Is a Malleable Protein Necessarily Highly Dynamic? The Hydrophobic Core
of the Nuclear Coactivator Binding Domain Is Well Ordered
Magnus Kjaergaard, Flemming M. Poulsen, and Kaare Teilum*
Department of Biology, University of Copenhagen, Copenhagen, Denmark
complex with different ligands. To understand the mechanism of the structural adaptability in the nuclear coactivator binding
domain (NCBD), we have compared the dynamics of the hydrophobic core of NCBD in the ligand-free state and in a well-folded
complex with the ligand activator for thyroid hormone and retinoid receptors using multiple NMR methods including methyl
chemical shifts, coupling constants, and methyl order parameters. From all NMR measures, the aliphatic side chains in the
hydrophobic core are slightly more dynamic in the free protein than in the complex, but have mobility comparable to the hydro-
phobic cores of average folded proteins. Urea titration monitored by NMR reveals that all parts of the protein, including the side-
chain packing in the hydrophobic core, denatures in a single cooperative process. The molten globule characteristics of NCBD
are thus restricted toa slowly fluctuating tertiary structure. Consequently, the conformationalplasticity of theprotein ismost likely
related to its low overall stability rather than an intrinsically flexible protein structure. The well-defined structure supports a model
of molecular recognition dominated by conformational selection, whereas only minor structural adjustments are necessary after
The nuclear coactivator binding domain of CREB binding protein folds into remarkably different structures in
Flexibility plays an important role in many protein-protein
interactions (1). An especially extreme case of flexibility
is the intrinsically disordered proteins (IDPs) that lack
a folded native structure (2–4). IDPs frequently interact
with many other proteins, where they often fold upon
binding to their partner proteins (5). The flexibility of the
native state of these proteins is believed to be key to their
binding promiscuity, as they can fold into different struc-
tures suited tobind structurally different ligands. An alterna-
tive mechanism to achieve binding to structurally different
proteins is the fold-switching mechanism. Fold-switching
proteins have at least two well-defined conformations with
comparable stability that are both significantly populated
in solution (6–8). Consequently, different ligands can bind
to either of the folded forms in a conformational selec-
tion-like reaction. Each of these conformations may be
well ordered and not extraordinary flexible as long as they
have a low thermodynamic stability (8).
The nuclear coactivator binding domain (NCBD) of
CREB binding protein (CBP) is a particularly interesting
example of plasticity in ligand recognition. NCBD forms
well-defined complexes with a number of ligands, where
it folds into remarkably different tertiary structures (9–12),
as illustrated in Fig. 1. There has been much interest in
understanding the basis of this structural plasticity (13–16)
and NCBD has emerged as a model system for studying
the mechanism of coupled folding and binding (14,17).
Three hypotheses have emerged so far that are not neces-
sarily mutually exclusive. The adaptability has either been
explained by the free protein being a molten globule
(9,13,18), having a low overall structural stability (14), or
being a downhill folder (15). NCBD has most of the charac-
teristics typical for a molten globule: 8-anilinonaphthalene-
1-sulphonic acid binding, broad urea denaturation profile,
the absence of an excess heat capacity peak in DSC, and
exchange broadening of the NMR spectra (18). The molten
globule state is frequently observed as a partially folded
intermediate of proteins under weakly denaturing conditions
(19,20). More recently, a number of proteins have been
discovered that behave as molten globules under native-
like conditions, which makes them the most folded
members of the family of IDPs (2). In the ligand-free state,
secondary chemical shifts show that the three helices of
NCBD are almost fully formed (13,14) and therefore the
protein have a high secondary structure content. All these
observations are characteristic of a molten globule. The
dynamics of the ligand-free state were examined using
15N relaxation techniques, which suggest that the backbone
experiences a uniform degree of dynamics that is compa-
rable to that of the most folded regions of the apomyoglobin
molten globule (13,21), although the free protein seems to
experience exchange broadening. Despite the dynamic
nature of NCBD, the structural ensemble is dominated by
a tertiary fold that resembles the conformations in complex
with the ligands activator for thyroid hormone and retinoid
receptors (ACTR), steroid receptor coactivator-1, and p53
(14). The fold appears to have well-defined tertiary
Submitted November 21, 2011, and accepted for publication February 6,
This article is dedicated to Flemming M. Poulsen, who passed away on
November 9, 2011.
