You are on page 1of 8

article

published online: 17 March 2013 | doi: 10.1038/nchembio.1213

A single-molecule dissection of ligand binding to a


protein with intrinsic dynamics
Eunkyung Kim1,8, Sanghwa Lee2,3,8, Aram Jeon4,5, Jung Min Choi1, Hee-Seung Lee4,5, Sungchul Hohng2,3,6* &
Hak-Sung Kim1,7*

Protein dynamics have been suggested to have a crucial role in biomolecular recognition, but the precise molecular mechanisms
remain unclear. Herein, we performed single-molecule fluorescence resonance energy transfer measurements for wild-type
maltose-binding protein (MBP) and its variants to demonstrate the interplay of conformational dynamics and molecular recog-
nition. Kinetic analysis provided direct evidence that MBP recognizes a ligand through an ‘induced-fit’ mechanism, not through
the generally proposed selection mechanism for proteins with conformational dynamics such as MBP. Our results indicated
that the mere presence of intrinsic dynamics is insufficient for a ‘selection’ mechanism. An energetic analysis of ligand binding
implicated the critical role of conformational dynamics in facilitating a structural change that occurs upon ligand binding.

T
© 2013 Nature America, Inc. All rights reserved.

he function of proteins depends on the recognition and bind- in a time-resolved manner at the single-molecule level19,20. Unlike
ing of specific ligands. Recent studies have revealed that many other experimental methods, single-molecule fluorescence reso-
proteins exist in different structural substates even without nance energy transfer (smFRET) measurements enable a kinetic
the existence of cognate ligands1–3, which suggests that conforma- analysis of the conformational dynamics of proteins. A comparison
tional dynamics has prominent roles in many biological processes, of the opening and closing kinetics in the absence and presence
including molecular recognition4–8. The presence of conformational of varying ligand concentrations provides crucial insight into the
substates has implicated a linkage between protein dynamics and molecular recognition process. As simply depicted in Figure 1b,
molecular recognition, leading to the development of dynamics- for the induced-fit model, the closing rate is expected to acceler-
coupled recognition mechanisms9–11. In the induced-fit model12–15, ate with an increase in the ligand concentration, whereas no change
a ligand binds a highly populated open conformation of a protein, in the closing rate is observed with the conformational selection
followed by a structural adjustment to a ligand-bound, fully closed mechanism. However, such a simple scheme has never been realized
state (Fig. 1a). In contrast, the conformational selection mechanism because intrinsic conformational dynamics are too rare and brief to
assumes that a ligand selectively binds a weakly populated single be directly observed through smFRET.
substate in the ensemble of conformations, and no further confor- Herein, we demonstrate the connection between the confor-
mational change occurs after ligand binding (Fig. 1a). In several mational dynamics and molecular recognition of proteins in the
cases, it has been shown that apo proteins sample a partially closed process of ligand binding through a real-time kinetic analysis using
form that resembles the ligand-bound form but can be distinguished smFRET measurements. A wild-type MBP and its variants were
from it by fine details. It has therefore been suggested that the con- used in this study. Ligand binding of MBP, which is involved in
formational selection and induced-fit mechanisms are not mutu- the active transport of maltodextrins in Escherichia coli, has been
npg

ally exclusive but are cooperative for the molecular recognition considered a classic example of the induced-fit model because of
process16–18. In the extended conformational selection model, which the structural differences between its ligand-free and ligand-bound
suggests a more general framework, the conformational selection forms21. However, recent paramagnetic NMR experiments and
mechanism works for the initial selective binding of a ligand to a molecular dynamics simulations have revealed that a ligand-free
partially closed form and is followed by a subsequent fine structural MBP undergoes dynamic structural transitions between a major
adjustment through an induced-fit mechanism. open form and a minor partially closed form22,23, raising an issue
Despite the evidence for protein dynamics and conceptual in the ligand recognition process. A previous study reported that
advances, however, molecular mechanisms that drive ligand recogni- ligands are accessible to the sugar-binding surface of the C-terminal
tion and the role of conformational dynamics in the molecular recog- domain in the partially closed conformation of an apo wild-type
nition of proteins remain controversial and elusive mainly owing to MBP22. It has therefore been suggested that ligands might pref-
experimental limitations. To truly understand the interplay between erentially bind the partially closed form of an MBP through a
protein dynamics and the molecular recognition mechanism, an conformational selection mechanism, followed by fine confor-
experimental method to directly correlate the binding of ligands to mational adjustment through the induced-fit mechanism4. We
the conformational dynamics of proteins in real time is necessary. In explored the molecular recognition mechanism of an MBP on the
this respect, a single-molecule analysis offers the potential to resolve basis of a kinetic analysis of the conformational dynamics using
the controversy by detecting transient ­intermediates or rare events smFRET measurements.

Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Korea. 2Department of Physics and Astronomy, Seoul
1

National University, Seoul, Korea. 3National Center for Creative Research Initiatives, Seoul National University, Seoul, Korea. 4Department of Chemistry,
Korea Advanced Institute of Science and Technology, Daejeon, Korea. 5Molecular-Level Interface Research Center, Korea Advanced Institute of Science and
Technology, Daejeon, Korea. 6Biophysics and Chemical Biology, Seoul National University, Seoul, Korea. 7Graduate School of Nanoscience and Technology,
Korea Advanced Institute of Science and Technology, Daejeon, Korea. 8These authors contributed equally to this work. *e-mail: shohng@snu.ac.kr or
hskim76@kaist.ac.kr

nature CHEMICAL BIOLOGY | vol 9 | MAY 2013 | www.nature.com/naturechemicalbiology 313


article Nature chemical biology doi: 10.1038/nchembio.1213

RESULTS a Induced fit


Intrinsic dynamics of MBPs Po L Po-L
To investigate the intrinsic dynamics of MBP in the absence of a
ligand using smFRET, we first labeled wild-type MBP with a biotin
+
at its N terminus and with a FRET pair (Cy3 and Cy5) on the internal
residues. We selected the dye-labeling positions (Lys34 and Arg354)
such that distinct changes in FRET efficiency arise during the con-
formational transitions between the apo and holo forms (Fig. 2).
Both a CD analysis and an amylose-binding assay confirmed that
the effect of cysteine introduction and dye labeling on the biological
activity of wild-type MBP was negligible (Supplementary Results, +

Supplementary Fig. 1). To observe the real-time conformational


dynamics of the proteins, we immobilized dye-labeled MBPs on
Pc L Pc-L
a PEG-coated quartz surface through a streptavidin-biotin inter- Conformational
selection
action (Supplementary Fig. 2) and imaged using a total internal
reflection fluorescence microscope (Supplementary Fig. 3). As the
paramagnetic NMR experiments revealed, a ligand-free wild-type b No ligand

MBP is expected to be in dynamic equilibrium between a major Tclosing

FRET
open state (~95%) and a weakly populated partially closed state
(~5%)22. However, we could not detect such fast conformational
dynamics in the ns-μs range22 through smFRET analysis with a Conformational selection Induced fit
15-ms time resolution (Fig. 3a). In an effort to increase the time + Ligand + Ligand
resolution, we conducted smFRET measurements using a 128 pixel
© 2013 Nature America, Inc. All rights reserved.

