J Mater Sci (2007) 42:8811–8823

DOI 10.1007/s10853-007-1917-y

NANO- AND MICROMECHANICAL PROPERTIES OF HIERARCHICAL BIOLOGICAL MATERIALS

A continuum model for remodeling in living structures

Ellen Kuhl Æ Gerhard A. Holzapfel

Received: 29 December 2006 / Accepted: 4 June 2007 / Published online: 14 July 2007

Springer Science+Business Media, LLC 2007

Abstract A new remodeling theory accounting for algorithm is applied to simulate the uniaxial loading of a
mechanically driven collagen fiber reorientation in car- cylindrical tendon and the complex multiaxial loading
diovascular tissues is proposed. The constitutive equations situation in a model artery. After investigating different
for the living tissues are motivated by phenomenologically material and spatial stress and strain measures as potential
based microstructural considerations on the collagen fiber driving forces, we conclude that the Cauchy stress, i.e., the
level. Homogenization from this molecular microscale to true stress acting on the deformed configuration, seems to
the macroscale of the cardiovascular tissue is performed be a reasonable candidate to drive the remodeling process.
via the concept of chain network models. In contrast to
purely invariant-based macroscopic approaches, the pres-
ent approach is thus governed by a limited set of physically
motivated material parameters. Its particular feature is the
underlying orthotropic unit cell which inherently incorpo- Introduction and motivation
rates transverse isotropy and standard isotropy as special
cases. To account for mechanically induced remodeling, Living tissues are able to adapt to physiological and path-
the unit cell dimensions are postulated to change gradually ophysiological stimuli in order to keep adequate perfusion
in response to mechanical loading. From an algorithmic according to the metabolic demand of the tissue. For
point of view, rather than updating vector-valued micro- example, changes in mechanical stimuli lead to altered
structural directions, as in previously suggested models, we cellular and extracellular activities, and typical observed
update the scalar-valued dimensions of this orthotropic unit biological responses are related to growth, remodeling,
cell with respect to the positive eigenvalues of a tensorial adaptation, and repair, i.e., mechanobiology (see, e.g.,
driving force. This update is straightforward, experiences Humphrey [33], Huang et al. [31], Ingber [34], Klein-Nu-
no singularities and leads to a stable and robust remodeling lend et al. [35], Wang and Thampatty [55], Holzapfel and
algorithm. Embedded in a finite element framework, the Ogden [28, 29], Mofrad and Kamm [47] or Lehoux et al.
[40]). Changes in the material (and structural) properties of,
for example, the artery wall through alterations in its
E. Kuhl (&)
Department of Mechanical Engineering, Stanford University, internal microstructure constitute an active process that
Stanford, CA 94305, USA occurs in response to changes of mechanical parameters, a
e-mail: [email protected] process called ‘arterial remodeling.’ It is the endothelium
cell that sense mechanical and humoral parameters, trans-
G. A. Holzapfel
Department of Solid Mechanics, School of Engineering duce signals to the underlying smooth muscle cells and to
Sciences, Royal Institute of Technology, Stockholm 100 44, the surrounding tissue, and relay mechanical and bio-
Sweden chemical changes into biomolecular events. Therefore, the
endothelium at the interface of the blood plays a crucial role
G. A. Holzapfel
Institute for Biomechanics, Center for Biomedical Engineering, in the initiation of arterial remodeling. In particular, in
Graz University of Technology, Graz 8010, Austria cardiovascular tissues, remodeling processes are important

123
8812 J Mater Sci (2007) 42:8811–8823

in the context of arterial development, atherosclerosis, and capture, predict and explain basic trends observed in col-
healing in response to arterial injury. A great deal of inter- lagen fiber remodeling and its impact on the structural
disciplinary research effort is devoted to the (mechanical) response at the tissue or organ level. There seems to be a
signaling pathways because they may enable the identification general agreement that the interplay between matrix stress,
of therapeutic targets and the development of new pharma- fibroblast alignment and stress in the actin network is
cological strategies. Moreover, understanding the interplay responsible for collagen fibril reorientation as reported by
between the architecture of the internal microstructure and the Stopak and Harris [50] and described in detail by Gariki-
mechanical loading is of fundamental importance to engineer, pati et al. [20]. However, we do not aim at explaining the
e.g., blood vessel substitutes, see Nerem and Seliktar [48]. origin of remodeling which is governed by many highly
The function and integrity of organs are maintained by complex interactive phenomena on the cellular level that
the tension in collagen fibers, which contribute signifi- involve altered gene expression in response to altered
cantly to the stability and strength of organs. Collagen loading (i.e., gene transcription, translation, protein syn-
fibers are typically considered as the main load bearing thesis, packing, and activation) which eventually results in
constituent of the extracellular matrix. Accordingly, altered rates of turnover of cells and matrix. Nor do we aim
changes in the material (and structural) properties can at following the classical continuum mechanics approach
primarily be attributed to variations in collagen content, and develop a purely invariant based macroscopic theory
type and thickness and, of course, in the orientation within governed by a number of abstract material parameters. Our
the tissue. Leung et al. [41, 42] were amongst the first to goal is to apply suitable homogenization techniques to
verify experimentally that mechanical forces relate to derive a sound phenomenologically and micromechani-
pressure and flow direct medial cell biosynthesis and cally based formulation with a limited number of param-
modulate structural adaptations to hemodynamic changes. eters that have a clear physical interpretation.
Based on in vitro studies of smooth muscle cells, they To this end, we begin our investigations on the
reported that aortic medial cells attached to elastic mem- microstructural or rather molecular level focusing on the
branes and subjected to cyclic stretching consistently syn- mechanical description of the individual collagen fibers.
thesized collagen of types I and III much more rapidly than The characteristic feature of typical collagen molecules
did cells growing on stationary membranes. In the present is their long, stiff, triple-stranded helical structure in
manuscript we use the word ‘remodeling’ exclusively with which three collagen polypeptide chains are wound
respect to collagen fiber reorientation, while the type and around one another in rope-like superhelical structures
thickness of collagen as well as its content and its con- which are stabilized by numerous hydrogen bonds. The
centration are assumed to be constant. In addition, we do mechanical properties of these helical structures are un-
not address adaptation in the form of volumetric growth like those of any other natural or synthetic polymers.
which is addressed in detail elsewhere in the literature, see, Collagen helices display a remarkable stiffness which
e.g., Rodriguez et al. [49], Lubarda and Hoger [43] or Kuhl may be characterized appropriately through the so-called
et al. [38]. There is, however, strong evidence that growth wormlike chain model. The wormlike chain, or rather
and remodeling can indeed be viewed as separate indi- Kratky and Porod model [36], imagines the polymer as a
vidual processes. Stopak and Harris [50] studied the ori- rod that bends smoothly under thermal fluctuations.
entation of collagen fibrils due to the forces exerted on Traditionally applied to model the DNA double helix,
them by fibroblast in gels. Fiber reorientation was found to see Bustamante et al. [7, 8] and Marko and Siggia [44],
take place in response to changes in the mechanical loading the Kradky and Porod model was recently adopted to
although no significant growth, resorption, and production simulate the behavior of the collagenous triple helix by
of new fibers was reported. Motivated by these findings, Bischoff et al. [3, 4], Garikipati et al. [19, 20] and Kuhl
Garikipati et al. [20] provided a theoretical framework that et al. [37, 39].
focuses exclusively on collagen fiber remodeling and After the collagen fibrils have formed in the extra-
supported their theory by a set of remodeling experiments. cellular space, they are greatly strengthened by the for-
For a more sophisticated theoretical approach that captures mation of covalent crosslinks between lysine residues of
the interaction of the individual phenomena of growth and the constituent collagen molecules. If cross-linking is
remodeling, the reader is referred to Menzel [45]. inhibited, the tensile strength of the fibrils is drastically
To gain further insight into the complex biomechanical reduced, the collagenous tissue becomes fragile and the
phenomena related to tissue remodeling we aim at formu- structure tends to tear, see Alberts et al. [1]. To incor-
lating and implementing a novel constitutive framework porate these characteristic cross-linking network effects,
for collagen fiber remodeling with particular emphasis on different isotropic chain network models have been pro-
the arterial wall, in which type I collagen is the major posed in the past, see, e.g., Flory [17], Treloar [53] and
constituent. As such, the central focus of this study is to Arruda and Boyce [2, 5, 6]. In order to account for the

