Fracture mechanics of

protein materials
Proteins are the fundamental building blocks of a vast array of
biological materials involved in critical functions of life, many of which
are based on highly characteristic nanostructured arrangements of
protein components that include collagen, alpha helices, or beta sheets.
Bone, providing structure to our body, or spider silk, used for prey
procurement, are examples of materials that have incredible elasticity,
strength, and robustness unmatched by many synthetic materials.
This is mainly attributed to their structural formation with molecular
precision. We review recent advances in using large-scale atomistic and
molecular modeling to elucidate the deformation and fracture mechanics
of vimentin intermediate filaments (IFs), which are hierarchical selfassembled protein networks that provide structure and stability to
eukaryotic cells. We compare the fracture and failure mechanisms of
biological protein materials (BPMs) with those observed in brittle and
ductile crystalline materials such as metals or ceramics. Our studies
illustrate how atomistic-based multiscale modeling can be employed to
provide a first principles based material description of deformation and
fracture, linking nano- to macroscales.
Markus J. Buehler* and Theodor Ackbarow
Laboratory for Atomistic and Molecular Mechanics, Massachusetts Institute of Technology, Department of Civil and Environmental Engineering,
77 Massachusetts Ave. Room 1-272, Cambridge, MA, 02139, USA
*E-mail: [email protected]

46

When materials are deformed, they display a small regime in

irreversible and remains even after the load is removed. This is

which deformation is reversible or elastic13. Once the forces

referred to as the plastic regime13. Plastic deformation is typically

on the material reach a certain level, deformation becomes

followed by fracture, when the material breaks and fails. Many

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

ISSN:1369 7021 Elsevier Ltd 2007 Open access under CC BY-NC-ND license.

Fracture mechanics of protein materials

REVIEW

Fig. 1 Overview of different material scales, from nano to macro, here exemplified for collagenous tissue59. BPMs such as collagen, skin, bone, spider silk, or
cytoskeletal networks in cells feature complex, hierarchical structures. The macroscopic mechanical material behavior is controlled by the interplay of properties
throughout various scales. In order to understand deformation and fracture mechanisms, it is crucial to elucidate atomistic and molecular mechanisms at each scale.

materials, including metals, ceramics, polymers, and biological

tissue show this generic behavior13. However, the details of the

(a)

response to mechanical load depend on the atomic and molecular

makeup of the material; from nano- to macroscales (see Fig. 1).
For example, ductile metals such as Cu or Ni can rather easily
undergo large permanent (or plastic) deformation without breaking3.
On the other hand, brittle materials like glass or Si cannot easily be
deformed, but instead fracture rapidly once the applied load exceeds a
threshold value1. Conversely, BPMs such as the cells cytoskeleton4 or

(b)

collagen networks in tendon or bone510 represent intriguing protein

networks that can dynamically adapt to load application by selforganization and self-arrangement. They develop stronger filaments
when needed and disposing of those that do not contribute to the
strength, making the material utilization more efficient and robust
against failure overall.

(c)

Understanding the origin of deformation and fracture of materials

has fascinated generations of material scientists. Currently, a major
challenge is the elucidation of mechanisms in increasingly complex
materials materials that consist of multiple components or
multiple hierarchies, or those whose atomic nano- and microstructures
contain a concurrent interplay of a variety of chemical bonds. An
important concept in understanding the deformation and failure
properties of materials is the underlying fundamental atomic
mechanisms1,3,11, as illustrated in Fig. 2 for ductile materials3, brittle
materials1, and BPMs.
While the basic deformation mechanisms of crystalline solids
are relatively well understood, analogous mechanisms have only

Fig. 2 Overview of the deformation and fracture behavior of different classes of

materials, including (a) ductile materials3, (b) brittle materials1, and (c) BPMs.
Each subplot shows a multiscale view of associated deformation mechanisms.
In ductile materials, deformation is mediated by the creation of dislocation
networks; each dislocation represents localized shear of the atomic lattice. In
brittle materials, fracture occurs by the spreading of cracks, which is mediated
by continuous breaking of atomic bonds. In BPMs, a complex interplay of
different protein structures controls the mechanical response. At the ultrascale,
the unfolding of individual protein molecules by rupture of hydrogen bonds
(HBs) represents the most fundamental deformation mechanism.

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

47

REVIEW

Fracture mechanics of protein materials

these materials is mediated by intermolecular slip, the unfolding of

Predictive computational experiments of

deformation and fracture

proteins12,13, the breaking of intermolecular cross-links14,15, or the

Materials failure processes begin with the erratic motion of individual

stretching of convoluted protein chains, as has been demonstrated in

atoms around flaws or defects within the material. These evolve into

multiscale simulations for collagenous materials and bone5,6,10,15,16 for

the formation of macroscopic fractures as chemical bonds rapidly

instance.

rupture, eventually compromising the integrity of the entire structure.

recently been discovered in BPMs. Permanent plastic deformation in

Because of the multiscale hierarchical structure of these materials,

Thus the behavior of chemical bonds under a large stretch on a small

different deformation mechanisms may occur at each scale, while

scale controls how structures respond to mechanical load and fail on

the inter- and intrahierarchal interactions might be of competing or

much larger material scales, as has, for instance, been demonstrated for

reinforcing character. The most fundamental deformation mode is,

model materials and Si1821.

however, often the breaking of weak hydrogen bonds (HBs). Even

Fracture mechanisms in brittle and ductile materials are

though HBs are 100 to 1000 times weaker than covalent bonds,

representative examples of an intrinsic multiscale problem that cannot

HBs are the most important type of chemical bonds that hold

be understood by considering one scale alone (see Figs. 1 and 2).