FRET
FRET
× 128 pixel electron-multiplying charge-coupled device camera with
a 2-ms time resolution. Nonetheless, no distinct transitions between
++ Ligand ++ Ligand
the two substates were detected for wild-type MBP in the absence or
presence of maltose (Supplementary Fig. 4).

FRET
FRET

It has been previously shown that mutations at the hinge region


of MBP affect its conformational equilibrium and ligand binding Time Time
affinity24,25; the mutants in this region had greatly increased ligand
binding affinities compared to the wild type, mainly because of the Figure 1 | Molecular recognition processes for induced-fit and conformational
changes in the off rates. In addition, the binding affinities of the selection mechanisms. (a) Thermodynamic cycles for conformational
Ile329 mutants for maltose show a strong correlation with the side selection and induced-fit mechanisms. Po, Pc and L indicate the protein’s
chain volume of the amino acid at position 329 (ref. 24). The bind- open conformation, closed conformation and a ligand, respectively. (b) The
ing affinity of maltose for an I329W mutant, whose side chain is expected time traces of the FRET efficiency in smFRET measurements for
larger than that of the I329F mutant but smaller than that of I329C- the respective recognition mechanisms. Tclosing indicates the time for a protein
thionitropyridine (I329C-TNP), lies between those measured for to undergo a conformational change from an open to a closed state in the
I329F and I329C-TNP24. On the basis of these findings, we expected absence or presence of varying ligand concentrations. Plus symbols represent
that introduction of bulky amino acids into the hinge region would the ligand concentration. Arrowheads indicate the binding of a ligand.
facilitate a kinetic analysis using smFRET by slowing down the
intrinsic conformational dynamics. Specifically, we attempted to use of conformational dynamics and the molecular recognition
single (I329W) and double (A96W I329W) mutants of the MBP. We mechanism of proteins4,26. If the ligands bind a highly populated
npg

constructed the I329W and A96W I329W mutants with two cysteine open form and induce a conformational transition of the protein
mutations, K34C and R354C, for dual dye labeling as described above to a closed form, the closing rates will be faster in the presence of
for wild-type MPB. The CD spectra of the wild type and the mutants a ligand than in its absence and will be accelerated by increasing
were almost identical (Supplementary Fig. 5), indicating negligible ligand concentration (Fig. 1b). In contrast, if the ligands selectively
differences in their secondary structures. In comparison to the wild bind a weakly populated, partially closed conformation, the clos-
type, which has a maltose dissociation constant of 2,180 ± 190 nM ing rates in the presence of a ligand will be the same as those with
(mean ± s.d.), the mutants showed increased ligand binding affini- no ligand because the population of a ligand-binding conformer
ties, as determined through isothermal titration calorimetry (ITC): is limited by the intrinsic conformational transition rate (Fig. 1b).
48 ± 19 nM for the single mutant (I329W) and 4.7 ± 3.3 nM for Therefore, to demonstrate the molecular recognition mechanism,
the double mutant (A96W I329W; Table 1). We tested whether the the intrinsic protein dynamics and the dependence of the closing
intrinsic conformational dynamics of the mutants could be detected rate on the ligand concentration should be determined. Figure 4a
in the absence of maltose, and we were able to clearly observe the shows the representative intensity and FRET time traces for the
transitions between the two distinct conformational states (Fig. 3b,c). single (I329W) mutant at varying maltose concentrations. The
The low- and high-FRET states were interpreted as open and par- high-FRET state of the single mutant became more populated with
tially closed forms, respectively. We estimated the partially closed increasing maltose concentrations.
states of the single and double mutants to be 14% and 42% of the It was noticeable that the ligand-free partially closed form
total populations, respectively (Fig. 3d). A kinetic analysis revealed and the ligand-bound closed form were indistinguishable in our
that the closing and opening rates of the double mutant are much smFRET assay. To verify that the FRET changes came from the
slower than those of the single mutant (Fig. 3e,f). conformational transitions between the two substates as predicted
by the crystal structures, we attempted to determine the inter-dye
Effects of ligand on the kinetic rates of MBPs distances on the basis of the FRET efficiencies. Although no crys-
We next examined the ligand recognition process of MBPs in con- tal structure of the single mutant is available, it was reported that
nection with intrinsic dynamics. A kinetic analysis in the absence mutation of the single mutant most likely predominantly affects
and presence of a ligand provides direct evidence for the ­interplay the ligand-free state of the protein, suggesting a closure angle of

314 nature chemical biology | vol 9 | MAY 2013 | www.nature.com/naturechemicalbiology


Nature chemical biology doi: 10.1038/nchembio.1213 article
K34C volume simulation. In the simulation, we used geometric constraints
K34C
based on the structural data for ligand-free and ligand-bound wild-
type MBPs and physical properties of dyes and flexible linkers (for
example, 1 nm for linker length, 0.5 nm for linker diameter and
0.7 nm for dye diameter). The inter-dye distances from the mea-
sured FRET efficiencies were comparable to those measured by the
­probability-based analysis method, supporting the conclusion that the

5.04 nm
5.92 nm
A96 A96
I329
I329
FRET changes in our study reflected the conformational transitions
between the two substates as predicted by the crystal structures.
The kinetic analysis of the closing and opening transitions for
the single mutant at varying ligand concentrations revealed that
Maltose the closing rate was faster in the presence of the ligand than in its
R354C absence and that this rate increased linearly with an increase in the
R354C
ligand concentration (Fig. 4b and Supplementary Fig. 7a). In con-
trast, the opening rates remained almost constant over the range of
Figure 2 | Experimental design for analyzing the conformational
tested ligand concentrations or in the absence of the ligand (Fig. 4c
dynamics of MBP. Structures of E. coli MBPs (Protein Data Bank codes
and Supplementary Fig. 7b). A similar behavior was observed with
1OMP (ligand-free, cyan) and 1ANF (ligand-bound, yellow)). The two
maltotriose (Fig. 4b,c and Supplementary Fig. 8) but not with
cysteine mutations (K34C, R354C) introduced for conjugating the dyes
sucrose, which does not bind MBP (Fig. 4b,c and Supplementary
are indicated. The distances between the Cα atoms of these residues in
the apo and holo proteins were estimated to be 5.92 nm and 5.04 nm,
Fig. 9). To test whether our smFRET analysis reliably monitors single
respectively. The mutation sites for the construction of single (A96W) and
ligand binding events, we determined the dissociation constant (Kd)
double (A96W I329W) mutants are also shown.
on the basis of the probability of the high FRET state (Online Methods
and Supplementary Fig. 10a–c). In the single mutant, we obtained
© 2013 Nature America, Inc. All rights reserved.