123
J Mater Sci (2007) 42:8811–8823 8813

anisotropic nature of cardiovascular tissues on the mes- principal stresses (or strains). Hence, as the stress (or
oscopic extracellular matrix level, we generalize the strain) state varies with the radial position, the orientations
cubic isotropic unit cell of the Arruda and Boyce model of the two collagen fiber families also vary across the
to obtain the orthotropic eight-chain model suggested thickness of the arterial wall, as reported by, e.g., Taber
recently by Bischoff et al. [3, 4]. and Humphrey [52] and Holzapfel et al. [30]. This varia-
Finally, it remains to incorporate the living nature of the tion across the wall thickness was successfully obtained
tissue and its ability to adapt its collagenous microstructure from the discrete collagen fiber reorientation models by
to the mechanical loading environment. Naturally, fiber Driessen et al. [13] and Hariton et al. [25]. In [13] it was
directions will evolve in vivo to optimize the load bearing assumed that the collagen fibers align along preferred
capacity while keeping the required compliance. Tradi- directions, situated in between the principal stretch direc-
tionally, remodeling theories in arteries can be classified tions, while in [25] the remodeling process was assumed to
into stress driven and strain driven approaches. The former be stress driven. Within the present manuscript, we com-
are typically based on the assumption that the cardiovas- bine these basic assumptions with the fundamental concept
cular tissue remodels its geometry to restore circumferen- of chain network models to obtain a three-dimensional
tial wall stress due to pressurization and wall shear stress remodeling theory which is general enough to predict
due to blood flow to ‘normal’ levels, see, e.g., Taber and remodeling in complex multiaxial loading situations. In
Humphrey [52], Gleason and Humphrey [22] or Hariton contrast to existing theories, which strongly rely on com-
et al. [25]. Alternatively, motivated by successful predic- plex rotational updates, this new approach can be algo-
tions in hard tissue mechanics, the authors of the latter type rithmically realized in terms of remarkably simple and
of models suggest that strain rather than stress is the rele- straightforward updates of scalar-valued spatial dimen-
vant driving force for the remodeling process, see Kuhl sions.
et al. [37], Himpel et al. [26] or Driessen et al. [13]. Either The manuscript is organized as follows: the governing
of the two theories is able to identify characteristic equations of the micromechanically motivated remodeling
microstructural directions which are allowed to reorient theory are derived in Sect. ‘‘Governing equations.’’ Starting
with respect to the eigendirections of a mechanically rel- from the molecular level, we derive the constitutive
evant second-order tensor. In terms of algorithmic proce- equations for anisotropic soft biological tissues based on
dures, this vector reorientation typically leads to complex the concept of orthotropic chain network models. Section
rotational updates which usually involve singularities due ‘‘Computational examples’’ then focuses on two particular
to the trigonometric nature of the underlying update model problems, a cylindrical tendon subject to uniaxial
equation, see, e.g., Menzel [45, 46]. Although very elegant tension, and a tube-like artery subject to uniaxial stretch in
from a mathematical point of view and maybe well-suited combination with a distending pressure. Section ‘‘Discus-
for microstructures with one predominant orientation, these sion’’ closes with some final remarks.
reorientation models seem rather cumbersome in the con-
text of arterial walls where multiple fiber families need to
be accounted for. Governing equations
When aiming to develop reliable constitutive theories
for remodeling in cardiovascular tissues it is crucial to have In what follows, we summarize our set of constitutive
detailed insight in the structural arrangement of the colla- equations for anisotropic cardiovascular tissues. To this
gen fiber distribution. By using the birefringent properties end, we apply the following hypotheses:
of collagen, Finlay et al. [16] elaborated tangential sections
• Hypothesis I: Large arteries seek to restore wall stress
of cerebral arterial walls to examine the integrated struc-
to within a range of homeostatic values.
tural order of the individual layers. Alternative techniques
• Hypothesis II: Collagen fibers as the main load bearing
providing information about the collagen fiber distribution
constituent of the extracellular matrix adapt their
in arterial walls were discussed recently by Elbischger
orientation and align with respect to the principal stress
et al. [14, 15]. Along these lines, continuously distributed
directions in order to minimize wall stress.
collagen fiber orientations were incorporated in the more
• Hypothesis III: Collagen fiber remodeling can be
recent models by Driessen et al. [11, 12], Freed et al. [18]
modeled and simulated phenomenologically to improve
and Gasser et al. [21].
the understanding of fiber orientation and provide
Experimental findings suggest that at biological equi-
further insight in the structural arrangement of the
librium two predominant fiber orientations can be identi-
individual arterial layers.
fied in each layer of the arterial wall. Typically, these two
discrete families of collagen fibers are found to be located It should be mentioned, however, that although the pres-
somewhere in between the directions of the two maximal ent model takes into account microstructural information,