together proteins, assemblies of proteins, and that control their

Experimentation, simulation, and development of theories must

adhesion behavior, and thereby play a vital role in controlling many

therefore consider a complex interplay of mechanisms at several

biological processes. Up to now, neither a systematic classification nor

scales. In particular in hierarchical BPMs, the development of a rigorous

fundamental theories exist for the different deformation and fracture

understanding of deformation depends critically on the elucidation of

mechanisms in BPMs.

the deformation mechanism at each scale, and how these mechanisms

The mechanical properties of BPMs have wide ranging implications

interact dynamically across the scales.

for biology. In cells for instance, mechanical sensation is used to

transmit signals from the environment to the cell nucleus or to control

Hierarchical multiscale modeling

tissue formation and regeneration4,17. The structural integrity and

Multiscale modeling is a particularly useful approach to gain insight

shape of cells is controlled by the cells cytoskeleton, which resembles

into these complex problems. In order to allow the best resolution

an interplay of complex protein structures and signaling cascades

at any length and time scale, a set of computational methods is

arranged in a hierarchical fashion. The transfer of concepts observed

integrated seamlessly, enabling one to bridge scales from nano to

in biology into technological applications and new material design a

macro (Fig. 3).

field referred to as biomimetics remains a significant challenge with,

Atomistic simulation, or molecular dynamics (MD)22, provides a

however, a potentially enormous payoff. In particular, the combination

fundamental view on materials deformation describing the patterns

of nanostructural and hierarchical features into materials developments

of fracture, yield, diffusion, and other mechanisms at resolutions that

could lead to breakthroughs in the design of new materials.

cannot yet be reached by experiments. After careful validation of these

Fig. 3 Schematic of the concept of hierarchical multiscale modeling. Hierarchical coupling of different computational tools can be used to traverse throughout
a wide range of length and time scales. Such methods provide a fundamental insight into deformation and fracture phenomena across various time and length
scales. Handshaking between different methods enables one to transport information from one scale to another. Eventually, results of atomistic, molecular, or
mesoscale simulation may feed into constitutive equations or continuum models. While continuum mechanical theories have been very successful for crystalline
materials, modeling the behavior of BPMs requires statistical theories, e.g. the extended Bell theory reviewed here. Experimental techniques such as atomic force
microscopy (AFM), molecular force spectroscopy (MFS), nanoindentation, or optical tweezers now overlap into atomistic and molecular approaches, enabling
direct comparison of experiment and simulation.

48

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

Fracture mechanics of protein materials

computational models with experiments, atomistic and multiscale

REVIEW

(a)

modeling have predictive power23. Predictive multiscale simulation

could play an important role in science, engineering, and materials
design in the coming decades.

Experimental techniques to probe ultrasmall scale

material mechanics

(b)

(c)

Recent advances in experimental techniques further facilitate analyses

of ultrasmall-scale material behavior. For instance, techniques such as
nanoindentation, optical tweezers, or atomic force microscopy (AFM)
can provide valuable insight to analyze the molecular mechanisms in
a variety of materials, including metals, ceramics, and proteins13,2427.
A selection of experimental techniques is summarized in Fig. 3,
illustrating the overlap with multiscale simulation methods.

Basics of atomistic modeling

The basic concept behind atomistic simulation via MD is to calculate
the dynamic trajectory of each atom in the material, by considering
their atomic interaction potentials and solving each atoms equation of
motion according to F = ma (where F is the force vector, m the atomic
mass, and a the acceleration)22. Numerical integration of this equation
by considering proper interatomic potentials enables simulation of

Fig. 4 (a) Tensile deformation of a Ni nanowire with a surface crack. The

geometry is a nanowire of face-centered cubic Ni with a small surface
crack, simulated by considering a cubic volume with 50 nm side length and
periodic boundary conditions in the wire direction (x-direction). The atomic
interactions are modeled using EAM28 adapted for Ni29. (b) Upon increase
of the tensile load to a few percent strain, dislocations start to emerge at the
small surface crack because of large local shear stresses. (c) Quickly, the entire
crystal is filled with a complex entanglement of dislocations. The increasing
entanglement of dislocations makes it more difficult for new dislocations to
nucleate and propagate into the crystal, thus making the material stronger.
This effect is known as strain hardening3.

a large ensemble of atoms that represents a larger material volume,

albeit typically limited to several nanoseconds. The availability of

materials to undergo significant shape change without breaking3 (see

interatomic potentials for a specific material is often a limiting factor in

Fig. 2a).

the applicability of this method.

Provided suitable interatomic potentials are available, MD is capable

A case study of large-scale atomistic modeling of deformation of

a Ni nanowire using the embedded atom method (EAM)28 adapted

of directly simulating a variety of materials phenomena, for instance

for Ni29 is presented in Fig. 4. The EAM has become a widely used

the response of an atomic lattice to applied loading under the presence

tool in the computational materials science of metals. For instance,

of a crack-like defect, or the unfolding mechanisms of proteins.

MD simulations have significantly contributed to the understanding of

In the following two sections, we briefly review atomistic simulation

deformation of nanocrystalline metals3032.

applied to ductile and brittle materials. We then move on to discuss

approach to address the fundamental deformation mechanisms in

Brittle materials: failure by spreading of

fractures

vimentin IF networks. We conclude with a discussion and outlook, in

The fracture of brittle solids proceeds by the breaking of atomic bonds,

particular highlighting similarities and differences between the classes

leading to the formation of new material surfaces. In contrast to ductile

of materials discussed here. We note that we primarily review specific

solids, the brittle character of materials arises from the energetic

research results from our own group.

favorability of bond rupture over shearing of the material1,3. This can

the behavior of BPMs, exemplifying a theoretical and numerical

be quantified by considering ratios of the stacking fault energies and

Deformation of ductile materials: dislocationmediated plasticity

surface energy, as discussed elsewhere33,34.

