases after nanodissection. The detailed process of AFM
nanosurgery will be discussed in Chapter 5.
Both the calcium depletion, as well as the nanorobotic manipulation experiments indicate that a
disruption of intercellular adhesion leads to a dependable decrease in cellular stiffness as measured
by AFM, and are consistent with the initial decrease in cell stiffness measured by AFM within one
hour after addition of autoantibody. Our data fits well with reports in the literature that see early
effects of PV IgG on cell adhesion [141, 142] and cytoskeleton disassembly [143] within 30 min
to 2 hs. While significant Pemphigus IgG induced acantholysis usually takes 12-24 hs to occur
(reviewed in [141]), complete loss of cell adhesion has been observed within 1-6 hs after antibody
treatment with pathogenic antibody in experimental models [144].
To define the (patho-) physiological correlates of the later increase in cell stiffness induced by
pathogenic-, but not non-pathogenic antibody binding, we next investigated whether the internal
remodeling of the cytoskeleton and nucleus in cells undergoing apoptosis is reflected in dynamic
changes in the structural properties of the cell that are measurable by AFM, as has been suggested [24, 139]. While the morphological characteristics of apoptosis are not necessarily detected
in the histopathology of PV lesions, increased expression of apoptotic signaling molecules has
been detected in both PV lesions and patient serum when compared to healthy controls, and PVIg
66

can trigger the activation of apoptotic signaling pathways in vitro [145]. Thus, we investigated
whether the late increase in cell stiffness detectable at 2 hs after pathogenic antibody treatment
has its pathophysiological correlate in an induction of apoptotic processes by exploring the extent
to which (i) apoptosis is induced by pathogenic anti-Dsg3 antibodies, (ii) similar levels of cell
stiffness increases can be observed with established inducers of apoptosis, and (iii) the antibody
induced stiffness increase can be reversed by blocking apoptotic signaling.

4.4

Two-hit hypothesis

We first measured whether pathogenic anti-Dsg3 antibodies induce an activation of apoptotic pathways in cultured keratinocytes. In cells treated with pathogenic anti-Dsg3 antibodies, we observe
an increase in combined Annexin V and PI positivity starting at 4h, with a peak at 12h and a return to baseline levels at 24h (Fig. 4.7). The non-pathogenic antibody induces a slight increase in
Annexin V and PI positivity as well, but this increase does not reach significant levels at any time
point, peaks later (around 16h) and returns to baseline levels earlier (at 20 hs) (Fig. 4.7). The return
to baseline levels within 24 hour after pathogenic antibody treatment indicates that the cells, while
exhibiting apoptosis-related phenomena, have not undergone terminal apoptosis. Our data support
previous findings by Pelacho et al. [146] that showed nuclear changes characteristic of apoptosis,
with a clear peak after an 8h incubation with PV-IgG, but a decrease thereafter.
Contrary to long held beliefs about the irreversible fate of apoptotic cells, it has recently been
shown that the initiation of apoptosis can be reversible with only transient exhibition of morphological signs of apoptosis at the cellular level [147]. Additionally, phosphatidylserine (PS) externalization (as measured by Annexin V binding in this study) is not restricted to apoptotic cells but
has also been described for human CD8 cytotoxic T lymphocytes that reversibly expose PS upon

67

PX4-4
PX4-3 *

+

40

+

% Annexxin and PI ce
ells
compared tto untreated c
cells

60

20

*
*

0

4h

8h

12h 16h
Time line

20h

24h

Figure 4.7 Apoptotic processes are induced by nonpathogenic and pathogenic anti-Dsg3 antibodies
T-cell receptor mediated antigen recognition [148]. Thus, the reversible PS externalization seen in
our study may indicate an antigen recognition-mediated initiation of apoptotic signaling cascades
that has the potential to, but does not necessarily progress to apoptotic cell death in all conditions.
To verify that the sustained increase in cellular stiffness observed after pathogenic anti-Dsg3
antibody binding can also be found as a result of apoptosis induction, we used the well established
apoptosis-inducer Fas ligand (FasL) [149]. Addition of FasL significantly increases the number of
apoptotic HaCaT cells in culture, while pretreatment of FasL-treated cultures with FasL neutralizing antibody significantly reduces FasL-induced apoptosis back to baseline levels. Parallel AFM
force measurements show that untreated HaCaT cells and Fas ligand treated cells have a low peak
in their stiffness distribution around 30 kPa and a high peak around 80 kPa. In FasL treated cells
the high stiffness peak sharply increases (28% of all cells, Fig. 4.8B2) when compared to untreated
cells (12 %, Fig. 4.8A2). When apoptosis is blocked by FasL neutralizing antibody, the second
peak of higher stiffness completely disappears (Fig. 4.8C2). The AFM measurements are echoed
by the data obtained from flow cytometry of the corresponding groups (Fig. 4.8A1,B1,C1). These
data indicate that the induction of apoptosis in HaCaT cell cultures leads to a sharp increase in cell
68

stiffness.
In order to provide evidence that the secondary increase in cellular stiffness produced by the
pathogenic autoantibodies is due to the induction of apoptosis, we attempted to block apoptosis
in pathogenic anti-Dsg3 antibody treated cells by (a) inhibiting all caspases, a family of cysteine
proteases that play essential roles in programmed cell death, or by (b) blocking the Fas/FasL pathway. Compared to baseline (Fig. 4.9A), addition of Px4-3 pathogenic antibody leads to a sharp
increase in the number of cells with higher stiffness (22.5%, Fig. 4.9B), that is similar to the distribution of cell stiffness (both in terms of percentage of cells showing cellular stiffening as well as
Young’s modulus values) as treatment with FasL (see Fig. 4.8B2). Pre-treatment of HaCaT cells
with caspase inhibitor before addition of Px4-3 antibody completely abrogates the development
of cells with a stiffer profile, even below background levels (Fig. 4.9C), indicating that anti-Dsg3
antibody-induced apoptosis and blocking thereof can be monitored by AFM stiffness measurements. Pre-treatment of HaCaT cells with FasL neutralizing antibody before addition of Px4-3
antibody, however, does not reduce the number of stiffer cells (Fig. 4.9D).
The disparate findings regarding the blocking of caspases or the FasL pathway indicate that
while inhibition of apoptotic processes can reverse pathogenic autoantibody mediated phenomena,
they are not mediated by the Fas/FasL signaling pathway. The inability of FasL neutralizing antibodies to reverse the pathogenic antibody-induced increase in stiffness is somewhat surprising,
as an activation of the Fas-FasL apoptotic pathway has repeatedly been implicated in PV pathomechanisms [150–152]. However, previous studies relied on observations from biopsy materials
or in vitro studies with patient sera or purified IgG. It has been shown that sera from untreated
PV patients contains high levels of FasL [150], which could conceivably lead to apoptosis independent of the effect of anti-Dsg antibodies. This notion is supported by the fact that inhibition
of FasL or caspase 8 can only partially inhibit PV sera induced apoptosis [150]. In addition to
69

no treatment #1

A

104

0.91

0.40
0.8

A1

A2

0.35
0.30
0.25

Percentile
e

FL3: PI Viability

103

102

0.20
0.15
0.10

101

0.05
95 5
95.5

100
FasL #1

B

10

0

0 00
0.00

2 84
2.84
10

1

2

10
FL1: Annexin

10

3

104

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

0.40

104

0.85

3.58

B1

B2

0.35
0.30
0.25

Percentile

FL3: PI Viability
y

103

102

0.20
0.15
0.10

101

0.05
85 3
85.3

100
10

0

0.00
0
00
0.0

10 3
10.3
10

1

2

10
FL1: Annexin

10

3

10

4

Young's modulus (Pa)

FasL + inhibitor #1
104

C

1.17

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k

0.40
0.6

C1

0.35

C2

0 30
0.30
Percentile

FL3: PI Viability

103

102

0.25
0.20
0.15
0.10

101

0.05
100

95.1
100

3.14
101

102
FL1: Annexin

103

0.00

104

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

Figure 4.8 Apoptotic processes are induced by pathogenic anti-Dsg3 antibodies detected by flow
cytometry and AFM nanomechanical analysis. A: No treatment; B: pathogenic antibody treatment
and C: pathogenic antibody treatment blocked by FasL inhibitor

70

containing FasL, serum autoantibody profiles vary among patients and may contain autoantibodies
to targets other than Dsg [153], whose effects have not been well studied. In this study, we use a
well-characterized and reproducible source of pathogenic PV anti-Dsg mAbs [131, 144] with nonpathogenic PV anti-Dsg mAbs as controls, in order to discern effects that can be attributed solely
to the action of anti-Dsg3 on its cellular target.
0.40

0.30

0.25

0.25

Percentile

0.30

0.20
0.15

0.15
0.10

0.05

0.05

0.5
0.4

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

0.3
0.2

0.00

0.40
0.35

C

B

0.20

0.10

0.00

Percentile

0.35

0.30
Percentile

Percentile

0.35

0.40

A

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

D

0.25
0.20
0.15
0.10

0.1

0.05
0.0

0.40

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k

0.00

Young's modulus (Pa)

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

0.40 increase in keratinocyte stiffness can be
Figure 4.9 The pathogenic anti-Dsg3 antibody-induced
0.35
blocked by caspase inhibitor, but not by Fas ligand neutralizer
0.35

0.30

E

0.30

F

0.25

inhibitors

Percentile

Percentile

Similar to FasL, activation of caspases has been0.25
observed in PV model systems, and caspase
0.20

0.15
have

0.20

been shown to block PV IgG-induced0.15
acantholysis in keratinocyte monolayers and

0.10

0.10

skin organ culture [154]. Our finding that a caspase inhibitor can completely abrogate anti-Dsg3 in0.05

duced

0.00
changes

0.05

in cell stiffness supports the notion that
0.00apoptotic signaling mechanisms are indeed
20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

20.0k 40.0k 60.0k 80.0k 100.0k 120.0k
Young's modulus (Pa)

71

involved in the downstream processes after antibody binding. However, as stated earlier, apoptotic signaling in PV may not necessarily progress to cell death, but exert its influence elsewhere,
such as by targeting proteins involved in the regulation of cell contacts and of the cytoskeleton
during [155]. As such, caspase 3 has been shown to specifically cleave the cytoplasmic tail of
Dsg3 and help release its extracellular domain from the cell surface, resulting in the disruption of
the desmosome structure and contributing to cell rounding and disintegration of the intermediate
filament system [154]. Along these lines, our ability to block anti-Dsg3 induced cell stiffening by
blocking caspases could simply reflect an inhibition of Dsg3 cleavage, and not apoptosis. However, in keeping with our own data on the disruption of i.c. adhesion structures presented above,
a disintegration of the keratin network should lead to cell softening and not stiffening. Thus, it is
conceivable that the anti-Dsg3 induced cell stiffening observed by AFM in this study can be viewed
as a sensitive marker for the early initiation of apoptotic signaling pathways that may contribute to
cellular dissociation, but not necessarily reach full apoptotic cell death.

4.5

An integrated model for antibody-mediated damage in the
skin

Three major hypotheses have been proposed regarding the mechanisms by which anti-Dsg antibody binding to the cell surface leads to acantholysis: (i) “steric hindrance”, the direct inhibition
of Dsg transinteractions [156, 157], (ii) depletion of desmosomal proteins from the keratinocyte
surface [158], and (iii) initiation of signal transduction events that lead to altered desmosome assembly, cytoskeleton derangement, cell cycle alterations, and apoptosis [159]. However, to date,
there is no conclusive theory of antibody-mediated acantholysis, and the role of apoptosis is unsettled [145]. While apoptotic phenomena have been observed in PV, there is considerable disagree72

ment regarding its role in acantholysis. Some groups have shown that acantholysis can occur in
the absence of apoptosis [160], and find that hallmarks of apoptosis, such as changes in nuclear
morphology and cell death, are detectable only late and subsequent to acantholysis [145]. Others
favor the theory that apoptotic signaling precedes acantholysis, but necessarily leads to apoptosis,
and have termed this paradigm “apoptolysis” [161]. The data presented in this study support the
notion that a dissociation of cell adhesion structures precedes the detection of apoptotic signaling
phenomena after pathogenic antibody attack.
While it has been suggested before that multiple steps are needed for PV mAb-initiated acantholysis [160, 161], we propose a novel “2-Hit hypothesis” of anti-Dsg3 autoantibody effects in
PV. “Hit 1” reflects structural changes in the keratinocyte that are induced both by pathogenic and
non-pathogenic antibodies. “Hit 2” involves a functional change in the keratinocyte biology that is
promoted by pathogenic, but not non-pathogenic antibodies. In this model, “Hit 1” has the pathophysiological correlate of an initial, but incomplete cellular dissociation, while “Hit 2” correlates
with the initiation of apoptotic signaling pathways. Given that the non-pathogenic antibody leads
to an initial cellular dissociation that is also seen with pathogenic antibody binding, but does not
alter the overall shape of the cell and does not induce the later cell structural changes seen with
pathogenic antibodies, it follows logically that “Hit 2” is needed to progress to clinical disease,
while “Hit 1” may be a reversible phenomenon.

4.6

Chapter summary

While AFM has been previously used to study the binding properties of anti-Dsg1 and 3 antibodies
to planar surfaces in a cell-free environment [158,162], we have for the first time established novel
nanorobotic AFM technologies for the investigation of PV pathomechanisms in vitro. We show

73

that, in addition to visually tracking the effects of anti-Dsg3 autoantibody binding on cellular adhesion in high resolution and real-time, AFM can reliably detect nanostuctural changes at the cell
surface in response to pathogenic and non-pathogenic antibody binding and downstream signaling
events that allow conclusions regarding disease pathomechanisms relevant to autoimmunity. We
demonstrate that AFM is an exquisitely sensitive technique capable of detecting cellular mechanostructural changes at a higher resolution and at earlier time-points than standard techniques, and
could, thus, serve as an excellent investigative tool and predictive screen to assess the impact disease associated disturbances in defined cellular parameters and functional pathways and blockage
of those pathways with potential novel therapeutic agents.

74

Chapter 5
Modeling and experimental verifications of
cell tensegrity

5.1

Introduction

The tensegrity cell model is a qualitative modeling technique for adherent cells [85]. It has been
proved to be consistent with the mechanical behavior of living cells [163]. The main concept in
delineating the mechanical properties of cells is that the cell body is a tensional integrity structure,
thus the name tensegrity. The cytoskeleton behaves as a discrete mechanical network composed
of three elements: microtubule, actin filament and intermediate filament [92]. The three elements
have different mechanical property themselves. In that, microtubule is the stiffest of the three and
thus would bear the compressional force when the cytoskeleton is compressed; while actin filament
and intermediate filament would assume more tension when the cell is stretched [85]. Through the
inter-connected cytoskeletal network, cells would maintain the structure integrity under different
force interactions.
It is generally accepted that cell can establish tensegrity force balance with its cytoskeletal
network and the ECM. The ECM provides the external support for the force balance, in which the
integrin of cells binds to ECM in the formation of the so-called focal adhesion complex [164]. The
focal adhesion sites are discrete, non-uniform protrusion regions of the cell periphery [59]. On
one hand, they provide the linkage between the cell and ECM, and at the same time serve as the
75

signaling transduction hub. They also host a cluster of proteins that behave as force sensors [59].
In tensegrity force balance, pre-stress in the filaments is a major factor governing cell mechanics. Under certain force balance or structural configuration, cell will have a corresponding
pre-stress, and ultimately a certain level of elasticity that can generally be revealed by AFM probing. Thus by measuring the elasticity of the cell, either locally or globally [89], the pre-stress of
the cell can be determined and its physical configuration can be predicted. Therefore to determine
whether cells can be modeled as tensegrity structures, one would need to tune the cell pre-stress by
biochemically modifying its living conditions and monitor whether the elasticity changes accordingly.
Metastatic adenocarcinoma cells show anchorage-independent growth patterns, and they become round in shape; whereas normal mesothelial cells show a large, and more stretched morphology. Experiments have verified that the cancerous type is seventy percent less stiff [76]. From
the force balancing perspective, cells with weak focal adhesions should less pre-stressed than cells
with stronger and denser focal adhesions. An analogy can be drawn for the cells with focal adhesion of different strength and the tent set up with varied number of anchors to the ground. Another
evidence that signifies the difference in pre-stress lies in the cytoskeleton structure itself, when
cells are less pre-stressed, they tend to have a disorganized actin filament network [165], a sign of
less tensional forces in individual actin filaments.
The tensegrity modeling of adherent cells is normally about single cells and only cell-ECM
interaction is considered in most cases. Cell-cell interactions, although often ignored when considering cell anchorage, could also be a source of external support in term of force balancing.
Being a special type of cell-cell adhesion structure, desmosome is one of the main cell-cell adhesion complexes in epithelial cells, provides mechanical strength to maintain the tissue integrity.
Desmosome links the intermediate filaments of neighboring cells through cadherin type of adhe76

sion. From the development point of view, desmosomes from different layers of epidermis vary in
size and appearance: basal layer epidermis have less-organized desmosomes with smaller dimensions compared with desmosome in the suprabasal layers [166]. Apparently, cells at the suprabasal
layers are tougher and stronger. We speculate that when desmosome becomes denser, cellular tension becomes stronger and the pre-stress becomes larger. All these evidences point to the idea
that cell-cell adhesions, especially desmosomes play an important role in providing the external
support just like focal adhesions do. In this study, we put this hypothesis into test by modulating
the cell-cell adhesions between neighboring cells using biochemical and mechanical methods, we
were ableFigure
to obtain
quantitative stiffness data validating the change of cellular stiffness.
1

A

B

C

Figure 5.1 Desmosome structure with different desmosomal proteins connecting each other all the
way to the intermediate filaments. AFM image and zoom-in (B) correlates with fluorescence image
(C)

5.2

Desmosome disruption leads to decrease in cell stiffness

Keratinocyte is a major constituent of the epidermis and is abundant in desmosome-based cellcell adhesions (Fig. 5.1). As mentioned above, desmosome provides the strength to maintain the
mechanical integrity of the tissue by “spot-welding” the intermediate filaments from neighboring
cells. Desmosome is composed of a cluster of proteins to form a molecular structure. The desmo77

some complex employs the cytoplasmic protein plaque to tether to the intermediate filament. To
disrupt the connection between neighboring cells would require the disassembly of the desmosome, which would release the intermediate filament linkage, which normally plays the role of
anchor points. The disassembly of the desmosome would therefore remove the anchor points from
the tensegrity structure and ultimately reduce the amount of stress in the system. This hypothesis
was verified by two biochemical experiments.

5.2.1

Desmosome disassembly results in stiffness decrease

The desmosome sometimes can be damaged in certain diseases, and one of them is called PV, a
potentially fatal autoimmune disease that targets one of the proteins in the complex, Dsg3 [145].
The binding of the autoimmune antibody would caused the release of the protein from the complex
and eventually leads to the disassembly of the desmosome structure. The damaged desmosome
leads to the loss of cellular adhesion between neighboring keratinocyte cells and eventually causes
blistering of the skin from the body fluid underneath the cells.
We tested the effect of the autoimmune antibody in vitro on keratinocytes. The administration
of anti-Dsg3 antibody induced the loss of cellular adhesion structures. This was confirmed by
comparing images obtained before and after the administration of the antibody. The AFM images
show the enlarged spacing between neighboring cells, a clear evidence of cell adhesion loss, as
well as retracted filament structures from the cell peripheral [10]. The topographical change is
accompanied by the decrease of the cell stiffness as measured by AFM based nanoindentation.
The Young’s modulus value of normal cells dropped from 29.8 ± 4.9 to 18.6 ± 4.6 kPa for antiDsg3 antibody treated cells (Fig. 5.2A).
Since the formation of desmosome requires the presence in the growth medium of Ca2+ , which
would form covalent bond with the desmosomal protein Dsg [167]. Depletion of the Ca2+ in the
78

Figure 7

A
Young's modulus (kPa)

60
50
40
30
20
10
0

Normal

Anti-Dsg3

2+

Ca depleted

B

C

Figure 5.2 Young’s modulus comparison for normal and calcium depleted cells
growth medium would hinder the formation of desmosome between cells. AFM imaging shows
a significant difference in terms of dimensions for cells cultured in normal growth medium and
the ones from mediums with Ca2+ depleted. Normal cells are stretched with a diameter around
25 µm and a height around 1 µm; while Calcium depleted cells are rounded up with a diameter
around 12 µm and a cell height around 3 µm. The stiffness of both types of cells were also
measured and compared. The Young’s modulus values agree well with previous measurements
for normal cells and antibody treated cells. The stiffness decreases from 36 ± 4.7 kPa to around
19 ± 3.9 kPa for normal and Calcium depleted26cells in Fig. 5.2A. These biochemical methods
indeed verified that the disruption of desmosome would lead to the decrease in cellular stiffness,
an indication of reduced pre-stress in the tensegrity-based cytoskeletal structure. The release of
intermediate filament connection was induced by desmosome disassembly or non-formation. We
then demonstrated that by severing the intermediate filament mechanically and directly, we could

79

also achieve the similar effect.