123
8814 J Mater Sci (2007) 42:8811–8823

it is still based on a rather phenomenological approach in the constant and h is the absolute temperature. In the case of
sense that it does not explain the mechanisms how the indi- living tissues, we suggest h = 310 K, i.e., h = 37 C.
vidual cells actually sense changes in loading and commu- The two parameters governing the chain behavior are
nicate this information. Many different receptors on the the contour length L and the persistence length A, as
surface of endothelial cells and vascular smooth muscle cells illustrated in Fig. 1.
are able to detect subtle changes in the mechanical envi- The force required to pull the ends of the chain away
ronment. They initiate various different mechanotransduc- from each other by a distance r thus follows straightfor-
tion cascades according to the nature of the mechanical wardly by taking the derivative of the free energy wchn with
stimulus perceived. The cytoskeleton and other structural respect to the end-to-end length.
components play an important role in mechanotransduction " #
as they are able to transmit and modulate tension between chn dwchn 1 r 1
f ¼ ¼ kh 4 þ
1 ð2Þ
focal adhesion sites, integrins and the extracellular matrix. dr 4A L ð1
r=LÞ2
Moreover, changes in the mechanical environment may also
initiate changes in the ionic composition of the cells, medi- Note that due to the particular nature of the free energy
ated by ion channels, stimulate various membrane receptors chn
w , the end-to-end length r of a wormlike chain cannot
and induce complex biochemical responses, see, e.g., Huang extend beyond its contour length L as 0 < r < L.
et al. [31], Mofrad and Kamm [47] or Lehoux et al. [40] for
excellent overviews. Since we do not aim at simulating these Remark 1 [Parameters on the collagen fiber level] The
molecular mechanisms of mechanotransduction, all these wormlike chain model is essentially a two-parameter
phenomena are modeled phenomenologically through a set model governed by the contour length L and the persistence
of continuum based remodeling equations which we describe length A. Figure 1 illustrates the physical meaning of the
in the sequel. persistence length. It shows the force-displacement curves
We begin on the microstructural level with the con- of a single collagen fiber indicating the increase in initial
stitutive description of the individual collagen fibers. On stiffness with increasing persistence length for, say,
the mesolevel, we then elaborate a representative volume A = 0.1, A = 0.4, and A = 0.8. Note that throughout the
element representing the extracellular matrix. On the entire manuscript, all lengths of the model have been
macroscopic level, we finally characterize the overall rendered non-dimensional by dividing them by the link
tissue behavior through a micromechanically motivated length of the chain, as proposed by Garikipati et al. [19].
constitutive model which is able to account for micro-
structural adaptation in response to changes in the Remark 2 [Specific data for the persistence length]
mechanical loading. F-actin (a filamentous protein responsible for the con-
traction and relaxation of muscle) has a persistence
On the collagen fiber level length A of approximately 16 lm. For nanotubes A is in

On the microscopic level, we assume that the microstruc-

ture of a collagen triple helix is represented through the
10
wormlike chain model. Wormlike chain models were
introduced within the context of DNA mechanics by Marko 9

and Siggia [44], and Bustamante et al. [7, 8], and recently 8
applied to collagen fibers by Bischoff et al. [3, 4], Gari- 7
f(r/L) times 1/(kθ)

kipati et al. [19, 20] and Kuhl et al. [37, 39]. In the sta-
6
tistical mechanics of long chain molecules such as collagen
fibrils, the key kinematic variable that characterizes the 5
conformation of the chain is the end-to-end length r. 4
According to the wormlike chain model, the free energy
3
wchn of a single collagen fiber can be expressed in terms of
the end-to-end length in the following form. 2

2  1
chn L r 1 r
w ¼ wchn
0 þ kh 2 2þ
ð1Þ 0
0 0.2 0.4 0.6 0.8 1
4A L 1
r=L L
r/L
Herein, wchn
0 is the value of the chain energy in the Fig. 1 Collagen fiber level s Single chain force vs. chain stretch for
unperturbed state, k = 1.381 · 10–23 J/K is the Boltzmann varying persistence lengths A

123
J Mater Sci (2007) 42:8811–8823 8815

the millimeter range. Note, however, that A for DNA The overall chain energy Wchn follows by summing up
in vivo has a value of ~50 nm (Hagerman [24]), and A the contributions wchn of eight individual chains weightened
for synthetic polymers is typically only a few nanome- by the overall chain number density nchn, i.e., the number of
ters. Hence, the persistence lengths A L of a typical chains per unit volume. According to the original eight-
DNA molecule and a synthetic molecule are considerably chain model by Arruda and Boyce, each of these chains
smaller than their contour lengths L. Recent studies connect the corners of a regular cuboid of dimensions 2 l1,
performed by means of optical tweezers seem to indicate 2 l2, and 2 l3 with its center, compare Fig. 2, left. The end-
that under physiological conditions collagen I molecules to-end length r0 in the undeformedpconfiguration, thus,
ffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
have a persistence length of ~14.5 nm which would be follows straightforwardly as r0 ¼ l2I ¼ l21 þ l22 þ l23 :
less than 5% of their contour length of ~309 nm, see Sun The unit cell is postulated to deform in the principal stretch
et al. [51]. Accordingly, collagen would be much more space. Accordingly, the end-to-end length r in the deformed
flexible than previously assumed, yet even more flexible configuration can be expressed in terms of the deformation
than DNA. Although we suggest to stick to the wormlike gradient F or rather in terms of the right Cauchy-Green
chain approach in the sequel, this is not a general limi- tensor C,
tation of the overall constitutive model as such. Due to qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffi
the modular structure of the overall framework, the free r ¼ l2I n

I  n
I ¼ l2I n0I  C  n0I ¼ l2I I
IC ð5Þ
energy function for the individual fibers (1) can easily be
modified, adapted and integrated straightforwardly in the with explicit summation over all three direction I = 1,2,3.
macroscopic model. Here, we have introduced the non-standard invariants
I
IC ¼ n
I ¼ n0I  C  n0I with the understanding that I
IC

I  n
On the extracellular matrix level represents the stretch in the n0I direction squared. Thereby,
n0I are the unit normal vectors of the unit cell axes in the
From a mechanical point of view, the extracellular matrix undeformed reference configuration, see figure 3. After the
is modeled as a surrounding substrate in which the indi- deformation, they map onto the vectors n
I ¼ F  n0I which
vidual collagen fibers are embedded. A representative are obviously no longer of unit length.
volume element of the extracellular matrix thus consists of Finally, the repulsive contribution W
rep ¼
lnðI
C I
C I
C Þ
1 2 3
a substrate of elastin, proteoglycans and cell, characterized is constructed to compensate for the chain stresses in the
through the isotropic free energy wiso, and an anisotropic reference configuration caused by non-vanishing initial
contribution due to the individual chains wchn. Moreover,
we introduce a repulsive chain contribution wrep to char-
acterize the tissue’s behavior of the initial configuration
such that the total free energy may be written in the fol-
lowing form.

