Académique Documents
Professionnel Documents
Culture Documents
Experiments have shown that mechanical stress can regulate many cellular processes. However, in most
cases, the exact regulatory mechanisms are still not well understood. One approach in improving our understanding of such mechanically induced regulation is the quantitative study of cell deformation under an
externally applied stress. In this paper, an axisymmetric finite-element model is developed and used to study
the deformation of single, suspended fibroblasts in an optical stretcher in which a stretching force is applied
onto the surface of the cell. A feature of our physical model is a viscoelastic material equation whose parameters vary spatially to mimic the experimentally observed spatial heterogeneity of cellular material properties.
Our model suggests that cell size is a more important factor in determining the maximal strain of the optically
stretched fibroblasts compared to the thickness of the actin cortical region. This result could explain the higher
deformability observed experimentally for malignant fibroblasts in the optical stretcher. Our model also shows
that maximal stress propagates into the nuclear region for malignant fibroblasts whereas for normal fibroblasts,
it does not. We discuss how this may impact the transduction of cancer signaling pathways.
DOI: 10.1103/PhysRevE.81.051924
I. INTRODUCTION
*chiamkh@ihpc.a-star.edu.sg
1539-3755/2010/815/05192412
experienced refer to 12 for detailed definition. This difference in the optical deformability has been exploited to
construct diagnostic devices to screen for cancer cells
1517.
In view of these experimental results, a numerical physical model that is able to describe the deformation of an individual cell, normal or cancerous, can give us important
insights into the response of the cell when it is subjected to
mechanical perturbations. The objective of this paper is to
describe a viscoelastic, axisymmetric finite-element model of
cell deformation and to discuss the results that we have obtained from applying this model to study the optical stretching of suspended fibroblast cells. While the use of viscoelastic models 18 to study cell deformation is not new,
especially in the bioengineering literature 1921, we consider a minimal physical model that yields features of the
experimental results and are not concerned with precise and
accurate models. In particular, we consider cellular heterogeneity in our model formulation. Previous viscoelastic models
considered the cell as one homogeneous viscoelastic continuum which can be characterized by a single set of viscoelastic parameters the number of parameters used varies
from model to model. In our model formulation, we recognize that mammalian eukaryotic cells are heterogeneous and
possess at least two distinct regions with different mechanical properties 2225: a cortical region the network of actin
filaments and associated cross-linking proteins that are tethered to the plasma membrane that exhibits a predominately
elastic response and a cytoplasmic region that is mainly free
of such polymer scaffolds and exhibits a viscous response.
Hence, we propose that the viscoelastic parameters in our
model vary spatially to represent these two distinct cellular
domains. Our approach is similar to that of the twocomponent model proposed by Laurent et al. 26. In their
model, the two components, namely. the submembranous
cortical cytoskeleton and the deep cytoskeleton, were
each represented by different sets of linear viscoelastic pa-
051924-1
TEO et al.
BTdV = f ext ,
where M is the mass matrix, B the strain-displacement matrix, the stress vector, f ext the external load vector, and u
the nodal displacement vector. To solve Eq. 1 for u, we
need to define a constitutive equation relating to u. For
this, we will assume that every discretized region within the
body is a viscoelastic three-parameter fluid consisting of a
linear dashpot connected in series to a Voigt element refer to
Fig. 1 and the Appendix for details of the formulation.
The choice of a viscoelastic fluid model is motivated by
experimental observations that suspended fibroblasts deformed in the optical stretcher exhibit residual strains i.e.,
do not return to their initial sizes on the time scale of obser-
FIG. 2. Model geometry and spatial variation of material properties. a Schematic of the axisymmetric geometry of a cell of radius unity
and definition of the radial coordinate used in the finite-element simulation. Making use of symmetry, we need to only model one-quarter
of the cell in our formulation. b Spatial variation of the material properties of the three-parameter fluid model with . The elasticity k of
the cell decreases exponentially away from the cortical region while the viscosities exhibit the opposite trend: 2 varies in a similar manner
as 1 with respect to . The cortical region is defined as the region 0.8 1.
051924-2
vation after the removal of the stretching force 14. For the
loading time scale that we are interested in 2.5 s, the residual strain after stress relaxation unloading is about 20%
of the peak strain during the loading phase refer to Fig. 1c
in Wottawah et al. 14. This observation contradicts the use
of a viscoelastic solid model to describe the deformation of
the fibroblast as such a model does not exhibit any residual
strain upon unloading. Instead, we chose a viscoelastic fluid
model because its unloading response is qualitative similar to
that shown in the optical stretcher experiment 14. This is
also another distinct difference between our model and other
viscoelastic solid models used to describe micropipette aspirations in former studies 19,20.