Ductile materials are characterized by their ability to undergo

fails because of the spreading of a large number of fractures. The

significant plastic deformation without breaking3. The fundamental

presence of a crack is vital for this process, since it provides a location

mechanism for this behavior is the formation of dislocations, which

with locally large stresses. Fig. 2b depicts the fundamental fracture

represent a localization of shear displacement in the atomic lattice.

mechanism of a brittle material, indicating the large stresses at the tip

Instead of shearing an entire crystal homogeneously, a dislocation

of the crack that cause repeated rupture of atomic bonds.

represents a localization of shear that propagates through the

Once the applied load exceeds a critical value, a brittle solid

A representative case study of atomistic modeling of the fracture

material, often nucleated at crack-like defects that represent locations

of Si is shown in Fig. 5, with a series of results obtained using a

of shear stress concentration3. Accumulation of dislocations enables

multiparadigm approach20 based on the combination of the ReaxFF

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

49

REVIEW

Fracture mechanics of protein materials

(a)

(a)

(b)

(c)
(b)

Fig. 6 Single-molecule pulling experiments carried out on a coiled-coil (CC)

protein structure. (a) An experimental setup based on AFM, and (b) a steered
molecular dynamics (SMD) analogue. In the SMD approach, the end of the
molecule is slowly pulled at a pulling velocity v vector. This leads to a slowly
increasing force F = k(v t x) (k = 10 kcal/mol/2), where t is the time and x
is the current position vector of the end of the molecule (F(x), schematically
shown in (c)). Both approaches, AFM and SMD, lead to force displacement
information. In addition to the F(x) curve, SMD provides detailed information
about associated atomistic deformation mechanisms. Because of the time scale
limitations of MD to several nanoseconds, there is typically a large difference
between simulation and experiment with respect to pulling rates. Whereas MD
simulations are limited to pulling rates of 0.1 ms1, experimental rates are six
to eight magnitudes smaller. This requires additional consideration in order to
interpret MD results in light of experimental findings.

mechanisms. Understanding these mechanisms poses a very

Fig. 5 Multiparadigm simulation of dynamic fracture of Si20,21. (a) The slab
geometry is approximately 50 nm by 100 nm in plane-strain geometry (outof-plane thickness is one unit cell), with an initial crack length a 17 nm. This
model describes the fracture mechanics of Si by a combination of a simple
Tersoff force field in regions far away from bond rupture events (blue region),
with the ReaxFF reactive force field used to describe the rupture processes at
the crack tip (yellow region). (b) The results of a computational experiment
with a slowly increasing mode I load, for the (111) oriented crystal. The strain
rate is 0.0005% strain increment per 1.5 fs integration step. The initial, static
regime is followed by a short period of crack growth during which a perfectly
flat, mirror-like surface is generated. Crack propagation becomes increasingly
erratic until the entire crystal is fractured.

challenging set of problems. Furthermore, the application of the insight

from such studies to address significant biological and biomedical
questions or to suggest new biomimetic material designs are exciting
intellectual and interdisciplinary opportunities.

Atomistic modeling of protein materials

All MD simulations reviewed here have been performed at a
temperature of 300 K (NVT ensemble), with temperature control
using a Berendsen thermostat22 and a time step of 1 fs. In all cases,
the molecular structure obtained from the Protein Data Bank is

reactive force

field20,35,36

with a Tersoff

potential37.

This method is

solved completely in a water skin that encompasses the entire

based on the strategy of integrating a set of computational tools, as

protein structure. Careful energy minimization and finite temperature

outlined in Fig. 3. Atomistic simulations of such fracture processes

equilibration of all structures are simulated before the protein is

provide a fundamental insight into the details of crack extension20. For

deformed. We use visual molecular dynamics (VMD) for visualization of

instance, the

simulations21

reveal that the steady-state fracture speed

has a minimum threshold speed of approximately 2

kms1,

a finding in

agreement with earlier experimental studies38.

protein structures39 and MD simulations are carried out using NAMD40

with the CHARMM22 force field with explicit TIP3 water41.
To apply forces to the molecule, we use steered molecular dynamics
(SMD)12. The SMD technique is equivalent to attaching one end of

Hierarchical BPMs

a virtual harmonic spring to the end of the system and pulling at

The two previous examples highlight two prominent mechanisms of

the virtual atom on the other end of the spring, with a pulling speed

deformation in crystalline materials: dislocation nucleation and brittle

v. The SMD method mimics an experiment where one end of the

crack extension. In this section, we discuss how the understanding that

molecule is fixed (e.g. on a Au surface), while the other end is pulled

exists for crystalline systems can be extended to biological materials.

with an AFM cantilever tip (see Fig. 6). The molecular strain is defined

BPMs have a significantly different molecular structure than

crystalline materials, which induces a variety of novel deformation

50

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

as = (x x0)/x0, where x0 is the undeformed length and x is the

deformed length of the protein structure.

Fracture mechanics of protein materials

REVIEW

Extended Bell theory: theoretical nanomechanics of

protein materials

processes exist for a protein structure, each of which has a specific

The key to comprehend the mechanical response of BPMs across

to the underlying bond-breaking mechanisms and rearrangements in

different time and length scales is to understand the rupture mechanics

the protein structure. Several theories exist that describe competing

of HBs under laterally applied load. Typically, a variety of unfolding

processes arising from mechanically induced instabilities of protein

reaction pathway and an associated energy barrier, mostly related

Extended Bell theory

Whereas continuum theories are suitable to describe the deformation
of macroscopic structures, statistical concepts must be employed
to accurately model the mechanical behavior of protein structures.
Bells theory is one of the most widely used models to describe the
statistical nature of bond breaking.
In Bells theory4246, the off rate is the product of the natural
bond vibration frequency 0 and the quasi-equilibrium likelihood of
reaching the transition state with an energy barrier Eb that is reduced
by mechanical energy F xb cos(), where F is the applied force, xb
is the distance between the equilibrated state (minimum of the well)
and the transition state, and is the angle between the direction of
the reaction pathway of bond breaking (x-direction) and the direction
of applied load F (see Fig. 7). The angle can be determined by
analyzing the molecular geometry. The off rate is given by:

