Force and function: probing proteins with AFM-based force spectroscopy.

Abstract

Forces play a pivotal role in life, and the response of live systems to forces requires molecules and molecular interactions with adequate properties to counteract both in a passive and also, if needed, in an active, dynamic manner. However, at the level of individual molecules these forces are so minute, that the development of sophisticated experiments to measure and control them was required. With the maturation of these techniques, particularly the AFM-based single-molecule force spectroscopy into commercial instruments, the scope has widened considerably and more and more studies shed light onto the different aspects of biomolecular mechanics. Many surprises turned up and more are waiting for us.

Full text

Page 1

Available online at www.sciencedirect.com
Force and function: probing proteins with AFM-based force
spectroscopy
Elias M Puchner1,2and Hermann E Gaub1
Forces play a pivotal role in life, and the response of live
systems to forces requires molecules and molecular
interactions with adequate properties to counteract both in a
passive and also, if needed, in an active, dynamic manner.
However, at the level of individual molecules these forces are
so minute, that the development of sophisticated experiments
to measure and control them was required. With the maturation
of these techniques, particularly the AFM-based single-
molecule force spectroscopy into commercial instruments, the
scope has widened considerably and more and more studies
shed light onto the different aspects of biomolecular
mechanics. Many surprises turned up and more are waiting
for us.
Addresses
1University of Munich and Center for Nanoscience, Amalienstr. 54,
80799 Munich, Germany
2Center for Integrated Protein Science Munich (CIPSM), Germany
Corresponding author: Puchner, Elias M (elias.puchner@physik.lmu.de)
and Gaub, Hermann E (gaub@lmu.de)
Current Opinion in Structural Biology 2009, 19:605–614
This review comes from a themed issue on
Biophysical methods
Edited by Keith Moffat and Andreas Engel
0959-440X/$ – see front matter
Published by Elsevier Ltd.
DOI 10.1016/j.sbi.2009.09.005
Introduction
Since all biological systems are subjected to forces,
diverse mechanical functions of proteins have evolved
in response to this selection pressure. Passive mechanical
adaptations range from rigid structures such as the cytos-
keleton, to flexible elastic regions, for example in the
giant muscle protein titin. For the active adaptation to
strain, molecular force sensors have evolved, which con-
vert force-induced conformational changes into biological
signals. Other examples are cell adhesion proteins that
change their affinity to other partners to regulate
adhesion.
Since force is a vector quantity, the direction of its
application is crucial but hard to control in ensemble
experiments at the molecular level. Therefore, it is
essential to study force-related functions of proteins on
a single-molecule level. Whereas optical [1??,2] and mag-
netic tweezers [3?,4] are ideal in the low-picoNewton
force regime, AFM-based [5] single-molecule force spec-
troscopy (SMFS) has become a widespread and powerful
tool to investigate the mechanical properties of proteins
and force-induced processes, which typically occur at
elevated forces of several tens to hundreds of picoNew-
tons (for a review see [6??]). Particularly the option to
select individual molecules in combination with the wide
range of forces has created new applications. Besides the
investigationofproteinswithmechanicalfunction,SMFS
gives insight into the energy landscape [7] of protein
folding and unfolding in general.
In this review we address AFM-based SMFS highlights
from recent years. We focus on the investigation of
the mechanical architecture of native and engineered
globular proteins, on structural insights, mechanical
detection of ligand binding, and on force-induced con-
formational changes of substrates and enzymes that
serve as molecular force sensors. In the last part we
give an outlook on how the precise mechanical one-by-
one assembly of single biomolecules guided by specific
DNA hybridization may serve as a toolbox for studying
the spatially regulated interplay of enzymes or other
functional units.
AFM-based force spectroscopy
In SMFS experiments forceis applied to themoleculesby
stretching them between the sample surface and thetip of
an AFM cantilever with a spring constant typically in the
10 pN/nm range (see Figure 1a). Attaching the molecules
covalently or through antibodies at specific sites provides
control over pickup points and higher efficiency, but is
often not necessary if multidomain linkers are used that
flank the investigated protein and that leave their finger-
printintherecordedtraces.Inthebeginningofaretraction
cycle, the unfolded part of the molecule is stretched
causing a characteristic elastic response. The most promi-
nent models describing the force-extension behavior are
the worm-like chain (WLC) [8], the freely jointed chain
(FJC) [9], the freely rotating chain (FRC) [10], and exten-
sions [11?]. At a certain force a molecular bond breaks or a
mechanical structural element unfolds, resulting in a sud-
den increase in contour length and in a drop in force (see
Figure1a).Becauseofthestochasticnatureofthisprocess,
the rupture force and the extension exhibit a distribution
and also depend on experimental parameters such as pull-
ingspeed[12,13?],temperature[14?],orsolvent [15??].To
gaindirectaccesstothecharacteristicmolecularparameter
www.sciencedirect.com
Current Opinion in Structural Biology 2009, 19:605–614