the dissociation constants of 99 nM for maltose and 56 nM for


9.5° ± 0.9° (mean ± s.d.) compared to wild-type MBP, with a subtle maltotriose (Supplementary Fig. 10a). These dissociation constants
change in the structure of the ligand-bound state25. It is therefore were similar to those obtained from bulk measurements (Table 1).
likely that both conformers of the single mutant will not differ from Qualitatively, we obtained the same kinetic effects of the ligands
those of wild-type MBP. Our previous study estimated the Förster from the double mutant (A96W I329W) for maltose and maltotriose
distance (R0) for a Cy3-Cy5 FRET pair to be 5.4 nm (ref. 27). Using (Fig. 4d,e and Supplementary Figs. 11 and 12), implying the gen-
the R0 value and FRET efficiencies in the presence of 100 nM malt- erality of our observation. We also obtained similar dissociation
ose for the MBP single mutant (Supplementary Fig. 6), we deter- constants for the double mutant from smFRET analysis and bulk
mined the inter-dye distances for both conformers (Supplementary measurements (Supplementary Fig. 10b and Table 1). For wild-
Table 1). We also calculated the inter-dye distances through a type MBP, we were not able to detect a transition between two sub-
probability-based analysis method using the Model Satellite Prior states through smFRET in the presence of maltose (Supplementary
Algorithm28 (Supplementary Table 1). This method enables the Fig. 13), mainly because the opening rate was too fast. In contrast,
computation of the three-dimensional probability density func- in the presence of maltotriose, the transition between the two sub-
tion for the position of a dye molecule through geometric accessible states was detectable, owing to a much longer dwell time in the

Figure 3 | Analysis of intrinsic dynamics a Cy3 Cy5 FRET d 0.6


of MBPs through smFRET. (a–c) The
Intensity

representative time traces of fluorescence 500


(AU)

Cy3 Cy5

partial closed form


0.4
npg

intensity (top) and FRET efficiency (bottom) 0


Ratio of

0.8
for wild-type MBP (a), the single mutant
FRET

0.4 0.2
(I329W) (b) and the double mutant (A96W
0
I329W) (c) in the absence of a ligand. The
0 2 4 6 8 10
decrease of FRET to 0 at around 9.7 s in Time (s) 0
a is mainly due to a bleaching of Cy5 dye.
SM

DM

Orange lines were added as a guide to make


1,000 0.08
the transitions clearer. Single and double b e
Intensity
(AU)

500
mutations were introduced in the hinge region,
closing rate (s )

0.06
–1

0
opposite the maltose-binding site of the MBP.
Intrinsic

0.8
(d) The relative population of the partially 0.04
FRET

0.4
closed state of the mutants in the absence 0
0.02
of maltose. The respective populations were 0 50 100 150
Time (s)
determined using the rates obtained from the 0
average dwell times for the open and partially
SM

DM

closed states. SM, single mutant; DM, double


mutant. (e,f) A comparison of the intrinsic c f 0.5
1,000
closing rates (e) and opening rates (f) of
Intensity
(AU)

500 0.4
opening rate (s )

the mutants in the absence of maltose. The


–1

0
intrinsic dynamics were analyzed from 48 and
Intrinsic

0.8 0.3
141 molecules showing a distinct transition
FRET

0.4
between two states among the single and 0.2
0
double mutants from three and four movies, 0 500 1,000 1,500 0.01
respectively. The error bars represent the s.d. Time (s) 0
of three or four data sets.
SM

DM

nature CHEMICAL BIOLOGY | vol 9 | MAY 2013 | www.nature.com/naturechemicalbiology 315


article Nature chemical biology doi: 10.1038/nchembio.1213

Table 1 | Energetic and kinetic parameters for wild-type MBP and its variants
Energy of conformational Binding energy Dissociation constants Dissociation constants
Protein change («c, kcal mol−1) (D«, kcal mol−1) (Kd, nM)a (Kd, nM)b
Wild type 1.77c N.D. (maltose) N.D. 2,180 ± 190
–8.91 (maltotriose) 715 1,950 ± 260
Single mutant (I329W) 1.09 –9.48 (maltose) 99 48 ± 19
–9.82 (maltotriose) 56 53 ± 15
Double mutant (A96W I329W) 0.19 –10.39 (maltose) 7.2 4.7 ± 3.3
–10.19 (maltotriose) 10 2.3 ± 1.0
a
The dissociation constants were determined from a hyperbolic Hill fit of the relative population of a closed state calculated from the transition rates (Hill coefficient = 1.0) (Supplementary Fig. 10).
b
The dissociation constants were determined through ITC. Data represent the mean ± s.d. in triplicate experiments. cThe energy of conformational change for wild-type MBP was calculated on the
basis of data indicating that the probability of an intrinsic closed state is 5%. N.D., not determined.

closed state. The closing and opening dynamics analyzed for vary- with binding in the partially closed form and dissociation in the
ing maltotriose concentrations (Supplementary Fig. 14) showed open form; and type IV (12%), with both binding and dissociation
a similar dependence of the ­binding and dissociation kinetics on in the partially closed form. Overall, more than 80% of binding and
ligand concentration (Fig. 4f,g and Supplementary Fig. 15). We dissociation events occurred in the open form, indicating a strong
were also able to obtain the dissociation constant of maltotriose for preference for the open conformation. In wild-type MBP, the par-
wild-type MBP from the probability of the high FRET state (Online tially closed form is much less populated than it is in the double
Methods), and it was comparable to that obtained through bulk mutant, and a much stronger preference for the open conformation
measurement (Supplementary Fig. 10c and Table 1). Even though is expected. The preference for the open conformation during bind-
the binding and dissociation kinetics could not be compared with ing seems to be due to the higher accessibility to the ligand.
© 2013 Nature America, Inc. All rights reserved.

the intrinsic closing and opening dynamics of the protein in this


case, we believe that the ligand recognition mechanism of the wild Ligand binding energy
type and the mutants are the same. The existence of a partially closed form itself has been interpreted
The analysis of the closing and opening kinetics of the MBPs as supporting evidence that this conformation has a critical role in
showed an acceleration of the closing rate with an increase in the the molecular recognition process4,29. However, our rigorous kinetic
ligand concentration as compared to the intrinsic closing rate. This analysis indicated a limited contribution of a partially closed form to
result therefore strongly supports the idea that the ligands pref- the ligand binding process, raising an issue regarding the biological
erentially bind the open conformation, followed by a structural roles of protein dynamics. We expected that consideration of the
transition of the proteins into a closed form, directly proving the
induced-fit mechanism, which has been the textbook explanation
a b No ligand c
for the molecular recognition process for the past 50 years. It is Cy3 Cy5 FRET Maltose
0.8

Opening rate (s–1)