(E F xb cos())
= 0 exp b

kb T

(1)

and describes how often a bond is broken per unit time (it equals the
reciprocal of the lifetime of a bond). The natural vibration frequency

Fig. 7 Statistical theory to predict the bond rupture mechanics. The graph
depicts the energy as a function of deformation along a deformation variable
(upper plot), along a particular pathway that leads to bond rupture; where F
is the applied force and xb is the displacement in the direction of the applied
force. In the schematic, three HBs (indicated by the red color, lower plot)
break simultaneously. Thus, xb corresponds to the lateral displacement
that is necessary to overcome the bond breaking distance of a HB, in the
particular atomistic geometry present in the CC. Given that xb* is the
distance to break a single HB, the distance xb* = xbcos denotes the lateral
displacement at bond breaking, with the angle as the angle between
pulling direction and orientation of the HB inside the molecule.

of a bond46 is 0 1 13 s-1.
with 0 as the natural bond breaking speed (when no load is applied),

However, eq 1 does not describe the dependence of the pulling

speed (the controlled parameter in experiment and MD simulation) at
which a bond breaks because of the applied pulling force. Instead, it
only provides an estimate of the time scale at which the bond will be
broken. In order to overcome this limitation, we modify eq 1 based on
the following idea: the speed at which a bond is broken equals the

defined as:

E
v0 = 0 xb exp b
kb T

(4)

This modified framework enables us to study the dependence

distance that needs to be overcome in order to break the bond (xb)

between the unfolding force and the bond breaking speed or to

divided by the time for bond breaking. Consequently, is the product

calculate the average force at which a bond breaks, at a certain

of xb. This leads to the following equation for the bond breaking

pulling rate. We can rewrite eq 3 as:

speed47:

(E F xb cos())
v = 0 xb exp b

kb T

F (v ) =
(2)

kb T
kb T
ln v
ln v0 = a ln v + b
xb cos()
xb cos()

(5)

where a = kb T/(xb cos ) and b = kb T/(xb cos ) ln 0.

Eq 5 predicts that the unfolding force depends logarithmically on
the pulling speed in a nonequilibrated system. We note that it

This equation can be rewritten in the following way:

contains two parameters a and b, which can be calculated from the

F xb cos()
v = v0 exp

kb T

parameters xb and Eb at a certain temperature and angle. The concept

(3)

is summarized in Fig. 7.

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

51

REVIEW

Fracture mechanics of protein materials

structures. These concepts stem primarily from a theory originally

postulated by Bell in 19784246.
Eq. 5 (see text box) provides an immediate link between the

Vimentin IFs have a variety of functions on several hierarchical

levels. On the cellular scale, vimentin networks act primarily as the
security belts of the cell51. Because of their architecture, the flexible

molecular geometry (via the angle ), the pulling rate, and the pulling

networks are very soft at small deformations and pulling rates, leading

force F that is necessary to lower the energy barrier in such a way that

to invisibility and nonresistance during cell movement. Conversely,

the HBs can be broken at the applied pulling velocity.

very stiff behavior is observed at large deformation and high

A possible strategy to determine the dependence of the unfolding

deformation rates, ensuring their function on the cellular as well as

force F on pulling speed, associated mechanisms, and energy barriers

on the tissue level49. The mechanical role of IFs is particularly evident

is to carry out MD simulations at different pulling rates, measure

in diseases in which the loss of mechanical function and integrity

the dependence of unfolding force on the pulling rate, and fit the

of various tissues is associated with intermediate-filament-protein

parameters a and b to this data. This information can then be used to

mutations, as reviewed elsewhere52,53.

calculate the parameters xb and Eb and thereby reconstruct the free

energy landscape.

Recently, additional functions have been found on the subnetwork

filament level, which control the location, shape, and stability of cell
organelles (e.g. mitochondria or golgi), their function, and the protein

Biological role of vimentin IFs

targeting process54. Yet other functions exist on the molecular level,

Together with globular protein microtubules (MTs) and microfilaments

consisting of different regulation mechanisms such as cell signaling

(MFs), IFs are one of the three major components of the cytoskeleton

(e.g. transcriptional effects, mechanotransduction), or associated

in eukaryotic cells48. The cytoskeleton plays a critical role not only in

protein organization (e.g. plectin or chaperones)54,55.

determining the shape and the mechanical properties of the cell, but
is also vital for many additional functions including protein synthesis,

Large deformation behavior of vimentin IFs

cell motility, and cell division or wound healing4850. Here we focus on

Here we focus on the large deformation behavior of vimentin

the mechanical behavior of vimentin IFs. Fig. 8 displays the hierarchical

proteins4,48,56,57. When cells and thus entire IF networks are deformed,

structure of vimentin networks, from the molecular to the cellular level.

each vimentin filament undergoes a tensile deformation, which

Fig. 8 Hierarchical structure of the IF network in cells. Through carefully following the various steps of assembly72,77 it has been shown that dimers associate to
fibrils, which form the second level of the hierarchy. In vivo, these fibrils can reach a length of up to several microns and consist of 16 dimers in cross section. The
third level of hierarchy consists of three-dimensional IF networks inside the cell, reinforcing the plasma membrane49,60,61. Inside the network, IF-associated proteins
such as plectins generate the connection between individual IFs as well as between other cytoskeletal components. The characteristic loading condition of full length
filaments is tensile loading. Because of this tensile load, each dimer is subject to a tensile load if the cell undergoes large deformation.