5.2.2

AFM based nanosurgery resulting in the decrease of stiffness

There have been reports of nanodissection on fixed rat aortic smooth muscle cells [21] and mouse
endothelial cells [168]; in both cases, a ‘scar’ was created successfully on top of the cell surface. The dissection of live cells proves to be more complicated than fixed cells. Once fixed, the
polymeric structure of the cell membrane as well as the cytoskeleton will be cross-linked, evidenced by the large increase of stiffness of fixed cells [113]. Thus fixed cellular bodies can be
regarded as a single material structure. However, live cells display distinct mechanical properties,
elasticity in their delicate membrane and the cytoskeleton structure and viscosity in their cytoplasm. It was shown that intermediate filaments have extremely high tensile strength by AFM
stretching [53, 169, 170]. Therefore, it behaves as a composite material structure. To dissect the
intermediate filament network inside the cell poses larger challenges (illustrated in Fig. 5.3A). The
cell body will deform when the AFM tip is pushed against the membrane. A tip sharp enough
is required to penetrate the cell membrane before it reaches the cytoplasm and the intermediate
filaments. Moreover, the lateral force, which is the primary driving force that causes the incision
should be controlled to cut through the filament and also monitored to confirm the dissection is
successfully done.
The force displacement curve in Fig. 5.3C shows the whole process when the AFM tip was
driven towards the cell membrane until it penetrates the cell membrane compared with the process
without penetration in Fig. 5.3B. Beginning with the contact point denoted as the circle when the
AFM tip starts to deform the cell membrane and when the laser detector starts to detect vertical
deflection; subsequently at certain point when tip deflection large enough, AFM tip generate a force
to penetrate the cell membrane. The penetration force was observed when there is a sharp drop
80

Figure 3

B 50

AFM tip
Cell membrane

Intermediate
filaments

40
Force (nN)

Cytoplasm

Retract
Approach

30
20
10
0

C 50
Force (nN)

A

0

1
2
Displacement (m)

Retract
Approach

40

15

30

10

20

5
1.5

10
0

2.0

Cutting
Depth

0

1
2
3
Displacement (m)

Figure 5.3 A: Illustration of composite cell structure during cell indentation and dissection; B:
Force-displacement curve without penetration of the cell membrane; C: Force displacement curve
with penetration of the cell membrane with the kink zoomed-in in the inset.

of repulsive force indicating a smaller resistance of cytoplasm compared with the cell membrane.
The measured penetration force for keratinocytes is around 1.9 ± 0.5 nN. The distance traveled by
the vertical piezo actuator between the contact point and penetration point can be used as a guide
to set the overall cutting depth for the nanomanipulation operation.
AFM force measurement has been applied in investigations of biological process. Most of
these applications focus on the normal force obtained from vertical deflection of the cantilever.
For AFM based nanodissection, we utilize the lateral force to cut through the elastic filament
structures. By maintaining a constant normal force and moving the AFM cantilever horizontally,
the lateral deflection of the cantilever can be obtained (as illustrated in Fig. 5.4A) from the PSD
signal. The normal force applied to the cantilever was determined by the depth of penetration and
depth of AFM tip contacting the substrate underneath as measured by force displacement in the
previous section.
The lateral position of the AFM tip was determined by a pre-defined trajectory around a specific
cell shown as the 3D construction of the original22AFM image from the custom-made nanomanipulation software in Fig. 5.4C. When moving along the trajectory, at each time t, the cutting force in
the planar direction can be determined by F = Fl / cos(θ ) , where Fl is the lateral force due to the

81

Figure 4

A

D

X position
Y position

10

m

5
0
0

B

5

10

15

20

25

30

35

C

nm

x

45

Z position

400

θ

40

200
0

y

-200

0

5

10

15

20

25

30

35

40

45

Lateral force
Filtered lateral force

30

nN

20
10
0
-10

0

5

10

15

20

25

30

35

20

40
45
time(sec)

15

nN

10

5

0

20

21

22

23

24

25

26

27

Figure 5.4 A:The lateral force was the laser position shift in the horizontal direction on the detector;
B: The tangent direction of the trajectory defines the angle (θ ) between the cutting force F and the
lateral force Fl ; C: The trajectory of the nanosurgery on the cell peripheral between neighboring
cells with the black triangle indicating the tip position; D: the XY position, cutting force and the
Z piezo position in the process of the nanosurgery along the path in C with a cutting depth of 100
23
nm.

twist of the cantilever perpendicular to its longitude direction and θ is the angle at each position
(illustrated in Fig.5.4B). The trajectory was then translated to position information, which directs
the lateral movement of the probe. A real time measurement data was shown in Fig. 5.4D, in which

82

the XY position, the cutting force F and the Z cutting depth were recorded. Three repeated cutting procedures were performed, the lateral movement is at 1.5 µm/s maximum for both X and Y;
while a constant Z position is maintained immediately when the probe engages with the cell sample. The maximum cutting force recorded during the operation was around 20 nN. Similar force
profile was observed for all three cutting procedures, it was maintained at higher level less than
20 nN for about 2 seconds before lowered to around 10 nN. Sudden release of force to extreme
low levels, or spikes, in the force profile should be caused by the snap of bundles of intermediate
filaments. Images in Fig. 5.5A and B show the AFM image before and after three repetitions of
nanosurgery with cutting depth of 100 nm. A line segment section at the working area shows the
topographical change in Fig. 5.5C and D. The filaments structures were removed as indicated by
the height decrease in the topography image, and the removal was controlled within 300 nm.
The stiffness data was collected before and after the dissection operations, which cut off the
connections between neighboring cells through repetition of nanosurgery. The Young’s modulus
value drops from 36.1 ± 3.5 kPa to 18.4 ± 1.9 kPa (Fig. 5.6). Thus it is conclusive that the stiffness
of cells follows the same pattern as antibody treated cells and Calcium depleted cells. The wellconnected cells are around 1.5-2 times stiffer than the cells whose intermediate filaments are cut
loose thus lose their anchoring point from neighboring cells. The well-connected cells with intact
cell-cell adhesion are around 1.5 times stiffer than the cells whose intermediate filaments are cut
loose. When cells are modeled as tensegrity, prestress determines the mechanical response of the
whole cytoskeleton structure to external load such as the nanoindentation force from the probe.
With less external support from intermediate filaments of their neighbors, cells would have to
establish a new force balance, one which presumably receives no contribution from tension in the
intermediate filaments. Theoretical calculation was then performed to verify the stiffness change
with and without intermediate filament tensions in the structural model.
83

Figure 5

A

C

800

600

Height (nm)

Height (nm)

800

B

400
200
0

0

2
4
6
Section (m)

D

600
400
200
0

8

0

2
4
6
Section (m)

8

Figure 5.5 Nanodissection of intermediate filaments by AFM (before and after) with cross section
to indicate
Figure
6 the cut-off bundle

A

B
Before

30
20
10
0

After

40
Cell count

Cell count

40

0
20 40 60 80 100
Young's modulus (kPa)

30
20
10
0

0

20 40 60 80 100
Young's modulus (kPa)

Figure 5.6 The stiffness response before and after AFM based nanodissection.
Therefore both modulations of desmosome yielded a cell sample without cell-cell adhesion,
and both cases the cell display a decrease in stiffness. Thus it confirms that the loss of cell-cell

24
84

adhesion either through mechanical dissection of intermediate filament or through desmosome
disassembly would result in the loss of cell stiffness. Conceptually, when cells are modeled as
tensegrity prestress becomes the main characteristics that determine the stiffness of the cytoskeleton structure; and it has been well documented that cell will become stiffer when prestress increases [93, 171, 172]. Similar conditions can be drawn for the keratinocytes with their intermediate filaments damaged. Since the intermediate filaments are tensional elements in the cytoskeleton
system, therefore the loss of tension due to the loss of intermediate filament based adhesions either
from antibody binding, calcium depletion or mechanical dissection would lower the prestress in
the tensional elements, so as the stiffness. Theoretical calculation was then performed to verify the
stiffness change with and without intermediate filament in the structural model, thus to confirm if
the dissection operation was successfully achieved.

5.3

Quantitative modeling based on six struts tensegrity structure

The structure model shown below with 6 struts (black) and 24 cables (blue) was originally developed by Ingber [173] as a conceptual model for cellular tensegrity (Fig. 5.7A). It is constructed by
connecting the ends of compressional struts with the tensional cables. The struts mimic the role
of microtubule in the cytoskeleton while the cables denote the actin filaments. Considering actin
filaments normally lie beneath the cell membrane and are abundant at the cell peripheral [92], the
design ensures that all the tensional cables for microfilament are at the boundary of the structure.
Since in each direction, we have a symmetric structure of two struts, thus the same notation is used
for both struts in each direction, as AA, BB and CC. The length of each strut is L and the distance
between them is s = L/2. The distance between each pair of notes (cable length of AB, BC or AC)
85

is l.
To quantify the relationship between the stress and the strain, a signature characteristics of
mechanical behavior, parallel loads are applied to both ends of the structure in one direction. A
force of

T
2

is applied to each node of C along the x direction to stretch the structure. At this

configuration and with the applied load, we want to find out the relationship between the strain and
the stress, or in other words, the modulus in the load direction.
Figure 2

A

B

A
A

B

C

B

T/2
T/2

C

C

T/2

T/2

B

C

B

X

A
A

Figure 5.7 The configuration of the tensegrity structure without intermediate filaments, the black
lines indicate the microtubule as the compressive elements and the blue line indicate the actin
filament as the tensional elements

By virtual work method, we define that the work of T on an incremental extension δsX per unit
reference volume V of the model equals the work of uniaxial stress on the incremental change in
uniaxial strain δeX , a relationship captured in Eq. 5.1.
T δ sX
= σX δeX
V

(5.1)

where eX and sX denote the strain and extension in X direction, and σX denotes stress in the same
direction. Since we know the dimension of the structure, we can calculate the volume (V ) enclosed

86

by it as in Eq. 5.2:
V=

5L3
16

(5.2)

Strain is defined as the unit extension, so we have:

δsX
s

δeX =

(5.3)

From Eq. 5.3 and considering the distance between the parallel struts, it follows that δeX = 2

δsX
L

.

By substituting it and Eq. 5.3 becomes:

16T δsX
2δs
= σX X
3
5L
L

(5.4)

thus,
σX =

8T
5L2

(5.5)

By definition of modulus as the ratio between strain and stress in X direction:

E∗ =

dσX
dσX
L dσX
=s
=
deX
dsX
2 dsX

(5.6)

Taking the derivative of Eq. 5.5, we obtain the relationship between strain and stress:

dσX
8 dT
= 2
dsX
5L dsX

(5.7)

By substituting Eq. 5.7 into Eq. 5.6, we finally arrive at the expression of modulus as:

E∗ =

L 8 dT
0.8 dT
=
2
2 5L dsX
L dsX

87

(5.8)

The E ∗ in Eq. 5.8 is the Young’s modulus obtained from the structural derivation, and it will be
applied to both structure with and without intermediate filaments by plug-in the

dT
dsX

from the two

structures.

5.3.1

Without intermediate filaments

From Eq. 5.8, we found the expression of E ∗ that is written in terms of

dT
dsX .

Then the only remain-

ing term is to find the relation between the force T and the distance between the struts sX . For
the configuration without the intermediate filaments, the calculation is as follows. The relationship
between the force T and the extension is:

T = 2FAB

sX
sX − LBB
+ 2FAC
lAB
lAC

(5.9)

where FAB and FAC are the forces in the corresponding cable. The kinematics of the configuration
in Fig. 5.7 is:
lAB =

1
2

(LBB − sX )2 + (sY )2 + (LAA )2

1
2
1
lBC =
2

lAC =

2
(LAA − sZ )2 + (sX )2 + LCC

(5.10)

2
(LCC − sY )2 + (sZ )2 + LBB

If we assume that the cables are linearly elastic of stiffness k and resting length lr . The force then
in each cable is given:
F=





k(l − lr ) if l > lr



0

(5.11)

if l < lr

By plugging the kinematics of the structure in Eq. 5.10 and the force expression in Eq. 5.11,
we have the relationship between T and sX . Taking the derivation of that, we can obtain

88

dT
dsX .

This

derivation will be evaluated in the initial state (the reference state) where the parameters are:

lAB = lAC = lBC = l0
LAA = LBB = LCC = L0
L0
sX = sY = sZ =
2
8
l0 =
L0
3

(5.12)

dT
|0 . Plug it into Eq. 5.8, we have the definition of
After plugging in these parameters, we obtain: ds
X

E ∗ in the configuration without the intermediate filament.

F 1 + 4ε0
.
E ∗ = 15.6 2
L0 1 + 12ε0
where ε0 is the initial cable strain, defined as ε0 =

5.3.2

(5.13)

l−lr
l .

With intermediate filaments

The structures with intermediate filaments are shown in Fig. 5.7B. The red lines connecting the
nodes in the middle are intermediate filaments. This structure is introduced by considering the fact
in biology that the intermediate filaments are projecting from the nucleus to the peripheral of the
cell [92]. Thus the designed structure has a tensional cable connecting each pair of struts and the
cable all crossing the center O. With this additional structural components, the kinematics for the

89

actin filament remains; while the kinematics for the intermediate filament can be characterized as:
1
2
1
rOB =
2
1
rOC =
2

2
s2x + LAA

rOA =

(5.14)

2
s2y + LBB
2
s2z + LCC

With the addition of intermediate filament structures, some force balance establishment has to
take into consideration the tension in the intermediate filament: forces in NOA , NOB and NOC . The
force balance now has been reconfigured and can be expressed as:
sx
sx
sx − LBB
+ FAC
+ NOA
)
lAB
lAC
rOA
LBB − sy
sy
sy
FBC
− FAB
− NOB
=0
lBC
lAB
rOB
sz
sz
LAA − sz
FAC
− FBC
− NOC
=0
lAC
lBC
rOC

T = 2(FAB

(5.15)

The force in each of the tensional elements including the actin filaments and the intermediate
filaments can then be expressed in Hooke’s law. For actin filaments, it remains the same as the
previous configuration in Eq. 5.11. For intermediate filaments, a second order elastic element was
added with elastic constant k2 on the basis of a first order elastic element:

F=





k1 (r − rr ) + k2 (r − rr )2

if r > rr




0

if r < rr

(5.16)

where rr is the reference length (zero tensional force) for the intermediate filaments.
To find he relationship between the applied load T and the displacement sX , we can plug the
kinematics into the force balance in Eq. 5.15 and replace the force with the expressions in Eq. 5.16,

90

we would have the static balance condition with the intrinsic parameter of the components and the
kinematic. With the same procedure as in the previous configuration, we can take the spatial
derivative of T with respect to sX to obtain

dT
dsX .

And this derivative will be evaluated at the initial

state with the following parameters:

lAB = lAc = lBC = l0
LAA = LBB = LCC = L0
(5.17)
sX = sY = sZ = L0 /2
l0 =

After evaluation under initial condition, the derivative

3/8L0
dT
sX |0

can be plugged into Eq. 5.8 , we have

∗ :
Ewith

E∗ =

L 8 dT
= EF0 + EF1 + EF2
2 5L2 dsx

(5.18)

where EF0 = 0.33 LF02 6εε00+1 defines the contribution from the stress in the actin filament; EF1 =
0

0.01 LF12 43εε00+2
0
and EF2 =

defines the contribution from the linear portion (F1 , k1 ) of the intermediate filament

F2 1.6ε0 +0.3
L0 ε0 (ε0 +1)

defines contribution from the quadratic portion (F2 , k2 ) of the intermediate

filament. Compared with EF0 and EF1 , EF2 is much smaller and therefore can be neglected. We
can assume that the intermediate filament can also be modeled as a first order Hooke relation.
However, EF1 and EF0 are comparable, and thus EF1 will make a difference for the modulus with
intermediate filament structure.
With the intrinsic parameters of individual filaments obtained from [173], we can calculate the
Young’s modulus values numerically. The parameters used in the calculation are shown in Tab.
5.1. The plotting in Fig. 5.8 shows the comparison of Young’s modulus values with respect to
the strain for three different situations, without intermediate filaments, with intermediate filaments
91

Table 5.1 Parameter list for the numerical evaluation
Variable
F0
F1
F1
L0
K
k1
k2

Physical meaning
Average force generated by a single actomyosin motor
Average force generated by a single intermediate filament from linear portion
Average force generated by a single intermediate filament from quadratic portion
Estimated length of an actin filament from
volume fraction
The stiffness coefficient of an actin filament
The linear portion of stiffness coefficient
of an intermediate filament
The quadratic portion of stiffness coefficient of an intermediate filament

value
1.6

unit
nN

3.2

nN

0.8

nN

0.7

µm

0.05

pN/nm

0.10

pN/nm

0.25

pN/nm

only the linear portion and with intermediate filaments both linear and quadratic terms. The curve
for the linear portion of intermediate filaments overlaps with the one containing both linear and
quadratic portions. This also confirms that EF2 is negligible. The Young’s modulus with intermediate filaments in the configuration increased with 18 percent at the upper boundary when ε0 → 1.
The trend is also indicative of the contribution from the intermediate filaments, that is when the
strain increases, the stiffness difference is larger. This agrees well with the functional contribution
of intermediate filaments in the cytoskeleton system. They are responsible for the shape maintaining when there is large strain or deformation [92].

5.4

Chapter summary

In this chapter, we validated that intercellular adhesion is part of the mechanism employed by
epithelial cells to achieve force balance in a cytoskeleton based tensegrity structure. We used

92

Figure 8

150

Young modulus (Pa)

With IFs
Without IFs

100

50

0

0.01

0.02

0.03
Strain

0.04

0.05

0.06

Figure 5.8 Young Modulus comparison with and without intermediate filaments from simulation
based on the structure model
AFM based nanorobotics to dissect the cell-cell adhesion between keratinocyte cells, primarily the
intermediate filament structures. The mechanical behavior data was subsequently obtained and a
significant decrease of cellular stiffness was observed. A linkage between the integrity of cell-cell
adhesion and the stiffness of the cell structure can be built. The cell adhesion loss was also induced
biochemically by antibody treatment and calcium depletion which all result a disrupted desmosome
structure. Both methods yield cell samples that behave similarly as when cell-cell adhesion was cut
loose mechanically, and stiffness decrease was observed expectedly. The findings were then taken
to test by a tensegrity structural model with and without intermediate filaments, the simulation
result verifies the reduction in Young’s modulus values when there is no extra tensional element
provided by the intermediate filament. Thus far, the hypothesis was validated that epithelial cells
employ cell-cell adhesion as a complementary mechanism to cell-ECM adhesion to sustain their
mechanical integrity during cell remodeling. This research work may provide some insights in
27
93

the development of new cell growth environment by taking into consideration of cell-cell adhesion
beyond cell-ECM interactions.

94

Chapter 6
Dynamic characterization of cell signaling
pathways mechanics

6.1

Introduction

EGF is a growth factor that binds to epidermal growth factor receptor, EGFR, a cell surface receptor [174] to promote cell growth and proliferation. The ligand binding can trigger a series of
downstream signaling events that regulate cell behaviors, resulting in cell proliferation, differentiation or migration. These different cell biological behaviors would result in different mechanical
behaviors. The A431 human epidermoid carcinoma cell line has been widely used as a biochemical
model for the investigation of signaling-pathway-related cell responses by EGF stimulation [57].
EGF treated A431 cells show significant rounding and swelling that have been reported by researchers [175]. It has long been postulated that signaling pathways could regulate the configuration of the cytoskeleton as a mechanism of facilitating the perspective cell behavior. For instance,
cancer cell would display loose attachment, i.e. cell adhesion, to make way for metastasis [76];
migration cells would have stronger cell adhesion in the filopodia, the leading edges of migration
direction than in the trailing edges [176]. The configuration of cytoskeleton thus will determine
the overall structure of the cell body; and one can use the cytoskeleton configuration to reveal the
biological behavior of living cells after certain biochemical cues. Since the cytoskeleton is deemed
as the main force bearer of the whole cell [85, 163], it can then be probed by mechanical force; the
95

mechanical response from the cell, or cytoskeleton would provide abundant information about its
mechanical status. We normally refer these information as biomechanical makers. Nanomechanical sensors are normally utilized to perform the force loading and measurement; the measurement
result would yield unique insight about the cell behavior and cell signaling pathways.
The imaging modality is based on force interactions at the molecular level between a cantilever
with a sharp tip attached at the free end and the sample beneath it. The force interaction is mainly
determined by the distance between the two, thus by controlling a constant interaction force, one
can build a topography map of the sample by raster scanning its entire surface. The nature of the
measurement process and its ability of working in liquid environment make AFM an ideal tool for
visualization of cellular and molecular structures in situ. Meanwhile, the probing interaction would
also provide the mechanical property of the cell samples [177]. We have observed ultra-structural
changes in real time at nanometer scale in cellular adhesions on keratinocyte cells [10]. Others have
also reported results from AFM imaging about cellular structural observation, either on animal cell
lines or bacteria cells [1]. Cellular elasticity can be measured by recording the force displacement
curves through AFM nanoindentation. Its effectiveness as biomarkers for physiological conditions
has been validated previously [178]. Cellular viscoelasticity can be obtained by fitting AFM force
displacement curves with Johnson-Kendall-Roberts (JKR) dynamic model [179].
QCM-D is another nanomechanical senor that could measure interactions at the molecular
level [47]. It has been used for molecular recognitions, such as membrane formations [48], protein
adsorptions [49], and cell spreading [50]. It operates at thickness-shear mode by oscillating a
quartz crystal disk with high frequency (normally 5 MHz) and low amplitude (less than 1 nm in
lateral direction). It collects the frequency and amplitude changes of the quartz crystal and reveals
the mechanical/structural status of the sample attached on top of the sensor disk: the shift in
resonance frequency indicates the mass absorption or loss; while the energy dissipation alteration
96

could be used to derive the viscoelasticity property of the adsorbed film on the disk [51, 52].
The combination of AFM and QCM-D for biological investigations has been achieved by previous works [180,181]. Functionally, QCM-D measures viscoelastic property change of biological
matters at different physiological conditions. The mechanical information can be verified by probing the sample surface topography using AFM under. This combination has been demonstrated in
the study by [180] where the effect of different substrates on cell shape and cell mechanics were
investigated. The topography and mechanical property of the cells are obtained by AFM imaging
and force measurement, and at the same time QCM-D characterizes the adhesion between the cells
and the substrates through the frequency and dissipation monitoring. Instrumentation of AFM
combined with QCM-D substrate has been developed by [181]. The system was equipped with an
oscillating quartz crystal as the substrate for AFM experiment. Therefore the AFM imaging and
QCM-D recording can be performed simultaneously on the same sample. However, the attention
for comparative characterization of cellular viscoelasticity by AFM and QCM-D has been lacking.
It has been shown that both AFM and QCM-D have the capability for viscoelasticity measurement.
While AFM scans and deforms cells from the top, QCM-D measures the sensor oscillation from
the bottom where cell adhesion with the substrate dominates the acoustic wave decay. AFM performs cellular investigations in the fashion of single cell analysis, and QCM-D studies the property
of a small population, a single cell layer spreading on the quartz crystal. In all these aspects, AFM
and QCM-D can verify and complement each other.

6.1.1

AFM energy dissipation and hysteresivity measurements

The nanoindentation process by AFM consists the loading and unloading of the vertical force
exerted by the cantilever as shown in Fig. 6.1. The vertical piezo firstly drives the cantilever
towards the cell sample and deforms it (P1 → P2 → P3 ); then it retracts back to the original height
97

Fig. 4

Approach
Retract

25
Deflection (nN)

B

A

30

Tip

A 431 Cell

P1
P2

20
15

P3

10

P4

5
0
0.0
P1

0.5

1.0
P5

1.5
P2

2.0
P4 P3

Z displacement(m)

P5

Figure 6.1 AFM energy dissipation

(P3 → P4 → P5 ). During the process, cantilever deflection will be picked up by position sensitive
device (PSD) to generate the applied forces at each location, and this recorded relationship is
the force-displacement curve. For each cycle, forces in the approach curve at each displacement
position is always larger than those in the retraction curve. This hysteresis in applied force over
the displacement defines the energy loss for the loading/unloading process. The energy required to
deform the cell surface at a nominal distance on loading is not entirely recovered from the cell in
withdrawing the same distance during unloading. If we define the energy under the approaching
curve as A, and the energy under the retraction curve as R, the area of the hysteresis loop (A − R)
therefore depicts the amount of energy dissipated during each cycle. Furthermore, we can define
ζ=

A−R
A

as the hysteresivity, a characteristics and measure of the viscosity property [182]. This

viscoelastic index ζ indicates the viscoelastic property of a material, with ζ = 0(A = R) indicating
pure elastic material and ζ = 1(R = 0) pure plastic material. Normal material will have a ζ value
0 < ζ < 1 showing both viscous and elastic properties [183].

98

6.1.2

QCM-D energy dissipation measurement

QCM uses a quartz crystal as an oscillator, and an alternating electrical field will generate a shear
deformation of the surface, as shown in Fig. 6.2A and 6.2B. The upper and lower surfaces will
move in parallel but opposite directions laterally, generating an acoustic wave that will propagate
in the direction perpendicular to the sensor disk surface. The acoustic wave has a frequency of
f = nv/(2tq ), where f is the resonant frequency, n is the overtone number and tq is the thickness of
the crystal. For the QCM working in gaseous condition, since the oscillation decays negligibly, the
Sauerbrey relationship defines the linear relationship between the mass change and the frequency
shift as [47]:
C
∆m = − ∆ f
n

(6.1)

where C = tq ρq / f0 , and ρq is the density of the crystal.
However, when the sensor is immersed in liquid as shown in Fig. 6.2C, the acoustic wave
will propagate through the liquid and the oscillation amplitude will decay exponentially over time.
The measurement circuit would periodically switch on and off the power input, the free decay of
oscillation can be obtained by recording the output voltage [47]. The voltage decay can be fitted
by:
A = A0 et/τ sin(2π f t + α)

(6.2)

The obtained parameters τ and f can then be used to evaluate the energy dissipation factor:

D=

1
π fτ

(6.3)

which can also be considered as energy loss ratio - loss energy over stored energy. It is related
to the viscoelasticity of the material on top of the sensor disk [184]. Several mechanical models
99

have been proposed to relate the viscosity and elasticity of the film with the measured D value.
Voinova and coworkers have developed a continuum mechanics model considering the oscillation
wave propagating through a viscoelastic material. The material was immersed in a Newtonian fluid
with either known or unknown properties [51]. The model will produce the relationship of ∆D and
∆ fFig.with
the material properties on the top.
1

Electrode

A

C

+
V
-

Liquid with
viscoelastic
material

B
V
+

Quartz crystal

Oscillation
decay

~

Figure 6.2 Schematic of QCM-D principle

In this work, we apply AFM probing and the QCM-D observations on the same set of cellular
signaling process induced by EGF stimulation. Both experiments monitor a monolayer of A431
cells that grow on a substrate (glass substrate for AFM and quartz crystal surface for QCM-D)
which will be stimulated by EGF. Whist QCM-D monitors the basal area of the cell layer close to
the quartz crystal surface, where cell adhesion dominates; AFM will probe the apical surface of a
single cell away from the substrate reveling cytoplasm behavior.
By analyzing and comparing the quantitative data from both techniques, the viscoelastic behavior of the cell with a more complete profile can be obtained. Once the mechanical profile is
obtained from both readings, a model of the signaling pathway dynamics following the EGF simulation could be built by considering the cell mechanical structure as the plant, which makes the
readings as output of the system, and the signaling pathway as the controller. We can address the
signaling events from the systematic control perspective and identify the pathway related dynamics
100

Fig. 3

d(t)
x(t)

Figure 6.3 The modeling of viscoelastic material under AFM based force displacement measurement during the retract process
model structures and parameters.

