of 8
Frequency Modulation Atomic Force
Microscopy Reveals Individual
Intermediates Associated with each Unfolded I27 Titin Domain
Michael J. Higgins,* John E. Sader,
y
and Suzanne P. Jarvis*
*Centre for Research on Adaptive Nanodevices and Nanostructures (CRANN), University of Dublin, Trinity College, Dublin 2, Ireland;
and
y
Department of Mathematics and Statistics, University of Melbourne, Victoria 3010, Australia
ABSTRACT In this study, we apply a dynamic atomic force microscopy (AFM) technique, frequency modulation (FM)
detection, to the mechanical unfolding of single titin I27 domains and make comparisons with measurements made using the
AFM contact or static mode method. Static mode measurements revealed the well-known force transition occurring at 100–120
pN in the first unfolding peak, which was less clear, or more often absent, in the subsequent unfolding peaks. In contrast, some
FM-AFM curves clearly resolved a force transition associated with each of the unfolding peaks irrespective of the number of
observed unfolded domains. As expected for FM-AFM, the frequency shift response of the main unfolding peaks and their
intermediates could only be detected when the oscillation amplitudes used were smaller than the interaction lengths being
measured. It was also shown that the forces measured for the dynamical interaction of the FM-AFM technique were significantly
lower than those measured using the static mode. This study highlights the potential for using dynamic AFM for investigating
biological interactions, including protein unfolding and the detection of novel unfolding intermediates.
INTRODUCTION
The reversible unfolding of immunoglobulin (Ig) domains in
the I-band region of the modular protein titin is believed to
function for the passive elasticity of muscle. Along with
other proteins (e.g., barnase) (1), the mechanical unfolding
of the well-defined tertiary structure of the wild-type titin Ig
27 domain (I27) has been studied extensively using atomic
force microscopy (AFM) in the static mode (i.e., the DC-
deflection of a nonoscillating cantilever is measured) (2–6).
By stretching recombinant proteins consisting of multiple
repeats of a single domain using AFM, a suite of information
on protein folding has been revealed, including the mechan-
ical unfolding forces of individual domains, kinetic param-
eters for both unfolding and refolding pathways, determination
of unfolding intermediates, and comparison of forced un-
folding with physiological or chemical denaturant unfolding
(for reviews, see Best and Clarke (7) and Fisher et al. (8)).
Related to this study, an unfolding intermediate due to the
disruption of hydrogen bonds between A and B
b
-strands of
the I27 domain was initially indicated using steered
molecular dynamic (SMD) simulations (9) and later con-
firmed in AFM measurements (3).
More recently, two major types of dynamic AFM (where
the cantilever is oscillated), amplitude modulated (AM) and
frequency modulated (FM), have been used to study various
biological force interactions, including ligand-receptor inter-
actions (10,11), polysaccharides elasticity (12–14), nucleic
acids/peptides (15,16), and proteins (17–19). These tech-
niques typically involve oscillating the cantilever well
below, or at, the resonance frequency to detect changes in
the amplitude, phase, and/or resonance frequency that occur
in response to changes in the interaction force. Attempts are
then made to quantify the force, though this can be complex
in most cases. In relation to protein unfolding, Okajima et al.
(18) were able to detect a suggested refolding response of the
hydrophobic core in a single monomeric globular protein,
and Janovjak et al. (19) revealed novel unfolding peaks in the
unfolding of bacteriorhodopsin protein from native purple
membrane. Pertinent to this study, Forbes and Wang (20)
used AM-AFM to measure the unfolding response of native
titin from skeletal muscle myofibrils. These authors revealed
the typical periodic sawtooth pattern, though they also de-
tected additional peaks in the stiffness measurements that
were suggested to correspond to structural transitions or in-
termediates during unfolding. Due to the heterogeneity of the

300 globular domains in the native titin and apparent lack
of correlation between the peaks in the simultaneous stiffness
and force measurements, the assignment of the additional
peaks to specific unfolding intermediates was not feasible.
However, the study importantly highlighted the ability of the
dynamic technique to detect transitions that could not easily
be detected in the force measurements alone. The recent
advance toward polymer pulling experiments using dynamic
methods is due to this possibility of achieving a greater force
resolution and ability to obtain additional information on the
dissipative components of the force.
In contrast, for this study, we use FM detection to in-
vestigate the unfolding of tandem repeats of the I27 domain
and make comparisons to static measurements that are also
performed in the study. As mentioned above, an intermediate
in the first unfolded peak has previously been observed using
static mode AFM (3), though the transition in the force due to
Submitted May 16, 2005, and accepted for publication September 27, 2005.
Address reprint requests to Dr. Michael Higgins, Centre for Research on
Adaptive Nanodevices and Nanostructures (CRANN), University of
Dublin, Trinity College, Dublin 2, Ireland. Tel.: 353-608-3088; Fax:
353-608-3027; E-mail: michael.higgins@tcd.ie.