Page 2

contour length without fitting, a method was developed
recentlythatdirectlytransformsforce-extensiontracesinto
force-contour length plots (see Figure 1b). The barrier
position histograms derived from these plots allow direct
comparison of proteins, pattern recognition of molecular
fingerprints, and averaging of traces [16?].
Whereas the easiest constant-velocity measurement
protocol provides control over the end-to-end distance
and high accuracy for the determination of barrier
positions, force-ramp [17] or force-clamp [18??] proto-
cols use feedback loops to control the force applied to
the molecule and allow an easier determination of the
rates associated with folding and unfolding or chemical
reactions.
Force resistance of native polyproteins
Comparing the response of proteins to biological forces is
important for the understanding how evolution adopted
material properties for different mechanical functions.
The first AFM-based SMFS unfolding and refolding
experiments [19??] were performed on the giant muscle
protein titin [20–23], which spans the half-sarcomere and
primarily consists of several hundred globular immuno-
globulin (Ig), fibronectin-type-3 (Fn3), and of elastic
PEVK regions. These mechanically stable domains resist
forces ranging from 150 pN to 300 pN, depending on the
pulling speed, and cause a regular sawtooth pattern in
force-extension traces due to successive unfolding of the
domains [19??]. Reversible pulling cycles allow the inves-
tigationoffolding.Inexperimentswithproteinconstructs
containing defined Ig and Fn3 domains from different
titin regions and different isoforms, it was found that
unfolding forces differ systematically [24??]. The combi-
nation of protein engineering with SMFS revealed the
hierarchical mechanical stability ranging from flexible
PEVK regions [25,26] to the weaker proximal and stron-
ger distal Ig-domains from cardiac titin, allowing the
explanation of the macroscopic mechanical properties
from its parts [27??]. Many other native or engineered
multidomain proteins subjected to biological forces have
been mechanically characterized in this way, including
the extracellular matrix proteins fibronectin [28] and
tenascin [29], the insect flight muscle proteins projectin
and kettin [30?] and filamin [31,32], which links the actin
cytoskeleton to the membrane.
Mechanical stability and structural reporters
To understand the mechanical stability of proteins, the
forceresistantstructuralelementshavetobeidentified.If
the structure is known, these elements can be identified
by point mutations that alter the stability of the protein
and thus serve as reporters (see Figure 2a). Molecular
dynamics simulations (MD) can give dynamic structural
explanations, as in the case of Ig27s unfolding intermedi-
ate [33]. The most common elements responsible for
the mechanical stability of proteins are force-bearing b-
sheets, in which the hydrogen bonds are simultaneously
loaded in shear geometry. Even protein domains without
mechanical function can have significant stability such as
the small B1-domain of the bacterial cell wall protein G
606
Biophysical methods
Figure 1
Principle of AFM-based SMFS experiments. (a) The protein construct, in
this case titin kinase flanked by Ig/Fn domains, is contacted with the tip
of an AFM cantilever and stretched while force-extension (fx) traces are
recorded. After each unfolding event the force drops and the sudden
increase in contour length can be measured by models for polymer
elasticity (dashed lines: WLC). The five even contour length increments
at the end (L1–L6) are due to unfolding of Ig/Fn domains, which serve as
spacers and provide their fingerprint. The initial low-force part (L0–L1) is
therefore caused by the framed protein. Reliable interpretation of force
spectroscopy data can also be achieved by other defined protein
constructs, for example containing multiple repeats of the investigated
protein. (b) To gain straight access to the characteristic variable of
barrier positions, a method was developed that directly transforms each
point of fx-traces into contour length space. The obtained barrier
position histograms allow averaging of several traces, the comparison of
proteins and the precise and direct determination of barrier positions
without fitting (L0, L1, etc.) [16?].
Current Opinion in Structural Biology 2009, 19:605–614www.sciencedirect.com