W ¼ Wiso þ Wchn þ Wrep ð3Þ

The individual terms of the free energy take on the

explicit representations. Fig. 2 Kinematics of the eight-chain model—orthotropic case (left),
transversely isotropic case (middle), and isotropic case (right)
1 1
Wiso ¼ k ln2 ðJÞ þ lðI1C
ndim Þ
l lnðJÞ
2

2
2
 F, u n01 =nσI
chn nchn L r 1 r
W ¼ kh 2 2þ
ð4Þ
4A L 1
r=L L
" # l1 → λσ+
n chn
1 1 1
rep I
rep
W ¼ kh þ
W n02 = nσII
4A L 4r0 ð1
r0 =LÞ2 4r0

n03 =nσIII l2 → λσ+

II
For the isotropic term Wiso, we apply a standard
F, u l3 → λσ+
III
neo-Hookean model expressed in terms of the first
invariant I1C ¼ C : I of the right Cauchy-Green tensor
Fig. 3 Remodeling based on changes of cell dimensions s
C ¼ Ft  F;where F ¼ rX u denotes the deformation Instantaneous alignment of cell axes n0I with eigenvectors nrI and
gradient and J ¼ detðFÞ > 0 is its determinant. Moreover, k gradual adaptation of cell dimensions lIwith respect to eigenvalues
and l are the standard Lamé constants. krþ
I

123
8816 J Mater Sci (2007) 42:8811–8823

end-to-end lengths r0. Classically, the free energy W assume the chain number density to be constant, however, its
introduced in (3) defines the Cauchy stress r in the fol- evolution in time due to changes in collagen content and
lowing form, thickness could be incorporated straightforwardly. The
degree of anisotropy is governed by the unit cell dimensions
1 dW t l1, l2,p
and
r¼  F ¼ riso þ rchn þ rrep ð6Þ ffiffiffiffi l3 which implicitly define the end-to-end length
J dF r0 ¼ l2I in the initial configuration.
whereby the individual stress contributions can be
expressed as follows.
On the tissue level
1
riso ¼ ½k lnðJÞ
lI þ lb In cardiovascular tissues, the collagen fibers are not arbi-
J " #
chn trarily distributed in space but follow rather a particular
n 1 1 1
rchn ¼ kh þ

chn
r ð7Þ
pattern. To account for remodeling in the form of micro-
4AJ L 4rð1
r=LÞ2 4r structural rearrangement, we allow the fiber direction to
" #
nchn 1 1 1 rotate in response to the current mechanical stress envi-
rep
r ¼ kh þ

rep
r ronment. We assume the Cauchy stress tensor
4AJ L 4r0 ð1
r0 =LÞ2 4r0
r ¼ krI nrI  nrI ð8Þ
Herein, the Finger deformation tensor b ¼ F  Ft repre-
sents a characteristic spatial strain measure. The tensorial to be the driving force of the remodeling process. Here, we
basis of the chain stress follows straightforwardly as have introduced its eigenvalue decomposition with krI and

chn ¼ l2I n
r
I :Note that the repulsive energy Wrep has

I  n nI being the corresponding eigenvalues and eigenvectors,
been constructed such that the corresponding tensorial respectively. For notational simplicity, we have implicitly

rep ¼
l2I =I
IC n
basis of the repulsive stress r
I  n

I ensures a assumed the summation over all I = 1,2,3 components. Our
stress free reference configuration as rrep ¼
rchn jr¼r0 for remodeling theory is based on two fundamental hypothe-
which I
IC ¼ 1 for all I = 1,2,3. ses:

Remarks 3 [Special cases of transverse isotropy and • Hypothesis I: The characteristic directions n0I of the
isotropy] The orthotropic eight-chain model, as introduced microstructure align instantaneously with respect to the
by Bischoff et al. [3, 4], can be understood as a kinematic eigenvectors nrI :
generalization of the original isotropic eight-chain model, • Hypothesis II: The unit cell dimensions lI adapt
as indicated in Fig. 2. gradually with respect to the positive eigenvalues krþ
I :

• Transverse The first postulate is motivated by the general idea

pffiffiffiffiffi isotropy l1 = r0 and l2 = l3 = 0 and thus
r ¼ I
1C r0 beyond all network models that the unit cell is taken to
pffiffiffi pffiffiffiffiffiffiffiffiffiffi
• Isotropy l1 ¼ l2 ¼ l3 ¼ r0 = 3 and thus r ¼ I1C =3 r0 deform in the principal stretch space, see, e.g., Boyce and
Arruda [6]. The second postulate closely follows the recent
The case of transverse isotropy follows by choosing approach of Hariton et al. [25] suggesting that the collagen
l2 = l3, its yet more special case is the consideration of one fibers in cardiovascular tissues are located between the
single fiber direction n01 with l1 = r0 and l2 p
= ffiffiffiffi
l3ffi = 0. In this directions of the two maximum principal stresses.
particular case, the end-to-end length r ¼ I
1C r0 is obvi- Accordingly, the essence of our remodeling theory can be
ously equivalent to the scaled stretch along the fiber summarized as follows,
direction n01 : The special case of isotropy follows by
þ
assuming that all pffiffiffi three cell dimensions are equal krI
l1 ¼ p ¼ l3ffi ¼ r0 = 3 such that the end-to-end length
l2 ffiffiffiffiffiffiffiffiffi n0I  nrI and lI ! r0 þ ð9Þ
jjkrI jj
r ¼ I1C =3 r0 can be expressed exclusively in terms of the
trace of the right Cauchy-Green tensor I1C ¼ C : I: compare with Fig. 3. For the sake of notational simplicity,
þ
we have introduced the notion krI with the understanding
þ þ
Remarks 4 [Parameters on the extracellular matrix level] that krI ¼ krI for positive eigenvalues krI > 0 and krI ¼ 0
On the extracellular matrix level, the mechanical behavior is for non-positive eigenvalues krI  0: From a physiological
characterized through eight parameters in total, i.e., Lamé point of view, we thus exclusively allow for tension
constants k and l characterizing the surrounding substrate, driven reorientation processes. It remains to formulate a
the micromechanically motivated chain parameters L and A reasonable evolution equation for the microstructural cell
and the chain number density nchn. For the moment, we shall dimensions lI which we assume to obey the following law

123
J Mater Sci (2007) 42:8811–8823 8817

rþ 
kI l0I dt n01 ¼ x  n01
dt lI ¼ j
expð
jtÞr0 ð10Þ
jjkrþ
I jj r0
which is obviously more cumbersome due to numerical
where j denotes a relaxation parameter and l0I ; I = 1,2,3, singularities introduced through trigonometric functions.
are the dimensions of the cuboid in the undeformed In contrast to the rather simple explicit equation for
configuration. Alternatively, the above equation could be the adaptation of the unit cell dimensions (11), the
integrated in time to render an explicit update equation for fiber direction can be expressed through the exponential
all cell dimensions lI. update