Closing rate (s )
Maltotriose
Intensity

noteworthy that the MBP recognizes a ligand through an induced- 0.5


–1
(AU)

500 0.4 Sucrose 0.6


fit mechanism and not through the generally proposed selection 0
0.8
0.3 0.4
0.2
mechanism for proteins with conformational dynamics.
FRET

0.4 0.1 0.2


0 0 0
0 50 100 150
0
20
40
60
100
0

0
20
40
60
100
0
Analysis of ligand binding by three-color smFRET
8

8
Time (s) Ligand (nM) Ligand (nM)
To directly quantify the binding preference of a ligand to an open d e
conformation, we devised the use of three-color smFRET mea- 0.03 0.015

Opening rate (s–1)


Closing rate (s )
Intensity

–1
(AU)

500
surements27 using Cy7-labeled maltose (Supplementary Note 1).
npg

0.02 0.010
0
In three-color smFRET, the opening and closing dynamics of 0.8 0.01 0.005
FRET

0.4
MBP are monitored through Cy3-Cy5 FRET or Cy5 intensity, and 0 0 0
binding of maltose is monitored through Cy7 intensity. Owing to 0 50 100
Time (s)
150 0 2 4 6 8 10
Ligand (nM)
0 2 4 6 8 10
Ligand (nM)
steric hindrance by the linker used for dye labeling, we expected
f g
a ­considerable reduction in the binding affinity of Cy7-maltose 12
Opening rate (s–1)
Closing rate (s )
–1

6 10
Intensity

compared to that of unlabeled maltose. Nonetheless, we were able 700


(AU)

8
0 4 6
to observe distinct binding events (jumps of Cy7 signal and con- 0.8 2 4
FRET

0.4
current drops of Cy5 signal) of Cy7-maltose to the double mutant, 0 0
2
0
A96W I329W, at high ligand concentrations (Fig. 5a). Binding 0 50 100 150
200
400
600
8 0
1,000
00
200
40 0
600
8 0
1,000
00

Time (s)
events were very short, suggesting that the modification on malt- Ligand (nM) Ligand (nM)
ose was a hindrance. The increase in Cy7 intensity indicates the
binding of Cy7-maltose to the MBP, and the subsequent decrease Figure 4 | Kinetic analysis of conformational dynamics for MBPs in the
reflects the release of the ligand from the MBP. It was noteworthy presence of a ligand. (a) Representative time traces of the fluorescence
that the binding and dissociation events of Cy7-maltose occurred intensity and FRET efficiency for the single mutant (I329W) at varying
in both the open and closed conformations of MBPs but happened maltose concentrations: top, no maltose; middle, 20 nM; bottom, 100 nM.
more frequently in the open conformation (Fig. 5a). To measure the Orange lines were added as a guide to make the transitions clearer.
preference of the ligand binding and dissociation events for the (b,c) Closing rates (b) and opening rates (c) as a function of ligand
open and the partially closed forms, we made a correlation plot of concentration for the single mutant (I329W). (d,e) Closing rates (d) and
Cy3-Cy5 FRET efficiencies both at the binding and dissociation opening rates (e) as a function of ligand concentration for the double mutant
points for the binding events of single Cy7-maltose (Fig. 5b). We (A96W I329W). (f,g) Closing rates (f) and opening rates (g) as a function
could classify single binding and dissociation events into four of ligand concentration for wild-type MBP. In the latter case, a ligand-free,
categories: type I (75% of 479 events), with both binding and the partially closed state was not directly detectable through smFRET, and the
dissociation in the open form; type II (6%), showing binding in the closing rate was set to 0 for the fitting. The error bars represent the s.d.
open form and dissociation in the partially closed form; type III (7%), All data sets used in this analysis are represented in Supplementary Table 2.

316 nature chemical biology | vol 9 | MAY 2013 | www.nature.com/naturechemicalbiology


Nature chemical biology doi: 10.1038/nchembio.1213 article
Figure 5 | Simultaneous analysis of Cy7-maltose a Cy5 signal Cy7 signal b
binding and conformational dynamics of the double 0.7
IV: 12%

(post-bound state)
mutant through three-color smFRET. 0.8

Normalized intensity
II: 6%

FRET efficiency
0.6
0.4
(a) Representative intensity time traces of Cy5
0
fluorescence (red, top) and Cy7 fluorescence (gray, 0.5
0.8
bottom) normalized to the sum of Cy3, Cy5 and Cy7 III: 7%
0.4 0.4
fluorescence intensities in the presence of 400 nM 0
I: 75%

Cy7-maltose under excitation at 532 nm. Yellow- 0 100 200 300 0.4 0.5 0.6 0.7
shaded regions represent the partially closed states Time (s) FRET efficiency
(pre-bound state)
of the double mutant (A96W I329W). The binding
of Cy7-maltose to an open state and a partially closed state are denoted by the open and solid arrowheads, respectively. (b) The relative distribution of
the FRET states before the binding and after the dissociation of Cy7-maltose from 479 binding and dissociation events of 52 molecules. Each data point is
represented through a gray-scaled two-dimensional Gaussian distribution. The shades of gray represent the relative density of the data points or events.
Red dashed lines were added for a clearer visualization of the four species (type I, binding to an open form followed by dissociation from the same state;
type II, binding to an open form and dissociation from a partially closed form; type III, binding to a partially closed form and dissociation from an open form;
and type IV, binding to a partially closed form followed by dissociation from the same state).

energetics of ligand binding for the three MBP variants would ­provide Recent studies with NMR dispersion relaxation and spectro-
some insight. Because ligand binding is associated with structural scopic techniques have revealed the existence of the conformational
changes in the MBPs, the dissociation constant of the ligand reflects substates of proteins29,32–35 and have suggested the necessary roles
both the binding energy and the energy cost required for the con- of the intrinsic dynamics of the protein in molecular recognition,
formational transition. When the binding energy is limited, as is leading to a development of dynamics-coupled recognition mech-
usually observed during small-ligand recognition processes, and if anisms. Much effort has been made to understand the interplay
© 2013 Nature America, Inc. All rights reserved.