Page 3

(GB1), which binds immunoglobulin G [34]. Point
mutations in the force-bearing b-strands, which disrupt
the formation of hydrogen bonds, destroy the mechanical
stability [35?]. In contrast to these b-sandwich topologies,
the newly designed protein TOP7 contains a force-trans-
ducing strand between the loaded b-sheet and also shows
significant mechanical stability [36?].
Recent studies show that the force response of loaded
b-sheets is not universal but context sensitive and de-
pendentontheotherpartsofaprotein.Shufflingstructural
elementsofthehomologousbutmechanicallydistincttitin
domains I32 and I27 (298 pN and 204 pN, respectively, at
400 nm/s pulling speed) results in various hybrid domains
with intermediate or extremely low mechanical stability
[37]. It was further demonstrated that exchanging the
hydrophobic core of the human fibronectin domain
FNfn10 by a mechanically stronger tenascin domain
TNfn3 results in 20% higher unfolding forces. Therefore,
the inner parts of proteins also contribute to the over all
mechanical stability of the protein [38?].
Besides point mutations weakening force-bearing struc-
tures, reporters have been developed that make them
stronger (see Figure 2a). The insertion of facing bi-
histidines in the two force-bearing b-strands provides a
chelation site, which enhances the force resistance of the
GB1 by as much as a factor of two. In contrast, chelation
sites introduced outside the force-bearing region do not
show any effect [39??]. This significant approach seems
very promising since the reversible binding of metal ions
allows switching between the ordinary and stabilized
state within the same experiment. Insertions of disulfide
bondsbycysteinemutationshaveevenmoreimpactsince
covalent bonds can resist much higher forces in the
nanoNewton range [40]. Therefore, the observed contour
length increments caused by unfolding of the protein
domains lack the part that is shortened by the disulfide
bond (see Figure 2b). By introducing single disulfide
bonds at four different locations in the titin domain
I27, and by measuring the shortened contour length,
the length per amino acid was precisely determined in
Ref. [41?], and it was further shown that the refolding
time depends upon the number of unfolded residues.
Larger proteins often exhibit several force resistant
elements so that unfolding does not take place in an
all or none fashion but rather shows intermediate states
such as GFP [42?], EYFP [43], T4 lysozyme [44?] mal-
tose-binding protein [45], or titin kinase [46???,47?]. The
insertion of mechanical reporters in the different struc-
tural elements of such proteins is of special interest since
itgives insight into themechanical architecture. With this
approach the mechanically stable elements of the mal-
tose-binding protein were identified in Ref. [45], and the
probability of two unfolding pathways of GFP could be
controlled [48?].
Force response depends not only on the topology of a
protein, but also on the pulling geometry as it was shown
for the small protein domain E2lip3 of pyruvate dehydro-
genase by means of different attachment points in SMFS
experiments and MD simulations [49]. Also ubiquitin
polyproteins, unfolding at 200 pN when loaded at their
N-terminus and C-terminus, showed a strong decrease in
Force and function Puchner and Gaub607
Figure 2
Mechanical reporters for structural and anisotropic investigations of
proteins. (a) Point mutations alter the mechanical stability of proteins if
inserted in the force resistant structural elements (shown in orange). This
approach is ideal to identify the parts of a polypeptide chain forming
force-bearing elements. Destabilizing mutations such as removal of
hydrogen bonds in a b-sheet lower the rupture forces (F1, green),
whereas stabilizing mutations such as bi-histidines result in higher forces
(F2, blue) if metal ions are present and form a chelate complex. (b) The
introduction of disulfide bonds by cysteine mutation reduces the
observed contour length increments (DL1, blue) by the length of the
shortened polypeptide chain also allowing the identification of
mechanically stable elements. (c) Crosslinking proteins at defined
cysteines have several interesting aspects. On the one hand, the force
can be applied at different points to study the anisotropic deformation
response of proteins. On the other hand, comparing the measured
contour length increments with the ones calculated from the number of
free amino acids reveals the distance d2of attachment points.
(Schematically depicted are GB1 domains, pdb: 1PGA .).
www.sciencedirect.com
Current Opinion in Structural Biology 2009, 19:605–614