6.2
6.2.1

Model development
AFM viscoelastic characterization

The following assumption was made for the force displacement based viscoelastic modeling: the
cellular material is modeled as a Voigt element, a parallel configuration of spring (k) and dashpot
(η) and there is only repulsive force between the AFM tip and the cell. Then we define the vertical
displacement of the Z piezo as d and the deformation of the cell surface as x (Fig. 6.3). The spring
constant of the cantilever is defined as kc and its deflection is dc .
For the retraction process, based on the Voigt element configuration, the dynamic force balance
can be defined as:
F = k(xm − x) + η x˙

(6.4)

where xm , a constant, is the maximum vertical deformation of the cellular body from the approach
process and x, a function of t, is the vertical contact point elevation due to the release of the can-

101

tilever pressure in the withdraw process. The force can also be calculated by cantilever deflection
as:
F = kc (dcm − ∆dc )

(6.5)

where dcm is the maximum cantilever deflection resulted from approach process and ∆dc is the
release of deflection and is defined as the difference between the distance of retraction from the
piezo (d) and the cell deformation recovery (x)

∆dc = d − x

(6.6)

Thus plug Eq. 6.5 and 6.6 into Eq. 6.4, and d = vt, where v is the velocity of the Z piezo motion.
We have:
η x˙ − (kc + k)x = f0 − kc vt

(6.7)

where f0 = kc dcm − kxm is a constant resulted from the approach process. The solution to the ordinary differential Eq. 6.7 can be used to evaluate deformation recovery x under different influence
of η.

6.2.2

QCM-D based cell membrane peeling model

The sensor disk motion poses the relative motion of the cell with regard to the fluid in the sensor
chamber. The effect is similar with the cell under liquid shear flow. In the leading edge, the
adhesion bonds between integrin receptors and the substrate ligands will be ruptured due to the
flow resulted cell strain; while at the other edge, there will be bond formation activities attributed
to the increased integrin and ligand affinity as shown in Fig. 6.4. The model in Fig. 6.4B describes
this peeling of focal adhesion between the cell and the substrate due to shear stress under liquid

102

Fig. 2

A

C

Shear flow

Cytoplasm layer
Bond
rapture

Contact
length

Substrate motion

B

Interfacial layer

Bond
formation

Sensor layer

Sensor
disk

Membrane
peeling
Cell
cytoplasm

Shear flow
S=0

Substrate motion
Figure 6.4 Cell membrane peeling model at the basal area (A, B); and the cell model as a whole
(C) in which the cytoplasm was modeled as a Voigt element.
flow. In the 2D model, focal adhesion sites are characterized by the contact length Lc between the
cell and the substrate. The substrate moves with speed vs and the peeling speed is proportional to it
as v p = κvs , where κ is the coefficient. The focal adhesion is modeled as a spring with coefficient
of ka , whereas the single focal adhesion bond strength is fa = ka (l − λ ), where l and λ are the
lengths of the bond in the stretched and unstretched states. The total focal adhesion strength then
becomes Fa = Na ka (l − λ ), where Na is the adhesion bond density. The adhesion kinetic equation
specifies the balance for the formation and dissociation of adhesion bonds which determines the

103

temporal- and spatial-dependent bond density Na (t, s) [185] [61]:

∂ Na
∂ Na
= vp
+ k f (Nl0 − Na )(Nr0 − Na ) − kr Na
∂t
∂s

(6.8)

kts (l − λ )2
0
k f = k f exp
2kb T

(6.9)

kr = kr0 exp

k f − kts (l − λ )2
2kb T

(6.10)

where Nl0 is the ligand density on the substrate; Nr0 is the cell surface receptor density; k f is
the bond formation rate with initial value of k0f ; kr is the bond formation rate with initial value kr0 ;
kts is the transient elastic constant of the adhesion bond and it is reasonable to assume that kts < ka
for this model; kb is the Boltzmann constant, and T is the temperature. For energy dissipation rate
calculations, we have

∂ Na
∂t

= 0, and Eqs. 6.8 to 6.10 will be evaluated numerically by fourth order

Runge-Kutta method with the initial value Na0 solved by:

0 = k f (Nl0 − Na )(Nr0 − Na ) − kr Na

(6.11)

The total energy dissipation rate for the focal adhesion dissociation can be defined as:

E˙a = wc

Lc
0

Fa v p

where wc is the effective width of the contact area.

104

∂l
dx
∂s

(6.12)

50

30
20

B

30
20
10

10
0

Approach
Retract

40
Force (nN)

Force (nN)

40

50

A

Approach
Retract

0

0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (Pm)

0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement (Pm)

Figure 6.5 AFM energy dissipation measurement on a single cell with and without EGF treatment

6.3

Results and Discussion

The AFM measurement result before and after the stimulation with EGF are shown in Fig. 6.5 A
and B. The displacement was controlled below 2 µm and the resulted loading force is within 30
nN, with the induced maximum cell deformation less than 500 nM (< 25% of cell height). These
parameters will be used throughout the force measurement process. The hysteresis area between
the loading and unloading force curve (A − R), is increased from 3.2 to 6.9 femto-Joules (fJ) due
to the stimulation, a 126 percent jump.

A

50

40

40

30

30

Count

Count

50

20

20
10

10
0

B

0
1

2

3 4 5 6 7 8
Energy dissipation (fJ)

9

1

2

3 4 5 6 7 8
Energy dissipation (fJ)

9

Figure 6.6 AFM energy dissipation measurement on 100 different cells before and after EGF treatment.

105

A statistical analysis was performed on a population of cells before and after EGF stimulation.
The result is shown in Fig. 6.6A and B. The cell sample was first measured under AFM on 100
randomly selected cells. EGF was then applied to the petri dish where the cell sample sits with final
concentration of 40 nM. The sample was then incubated for 30 min till another AFM measurement
on 100 randomly selected cells. The result was plotted with normal distribution for energy dissipation before and after EGF addition. It can be observed that energy dissipation increased from
3.26 ± 0.95 fJ to 5.16 ± 0.89 fJ (Mean ± SD). This demonstrates the cells display a different
Figure 2-1: New result at 37 degree

9
8
7
6
5
4
3
2
1

0.6

Control
10 nM
20 nM
40 nM

EGF

0

Control
10 nM
20 nM
40 nM

0.5
Hysteresivity 

Energy dissipation (fJ)

mechanical property after EGF stimulation.

20 40 60 80 100
Time (Minute)

0.4
0.3
0.2
0.1

EGF

0

20 40 60 80 100
Time (Minute)

Figure 6.7 Dynamic mechanical responses (Energy dissipation) of A431 cells to stimulations with
four different concentrations of EGF: 0, 10 nM, 20 nM, and 40 nM

This is a general trend on the cell population, and then we want to define how cells dynamically
adjust their cytoskeleton structure throughout the stimulation process. We then monitor the stimulation process dynamically in real time. After the EGF stimulation, force displacement curves
were taken at a 1Hz frequency. Fig. 6.7 summarizes the dynamic response of the cells after EGF
treatment. Three different EGF concentrations were tested with 10 nM, 20 nM and 40 nM. The
same volume of buffer solution was added in a separated experiment as control. The baseline
measurement was longer at 40 min before EGF or control buffer was added. For all the three
106

Page 16 of 21

dosages, a few minutes after the administration of EGF, there is an increase in energy dissipation,
suggesting the effect of EGF is immediate; this increase continues until around 20 min when the
response starts to level off. At the end of the observation, the energy dissipation is two fold the
amount of baseline value. The experiment result also shows that energy dissipation increased with
dosage dependent by comparing the three dosages. Larger energy dissipation was observed on 20
nM and 40 nM EGF experiments than 10 nM EGF. The 40 nM dose has a slightly large increase
than 20 nM, although the difference is small (Fig. 6.7). Similar dynamics was observed for the
hysteresivity measurement (Fig. 6.8).

9
8
7
6
5
4
3
2
1

0.6

Control
10 nM
20 nM
40 nM

EGF

0

Control
10 nM
20 nM
40 nM

0.5
Hysteresivity 

Energy dissipation (fJ)

re 2-1: New result at 37 degree

20 40 60 80 100
Time (Minute)

0.4
0.3
0.2
0.1

EGF

0

20 40 60 80 100
Time (Minute)

Figure 6.8 Dynamic mechanical responses (Hysteresivity) of A431 cells to stimulations with four
different concentrations of EGF: 0, 10 nM, 20 nM, and 40 nM

The QCM-D experiment results for a continuous monitoring of 100 min after stimulation of
EGF of dosage variation confirmed that the energy dissipation decreased over time. After stimulation with EGF of different concentrations, the dissipation decrease is clearly dosage-dependent.
Immediately after stimulation, the dissipation shows a sharp increase, then after about half hour
it drops down. The initial sharp increase of dissipation is caused by the liquid flush during the
change of EGF containing medium. Overall the energy dissipation is decreased due to the stimulation. Fig. 6.9 shows representative dose-dependent and time-dependent changes in ∆D, of the
Page 16 of 21

107

Figure 3

0 nM
10 nM
20 nM
40 nM

20

0

-6

D (x10 )

10

-10
-20
-30
0

10 20 30 40 50 60
Time (Minute)

Figure 6.9 QCM-D measurement of energy dissipation over time after EGF treatment with concentration of 0 nM, 10 nM, 20 nM and 40 nM
Control

40
monolayer of A431 cells in response
to EGF. Unlike the20AFM
nM results shown in Fig. 6.7 and 6.8,
40 nM

20

where ζ increased with time and with dose, the QCM-D100
detected
nM a decrease in ∆D with time and
D (x10-6)

0

with dose. As a mechanical property,
the change in energy dissipation factor ∆D measured directly
-20
by the QCM-D is analogous-40
to hysteresivity, ζ , computed from the AFM measurements, where
-60

both quantities correspond to the mechanical energy loss relative to the energy input per measure-80

0
20 probe
40 different
60 80 regions
100 of the deposited cells; the AFM
ment cycle. However, the two techniques
Time (min)

tip probes 500 nm into the cell from its top surface while the QCM-D senses approximately 100
nm into the basal area up from the bottom surface of the sensor crystal.
Interestingly, responses in D and ζ reached their plateaus within a similar time frame after the
addition of the EGF (40-50 minutes for the decrease in ∆D and 50 minutes for the increase in ζ ).
Both changes showed signs of saturation in a similar dose range (responses at 20 nM and 40 nM at
very similar levels). These similarities indicate that both mechanical responses (D and ζ ) that were
developed locally might be synchronously mediated by the same globe cell signaling pathway(s).
For pure elastic materials, there will be no energy dissipation. Accordingly, when an AFM
Page 19 of 21

108

Figure 8

Deformation (m)

0.20




0.15

0.10

0.05

0.00
0.0

0.1

0.2

0.3

0.4

0.5

Time (sec)

Figure 6.10 AFM energy dissipation simulation result shows that it takes longer time for the deformation to recover for a cell with higher viscosity η as indicated with η = 0.45 showing larger
deformation overall.
force measurement is performed on it, the approach and retract portions of the force displacement
curve would overlap. However, for biological samples, viscosity is inherent and researches have
shown that cellular components inside are mostly viscoelastic such as the cytoplasm fluid, even
cytoskeleton [186]. Therefore, normal force curves on biological samples in liquid display large
hysteresis between the approach and the retract curves.
The hysteresis between loading and unloading force curves indicates that the repulsive force
at each displacement position in unloading is lower than in loading. The origin of the decreased
repulsion on the unloading curve is that at any given nominal displacement the actual local separation is larger on retraction than on approach, because it takes a finite time for the deformed
cell surfaces to relax and to recover to its original shape. According to the modeling of force displacement process as described in the previous section, the deformation relaxation process can be
plotted in Fig. 6.10. At the given deformation xm = 170 nm, the increase of viscous factor η (from
0.35 to 0.45) will result in the increased relaxation time. In other words, it takes longer time for

109

Figure 9

2000

0

12

2

0

12

2

0

12

2

Energy dissipation rate (fJ/s)

Nf =2 x10 sites/cm
Nf =4 x10 sites/cm

1500

Nf =6 x10 sites/cm

1000

500

0
0

2

4

6

8

10

12

14

2

Contact area Lcwc (m )

Figure 6.11 Cell peeling model simulation
a cell with higher viscosity to recover its deformation, meaning at any given displacement, there
will be smaller repulsive force, resulting in a larger hysteresivity (ζ ) and larger energy dissipation (A − R). This means the cellular body would display increased viscous characteristics to have
increased energy dissipation after EGF stimulation.
For QCM-D measurement, energy dissipation measurement from QCM-D is the sum of the
energy dissipated in disruption of cellular adhesion Ea and the energy loss in the viscoelastic
cytoplasm Ec :E = Ea + Ec . The EGF reduces the focal adhesion area and strength in A431
cells [175, 187], which results in the decrease in Ea as confirmed by the adhesion peeling model
described in previous section. The simulation result in Fig. 6.11 shows that the decrease in contact
area A (X axis) as well as the reduced substrate adhesion ligand density (from 6 × 1012 to 2 × 1012
site/cm2 ) will reduce the energy dissipation rate E˙a (Y axis). On the other hand, the loss of focal
adhesion will cause the cell rounding and swelling [175]. Our AFM nanoindentation measurements revealed that the cell body becomes more viscous after EGF stimulation. The increase in
viscosity η in cell cytoplasm would induce relative high damping thus increase the cytoplasm en110

ergy loss during QCM-D measurement [47]. However, the QCM-D measured energy dissipation E
decreases after EGF treatment, indicating that Ea dominates the energy dissipation process and Ec
has minimal effect. The modeling of the wave propagation through the cellular material therefore
should employ the multilayer configuration (Fig. 6.4C). Cellular adhesion through focal adhesion
complex linking cytoskeleton with ECM creates an interfacial layer between the cell membrane
and the ECM with gap thickness ranging from 30 nm in closest contact area to 100-250 nm in the
rest areas [165, 188, 189].
Evidence shows that the acoustic wave decays almost completely before reaching the cell cytoplasm [189], rendering the energy dissipation in cellular adhesion become the primary energy
loss mechanism during the QCM-D measurement, which agrees with our experimental findings.
Therefore the characterization process benefits from the difference in measurement mechanisms
of the two instruments. AFM performs nanoindentation on cell membranes with less than 500
nm deformation in a single-cell analysis fashion reveling cell cytoplasm properties, while QCM-D
monitors the cell-ECM interfacial layer where focal adhesion dominates the energy dissipation.
Therefore, AFM and QCM-D work in a complemented manner in the characterization of the viscoelastic property changes.

6.4

Chapter summary

Cell signaling is one of the fundamental processes that control the cell fate. It regulates the cell
shape, and through which cell mechanics. To identify the dynamic signaling pathway in situ, instruments and techniques that are capable of monitoring the real-time mechanical property changes
(such as viscoelasticity) as well as structural rearrangements are crucial. AFM has been proved
to be an effective instrument to visualize cell membrane and cytoskeleton structures. It can also

111

provide the mechanical information with high temporal and spatial resolution. Meanwhile, the viscoelasticity change can also be measured as the change of energy dissipation of a cell monolayer
by means of QCM-D. Our study pioneered a novel approach where AFM and QCM-D can work in
complemented fashion for the characterization of cellular viscoelastic properties during signaling
induced cell cytoskeleton remodeling and thus facilitate the investigation of structural and mechanical biomarkers in cellular signaling processes. These mechanical readings will provide us more
insight into the dynamics of the cell signaling.

112

Chapter 7
Plant model derivation

7.1

Introduction

The QCM-D is a mechanical sensor that measures near surface changes by recording the mechanical oscillation of a quartz crystal. The shear mode oscillation is generated by applying an
alternating voltage to a thin AT-cut quartz crystal sandwiched between a pair of electrodes. The
piezoelectric property of the quartz crystal enables the fine displacement difference between electrodes, thus a high frequency shear-stress at the measuring surface. Because of the fine displacement in nanometer scale and the ultra-high oscillation frequency commonly at 5MHz in resonance,
high sensitivity to the changes on the crystal surface can be achieved. It has garnered attentions
from different fields ranging from monitoring chemical depositions [190] to detecting protein absorptions [191] and nucleic acid hybridizations [192]. Recently it has been used in probing the
interactions between the cell and the sensor disk as a culture substrate, making it a regular cell
adhesion assay of label-free manners [193–195].
When animal cells cultured on the sensor disk, the QCM-D device can detect the subtle changes
to the cell adhesion, i.e. cytoskeleton and focal adhesions under different physiological/pathological
conditions [196,197]. Specifically, it has been demonstrated by measurements of frequency changes,
the attachment and spreading of initially suspended cells can be monitored in real time [189, 198,
199]. We have previously characterized cell adhesion change following EGF treatment by energy
dissipation factor recordings [67, 68, 200]. The measurement and interpretation of energy dissi113

pation factor as an effective marker for cell adhesion has received ever increasing attention, as
mechanical force in this case plays an important role in various cellular processes. The mechanical
interaction at the protein/sensor disk interface contains valuable information strongly and uniquely
linked to focal adhesion. To characterize the related cell property requires extracting these information from the measurement data, thus a physical model that describes the experimental condition,
in this case a quartz crystal sensor in shear mode oscillation, emphasizes the role of mechanical
interaction and contains cellular properties as model parameters needs to be established [189].
Most of the models to date have been based on the equivalent circuit and impedance analysis
method [189, 201]. By constructing an equivalent circuit to model the electrical characteristics of
a quartz crystal resonator in contact with a viscoelastic material, the recorded impedance data can
then be analyzed to obtain the viscoelastic property change of the material, i.e. the cultured cell. A
commonly accepted model developed by Voinova and coworkers [51] took a continuum mechanics approach and modeled the mechanical wave generated by the sensor oscillation propagating
through a viscoelastic material covered by a Newtonian fluid. By setting up appropriate boundary
conditions, the change of frequency and energy dissipation factor can be expressed in terms of the
physical properties of the viscoelastic thin film and the fluid. These modeling approaches along
with others reviewed by Johannsmann [202], though popular and easy to use, treat the cells as a
uniform material and neglect the composite nature of the cell structure. In fact, more evidence is
pointing towards the focal adhesion complex as one of the main units that physically interact with
the sensor disk during the oscillation; there exists coupled multi-dimensional energy dissipation
sources at the fluid-solid interface. The models mentioned failed to take into consideration of the
physical interactions that governs these energy dissipative processes.
Thus a mechanical model that uses the mechanical oscillation as the measurement energy
source and physically mimics the cell/substrate interaction in detailed fashion down to the molec114

ular level is better positioned to meet the demand. Considering the strong correlation between
the focal adhesion and the energy dissipation factor [68, 203], we believe the frictional bond slip
between the integrin and ECM, the viscous damping between basal membrane and the trapping
liquid between the basal membrane and the substrate, as well as the frictional slip between the
focal adhesion complex and the stress fiber, plus the stress fiber and the cytoplasm are all sources
of energy dissipation, and they contribute in different manner and magnitude during each measurement cycle. Thus we propose a mechanical model based on the dynamics of the oscillation sensor
disk as well as the kinetics of the focal adhesion and stress fiber under the experimental conditions.
The model will consider these energy dissipative mechanisms and quantify their contributions, and
ultimately serve as an analytical tool for QCM-D related cell adhesion assays.

7.2

Model development

The oscillation of the sensor disk experiences the friction from the attached cell and the fluid which
dissipates the kinetic energy and slows down the movement. Modern tribology considers friction
at the macroscopic level originating from the formation and rupture of molecular bonds [204,
205]. Filippov and coworkers presented a simple model containing two plates with relative motion
connected by molecular bonds to address the frictional energy dissipation through bond rupture and
viscous component [206]. The energy dissipation through the viscous damping at the liquid-solid
interface can be easily comprehended [207, 208]; while the model dealt mainly with the energy
dissipative process through the molecular bond stick-slip motion [209]. Mechanical forces exerted
from the oscillation break the energy equilibrium of bond formation/dissociation. The bond length
and unbinding probability becomes dependent or a function of the applied force; thus it lowers
the barrier height of the unbinding process and elevated the dissociation probability and ultimately

115

FIGURES.
B

A

Voigt stress fiber
Stress fiber
Focal adhesion

V

Trapping liquid

1

2 … li=l0+Δli

l0
xi

Sensor disk

N bonds

Figure 1. The
QCM-D/cell
focal adhesion
energy dissipation
model: The
mechanical
Figure
7.1 The QCM-D/cell
focal adhesion
energy dissipation
model
structural model describes the sensor disk as a harmonic oscillator (A); while the physical
dissipates
kineticthe
energy
[210].
This was
by Bell’s
of bond bond
kinetics
under
interactionthe
between
cell and
the sensor
diskdescribed
is characterized
bymodel
the molecular
friction

(B). The bond formation and rupture events were calculated as the Bell model 26-27. The force in
mechanics influences [60, 211]. Base on friction model and combined with the Bell’s model of
the focal adhesion complex is balanced by the stress fiber which displays the viscoelastic
molecular bond kinetics, we designed a mechanical model to mimic the QCM-D measurement
property.
of cell adhesion. The model considers the sensor disk interacts with the cell through the focal
adhesion complex with molecular bonds.

7.2.1

The dynamics model for the sensor disk

The proposed mechanical model treat the quartz crystal as a harmonic oscillator [190] and the
oscillation is maintained by a spring (K) shown in Fig. 7.1A. The sensor disk anchored on the
spring will have an initial displacement as the energy input. The physical interaction between the
sensor disk and the cell is characterized by the molecular bond forces (Fb ) at the interfacial layer
denoting the integrin-ECM adhesion, plus the viscous damping (ηl ) from the liquid trapped in
between the basal membrane and the sensor disk. Thus the dynamics of the disk oscillation can be
defined:
M X¨ + ηl X˙ + Fb + K(X + Ls0 − Lsr) = 0
116

(7.1)

17

where M is the mass of the disk and ηl is the viscous damping coefficient; X is the index of
the disk horizontal motion; K is the spring constant; Lso and Lsr are the initial and rest length of
the spring. Therefore, η X˙ represents damping from the intrinsic decay and by the trapping fluid
and the interactive bond force is defined by Fb . K(X + Ls0 − Lsr) is the force in the spring and M X¨
is the force resulting from the acceleration of the sensor disk. The adhesion bond force Fb will be
derived in the next section.
Apparently, without the friction between the cell and the sensor disk caused by the relative
movement between the two or the intracellular fluid viscous damping, the sensor disk would be in
the oscillation motion at all time. With the friction in mind, we have the energy loss of the sensor
disk that was captured in energy dissipation factor(D), which is defined as the ratio between the
input energy and the output energy:

D=

1 Din − Dout
2π
Din

(7.2)

The input energy is the initial kinetic energy of the sensor disk defined as: Din = 21 MV02 ; while
the output energy is the current kinetic energy of the sensor disk defined as: Dout = 12 M X˙ 2 . Eq. 7.2
defines the total energy dissipation factor during a given time period; it will be used to evaluate
different energy dissipation channels and to compare with the experimental data from the QCM-D
measurement.