Ó
2006 by the Biophysical Society
0006-3495/06/01/640/08 $2.00
doi: 10.1529/biophysj.105.066571
640
Biophysical Journal Volume 90 January 2006 640–647
the intermediate becomes very unclear, or is more often ab-
sent, with subsequent unfolded domains. By performing
dynamic measurements on a well-defined modular protein,
such as titin I27, that has a known intermediate, we were able
to detect corresponding individual unfolding intermediates
for each peak in the force-extension curves. In addition, the
intermediate could clearly be observed in the final unfolded
domain (eighth peak) for an I27 construct with eight do-
mains. This highlights the potential of dynamic techniques
for future studies on protein folding, including the detection
of novel unfolding intermediates.
MATERIAL AND METHODS
Construction of wild-type TI I27 multimer
Recombinant methods used to express and synthesize direct tandem repeats
of wild-type TI I27 consisting of eight individual modules or domains were
performed according to the recent study by Steward et al. (21).
Static mode force measurements on titin protein
Cleaved round 1 cm mica surfaces on Teflon were coated with titin protein
by pipetting 100
m
Lof30
m
g/ml protein suspension in phosphate-buffered
saline (PBS) buffer onto the mica surface. The mica surface was immersed
for 15 min to allow for sufficient adsorption of the protein and then washed
by exchanging the protein solution (three times) with fresh PBS buffer. The
protein-coated mica surface was then mounted on the AFM sample stage
and static mode force measurements performed using an Asylum Research
MFP-3D AFM (Santa Barbara, CA). Force measurements were taken using
Mikromasch silicon nitride cantilevers that were calibrated using the plan
view method and had measured spring constants 40–100 pN/nm. Force
measurements were taken with a piezo velocity of 1
m
m/s and numerous
force curves obtained on different positions on the protein substrate. Anal-
ysis of the force curves, including worm-like chain model (WLC) fitting to
the elastic response of the extended protein, was performed using the Asylum
Research AFM IGOR Pro software (Wavemetrics, Lake Oswego, OR).
Frequency modulation detection (dynamic)
measurements on titin protein
Preparation of cantilevers for magnetic activation
First, a nanosensor silicon cantilever was calibrated as above and had a
measured spring constant of 1.2 N/m. A glass encapsulated NdBFe
(neodymium/boron/iron) particle was then glued (Epotek 41) onto the
back of the cantilever, directly behind the tip, with the aid of an optical
microscope and micromanipulator. The particle was then magnetized using
an impulse magnetizer (ASC Scientific, Carlsbad, CA), Model IM-IO-ZO)
by positioning the cantilever at an angle of 12
°
from the normal surface (i.e.,
tip angle in AFM holder) and applying a charging voltage of 300 V,
corresponding to a magnetic field strength of 37.3 kG, for 30 s.
Modification of AFM for frequency modulation detection
Force measurements were taken using a modified Asylum Research MFP-
3D to enable frequency modulation (FM) detection in liquid. This was
achieved by implementing Magnetic Activation Dynamic (MAD) mode
(22–24), whereby the cantilever with attached magnetic particle was
oscillated by applying an external magnetic field via a solenoid positioned
underneath the sample stage. For this setup, the voltage to the coil was
amplified using a homemade ‘coil driver’, and the solenoid replaced the
position of the objective lens in the AFM base. To regulate the FM detection
scheme, we used a Nanosurf (Liestal, Switzerland) Phase-Loop-Lock
controller/detector (PLL). The PLL used two feedback systems to control the
dynamic force measurements. One feedback system kept the oscillation
amplitude of the cantilever constant by varying the driving voltage to the
coil. A second feedback system shifted the phase signal of the lever response
to 90
°
, which was then used as the excitation signal to keep the lever
oscillating at its fundamental eigenfrequency. By monitoring changes in the
resonance frequency and excitation amplitude required to keep a constant
oscillation amplitude, we were able to measure frequency shift and
dissipation caused by the tip sample interaction. The AFM was controlled
using a modified version of the Asylum Research software (IGOR Pro,
Wavemetrics).
In contrast to previous dynamical techniques (12–14), a DC-deflection
signal was not acquired simultaneously with the frequency shift and dis-
sipation. This was because the FM-AFM technique in liquid required the use
of stiffer levers (i.e., 1.2 N/m) with higher resonant frequencies to reduce
frequency noise and maintain a stable cantilever oscillation. Thus, the in-
crease in the cantilever stiffness dramatically reduced the DC deflection
sensitivity and made obtaining simultaneous DC measurements difficult.
More importantly, by using relatively large oscillation amplitudes (i.e.,
4.5 nm, 6.2 nm, 11.5 nm, 26.5 nm), the DC deflection probes the average
force experienced by the tip during one oscillation cycle, leading to com-
plexity in its interpretation and comparison to standard DC measurements.
Frequency modulation measurements on titin
and data analysis
Titin protein-coated mica surfaces were prepared for AFM, as above. Before
taking measurements, the resonance frequency of the cantilever (18.47 kHz)
in PBS buffer, with attached magnetic, was obtained by performing a thermal
power spectrum. Frequency shift curves were taken with a piezo velocity of
1
m
m/s and varying oscillation amplitudes of 4.5 nm, 6.2 nm, 11.5 nm, and
26.5 nm. Importantly, physisorption of the protein to the cantilever tip could
be achieved only when the tip was set to dwell for 3–5 s during intermittent
contact with the mica surface before retraction.