52

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

Fracture mechanics of protein materials

REVIEW

Fig. 9 Geometry of the molecular building block of vimentin. The dimers, approximately 45 nm long, are the elementary building blocks of vimentin IFs. A dimer
consists of a head, tail (plotted in red), and an elongated rod domain that is divided into four alpha-helical CCs (1A, 1B, 2A, 2B) connected through linkers L1, L12,
L2 (also red)78. We also indicate the segments of the dimer that correspond to the CC without stutter (model A) and the CC with stutter defect (model B).

is carried by individual dimers at the molecular level. Therefore, a

with = 23 10.2, suggesting that the unfolding mechanism is

detailed understanding of coiled-coil (CC) dimers and their mechanical

characterized by sequential breaking of single HBs47 (this is concluded

properties under small and large tensile deformations is vital to provide

since the energy for a single HB is in the range of 5 kcal mol1).

insight into the function and mechanisms of vimentin filaments and

networks49,5861. Fig. 9 depicts the molecular geometry of the CC

It has been found that deformation mechanisms depend on the

pulling speed47, and that there is a change in deformation mechanism

dimer. The following simulations are carried out by stretching a short

segment in the 2B domain of the dimer, as indicated in the schematic.

(a)

Fig. 10 depicts force-strain curves of an IF CC dimer, at different

pulling rates, obtained by MD simulations47,6264. The force-extension
curves depend strongly on the pulling velocity. Fig. 11a depicts the
unfolding force, which determines the force level during the second
regime, as a function of pulling speed. These results clearly confirm the
logarithmic dependence of the unfolding force on the pulling speed,
as predicted by the extended Bell theory (eq 5). MD simulations in the
fast pulling regime show that Eb = 5.59 kcal mol1 and xb = 0.185 ,

(b)

Fig. 10 Force-strain curves for mechanical tensile deformation of a CC

vimentin dimer47. The continuous curves represent results obtained from a
theoretical model based on the Bell theory that enables one to predict the
unfolding mechanics for a wide range of pulling velocities. The first regime
represents elastic deformation, up to approximately 1025% tensile strain.
This regime is followed by a plateau region during which unfolding of the
CC structure occurs at approximately constant force level. The last regime
displays significant strain hardening, during which the coiled superhelix is
lost and the protein backbone is stretched, leading to a significant increase in
stiffness.

Fig. 11 (a) Unfolding force of a vimentin dimer (at the angular point), as a
function of the pulling velocity47. All simulation results display a logarithmic
dependence of the unfolding force on the pulling velocity, in agreement with
the predictions by the extended Bell theory. (b) Unfolding wave speed as a
function of lateral pulling speed. The unfolding wave is predicted to disappear
at a pulling velocity of approximately 0.161 ms1; this leads to a change
of strain distribution from localization of strain at the ends of the molecule
to a homogeneous strain along the molecular axis. This induces a different
molecular unfolding mechanism, which is characterized by breaking of three
HBs, at a lateral applied displacement of 1.25 . The second regime is
plotted in (a) as a red line. The predictions based on this model agree well with
experimental results.

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

53

REVIEW

Fracture mechanics of protein materials

(a)

(b)

Fig. 12 Visualization of the atomistic-scale unfolding process of the CC structure. The yellow lines in (a) indicate the HB that forms between the backbone O atom
and the H atom connected to the N atom in the protein backbone, and which have broken in (b). The unfolding wave destroys these HBs, leading to loss of the
helical structure.

because of a change in strain distribution. At large pulling velocities,

the prediction according to this model in Fig. 11a. A comparison with

an unfolding wave propagates throughout the molecule, sequentially

experiments52,65 reveals reasonable agreement.

breaking HBs one by one. Physically, the unfolding wave represents a

Fig. 12 depicts details of the atomistic rupture mechanisms as

localization of strain that leads to the local rupture of HBs before the

the CC structure is unfolded. Fig. 13a displays the overall molecular

entire protein is equilibrated. As shown in Fig. 11b, the speed of this

unfolding dynamics. We note that the intersection of the unfolding

unfolding wave reaches zero at a pulling speed of 0.161

ms1.

Thus, at

force corresponding to the two modes is very close to the predicted

lower pulling rates, this localization of strain in a convolution does not

transition speed of 0.161 ms1. This provides an independent

occur since there is no unfolding wave and all HBs in a convolution

confirmation of the hypothesis47.

are stretched equally. Failure then occurs by simultaneous rupture of

three HBs in one arbitrary convolution of the alpha helix (AH).
Thus, in order to apply the extended Bell theory (eq 6) to the

Molecular defects in vimentin dimers: soft spots

A variety of discontinuities or defects exist that interrupt the CC

slow pulling regime, the parameters Eb and xb must be modified. We

periodicity locally without destroying the overall molecular structure.

multiply Eb by a factor of three to account for the fact that three

Skips are insertions of one residue into the heptad pattern (a

HBs rupture simultaneously instead of one, and use a value for xb of

characteristic seven residue repeat in CCs)66, stammers result

1.25 (corresponding to the transition distance of

HBs47).

We plot

(a)

through an insertion of three additional residues6769, and stutters

(b)

Fig. 13 Snapshots of the unfolding dynamics, (a) for a perfect CC, corresponding to model A in Fig. 9, and (b) the CC with the stutter domain, corresponding to
model B in Fig. 964. Unfolding proceeds sequentially, beginning from the end where the load is applied. It is apparent that the stutter represents a location where
unfolding starts as well (see the 2nd and 3rd snapshot from the top, where it is evident that the AH motif begins to disappear at the location of the stutter). The red
arrow indicates the location of the stutter. The structure is rendered using the VMD ribbons method; color is assigned according to the secondary protein structure
(thus loss of the AH structure is visible by change towards a white color). The initial length of both structures is 70 .

54

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

Fracture mechanics of protein materials

REVIEW

appear if four additional residues interrupt the heptad sequence70.

MD simulation results shown in Fig. 13b confirm this hypothesis, as

Presence of a stutter results in an almost parallel run of both AHs

they show that the CC unfolds at the stutter region.