7.2.2

The mechanical model of cell adhesion

The focal adhesion is characterized as the total effect of the molecular bonds. Each molecular bond
is defined as a single spring connecting the cell and the moving disk [206] as shown in Fig. 7.1B,
and it has two states: a closed state indicating an intact bond that enables the connection while
117

an open state meaning the bond is ruptured and the connection is lost. The adhesion force of the
bonds in the lateral direction thus is defined as the sum of forces in each individual bond based on
the assumption that all the bonds are uniformly distributed and contribute to the total force equally:
N

Fb = ∑ qi fix

(7.3)

i=1

where Fb is the total force, i is the bond index and N is the number of bonds in the focal
adhesion complex, fix is the force in the ith bond projected in the X direction, qi is the state of the
bond with Eq. 7.3 will be evaluated based on the status of each bond and be used to update Fb in
the master equation Eq. 7.1. The transition between a closed bond and an open bond is a stochastic
process defined by:

qi (t + ∆t) = qi (t) − qi (t)H(ξi − ∆tko f f ) + (1 − qi (t))H(ξi − ∆tkon )

(7.4)

where H() is the Heaviside step function, ξi is a random number in (0, 1). Eq. 7.4 defines bond
transition and will be used to update each bond status with the formation and dissociation rate kon
and ko f f defined later in the section.
If it is a closed bond, then the force in the individual bond is defined according to Hook’s law:
fi = kb (li − l0 ), where kb is the spring constant of the bond, with the projection in the X direction
fix = fi xi /li ; li is the bond length and xi is its projection in the X direction. The geometrical relation
defines: xi =

li2 − l02 and l0 is the initial bond length and the distance between the sensor disk

and the adhesion plaque. Thus,

fix = fi

1 − (l0 /(l0 + ∆li ))2

118

(7.5)

˙ as:
The bond movement (x˙i ) can be defined with respect to the substrate movement (X)

x˙i = qi X˙ + λ (1 − qi )xi

(7.6)

Eq. 7.6 builds the position relationship between the bonds and the substrate and xi will be used
to update the length of each bond and eventually the formation and dissociation rate kon and ko f f .
The bond formation and dissociation is defined in Bell’s model [60, 211], which describes the
impact of mechanical forces on the kinetics of bond formation/dissociation, especially on the rate
of dissociation as:
ko f f (li ) = ko0 f f eβ fi ∆li

kon =

where ∆li is the bond stretch; β =





0
kon

ifτc > τb




0 τc
kon
τb

ifτc < τb

1
kB T ; τc

(7.7)

(7.8)

is the contact time defined by the time during which

the free end of an integrin is exposed to a moving contact area, and it is inversely proportional to
the moving speed: τc = Lc (X˙ − ∆l˙s ) , where Łc is the contact length and X˙ and ∆l˙s are speed of the
adhesion plaque motion and the stress fiber extension, respectively. τb is the intrinsic association
time of integrin and ligand molecules on the order between 0.01 and 1 sec, and we used a value of
0.01 sec [212]. Eqs. 7.7 and 7.8 will update the status of each bond in association with Eq. 7.4.
The actin filament is represented by a viscoelastic Voigt element, i.e., spring-dashpot combination in parallel. The force balance in the actin filament with respect to the binding forces is
defined:

119

N

∑ qi fi = ks∆ls + µs∆l˙s

(7.9)

i=1

where ks and µs are the elastic and damping coefficient of the stress fiber; ∆ls is the extension
of the stress fiber. The total force from the closed bonds is balanced by the stress fiber with elastic
component ks ∆ls and viscous component µs ∆l˙s . Eq. 7.9 uses the force information in all the bonds
and calculates the extension of the stress fiber ∆ls . The energy dissipation of the mechanical model
mainly comes from these two channels: the friction at the interface of cell-disk contact (Ea ) and
N

the viscous damping of the stress fiber (Ed =

∑ qi fi ∆ls dt). The cell-disk friction (Ea = Eb + Ev )
i=1

at the microscopic level also comes with two components: the rupture of adhesion bonds between
N

the cell and the sensor disk (Eb =
liquid at the interface (Ev = ηl

7.2.3

∑ fi ∆li dt) [185] and the viscous damping through the trapped

i=1
2
X˙ dt).

Gaussian compensation of oscillation amplitude

The quartz crystal of the sensor disk oscillates in different amplitude from different regions; as
studies have shown that a bell-shaped curve could be drawn for the relationship of the amplitude
with respect to the position on the sensor disk [208]. The amplitude is largest in the center and
fades away at the peripheral. Gaussian distribution compensation was applied to the simulation
to mimic this phenomenon. A Gaussian distribution profile for the oscillation of the sensor disk
was generated according to the equation:A =

A0 −
2π e

(x−x0 )2 +(y−y0 )2
σ2

, where A denotes the oscillation

amplitude in nanometer, i.e. the initial spring displacement (Fig. 7.2A); A0 is a coefficient to
normalize the amplitude values to render the largest amplitude to be 1 nm; x and y are indexes
for the dimension on the disk and σ is the variance defining the width of the Gaussian. The
profile generated with x0 = 5 and y0 = 5 in a sensor disk with a diameter of 10 mm, and a σ =

120

A
1.2
1
0.8
0.6
0.4
0.2
0
15
10

15

10

5

5

y (mm)
B

x (mm)

0

0

1.4
1.2
1
0.8
0.6
0.4
0.2
0

0

5

10

15

Sensor disk
Figure 7.2 A: The initial spring displacement L − L

will be varied according to a Gaussian

be varied according to a Gaussian
Figure
2. (A) The initial spring displacement Ls0-Lsr srwill
profile of the oscillation amplitude A = L − L s0 with
the maximum amplitude A0/2π of 1 nm
s0

sr

andof
a variance
of 3.16; amplitude
B: Illustration
of s0
the
manuallyamplitude
divided areas
withofequal
profile
the oscillation
A=L
-Lsensor
thewith
maximum
A0/2π
1 nm
sr withdisk
distribution of bonds, each was assigned with one amplitude.

and a variance of 3.16; (B) illustrates the sensor disk with manually divided areas with
equal in
distribution
of bonds,
was assigned
with one
amplitude.
3.16 is shown
Fig. 7.2B with
10 by 10each
resolution.
Each amplitude
will
be evaluated once by
numerical simulation, and the 100 simulation data would be compiled together to obtain the energy
dissipation data.

121

7.2.4

Numerical simulation

The calculations were achieved numerically with a Matlab program. The general strategy for the
simulation is like the following. The states of all the bonds were then initialized along with the
initial bond stretching ∆li using Eq. 7.6 along with the geometrical relationship, bond formation
(kon ) and dissociation rate (ko f f ) as in Eqs. 7.7 and 7.8. The system equilibrium was first achieved
by a continuous 10 µs running. For each simulation step (h = 10−8 s), after the bond kinetics
was initialized, the total bond force Fb can be calculated as in Eq. 7.3. The relationship between
˙ was defined subsequently as in Eq. 7.6.
the bond movement (x˙i ) and the substrate movement (X)
The master differential equation of the sensor disk dynamics was then solved numerically using
the 4th order Runge-Kutta method [213]. The stress fiber stretching could then be updated as in
Eq. 7.9. The total observation time is 50 µs and all the simulation parameters are listed in Tab. 7.1.
The sensor disk has a thickness of 0.33 mm and a diameter of 14 mm. With the density of quartz
comes around 2.65 g/cm3 , we can calculate the mass of the sensor disk around 0.134 g. Since the
frequency of the harmonic oscillator is determined by the mass and spring constant as f =

K
M,

to

achieve nature frequency of 15 MHz, the spring constant is designed to be 3.05 ×1013 N/m. Other
parameters were based on values provided in previous literature.

7.3
7.3.1

Result and discussion
The dynamics of the sensor disk and the kinetics of the bond

The disk model behaves in harmonic oscillation rather than the thickness-shear mode that QCMD
operates. Only the stored oscillation energy is twice than in thickness-shear mode, while other
characteristics remain [190, 220]. The oscillation amplitude diminishes over time due to the direct

122

Table 7.1 The parameters in the model with their physical meanings used for the simulation
Symbol
M
K
kB
T
η
0
kon
ko0 f f
kb
ks
µs
l0
lc

Physical meaning
Mass of the sensor disk
Spring constant of the main spring
Boltzmann constant
Room temperature
Damping coefficient
Forward rate constant
Reverse rate constant
Bond spring constant
Spring constant of stress fiber
Viscosity of stress fiber
Initial bond length
Contact length of the bond

value
0.134 g
3.05 × 1013 N/m
1.38 × 10−23m2 kgs−2 K −1
310 K
10−3 Ns/m
100s−1
1s−1
10nN/µm
45 nN/µm
45nNs/µm
10 nm
20 nm

Ref.

[214]
[212, 215]
[215, 216]
[211, 212]
[66]
[66, 217]
[211, 218]
[219]

mechanical coupling between the cell and extracellular matrix, i.e. the sensor disk. This mechanical coupling is decomposed into the pulling/pushing of the focal adhesion to the cytoskeleton [221]
and the viscous damping at the interface [207,208]. Shown in Fig. 7.3 is the dynamics of the sensor
disk at a given initial displacement of 1 nm for a given simulation time of 5 µs. The displacement
oscillates between -1 nm and 1 nm with the amplitude gradually decreases over cycles while the
maximum speed diminishes correspondingly; and the kinetic energy is dissipated at a factor of
around 10−5 (1 − Et /E0 ) shown in Fig. 7.3D.
The spontaneous bond association and rupture processes are strongly influenced by the relative
motion between the sensor disk and the cell. As shown in Fig. 7.3 third panel, in each oscillation
cycle when the speed goes beyond a certain value (VC

6.0 × 10−4 ) determined from simula-

tion shown in the Fig. 7.3B and C, the bond formation will be interrupted and the ratio of intact
bond decreases dramatically to as low as 0.2; as more and more bond breaks, the force in the remaining bond increases, promoting the ensuing bond rupture probability. The synchronized bond
dissociation in the stochastic bond behavior displayed the characteristics of the stick-slip motion

123

1

2

3

4

5

Position

1
0

A

-1

Veleocity

5

Vc

0
-5

B

1.2
1.0
0.8
0.6
0.4
0.2
0.0
2.0

Closed bond ratio

R1

C

R2

E0

Energy

*

1.5

E*t

-6

Energy (10 J)

Closed bond ratio Velocity (10-3 m/s)

-9

Position (10 m)

0

1.0
0.5

D

0.0

0

1

2
3
-6
Time (10 s)

4

5

Figure 7.3 The simulation result at a given initial displacement of 1 nm for a period of 5 µs.
Figure 3.The simulation result at a given initial displacement of 1 nm for a period of 5 μs:
regime (R(A)
Fig. 7.3C)theat displacement
a relative highoscillation
speed of movement.
the shows
stick-slip
Sub-figure
between -1 At
nmthis
andregime,
1 nm; (B)
theof
2 in indicates
the bond/disk
lower the
energy
for unbinding
raises
velocity
change interaction
during eachwould
oscillation
cycle;
The barrier
kineticsrequired
of the bond
formationand
andthus
rupture
the frequency
of bond
dissociation,
resulting kinetic
energy
is defined
shows
two distinct
regimes:
the low-velocity
sliding
in loss.
whichThis
theenergy
bond loss
kinetics
was as
notthe
N

energy summation
of every bond
breaking
events:
At rupture
speed lower
VC , the
bond
∑ fi ∆li [185].
influenced
and the stick-slip
regime
in which
synchronized
bond
was than
clearly
visible
i=1
-4
formation
and rupture
was
not influenced
the closed
when
the speed
reachesprocess
a critical
value
V>Vc (VCand
≈6.0×10
)(C).bond
The ratio
total hovers
kineticaround
energythe
of value
the

of kondisk
/(konoscillates
+ ko f f ), and
which
a little
1; denoting
and the absence
correlation
between
individual
sensor
theispeak
in less
eachthan
cycle
the totalofsystem
energy
declines
as the
-5
bond formation
and The
rupture
events
is a distinct
feature
bond
defined
bydissociation
(1-Et/E0) inin
simulation
proceeds.
energy
dissipation
factor
can of
bespontaneous
is around 10thermal

the low-velocity sliding regime (R1 in Fig. 7.3C) [206]. The binding/unbinding probability is not
(D).

124

19

affected at this regime and thus no energy is consumed in this period of every oscillation cycle.

7.3.2

Contributions from different energy dissipation source
0

1

2

3

4

5

4

2
1
0
4

A
Ev

3
2
1
0

B

-1

Ed

4

Stress fiber (10

-17

J)

Trapping fluid (10

-10

J)

Bond (10

-13

J)

Eb
3

2

0

C
0

1

2
3
-6
Time (10 s)

4

5

Figure 7.4 Dynamic energy dissipation for the three channels
Figure 4. Dynamic energy dissipation for the three channels: The energy consumed in a
-8
The total energy
dissipation model.
for the observation
periodThere
of one
simulation
stepof(h=10
aresingle
three main
channels
energy). dissipation
in the mechanical
The dynamic

period
can bebond
obtained
by the
summation
energydamping
consumption
forby
each
Thefluid
stress
fiber
molecular
rupture
events
and the of
viscous
caused
theperiod.
trapping
dissipate
(C)
dissipates
the least
(10-17disk
) of interface.
the three The
channels,
less than level
the energy
energy
as friction
at theenergy
cell-sensor
frictionsixat orders
the intracellular
is chardissipated
the trapping
fluid
(B)stress
and three
orders
less in
than
bond
rupture
energyis
acterized through
by the viscous
damping
of the
fiber, as
the force
thethe
focal
adhesion
complex
dissipation
(A)the
(atstress
bondfiber.
number=1k).
balanced by
These three energy dissipation channels consumed the dynamic energy

125

of the oscillating disk, and the simulation also showed that the sum of the three channels equal the
total kinetic energy loss.
Of the three energy dissipation channels, the stress fiber dissipates the least energy; six orders
lower than the viscous damping of the trapping liquid at the interface. The bond rupture energy
dissipation is three orders less than the trapping liquid dissipation simulated when bond number
is 105 . The three dynamic energies shown in Fig. 7.4 is calculated based on the energy consumed
in a period of each simulation time step (h = 10−8 s). As the number of bond increase, both
the viscous damping by the trapping fluid and the stress fiber remain in the same level (10−6
and 10−11 , respectively). However, the energy dissipation through the synchronized bond rupture
events increased almost linearly with respect to the bond number (Shown in Fig. 7.5). At the
bond number around 4 × 108 , the bond rupture energy levels with the viscous damping of the
trapping liquid; and it dominates the total energy dissipation at bond number 109 and beyond. It
is worth noting that these energy dissipation values and contribution relationships were calculated
for a simulation time of 5 µs. At longer observation time, these values vary but the relationship
remains.

7.3.3

Relationship of bond number and energy dissipation

Since the viscous damping of the trapping liquid will remain the same as the bond number increase causing the increase in the bond rupture dissipation, thus we can expect a linear increase in
total energy dissipation, considering the fact that stress fiber energy dissipation is negligible. This
was confirmed by the simulation when the bond number goes from 3 × 107 to 3 × 109 shown in
Fig. 7.6A. A further increase in bond number to 108 shows that the linear relationship transforms
to an exponential decay at bond number 4 × 109 and beyond as shown in Fig. 7.6B, with the linear
portion correlating with bond number range in Fig. 7.6A.
126

Bond Energy
Trapping fluid damping
Stress fiber viscous

1E-4

Energy (J)

1E-5
1E-6
1E-7
1E-8
1E-9
1E-10
10

5

10

6

7

10
10
Bond number

8

10

9

Figure 7.5 The energy dissipation channel contribution with respect to the bond number.
Figure 5. The energy dissipation channel contribution with respect to the bond number: As
5
9
Thebond
documented
data aboutfrom
bond10
density
adhesion
supports
thefrom
concept
molecular
the
number increases
to 10in
, focal
both the
viscous
damping
the of
trapping
liquid

bonds
number
than 10
Basedlevel;
on the
size the
of the
sensor
disk energy
and considering
theshows
cells a
and atthea higher
stress fiber
remain
the9 .same
while
bond
rupture
dissipation
5 cells. The
8
with
area increase.
around 1000-2000
µm2 [221,
222],
the disk
would
contain
a monolayer
of 10the
linear
At bond number
4×10
, bond
rupture
energy
dissipation
reaches
same level
2 [223], thus we could estimate
bond
density
for focal
adhesion
is around
200-500
bonds
and µm
beyond,
the bond rupture dissipation
with
the trapping
liquid
damping;
at bond
number
109 per

thatdominates
the total effective
bond number
would be in the order of 1010 . The measurement data from
the total energy
dissipation.
QCM-D for energy dissipation has D around 400 × 10−6 , from Fig. 7.6A it can be determined
that the effective bond number that are contributing to the energy dissipation process is around
1.6 × 109 . Therefore reasonably we could state that the energy dissipation arises predominantly
from the bond rupture energy and the fluid friction at the interface, and thus the energy dissipation
factor would appear linearly correlated with the bond number or further size of the focal adhesion
if we assume the bond density over the entire focal adhesion is uniform.

127

21

A

800

D
Linear fit of D

-6

D (x10 )

600
400
200
0
0

5

10 15 20 25
8
Bond number (x10 )

30

B
2.5

D
Exp fit of D

2.0

-3

D (x10 )

1.5
1.0
0.5
0.0
0

2
4
6
8
9
Bond number (x10 )

10

Thedissipation
energy dissipation
linearly
with number:
the bondA:
number.
FigureFigure
6. The 7.6
energy
increasesincreases
linearly with
the bond
As the number
of bonds goes from 105 to 108, the total energy dissipation factor went from under 10-6 to

7.3.4

Experimental observations

800×10-6. The simulation data was fitted with linear curve; fitting equation: D= 5.1×10-5+2.2

The EGF
reduced
the size
of theinfocal
the change
is in a time-dependent
×10-13treatment
*B (R2 >0.95);
A further
increase
bond adhesion,
number to and
108 shows
an exponential
growth (B)
manner
it isequation:
correlated
the real-time
dissipation factor measurement from the
withand
fitting
D with
0.0016*(1
 e  B /3.510 energy
) (R2 >0.95).
9

QCM-D [68]. The simulation uses the time-lapse normalized focal adhesion sizes as the reference
input, and linearly converts them to bond number (100 percent correlates with bond number of
1.6 × 109 ), and the energy dissipation factors corresponding to each bond number would then be
9

calculated based on the fitting equation for Fig. 7.6B: D = 0.0016(1−e−B/3.5×10 ). A strong corre22

128

D experiment
D simulation

5
0

-100
-6

D (x10 )

-6

D (x10 )

-5

0

-10
-200

-15
-20

-300

-25
-400
-30
-35
-20

-500
0

20

40

60

80 100 120 140 160

Time (Min)

Figure 7.7 The simulation data comparison with the experimental observation
Figure 7. The simulation data comparison with the experimental observation: The
lation between the energy dissipation factor from simulation (black) and from QCM-D experiment
simulation uses the normalized focal adhesion size as input reference and converts it to bond
(blue) is clearly visible as shown in Fig. 7.7.
number. The corresponding energy dissipation (black) shows strong correlation with the real-

time QCM-D measurement of energy dissipation factor (blue).

7.4

Chapter summary

In this chapter, we designed a mechanical model based on the physical interaction between the
QCM-D disk and the cell focal adhesion. The proposed model considers the sensor crystal as a
harmonic oscillator and the friction at the cell/disk interface due to the dynamic molecular bond
rupture as well as the viscous damping at stress fiber/cytoplasm causes the kinetic energy loss
during each oscillation cycle. We found that the QCM-D energy dissipation comes mainly from
the cell focal adhesion and sensor interaction, and the viscous damping caused by the trapping fluid
at the interface and by the stress fiber contribute insignificantly. The energy dissipation almost
linearly correlates with the bond number and thus the focal adhesion size. The finding confirms
the experimental results obtained early with the similar correlation. The proposed model, different

129

23

from present models which are overwhelmingly relied on equivalent circuit method, emphasizes
the importance of the physical interaction between the focal adhesion and the sensor disk and
its role in the energy dissipation process of sensor oscillation. Proved valid, this model and its
simulation result could well serve as an analytical tool for QCM-D based measurements of cell
adhesion.

130

Chapter 8
Modeling of signaling pathway dynamics

8.1

Introduction

EGFR is a transmembrane receptor that regulates cell growth, proliferation, motility, and differentiation through its downstream signaling pathways [224, 225]. These pathways are activated
by the binding of EGF to the extracellular domain of EGFR, followed by phosphorylation of the
cytoplasmic tyrosine kinase. Because of its critical role in regulation of cell growth, proliferation,
motility, and differentiation, an abnormal level of EGFR can often lead to aberrant cell signaling,
which may induce uncontrolled cell growth and malignant phenotypes. This is evidenced by the
fact that EGFR is highly expressed in a variety of human tumors [226]. In particular, overexpression of EGFR appears to promotes aberrant cell motility that has often been associated with tumor
invasion and metastasis [227].
EGF-induced cell motility is a highly coordinated cellular function involving precise regulation of cell adhesion and de-adhesion to extracellular matrix (ECM) proteins by multiple downstream pathways of EGFR signaling [228]. Three pathways, the phosphoinositide 3-kinase (PI3K)
pathway, the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ ERK)
pathway, and the phospholipase C (PLC) pathway, have previously been linked to the mediation
of the assembly and disassembly of focal adhesions that lead to changes in cell adhesion in various cell lines [229, 230]. Although many of the key signaling molecules in each pathway have
been elucidated and characterized, the understanding of how these pathways regulate cell motility
131

through coordination of cell adhesion and de-adhesion, is still limited by the complexity of this cell
signaling system, particularly its inherent nonlinearity (e.g., crosstalk, feedback loop, feedforward
loop) and dynamic complexity.
By far the traditional reductionist approach has failed to elucidate the decisive signaling reactions and molecules that are responsible for the cellular functions, which are often integrated
outputs of multiple signaling pathways [231]. A quantitative understanding of regulation of cellular functions is also seriously lacking. Thus, it becomes necessary to introduce computational
modeling approach at the system level to examine cellular functions in order to determine the responsible properties and interactions of the cellular components and assess how these components
work together to generate an integrated downstream signal output in response to various extracellular cues.
Over the years, mathematical models have been developed to quantitatively describe various aspects of EGFR signaling [232, 233] including receptor dynamics and trafficking [234, 235]
and downstream signaling events [236–238]. The transient versus the sustained responses of the
MAPK cascade [239–241], autocrine positive-feedback loops [242, 243], the cross-talk between
the downstream pathways of EGFR [243] have also been the areas of interests. Most of these models are described by a set of ODEs derived from a set of stoichiometric reactions and the dynamics
of the involved species based on the prior knowledge of their kinetic constants and concentrations [244].
Computational models of dynamic cell adhesion based on the receptor-ligand bonds between
the cytoskeleton and ECM have also been the focus of growing research activities. These models have provided quantitative analyses that have led to mechanistic insights into a wide range of
cellular processes such as cell spreading, migration, proliferation and differentiation. The thermodynamic framework of cell adhesion models was established by Bell back in 1978 [60]. Based
132

on his seminal work, peeling models were developed to reveal the kinetics of cell adhesion and
de-adhesion through quantitative mechanics of cell membrane [61,218,245,245]. A statistical mechanics model was also established to probe the adhesion of multicomponent membranes mediated
by specific adhesion molecules [246, 247]. The stochastic effects of adhesion bonds switching between open and closed states under constant force were investigated [248], so were the behaviors
of focal adhesions under various exogenous mechanical stimuli [249–252]. Some of the recent
studies have been focused on dynamics of integrins which serve as clutches that link F-actin and
ECM [214, 253, 254].
Although the importance of EGFR signaling in mediating cell adhesion has attracted a significant research interest in recent years, particularly in cancer research [58, 255–257], none of
signal-to-response models describing the coupling between multi biochemical pathways mediated
by EGFR signaling and adhesion mechanics of the cell has been reported. To our knowledge, the
only model that has made an attempt for a coupling between cell signaling and cell mechanics
is the one developed by Besser and Schwarz, where they connect the Rho pathway to adhesion
mechanics by treating the mechanical force as enzyme in the framework of Michaelis-Menten kinetics [66]. Besides only a single pathway was considered for the model, the validity of treating
the mechanical force as enzyme is also open to question. This could be attributed to the lack of
understanding of the molecular mechanism concerning the mechanotransduction. It is clear that
the development of an EGFR signaling-to-cell adhesion model that provides insight into how signals generated by environmental stimuli lead to phenotypic cell behavioral responses is one of the
major challenges in system biology [258]. This requires the establishment of a molecular element
that links between EGFR signaling and the mechanical response of the cell during alteration of cell
adhesion.
We hypothesize that such a cellular element is focal adhesions, which are a type of adhesion
133

protein complexes located across the bottom surface of cells that are primarily responsible for
cell adhesion to ECM and substratum [259]. The assembly and disassembly of focal adhesions
are known to be regulated by the pathways mediated by EGFR signaling [229, 230]. Therefore
biochemical model that describes the regulation of the restructuring of focal adhesions by EGFR
signaling and biomechanical model that depicts the cell-substrate adhesion through focal adhesions
could be linked together through focal adhesions. Based on this coupling, we propose to model the
regulation of dynamic changes in cell adhesion by EGFR signaling.
The dynamic change in cell adhesion can be measured with the use of a nanomechanical sensor
in QCM-D. The QCM-D is a highly sensitive nanomechanical device that provides simultaneous
measurements of changes in frequency (∆ f ) and energy dissipation factor (∆D) of an oscillating AT-cut quartz crystal [260]. ∆D is defined as the ratio of the loss of mechanical energy per
oscillation period of the sensor crystal to the total mechanical energy input to the system. The
QCM-D measurements provide an assessment of the changes in mass (from ∆ f ) and mechanical
properties (from ∆D) of a layer of biomolecules attached to the surface of the quartz crystal[45].
Because the acoustic signal diminishes exponentially with distance above the surface of the quartz
crystal oscillator on which the cells are deposited, the QCM-D probes primarily the basal area of
the cell monolayer [203, 261], which specifically relates to cell adhesion. This technique has been
shown to be non-invasive to mammalian cells when the amplitude of shear oscillation is kept under
1 nm [199]. We have recently demonstrated a linear correlation between the level of focal adhesions, quantified based on number and size of vinculin (the key component of focal adhesions), and
the ∆D -response of cells [68]. Because the level of focal adhesions is quantitatively related to the
strength of cell adhesion [262–265], the time-dependent ∆D response could serve as a real-time,
quantitative indicator of the strength of cell-substrate adhesion [193, 194].