B. Model parameters
The initial geometry of the cell in our model is represented by a sphere of radius unity in nondimensional units.
We assume that the material properties of the cell are azimuthally isotropic. Then, we can introduce the radial polar
coordinate of a material point from the origin of the cell
Fig. 2a so that the material property of each discretized
region is a function of Fig. 2b.
For normal cells, we represent the actin cortical region as
the region 0.8 1, i.e., with a cortical thickness of 20%,
and the cytoplasm as the region 0 0.8 15. The thickness of the actin cortical region in our model is chosen to
represent the average cortical thickness of normal fibroblasts
17%23% based on the analysis of fluorescence images
straining for actin 15. From these fluorescence confocal
images refer to Fig. 4 in Guck et al. 12, we can see that
the fibroblasts retains an extensive polymeric network
even when in suspension 12 and that the actin filaments
are highly localized to the cortical layer underneath the
membrane. The intensity of the fluorescence in these images
provides the spatial distribution of filamentous actin within
the suspended fibroblasts and we can see that actin concentration decreases toward the center of the fibroblasts away
from the cortical region. Based on these images, we assumed that the actin filaments concentration in the cortical
region to be constant and decreases linearly toward the center of the cell 0.
The elasticity of an isotropic, cross-linked network comprising semiflexible biopolymers such as actin have been
i =
for c
kcort
and
051924-3
TEO et al.
TABLE I. Rheological properties of different cell types obtained experimentally. The cellular elasticity and viscosity reported spreads
over a huge range, with about 23 orders of magnitude difference, depending on the cell type, the experimental techniques, and the model
used to extract the rheological properties from the experimental data. The parameters 0, 1, 1 defined for the SLS below follows this
convention: 0 is the elasticity of the spring element in parallel connection with the Maxwell element, 1 is the elasticity of the spring
element in the Maxwell element, and 1 is the viscosity of the dashpot in the Maxwell element.
Technique
Cell type
Elasticity
Pa
Viscosity
Pa s
Model
Endothelial
Fibroblast
Endothelial
Fibroblast
Chondrocytes
Nucleus
Leukocyte
Neutrophil
D. discoideum
Platelet
Fibroblast
Fibroblast
Chondrocytes
210
G = 70 100
0 = 45, 1 = 75
0 = 960, 1 = 510
0 = 200, 1 = 250
0 = 550 700, 1 = 500 750
0 = 0.75, 1 = 24
118
G = 55
100050000
1700
30000
0 = 170 180, 1 = 190 200
n.a.
G = 25 50 Pa
1 = 3400
1 = 13, 000
1 = 2500
1 = 5000 6000
1 = 33
n.a.
G = 25 Pa
n.a.
400000
2000
1 = 7500 8000
Secant
Three-parameter fluid deduced
SLS
SLS
210
G = 38 Pa
2000
Four-parameter fluid
Direct 10 rad/s
Secant
SLS
SLS
Secant
Direct 10 rad/s
Hertz theory
Kelvin-Voigt
Four-parameter fluid
SLS
Intracellular measurements
Magnetic tweezers 22
LTM 25
Magnetic twist 38
Macrophage
Epithelial
Macrophage
20735
G = 72
15
the cellular material properties holds only for cells in suspension and not for adhered cells.
The cell nucleus is omitted in our model formulation for
simplicity given that the peak strain experienced by the cells
in the optical stretcher for the 2.5 s loading time scale that
we are considering is less than 6% 14. We recognize that
the cell nucleus possesses a very different set of mechanical
properties from the other part of the cell and it plays a significant role in determining the overall response of the cell
under any form of mechanical loading. However, under the
small deformation regime that we are modeling, the contribution of the nucleus to the overall cellular response will
likely be minimal and will not affect the qualitative trend of
the results presented. The contribution of the nucleus to the
overall cellular deformation will become increasingly important and significant at higher strain regime such as that observed for the 10 s loading time scale in the optical stretcher
where the cellular strain reaches a quasistatic plateau. Wottawah et al. 14 attributed this observation to further
fluidlike extension is prevented by an elastic component
evidence of significant contribution from the nucleus,
which now dominates the temporal deformation.