Magnus Kjaergaard’s present address is Department of Chemistry, Univer-
sity of Cambridge, Cambridge, UK.
Editor: Josh Wand.
? 2012 by the Biophysical Society
Biophysical Journal Volume 102 April 2012 1627–16351627
interactions and to fold in a cooperativemanner even though
it is only marginally stable (14), which led to the suggestion
that the low stability might be important for the binding
plasticity. The third explanation of NCBDs plasticity is
that it is a downhill folder. This conclusion was based on
its broad denaturation profile and coarse-grained molecular
simulations (15). According to these simulations, the bar-
rierless nature of NCBD extends to the native ensemble of
NCBD, where it samples highly diverse structures without
passing significant energy barriers (15). The absence of
energy barriers allows NCBD to sample a continuum of
conformations that the protein’s diverse ligands can bind to.
To clarify what role the intrinsic flexibility of NCBD
plays for its ability to adopt different conformations in
complex with ligands, we have characterized the dynamics
of the methyl groups in NCBD by NMR relaxation, and
the unfolding process of NCBD by following chemical
shifts in a urea titration. Collectively, the results show that
NCBD has a well-ordered core that is slightly more dynamic
in the free state than in complex with a ligand, and that all of
NCBD cooperatively loses its structure during unfolding.
Despite having many of the characteristics of a molten
globule, NCBD thus behaves much as a fully compact
protein with a hydrophobic core that fluctuates slowly.
This suggests that the conformation of NCBD observed
both in the free state and in complex with ligand indeed
represent a minimum in conformational space surrounded
by energy barriers.
MATERIALS AND METHODS
The plasmid for coexpression of the NCBD (UniProt: P45481, residues
2059–2117) of murine CBP and the activation domain of ACTR (UniProt:
Q9Y6Q9, residues 1018–1088) was a gift from Peter E. Wright (The
Scripps Research Institute, San Diego, CA) and was described previously
(9). Isotope labeling was achieved by growing the cells in M9 medium
with 4 g/L13C glucose and 1.5 g/L15N ammonium sulfate as the only
carbon and nitrogen sources. Four different labeling schemes were used:
Uniform15N labeling, uniform13C and15N labeling, 10%13C and uniform
15N labeling [10%13C], and uniform13C and15N labeling combined with
50% random2H labeling [50%2H]. Random2H labeling was achieved by
growing the bacteria in 50%2H2O. Protein purification was performed as
described previously for ACTR (22) except that a FastFlow SP column
was used for the ion exchange step.
Lyophilized NCBD was dissolved into 20 mM phosphate in 10% D2O and
adjusted to pH 6.5. Relaxation experiments were recorded at 31?C on Var-
ian VnmrS 500 MHz and Varian Inova 750 MHz spectrometers equipped
with conventional triple resonance probes. Chemical shift and coupling
constant measurements were recorded on a Varian Inova 800 MHz spec-
trometer equipped with a cold probe. NMR spectra were processed using
nmrPipe (23) and analyzed in CCPNMR analysis (24). Constant time
13C-HSQC spectra were recorded on samples containing 1 mM [10%
13C] NCBD, and 1 mM [10%13C] NCBD complexed with 1.5 mM unla-
beled ACTR. LRCC (25) spectra were recorded on a sample containing
0.7 mM uniformly labeled NCBD and 0.7 mM uniformly labeled NCBD
in complex with 1.2 mM unlabeled ACTR.
Deuterium relaxation experiments were recorded on a sample containing
in complex with 1.8 mM ACTR. Relaxation times of 0.2, 1.3, 2.8, 4.4, 6.2,
of 0.05, 2.8, 4, 10, 16*, 22, 35, and 40 ms were used for measurement of the
Dz, 3Dz2-2, Dþ2, and DþDzþ DzDþcoherences as described (26,27).