134

sensor in QCM-D, which is capable of generating real-time, continuous and measurable signals,
will be used for evaluating the system output.

Δy(t)
y1 (t)
u(t)
Pw(s)

C(s)

+

y(t)
m

Fig.
2. The
control
model for
EGFR-mediated
adhesion.
Figure
8.1system
Diagram
of control
model
of pathwaycell
dynamics

The proposed system structure is illustrated in Figure 2. The signaling network that is known to

8.2

Model development

regulate the EGF-induced cell adhesion is designated as the controller (Pw(s)) of the system. The
We propose to use quantitative modeling to dissect the signaling pathways involved in the process

controller can convert the input signal u(t) (e.g., EGF stimulation) to the execution signal y1(t),
of the EGF-induced dynamic change of cell adhesion. The model will be constructed based on a

which is responsible for controlling the cell adhesion complex, the plant C(s) of the system.
system identification process, which is regularly employed in control system design to elucidate

Feeding the plant C(s) with a measurement signal m from the QCM-D will result in the dynamic
the unknown structures and parameters of some of the components in the system based on the

output y(t), a quantity that correlates with the time-dependent ∆D signal of the QCM-D
prior knowledge and the input output information of the system [266]. A nanomechanical sensor

measurement.
the cells
can alsoreal-time,
activelycontinuous
adjust to and
exogenous
stimulation
through
a
in QCM-D,Meanwhile,
which is capable
of generating
measurable
signals, will
be
self-imposed
adaption the
by system
sendingoutput.
the feedback signal Δy(t) to the controller.
used for evaluating
Themodel
proposed system structure is illustrated in Fig. 8.1. The signaling network that is known
The plant
to regulate the EGF-induced cell adhesion is designated as the controller Pw(s) of the system. The
controller can convert the input signal u(t) (e.g., EGF stimulation) to the execution signal y1 (t),

7

which is responsible for controlling the cell adhesion complex, the plant C(s) of the system. Feeding the plant C(s) with a measurement signal m from the QCM-D will result in the dynamic output
y(t), a quantity that correlates with the time-dependent ∆D signal of the QCM-D measurement.
Meanwhile, the cells can also actively adjust to exogenous stimulation through a self-imposed
adaption by sending the feedback signal ∆y(t) to the controller.

135

8.2.1

Plant model

As for the plant C(s), the adhesion complex of adherent cells can receive the input signal u2 from
the EGF-activated signaling pathways. In response, the adhesion complex will undergo restructuring through dynamic assembly or disassembly of focal adhesions, resulting in the dynamic change
in cell adhesion. This will be assessed based on the output of the system. The correlation between
the level of cell adhesion and the ∆D response has been established in our previous study [68] .
The cell layer attached to a QCM-D sensor can be modeled as a sensor disk anchored to a spring
with an initial momentum [190] (Fig. 7.1). The dynamic equation for the movement of this sensor
disk is defined as:

MY¨ + η Y˙ + f (u2 ) + KY = 0

(8.1)

where M is the mass of the disk, η is the damping coefficient, Y is the index of the disk horizontal motion, K is the spring constant, and f is the mapping coefficient from the focal adhesion to the
sum of bond forces between the sensor disk and the cell adhesion complex. Therefore, represents
damping by the trapping fluid underneath of the cell body and the interactive bond force is defined
by the mapping f (u2 ). KY is the force in the spring and is the force resulting from the acceleration
of the sensor disk. This equation is derived based on the assumption that the friction between the
cell and the sensor disk and the viscous damping between the cell body and the liquid trapped underneath the cell body are the primary causes of the energy loss of the sensor disk. This assumption
can be verified by comparing the sum of the friction and the viscous damping to the quantity of
energy dissipation measured with the QCM-D. The system dynamics can be transformed to a state
space model as:

136

  

  
0
1  Y1
˙
   0 
Y1  


=
u
  
 +
f
η
K
Y2
Y˙2
−M
−M −M

(8.2)

with the two state variables Y1 and Y2 denoting the position and velocity of the oscillating
sensor disk. These two states specify the input of the QCM-D measurement signal m = [Y1 ,Y2 ]T .
The output from the mechanical model is defined as the rate of dissipation of kinetic energy:

Y=

d 1
( MY2 2 )
dt 2

(8.3)

which is quantitatively related to the ∆D response measured with the QCM-D.
The detailed analysis of the plant model yields an exponential relationship between the input
for the plant, the number of integrin-ECM bonds, and the output of the plant, the energy dissipation
factor, ∆D as shown in Fig. 8.2. The fitted curve shows a relationship that is defined by equation:

−

D = 0.0016(1 − e

B
3.5×109

)

(8.4)

which would then be converted to a transfer function of:

c(s) =

1
625s(3.5 × 109 s + 1)

(8.5)

The detail of the characterization process can be found in Chapter 7. Briefly, the mechanical
oscillation by the sensor disk would be damped by the interaction of the molecular bonds between
the cell and the disk. The process of bond friction relies on breaking of bonds to dissipate the
kinetic energy of the disk, and this process is a stochastic process defined by Bell’s model [60],
which in the overall control architecture determines the bond number conversion to the friction
137

c( s ) 

1
.
625s (3  109 s  1)

2.5

D
Exp fit of D

2.0

-3

D (x10 )

1.5
1.0
0.5
0.0
0

2
4
6
8
9
Bond number (x10 )

10

Fig.4.exponential
The exponential
relationship ofof
thethe
inputinput
and output
the plant
Figure 8.2 The
relationship
and for
output
formodel.
the plant model.

force or f (u2 ). Plug the friction force to the second order ODE would result in the relationship
9
between the number of bonds u2 and the output Y which is the energy dissipation factor, the
quantity that measured by QCM-D in the real-time experiment.

8.2.2

The controller model

We define the controller as the EGF-mediated signaling network that regulates cell adhesion (i.e.,
restructuring of focal adhesions). This controller includes the PLC pathway (Pwa ), the PI3K pathway (Pwb ), and the MAPK/ERK pathway (Pwc ). Some of the signaling molecules that are involved
in EGFR activation and its downstream cascades will not be included in the initial model, but will
be considered during the model refinement [236, 237]. The state variable of Pw is designated as
the concentration of a specific downstream effector molecule that is directly involved in regulation of restructuring of focal adhesions (e.g., calpain in the MAPK/ERK pathway). The change in
concentration of a species involved in a signaling reaction is usually described with an ODE. For
example, when EGF binds to the EGF receptor on the membrane of the cell: [EGF] + [FreeEGFreceptor]

[Bound-EGF-Receptor] (BEGFR), and the reaction is regarded as a mass reaction,

138

thus the time-dependent change in concentration of BEGFR can be described mathematically by
an ODE:
d[BEGFR]
= krbEGF [EGFR][EGF] − kruEGF [BEGFR]
dt

(8.6)

where [EGFR] is the concentration of EGF receptors on the cell membrane, [EGF] is the concentration of EGF and is the overall input of the system u, krbEGF and kruEGF are the forward and reverse
reaction rate constants (binding/unbinding of EGF to the free EGF receptor). The concentration
of [EGFR] can then be determined based on these parameters, which are available from previous
biochemical, biophysical, and modeling studies [236, 243, 267–269]]. For a Michaelis-Menten
reaction, like the Rac molecule activates Rho, the process can be defined by:

[InActiveRho]
d[ActiveRho]
= +kRho [ActiveRac]
dt
[InactiveRho] + km Rho

(8.7)

where [ActiveRho] and [InActiveRho] are the concentrations of active and inactive Rho, and
[ActiveRac] is the concentration of active Rac; kRho is the rate constant for the reaction and kkmRho
is the Michaelis constant for the Michaelis-Menten reaction.
To determine the concentration of a downstream effector may require solving a series of ODEs
derived from sequential signaling reactions upstream in the pathway. The detailed system consists
of 14 molecules that primarily determine the output of the effector molecule for cell adhesion
regulation. From the 14 molecules, we employed 26 state variables based on their active and
inactive status (the list of variables is shown in Tab. 8.1). The structure is shown in Fig. 8.3 on
which the group of ODEs was based upon.
If the concentration or the rate constant of a particular species is not readily available, an
estimate based on the information on a related species may be used for the calibration of the
initial model. Such estimated parameters can be determined eventually through model fitting of
139

Table 8.1 State variables and ODEs

Variables
X1
X2
X3
X4

Physical Meaning
[EGF]
[Free EGF Rec]
[Bound EGF Rec]
[Active Ras]

X5

[InActive Ras]

X6

[Active Raf]

X7

[InActive Raf]

X8
X9
X10
X11
X12
X13
X14

[Active PI3K]
[InActive PI3K]
[Active Akt]
[InActive Akt]
[Active Rac]
[InActive Rac]
[Active Rho]

X15

[InActive Rho]

X16
X17
X18
X19
X20
X21

[Active MEK1/2]
[InActive MEK1/2]
[Active Erk1/2]
[InActive Erk1/2]
[Active PP2A]
[Active PLCγ]

X22

[InActive PLCγ]

X23
X24
X25

[Active PIP2]
[InActive PIP2]
[Active Cofilin]

X26

[InActive Cofilin]

ODEs
dX1
dt = −krbEGF X1 X2 + kruEGF X3
dX2
dt = −krbEGF X1 X2 + kruEGF X3
dX3
dt = +krbEGF X1 X2 − kruEGF X3
X5
dX4
X4
dt = +kRas X3 X5 +kmRas − kERKRas X18 X4 +kmERKRas

dX5
X5
X4
dt = −kRas X3 X5 +kmRas + kERKRas X18 X4 +kmERKRas
dX6
X6
X7
dt = +kRa f X4 X7 +kmRa f − kAktRa f X10 X6 +mAktRa f
X6
dX7
X7
dt = −kRa f X4 X7 +kmRa f + kAktRa f X10 X6 +mAktRa f
X9
X9
dX8
dt = +kPI3K X3 X9 +kmPI3K + kRasPI3K X4 X9 +kmRasPI3K
dX9
X9
X9
dt = −kPI3K X3 X9 +kmPI3K − kRasPI3K X4 X9 +kmRasPI3K
dX10
X11
dt = +kAkt X8 X11 +kmAkt
dX11
X11
dt = −kAkt X8 X11 +kmAkt
X13
dX12
dt = +kRac X8 X13 +kmRac
dX13
X13
dt = −kRac X8 X13 +kmRac
X15
dX14
dt = +kRho X1 2 X15 +kmRho
X15
dX15
dt = −kRho X1 2 X15 +kmRho
dX16
X16
X17
dt = +kMEK X6 X17 +kmMEK − kdPP2AMEK X20 X16 +kmPP2AMEK
X16
X17
dX17
dt = −kMEK X6 X17 +kmMEK + kdPP2AMEK X20 X16 +kmPP2AMEK
X19
dX18
X18
dt = +kERK X1 6 X19 +kmERK − kdPP2AERK X20 X18 +kmPP2AERK
dX19
X19
X18
dt = −kERK X1 6 X19 +kmERK + kdPP2AERK X20 X18 +kmPP2AERK
dX20
dt = 0
X22
dX21
dt = +kPLCγ X3 X22 +kmPLCγ
dX22
X22
dt = −kPLCγ X3 X22 +kmPLCγ
dX23
X24
dt = +kPIP2 X21 X24 +kmPIP2
X24
dX24
dt = −kPIP2 X21 X24 +kmPIP2
dX25
X26
dt = +kCo f ilin X23 X26 +kmCo f ilin
dX26
X26
dt = −kCo f ilin X23 X26 +kmCo f ilin

140

regulation. From the 14 molecules, we employed 26 state variables based on their active and
inactive status (the list of variables is shown in Table 1). The structure is shown in Figure 5 on
which the group of ODEs was based upon.

Figure
systemofofmolecules
molecules
ODEs
Fig.8.3
5. The
The system
thatthat
ODEs
werewere
builtbuilt
upon.upon.

11 the experimental data. The rate constants for the 26 ODEs are listed in Tab. 8.2. A combination of
equations can easily be built for a more specific and complicated signaling network based on the
registered items in the database. The overall signal output from the entire signaling network will
be represented as a weighted linear combination of three pathways:

y1 = αPwa + β Pwb + γPwc

where α, β and γ are mapping parameters.

141

(8.8)

time energy dissipation factor data were then fitted. The data would then be tested when one or
several of the pathways were to be blocked.

10

D

0
0nM
0nM fit
1nM
1nM fit
5nM
5nM fit
10nM
10nM fit

-10
-20
-30

0 20 40 60 80 100 120 140 160 180
Time (Min)

Fig.fitting
6. The fitting
EGFtreatment
treatment of of
A431
cells cells
based on
the control
model.
Figure 8.4 The
resultresult
forfor
EGF
A431
based
on the
control model.

8.312 Result
The parameter for the control model Pw(s) was obtained mostly from the paper by Brown et
al. [268]. For the unknown parameters, we used a fitting algorithm to fit the experimental data with
the model and obtain an optimal parameter set. For the EGF treatment of the A431 cells, the real
time energy dissipation factor data were then fitted. The data would then be tested when one or
several of the pathways were to be blocked.
The real-time measurement of energy dissipation variations is compared against the simulation
data from the model in Fig. 8.4. Based on the parameters from literature, the model fits the experimental data with precision for concentrations of 1, 5 and 10 nM. The dynamics model shows
discrepancies with the experimental data only at the first phase of the cell adhesion transition
(Phase I) when the cell deadhesion dominates the process with drastic restructuring. Features in
Phase II and III with stable configuration and gradual recovery of cell adhesion are well captured.
Of equal importance, the expression of these key molecules in the pathway can be analyzed. It
has been reported [68] that the three branches of pathway cascade induce different outcomes of cell

142

adhesion. Specifically, the PLC pathway strengthens the focal adhesion complex and promotes the
assembly of focal adhesion while inhibits the disassembly of cell adhesion. On the contrary, both
the MAPK pathway and the PI3K pathway destabilize the focal adhesion structure and compel the
disassembly of the focal adhesion. As the cell experience the rise and fall of the cell adhesion both
in size and strength as indicated by the energy dissipation and fluorescence labeling, the expression
of these key molecules also fluctuates. The model could also be tested by blocking certain branches
of the pathway and monitor the response.
As shown in Fig. 8.5, the expression levels of Rho, MEK and PLCγ over the course of preservation are plotted. Both the levels of Rho and MEK went up in Phase I, though the temporal
dynamics of these two molecules did not overlap. The upregulation of MEK experienced a more
drastic boost during the first 20 minutes of EGF stimulation and reached plateau 15 minutes into
the observation; while the deregulation had a similar temporal dynamics. The expression level of
Rho increased at a pace shared with the overall deadhesion. It plateaued at around the end of Phase
I and the inception of Phase II, and subsequently dropped at a slower rate than MEK. Since both
pathways are responsible for the disassembly of focal adhesion, the increase of both would shadow
or at least neutralize the promoting effect of the PLC pathway. Additionally, the downregulation
of PLCγ begins slowly after receiving EGF signal, which exacerbates the disassembly of the focal
adhesion. After Phase II, however, the combined effect of both decreased levels of MEK and Rho,
plus the upswing of the PLCγ would essentially bring back the focal adhesions that were previously disrupted. A concerted effort of these key molecules in the pathway dynamics accomplished
the delicate task of cell adhesion regulation.
The overall effect of the molecule dynamics after EGF stimulation is manifested through the
biophysical properties of the focal adhesions as well as the cytoskeleton mechanics, as the realtime measurement data from QCM-D tracks the change of energy dissipation that are directly
143

1 5 .0

M E K 1 /2
R h o
P L C γ

1 2 .5
1 0 .0
7 .5
5 .0
2 .5
0 .0
0

2 0

4 0

6 0

8 0

1 0 0

1 2 0

1 4 0

1 6 0

1 8 0

T im e ( M in )

Figure 8.5 Expression levels of MEK1/2, Rho and PLCγ

related to the size and strength of focal adhesion. Leveraging this functional study, we were able
to provide a multifaceted characterization of the EGF induced cell adhesion change. The analysis
from the signaling pathway control model yields a unique peek into the biochemical dynamics
of the regulation as well as the biomechanical dynamics of the adhesion. The simulation result
compares well with the realtime measurement data. More importantly, the simulation provides
the molecule dynamics in the process of regulation which can be verified by biochemical assays,
such as Western blot. At the minimum, from the biochemical assay, we can build a calibration
standard for each key molecule over the time course and we’ll prevent the messy experiments from
repeating. If the prediction from the model proves correct, we will find similar levels of protein
expression during the EGF stimulation process.

144

8.4

Chapter summary

This chapter proposed a signaling pathway control model that aims to decipher intricate signaling
pathways that regulate cell adhesion during EGF stimulation. The model leverages on the capabilities of nanomechanical senor in QCM-D which provides real time measurement of cell adhesion
strength. A mechanical model was built to quantitatively relate the measurement value, energy dissipation (∆D) with the dynamic bond formation/dissociation. This mechanical model was considered the plant controlled by the signaling pathway model. Upon obtaining the measurement output
in ∆D and with prior knowledge of EGF concentration, we identified the key steps and branches of
signaling pathways in the regulation process, or more specifically the dynamics of key molecules.
The study initiates a unique methodology in delineating pathway dynamics that essentially eliminates the necessity of the messy, time-consuming biochemical assays, at least significantly reduces
the amount of them, which will shed insight on the system biologists’ approaches, as well as be of
importance to the the modeling and control community. The result of the study will advance the
understanding of the EGF regulated cell adhesion and benefit the development of therapeutics in
cancer treatment with overexpression of EGFR.

145

Table 8.2 Rate constants
Reaction name
EGF binding
Ras activation by bound
EGFR

Reaction
EGF+
FreeEGFRec
⇔
BoundEGFRec
RasInactive → RasActive

Ras deactivation by ERK1/2

RasActive → RasInactive

Raf activation by Ras

RafInactive → RafActive

Raf deactivation by Akt

RafActive →RafInactive

PI3k activation by bound
EGFR

PI3kActive → PI3kInactive

PI3K activation by Ras

PI3kActive→ PI3kInActive

Akt activation by PI3k

AktActive→ AktInactive

Rac activation by PI3k

RacActive → RacInactive

Rho activation by Rac

RhoActive → RhoInactive;

MEK1/2 activation by Raf

MEK1/2Active
MEK1/2Inactive;

MEK1/2
PP2A
ERK1/2
MEK1/2
ERK1/2
PP2A

deactivation

activation

deactivation

by

by

by

PLCγ activation by bound
EGFR
PIP2 activation by PLCγ

MEK1/2Active
MEK1/2Inactive
ERK1/2Active
ERK1/2Inactive
ERK1/2Active
ERK1/2Inactive
PLCγActive
PLCγInactive;

→

Value
2.185E-5

kRas

694.731

kmRas
kERKRas
kmERKRas
kRa f
kmRa f
kAktRa f
kmAktRa f
kPI3K

6086070
1611.97
896896
1509.36
1432410
15.1212
119355
10.6737

kmPI3K
kRasPI3K
kmRasPI3K
kAkt
kmAkt
kRac
kmRac
kRho
kmRho
kMEK1/2

184912
0.0771067
272056
0.0566279
653951
Est.
Est.
Est.
Est.
185.759

kmMEK1/2
→

kPP2AMEK1/2

4768350
2.83243

→

kmPP2AMEK1/2
kERK1/2

518753
9.85367

kmERK1/2
→

kPP2AERK1/2

1007340
8.8912

→

kmPP2AERK1/2
kPLC

3496490
Est.

kmPLCg
kPIP2
kmPIP2

Est.
Est.
Est.

PIP2Active → PIP2Inactive

146

Parameter
krbEGF

Chapter 9
Conclusion and future work

9.1
9.1.1

Conclusion
Development of AFM nanorobotics for biomedical applications

The use of AFM as a robotic platform in changing materials especially biological materials at
their native states has transformed the biomedical research significantly in numerous aspects. The
unprecedented level at which AFM can visualize biological samples combined with the dexterity
of the robotic control schemes makes the operation at the cellular level a possibility. With the
AFM nanorobotic platform, cellular surgeries at nanometer precision were carried out to mimic
the effect of antibody treatment; while the flexibility of a haptic device provide a convenient and
efficient means for capturing the mechanical property changes at a survey mode.
The development of the platform requires the precise modeling of the interaction of the AFM
end-effector, or cantilever with the biological samples, especially live cells, with complex material
properties, and the control of the positioning of the end-effector monitors the interaction force and
feedback to the operator. Living cells normally display a high level of viscosity, therefore, the
operation frequency of the AFM manipulation should be high enough to provide the extra rounds
of position and force feedback. We have developed a number of control schemes to increase the
frequency so that the operation could be performed smoothly.