The stresses generated by the optical stretcher on the surface of the cell = 1 can be approximated by 16
= 0 cosn ,
051924-4
Cortical
Cytoplasm
1 Pa s
2 Pa s
1500
7.5
150
0.49
3000
75
30
0.49
k Pa
Poissons ratio
051924-5
TEO et al.
0.08
0.06
experimental
=0
=1
=2.5
=5
=10
0.06
strain
0.05
experimental
=0
=2.5
linear
log
0.05
0.04
0.04
strain
0.07
0.03
0.03
0.02
0.02
0.01
0.01
0
0
(a)
0
0
time (s)
(b)
time (s)
FIG. 4. Color online Sensitivity analysis to determine if our simulation results the strain response along the long axis of the cell are
dependent on the values of the parameter . a We observe that our results are dependent on the value of chosen and the strain response
curve obtained using a value of = 2.5 gives us the best fit to the experimental data. b The strain response obtained from using the
exponential variation gives the best fit to the experimental data as compared to the linear and logarithm variations.
III. RESULTS
A. Maximal strain developed depends on size of suspended
fibroblasts being stretched
From experiments 12,15, we know that the optical deformability of malignantly transformed fibroblasts is significantly higher than that of normal fibroblasts. From the data
in these papers, we estimate that the mean optical deformability of malignantly transformed fibroblasts to be about
40% higher compared to normal fibroblasts. Guck et al. 12
suggested that the thickness of the actin cortical region plays
an important role in determining this difference. We find that
the thickness of the cortical region does indeed affect the
cellular deformability but that cell size, through its effect on
the stress profile is a more important factor in determining
the deformation of the cell instead.
Table III lists some of the differences between malignantly transformed and normal fibroblasts. Normal fibroblasts generally have a larger radius on average, 10% larger
and also a thicker actin cortical region as compared to malignantly transformed fibroblasts 15. The larger radius of
the normal fibroblasts means that the value of the parameter
n in Eq. 4, which characterizes the spread of the stress
distribution, is higher. Thus, under the same experimental
setup, these normal fibroblasts are being subjected to a more
localized stress distribution compared to the malignantly
transformed fibroblasts 15. The range of the parameter n
051924-6
Parameters
Normal
Malignantly
transformed
1723
1424
1215
414
n=6
z
r
z
r
z
r
0.62
0.08
0.024
0.022
strain
Cell type
n=4
0.02
0.018
0.016
0.014
0
10
12
14
16
18
20
22
24
26
051924-7
TEO et al.
0.09
0.08
strain
0.1
Effect of the
parameter n
0.07
Effect of the
cortical
thickness
0.06
0.05
malignant
0.04
0
normal
10
12
14
16
18
20
22
24
IV. DISCUSSION
A. Influence of cell size on the maximal strain
051924-8
FIG. 7. Color online Plot of von Mises stress distribution for normal and malignant fibroblasts. Lighter colors denote higher stress
magnitudes while darker colors denote lower stress magnitudes. The von Mises stress at 2.5 s is obtained from finite-element simulation for
a normal fibroblasts subjected to a localized surface stretching n = 24 and b malignant fibroblasts subjected to a broad surface stretching
n = 4. For the localized stress distribution, we see that the regions with the highest von Mises stress concentration occur at a region just
beneath the cortical region. This observation is distinctly different from that observed for the broad stress distribution in which the region of
highest von Mises stress is in the perinuclear region of the cell.
We have described a viscoelastic, axisymmetric finiteelement model that takes into account the spatial heterogeneity of cells and have shown that our model can be used to
simulate the stretching of suspended fibroblast cells in the
optical stretcher. The first result is that the higher optical
deformability observed for malignant fibroblasts 12 can be
explained by the difference in the surface stress profile
051924-9
TEO et al.
We thank J. Guck and F. Wottawah for insightful discussions about their experiments and acknowledge the financial
support of the Agency for Science, Technology, and Research A*STAR of Singapore under the A*STAR-Imperial
Partnership Ph.D. Program.
FIG. 8. Representation of a standard linear solid model. It consists of a single-spring element 0 in parallel connection with a
Maxwell model. The Maxwell model consists of another spring
element 1 in series connection to a dashpot element 1.
dimensions refer to Fig. 9. We used a different set of notation here to represent the parameters of the two SLS in series
to prevent confusion with the parameters for the threeparameter fluid model.