Frequency shift (volts) versus extension (nanometer) curves were
obtained and converted to frequency shift (hertz) curves using the sensitivity
value (73.4 Hz/V) of the PLL. To convert the observed frequency shift into
an interaction force
F
ð
z
Þ
;
the formulation recently proposed by Sader and
Jarvis (25) was used,
F
ð
z
Þ¼
2
k
Z
N
z
1
1
A
1
=
2
8
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p
ð
t

z
Þ
p
!
V
ð
t
Þ
A
3
=
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
ð
t

z
Þ
p
d
V
ð
t
Þ
dt
dt
;
(1)
where
k
is the spring constant,
A
is the oscillation amplitude of the tip,
V
ð
z
Þ¼
D
v
ð
z
Þ
=
v
res
;
v
res
is the natural resonant frequency of the cantilever
in the absence of an interaction force, and
z
is the distance of closest
approach between the tip and the surface. Note that
F
ð
z
Þ
is the conservative
component of the interaction force between tip and sample and does not
include any contribution due to dissipative effects. This formula is valid for
any
A
and requires that the interaction force be continuous throughout the
measurement. However, our curves contained an apparent discontinuity in
the interaction force at the moment of unfolding of each I27 domain (i.e.,
peak-to-trough transition). To satisfy the requirement of the formula, we
investigated only the region where the force was continuous, which cor-
responded to the elastic response of the unfolded polypeptide. Equation 1
also required that the force decay to zero as the tip-surface distance ap-
proaches infinity, which is not the case in our measurements, i.e., the
interaction force increases with increasing separation, with the maximum
attractive force experienced by the tip being on the retract portion of the
FM-AFM for Unfolding Proteins
641
Biophysical Journal 90(2) 640–647
oscillation cycle. To address this, we effectively treated the elastic response
of the unfolded region as we would for a repulsive interaction that is
experienced during an approach to the surface. In doing so, the data to be
analyzed were reversed so that the peak maximum of the unfolding event
corresponded to the origin of a hypothetical surface. The effect of frequency
noise, exacerbated by a derivative step in the force conversion, was
eliminated by using a polynomial fit of the unfolding region for the analysis.
It is noted that this technique relies on measuring the change in reso-
nance frequency of the cantilever upon application of an interaction force.
Equation 1 unifies previously established theoretical work (26) by allowing
for the unequivocal determination of the interaction force, for any oscillation
amplitude. The constant-amplitude approach implicitly decouples the in-
fluence of conservative and dissipative force. Equation 1 thus gives a direct
connection between the change in frequency and the true interaction force,
with no ambiguity, and is a dynamic analogy of Hooke’s law for FM-AFM
constant amplitude measurements. The above analysis procedure has been
explained in detail previously (11) and used to calculate unbinding forces of
single ligand receptors interactions using a flexible linker.
RESULTS
Static mode force curves revealing protein
unfolding and force transitions
Static mode measurements performed to explain the prin-
ciple of force-induced modular protein unfolding, and for
comparison with FM detection, revealed that the protein had
sufficiently adsorbed to both the cantilever tip and mica to
enable the unfolding of the I27 domains during extension.
Typical force-extension curves for the protein were observed
(Fig. 1
A
), whereby the sawtooth pattern with periodic
spacing between adjacent peaks indicated the sequential
unfolding of modular domains (six domains in this case)
within the polymer construct. The mean spacing between
peaks was found to be 25.4 nm
6
2.56 nm (mean
6
SD,
n
¼
52), indicating the nonfully extended length of the unfolded
polypeptide. The error in the spacing may be due to the
measurements being taken on a mica surface compared to
a gold surface that is typically used to promote good binding.
A reduction in the binding on the mica would lower the
probability of observing an idealized protein unfolding curve
(many curves are convoluted by multiple protein interac-
tions) and thus may increase the measured error. WLC fitting
to the elastic response of the unfolded polypeptide regions in
Fig. 1, using the persistence length
p
as an adjustable pa-
rameter, gave an average
p
of 0.29 nm and change in contour
length
D
L
of 28.5 nm for each unfolding event. It is noted
that the elastic response in the first peak does not represent an
unfolded polypeptide, but the elastic properties of the poly-
mer construct with fully folded domains (i.e., no domains
have yet unfolded). The
p
and
D
L
were in agreement with the
stretching of a single titin I27 molecule and were consistent
throughout all curves that showed similar force-extension
profiles. The average peak force was measured to be 208
6
43 pN (
n
¼
93), which was in the range for unfolding forces
previously measured at similar tip velocities. As previously
reported, the elastic response of the first unfolding peak was
not well described by the WLC due to a known force
transition, observed as a ‘hump’, occurring at
;
100 pN (Fig.
1,
A
and
B
). This force transition was regularly observed in
the first peak (Fig. 1
A
) but became less clear, or was mostly
absent, with increasing peak number (data not shown). For
the force curve shown in Fig. 1
A
, the transition could not be
observed past the first peak. Fig. 1
B
shows an expanded
image of the first unfolding peak from Fig. 1
A
, which
consists of two WLC fits, one to the elastic response of the
fully folded polymer construct before the force transition
(WLC 1) and after the transition (WLC 2). With
p
set at
0.4 nm (3),
D
L
between both fits was measured to be 3.9 nm.