the energy cost required for the structural transition is too large, between protein dynamics and molecular recognition. In particular,
physiologically relevant binding affinities cannot be achieved. MBP, a member of the bacterial periplasmic-binding protein (PBP)
Therefore, we determined both the conformational energy costs family, has been the subject of extensive studies owing to its large-
for the closing transition and the ligand binding energies for the scale domain rearrangement and the existence of conformational
wild type and single and double mutants (Online Methods). The substates and exemplifies the controversy regarding the recognition
correlation between the structural energy costs and the dissocia- mechanism22,23. A large conformational change in an MBP by ligand
tion constants observed in the three MBP variants supports this binding is considered a classic example of the induced-fit mecha-
hypothesis (Table 1). It is more likely that the intrinsic dynamics nism. The dynamic structural transitions of MBP between a major
of MBP facilitate a large conformational change that occurs upon open (~95%) form and minor partially closed (~5%) form suggest
sugar binding, which can be achieved with less energy expenditure. that the initial binding of a ligand through conformational selection
This view is consistent with recent studies suggesting that confor- is followed by a conformational adjustment4. Meanwhile, simulation
mational dynamics may have a role in facilitating the transition studies have indicated that the induced-fit mechanism mainly
to a holo structure18,22 or in influencing the chemical step during ­governs the binding of a ligand to MBP, but both conformational
enzyme catalysis30,31. selection and induced-fit mechanisms have a major role in molecu-
lar recognition23. Similarly, two possible pathways were proposed for
DISCUSSION MBP on the basis of accelerated molecular dynamics simulations36:
Our approach using smFRET measurements enabled a kinetic either an induced-fit model or a model involving both induced-
analysis that interconnects the ligand binding and intrinsic con- fit and conformational selection mechanisms. Thus, to dissect the
formational dynamics in the protein, providing direct evidence fundamental aspects of the recognition mechanisms, the intrinsic
npg

demonstrating the molecular mechanism of MBP that drives ligand dynamics and structural changes of the protein should be traced
recognition. Our results showed that the binding of a ligand to the over the entire course of the ligand binding process. Our kinetic
MBP proceeds through the induced-fit mechanism and not through analysis of conformational dynamics in the absence and presence of
the generally proposed conformational selection mechanism, increasing ligand concentration provided direct evidence to dissect
although different conformational substates exist. Our study also the molecular recognition mechanism of MBP. The results of this
revealed that the question of whether the recognition of the interact- study indicated an induced-fit mechanism for MBP, but one that
ing molecules is ‘induced’ or ‘selected’ is a problem of kinetics rather differs from Koshland’s pure induced-fit model, where the protein is
than of thermo­dynamics, which deals with the relative probabilities considered to exist in a single, stable conformation. Taken together
of the conformational substates. In addition, the mere presence of with the three-color smFRET results, we propose an ‘extended
conformational dynamics was insufficient evidence to indicate the induced fit’ model in which the protein exists in the ensemble of
molecular recognition mechanism. An important question to be conformations and the ligands bind the existing substates, inducing
answered is whether the ligand binds both open and partially closed a structural transition to a ligand-bound, closed form. Our model
conformations of the MBP. Our three-color smFRET experiments can be distinguished from the extended conformational selection
using Cy7-labeled maltose as a ligand showed that Cy7-maltose can model, in which the ligands bind only a weakly populated, higher-
bind both open and partially closed conformations of the double energy ‘bound-like’ conformation followed by structural adjustment
mutant and shows a preference for the open conformation, although to a complete ligand-bound form.
a considerable reduction in the binding affinity of Cy7-maltose was The conformational diversity of proteins has been suggested to be
expected, owing to the steric hindrance by the linker used for dye important in the evolution of molecular recognition4. The binding of
labeling. We therefore expect that intact ligands can bind both con- a ligand to all conformations is likely to contribute to low ligand speci-
formations of an apo wild-type MBP but prefer an open conforma- ficity of the protein, whereas the conformational selection mechanism
tion. Because the ligand binds both substates of the MBP, followed gives rise to narrow ligand specificity because generally a single sub-
by a conformational change, the results of three-color smFRET state is engaged in the molecular recognition process. Considering
measurements also indicate that the molecular recognition of MBP MBP’s role in E. coli and its broad specificity for sugars, its molecular
follows an induced-fit mechanism. recognition mechanism will be advantageous for the survival of

nature CHEMICAL BIOLOGY | vol 9 | MAY 2013 | www.nature.com/naturechemicalbiology 317


article Nature chemical biology doi: 10.1038/nchembio.1213

microorganisms on diverse sugars. The presence of different substates 22. Tang, C., Schwieters, C.D. & Clore, G.M. Open-to-closed transition in apo
in an MBP may allow an interaction with more diverse types of sug- maltose-binding protein observed by paramagnetic NMR. Nature 449,
1078–1082 (2007).
ars. It has been suggested that less specific PBPs, such as glutamine-
23. Kondo, H.X., Okimoto, N., Morimoto, G. & Taiji, M. Free-energy landscapes
binding protein (GlnBP), are unlikely to have a partially closed con- of protein domain movements upon ligand binding. J. Phys. Chem. B 115,
formation37, whereas PBPs showing broad specificity, such as lysine-, 7629–7636 (2011).
arginine-, ornithine-binding (LAO) protein38; ribose-binding protein 24. Marvin, J.S. & Hellinga, H.W. Manipulation of ligand binding affinity by
(RBP)39; glucose/galactose-binding protein (GGBP)40; and choline- exploitation of conformational coupling. Nat. Struct. Biol. 8, 795–798 (2001).
25. Millet, O., Hudson, R.P. & Kay, L.E. The energetic cost of domain
binding protein (ChoX)41 seem to have a partially closed substate. reorientation in maltose-binding protein as studied by NMR and fluorescence
Molecular recognitions are central to all biological processes. Both spectroscopy. Proc. Natl. Acad. Sci. USA 100, 12700–12705 (2003).
the induced-fit and conformational selection mechanisms are likely to 26. Bae, S., Kim, D., Kim, K.K., Kim, Y.G. & Hohng, S. Intrinsic Z-DNA is
have crucial roles in molecular recognition. The approach presented stabilized by the conformational selection mechanism of Z-DNA–binding
in this study will provide a new paradigm in elucidating the molec- proteins. J. Am. Chem. Soc. 133, 668–671 (2011).
27. Lee, S., Lee, J. & Hohng, S. Single-molecule three-color FRET with both negligible
ular mechanisms by which other proteins showing conformational spectral overlap and long observation time. PLoS ONE 5, e12270 (2010).
dynamics recognize and bind their cognate ligands, assisting in the 28. Muschielok, A. et al. A nano-positioning system for macromolecular
design of more potent drugs and even proteins with new functions. structural analysis. Nat. Methods 5, 965–971 (2008).
29. Baldwin, A.J. & Kay, L.E. NMR spectroscopy brings invisible protein states
Received 6 August 2012; accepted 21 February 2013; into focus. Nat. Chem. Biol. 5, 808–814 (2009).
published online 17 March 2013 30. Eisenmesser, E.Z. et al. Intrinsic dynamics of an enzyme underlies catalysis.
Nature 438, 117–121 (2005).
31. Bhabha, G. et al. A dynamic knockout reveals that conformational fluctuations
Methods influence the chemical step of enzyme catalysis. Science 332, 234–238 (2011).
Methods and any associated references are available in the online 32. Mulder, F.A., Mittermaier, A., Hon, B., Dahlquist, F.W. & Kay, L.E. Studying
version of the paper. excited states of proteins by NMR spectroscopy. Nat. Struct. Biol. 8, 932–935
(2001).
33. Vallurupalli, P. & Kay, L.E. Complementarity of ensemble and single-molecule
© 2013 Nature America, Inc. All rights reserved.