147

9.1.2

Dynamical monitoring of cellular structural and mechanical changes
under patho/physiological conditions

The mechanical profile of cells gives out a wide spectrum of information regarding on their status,
not only biophysical status, but biochemical status as well. The real-time observation and monitoring of these properties provides us a unique peek into the complex dynamics of the cell structural
and the change of their biological behavior. We have demonstrated this concept by different sets
of experiments on a number of biological stimuli-response processes.
The experiment with the insulinoma cells showed that the ionic response from the cell after
stimulation from different chemical species can be revealed and delineated by a continuous monitoring of the subtle changes of the membrane heights in the order of a few tens of nanometers,
and the change of cellular elasticity. The structural and mechanical alterations upon stimulation
of glucose and capsaicin tracked by the AFM nanorobotics platform echoed the response from the
ion channels, or the change of membrane capacitance. Thus the structure and mechanical profiles
give us a signature of the ion channel activities of the cell, and we were able to establish a causal
relationship of a single-phase biological process.
The experiment with the HaCaT cells provided us some insights in the cellular response when
the cellular structures, specifically the adhesion structure, were damaged. We found that the cellular stiffness experienced a two-phase response when the keratinocytes were treated with an autoimmune antibody found in patients with Pemphigus Vulgaris. The stiffness dropped in the first
hour of treatment, which was followed by an increase in the ensuing four hours of observation.
We postulate that the loss of adhesion was directly caused by the loss of cellular adhesion between
neighboring cells. The cellular adhesions as we consider are tensional anchoring sites to provide
the mechanical support and maintain the integrity of the cell structures, the loss of which will result

148

in the decrease in the pre-stress, which determines the stiffness in a tensegrity cellular structure,
and ultimately resulted in the reduced stiffness. We conclude that the increase of stiffness in the
second phase can find its roots in the initiation of the apoptosis pathway, which leads to the increased density of the filamentous structures of the cell, and thus stiffness. These hypothesis were
verified by a series of experiments, and one of the important piece is to mimic the loss of adhesion
by nanorobotic surgery.

9.1.3

Nanomechanical characterization and nanorobotic surgery delineate
cell signaling pathway

The nanorobotic surgery severs the ties of the intermediate filaments between neighboring keratinocytes. The dissection processes were controlled at nanometer precision. We were able to
monitor the snap of bundles of intermediate filaments by looking at the profile of the lateral force
from the cantilever deflection during the nanorobotic surgery. The filamentous structures were
imaged, sectioned and compared before and after surgery. We could confirm that the snap of the
bundles were due to the severing force of the end-effector.
The repeated surgery operation could ultimately isolate a single cell from its neighbors, a physiological states similar with the antibody treatment, or non-formation of cellular adhesions. The
isolation of cells, or the damage of the cell adhesion led to the decreased cellular stiffness, just as
we expected. It verifies our conjecture that the cell behaves as a tensegrity structure, any single
change at its cytoskeleton will trigger its active rearrangement globally. More importantly, the
surgery operations that isolated cells did not lead to any increase of cellular stiffness. That, as we
conclude should be a direct result of antibody-triggered, well-synchronized signaling pathways,
apoptosis pathway in this case. Thus we set out to see whether we could find a quantitative method

149

of looking at signaling pathways from the mechanical measurements we took using the integrated
platform, thus the transition from mechanophenotying to mechanoproteomics.

9.1.4

Cell adhesion model with nanomechanical oscillator

To take a closer look at the cell adhesion at the basal area, we introduced into out platform a
nanomechanical oscillator to detect the change of the cell-ECM adhesion strength during cellular
events. The oscillator monitors the mechanical response from the bottom of the cell, while the
AFM nanorobot takes measurement on the top. The result of the experiment showed a well synchronized profile from both measurement data. The A431 cell responded in a synchronized fashion
temporally after the stimulation of EGF: the increased stiffness and increased hysteritivity from the
AFM data, and the decreased energy dissipation from the oscillator. The profile pointed to us that
the cell went through a transition that requires it to reduce its grasp onto the substrate underneath,
while at the same time, direct its cytoskeleton to transit towards a more liquid like states.
A mechanical model of the cell adhesion structure interacting with the oscillating sensor was
developed, with implications to the cytoskeleton. The model qualitatively evaluates the response
from both data and analyzed the change of cell-ECM adhesion and the effect of this change on the
actin filaments. We conclude that the reduced cell-ECM strength was due to the decreased number
of integrin based bonds, integrin-RGD. The decreased force equilibrium led to the reduced energy
dissipation level at each cycle of the oscillation. The loss of integrin bonds eventually led to the
decreased cluster strength at the focal adhesion sites and thus broke the old force equilibrium
between the focal adhesion and the actin filaments and established a new equilibrium at a lower
force in the filaments. Quantitatively, the model provides the input and output relationship of the
oscillation sensor to convert the molecular levels to the energy dissipation readouts that can be
deduced for the comprehensive model of signaling pathway.
150

9.1.5

A systematic control model for cell signaling pathway

We think that these transition of cellular structure must have been orchestrated by the same EGF
triggered pathway. Thus we took a systematic approach in the dissection of the signaling pathway.
It employs the common concept of system identification process in a control practice. We took the
input and output information and deduced the regulators in the middle. In our case, the input is the
biochemical specie of EGF and the output is the mechanical readout of energy dissipation, a set of
realtime quantitative data captured dynamically during the cellular transition process.
We consider the complex signaling network as the controller that regulates the mechanical behavior, or mechanical interaction of cells with the sensing elements. The pathway was modeled by
the traditional differential equation approach and the plant is the mechanical model of the interaction between the adhesion bonds and the oscillator. We simulated a range of inputs from the EGF
and look at the best fit with the experimental data to identify the key parameters in the signaling
pathway. At these parameters, we looked at the levels of a few important regulating proteins, the
rise and fall of levels of these molecules were identified and verified in literature during similar
biological processes.

9.2
9.2.1

Future work
Signal transduction and propagation study

A key aspect of the signaling transduction is the intra- and intercellular signal propagation. For intracellular propagation, local concentrated EGF would cause dimmerization of EGFR, a signature
reaction due to activation which would lead to downstream signaling events. Besides, this would
trigger reactions from neighboring receptors, and thus a wave of dimerization would propagate

151

throughout the entire cell (Fig. 9.1A1,2). A study can be carried out to see whether the pathway
can be triggered locally as by a AFM nanorobotic delivery of the EGF molecular to a specific
location on the cell membrane (Fig. 9.1A3). We can also design experiments to test the signaling propagation through the physical contacts between neighboring cells. The study will tell us
whether one stimulated cell will lead to the response from its neighbors, and whether the signal
can be passed on to a large group of cells to provoke a synchronized response (Fig. 9.1B1,2).

A

EGF

EGFR
A1

A2

A3

B

B1

B2

Figure 9.1 EGF stimulated intra and intercellular signaling propagation: A1, no activation; A2,
global activation; A3, Local activation. B1, Local stimulation; B2, Intercellular propagation

9.2.2

Functional model of the pathway based on mechanical profile

A function model can be built to take the mechanical profile of any nature, elasticity, viscosity,
energy dissipation, surface roughness and so on. The mechanical profile will work as a unified
marker of physio/pathological events. An extension of the current signaling pathway model can be
developed to take advantage of the unified maker and derive the molecular expressions of cellular
events.

152

BIBLIOGRAPHY

153

BIBLIOGRAPHY
[1] Daniel J M¨uller and Yves F Dufrene. Atomic force microscopy as a multifunctional molecular toolbox in nanobiotechnology. Nature nanotechnology, 3(5):261–269, 2008.
[2] JKH H¨orber and MJ Miles. Scanning probe evolution in biology. Science, 302(5647):1002–
1005, 2003.
[3] Gerd Binnig, Calvin F Quate, and Ch Gerber. Atomic force microscope. Physical review
letters, 56(9):930–933, 1986.
[4] Vladimir A Ukraintsev, Christopher Baum, Gary Zhang, and Craig L Hall. The role of
afm in semiconductor technology development: the 65 nm technology node and beyond. In
Proc. SPIE, volume 5752, pages 127–139, 2005.
[5] Todd D Krauss and Louis E Brus. Charge, polarizability, and photoionization of single
semiconductor nanocrystals. Physical Review Letters, 83(23):4840–4843, 1999.
[6] Andre K Geim and Konstantin S Novoselov. The rise of graphene. Nature materials,
6(3):183–191, 2007.
[7] King Wai Chiu Lai, Carmen Kar Man Fung, Hongzhi Chen, Ruiguo Yang, Bo Song, and
Ning Xi. Manipulation and assembly methods for graphene based nano devices. In Nanotechnology (IEEE-NANO), 2010 10th IEEE Conference on, pages 623–626. IEEE, 2010.
[8] Guangyong Li, Ning Xi, and Donna H Wang. In situ sensing and manipulation of molecules
in biological samples using a nanorobotic system. Nanomedicine: nanotechnology, biology,
and medicine, 1(1):31, 2005.
[9] Guangyong Li, Ning Xi, and Donna H Wang. Probing membrane proteins using atomic
force microscopy. Journal of cellular biochemistry, 97(6):1191–1197, 2006.
[10] Carmen Kar Man Fung, Kristina Seiffert-Sinha, King Wai Chiu Lai, Ruiguo Yang, Dan Panyard, Jiangbo Zhang, Ning Xi, and Animesh A Sinha. Investigation of human keratinocyte
cell adhesion using atomic force microscopy. Nanomedicine: Nanotechnology, Biology and
Medicine, 6(1):191–200, 2010.

154

[11] Ruiguo Yang, Ning Xi, Carmen Kar, Man Fung, King Wai, Chiu Lai, K Seiffert-Sinha, and
AA Sinha. Analysis of keratinocytes stiffness after desmosome disruption using atomic
force microscopy based nanomanipulation. In Nanotechnology, 2009. IEEE-NANO 2009.
9th IEEE Conference on, pages 640–643. IEEE, 2009.
[12] Ruiguo Yang, Ning Xi, Carmen Kar Man Fung, King Wai Chiu Lai, Kristina Seiffert-Sinha,
Animesh A Sinha, and Weijing Zhang. Investigations of bio markers for human lymphoblastoid cells using atomic force microscopy. In Nano/Micro Engineered and Molecular Systems
(NEMS), 2010 5th IEEE International Conference on, pages 129–132. IEEE, 2010.
[13] Carmen Kar Man Fung, Ruiguo Yang, King Wai Chiu Lai, Ning Xi, K Seiffert-Sinha, and
AA Sinha. Micro fixture enabled in-situ imaging and manipulation of cell membrane protein. In Nano/Micro Engineered and Molecular Systems (NEMS), 2010 5th IEEE International Conference on, pages 165–168. IEEE, 2010.
[14] Daniel J Muller. Afm: A nanotool in membrane biology. Biochemistry, 47(31):7986–7998,
2008.
[15] Guangyong Li, Ning Xi, Lianqing Liu, Jiangbo Zhang, and K Lai. Study of dna properties under controlled conditions using afm based nano-robotics. In Nanotechnology, 2007.
IEEE-NANO 2007. 7th IEEE Conference on, pages 1018–1021. IEEE, 2007.
[16] Andreas Engel and Daniel J M¨uller. Observing single biomolecules at work with the atomic
force microscope. Nature Structural & Molecular Biology, 7(9):715–718, 2000.
[17] King Wai Chui Lai, Ning Xi, Carmen Kar Man Fung, Jiangbo Zhang, Hongzhi Chen, Yilun
Luo, and Uchechukwu C Wejinya. Automated nanomanufacturing system to assemble carbon nanotube based devices. The International Journal of Robotics Research, 28(4):523–
536, 2009.
[18] S´ebastien Decossas, Fr´ed´eric Mazen, Thierry Baron, Georges Br´emond, and Abdelkader
Souifi. Atomic force microscopy nanomanipulation of silicon nanocrystals for nanodevice
fabrication. Nanotechnology, 14(12):1272, 2003.
[19] Ning Xi, CKM Fung, Ruiguo Yang, Kristina Seiffert-Sinha, King Wai Chiu Lai, and Animesh A Sinha. Bionanomanipulation using atomic force microscopy. Nanotechnology Magazine, IEEE, 4(1):9–12, 2010.
[20] Harald Herrmann, Harald B¨ar, Laurent Kreplak, Sergei V Strelkov, and Ueli Aebi. Intermediate filaments: from cell architecture to nanomechanics. Nature Reviews Molecular Cell
Biology, 8(7):562–573, 2007.

155

[21] JD Beard, DJ Burbridge, AV Moskalenko, O Dudko, Polina L Yarova, SV Smirnov, and
SN Gordeev. An atomic force microscope nanoscalpel for nanolithography and biological
applications. Nanotechnology, 20(44):445302, 2009.
[22] Matthias Rief, Filipp Oesterhelt, Berthold Heymann, and Hermann E Gaub. Single
molecule force spectroscopy on polysaccharides by atomic force microscopy. Science,
275(5304):1295–1297, 1997.
[23] Keir C Neuman and Attila Nagy. Single-molecule force spectroscopy: optical tweezers,
magnetic tweezers and atomic force microscopy. Nature methods, 5(6):491–505, 2008.
[24] Andrew E Pelling, Farlan S Veraitch, Carol Pui-Kei Chu, Chris Mason, and Michael A
Horton. Mechanical dynamics of single cells during early apoptosis. Cell motility and the
cytoskeleton, 66(7):409–422, 2009.
[25] Dimitrios Fotiadis, Simon Scheuring, Shirley A M¨uller, Andreas Engel, and Daniel J M¨uller.
Imaging and manipulation of biological structures with the afm. Micron, 33(4):385–397,
2002.
[26] PK Hansma, JP Cleveland, M Radmacher, DA Walters, PE Hillner, M Bezanilla, M Fritz,
D Vie, HG Hansma, CB Prater, et al. Tapping mode atomic force microscopy in liquids.
Applied Physics Letters, 64(13):1738–1740, 1994.
[27] Simon Scheuring and Yves F Dufrˆene. Atomic force microscopy: probing the spatial organization, interactions and elasticity of microbial cell envelopes at molecular resolution.
Molecular microbiology, 75(6):1327–1336, 2010.
[28] QS Li, GYH Lee, CN Ong, and CT Lim. Afm indentation study of breast cancer cells.
Biochemical and biophysical research communications, 374(4):609–613, 2008.
[29] Martin Stolz, Riccardo Gottardi, Roberto Raiteri, Sylvie Miot, Ivan Martin, Rapha¨el Imer,
Urs Staufer, Aurelia Raducanu, Marcel D¨uggelin, Werner Baschong, et al. Early detection of aging cartilage and osteoarthritis in mice and patient samples using atomic force
microscopy. Nature nanotechnology, 4(3):186–192, 2009.
[30] Andrew Hards, Chunqing Zhou, Markus Seitz, Christoph Br¨auchle, and Andreas Zumbusch. Simultaneous afm manipulation and fluorescence imaging of single dna strands.
ChemPhysChem, 6(3):534–540, 2005.
[31] Chiara Callies, Peter Sch¨on, Ivan Liashkovich, Christian Stock, Kristina Kusche-Vihrog, Johannes Fels, Alexandra S Str¨ater, and Hans Oberleithner. Simultaneous mechanical stiffness
156

and electrical potential measurements of living vascular endothelial cells using combined
atomic force and epifluorescence microscopy. Nanotechnology, 20(17):175104, 2009.
[32] Christoph Riethm¨uller, Tilman E Sch¨affer, Ferry Kienberger, Werner Stracke, and Hans
Oberleithner. Vacuolar structures can be identified by afm elasticity mapping. Ultramicroscopy, 107(10):895–901, 2007.
[33] Michael J Rust, Mark Bates, and Xiaowei Zhuang. Stochastic optical reconstruction
microscopy (storm) provides sub-diffraction-limit image resolution. Nature methods,
3(10):793, 2006.
[34] Susana Moreno Flores and Jos´e L Toca-Herrera. The new future of scanning probe microscopy: Combining atomic force microscopy with other surface-sensitive techniques, optical microscopy and fluorescence techniques. Nanoscale, 1(1):40–49, 2009.
[35] Gennaro Picardi, Marc Chaigneau, Razvigor Ossikovski, Christophe Licitra, and Guillame
Delapierre. Tip enhanced raman spectroscopy on azobenzene thiol self-assembled monolayers on au (111). Journal of Raman Spectroscopy, 40(10):1407–1412, 2009.
[36] Philip S Waggoner and Harold G Craighead. Micro-and nanomechanical sensors for environmental, chemical, and biological detection. Lab on a Chip, 7(10):1238–1255, 2007.
[37] Martino Poggio. Nanomechanics: Sensing from the bottom up. Nature nanotechnology,
8(7):482–483, 2013.
[38] Gajendra S Shekhawat and Vinayak P Dravid. Nanomechanical sensors: Bent on detecting
cancer. Nature nanotechnology, 8(2):77–78, 2013.
[39] Hans Peter Lang, Martin Hegner, and Christoph Gerber. Cantilever array sensors. Materials
today, 8(4):30–36, 2005.
[40] F Huber, HP Lang, N Backmann, D Rimoldi, and Ch Gerber. Direct detection of a braf
mutation in total rna from melanoma cells using cantilever arrays. Nature nanotechnology,
8(2):125–129, 2013.
[41] G Longo, L Alonso-Sarduy, L Marques Rio, A Bizzini, A Trampuz, J Notz, G Dietler,
and S Kasas. Rapid detection of bacterial resistance to antibiotics using afm cantilevers as
nanomechanical sensors. Nature nanotechnology, 8(7):522–526, 2013.
´
[42] Montserrat Calleja, Priscila M Kosaka, Alvaro
San Paulo, and Javier Tamayo. Challenges
for nanomechanical sensors in biological detection. Nanoscale, 4(16):4925–4938, 2012.
157

[43] Christian Franck, Stacey A Maskarinec, David A Tirrell, and Guruswami Ravichandran.
Three-dimensional traction force microscopy: a new tool for quantifying cell-matrix interactions. PLoS One, 6(3):e17833, 2011.
[44] Venkat Maruthamuthu, Benedikt Sabass, Ulrich S Schwarz, and Margaret L Gardel. Cellecm traction force modulates endogenous tension at cell–cell contacts. Proceedings of the
National Academy of Sciences, 108(12):4708–4713, 2011.
[45] Jianping Fu, Yang-Kao Wang, Michael T Yang, Ravi A Desai, Xiang Yu, Zhijun Liu, and
Christopher S Chen. Mechanical regulation of cell function with geometrically modulated
elastomeric substrates. Nature methods, 7(9):733–736, 2010.
[46] Adam J Engler, Shamik Sen, H Lee Sweeney, and Dennis E Discher. Matrix elasticity
directs stem cell lineage specification. Cell, 126(4):677–689, 2006.
[47] Fredrik H¨oo¨ k and Bengt Kasemo. The qcm-d technique for probing biomacromolecular
recognition reactions. In Piezoelectric Sensors, pages 425–447. Springer, 2007.
[48] Malin Edvardsson, Michael Rodahl, and Fredrik H¨oo¨ k. Investigation of binding event perturbations caused by elevated qcm-d oscillation amplitude. Analyst, 131(7):822–828, 2006.
[49] Karin Glasm¨astar, Charlotte Larsson, Fredrik H¨oo¨ k, and Bengt Kasemo. Protein adsorption
on supported phospholipid bilayers. Journal of colloid and interface science, 246(1):40–47,
2002.
[50] Carine Galli Marxer, Martine Collaud Coen, Thomas Greber, Urs F Greber, and Louis
Schlapbach. Cell spreading on quartz crystal microbalance elicits positive frequency shifts
indicative of viscosity changes. Analytical and bioanalytical chemistry, 377(3):578–586,
2003.
[51] Marina V Voinova, M Rodahl, M Jonson, and B Kasemo. Viscoelastic acoustic response
of layered polymer films at fluid-solid interfaces: Continuum mechanics approach. Physica
Scripta, 59(5):391, 1999.
[52] Shannon M Notley, Malin Eriksson, and Lars W˚agberg. Visco-elastic and adhesive properties of adsorbed polyelectrolyte multilayers determined in situ with qcm-d and afm measurements. Journal of colloid and interface science, 292(1):29–37, 2005.
[53] L Kreplak, H B¨ar, JF Leterrier, H Herrmann, and U Aebi. Exploring the mechanical behavior of single intermediate filaments. Journal of molecular biology, 354(3):569–577, 2005.

158

[54] Aleksandr Noy. Handbook of molecular force spectroscopy. Springer, 2008.
[55] Graham Carpenter and Stanley Cohen. Epidermal growth factor. Annual review of biochemistry, 48(1):193–216, 1979.
[56] Sreenath V Sharma, Daphne W Bell, Jeffrey Settleman, and Daniel A Haber. Epidermal
growth factor receptor mutations in lung cancer. Nature Reviews Cancer, 7(3):169–181,
2007.
[57] Kanae Oda, Yukiko Matsuoka, Akira Funahashi, and Hiroaki Kitano. A comprehensive
pathway map of epidermal growth factor receptor signaling. Molecular systems biology,
1(1), 2005.
[58] Zhimin Lu, Guoqiang Jiang, Peter Blume-Jensen, and Tony Hunter. Epidermal growth
factor-induced tumor cell invasion and metastasis initiated by dephosphorylation and downregulation of focal adhesion kinase. Molecular and cellular biology, 21(12):4016–4031,
2001.
[59] Michele A Wozniak, Katarzyna Modzelewska, Lina Kwong, and Patricia J Keely. Focal
adhesion regulation of cell behavior. Biochimica et Biophysica Acta (BBA)-Molecular Cell
Research, 1692(2):103–119, 2004.
[60] George I Bell. Models for the specific adhesion of cells to cells. Science, 200(4342):618–
627, 1978.
[61] M Dembo, DC Torney, K Saxman, and D Hammer. The reaction-limited kinetics of
membrane-to-surface adhesion and detachment. Proceedings of the Royal Society of London. Series B. Biological Sciences, 234(1274):55–83, 1988.
[62] Salvatore Pece and J Silvio Gutkind. Signaling from e-cadherins to the mapk pathway by
the activation of epidermal growth factor receptors upon cell-cell contact formation. Journal
of Biological Chemistry, 275(52):41227–41233, 2000.
[63] Elizabeth J Joslin, Harish Shankaran, Lee K Opresko, Nikki Bollinger, Douglas A Lauffenburger, and H Steven Wiley. Structure of the egf receptor transactivation circuit integrates
multiple signals with cell context. Molecular BioSystems, 6(7):1293–1306, 2010.
[64] Melody K Morris, Julio Saez-Rodriguez, David C Clarke, Peter K Sorger, and Douglas A
Lauffenburger. Training signaling pathway maps to biochemical data with constrained fuzzy
logic: quantitative analysis of liver cell responses to inflammatory stimuli. PLoS computational biology, 7(3):e1001099, 2011.
159

[65] Neda Bagheri, Marisa Shiina, Douglas A Lauffenburger, and W Michael Korn. A dynamical
systems model for combinatorial cancer therapy enhances oncolytic adenovirus efficacy by
mek-inhibition. PLoS computational biology, 7(2):e1001085, 2011.
[66] Achim Besser and Ulrich S Schwarz. Coupling biochemistry and mechanics in cell adhesion: a model for inhomogeneous stress fiber contraction. New Journal of Physics,
9(11):425, 2007.
[67] Ruiguo Yang, Jennifer Y Chen, Ning Xi, King Wai Chiu Lai, Chengeng Qu, Carmen
Kar Man Fung, Lynn S Penn, and Jun Xi. Characterization of mechanical behavior of
an epithelial monolayer in response to epidermal growth factor stimulation. Experimental
cell research, 318(5):521–526, 2012.
[68] Jennifer Y Chen, Ammar Shahid, Marcela P Garcia, Lynn S Penn, and Jun Xi. Dissipation
monitoring for assessing egf-induced changes of cell adhesion. Biosensors and Bioelectronics, 38(1):375–381, 2012.
[69] Toshio Ando, Noriyuki Kodera, Daisuke Maruyama, Eisuke Takai, Kiwamu Saito, and Akitoshi Toda. A high-speed atomic force microscope for studying biological macromolecules
in action. Japanese Journal of Applied Physics, 41(part 1):4851–4856, 2002.
[70] H Xie, DS Haliyo, and S R´egnier. Parallel imaging/manipulation force microscopy. Applied
Physics Letters, 94(15):153106–153106, 2009.
[71] Yoomin Ahn, Takahito Ono, and Masayoshi Esashi. Micromachined si cantilever arrays
for parallel afm operation. Journal of Mechanical Science and Technology, 22(2):308–311,
2008.
[72] Atsushi Ikai. The world of nano-biomechanics: mechanical imaging and measurement by
atomic force microscopy. Access Online via Elsevier, 2007.
[73] H Hatze. The meaning of the term biomechanics. Journal of Biomechanics, 7(2):189–190,
1974.
[74] Michael Fitzgerald. Nanobiomechanics”. technology review. mit. retrieved february 23,
February 20011.
[75] Brian M Cooke, Narla Mohandas, and Ross L Coppel. The malaria-infected red blood cell:
structural and functional changes. Advances in parasitology, 50:1–86, 2001.
[76] Sarah E Cross, Yu-Sheng Jin, Jianyu Rao, and James K Gimzewski. Nanomechanical analysis of cells from cancer patients. Nature nanotechnology, 2(12):780–783, 2007.
160

[77] Kathryn Rosowski. Introduction to cell mechanics and mechanobiology. The Yale journal
of biology and medicine, 86(3):436, 2013.
[78] Bernd W Koenig, Helmut H Strey, and Klaus Gawrisch. Membrane lateral compressibility
determined by nmr and x-ray diffraction: effect of acyl chain polyunsaturation. Biophysical
journal, 73(4):1954–1966, 1997.
[79] Daniel Reeves, Tristan Ursell, Pierre Sens, Jane Kondev, and Rob Phillips. Membrane
mechanics as a probe of ion-channel gating mechanisms. Physical Review E, 78(4):041901,
2008.
[80] Debbie C Thurmond. Insulin-regulated glucagon-like peptide-1 release from l cells: actinout. Endocrinology, 150(12):5202–5204, 2009.
[81] Chauying J Jen, Shuo-Ju Jhiang, and Hsiun-Ing Chen. Invited review: effects of flow on
vascular endothelial intracellular calcium signaling of rat aortas ex vivo. Journal of Applied
Physiology, 89(4):1657–1662, 2000.
[82] Martin A Schwartz and Mark H Ginsberg. Networks and crosstalk: integrin signalling
spreads. Nature Cell Biology, 4(4):E65–E68, 2002.
[83] CT Lim, EH Zhou, and ST Quek. Mechanical models for living cellsa review. Journal of
biomechanics, 39(2):195–216, 2006.
[84] Donald E Ingber. Tensegrity: the architectural basis of cellular mechanotransduction. Annual review of physiology, 59(1):575–599, 1997.
[85] Donald E Ingber. Tensegrity i. cell structure and hierarchical systems biology. Journal of
Cell Science, 116(7):1157–1173, 2003.
[86] J Li, M Dao, CT Lim, and S Suresh. Spectrin-level modeling of the cytoskeleton and optical
tweezers stretching of the erythrocyte. Biophysical Journal, 88(5):3707–3719, 2005.
[87] Olivier Thoumine and Albrecht Ott. Time scale dependent viscoelastic and contractile regimes in fibroblasts probed by microplate manipulation. Journal of cell science,
110(17):2109–2116, 1997.
[88] A Yeung and E Evans. Cortical shell-liquid core model for passive flow of liquid-like spherical cells into micropipets. Biophysical journal, 56(1):139–149, 1989.