The analysis of the intermediate using two WLC fits was
performed according to previous work first describing the
intermediate (3), and the
D
L
value obtained is used to de-
scribe the accumulated lengthening of the folded domains in
their native state during extension. Our
D
L
value fell on the
linear regression slope for the previously observed relation-
FIGURE 1 (
A
) Typical force-extension curve for the unfolding of
repeating titin I27 domains in recombinant protein. The sawtooth pattern
with periodic spacing of 25.4 nm
6
2.56 nm for all recorded peaks (
n
¼
52)
indicated the sequential unfolding of modular domains. WLC fitting to the
elastic response of the unfolded polypeptide region (with the exception of
the first peak) of each peak in the figure gave a mean persistence length of
0.29
6
0.04 nm and change in the contour length of 28.5
6
1.1 nm. The first
peak (
leftmost
peak beginning at zero extension) was not fitted well by the
WLC due to the force transition (
arrow
) occurring at a force of
;
100 pN.
The force transition was not observed in the subsequent unfolded domains.
(
B
) Expanded image of the first unfolded peak from
A
clearly showing the
occurrence of the force transition. WLC fitting, with the persistence length
set at 0.4 nm, to the elastic response of the polymer construct with fully
folded domains before the transition (WLC 1) and after the transition region
(WLC 2) gave contour lengths of 16.4 nm and 20.3 nm, respectively. A
previous relationship shows that the
D
L
for these two WLC fits to the first
unfolded peak increases with an increase in the observed number of
unfolded domains. A
D
L
of 3.9 nm for the WLC fits in the figure agrees with
the described relationship for the observed unfolding of six domains, as in
A.
642
Higgins et al.
Biophysical Journal 90(2) 640–647
ship between
D
L
in the first unfolding peak and number of
unfolded domains observed in the force-extension curve (3).
A full statistical analysis is still required to make a com-
parison with the previous work mentioned, as the first peak is
known to be convoluted by the effects of various protein-
surface and protein-protein adhesive interactions. For the
purpose of this work, we give only an example of this WLC
fitting procedure to provide background detail on the inter-
mediate.
Frequency modulation detection revealing
individual force transitions of each module
FM detection measurements revealed sawtooth patterns in
the frequency shift curves, indicating that the dynamic tech-
nique was also able to detect the sequential unfolding of
domains in the I27 protein construct (Fig. 2
A, light gray
trace
). The mean spacing between adjacent peaks in curves
for
A
values of 4.5 nm, 6.2 nm, and 11.5 nm was found to be
25.2
6
2.9 nm (
n
¼
35), 25. 3
6
3.4 nm (
n
¼
61), and 24.6
6
3.1 nm (
n
¼
13), respectively. Thus, irrespective of
A
, the
mean spacing agreed with the nonfully extended length of
the unfolded polypeptide region measured in the static force
measurements. In addition to the main frequency peaks
representing the onset of an unfolded domain, a second
smaller peak was also observed to be associated with each of
the main unfolding peaks (Fig. 2
A
). These secondary peaks
were very difficult to observe in the raw frequency shift
curves due to the significant noise level. A smoothed curve
of the frequency shift (
dark line
) also revealed the secondary
peaks, which appeared as a small increase in the frequency
shift to
;
15–18 Hz to form a discontinuity in the initial
frequency shift (Fig. 2
A
,
arrows
), followed by a further
increase in the frequency shift corresponding to the main
unfolding peak. This discontinuity in the frequency shift of
the main unfolding peaks appeared to indicate the presence
of a force transition, presumably that observed in the static
measurements. However, we stress that the secondary peaks
observed in the smoothed frequency curve do not represent
the true structure of the transition and must be further
validated, as the binominal smoothing process applied
consists of both the feature of the transition and frequency
noise. To validate the true presence and structure of the
transition for the different unfolded domain numbers, we
performed an averaging process for all measured peaks
obtained for
A
of values 4.5 nm and 6.2 nm and according to
their unfolding number (Fig. 2
B
). This is possible because
the frequency noise had a mean of zero and is additive to the
true signal (26). This property enabled the noise to be
minimized by averaging multiple measurements of the same
interaction to leave the underlying true signal undistorted.
The averaged frequency shifts for the different folded
domains are shown in Fig. 2
B
. We emphasize that these
measurements were highly reproducible and independent of
the number of averages taken. In addition, averaging a greater
number of peaks reduced the noise level without distorting
the underlying curve and thus demonstrated the validity of
this procedure. Fig. 2
B
revealed that the intricate structure of
the transition appeared to be hump-like or a plateau that
occurred at a consistent frequency shift of
;
15–18 Hz and
had a slight tendency to increase in length with an increase in
the unfolding number. In some curves, the intermediate was
not always present for every unfolded domain in the
frequency shift curves, as highlighted in the first few peaks
for the top curve (4.5 nm) in Fig. 3
A
. The absence of the
intermediate may have been due to a reduction in the signal
because of the apparent frequency noise highlighted in Fig. 2
A
. Interestingly, the intermediate was always clearly resolved
in the final unfolded domain for the I27
8
construct.