References measures of protein motion: a relaxation dispersion NMR study of an enzyme


1. Henzler-Wildman, K. & Kern, D. Dynamic personalities of proteins. Nature complex. Proc. Natl. Acad. Sci. USA 103, 11910–11915 (2006).
450, 964–972 (2007). 34. Boehr, D.D., McElheny, D., Dyson, H.J. & Wright, P.E. The dynamic energy
2. Korzhnev, D.M., Religa, T.L., Banachewicz, W., Fersht, A.R. & Kay, L.E. landscape of dihydrofolate reductase catalysis. Science 313, 1638–1642 (2006).
A transient and low-populated protein-folding intermediate at atomic 35. Henzler-Wildman, K.A. et al. Intrinsic motions along an enzymatic reaction
resolution. Science 329, 1312–1316 (2010). trajectory. Nature 450, 838–844 (2007).
3. Bouvignies, G. et al. Solution structure of a minor and transiently formed 36. Bucher, D., Grant, B.J. & McCammon, J.A. Induced fit or conformational
state of a T4 lysozyme mutant. Nature 477, 111–114 (2011). selection? The role of the semi-closed state in the maltose binding protein.
4. Boehr, D.D., Nussinov, R. & Wright, P.E. The role of dynamic conformational Biochemistry 50, 10530–10539 (2011).
ensembles in biomolecular recognition. Nat. Chem. Biol. 5, 789–796 (2009). 37. Bermejo, G.A., Strub, M.P., Ho, C. & Tjandra, N. Ligand-free open-closed
5. Akke, M. NMR methods for characterizing microsecond to millisecond dynamics transitions of periplasmic binding proteins: the case of glutamine-binding
in recognition and catalysis. Curr. Opin. Struct. Biol. 12, 642–647 (2002). protein. Biochemistry 49, 1893–1902 (2010).
6. Popovych, N., Sun, S., Ebright, R.H. & Kalodimos, C.G. Dynamically driven 38. Oh, B.H. et al. Three-dimensional structures of the periplasmic lysine/
protein allostery. Nat. Struct. Mol. Biol. 13, 831–838 (2006). arginine/ornithine-binding protein with and without a ligand. J. Biol. Chem.
7. Masterson, L.R. et al. Dynamics connect substrate recognition to catalysis in 268, 11348–11355 (1993).
protein kinase A. Nat. Chem. Biol. 6, 821–828 (2010). 39. Ravindranathan, K.P., Gallicchio, E. & Levy, R.M. Conformational equlibria
8. Carroll, M.J. et al. Evidence for dynamics in proteins as a mechanism for and free energy profiles for the allosteric transition of the ribose-binding
ligand dissociation. Nat. Chem. Biol. 8, 246–252 (2012). protein. J. Mol. Biol. 353, 196–210 (2005).
9. Frauenfelder, H., Sligar, S.G. & Wolynes, P.G. The energy landscapes and 40. Flocco, M.M. & Mowbray, S.L. The 1.9 Å X-ray structure of a closed
motions of proteins. Science 254, 1598–1603 (1991). unliganded form of the periplasmic glucose/galactose receptor from
10. Tsai, C.J., Kumar, S., Ma, B. & Nussinov, R. Folding funnels, binding funnels, Salmonella typhimurium. J. Biol. Chem. 269, 8931–8936 (1994).
and protein function. Protein Sci. 8, 1181–1190 (1999). 41. Oswald, C., Smits, S.H., Hoing, M., Bremer, E. & Schmitt, L. Structural
npg

11. Wolynes, P.G. Recent successes of the energy landscape theory of protein analysis of the choline-binding protein ChoX in a semi-closed and ligand-free
folding and function. Q. Rev. Biophys. 38, 405–410 (2005). conformation. Biol. Chem. 390, 1163–1170 (2009).
12. Koshland, D.E. Application of a theory of enzyme specificity to protein
synthesis. Proc. Natl. Acad. Sci. USA 44, 98–104 (1958).
13. Goh, C.S., Milburn, D. & Gerstein, M. Conformational changes associated
with protein-protein interactions. Curr. Opin. Struct. Biol. 14, 104–109 (2004). Acknowledgments
14. Evenäs, J. et al. Ligand-induced structural changes to maltodextrin-binding This work was supported by the Pioneer Research Program for Converging
protein as studied by solution NMR spectroscopy. J. Mol. Biol. 309, 961–974 Technology (2008-2000218), Advanced Biomass R&D Center (2011-0031363),
(2001). the Intelligent Synthetic Biology Center (2011-0031950) and the Brain Korea 21
15. Sooriyaarachchi, S., Ubhayasekera, W., Park, C. & Mowbray, S.L. program of the Ministry of Education, Science and Technology (H.-S.K.) and Creative
Conformational changes and ligand recognition of Escherichia coli D-xylose Research Initiatives (2009-0081562) and the World-Class University program of the
binding protein revealed. J. Mol. Biol. 402, 657–668 (2010). National Research Foundation of Korea (S.H.). We thank S. Mowbray and D.D. Boehr
16. Hammes, G.G., Chang, Y.C. & Oas, T.G. Conformational selection or induced for helpful discussions.
fit: a flux description of reaction mechanism. Proc. Natl. Acad. Sci. USA 106,
13737–13741 (2009).
17. Silva, D.A., Bowman, G.R., Sosa-Peinado, A. & Huang, X. A role for both
Author contributions
E.K. and J.M.C. prepared and characterized the MBP variants. E.K. and S.L. designed,
conformational selection and induced fit in ligand binding by the LAO
performed and analyzed the single-molecule experiments. A.J and H.-S.L prepared and
protein. PLoS Comput. Biol. 7, e1002054 (2011).
characterized the Cy7-maltose. E.K., S.L., S.H. and H.-S.K. interpreted the data and wrote
18. Csermely, P., Palotai, R. & Nussinov, R. Induced fit, conformational selection
the manuscript.
and independent dynamic segments: an extended view of binding events.
Trends Biochem. Sci. 35, 539–546 (2010).
19. Antikainen, N.M., Smiley, R.D., Benkovic, S.J. & Hammes, G.G. Conformation Competing financial interests
coupled enzyme catalysis: single-molecule and transient kinetics investigation The authors declare no competing financial interests.
of dihydrofolate reductase. Biochemistry 44, 16835–16843 (2005).
20. Hanson, J.A. et al. Illuminating the mechanistic roles of enzyme
conformational dynamics. Proc. Natl. Acad. Sci. USA 104, 18055–18060 (2007). Additional information
21. Sharff, A.J., Rodseth, L.E., Spurlino, J.C. & Quiocho, F.A. Crystallographic Supplementary information and chemical compound information is available in the
evidence of a large ligand-induced hinge-twist motion between the two online version of the paper. Reprints and permissions information is available online at
domains of the maltodextrin binding protein involved in active transport and http://www.nature.com/reprints/index.html. Correspondence and requests for materials
chemotaxis. Biochemistry 31, 10657–10663 (1992). should be addressed to S.H. or H.-S.K.