161

[89] Philippe Carl and Hermann Schillers. Elasticity measurement of living cells with an atomic
force microscope: data acquisition and processing. Pfl¨ugers Archiv-European Journal of
Physiology, 457(2):551–559, 2008.
[90] RE Mahaffy, S Park, E Gerde, J K¨as, and CK Shih. Quantitative analysis of the viscoelastic
properties of thin regions of fibroblasts using atomic force microscopy. Biophysical journal,
86(3):1777–1793, 2004.
[91] Nan Yang, Kenneth Kar Ho Wong, John R de Bruyn, and Jeffrey L Hutter. Frequencydependent viscoelasticity measurement by atomic force microscopy. Measurement Science
and Technology, 20(2):025703, 2009.
[92] Bruce Alberts, Dennis Bray, Julian Lewis, Martin Raff, Keith Roberts, and James D Watson.
Molecular biology of the cell. 1994. Garland, New York.
[93] Ning Wang, Iva Marija Toli´c-Nørrelykke, Jianxin Chen, Srboljub M Mijailovich, James P
Butler, Jeffrey J Fredberg, and Dimitrije Stamenovi´c. Cell prestress. i. stiffness and prestress
are closely associated in adherent contractile cells. American Journal of Physiology-Cell
Physiology, 282(3):C606–C616, 2002.
[94] JM Mitchison and MM Swann. The mechanical properties of the cell surface i. the cell
elastimeter. Journal of Experimental Biology, 31(3):443–460, 1954.
[95] CT Lim, M Dao, S Suresh, CH Sow, and KT Chew. Large deformation of living cells using
laser traps. Acta Materialia, 52(7):1837–1845, 2004.
[96] Geoffrey N Maksym, Ben Fabry, James P Butler, Daniel Navajas, Daniel J Tschumperlin,
Johanne D Laporte, and Jeffrey J Fredberg. Mechanical properties of cultured human airway
smooth muscle cells from 0.05 to 0.4 hz. Journal of Applied Physiology, 89(4):1619–1632,
2000.
[97] Sung Woong Han, Chikashi Nakamura, Ikuo Obataya, Noriyuki Nakamura, and Jun
Miyake. A molecular delivery system by using afm and nanoneedle. Biosensors and Bioelectronics, 20(10):2120–2125, 2005.
[98] Steven D Hughes, Christian Quaade, John H Johnson, Sarah Ferber, and CB Newgard.
Transfection of att-20ins cells with glut-2 but not glut-1 confers glucose-stimulated insulin secretion. relationship to glucose metabolism. Journal of Biological Chemistry,
268(20):15205–15212, 1993.
[99] George G Holz. Epac: a new camp-binding protein in support of glucagon-like peptide-1
receptor-mediated signal transduction in the pancreatic β -cell. Diabetes, 53(1):5–13, 2004.
162

[100] Martin D Meglasson and Franz M Matschinsky. Pancreatic islet glucose metabolism and
regulation of insulin secretion. Diabetes/metabolism reviews, 2(3-4):163–214, 1986.
[101] Daniel L Cook and Nicholas Hales. Intracellular atp directly blocks k&plus; channels in
pancreatic b-cells. 1984.
[102] LS Satin and DL Cook. Voltage-gated ca2+ current in pancreatic b-cells. Pfl¨ugers Archiv,
404(4):385–387, 1985.
[103] P Rorsman. The pancreatic beta-cell as a fuel sensor: an electrophysiologist’s viewpoint.
Diabetologia, 40(5):487–495, 1997.
[104] Susanne G Straub and Geoffrey WG Sharp. Glucose-stimulated signaling pathways in
biphasic insulin secretion. Diabetes/metabolism research and reviews, 18(6):451–463,
2002.
[105] Mica Ohara-Imaizumi, Tomonori Fujiwara, Yoko Nakamichi, Tadashi Okamura, Yoshihiro
Akimoto, Junko Kawai, Satsuki Matsushima, Hayato Kawakami, Takashi Watanabe, Kimio
Akagawa, et al. Imaging analysis reveals mechanistic differences between first-and secondphase insulin exocytosis. The Journal of cell biology, 177(4):695–705, 2007.
[106] Michael J Caterina, Mark A Schumacher, Makoto Tominaga, Tobias A Rosen, Jon D Levine,
and David Julius. The capsaicin receptor: a heat-activated ion channel in the pain pathway.
Nature, 389(6653):816–824, 1997.
[107] Arpad Szallasi, Daniel N Cortright, Charles A Blum, and Samer R Eid. The vanilloid receptor trpv1: 10 years from channel cloning to antagonist proof-of-concept. Nature Reviews
Drug Discovery, 6(5):357–372, 2007.
[108] Pablo Juan-Pic´o, Esther Fuentes, F Javier Bermudez-Silva, F Javier Diaz-Molina, Cristina
Ripoll, Fernando Rodr´ıguez de Fonseca, and Angel Nadal. Cannabinoid receptors regulate ca¡ sup¿ 2+¡/sup¿ signals and insulin secretion in pancreatic β -cell. Cell calcium,
39(2):155–162, 2006.
[109] Luciano De Petrocellis, Pietro Marini, Isabel Matias, Aniello Schiano Moriello, Katarzyna
Starowicz, Luigia Cristino, Santosh Nigam, and Vincenzo Di Marzo. Mechanisms for the
coupling of cannabinoid receptors to intracellular calcium mobilization in rat insulinoma
β -cells. Experimental cell research, 313(14):2993–3004, 2007.
[110] Sven Karlsson, AJ Scheurink, Anton B Steffens, and Bo Ahr´en. Involvement of capsaicinsensitive nerves in regulation of insulin secretion and glucose tolerance in conscious
163

mice. American Journal of Physiology-Regulatory, Integrative and Comparative Physiology, 267(4):R1071–R1077, 1994.
[111] Jooyoung Jung, Sun Wook Hwang, Jiyeon Kwak, Soon-Youl Lee, Chang-Joong Kang,
Won Bae Kim, Donghee Kim, and Uhtaek Oh. Capsaicin binds to the intracellular domain of the capsaicin-activated ion channel. The Journal of neuroscience, 19(2):529–538,
1999.
[112] Sven-Eric Jordt and David Julius. Molecular basis for species-specific sensitivity to hot chili
peppers. Cell, 108(3):421–430, 2002.
[113] Tatyana G Kuznetsova, Maria N Starodubtseva, Nicolai I Yegorenkov, Sergey A Chizhik,
and Renat I Zhdanov. Atomic force microscopy probing of cell elasticity. Micron,
38(8):824–833, 2007.
[114] Y Akiba, S Kato, K Katsube, M Nakamura, K Takeuchi, H Ishii, and T Hibi. Transient
receptor potential vanilloid subfamily 1 expressed in pancreatic islet beta cells modulates
insulin secretion in rats. BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS, 321(1):219–225, AUG 11 2004.
[115] CE Lewis, A Clark, SJH Ashcroft, GJS Cooper, and JF Morris. Calcitonin gene-related
peptide and somatostatin inhibit insulin release from individual rat b cells. Molecular and
cellular endocrinology, 57(1):41–49, 1988.
[116] JF O’Neill, RC Johnston, L Halferty, GP Brennan, J Keiser, I Fairweather, et al. Adult
triclabendazole-resistant fasciola hepatica: morphological changes in the tegument and gut
following in vivo treatment with artemether in the rat model. Journal of helminthology,
83(2):151, 2009.
[117] WILLY J Malaisse, Abdullah Sener, MAJDA Koser, M Ravazzola, and F Malaisse-Lagae.
The stimulus-secretion coupling of glucose-induced insulin release. insulin release due to
glycogenolysis in glucose-deprived islets. Biochem. J, 164:447–454, 1977.
[118] Zhanxiang Wang, Eunjin Oh, and Debbie C Thurmond. Glucose-stimulated cdc42 signaling
is essential for the second phase of insulin secretion. Journal of Biological Chemistry,
282(13):9536–9546, 2007.
[119] Zhanxiang Wang and Debbie C Thurmond. Mechanisms of biphasic insulin-granule
exocytosis–roles of the cytoskeleton, small gtpases and snare proteins. Journal of cell science, 122(7):893–903, 2009.

164

[120] Yoichi Hamada, Hiroshi Ikegami, Hironori Ueda, Yoshihiko Kawaguchi, Eiji Yamato, Koji
Nojima, Kazunori Yamada, Naru Babaya, Masao Shibata, and Toshio Ogihara. Insulin
secretion to glucose as well as nonglucose stimuli is impaired in spontaneously diabetic
nagoya-shibata-yasuda mice. Metabolism, 50(11):1282–1285, 2001.
[121] Seigo Fujimoto, Mayumi Mori, Hiromi Tsushima, and Mitoshi Kunimatsu. Capsaicininduced, capsazepine-insensitive relaxation of the guinea-pig ileum. European journal of
pharmacology, 530(1):144–151, 2006.
[122] Frances M Ashcroft and Patrik Rorsman. Electrophysiology of the pancreatic β -cell.
Progress in biophysics and molecular biology, 54(2):87–143, 1989.
[123] Patrik Rorsman, Lena Eliasson, Takahiro Kanno, Quan Zhang, and Sven Gopel. Electrophysiology of pancreatic β -cells in intact mouse islets of langerhans. Progress in biophysics
and molecular biology, 107(2):224–235, 2011.
[124] S Gopel, Q Zhang, L Eliasson, XS Ma, J Galvanovskis, T Kanno, A Salehi, and P Rorsman.
Capacitance measurements of exocytosis in mouse pancreatic alpha-, beta- and delta-cells
within intact islets of Langerhans. JOURNAL OF PHYSIOLOGY-LONDON, 556(3):711–
726, MAY 1 2004.
[125] Arthur Beyder and Frederick Sachs. Electromechanical coupling in the membranes
of shaker-transfected hek cells. Proceedings of the National Academy of Sciences,
106(16):6626–6631, 2009.
[126] Dong-Qi Tang, Li-Zhen Cao, Brant R Burkhardt, Chang-Qi Xia, Sally A Litherland, Mark A
Atkinson, and Li-Jun Yang. In vivo and in vitro characterization of insulin-producing cells
obtained from murine bone marrow. Diabetes, 53(7):1721–1732, 2004.
[127] Jean-Claude Henquin, Myriam Nenquin, Patrick Stiernet, and Bo Ahren. In vivo and in vitro
glucose-induced biphasic insulin secretion in the mouse pattern and role of cytoplasmic ca2+
and amplification signals in β -cells. Diabetes, 55(2):441–451, 2006.
[128] BA Becker and AA Gaspari. Pemphigus vulgaris and vegetans. Dermatologic clinics,
11(3):429, 1993.
[129] Ken Ishii, Reiko Harada, Itsuro Matsuo, Yuji Shirakata, Koji Hashimoto, and Masayuki
Amagai. In vitro keratinocyte dissociation assay for evaluation of the pathogenicity of antidesmoglein 3 igg autoantibodies in pemphigus vulgaris. Journal of investigative dermatology, 124(5):939–946, 2005.

165

[130] Grant J Anhalt, Ramzy S Labib, John J Voorhees, Theodore F Beals, and Luis A Diaz.
Induction of pemphigus in neonatal mice by passive transfer of igg from patients with the
disease. The New England journal of medicine, 306(20):1189, 1982.
[131] Aimee S Payne, Don L Siegel, and John R Stanley. Targeting pemphigus autoantibodies
through their heavy-chain variable region genes. Journal of Investigative Dermatology,
127(7):1681–1691, 2007.
[132] Marilyn G Farquhar and George E Palade. Junctional complexes in various epithelia. The
Journal of cell biology, 17(2):375–412, 1963.
[133] Anthea Scothern and David Garrod. Visualization of desmosomes in the electron microscope. Methods in cell biology, 88:347–366, 2008.
[134] George F Obland. The fine structure of the interrelationship of cells in the human epidermis.
The Journal of biophysical and biochemical cytology, 4(5):529–538, 1958.
[135] Hiroshi Shimizu, Takuji Masunaga, Akira Ishiko, Arata Kikuchi, Takashi Hashimoto, and
Takeji Nishikawa. Pemphigus vulgaris and pemphigus foliaceus sera show an inversely
graded binding pattern to extracellular regions of desmosomes in different layers of human
epidermis. Journal of investigative dermatology, 105(2):153–159, 1995.
[136] Xuming Mao, Eun Jung Choi, and Aimee S Payne. Disruption of desmosome assembly
by monovalent human pemphigus vulgaris monoclonal antibodies. Journal of Investigative
Dermatology, 129(4):908–918, 2008.
[137] Paula Berkowitz, Peiqi Hu, Zhi Liu, Luis A Diaz, Jan J Enghild, Michael P Chua, and
David S Rubenstein. Desmosome signaling inhibition of p38mapk prevents pemphigus vulgaris igg-induced cytoskeleton reorganization. Journal of Biological Chemistry,
280(25):23778–23784, 2005.
[138] Subra Suresh. Nanomedicine: Elastic clues in cancer detection. Nature nanotechnology,
2(12):748–749, 2007.
[139] Wilbur A Lam, Michael J Rosenbluth, and Daniel A Fletcher. Chemotherapy exposure
increases leukemia cell stiffness. Blood, 109(8):3505–3508, 2007.
[140] Tomomi E Kimura, Anita J Merritt, and David R Garrod. Calcium-independent desmosomes
of keratinocytes are hyper-adhesive. Journal of Investigative Dermatology, 127(4):775–781,
2006.

166

[141] Jens Waschke. The desmosome and pemphigus.
130(1):21–54, 2008.

Histochemistry and cell biology,

[142] Yuzo Takahashi, Harish P Patel, Ramzy S Labib, Luis A Diaz, and Grant J Anhalt. Experimentally induced pemphigus vulgaris in neonatal balb/c mice: a time-course study of
clinical, immunologic, ultrastructural, and cytochemical changes. Journal of investigative
dermatology, 84(1):41–46, 1985.
[143] Aimee S Payne, Ken Ishii, Stephen Kacir, Chenyan Lin, Hong Li, Yasushi Hanakawa,
Kazuyuki Tsunoda, Masayuki Amagai, John R Stanley, Don L Siegel, et al. Genetic and
functional characterization of human pemphigus vulgaris monoclonal autoantibodies isolated by phage display. Journal of Clinical Investigation, 115(4):888–899, 2005.
[144] Alex I Chernyavsky, Juan Arredondo, Yasuo Kitajima, Miki Sato-Nagai, and Sergei A
Grando. Desmoglein versus non-desmoglein signaling in pemphigus acantholysis characterization of novel signaling pathways downstream of pemphigus vulgaris antigens. Journal
of Biological Chemistry, 282(18):13804–13812, 2007.
[145] Enno Schmidt and Jens Waschke. Apoptosis in pemphigus. Autoimmunity reviews, 8(7):533,
2009.
[146] Beatriz Pelacho, Cristina Natal, Agust´ın Espa˜na, I Sanchez-Carpintero, Mar´ıa J Iraburu, and
Mar´ıa Jes´us Lopez-Zabalza. Pemphigus vulgaris autoantibodies induce apoptosis in hacat
keratinocytes. FEBS letters, 566(1):6–10, 2004.
[147] HL Tang, KL Yuen, HM Tang, and MC Fung. Reversibility of apoptosis in cancer cells.
British journal of cancer, 100(1):118–122, 2008.
[148] Karin Fischer, Simon Voelkl, Jana Berger, Reinhard Andreesen, Thomas Pomorski, and
Andreas Mackensen. Antigen recognition induces phosphatidylserine exposure on the cell
surface of human cd8+ t cells. Blood, 108(13):4094–4101, 2006.
[149] Marcus Lettau, Maren Paulsen, Hendrik Schmidt, and Ottmar Janssen. Insights into the
molecular regulation of fasl (cd178) biology. European Journal of Cell Biology, 90(6):456–
466, 2011.
[150] Mario Puviani, Alessandra Marconi, Emanuele Cozzani, and Carlo Pincelli. Fas ligand in
pemphigus sera induces keratinocyte apoptosis through the activation of caspase-8. Journal
of investigative dermatology, 120(1):164–167, 2003.
[151] MG Pacheco-Tovar, E Avalos-D´ıaz, E Vega-Memije, JJ Bollain-y Goytia, E L´opez-Robles,
MT Hojyo-Tomoka, L Dom´ınguez-Soto, and R Herrera-Esparza. The final destiny of acan167

tholytic cells in pemphigus is fas mediated. Journal of the European Academy of Dermatology and Venereology, 23(6):697–701, 2009.
ˇ C-ZLOTKIN,
´
[152] MARINA FRUSI
Rochel Pergamentz, Beno Michel, Michael David, Daniel
Mimouni, Franc¸ois Br´eg´eg`ere, and Yoram Milner. The interaction of pemphigus autoimmunoglobulins with epidermal cells: activation of the fas apoptotic pathway and the use
of caspase activity for pathogenicity tests of pemphigus patients. Annals of the New York
Academy of Sciences, 1050(1):371–379, 2005.
[153] Vu Thuong Nguyen, Assane Ndoye, Leonard D Shultz, Mark R Pittelkow, Sergei A Grando,
et al. Antibodies against keratinocyte antigens other than desmogleins 1 and 3 can induce
pemphigus vulgaris–like lesions. Journal of Clinical Investigation, 106(12):1467–1479,
2000.
[154] X Wang, F Bregegere, M Fruˇsi´c-Zlotkin, M Feinmesser, B Michel, and Y Milner. Possible apoptotic mechanism in epidermal cell acantholysis induced by pemphigus vulgaris
autoimmunoglobulins. Apoptosis, 9(2):131–143, 2004.
[155] J¨org Weiske, Torsten Sch¨oneberg, Werner Schr¨oder, Mechthild Hatzfeld, Rudolf Tauber,
and Otmar Huber. The fate of desmosomal proteins in apoptotic cells. Journal of Biological
Chemistry, 276(44):41175–41181, 2001.
[156] Claudio Brancolini, Andrea Sgorbissa, Claudio Schneider, et al. Proteolytic processing
of the adherens junctions components beta-catenin and gamma-catenin/plakoglobin during
apoptosis. Cell death and differentiation, 5(12):1042, 1998.
[157] Atsushi Shimizu, Akira Ishiko, Takayuki Ota, Kazuyuki Tsunoda, Masayuki Amagai, and
Takeji Nishikawa. Igg binds to desmoglein 3 in desmosomes and causes a desmosomal split
without keratin retraction in a pemphigus mouse model. Journal of investigative dermatology, 122(5):1145–1153, 2004.
[158] Wolfgang-Moritz Heupel, Detlef Zillikens, Detlev Drenckhahn, and Jens Waschke. Pemphigus vulgaris igg directly inhibit desmoglein 3-mediated transinteraction. The Journal of
Immunology, 181(3):1825–1834, 2008.
[159] Cath´arine C Calkins, Shannon V Setzer, Jean Marie Jennings, Susan Summers, Kazuyuki
Tsunoda, Masayuki Amagai, and Andrew P Kowalczyk. Desmoglein endocytosis and
desmosome disassembly are coordinated responses to pemphigus autoantibodies. Journal
of Biological Chemistry, 281(11):7623–7634, 2006.

168

[160] Xuming Mao, Yasuyo Sano, Jin Mo Park, and Aimee S Payne. p38 mapk activation is
downstream of the loss of intercellular adhesion in pemphigus vulgaris. Journal of Biological Chemistry, 286(2):1283–1291, 2011.
[161] Sergei A Grando, Jean-Claude Bystryn, Alexander I Chernyavsky, Marina Fruˇsi´c-Zlotkin,
Robert Gniadecki, Roberta Lotti, Yoram Milner, Mark R Pittelkow, and Carlo Pincelli.
Apoptolysis: a novel mechanism of skin blistering in pemphigus vulgaris linking the apoptotic pathways to basal cell shrinkage and suprabasal acantholysis. Experimental dermatology, 18(9):764–770, 2009.
[162] Jens Waschke, Paola Bruggeman, Werner Baumgartner, Detlef Zillikens, Detlev Drenckhahn, et al. Pemphigus foliaceus igg causes dissociation of desmoglein 1–containing junctions without blocking desmoglein 1 transinteraction. Journal of Clinical Investigation,
115(11):3157–3165, 2005.
[163] Ning Wang, Keiji Naruse, Dimitrije Stamenovi´c, Jeffrey J Fredberg, Srboljub M Mijailovich, Iva Marija Toli´c-Nørrelykke, Thomas Polte, Robert Mannix, and Donald E Ingber.
Mechanical behavior in living cells consistent with the tensegrity model. Proceedings of the
National Academy of Sciences, 98(14):7765–7770, 2001.
[164] George E Plopper, Helen P McNamee, Laura E Dike, Krzysztof Bojanowski, and Donald E
Ingber. Convergence of integrin and growth factor receptor signaling pathways within the
focal adhesion complex. Molecular Biology of the Cell, 6(10):1349, 1995.
[165] Fang Li, James H-C Wang, and Qing-Ming Wang. Thickness shear mode acoustic wave sensors for characterizing the viscoelastic properties of cell monolayer. Sensors and Actuators
B: Chemical, 128(2):399–406, 2008.
[166] Kathleen J Green and Claire A Gaudry. Are desmosomes more than tethers for intermediate
filaments? Nature reviews Molecular cell biology, 1(3):208–216, 2000.
[167] Reinhard Windoffer, Monika Borchert-Stuhltr¨ager, and Rudolf E Leube. Desmosomes: interconnected calcium-dependent structures of remarkable stability with significant integral
membrane protein turnover. Journal of cell science, 115(8):1717–1732, 2002.
[168] Christoph Riethmuller, Ines Nasdala, and Dietmar Vestweber. Nano-surgery at the
leukocyte–endothelial docking site. Pfl¨ugers Archiv-European Journal of Physiology,
456(1):71–81, 2008.
[169] Zhao Qin, Laurent Kreplak, and Markus J Buehler. Nanomechanical properties of vimentin
intermediate filament dimers. Nanotechnology, 20(42):425101, 2009.