 1
krþ l0I n0kþ1
3
¼ expð
Dt e xÞ  n0k x ¼ n0kþ1  nrmax
lI ¼ I

½1
expð
jtÞr0 þ l0I ð11Þ 1 1
j 1
jjkrþ
I jj r 0
3
where x and e denote the time discrete rotation vector and
Recall that the reorientation
pffiffiffiffi process itself leaves the
the third-order permutation tensor, respectively. Recall that
initial chain length r0 ¼ l2I unaltered.
in addition to potential numerical difficulties, this previous
reorientation approach is restricted to transversely isotropic
Remarks 5 [Remodeling based on fiber rotation] Previous
biological tissues and its extension to reorienting multiple
remodeling approaches in the literature focused on the
directions seems to be a rather complex task.
reorientation of a single microstructural direction, see, e.g.,
Menzel [46], Kuhl et al. [37] and Himpel et al. [26].
Remarks 6 [Dissipation] It is a well-accepted fact that a
Applications thus focused primarily on transversely
purely mechanical theory is thermodynamically inadmis-
isotropic biological tissues with one distinguished fiber
sible for remodeling processes that stiffen the material, see,
orientation such as muscles, tendons or ligaments for which
e.g., the discussions in Menzel [46], Kuhl et al. [37], and
l2 = l3  l1. Similar to the previous approach, reorientation
Himpel et al. [26] or Garikipati et al. [20]. The fact that the
was assumed to be driven by either strain or stress, based on
dissipation would be positive for stiffening materials in a
an eigenvalue decomposition, as given in (8). In contrast to
purely mechanical theory indicates that other thermody-
the new model suggested in this contribution, the following
namic phenomena, e.g., of chemo-mechanical nature,
assumptions were postulated for the remodeling process:
should indeed be taken into account. Alternatively, mixture
• The characteristic direction n01 of the microstructure theories could be considered in which energy and entropy
aligns gradually with respect to the eigenvector nrmax are exchanged amongst the individual constituents of the
related to the maximum positive eigenvalue krþ
max : tissue, see, e.g., Gleason and Humphrey [22, 23].
• The unit cell dimensions lI remain unaffected by the Accordingly, to date, there is no general agreement of how
process of remodeling. evolution laws for reorientation of microstructural direc-
tions should be formulated. In the context of linear elas-
Consequently, the reorientation-based analogue of Eq.
ticity, it has been shown that the free energy attains an
(9) could be expressed as follows, compare Fig. 4.
extremum if strain and stress share the same principal
n01 ! nrmax and lI ¼ const. directions, see, e.g., Cowin [10] or Vianello [54]. In non-
linear elasticity, however, it is not even clear to date
Rather than updating the scalar-valued unit cell lengths whether stress or strain is the relevant driving force for the
lI, this reorientation approach requires an update of the remodeling process. The previous n01 ! nrmax non-linear
vector-valued fiber direction n0I reorientation model addressed in Remark 5 is based on the
general paradigm that nature always tries to find the
extremum, see Menzel [46], Kuhl et al. [37] and Himpel
et al. [26]. However, the recent reorientation model for
arteries by Hariton et al. [25] just postulated a general
stress driven remodeling process and so does the model
presented herein. For detailed discussions about the impact
of the dissipation inequality in remodeling, the reader is
referred to the excellent discussion by Garikipati et al.
[20].
Fig. 4 Remodeling based on changes of cell orientation s Gradual
alignment of the cell axes n01 with eigenvector nrmax of related Remarks 7 [Parameters on the tissue level] An additional
maximum principal eigenvalue krþ
max and constant cell dimensions
lI = const convincing benefit of the incorporation of remodeling is

123
8818 J Mater Sci (2007) 42:8811–8823

that the number of parameters could even be reduced, as Computational examples

compared to the fixed cell dimension model introduced in
Sect. ‘‘Governing equations.’’ For the overall model of Finally, the features of the proposed remodeling strategy is
adapting living tissue, seven material parameters are elaborated by means of two benchmark problems, i.e., a
required altogether, i.e., the two Lamé constants k and l cylindrical tendon subject to uniaxial tension, and a tube-
for the extracellular matrix, the chain number density nchn, like artery subject to uniaxial stretch in combination with
the micromechanically motivated contour length L and the internal pressure.
persistence length A. However, now, instead of prescribing Since we aim at elaborating inhomogeneous structures
fixed cell dimensions l1, l2, and l3, we just have to define with an initially random fiber orientation, the remodeling
the initial chain length r0, while the individual dimensions strategy is embedded in a standard finite element algorithm
of the cell evolve naturally in response to the given with the remodeling equation being evaluated at the inte-
mechanical loading environment. gration point level. A Newton–Raphson solution strategy
Nevertheless, one additional parameter j is introduced based on a consistent linearization of the governing equa-
to account for the speed of adaptation. From a biological tions is applied throughout to solve the non-linear biome-
point of view, the relaxation parameter j reflects the chanical problem efficiently in an incrementally iterative
turnover rate or rather the speed of adaptation. Note that we way.
have implicitly assumed that the time scale for fiber In the following examples, we choose the Lamé param-
reorientation is orders of magnitude larger than the typical eters to k = 27.293 and l = 3.103 according to Menzel
time scale of a diastolic-systolic pressure circle. The role of [45]. The Boltzmann constant is k = 1.381 · 10–23, and the
j is illustrated in Fig. 5 for a simulation with absolute temperature is h = 310, measured in Kelvin.
k = 27.293, l = 3.103, nchn = 7.0 · 1021 chains per unit Recall that all chain lengths are normalized by a length
volume, L = 2.125, A = 1.82, and r0 = 1.0. In the initial scale originating from the statistical model. Following
state, we assumepffiffian ffi isotropic fiber distribution with Garikipati et al. [19], we scale all lengths by the link length
l1 ¼ l2 ¼ l3 ¼ r0 = 3: The curves illustrate the angle of in a chain. Accordingly, the contour length is chosen to be
alignment for a cubic specimen of unit length subject to L = 1.594, the corresponding persistence length that ac-
uniaxial tension with an incremental loading of DF = 1.0. counts for a reasonable initial stiffness
pffiffiffiffi is A = 1.365, and the
The load is applied in 10 steps in the dimensionless time initial end-to-end length is r0 ¼ l2I ¼ 1:0: The degree of
interval 0 < t £ 10. Then, at F = 10.0, the load is held anisotropy is reflected through the chain number density
constant while the cell dimensions are allowed to remodel nchn, i.e., the number of chains per unit volume, which we
progressively at 10 < t £ 100. Figure 5 demonstrates that choose in the order of 1021, as suggested by Bischoff et al.
the collagen fibers align gradually with respect to the [3] or Garikipati et al. [19], thus nchn = 2 · 1021. The
loading axis. As time evolves, l1 fi r0 while l2 = l3 fi 0 relaxation parameter is chosen to be j = 0.025 per unit
whereby n01 is instantaneously aligned with the loading time. In all examples, we start with an initially random fiber
axis. As expected, the speed of the adaptation process orientation which is realized by assigning different values to
increases for larger values of the relaxation parameter j. the unit cell dimensions lI in each element of the mesh. Due
to the geometric and material non-linearities of the problem,
the load is applied incrementally until the final load level is
reached. Then, the load is held constant while remodeling
60
occurs until convergence towards a biological equilibrium
state.
50