318 nature chemical biology | vol 9 | MAY 2013 | www.nature.com/naturechemicalbiology


ONLINE METHODS Trolox were used42. Single-molecule fluorescence images were taken in imaging
Reagents. Sugars (maltose, maltotriose and sucrose) used as ligands were buffer (10 mM Tris-HCl, pH 8.0, 50 nM PCD, 2.5 mM PCA, 1% (v/v) Trolox)
purchased from Sigma. Cy3-maleimide, Cy5-maleimide and Cy7-mono using a home-built prism-type total internal reflection fluorescence micro-
NHS ester fluorophores were purchased from GE Healthcare. Cy7-labeled scope. Three-color smFRET experiments were performed using Cy7-labeled
maltose (10) was obtained through conjugation between the poly(ethylene maltose. Cy3 was excited by a 532-nm laser (Compass215M; Coherent) for
glycol)-amine linked maltose (9, Medigen, Korea) and Cy7-mono NHS ester two-color FRET, and the alternative excitation of Cy3 and Cy5 was performed
(GE Healthcare). Details on their synthesis and preparation are described in by a 532-nm and 633-nm laser (LGK 7626S; Lasos) using a mechanical shut-
Supplementary Note 1. ter for three-color smFRET27. The fluorescence signals of the three dyes were
collected using a water immersion objective lens (UPlanSApo 60×; Olympus)
Protein expression and purification. MBP from pMAL-c2x (New England and filtered through a 540-nm long-pass filter (LP03-532RU-25; Semrock) and
Biolabs) was linked with a His6 tag and the biotin acceptor peptide for a 633-nm notch filter (NF03-633E-25; Semrock) to reject scattered excitation
biotinylation at the N terminus. The resulting gene was cloned into the pET21a laser lines. The emissions were then separated by dichroic mirrors (635dcxr;
vector (Novagen). Single and double mutants of MBP were constructed by Chroma Technology) and detected by an electron-multiplying charged-
introducing one (I329W) and two (A96W I329W) mutations, respectively, coupled-device (Ixon DV897; Andor).
using site-directed mutagenesis. For dye labeling, two cysteine residues were
introduced in the N-terminal (K34C) and the C-terminal (R354C) domains of Analysis of smFRET data. To collect only the dual-labeled proteins from
the wild type and mutants. MBP variants were expressed in E. coli BL21(DE3) heterogeneous populations, a few hundred single-molecule spots from the
cells harboring pProTrc-BirA, which contains the biotin ligase gene from images of an acceptor detection channel were picked following 532-nm
pBIRAcm (Avidity). Proteins were purified by immobilized metal affinity laser excitation. Time traces of Cy3 and Cy5 signals from a single molecule
chromatography using Ni-NTA superflow matrix (Qiagen). The purified were extracted from a recorded movie file using IDL software (ITT Visual
MBP was further desalted with PBS (pH 7.2) using a PD-10 desalting column Information Solutions) and analyzed using custom software written in
(GE Healthcare) for dye labeling. Protein concentration was measured by the MATLAB (MathWorks). The donor (ID) and acceptor (IA) fluorescence inten-
Bradford method. HABA/Avidin reagent (Sigma) was used to quantify the sities from each single molecule were measured, and the FRET efficiencies
amount of biotinylated protein. (EFRET) were calculated using the intensity-based equation; IA / (IA + ID)42. After
© 2013 Nature America, Inc. All rights reserved.

background subtraction and cross-talk correction, we selected real molecule


Dye labeling. MBP variants containing two cysteine residues were labeled traces showing anticorrelations between the donor and acceptor or single-step
with Cy3- and Cy5-maleimide dyes (GE Healthcare). After reducing the sur- photobleaching. To obtain the kinetic times from the time traces, FRET effi-
face exposed–cysteine residues with 100-fold excess of Tris(2-carboxyethyl) ciency levels were identified as orange lines within all figures by the standard
phosphine (TCEP) hydrochloride for 10 min, proteins were incubated with a threshold method. The dwell times were obtained in the absence and presence
ten-fold molar excess of the two dyes in anhydrous dimethylformamide (DMF) of varying ligand concentrations by fitting dwell-time histograms of each state
in equimolar ratio for 2 h at room temperature. Unreacted dyes were removed to a single-exponential decay curve42, and the transition rates were determined
from the protein solution by desalting with a PD-10 column (GE Healthcare) by inverting the dwell times.
using TNG buffer (10 mM Tris, pH 8.0, 50 mM NaCl, 5% glycerol). The
labeling efficiency was determined by measuring the protein concentration Determination of dissociation constant. The reaction observed through the
(absorbance at 280 nm) and the dye concentrations (absorbance at 552 nm and smFRET assay is depicted as follows.
at 650 nm for Cy3 and Cy5, respectively). k
f

O C (1)
kb
CD analysis. CD measurements were conducted before and after dye labeling
for wild-type MBP and its variants using a JASCO J-815 (JASCO) with a quartz where kf is the transition rate from an open form (O) to a closed form (C),
cell of path length 0.01 cm at 25.0 °C. Protein solutions were prepared in and kb is the transition rate from a closed form to an open form. In the case of
sample buffer (10 mM Tris-Cl, 50 mM NaCl, 5% glycerol, pH 7.5), and the the single (I329W) and double (A96W I329W) mutants, kf and kb at varying
protein concentration ranged from 0.1 mg/ml to 0.5 mg/ml. CD spectra were ligand (L) concentrations are given by
taken between 200 nm and 260 nm with 1-nm increments, and the background
spectrum of the buffer solution was subtracted.
npg

kf = kf0 + a ⋅[L]
Assay of maltose-binding activity. Maltose-binding activities of MBP variants
were tested by the amylose-binding assay using the MBP Excellose Spin Kit kb = kb0 (2)
(TaKaRa Bio). Purified proteins were added on to a MBP Excellose spin column,
incubated at room temperature for 2 min and centrifuged at 700g for 2 min. Our empirical result for the closing rate is consistent with a picture that a
Following washing and elution using binding buffer (50 mM Tris, pH 7.0, ligand forms an encounter complex with the open form and subsequently
100 mM NaCl) and elution buffer (binding buffer containing 10 mM maltose), induces a conformational transition to the closed form44:
all filtrates were subjected to SDS-PAGE (12%) analysis.
r [ L] r
b