Page 4

stability by more than a factor of two in a Lys48 and C-
terminus crosslinked polymer at a pulling speed of
300 nm/s. The different unfolding pathways were also
reproduced with SMD simulations [50?]. A more general
and now widely used approach for establishing different
pulling geometries is based on crosslinking proteins
through cysteines located at their surface [51] (see
Figure 2c). In the group of Rief, GFPs were crosslinked
at various positions resulting in unfolding forces up to
sixfold higher and in directional spring constants ranging
from 1 N/m to 17 N/m [52?]. The comparison of the
measured contour length increments with the free length
of the polypeptide chain revealed the distances between
the attachment points on the protein surface [53??]. This
triangulation method is a promising approach to gain
structural information on proteins with unknown struc-
ture. The anisotropy of the energy landscape was further
demonstrated with the yellow photoactive protein acting
as a photoreceptorandthe Per-Arnt-Sim domain involved
in regulatory and sensory function [54,55]. SMD simu-
lations revealed the different geometries of pulling the
force-bearing b-strands apart. Even more interesting in
these studies is the experimental mechanical detection of
light induced structural changes of these proteins.
Having summarized SMFS experiments leading to the
understanding of the force response of proteins, to their
anisotropic stretching response and to structural infor-
mation, we now focus on the binding of ligands that
modify the energy landscape and may be mechanically
detected.
Ligand-induced changes in the mechanical
properties of proteins
Specific bindingof ligands isoneofthe mostfundamental
biological processes. It is clear that binding of a ligand to a
protein reduces the total free energy by the binding
energy. However, this change in the energy landscape
of a protein must not necessarily influence the barriers
along the mechanical unfolding pathway. Some recent
experiments revealed cases in which binding to a protein
alters its mechanical properties. NuG2, a mutant of the
GB1 domain, unfolds at 105 pN at a pulling speed of
400 nm/s. However, in the presence of the fc fragment of
the IgG antibody, a certain fraction of domains is stabil-
ized by more than 100 pN [56]. Since the high force
fraction is caused by binding and stabilization of the
antibody fragments, variation of their concentration and
counting the bound fraction revealed the binding con-
stant. The same principle was applied to the mutant
GT18P of GB1 in which the force-bearing b-strands
are disrupted. Whereas the domains themself show
almost no mechanical stability and behave as entropic
springs, the addition of antibodies resulted in rupture
forces of the domains centered around 100 pN [35?].
Other recent examples include dihydrofolate reductase
[57], binding of GDP to the GTPase Ran [58?], inhibitor
binding to an antiporter [59?], binding of nickel to engin-
eered chelation sites [39??], and mechanically induced
binding of ATP to titin kinase [46???] or protein–DNA
interaction [60]. In a recent and very interesting work,
binding of Ca/MAS and of a peptide sequence of myosin
light-chain kinase to calmodulin was observed with a
different approach [61???]. Calmodulin, being able to
switch between two conformations, was stretched much
slower than the time constant of its fluctuations. The
transition rates between the conformations were deduced
from the force-time traces and it was shown that they
depend on the presence of the ligands [61???].
Force-induced mechanoenzymatics
In the next level of force-functionality we discuss
active adaptation to force by the conversion of mech-
anical stress to biochemical signals. The interplay of
mechanical stability, ligand binding, and force-induced
conformational changes finally leads to biological force
sensors.
Using force-clamp SMFS, the force-induced reduction of
hidden disulphide bonds in an engineered substrate by
the enzyme thiredoxin was studied in Fernandez’s lab in
great detail [62???]. The disulphide bonds were inserted
in a polyprotein of I27 domains in such a way, that they
are not accessible in the folded domain but become
exposed upon force-induced unfolding. Measuring the
force-dependent cleavage rates at different enzyme con-
centrations, a kinetic model was developed containing an
intermediate state and two paths with two different force-
dependent rate constants. Interestingly, the correspond-
ing potential widths were structurally interpreted and
suggested to be reorientations of the disulphide bond
during catalysis. In another study, thiredoxin was shown
to selectively reduce the natural disulfide bonds of the
multimodular protein angiostatin during force-induced
unfolding [63?].
How are these mechanisms used by nature? A prime
example of a biological system that needs adaptation to
mechanical stress is the vertebrate muscle. Close to the
M-band structure of the muscle sarcomere [64,65] a
kinase domain (TK) is incorporated in the crosslinking
titin at an ideal position for the detection of shear forces
between neighboring myosin filaments. It was shown,
that the ATP binding site of this kinase is blocked by the
C-terminal regulatory tail (red in Figure 3a), preventing
binding of the cosubstrate ATP [66].
To mimic the biological force in the M-band, the natural
protein construct consisting of the kinase domain and the
surrounding immunoglobulin and fibronectin (Ig/Fn)
domains was stretched and unfolded in SMFS exper-
iments [46???]. The force-extension traces, as shown in
Figure 1a, exhibit a hierarchical mechanical stability of
the domains: first TK unfolds over five energy barriers
608
Biophysical methods
Current Opinion in Structural Biology 2009, 19:605–614 www.sciencedirect.com