Cylindrical tendon under uniaxial tension

angle of alignment

40

30 To elaborate the proposed remodeling algorithm in the

simple case of uniaxial loading, we analyze a cylindrical
20 tendon under uniaxial tension which has been studied
earlier in the context of tissue engineering of tendon
10
constructs by Kuhl et al. [37]. When strained along its
load-bearing axis, tendon shows a microscopical straight-
0
0 20 40 60 80 100 ening of the low amplitude wavy collagen fibers accom-
time
panied by an exponentially increasing resistance to
Fig. 5 Tissue level s Angle between collagen fibers and direction of elongation. The model tendon we aim at analyzing quali-
uniaxial tension versus time for varying relaxation parameters j tatively to elaborate whether the suggested model is able to

123
J Mater Sci (2007) 42:8811–8823 8819

capture these general trends is 10 units long and has an combination with an internal pressure representing the blood
initial cross-section of 10 units2. It is discretized with 10 flow. It has long been recognized that the layered structure of
elements in height and 48 elements per cross-section, thus cardiovascular tissues displays a geometric complexity way
with 480 tri-linear finite elements in total. The load is beyond the highly organized parallel bundles of collagen
increased incrementally with DF = 25 from 0 < t £ 12 until fibers that constitute the mechanical backbone of tendons.
a final load level of F = 300 is reached. From 12 < t £ 150, Motivated by earlier theoretical studies of Driessen et al.
we apply another 138 remodeling steps until biological [13] and Hariton et al. [25], and supported by the experi-
equilibrium occurs. mental findings of Finlay et al. [16], we analyze the
Figure 6 illustrates the remodeling history of the cylin- remodeling history in a tube-like artery. At this point, we
drical tendon. At first, when the load is applied, the tendon restrict ourselves to a rather qualitative analysis aiming at
is stretched in the axial direction up to about 175% of its including physiologically realistic data at later stages of this
original length, compare with Fig. 6, left. However, as time project. The initial dimensions are given through a length of
evolves, the collagen fibers tend to align gradually with the 8 units and an inner and outer radius of 1 and 3 units,
loading axis. Accordingly, the structure stiffens signifi- respectively. Along the height we apply 12 elements, we use
cantly and the overall stretch decreases to about 140%. At 8 elements across the thickness and 16 elements in circum-
the final and yet stiffest state, all fibers are aligned with the ferential direction. Here, we apply standard tri-linear finite
direction of the mechanical load. elements. Note, however, that shell elements would be more
During the remodeling process, the unit cell lengths appropriate when it comes to quantitative studies. The load is
pffiffiffiffi evolved from initially random values 0 £ lI
have obviously applied incrementally in 25 load steps from 0 < t£25 with
£ r0 with l2I ¼ r0 to l1 = r0 and l2 = l3 = 0 representing Dp = 0.192 p0 and Dl = 0.032 until a final state of p = 4.8 p0
the special case of transverse isotropy. Selected snapshots and Dl = 0.8, i.e., an axial stretch of 10%, is reached. Again,
of the remodeling history are depicted in Fig. 6, right. For the load is then held constant for another 125 time steps for
each element, the end-to-end vectors of each chain have the time period 25 < t£150 to allow for remodeling towards
been projected on the tendon’s surface. The individual a final state of biological equilibrium. Figure 7 shows five
colors represent the collagen fiber angle measured against selected snapshots of the remodeling process. In between the
the loading axis. Red colors indicate a full alignment with an first and the second snapshot, the internal pressure and the
angle zero, while blue colors indicate that the collagen fibers prescribed axial stretch are increased incrementally.
are oriented orthogonal to the loading axis. In particular, in Accordingly, the tube blows up and stretches along its axial
the second snapshot of the series at t = 12 corresponding to direction occur. Due to the nearly incompressible behavior of
the final loading state, the spatial inhomogeneity is nicely soft biological tissues, the tube thickness decreases drasti-
visible. The red elements with aligned collagen fibers display cally in response to loading. As time evolves, the collagen
very stiff response and deform only marginally while the soft fibers tend to reorient so that they are finally located between
blue elements undergo significant stretches. As expected, the two directions of maximum principal stress. For this
this initial inhomogeneity tends to vanish gradually in the particular case of loading, these are obviously the longitu-
course of remodeling. dinal and the circumferential directions.
Starting from an initial random orientation
pffiffiffiffi at random
Tube-like artery subject to uniaxial stretch and internal unit cell lengths 0 £ lI £ r0 with l2I ¼ r0 in the first
pressure snapshot of the series, the unit cell lengths evolve pro-
gressively driven by the positive eigenvalues of the Cauchy
We turn now to the more challenging example of a multi- stress krþ
I : As a natural consequence, the chain directions
axial loading state induced by a uniaxial stretching in remodel gradually and the tube stiffens with respect to the

Fig. 6 Cylindrical tendon 1.8

under uniaxial tension s Stretch 1.7
versus time (left) and selected 1.6
snapshots of the remodeling 1.5
stretch

process at different times with

1.4
colors representing the
1.3
individual collagen fiber angles
(right) 1.2
1.1
1
0 50 100 150
time

123
8820 J Mater Sci (2007) 42:8811–8823

Fig. 7 Tube-like artery subject

to axial stretch and internal
pressure s Selected snapshots
of the remodeling process at
different times with colors
representing the individual
collagen fiber angles

applied loading. Accordingly, the outer radius, which had The predicted inhomogeneity in the radial direction with
increased during the first stages of loading, decreases an almost circumferential fiber orientation at the luminal
remarkably in response to remodeling, compare time t = 24 side of the artery and an increase of the inclination towards
with t = 150. While the chains had originally been oriented the outer side is closely related to the phenomenon of
randomly in space at t = 1, their component in radial prestress, typically encountered in arterial specimens. For
direction vanishes gradually as l3 goes to zero in response the considered model problem, the degree of inhomoge-
to the compressive stresses kr3  0 in the radial direction at neity, obviously, strongly depends on the stretch-to-pres-
t = 150. Thus, the suggested remodeling algorithm proves sure ratio. Motivated by a recent study by Gleason and
able to produce not only transversely isotropic but also Humphrey [23] who analyzed the elastin, collagen and
orthotropic microstructures in a natural way. smooth muscle cell turnover and remodeling in response to
Recall that the final collagen fiber angles developed different loading scenarios, we systematically elaborate
naturally in response to the positive eigenvalues krþI of the combinations of transmural pressure and axial stress. To
Cauchy stress r which had been postulated to be the illustrate the influence of the mechanical loading situation
driving force of the remodeling process. Due to the inho- on the remodeling process, we display Fig. 9 to show the
mogeneous stress state across the radial direction, different final biological equilibrium stages for different stretches at
fiber angles arise in the individual layers. different pressure levels. For the sake of visibility, we have
Figure 8 depicts the final collagen fiber orientations at virtually peeled off the individual layers after the calcula-
three different radial locations representing the intima, the tion of the entire tube had been performed. Experimentally,
media and the adventitia. The results suggest that the the definition of these specific layers could be made
transmural pitch of the fiber orientation increases from the apparent by the use of stains differentiating smooth mus-
inner to the outer wall. The analytically predicted double- cles cells from collagen fibers. For pure stretching, as
helix architecture of the collagen fibers agrees very well depicted in Fig. 9 (left), the fibers of all layers are aligned
with experimental observations by Finlay et al. [16] and with the loading axis, i.e., l1 = r0 and l2 = l3 = 0 with
Holzapfel et al. [27, 30]. In addition, our results nicely n01 pointing in the axial direction. For pure pressure,
agree with the more recent numerical studies by Driessen as depicted in Fig. 9 (right), all fibers are oriented in
et al. [13] and Hariton et al. [25]. the circumferential direction, i.e., again, l1 = r0 and
l2 = l3 = 0, however, now with n01 pointing in the circum-
ferential direction. These computational results agree well
with the experimental findings of human brain arteries by
Finlay et al. [16] who reported a significant realignment of
collagen with increasing distending pressure for all arterial
layers and a recruitment towards the circumferential
direction. Intermediate stages of combined stretch and
pressure loading, as shown in the three images between the
left and the right image of Fig. 9, reveal the usually
observed helical fiber arrangement. Based on these results,
it is obvious to suggest that the collagenous architecture
strongly depends on the mechanical environment.
In the biomechanics community there is an ongoing
debate over which quantity drives the remodeling process.
Fig. 8 Tube-like artery subject to axial stretch and internal pressure
s Final stage of the remodeling process with collagen fiber To this end, we analyze the influence of different driving
orientations projected on the individual layers forces of tensorial nature. For purpose of comparison with