Introducing the stutter into the dimer structure thus corresponds

without interrupting the CC geometry. To date, little is known

about the biological, mechanical, or physical reasons for the presence

to the deliberate addition of a weakening defect into the

of these molecular defects. It has been shown recently that mutations

molecular structure. The biological role of the stutter may be to

in desmin IFs related to different myopathies effectively leading to

provide a predefined, and, most importantly, controlled location where

muscle weakness result in additional stutters or even the creation

unfolding occurs under large deformation. As pointed out above,

of stammers in the dimer structure71. Vimentin dimers have a

vimentin IFs have their most significant role in large deformation

highly conserved stutter in the 2B segment of the dimer, as shown in

behavior.

Fig. 9.

Supermolecular deformation mechanisms of tetramers

Based on analysis of the geometry of the CC structure, the

extended Bell theory predicts that the stutter represents a molecular

The studies reviewed in the previous sections were performed at the

defect at which unfolding occurs at lower forces than in the rest of the

molecular level. Here we provide a simple analysis of the deformation

protein. Geometrically, the difference of a perfect CC and the segment

mechanisms at the next highest scale, describing the interaction

that includes the stutter region is the angle . In the perfect CC, the

between two dimers.

angle cc = 23 10.2, whereas within the stutter region, the angle

The analysis of the nanomechanics of a vimentin dimer reveals that

St = 16 8.5 (the angle St 0 since the HBs of a straight AH are

at pulling velocities that range between 1 x 108 ms1 and 1 x 106

slightly tilted in the direction of the twisting backbone).

ms1 (rates that appear in vivo and in experiments), the unfolding force

Provided that xb and Eb are equal that is, assuming the same

of the CC structure is predicted to range between approximately 50 pN

unfolding mechanism the extended Bell theory predicts that the

and 200 pN. This estimate for the unfolding force enables elucidation

structure with the larger angle is stronger (for 2 > 1, cos2 < cos1):

of the deformation mechanism of a tetramer structure, that is, the

combination of two dimers (see Fig. 8). Experimental analyses of the

F2 ( v ) =

cos 1
F1 ( v )
cos 2

assembly processes suggest that assembly of two dimers occurs via

the interaction of the head and tail domains with the CC structure, as

(6)

shown in Fig. 14a60,72.

Here we present a simple analysis considering the random coil

Therefore, in a system that consists of a combination of a perfect CC

and a stutter defect, the stutter defect represents the weakest segment,

domain of the head as an entropic chain, modeled using the worm-

which will unfold most easily under increasing applied tensile load. The

like chain (WLC) theory73 (see Fig. 14b). The model reveals that that

(a)

(b)

Fig. 14 (a) Assembly of two dimers into a tetramer (see also Fig. 9). (b) The interaction between two dimers is considered as a random coil polypeptide chain and
modeled with the WLC theory (see inset equation, where p 0.4 nm is the persistence length of a single polypeptide, x the end-to-end distance, and L the contour
length). We assume that 1/3 of the 80 residues are not bonded to either of the two molecules, and thus deform according to entropic elasticity, defining L. This
simple analysis reveals that that unfolding of the dimer begins when the shear displacement ranges between 70 and 80 (note that the interdimer shear force
induced by the two entropic chains combined reaches the range of the critical CC unfolding force between 25 pN and 100 pN, marked here with the red shaded
area). This simple analysis predicts a change in deformation mechanism from entropic dominated elastic stretching of the head/tail domains to uncoiling of the CC
structures.

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

55

REVIEW

Fracture mechanics of protein materials

unfolding of the dimer begins when the shear displacement between

hand, there is rupture of HBs on the dimer scale; on the other hand,

the molecules equals between 70 and 80 , or, equivalently, between

there are entropic effects between dimers on the supermolecular level.

12% and 14% tensile strain, because at this point the resistance

This particular molecular architecture effectively results in controlled

induced by the random coil then reaches the force necessary to induce

yield over hundreds of percent of strain. We emphasize that the

unfolding of the CC structure (see Fig. 11). We emphasize that this is a

understanding of these mechanisms is currently only in its infancy, in

simple calculation. However, it clearly illustrates that entropic elasticity

particular the competition between intermolecular adhesion, entropic

can lead to forces that are sufficiently large to induce unfolding of the

elasticity, and molecular unfolding.

protein structure.
This analysis suggests a possible change in the deformation

Discussion and conclusion

mechanism from entropic-dominated elastic stretching of the head/

Deformation and fracture are intimately linked to the atomic

tail domains to uncoiling of the CC structures. This is an example of

microstructure of the material. Whereas crystalline materials show

concurrent mechanisms that operate at different scales. On the one

mechanisms such as dislocation spreading or crack extension

Table 1 Selection of various material defects, their appearance, and how they control the deformation and fracture
properties.

56

Defect

Description

Materials in which defect

is found

Consequences for
mechanics

Remarks

Crack

Void-like inclusion or
surface crack; region
with reduced traction
across a molecular plane

Many crystals (metals,

ceramics), in tissues
(larger scales > several
micrometers)

Location of stress
concentration; applied
stress is multiplied many
times at a crack tip;
can lead to breaking or
shearing of atomic bonds

Cracks at various scales

largely control the
strength of crystalline
materials

Dislocation

Localized shear
displacement in crystals

Predominantly metals
and other crystalline
materials

Mediates plastic
deformation

Nucleation of
dislocations often
competes with crack
extension; determines
whether a material is
brittle or ductile

Modifications in
secondary protein
structure, e.g. stutter
defect in CCs

Two AHs in CC run in

parallel, thus loss of
helical geometry

BPMs with distinct

secondary structure,
e.g. IFs

Weaker molecular
structure, unfolds first
under load

Stutter defect provides

a means to control
unfolding locations with
molecular precision.