Page 5

below 50 pN at pulling speeds of 1 mm/s. After complete
unfolding, five force peaks are observed whose number
and contour length increments correlate exactly with the
expected values of the structural Ig/Fn domains. In con-
trast to the independently unfolding Ig/Fn domains,
unfolding of TK is, despite the comparable levels of
the peaks, strictly ordered and thus follows a well-defined
pathway through the energy landscape (see Figure 3a). In
the presence of the cosubstrate ATP at a physiological
concentration, a certain fraction of unfolding traces shows
an additional energy barrier caused by binding of ATP to
the sequential and force-induced opening of the binding
pocket (see Figure 3b).
Since the different conformational states of TK during
force-induced activation are controlled by its end-to-end
distance, the opening time of the mechanically opened
binding pocket is determined by the unfolding speed. In
addition to the equilibrium binding constants discussed
in the section above, the access to the time domain
uncovers the kinetics of the binding process. Exper-
imentsatdifferentopeningtimesrevealanonequilibrium
binding kinetics (see Figure 3c) with the corresponding
parameters, which are kon= 1.8 ? 0.3 ? 1041/ms, koff=
6 ? 3 1/s, and Kd= 347 mM. To identify the element of
TK involved in ATP binding, lysine-36, which plays a
crucial role in homologous kinases, was replaced by
alanine (K36A). In contrast to the mutations in force-
bearing structuralelements
mutation does not influence mechanical stability but
results in a more than sixfold higher off rate and a Kd
in the millimolar range. Apparently force response
and binding of the cosubstrate are well-separated func-
tions of TK.
described above,this
Force and function Puchner and Gaub609
Figure 3
Mechanoenzymatics of the muscular force sensor titin kinase. (a) The superposition of 66 force-extension traces (such as the one in Figure 1) shows
the complex mechanical architecture of TK which unfolds over five distinct barriers always in the same order and below 50 pN at a pulling speed of
720 nm/s. (b) In the presence of the cosubstrate ATP (2 mM) a certain fraction of traces (44) show an additional barrier (2* red) caused by binding of
ATP to the mechanically opened binding pocket and its interaction with LYS36 and MET34. (c) Varying the opening time of the binding pocket by
different pulling speeds and counting the fraction of binding events reveals the nonequilibrium binding kinetics and the force-triggered nature of this
process. Mutation of LYS36 to alanine enlarges to off rate more than sixfold and shifts the dissociation constant to a millimolar range. (d) Correlation of
the experimental data with MD simulations reveals the structural changes during sequential force-induced activation and the structural elements
responsible for the observed barriers (pdb: 1TKI ). Data adapted from Ref. [46???].
www.sciencedirect.com
Current Opinion in Structural Biology 2009, 19:605–614