FIGURE 2 (
A
) Frequency shift versus extension curve (
light gray trace
)
for the sequential unfolding of titin I27 domains taken using a resonance
frequency
f
of 18.5 kHz and oscillation amplitude
A
of 6.5 nm. The mean
spacing for
A
values of 4.5 nm, 6.5 nm, and 11.5 nm were 25.2
6
2.9 (
n
¼
35), 25.3
6
3.4 (
n
¼
61), and 24.6
6
6(
n
¼
13), respectively, indicating the
successful unfolding modular domains using this dynamic technique. The
curve shows 10 peaks with each peak recording a positive frequency shift of

25 Hz. The first two peaks, which are slightly greater in magnitude than the
remaining eight, may be due to multiple binding of proteins (i.e., the peaks
correspond to an additional protein molecule on the tip). During extension,
the increasing stiffness of the unfolded polypeptide region restricts the
cantilever movement, causing an increase in the effective spring constant
k
c
of the cantilever. Thus, an increase in
k
c
causes a direct increase in the
cantilever resonance frequency, which is observed as a positive frequency
shift in the unfolding peaks. A smoothed curve (
dark line
) clearly shows
a discontinuity (
arrows
) in the positive frequency shift of the main unfolding
peak, indicating the unfolding intermediate. (
B
) Graph showing averaging of
all peaks showing intermediates (from
A
¼
4.5 nm and 6.5 nm) and
according to their unfolded domain number, which was done by
superimposing and aligning each peak at their peak maximum. The distance
between peaks is negligible, and the noise in each peak increases with the
unfolded domain number due to the fewer numbers of samples averaged.
The distance from the beginning of the force transition (
arrow
) to the peak
maximum showed a slight increase with an increase in the unfolded domain
number. These distances were calculated to be 5.6 nm (peak 1), 6.6 nm (peak 2),
7.9 nm (peak 3), 8.1 nm (peak 4), 10.9 nm (peak 5), and 14.4 nm (peak 6).
FM-AFM for Unfolding Proteins
643
Biophysical Journal 90(2) 640–647
In comparison to smaller
A
values of 4.5 nm and 6.2 nm,
force transitions were not observed in the unfolding peaks,
or were albeit very unclear, for the
A
of 11.5 nm (Fig. 3),
whereas only some irregular peaks were observed for the
A
of 26.5 nm (Fig. 3). In addition, frequency shifts were greater
for smaller amplitudes, which was expected due to
D
f

D
F
/
A
1.5
, as given in Eq. 1 (Fig. 3). Conversion of the frequency
shift curves into an interaction force, as outlined in the
Material and Methods section and shown in Fig. 4,
A
and
B
,
surprisingly resulted in much lower unfolding forces com-
pared to static mode measurements (Fig. 4
B
). Average peak
unfolding forces were found to be 75
6
19.9 pN (
n
¼
37),
63.1
6
20.2 pN (
n
¼
54), and 63.9
6
7.9 (
n
¼
14) pN for
A
values of 4.5 nm, 6.2 nm, and 11.5 nm, respectively,
indicating that the same force law was recovered irrespective
of the oscillation amplitude used.
DISCUSSION
The Ig I27 domain in native titin or as tandem repeats in
recombinant polypeptides is unique compared to other mod-
ular proteins in that individual domains unfold via an in-
termediate when under an applied force. SMD simulations
(9) show that this intermediate results from the rupture of two
hydrogen bonds bridging the A and B
b
-strands, which is
expected to proceed before the detachment of the A
9
and G
strands. The latter process presumably renders the C- and
N-terminal free to unravel, causing the domain to completely
unfold. This intermediate has been directly observed pre-
viously using AFM (3) with similar results repeated in this
study for comparison with dynamic techniques. In these
static mode AFM studies, the intermediate is observed as
a prominent ‘hump’ or force transition in the first unfolding
peak occurring at

100–120 pN that subsides, or is not
present, in subsequent unfolding events. Data showing an
increase in the change of contour length in the first unfolding
peak, as described in Fig. 1
B
, as a function of an increase in
the number of observed unfolded domains are used to
explain a 15% lengthening or 6.6 A
̊
extension of a resting
domain during the force transition (3), which agrees with
previous SMD observations (9). It is suggested that the force
transition in the first peak corresponds to the accumulation
transitional lengthening of all the folded domains, whereas
the reappearance of smaller transitions in subsequent peaks
describes the process as being reversible before the next
FIGURE 3 Comparison of unfolding peaks obtained using a range of
oscillation amplitudes. Peaks for
A
¼
4.2 nm and 6.2 nm mostly show a clear
unfolding intermediate (
arrows
) in the frequency shift (the intermediate is
not clear for three peaks in
A
¼
4.2 nm,
no arrows
) compared to
A
¼
11.5
nm. For
A
¼
26.5 nm, the peaks do not correspond to a specific unfolded
domain number, as measurements for this amplitude showed irregularities in
the peak spacing. The frequency shift of the peaks increases with a decrease
in
A
according to
D
f

D
F
/
A
1.5
, as given in Eq. 1.