The representation of the three-parameter fluid model can
be done through the appropriate selection of parameters in
the SLS models to mimic both the fluidlike response of the
dashpot and the creep response of the Voigt element for the
time intervals that we are interested in. First, to mimic the
fluidlike response of the dashpot, we need to set the value of
the single spring element 01 in the first SLS refer to Fig.
9 to approaching zero and set the value of the spring element in the Maxwell model 11 to approaching infinity.
This has the effect of short-circuiting the contributions of
these two spring elements in the first SLS for the time intervals that we are interested in. Next, to mimic the creep response of the Voigt element, we will need to set the value of
the spring element in the Maxwell model 12 in the second
SLS refer to Fig. 9 to approaching infinity. This also has
the effects of short-circuiting the contribution of that spring
element in the second SLS. Using Boltzmanns principle of
superposition 18, we are able to sum the response of the
two SLS to mimic the response of the three-parameter fluid
model.
The development of the finite-element formulation for the
SLS model refer to Fig. 8 starts from the general integral
representation of linear viscosity as a one-dimensional equation as shown below:
1 + 2 t
t 12 2t
,
+
2 = t +
t
t
t
k
k
051924-10
t =
t s
s
ds,
s
t s = 0 + 1ets/1 ,
1 =
1
,
1
where t s is the relaxation function and 1 is the relaxation time for the Maxwell element in the SLS. Following
the approach outlined in Kaliske and Rothert 44, splitting
the integral into an elastic and a viscoelastic contribution
leads to the elastic stress component 0t and the internal
stress equivalent variable h1t as shown below:
s
t = 0
ds +
s
0
t
= 0t +
11
requires the
This recursive determination of the current hn+1
1
quantities n0 and hn1 of the preceding time step n and, therefore, they have to be stored in a database. The shown straindriven integration algorithm is unconditionally stable for
small and large time steps and it is second-order accurate.
The extension of Eq. 9 to a fully three-dimensional approach is easily performed by introducing tensor quantities.
In a multiaxial stress state, the total stress tensor n+1 for a
linear elastic Maxwell material is determined from the elastic
and from the internal stress variables hn+1
contribution n+1
0
1 .
Thus, we can write the stress tensor n+1 in a multiaxial
stress state as follows:
s
ds
1ets/1
s
0
n+1
n+1 = n+1
0 + h1 ,
12
e n+1
n+1
,
0 =D
13
t/1 n
h1 + 1
hn+1
1 =e
s
ds
s
= 0t + h1t.
The internal stress equivalent variable h1t can be evaluated using the recursive formula given by Ref. 44 and as
shown below:
t/1 n
h1 + 1
hn+1
1 =e
1
.
0
1ets/1
1 =
1 et/1 n+1
0 n0,
t
1
10
1 et/1 n+1
0 n0,
t
1
14
051924-11
TEO et al.
22 A. R. Bausch, W. Moller, and E. Sackmann, Biophys. J. 76,
573 1999.
23 B. D. Hoffman, G. Massiera, K. M. Van Citters, and J. C.
Crocker, Proc. Natl. Acad. Sci. U.S.A. 103, 10259 2006.
24 Y. Tseng, T. P. Kole, and D. Wirtz, Biophys. J. 83, 3162
2002.
25 S. Yamada, D. Wirtz, and S. C. Kuo, Biophys. J. 78, 1736
2000.
26 V. M. Laurent, R. Fodil, P. Canadas, S. Fereol, B. Louis, E.
Planus, and D. Isabey, Ann. Biomed. Eng. 31, 1263 2003.
27 M. L. Gardel, J. H. Shin, F. C. MacKintosh, L. Mahadevan, P.
Matsudaira, and D. A. Weitz, Science 304, 1301 2004.
28 M. L. Gardel, F. Nakamura, J. Hartwig, J. C. Crocker, T. P.
Stossel, and D. A. Weitz, Phys. Rev. Lett. 96, 088102 2006.
29 O. C. Zienkiewicz and R. L. Taylor, The Finite Element
Method. Basic Formulation and Linear Problems, 4th ed.
McGraw-Hill, London, 1989, Vol. 1.
30 F. C. MacKintosh, J. Kas, and P. A. Janmey, Phys. Rev. Lett.
75, 4425 1995.
31 K. L. Sung, C. Dong, G. W. Schmid-Schonbein, S. Chien, and
R. Skalak, Biophys. J. 54, 331 1988.
32 G. I. Zahalak, W. B. McConnaughey, and E. L. Elson, J. Bio-
051924-12