123
J Mater Sci (2007) 42:8811–8823 8821

Fig. 9 Tube-like artery subject

to axial stretch and internal % % % % %
pressure s Variations of
collagen fiber orientations in
response to different stretch-to-
pressure ratios with colors
representing the individual
collagen fiber angles

the (spatial) Cauchy stress r used in the previous examples, Finally, we elaborate the difference of a stress and
we now elaborate the use of the (material) second Piola– strain-based remodeling criterion, as used by, e.g., Driessen
Kirchhoff stress S ¼ JF
1  r  F
t ; as the representative et al. [13], Menzel [46] and Humphrey [32], Hariton et al.
material stress measure. Figure 10 compares now the out- [25], respectively. To this end, we compare the outcome of
come of a remodeling process driven by spatial and two different remodeling processes based on either the
material stresses. The diagram illustrates the fiber orien- Cauchy stress r as the characteristic spatial stress measure,
tation angle over the radius. Variations in fiber orientation see Fig. 11 (left), or on the Finger deformation tensor b as
seem to be slightly more pronounced when using the sec- the characteristic spatial strain measure, see Fig. 11 (right).
ond Piola–Kirchhoff stress as the driving force. Neverthe- Both the curves as well as the plots of the structure dem-
less, for this particular problem, it appears that differences onstrate that strain-based remodeling is much less sensitive
are rather marginal. We suggest that the Cauchy stress, to the radial variation. The difference between an almost
which is the true stress experienced by the deformed circumferential arrangement at the luminal side and a
structure, is the more reasonable choice. rather helical structure towards the outermost layer is less

Fig. 10 Tube-like artery

subject to axial stretch and 40
internal pressure s Collagen 35
fiber orientation versus radial
30
position (left) and outcome of
angle

spatial versus material stress- 25

based remodeling (right) 20
15
10
5
0
2.2 2.4 2.6 2.8 3 3.2 3.4
radius

Fig. 11 Tube-like artery 40

subject to axial stretch and
35
internal pressure s Collagen
fiber orientation versus radial 30
position (left) and outcome of 25
angle

spatial stress versus spatial

20
strain-based remodeling (right)
15

10

0
2 2.2 2.4 2.6 2.8 3 3.2 3.4
radius

123
8822 J Mater Sci (2007) 42:8811–8823

pronounced for the strain driven case. This finding is, of these more complex inhomogeneous (living) structures, the
course, rooted in the non-linear nature of the force- remodeling algorithm was embedded in a non-linear finite
displacement behavior of the individual chains, recall element program, the equations were linearized consis-
Fig. 1. While the strain varies slowly close to the locking tently and solved with an incremental iterative Newton–
stretch, corresponding stresses may vary significantly. Raphson solution strategy. Starting with initially random
Accordingly, the stress-based remodeling criterion predicts fiber orientation, the algorithm generated collagenous
a wider variation in collagen fiber angles than the strain architectures which qualitatively resembled experimental
driven remodeling algorithm. Although for hard tissues, observations. In all computational examples, the suggested
such as bone, several studies including the one by Cowin algorithm convinced through its remarkable stability and
[9] suggest that strain is the critical mechanical factor in robustness. What remains, however, is the realization of
growth and adaptation, we postulate that stress is a more quantitative validations of the suggested theory and its
reasonable choice as the driving force in the context of soft algorithmic realization. In addition to the collagen fiber
tissues such as arteries, tendons or ligaments. This obser- orientation, which could eventually be determined through
vation is along with Taber and Humphrey [52] who suggest techniques such as microscopy, histology, magnetic reso-
that stress and not strain correlates well with growth in nance diffusion tensor imaging, the chain number density
arteries. and the turnover rate remain to be classified. The latter is
assumed to be of particular relevance in the context of
improving cardiovascular surgery and predicting patient-
Discussion specific remodeling in response to medical treatment.
Finally we would like to point out once again, that
A continuum theory of remodeling for soft biological tis- although the reorientation approach outlined in the present
sues has been proposed with a particular focus on cardio- manuscript is motivated by micromechanical consider-
vascular tissue. In this context, remodeling is attributed to ations, it is still rather phenomenological to most extend. In
the collagen fiber reorientation as a natural consequence of contrast to purely invariant based formulations, it intro-
changes in the mechanical loading environment. Hence, in duces material parameters such as the contour length or the
our approach the optimal configuration of collagen fibers persistence length which have a clear physical interpreta-
depends solely on the external loading acting upon the tion. The computational mechanics community would thus
living tissue. Based on a homogenization strategy from the certainly classify the model as bottom-up or rather micro-
molecular microscale via the mesoscale of the extracellular mechanically based and non-phenomenological. In the cell
matrix to the macroscale of the overall tissue, we derived a biology community, however, the same model would most
statistical mechanics-based chain network model governed probably be considered rather phenomenological or top-
by a limited set of physically motivated material parame- down since it does not explicitly address cell level
ters. In addition to the two (classical) Lamé constants, five phenomena like mechanotransduction or the biochemical
additional parameters sufficed to characterize the highly origin of the remodeling process as such. Nevertheless, to
non-linear, exponentially stiffening of the collagen mor- date it seems unmanageable to simulate large tissue
phology: the contour length, the persistence length and the structures or organs and yet at the same time to account for
initial end-to-end length of the collagen chains, the chain every single biochemical phenomenon individually. To the
number density accounting for the degree of anisotropy and most extend, these complex mechanisms are not even fully
the turnover rate. The concept of chain network models understood at this point.
was applied to transmit information from the molecular Successful constitutive models require an understanding
level to the extracellular matrix level. In contrast to exist- of the functional interactions between the key components
ing remodeling theories which are based on complex of cells up to the organs, and how these interactions change
rotational updates of characteristic microstructural direc- from a physiological to a pathological state. Such infor-
tions, our theory essentially captures remodeling in the mation resides neither in the individual genome nor in the
form of changes of the dimensions of this representative protein. This information is contained in the interactions of
element. By means of simple model problems, we were proteins with cellular, organ and system structures. The
able to show that the theory indeed succeeded in charac- identification of these interactive relationships is clearly
terizing mechanically introduced remodeling of collagen within the focus of intense current research. Continuum
fibers. models like the one presented herein will certainly benefit
Finally, we elaborated two biomechanically relevant from gradually incorporating more and more information
boundary-value problems, a cylindrical tendon subject to from the protein, subcellular and cellular level in the future
uniaxial tension and a tube-like artery loaded by an axial to define precisely how mechanical forces translate into
stretch in combination with different internal pressures. For chemical signals that initiate the process of remodeling.