Amino acid sequence

mutation

Replacement, deletion
or addition of a specific
amino acid

Any protein based

material

May strengthen or
weaken the material

Some mutations lead

to diseases; other
mutations are necessary
to improve the function
(e.g. antibodies are
generated through
random mutations and
thus enable resistance)

Grain boundary (GB)

Interface between two

differently oriented
crystals

Crystalline materials
(metals, ceramics)

Mediates deformation,
e.g. by GB sliding,
GB diffusion at high
temperatures

GBs are particularly

important at small grain
sizes, e.g. nanocrystalline
metals

Vacancy

Missing atom in crystal

lattice

Crystalline materials

Enhances diffusive
material transport, which
mediates deformation

Critical for hightemperature material

behavior

Craze

Region of localized
yielding, formation of
microvoids and fibrils

Polymers, plastics

Process prior to cracking,

dissipates energy

Increases toughness of
plastics

Shear band

Small region inside

material in which
localized shear has
occurred

Polymers, metallic
glasses

Reduction of the
material strength

Mediates plasticity

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

Fracture mechanics of protein materials

REVIEW

(Figs. 4 and 5), hierarchical BPMs feature molecular unfolding or sliding,

since their formation and rupture can be directed by changes in the

where HB rupture is of particular significance (see Figs. 2 and 12).

chemical environment at minimal energy use. Cells or organisms

The dominance of specific mechanisms is controlled by geometrical

maintain control over structural assemblies based on such bonds.

parameters as well as by the structural arrangement of the protein

Controllability is vital in the context of complex biological signaling

elementary building blocks across many hierarchical scales, from

pathways.

nano to macro (Figs. 1 and 2). It is known that nano- or microscopic

Another distinction between traditional and biological materials is

structures often control the macroscopic material behavior: for

the geometrical occurrence of defects. While defects are often

example, grain size reduction or confinement leads to an increase

distributed randomly over the volume in crystalline materials,

of the strength in crystalline metals30,7476. Similar mechanisms are

biological materials consist of an ordered structure that reaches down

observed in biological materials. For instance the alteration of the

to the nanoscale. Defects are placed with atomistic or molecular

molecular length of tropocollagen molecules in collagen fibrils directly

precision, and may play a major role in the material behavior

controls the dominating deformation mechanism and influences the

observed at larger scales. In IF, the addition of molecular stutter defects

strength6,15.

leads to soft spots that represent preferred locations of protein

A major trait of BPMs is the occurrence of hierarchies and the

abundance of weak interactions. The presence of hierarchies in

unfolding64.
These results suggest that analogies can be drawn between

biological materials may be vital to take advantage of molecular

biological and synthetic materials. Table 1 summarizes a few crystal

and submolecular features, often characterized by weak interactions,

and molecular defects, their impact on the deformation behavior and

and multiply their properties so that they become visible at larger

how they control the material properties.

scales. Use of weak interactions makes it possible to produce

In addition to the long-term impact in biology, bioengineering,

strong materials at moderate temperatures and thus with limited

and medicine, this research may eventually contribute to our

energy use. Weak interactions are also critical for controllability,

understanding of how different scales interact with one another.

Fig. 15 The long term impact of our work is that it will extend our ability to perform structural engineering at the macroscale, to the ultimate scale, the nanoscale.
Opening the material scale as design space for new material development may open endless possibilities for development of robust, adaptive, active, and smart
materials.

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

57

REVIEW

Fracture mechanics of protein materials

It may also enable synthesis of novel complex structural materials,

Acknowledgments

designed from the nano- to the macroscale, as shown in Fig. 15. In

This research was partly supported by the Army Research Office (ARO),
grant number W911NF-06-1-0291 (program officer Bruce LaMattina),
as well as a National Science Foundation CAREER Award, grant number
CMMI-0642545 (program director Jimmy Hsia). TA acknowledges
the support of the German National Academic Foundation and the
Juergen-Ulderup Foundation. We acknowledge discussions with Harald
Herrmann (University of Heidelberg, Germany) and Laurent Kreplak
(University of Basel, Switzerland), as well as continued support by Lothar Gaul
(University of Stuttgart, Germany).

order to achieve these goals, major challenges must be overcome, in

particular in relating molecular processes to larger-scale phenomena. As
illustrated in Fig. 8, protein materials constitute exceedingly complex
structures. While the behavior of individual proteins is reasonably
well understood, the properties of large assemblies of proteins remain
largely unknown.