O c→C ⋅ L
 O ⋅ L 
 (3)
ITC. All calorimetric titrations were performed using the iTC-200 ru
Microcalorimeter (Microcal) at 7.0 °C. The protein concentration in the calo-
where rb, ru and rc are binding, unbinding and closing rate constants, respec-
rimeter cell ranged from 0.02 mM to 0.2 mM, and the ligand concentration in
tively. In this scheme,
the injector varied from 0.2 mM to 2 mM such that the molar ratio of ligand to
macromolecule, following the last injection, was approximately 2. Protein and rbrc
a=
ligand solutions were prepared in sample buffer (50 mM NaPO4, 100 mM NaCl, ru
pH 7.0). Data were analyzed using the software provided by the supplier.
At equilibrium, therefore,
smFRET experiments. Cleaned quartz slides and coverslips were coated with
polyethylene glycol and biotinylated polyethylene glycol (Laysan Bio) to pre- [C] kfo + a ⋅[L]
(kf0 + a ⋅[L]) ⋅[O] = kbo ⋅[C] → = (4)
vent nonspecific adsorption of proteins to the surface42. A sample chamber [O] kbo
was created between a cleaned quartz slide and a coverslip using double-sided
adhesive tape. Proteins were immobilized on the surface by the successive From equation (4), the probability of a closed state (Pclosed) is expressed as
addition of streptavidin (Invitrogen) and 0.5–1.5 nM biotinylated protein in follows:
TN buffer (10 mM Tris-HCl, pH 8.0, 50 mM NaCl). To reduce photobleach-
ing and blinking of dyes, a protocatechuic acid (PCA)/protocatechuate-3,4- [C] k o + a ⋅[L]
Pclosed = = o f o (5)
dioxygenase (PCD) oxygen-scavenging system43 and the blinking suppressant [O]+[C] kb + kf + a ⋅[L]

doi:10.1038/nchembio.1213 nature CHEMICAL BIOLOGY


which can be modified as follows: The chemical potential of an ideal solution is generally written as
o o [L] (6) 1 [L] 1
Pclosed = Pclosed + (1 − Pclosed )⋅ m = mo + ln = mo + ln[L]
K d + [L] b [L]o b (10)

where μo is the chemical potential of the reference state ([L]o = 1 M).


where
By substituting equation (10) into equation (8) and (9), we obtain
o kfo
Pclosed =
(kbo + kfo )
− b  e b − mo − 1 ln[ L]+ e c 
b
e − be c + e   e − be c + [L] ⋅ e − b ( Δe +e c )
Pclosed = =
is the probability of a partially closed state in the absence of a ligand, and − b  e b − mo − 1 ln[ L]+ e c 
b
1 + e − be c + [L] ⋅ e − b ( Δe +e c )
1 + e − be c + e  

kbo + kfo
Kd = e b Δe + [L] e b Δe + [L]
(a ) = =
e b ( Δe + e c )+ b Δe
e + [L] e b Δe
⋅ (1 + e be c ) + [L] (11)
is the dissociation constant of a ligand. The dissociation constants
for the single and double mutants were obtained by fitting the closed
state probabilities at varying ligand concentrations to equation (6). − b  e b − mo − 1 ln[ L]+ e c 
b
e   [L] ⋅ e − b ( Δe +e c )
In the case of wild-type MBP, a ligand-free partially closed state was Pbound = = − be c
not directly detectable through smFRET, and Pclosed
 was set to 0 for − b  e b − mo − 1 ln[ L]+ e c 
b
1+ e + [L] ⋅ e − b ( Δe +e c )
1 + e − be c + e  
the fitting.
[L] [L]
Calculation of the binding and conformational energies. On the basis of = =
e b ( Δe +e c )+ e b Δe + [L] e b Δe ⋅ (1 + e be c ) + [L] (12)
the concept that dissociation constant (Kd) reflects the binding energy and
the conformational change energy, we calculated respective energy values
© 2013 Nature America, Inc. All rights reserved.

where we introduced the notation for the binding free energy (the energy
from the dissociation constant (Kd) and the probability of an intrinsic
 difference upon taking a ligand from solution and placing it on the protein),
closed state, Pclosed . In statistical mechanics, the relative probability of a state
Δε=εb−μo. Using equation (7), Pclosed can be rewritten as follows.
is determined by a Boltzmann factor of the state, e−βε, where β is 1/(kBT),
and ε is the free energy of the state45. In the absence of a ligand, there exist two
o e b Δe + [L] 1
distinct states: an open state and a partially closed state. The probability of a Pclosed − Pclosed = −
partially closed state ( Pclosed
 ) is given by e b Δe
⋅ (1 + e be c ) + [L] 1 + e be c
(13)
e be c [L]
− be c = be c
⋅ b Δe be c
o e 1 (7) 1+ e e ⋅ (1 + e ) + [L]
Pclosed = =
1 + e − be c 1 + e be c
o o [L] (14)
where εc is the free energy of the partially closed state relative to the open → Pclosed = Pclosed + (1 − Pclosed )⋅
e b Δe ⋅ (1 + e be c ) + [L]
state. Therefore, the conformational energy (εc, the energy required for the

transition form an open state to a partially closed state) is obtained from Pclosed . By comparing equation (14) with equation (6), we obtain the following

Pclosed was directly obtained from smFRET experiments for the single and equation:
double mutants, whereas a reported value (5%) was used for wild-type
MBP. In contrast, there exist three states in the presence of a ligand: (i) a K d = e b Δe ⋅ (1 + e be c ) (15)
ligand-free open state (ε=0), (ii) a ligand-free partially closed state (ε=εc) and
(iii) a ligand-bound closed state (ε=εb−μ+εc, where εb and μ are the ligand- which allows us to determine the binding free energy (Δε) from the dissocia-
receptor interaction energy and the chemical potential of the ligand, respectively; tion constant (Kd). For wild-type MBP, the same relation between Kd and Δε
(equation (15)) is obtained from equation (12).
npg

we assumed that the conformational energy difference between a ligand-free


partially closed state and a ligand-bound closed state was negligible). Therefore,
the probabilities in either a partially closed state or a closed state (Pclosed) and 42. Roy, R., Hohng, S. & Ha, T. A practical guide to single-molecule FRET.
the probability in a ligand-bound closed state (Pbound ) are given by Nat. Methods 5, 507–516 (2008).
43. Aitken, C.E., Marshall, R.A. & Puglisi, J.D. An oxygen scavenging system for
e − be c + e − b (e b − m +e c ) improvement of dye stability in single-molecule fluorescence experiments.
Pclosed = (8) Biophys. J. 94, 1826–1835 (2008).
1 + e − be c + e − b (e b − m +e c )
44. Weikl, T.R. & von Deuster, C. Selected-fit versus induced-fit protein binding:
kinetic differences and mutational analysis. Proteins 75, 104–110 (2009).
e − b (e b − m + e c ) 45. Phillips, R., Kondev, J. & Theriot, J. Physical Biology of the Cell.
Pbound = (9)
1 + e − be c + e − b (e b − m +e c ) Ch. 6 (Garland Science, New York, 2009).

nature chemical biology doi:10.1038/nchembio.1213

You might also like