FIGURE 4 (
A
) Diagram showing method for the conversion of the
frequency shift curves to a force. The elastic response of the unfolded
polypeptide region in the frequency shift curves was fit to a polynomial
(
smooth trace
), which was then reversed so that the peak maximum occurred
at zero extension distance that corresponded to the origin of a hypothetical
surface. (
B
) The force (
solid line
) was then calculated from values of the fit
using Eq. 1. The mean forces were 75
6
19.9 pN (
n
¼
37), 63.1
6
20.2 pN
(
n
¼
54), and 63.9
6
7.9 (
n
¼
14) pN for
A
values of 4.5 nm, 6.2 nm, and
11.5 nm, respectively. The intermediate unfolding forces were typically half,
or just below half, of the unfolding force, as observed in the force curve. The
corresponding reversed frequency shift curve (
dotted line
) is plotted for
comparison and shown to reflect the force profile.
644
Higgins et al.
Biophysical Journal 90(2) 640–647
unfolding event takes place. However, the latter point is still
vague, as a review of the literature reveals that it is very
difficult to discern the transition past the second unfolding
peak, or third peak at most, even when many domains are
observed to have unfolded. This observation is also apparent
in the static mode curves measured in this study. To address
this issue, we have shown here that FM detection is able to
detect individual transitions associated with each domain
regardless of the number observed in the force-extension
profile. This is perhaps in most part due to the FM detection
technique being more sensitive to the force in the
z
direction
compared to the static mode. The transition cannot be
explained due to refolding due to the following: partial
refolding (i.e., the entire domain is not relaxed) would only
be possible on the downswing of an oscillation cycle and
where the cantilever spends most of its time at the bottom of
the cycle. However, the time period for a complete
oscillation cycle (1/
f
¼
54
m
s) is much smaller than the
time required for domain refolding (5). In addition, a greater
opportunity for refolding should be the case for the larger
amplitude of 11.5 nm due to sampling farther back (i.e.,
during the downswing of an oscillation) along the refolding
pathway, though these measurements showed only peaks
corresponding to the unfolding of the domains, indicating
that refolding was not evident.
The ability to detect both discrete unfolding events and
their intermediate in titin I27 is dependent on the size of the
A
being smaller than the length scale of the interactions in-
tended for measurement. In this case, the predominant length
scales for an individual unfolding event and its intermediate
are
;
25 nm and 10–15 nm, respectively. For amplitudes
much greater than these length scales, such as the 26.5 nm
used here, it is apparent that the detected frequency shifts
become smaller and less clear. This decrease in the frequency
shift is expected as the interaction force of a single unfolding
domain is averaged over a larger distance in comparison to
smaller amplitude measurements, i.e., a shorter portion of the
oscillation cycle experiences the interaction force for larger
amplitudes. Furthermore, in the case of modular proteins,
a single oscillation cycle with large amplitude may contain
different components of the unfolding force profile, in-
cluding unfolding one domain while stretching another. This
can lead to discontinuities in the force and make inter-
pretation of the data difficult. For an
A
value of 11.5 nm,
discrete frequency shifts can be detected for each unfolded
domain, though the intermediate is smoothed out due to a
decrease in the frequency shift when sampling this com-
parable length scale. In contrast, both of the smaller
A
values
(4.5 nm and 6.2 nm) are most sensitive to both the length
scales in question. Due to the complexity of protein structure
and composition, the length scales of interactions de-
termining unfolding pathways will be highly variable and
be protein specific. Therefore, the ideal situation for these
types of measurements would be to use the smallest achiev-
able amplitude to account for all possible interactions. Here
in this study, the use of smaller amplitudes was limited due to
the stability of the oscillation and size of the frequency noise,
though further AFM modifications are being made in an
attempt to reduce amplitudes to subnanometer values.
As predicted by Eq. 1, similar unfolding peak forces were
obtained, irrespective of
A
, which supports the validity of the
approach. Thus, it remains unclear why the measured forces
(

70 pN) are significantly lower than the forces obtained in
static force measurements (

208 pN). This observation is in
contrast to previous dynamic studies where measured forces
are comparable to the statically measured forces of the same
system. For example, two different dynamic approaches were
able to measure the chair-to-boat transition forces (700–1000
pN) of single dextran polysaccharides (12,14), which were
comparable to those measured independently in earlier static
mode studies (27). For the latter dynamic studies, the forces
were obtained by filtering out the dynamic signal to obtain
the static deflection of the cantilever. In contrast, here, the
force is determined directly from the dynamic signal
(frequency shift). We now discuss possible reasons for the
lower forces observed with our dynamic technique. Errors in
the measured frequency shift may arise if the interaction
response of the molecule occurs on a shorter timescale than
the time constant of the PLL. In which case, a positive fre-
quency shift would be underestimated (as too the force) due
to the delayed response of the PLL not being able to track the
frequency shift increase. However, for the measurements
here, the 400 Hz bandwidth of the PLL is sufficient to
correctly track significantly larger frequency shifts during the

35 ms timescale of the elastic response of the unfolded poly-
peptide. Therefore, the frequency shifts recorded here would
pertain to being real. The converted forces were obtained
using a recently developed arbitrary amplitude formula (Eq. 1)
(25) that unifies two previous well-established formulas that
describe the relationship between the frequency shift and force
(26,28,29). The fact that the unfolding forces scale correctly
with respect to
A
validates the unification of the two previous
formulas within Eq. 1. This arbitrary amplitude formula has
also been experimentally validated by quantifying the struc-
tural forces of liquids whereby the forces scaled appropriately
with
A
and, importantly, agreed with the forces previously
measured for the same liquid using a different technique (i.e.,
surface force apparatus) (30). Without any previous constant
amplitude FM measurements on modular proteins for
comparison, we speculate that the lower unfolding forces
may relate to the discontinuity in the force interaction during
a single oscillation cycle of the tip as the domain unfolds (e.g.,
during the intermediate). This discontinuity in the force may
occur if bonds are broken during the upswing of an oscillation
cycle and are unable to reform on the timescale of the
cantilever oscillation, even when the domains are allowed to
relax during the downswing. As a result, the oscillating can-
tilever would be repeatedly sampling through the force
interaction of already unfolded domain regions, with the
effective averaging of the force resulting in potentially lower
FM-AFM for Unfolding Proteins
645
Biophysical Journal 90(2) 640–647
values. In addition, the potential for the force to vary as
a function of the measurement technique has also recently
been observed for the folding and refolding of individual
proteins (31). To alleviate possible discontinuities in the force,
appropriate modifications to the force analysis have been
outlined in the Material and Methods section and other studies
(11), though further investigation is required to assess the
possible additional effects due to the disruption of hydrogen
bonds during the intermediate.