REFERENCES

39. Humphrey, W., et al., J. Mol. Graphics (1996) 14, 33

1. Broberg, K. B., Cracks and Fracture, Academic Press, London, (1990)

2. Courtney, T. H., Mechanical behavior of materials, McGraw-Hill, New York, (1990)

40. Nelson, M. T., et al., Int. J. Supercomputer Appl. High Performance Computing
(1996) 10, 251

3. Hirth, J. P., and Lothe, J., Theory of Dislocations, Wiley, New York, (1982)

41. MacKerell, A. D., Jr., et al., J. Phys. Chem. B (1998) 102, 3586

4. Alberts, B., et al., Molecular Biology of the Cell, Garland, New York, (2002)

42. Evans, E., and Ritchie, K., Biophys. J. (1997) 72, 1541

5. Buehler, M. J., J. Mater. Res. (2006) 21, 1947

43. Dudko, O. K., et al., Phys. Rev. Lett. (2006) 96, 108101

6. Buehler, M. J., Proc. Natl. Acad. Sci. USA (2006) 103, 12285

44. Wiita, A. P., et al., Proc. Natl. Acad. Sci. USA (2006) 103, 7222

7. Fratzl, P., et al., J. Mater. Chem. (2004) 14, 2115

45. Gilli, P., et al., J. Am. Chem. Soc. (2004) 126, 3845

8. An, K.-N., et al., Biorheology (2004) 41, 239

46. Bell, G. I., Science (1978) 200, 618

9. Ramachandran, G. N., and Kartha, G., Nature (1955) 176, 593

47. Ackbarow, T., and Buehler, M. J., J. Mater. Sci. (2007) doi: 10.1007/s10853-0071719-2

10. Buehler, M. J., Nanotechnology (2007) 18, 295102

11. Freund, L. B., Dynamic Fracture Mechanics, Cambridge University Press,
Cambridge (1990)

49. Mcke, N., et al., J. Mol. Biol. (2004) 335, 1241

12. Lu, H., et al., Biophys. J. (1998) 75, 662

50. Helfand, B. T., et al., J. Cell Sci. (2004) 117, 133

13. Rief, M., et al., Science (1997) 276, 1109

51. Burkhard, P., et al., Struct. Folding Design (2000) 8, 223

14. Fantner, G. E., et al., Nat. Mater. (2005) 4, 612

52. Kiss, B., et al., J. Struct. Biol. (2006) 155, 327

15. Buehler, M. J., J. Mech. Behavior Biomed. Mater. (2007), doi: 10.1016/
j.jmbbm.2007.04.001

53. Omary, M. B., et al., N. Engl. J. Med. (2004) 351, 2087

54. Toivola, D. M., et al., Trends Cell Biol. (2005) 15, 608

16. Buehler, M. J., and Wong, S. Y., Biophys. J. (2007) 93, 37

55. Bruck, H. A., et al., Exp. Mech. (2002) 42, 361

17. Engler, A. J., et al., Cell (2006) 126, 677

56. Janmey, P. A., et al., J. Cell Biol. (1991) 113, 155

18. Buehler, M. J., and Gao, H., Nature (2006) 439, 307

57. Schietke, R., et al., Eur. J. Cell Biol. (2006) 85, 1

19. Buehler, M. J., et al., Nature (2003) 426, 141

58. Moir, R. D., and Spann, T. P., Cell. Mol. Life Sci. (2001) 58, 1748

20. Buehler, M. J., et al., Phys. Rev. Lett. (2006) 96, 095505

59. Wilson, K. L., et al., Cell (2001) 104, 647

21. Buehler, M. J., et al., Threshold crack speed in dynamic fracture of silicon, In
Multiscale Modeling of Materials, Devanathan, R., et al., (eds.), Mater. Res. Soc.
Symp. Proc., Warrendale, USA (2007), 978E, 0978-GG02-03
22. Allen, M. P., and Tildesley, D. J., Computer Simulation of Liquids, Clarendon Press,
New York, (1989)
23. Goddard, W. A., A Perspective of Materials Modeling. In Handbook of Materials
Modeling, Yip, S., (ed.), Springer, Berlin, (2005)
24. Tai, K., et al., Nano Lett. (2006) 6, 2520
25. Sun, Y.-L., et al., J. Biomech. (2004) 37, 1665
26. Dao, M., et al., J. Mech. Phys. Solids (2005) 53, 493
27. Lim, C. T., et al., Mater. Sci. Eng., C (2006) 26, 1278
28. Daw, M. S., and Baskes, M. I., Phys. Rev. B (1984) 29, 6443
29. Angelo, J. E., et al., Modell. Simul. Mater. Sci. Eng. (1995) 3, 289
30. Wolf, D., et al., Metallkunde (2003) 94, 1052
31. Van Swygenhoven, H., et al., Nat. Mater. (2004) 3, 399
32. Yamakov, V., et al., Nat. Mater. (2004) 3, 43
33. Rice, J. R., and Thomson, R. M. Philos. Mag. (1974) 29, 73
34. Rice, J. R., J. Mech. Phys. Solids (1992) 40, 239
35. van Duin, A. C. T., et al., J. Phys. Chem. A (2001) 105, 9396
36. van Duin, A. C. T., et al., J. Phys. Chem. A (2003) 107, 3803
37. Tersoff, J., Phys. Rev. Lett. (1988) 61, 2879
38. Hauch, J. A., et al., Phys. Rev. Lett. (1999) 82, 3823

58

48. Wang, N., and Stamenovic, D., J. Muscle Res. Cell Motil. (2002) 23, 535

SEPTEMBER 2007 | VOLUME 10 | NUMBER 9

60. Herrmann, H., and Aebi, U., Annu. Rev. Biochem. (2004) 73, 749
61. Strelkov, S. V., et al., J. Mol. Biol. (2004) 343, 1067
62. Ackbarow, T., and Buehler, M. J., Hierarchical nanomechanics of vimentin alpha
helical coiled-coil proteins. In Multiscale Modeling of Materials, Devanathan R.,
et al., (eds.), Mater. Res. Soc. Symp. Proc., Warrendale, USA (2007), 978E, 0978GG10-05
63. Ackbarow, T., and Buehler, M. J., unpublished results
64. Ackbarow, T., and Buehler, M. J., unpublished results
65. Schwaiger, I., et al., Nat. Mater. (2002) 1, 232
66. McLachlan, A. D., and Karn, J., J. Mol. Biol. (1983) 164, 605
67. Brown, J. H., et al., Proteins: Struct., Funct., Genet. (1996) 26, 134
68. Strelkov, S. V., et al., EMBO J. (2002) 21, 1255
69. Herrmann, H., and Aebi, U., Cell. Mol. Life Sci. (1999) 55, 1416
70. Parry, D. A. D., J. Mol. Biol. (1978) 120, 545
71. Kaminska, A., et al., Hum. Genet. (2004) 114, 306
72. Kreplak, L., et al., Exp. Cell Res. (2004) 301, 77
73. Bustamante, C., et al., Science (1994) 265, 1599
74. Yip, S., Nature (1998) 391, 532
75. von Blanckenhagen, B., et al., Modell. Simul. Mater. Sci. Eng. (2001) 9, 157
76. Nieh, T. G., and Wadsworth, J., Scripta Metall. Mater. (1991) 25, 955
77. Smith, T. A., et al., Proteins: Struct., Funct., Genet. (2003) 50, 207
78. Strelkov, S. V., et al., Bioessays (2003) 25, 243