169

[170] Laurent Kreplak, Harald Herrmann, and Ueli Aebi. Tensile properties of single desmin
intermediate filaments. Biophysical journal, 94(7):2790–2799, 2008.
[171] Dimitrije Stamenovi´c, Srboljub M Mijailovich, Iva Marija Toli´c-Nørrelykke, Jianxin Chen,
and Ning Wang. Cell prestress. ii. contribution of microtubules. American Journal of
Physiology-Cell Physiology, 282(3):C617–C624, 2002.
[172] Shaohua Hu, Jianxin Chen, and Ning Wang. Cell spreading controls balance of prestress by
microtubules and extracellular matrix. Front Biosci, 9:2177–2182, 2004.
´ and MARK F COUGHLIN. The role of prestress and architec[173] DIMITRIJE STAMENOVIC
ture of the cytoskeleton and deformability of cytoskeletal filaments in mechanics of adherent
cells: a quantitative analysis. Journal of theoretical biology, 201(1):63–74, 1999.
[174] Pavel S Grudinkin, Valery V Zenin, Andrey V Kropotov, Viktoria N Dorosh, and Nikolay N Nikolsky. Egf-induced apoptosis in a431 cells is dependent on stat1, but not on stat3.
European journal of cell biology, 86(10):591–603, 2007.
[175] Michael Chinkers, James A McKanna, and Stanley Cohen. Rapid rounding of human epidermoid carcinoma cells a-431 induced by epidermal growth factor. The Journal of cell
biology, 88(2):422–429, 1981.
[176] Anne J Ridley, Martin A Schwartz, Keith Burridge, Richard A Firtel, Mark H Ginsberg,
Gary Borisy, J Thomas Parsons, and Alan Rick Horwitz. Cell migration: integrating signals
from front to back. Science, 302(5651):1704–1709, 2003.
[177] Emilios K Dimitriadis, Ferenc Horkay, Julia Maresca, Bechara Kachar, and Richard S Chadwick. Determination of elastic moduli of thin layers of soft material using the atomic force
microscope. Biophysical Journal, 82(5):2798–2810, 2002.
[178] Sarah E Cross, Yu-Sheng Jin, Julianne Tondre, Roger Wong, JianYu Rao, and James K
Gimzewski. Afm-based analysis of human metastatic cancer cells. Nanotechnology,
19(38):384003, 2008.
[179] Etienne Barthel. Adhesive elastic contacts: Jkr and more. Journal of Physics D: Applied
Physics, 41(16):163001, 2008.
[180] Veronica Saravia and Jos´e L Toca-Herrera. Substrate influence on cell shape and cell mechanics: Hepg2 cells spread on positively charged surfaces. Microscopy research and technique, 72(12):957–964, 2009.

170

[181] Oliver Hayden, Roland Bindeus, and Franz L Dickert. Combining atomic force microscope
and quartz crystal microbalance studies for cell detection. Measurement Science and Technology, 14(11):1876, 2003.
[182] Paul Hansma, Hongmei Yu, David Schultz, Azucena Rodriguez, Eugene A Yurtsev, Jessica Orr, Simon Tang, Jon Miller, Joseph Wallace, Frank Zok, et al. The tissue diagnostic
instrument. Review of Scientific Instruments, 80(5):054303–054303, 2009.
[183] O Klymenko, J Wiltowska-Zuber, M Lekka, and WM Kwiatek. Energy dissipation in the
afm elasticity measurements. Acta Physica Polonica-Series A General Physics, 115(2):548,
2009.
[184] Diethelm Johannsmann. Viscoelastic, mechanical, and dielectric measurements on complex samples with the quartz crystal microbalance. Physical Chemistry Chemical Physics,
10(31):4516–4534, 2008.
[185] Cheng Dong and Xiao X Lei. Biomechanics of cell rolling: shear flow, cell-surface adhesion, and cell deformability. Journal of biomechanics, 33(1):35–43, 2000.
[186] Marina Puig-de Morales-Marinkovic, Kevin T Turner, James P Butler, Jeffrey J Fredberg,
and Subra Suresh. Viscoelasticity of the human red blood cell. American Journal of
Physiology-Cell Physiology, 293(2):C597–C605, 2007.
[187] Heng Xie, Manuel A Pallero, Kiran Gupta, Philip Chang, Margaret F Ware, Walter Witke,
David J Kwiatkowski, Douglas A Lauffenburger, Joanne E Murphy-Ullrich, and Alan Wells.
Egf receptor regulation of cell motility: Egf induces disassembly of focal adhesions independently of the motility-associated plcgamma signaling pathway. Journal of cell science,
111(5):615–624, 1998.
[188] K-F Giebel, Clemens Bechinger, Stephan Herminghaus, M Riedel, Paul Leiderer, Ulrich
Weiland, and Martin Bastmeyer. Imaging of cell/substrate contacts of living cells with
surface plasmon resonance microscopy. Biophysical journal, 76(1):509–516, 1999.
[189] Joachim Wegener, Jochen Seebach, Andreas Janshoff, and Hans-Joachim Galla. Analysis
of the composite response of shear wave resonators to the attachment of mammalian cells.
Biophysical journal, 78(6):2821–2833, 2000.
[190] Michael Rodahl and Bengt Kasemo. On the measurement of thin liquid overlayers with the
quartz-crystal microbalance. Sensors and Actuators A: Physical, 54(1):448–456, 1996.

171

[191] F H¨oo¨ k, M Rodahl, P Brzezinski, and B Kasemo. Energy dissipation kinetics for protein
and antibody-antigen adsorption under shear oscillation on a quartz crystal microbalance.
Langmuir, 14(4):729–734, 1998.
[192] L a´ Michelle Furtado et al. Hybridization of complementary strand and single-base mutated
oligonucleotides detected with an on-line acoustic wave sensor. Analyst, 123(10):1937–
1945, 1998.
[193] C Fredriksson, S Kihlman, M Rodahl, and B Kasemo. The piezoelectric quartz crystal mass
and dissipation sensor: a means of studying cell adhesion. Langmuir, 14(2):248–251, 1998.
[194] Michael Rodahl, Fredrik H¨oo¨ k, Claes Fredriksson, Craig A Keller, Anatol Krozer, Peter
Brzezinski, Marina Voinova, and Bengt Kasemo. Simultaneous frequency and dissipation
factor qcm measurements of biomolecular adsorption and cell adhesion. Faraday discussions, 107:229–246, 1997.
[195] Megan Susan Lord, Charlotte Modin, Morten Foss, Mogens Duch, Anne Simmons, Finn S
Pedersen, Bruce K Milthorpe, and Flemming Besenbacher. Monitoring cell adhesion on
tantalum and oxidised polystyrene using a quartz crystal microbalance with dissipation.
Biomaterials, 27(26):4529–4537, 2006.
[196] Megan S Lord, Charlotte Modin, Morten Foss, Mogens Duch, Anne Simmons, Finn S
Pedersen, Flemming Besenbacher, and Bruce K Milthorpe. Extracellular matrix remodelling during cell adhesion monitored by the quartz crystal microbalance. Biomaterials,
29(17):2581–2587, 2008.
[197] Kenneth A Marx, Tiean Zhou, Anne Montrone, Donna McIntosh, and Susan J Braunhut.
A comparative study of the cytoskeleton binding drugs nocodazole and taxol with a mammalian cell quartz crystal microbalance biosensor: Different dynamic responses and energy
dissipation effects. Analytical biochemistry, 361(1):77–92, 2007.
[198] Joachim Wegener, Andreas Janshoff, and H-J Galla. Cell adhesion monitoring using a quartz
crystal microbalance: comparative analysis of different mammalian cell lines. European
biophysics journal, 28(1):26–37, 1998.
[199] Vanessa Heitmann and Joachim Wegener. Monitoring cell adhesion by piezoresonators:
impact of increasing oscillation amplitudes. Analytical chemistry, 79(9):3392–3400, 2007.
[200] Jennifer Y Chen, Minghong Li, Lynn S Penn, and Jun Xi. Real-time and label-free detection
of cellular response to signaling mediated by distinct subclasses of epidermal growth factor
receptors. Analytical Chemistry, 83(8):3141–3146, 2011.

172

[201] Jonathan S Ellis and Michael Thompson. Viscoelastic modeling with interfacial slip
of a protein monolayer electrode-adsorbed on an acoustic wave biosensor. Langmuir,
26(13):11558–11567, 2010.
[202] Diethelm Johannsmann. Studies of viscoelasticity with the qcm. In Piezoelectric Sensors,
pages 49–109. Springer, 2007.
[203] Le Guillou-Buffello, Marcel Gindre, Paul Johnson, Pascal Laugier, V´eronique Migonney,
et al. An alternative quantitative acoustical and electrical method for detection of cell adhesion process in real-time. Biotechnology and bioengineering, 108(4):947–962, 2011.
[204] E Gnecco, R Bennewitz, and E Meyer. Abrasive wear on the atomic scale. Physical review
letters, 88(21):215501, 2002.
[205] C Drummond, J Israelachvili, and Ph Richetti. Friction between two weakly adhering
boundary lubricated surfaces in water. Physical Review E, 67(6):066110, 2003.
[206] AE Filippov, J Klafter, and M Urbakh. Friction through dynamical formation and rupture
of molecular bonds. Physical review letters, 92(13):135503, 2004.
[207] Jonathan S Ellis and Michael Thompson. Slip and coupling phenomena at the liquid–solid
interface. Physical Chemistry Chemical Physics, 6(21):4928–4938, 2004.
[208] Michael Rodahl and Bengt Kasemo. Frequency and dissipation-factor responses to localized
liquid deposits on a qcm electrode. Sensors and Actuators B: Chemical, 37(1):111–116,
1996.
[209] Jacob Klein.
Frictional dissipation in stick-slip sliding.
98(5):056101, 2007.

Physical review letters,

[210] Julia Schmitz and Kay-Eberhard Gottschalk. Mechanical regulation of cell adhesion. Soft
Matter, 4(7):1373–1387, 2008.
[211] GEORGE I Bell, MICAH Dembo, and PIERRE Bongrand. Cell adhesion. competition
between nonspecific repulsion and specific bonding. Biophysical Journal, 45(6):1051–1064,
1984.
[212] Dong Kong, Baohua Ji, and Lanhong Dai. Stability of adhesion clusters and cell reorientation under lateral cyclic tension. Biophysical journal, 95(8):4034–4044, 2008.

173

[213] John Charles Butcher. The numerical analysis of ordinary differential equations: RungeKutta and general linear methods. Wiley-Interscience, 1987.
[214] Benedikt Sabass and Ulrich S Schwarz. Modeling cytoskeletal flow over adhesion sites:
competition between stochastic bond dynamics and intracellular relaxation. Journal of
Physics: Condensed Matter, 22(19):194112, 2010.
[215] Dong Kong, Baohua Ji, and Lanhong Dai. Stabilizing to disruptive transition of focal adhesion response to mechanical forces. Journal of biomechanics, 43(13):2524–2529, 2010.
[216] Linda J Rinko, Michael B Lawrence, and William H Guilford. The molecular mechanics
of p-and l-selectin lectin domains binding to psgl-1. Biophysical journal, 86(1):544–554,
2004.
[217] Sanjay Kumar, Iva Z Maxwell, Alexander Heisterkamp, Thomas R Polte, Tanmay P Lele,
Matthew Salanga, Eric Mazur, and Donald E Ingber. Viscoelastic retraction of single living
stress fibers and its impact on cell shape, cytoskeletal organization, and extracellular matrix
mechanics. Biophysical journal, 90(10):3762, 2006.
[218] Michael D Ward, Micah Dembo, and Daniel A Hammer. Kinetics of cell detachment: peeling of discrete receptor clusters. Biophysical journal, 67(6):2522–2534, 1994.
[219] Elisabetta Ada Cavalcanti-Adam, Tova Volberg, Alexandre Micoulet, Horst Kessler, Benjamin Geiger, and Joachim Pius Spatz. Cell spreading and focal adhesion dynamics are
regulated by spacing of integrin ligands. Biophysical journal, 92(8):2964–2974, 2007.
[220] BNJ Persson. Applications of surface resistivity to atomic scale friction, to the migration of
hotadatoms, and to electrochemistry. The Journal of chemical physics, 98:1659, 1993.
[221] Nina Tymchenko, Erik Nileb¨ack, Marina V Voinova, Julie Gold, Bengt Kasemo, and Sofia
Svedhem. Reversible changes in cell morphology due to cytoskeletal rearrangements measured in real-time by qcm-d. Biointerphases, 7(1-4):1–9, 2012.
[222] Carlos Drummond and Jacob Israelachvili. Dynamic phase transitions in confined lubricant
fluids under shear. Physical Review E, 63(4):041506, 2001.
[223] Paul W Wiseman, Claire M Brown, Donna J Webb, Benedict Hebert, Natalie L Johnson,
Jeff A Squier, Mark H Ellisman, and AF Horwitz. Spatial mapping of integrin interactions and dynamics during cell migration by image correlation microscopy. Journal of cell
science, 117(23):5521–5534, 2004.

174

[224] Mark A Lemmon and Joseph Schlessinger. Cell signaling by receptor-tyrosine kinases. Cell,
141(7):1117, 2010.
[225] Graham Carpenter. Receptors for epidermal growth factor and other polypeptide mitogens.
Annual review of biochemistry, 56(1):881–914, 1987.
[226] AP Dei Tos and I Ellis. Assessing epidermal growth factor receptor expression in tumours:
what is the value of current test methods? European Journal of Cancer, 41(10):1383–1392,
2005.
[227] Alan Wells, Jareer Kassis, James Solava, Timothy Turner, and Douglas A Lauffenburger.
Growth factor-induced cell motility in tumor invasion. Acta Oncologica, 41(2):124–130,
2002.
[228] Douglas A Lauffenburger and Alan F Horwitz. Cell migration: Review a physically integrated molecular process. Cell, 84:359–369, 1996.
[229] Robert J Cain and Anne J Ridley. Phosphoinositide 3-kinases in cell migration. Biology of
the Cell, 101(1):13–29, 2009.
[230] Fred D Allen, Clara F Asnes, Philip Chang, Elliot L Elson, Douglas A Lauffenburger,
and Alan Wells. Epidermal growth factor induces acute matrix contraction and subsequent
calpain-modulated relaxation. Wound repair and regeneration, 10(1):67–76, 2002.
[231] Jeongah Yoon and Thomas S Deisboeck. Investigating differential dynamics of the mapk
signaling cascade using a multi-parametric global sensitivity analysis. PloS one, 4(2):e4560,
2009.
[232] H Steven Wiley, Stanislav Y Shvartsman, and Douglas A Lauffenburger. Computational
modeling of the egf-receptor system: a paradigm for systems biology. Trends in cell biology,
13(1):43–50, 2003.
[233] Ami Citri and Yosef Yarden. Egf–erbb signalling: towards the systems level. Nature reviews
Molecular cell biology, 7(7):505–516, 2006.
[234] C Wofsy, B Goldstein, K Lund, and HS Wiley. Implications of epidermal growth factor
(egf) induced egf receptor aggregation. Biophysical journal, 63(1):98–110, 1992.
[235] Haluk Resat, Jonathan A Ewald, David A Dixon, and H Steven Wiley. An integrated model
of epidermal growth factor receptor trafficking and signal transduction. Biophysical journal,
85(2):730–743, 2003.
175

[236] Boris N Kholodenko, Oleg V Demin, Gisela Moehren, and Jan B Hoek. Quantification of
short term signaling by the epidermal growth factor receptor. Journal of Biological Chemistry, 274(42):30169–30181, 1999.
[237] Birgit Schoeberl, Claudia Eichler-Jonsson, Ernst Dieter Gilles, and Gertraud M¨uller. Computational modeling of the dynamics of the map kinase cascade activated by surface and
internalized egf receptors. Nature biotechnology, 20(4):370–375, 2002.
[238] Neil Kumar, Raffi Afeyan, Hyung-Do Kim, and Douglas A Lauffenburger. Multipathway model enables prediction of kinase inhibitor cross-talk effects on migration of her2overexpressing mammary epithelial cells. Molecular pharmacology, 73(6):1668–1678,
2008.
[239] Frances A Brightman and David A Fell. Differential feedback regulation of the mapk cascade underlies the quantitative differences in egf and ngf signalling in pc12 cells. FEBS
letters, 482(3):169–174, 2000.
[240] NI Markevich, G Moehren, OV Demin, A Kiyatkin, JB Hoek, and BN Kholodenko. Signal
processing at the ras circuit: what shapes ras activation patterns? Systems biology, 1(1):104–
113, 2004.
[241] Satoru Sasagawa, Yu-ichi Ozaki, Kazuhiro Fujita, and Shinya Kuroda. Prediction and validation of the distinct dynamics of transient and sustained erk activation. Nature Cell Biology,
7(4):365–373, 2005.
[242] Stanislav Y Shvartsman, Cyrill B Muratov, and Douglas A Lauffenburger. Modeling and
computational analysis of egf receptor-mediated cell communication in drosophila oogenesis. Development, 129(11):2577–2589, 2002.
[243] Mariko Hatakeyama, Shuhei Kimura, Takashi Naka, Takuji Kawasaki, Noriko Yumoto, Mio
Ichikawa, Jae-Hoon Kim, Kazuki Saito, Mihoro Saeki, Mikako Shirouzu, et al. A computational model on the modulation of mitogen-activated protein kinase (mapk) and akt pathways in heregulin-induced erbb signalling. Biochemical Journal, 373(Pt 2):451, 2003.
[244] Regina Samaga, Julio Saez-Rodriguez, Leonidas G Alexopoulos, Peter K Sorger, and Steffen Klamt. The logic of egfr/erbb signaling: theoretical properties and analysis of highthroughput data. PLoS computational biology, 5(8):e1000438, 2009.
[245] EVAN A Evans. Detailed mechanics of membrane-membrane adhesion and separation. i.
continuum of molecular cross-bridges. Biophysical journal, 48(1):175–183, 1985.

176

[246] Reinhard Lipowsky. Adhesion of membranes via anchored stickers. Physical review letters,
77(8):1652–1655, 1996.
[247] Thomas R Weikl and Reinhard Lipowsky. Adhesion-induced phase behavior of multicomponent membranes. Physical Review E, 64(1):011903, 2001.
[248] Thorsten Erdmann and Ulrich S Schwarz. Stochastic dynamics of adhesion clusters under
shared constant force and with rebinding. The Journal of chemical physics, 121:8997, 2004.
[249] Alice Nicolas, Benjamin Geiger, and Samuel A Safran. Cell mechanosensitivity controls
the anisotropy of focal adhesions. Proceedings of the National Academy of Sciences of the
United States of America, 101(34):12520–12525, 2004.
[250] Tom Shemesh, Benjamin Geiger, Alexander D Bershadsky, and Michael M Kozlov. Focal
adhesions as mechanosensors: a physical mechanism. Proceedings of the National Academy
of Sciences of the United States of America, 102(35):12383–12388, 2005.
[251] Vikram S Deshpande, Robert M McMeeking, and Anthony G Evans. A model for the
contractility of the cytoskeleton including the effects of stress-fibre formation and dissociation. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Science,
463(2079):787–815, 2007.
[252] Ana-Sunˇcana Smith, Kheya Sengupta, Stefanie Goennenwein, Udo Seifert, and Erich Sackmann. Force-induced growth of adhesion domains is controlled by receptor mobility. Proceedings of the National Academy of Sciences, 105(19):6906–6911, 2008.
[253] Clarence E Chan and David J Odde. Traction dynamics of filopodia on compliant substrates.
Science Signaling, 322(5908):1687, 2008.
[254] Alice Macdonald, A Rick Horwitz, and Douglas A Lauffenburger. Kinetic model for lamellipodal actin-integrin’clutch’dynamics. Cell adhesion & migration, 2(2):95–105, 2008.
[255] Sourabh Kharait, Kien Tran, Clayton Yates, and Alan Wells. Cell motility in prostate tumor
invasion and metastasis. In Cell Motility in Cancer Invasion and Metastasis, pages 301–338.
Springer, 2006.
[256] Susanne Rogers and Suzanne Eccles. Egfr signaling in invasion, angiogenesis and metastasis. In EGFR Signaling Networks in Cancer Therapy, pages 245–264. Springer, 2008.
[257] Charles Kumar Thodeti and Kaustabh Ghosh. Mechanisms of tumor cell migration and
invasion in lung cancer metastasis. In Lung Cancer Metastasis, pages 93–109. Springer,
2010.
177

[258] Matthew J Lazzara and Douglas A Lauffenburger. Quantitative modeling perspectives on
the erbb system of cell regulatory processes. Experimental cell research, 315(4):717–725,
2009.
[259] Allison L Berrier and Kenneth M Yamada. Cell–matrix adhesion. Journal of cellular physiology, 213(3):565–573, 2007.
[260] Matthew C Dixon. Quartz crystal microbalance with dissipation monitoring: enabling realtime characterization of biological materials and their interactions. Journal of biomolecular
techniques: JBT, 19(3):151, 2008.
[261] Vanessa Heitmann, Bj¨orn Reiß, and Joachim Wegener. The quartz crystal microbalance in
cell biology: basics and applications. In Piezoelectric Sensors, pages 303–338. Springer,
2007.
[262] Nathalie Q Balaban, Ulrich S Schwarz, Daniel Riveline, Polina Goichberg, Gila Tzur, Ilana
Sabanay, Diana Mahalu, Sam Safran, Alexander Bershadsky, Lia Addadi, et al. Force
and focal adhesion assembly: a close relationship studied using elastic micropatterned substrates. Nature cell biology, 3(5):466–472, 2001.
[263] Karen A Beningo, Micah Dembo, Irina Kaverina, J Victor Small, and Yu-li Wang. Nascent
focal adhesions are responsible for the generation of strong propulsive forces in migrating
fibroblasts. The Journal of cell biology, 153(4):881–888, 2001.
[264] Nathan D Gallant, Kristin E Michael, and Andr´es J Garc´ıa. Cell adhesion strengthening:
contributions of adhesive area, integrin binding, and focal adhesion assembly. Molecular
biology of the cell, 16(9):4329–4340, 2005.
[265] John L Tan, Joe Tien, Dana M Pirone, Darren S Gray, Kiran Bhadriraju, and Christopher S
Chen. Cells lying on a bed of microneedles: an approach to isolate mechanical force.
Proceedings of the National Academy of Sciences, 100(4):1484–1489, 2003.
[266] Lennart Ljung. Aspects on the system identification problem. Signal Processing, 4(5):445–
456, 1982.
[267] RP Araujo, EF Petricoin, and LA Liotta. A mathematical model of combination therapy
using the egfr signaling network. Biosystems, 80(1):57–69, 2005.
[268] Kevin S Brown, Colin C Hill, Guillermo A Calero, Christopher R Myers, Kelvin H Lee,
James P Sethna, and Richard A Cerione. The statistical mechanics of complex signaling
networks: nerve growth factor signaling. Physical Biology, 1(3):184, 2004.
178

[269] William W Chen, Birgit Schoeberl, Paul J Jasper, Mario Niepel, Ulrik B Nielsen, Douglas A
Lauffenburger, and Peter K Sorger. Input–output behavior of erbb signaling pathways as
revealed by a mass action model trained against dynamic data. Molecular systems biology,
5(1), 2009.

179