As mentioned, previous dynamic AFM studies have shown
that extended single polysaccharide chains also display force
transitions that are comparable to those previously observed
in static mode measurements (12–14), and novel transitions
and intermediates have also been reported in recent experi-
ments on proteins. Both Mitsui et al. (17) and Okajima et al.
(18) measured an out-of-phase response in the cantilever
deflection signal for a partially unfolded protein and attrib-
uted the response to refolding. Recently, Janovjak et al. (19)
dynamically unfolded single bacteriorhodopsin proteins from
native purple membranes to reveal novel force peaks that
were ascribed to an intermediate involving kinks in the
a
-helices, though the possibility of refolding due to relax-
ation was noted.
CONCLUSION
The aim of this study was to investigate the possibility of
detecting unfolding intermediates in I27 domains using FM
detection, as the dynamic technique was proposed to be more
sensitive in regions where tip instabilities normally occur for
typical static mode measurements and to changes in the force
gradient along the force-extension profile. Although the pro-
tein unfolding response in dynamic measurements was more
complex compared to static measurements, it was clearly
shown that the unfolding intermediate could be observed for
each unfolded domain. Thus, this approach opens up the
possibility of detecting novel unfolding intermediates in
similar protein studies. Future work is ongoing to make a full
statistical analysis of the effects of the oscillation amplitude
and resonance frequency on the unfolding response of the
protein. In addition, the implementation of subangstrom
oscillation amplitude measurements is currently being de-
veloped and modified for operation in the liquid environment.
We gratefully acknowledge Annette Steward and Jane Clarke (University of
Cambridge, UK) as the source of the wild-type TI I27 multimer protein. We
also thank Irene Revenko (Asylum Research) for her assistance.
This research was supported by Science Foundation Ireland Research Grant
(01/PI.2/C033) and the Human Frontier Science Program.
REFERENCES
1. Best, R. B., B. Li, A. Steward, V. Daggett, and J. Clarke. 2001. Can
nonmechanical proteins withstand force? Stretching barnase by atomic
force microscopy and molecular dynamics simulation.
Biophys. J.
81:2344–2356.
2. Rief, M., M. Gautel, F. Oesterhelt, J. M. Fernandez, and H. E. Gaub.
1997. Reversible unfolding of individual titin immunoglobulin do-
mains by AFM.
Science.
276:1109–1112.
3. Marszalek, P. E., H. Lu, H. Li, M. Carrion-Vazquez, A. F. Oberhauser,
K. Schulten, and J. M. Fernandez. 1999. Mechanical unfolding inter-
mediates in titin modules.
Nature.
402:100–103.
4. Carrion-Vazquez, M., A. F. Oberhauser, S. B. Fowler, P. E. Marszalek,
S. E. Broedel, J. Clarke, and J. M. Fernandez. 1999. Mechanical and
chemical unfolding of a single protein: a comparison.
Proc. Natl. Acad.
Sci. USA.
96:3694–3699.
5. Best, R. B., S. B. Fowler, J. L. Toca-Herrera, A. Steward, E. Paci, and
J. Clarke. 2002. Mechanical unfolding of a titin Ig domain: structure
of unfolding intermediate revealed by combining AFM, molecular
dynamics simulations, NMR and protein engineering.
J. Mol. Biol.
322:841–849.
6. Best, R. B., S. B. Fowler, J. L. Toca-Herrera, and J. Clarke. 2002.
A simple method for probing the mechanical unfolding pathway
of proteins in detail.
Proc. Natl. Acad. Sci. USA.
99:12143–
12148.
7. Best, R. B., and J. Clarke. 2002. What can atomic force microscopy tell
us about protein folding?
Chem. Commun.
3:183–192.
8. Fisher, T. E., A. F. Oberhauser, M. A. Carrion-Vazquez, P. E.
Marszalek, and J. M. Fernandez. 1999. The study of protein me-
chanics with the atomic force microscope.
Trends Biochem. Sci.
24:
379–384.
9. Lu, H., B. Isralewitz, A. Krammer, V. Vogel, and K. Schulten. 1998.
Unfolding of titin immunoglobulin domains by steered molecular dy-
namic simulations.
Biophys. J.
75:662–671.