123
J Mater Sci (2007) 42:8811–8823 8823

References 27. Holzapfel GA, Gasser TC, Stadler M (2002) European J Mech A
Solids 21:441
1. Alberts B, Johnson A, Lewis J, Raff M, Roberts K, Walter P 28. Holzapfel GA, Ogden RW (2003) Biomechanics of soft tissue in
(2002) Molecular biology of the cell, 4th edn. Garland Science cardiovascular systems CISM courses and lectures no 441.
Taylor Francis Group, New York Springer Verlag, Wien–New York
2. Arruda EM, Boyce MC (1993) J Mech Phys Solids 41:389 29. Holzapfel GA, Ogden RW (2006) mechanics of biological tissue.
3. Bischoff JE, Arruda EM, Grosh K (2002) J Appl Mech 69:570 Springer, Berlin–Heidelberg–New York
4. Bischoff JE, Arruda EM, Grosh K (2002) J Appl Mech 69:198 30. Holzapfel GA, Stadler M, Schulze-Bauer CAJ (2002) Ann Bio-
5. Boyce MC (1996) Rubber Chem Technol 69:781 med Eng 30:753
6. Boyce MC, Arruda EM (2000) Rubber Chem Technol 73:504 31. Huang H, Kamm R, Lee RT (2003) Am J Physiol Cell Physiol
7. Bustamante C, Bryant Z, Smith SB (2003) Nature 421:423 287:C1
8. Bustamante C, Smith S, Marko JF, Siggia ED (1994) Science 32. Humphrey JD (2001) J Biomech Eng 123:638
265:1599 33. Humphrey JD (2002) Cardiovasular solid mechanics. Springer
9. Cowin SC (1984) Calc Tissue Int 36:S99 Verlag, Berlin–Heidelberg–New York
10. Cowin SC (1995) J Elast 34:45 34. Ingber DE (2003) Ann Intern Med 35:564
11. Driessen NJB, Bouten CVC, Baaijens FPT (2005) J Biomech Eng 35. Klein-Nulend J, Bacabac RG, Mullender MG (2005) Pathol Biol
127:494 (Paris) 53:576
12. Driessen NJB, Peters GWM, Huyghe JM, Bouten CVC, Baaijens 36. Kratky O, Porod G (1949) Recl Trav Chim 68:1106
FPT (2003) J Biomech 36:1151 37. Kuhl E, Garikipati K, Arruda EM, Grosh K (2005) J Mech Phys
13. Driessen NJB, Wilson W, Bouten CVC, Baaijens FPT (2004) J Solids 53:1552
Theor Biol 36:53 38. Kuhl E, Maas R, Himpel G, Menzel A (2006) Biomechanics and
14. Elbischger PJ, Bischof H, Holzapfel GA, Regitnig P (2005) In: modeling in mechanobiology, pp available online, DOI 10.1007/
Suri JS, Yuan C, Wilson DL, Laxminarayan S (eds) Plaque s10237-006-0062-x
imaging: pixel to molecular level, vol 113. Studies in Health 39. Kuhl E, Menzel A, Garikipati K (2006) Phil Mag 86:3241
Technology and Informatics, IOS Press, p 97 40. Lehoux S, Castier Y, Tedgui A (2006) J Intern Med 259:381
15. Elbischger PJ, Bischof H, Regitnig P, Holzapfel GA (2004) 41. Leung DYM, Glagov S, Mathews MB (1976) Science 191:475
Pattern Anal Appl 7:269 42. Leung DYM, Glagov S, Mathews MB (1977) Circ Res 41:316
16. Finlay HM, Mc Cullough L, Canham PB (1995) J Vasc Res 43. Lubarda VA, Hoger A (2002) Int J Solids Struct 39:4627
32:301 44. Marko JF, Siggia ED (1995) Macromolecules 28:8759
17. Flory PJ (1969) Statistical mechanics of chain molecules. John 45. Menzel A (2006) In: Holzapfel GA, Ogden RW (eds) IUTAM
Wiley Sons, Chichester–New York Symposium mechanics of biological tissue. Springer Verlag, p 91
18. Freed AD, Einstein DR, Vesely I (2005) Biomech Model 46. Menzel A (2005) Biomech Model Mechanobiol 3:147
Mechanobiol 4:100 47. Mofrad MRK, Kamm R (eds) (2006) Cytoskeletal mechanics
19. Garikipati K, Arruda EM, Grosh K, Narayanan H, Calve S (2004) models and measurements. Cambridge University Press
J Mech Phys Solids 52:1595 48. Nerem RM, Seliktar D (2001) Annu Rev Biomed Eng 3:225
20. Garikipati K, Olberding JE, Narayanan H, Arruda EM, Grosh K, 49. Rodriguez EK, Hoger A, Mc Culloch AD (1994) J Biomech
Calve S (2006) J Mech Phys Solids 54:1493 27:455
21. Gasser TC, Ogden RW, Holzapfel GA (2006) J R Soc Interface 50. Stopak D, Harris AK (1982) Dev Biol 90:383
3:15 51. Sun YL, Luo ZP, Fertala A, An KN (2002) Biochem Biophys Res
22. Gleason RL, Humphrey JD (2004) J Vasc Res 41:352 Commun 295:382
23. Gleason RL, Humphrey JD (2005) Math Med Biol 22:347 52. Taber LA, Humphrey JD (2001) J Biomech Eng 123:528
24. Hagerman PJ (1988) Annu Rev Biophys Biophys Chem 17:265 53. Treloar LRG (1975) The physics of rubber elasticity. Clarendon
25. Hariton I, de Botton G, Gasser TC, Holzapfel GA (2006) Bio- Press, Oxford
mechanics and modeling in mechanobiology, pp available online, 54. Vianello M (1996) J Elast 42:283
DOI 10.1007/s10237-006-0049-7 55. Wang JH, Thampatty BP (2006) Biomech Model Mechanobio 5:1
26. Himpel G, Menzel A, Kuhl E, Steinmann P (2007) Int J Num
Meth Eng, in press

123