10. Chtcheglova, L. A., G. T. Shubeita, S. K. Sekatskii, and G. Dietler.
2004. Force spectroscopy with a small dithering of AFM tip: a
method of direct and continuous measurement of the spring constant
of single molecules and molecular complexes.
Biophys. J.
86:
1177–1184.
11. Higgins, M. J., C. Riener, T. Uchihashi, J. Sader, R. A. McKendry, and
S. P. Jarvis. 2005. Frequency modulation atomic force microscopy:
a dynamic measurement technique for biological systems.
Nanotech-
nology.
16:85–89.
12. Humphris, A. D. L., J. Tamayo, and M. J. Miles. 2000. Active quality
factor control in liquids for force spectroscopy.
Langmuir.
16:7891–
7894.
13. Humphris, A. D. L., M. Antognozzi, T. J. McMaster, and M. J. Miles.
2002. Transverse dynamic force spectroscopy: a novel approach to
determining the complex stiffness of a single molecule.
Langmuir.
18:1729–1733.
14. Kawakami, M., K. Byrne, B. S. Khatri, T. C. B. Mcleish, S. E.
Radford, and D. A. Smith. 2005. Viscoelastic measurements of single
molecules on a millisecond time scale by magnetically driven os-
cillation of an atomic force microscope cantilever.
Langmuir.
21:
4765–4772.
15. Liu, Y. Z., S. H. Leuba, and S. M. Lindsay. 1999. Relationship
between stiffness and force in single molecule pulling experiments.
Langmuir.
14:8547–8548.
16. Lantz, M. A., S. P. Jarvis, H. Tokumoto, T. Martynski, T. Kusumi, C.
Nakamura, and J. Miyake. 1999. Stretching the a-helix—a direct
measure of the hydrogen bond energy of a single peptide molecule.
Chem. Phys. Lett.
315:61–68.
17. Mitsui, K., K. Nakajima, H. Arakawa, M. Hara, and A. Ikai. 2000.
Dynamic measurement of single protein’s mechanical properties.
Biochem. Biophys. Res. Commun.
272:55–63.
18. Okajima, T., H. Arakawa, M. T. Alam, H. Sekiguchi, and A. Ikai.
2004. Dynamics of a partially stretched protein molecule studied using
an atomic force microscope.
Biophys. Chem.
107:51–61.
19. Janovjak, H., D. Mu
̈ller, and A. D. L. Humphris. 2005. Molecular force
modulation spectroscopy revealing the dynamic response of single
bacteriorhodopsins.
Biophys. J.
88:1423–1431.
646
Higgins et al.
Biophysical Journal 90(2) 640–647
20. Forbes, J. G., and K. Wang. 2004. Simultaneous dynamic stiffness and
extension profiles of single titin molecules: nanomechanical evidence
for unfolding intermediates.
J. Vac. Sci. Technol. A.
22:1439–1443.
21. Steward, A., J. L. Toca-Herrera, and J. Clarke. 2002. Versatile cloning
system for construction of multimeric proteins for use in atomic force
microscopy.
Protein Sci.
11:2179–2183.
22. Jarvis, S. P., A. Oral, T. P. Weihs, and J. B. Pethica. 1993. A novel
force microscope and point contact probe.
Rev. Sci. Instrum.
64:3515–
3520.
23. Jarvis, S. P., T. Uchihashi, T. Ishida, Y. Nakayama, and H. Tokumoto.
2000. Local solvation shell measurement in water using a carbon
nanotube probe.
J. Phys. Chem. B.
104:6091–6094.
24. Jarvis, S. P., T. Ishida, T. Uchihashi, Y. Nakayama, and H. Tokumoto.
2001. Frequency modulation detection atomic force microscopy in the
liquid environment.
Appl. Phys. A.
72:S129–S132.
25. Sader, J. E., and S. P. Jarvis. 2004. Accurate formulas for interaction
force and energy in frequency modulation force spectroscopy.
Appl.
Phys. Lett.
84:1801–1803.
26. Giessibl, F. 2003. Advances in atomic force microscopy.
Rev. Mod.
Phys.
75:949–983.
27. Marszalek, P. E., H. Li, and J. M. Fernandez. 2001. Fingerprinting
polysaccharides with single-molecule atomic force microscopy.
Nat.
Biotechnol.
19:258–262.
28. Du
̈rig, U. 1999. Relations between interaction force and frequency shift
in large amplitude dynamic force microscopy.
Appl. Phys. Lett.
75:
433–435.
29. Albrecht, T. R., P. Gru
̈tter, D. Horne, and D. Rugar. 1991. Frequency
modulation detection using high-Q cantilevers for enhanced force
microscope sensitivity.
J. Appl. Phys.
69:668–673.
30. Uchihashi, T., M. J. Higgins, Y. Nakayama, J. E. Sader, and S. P.
Jarvis. 2005. Quantitative measurement of solvation shells using fre-
quency modulation atomic force microscopy.
Nanotechnology.
16:
49–53.
31. Cecconi, C., E. A. Shank, C. Bustamante, and S. Marqusee. 2005.
Direct observation of the three-state folding of a single protein
molecule.
Science.
309:2057–2060.
FM-AFM for Unfolding Proteins
647
Biophysical Journal 90(2) 640–647