* denotes points determined in triplicate. Experiments were recorded with
64–76 increments in the indirect dimension, a recycle delay of 2 s and 16
transients collected for each increment. For both free NCBD and NCBD in
complex with ACTR the five deuterium relaxation experiments were
ing 1.3 mM15N-NCBD and a sample containing 1.3 mM15N-NCBD in
complex with 1.8 mM ACTR at 750 MHz using previous published pulse
sequences (28). All relaxation data were analyzed in CCPNMR analysis.
on the optimized parameters were determined by bootstrap analyses.
From the five relaxation rates measured at 750 MHz three spectral densi-
ties, J(0), J(uD), J(2uD), were extracted by optimizing their values in the
expressions describing the relations between the relaxation rates and the
spectral densities using a quadrupolar coupling constant of 167 kHz (26).
For free NCBD J(uD) and J(2uD) were also determined at 500 MHz. For
NCBD in complex with ACTR the correlation time, tr, for the overall rota-
tional diffusion was estimated from the15N R1and R2relaxation rates by
fitting their trimmed mean to the following expression as described by
Fushman et al. (29):
For NCBD in complex with ACTR three spectral densities for each methyl
group were fit to the Lipari-Szabo two-parameter model (LS2) for the spec-
1 þ ðutrÞ2þð1 ? S2Þt
where 1/t ¼ 1/trþ 1/tf. tris the correlation time for the overall rotational
diffusion of the complex and tfis the correlation time for the fast (ps)
1 þ ðutÞ2;
ture of the ligand-free NCBD (2KKJ) (14), the complex with ACTR
(1KBH) (9), and the complex with IRF-3 (1ZOQ) (10). The NCBD struc-
tures are shown in black, aligned along helix 1, and display remarkably
different conformations in the two complexes.
Representative structures of NCBD. The dominating struc-
Biophysical Journal 102(7) 1627–1635
1628 Kjaergaard et al.
internal motion of the methyl group. trwas kept constant in the fit of S and
tf. This model fit all data well and the current data set does not warrant the
use of a more complex model. The order parameter describing the fluctua-
tion of the average methyl axis is defined as S2axis¼ 9 S2.
As the backbone resonances of free NCBD are significantly broadened
by exchange (13) a reliable correlation time for the rotational diffusion
cannot be estimated from15N relaxation rates. Therefore, the five spectral
densities were fit to a model with an individual efficient correlation time for
each methyl group, i.e., a three-parameter model (LS3):
?2þð1 ? S2Þt
1 þ ðutÞ2;
where 1/t ¼ 1/tCeffþ 1/tf. tCeffis the local efficient correlation time
including both the overall rotational diffusion and internal motions on the
same (ns) timescale. tfis the correlation time for the fast (ps) internal
motion of the methyl group. Again the axial order parameter is defined
as S2axis¼ 9 S2.
Error bars on plots of spectral densities and order parameters indicate the
95% confidence intervals determined from the distribution of the parame-
ters in 100 Monte Carlo simulations. The average order parameters for
methyl groups derived from Carbonell and del Sol (30) are Ab¼ 0.81,
Ig2¼ 0.78, Id¼ 0.60, Ld¼ 0.58, Mε¼ 0.35, Tg¼ 0.74, and Vg¼ 0.70.
The denaturation of NCBD was followed by NMR using a urea titration at
31?C. Starting from a uniformly labeled 1 mM sample of NCBD, the urea
concentration was increased in 0.5 M steps by addition from a stock con-
taining 10 M urea but similar in other respects. At each denaturant concen-
tration the following experiments were acquired:15N HSQC, constant time
of urea concentration was fitted to a two-state transition with sloping base-
lines using Igor Pro (WaveMetrics, Lake Oswego, OR). The chemical shifts
were fitted to a two-state equilibrium by considering either each curve indi-
vidually, all backbone chemical shifts from each residue together or three
clusters corresponding to the three helices in free NCBD. Methyl chemical
shifts are left out of the global fits as the posttransition baseline typically is
poorly defined due to spectral overlap. In the global fits the m-values and
denaturation midpoints are treated as global parameters and errors on the
parameters are determined from the asymptotic standard errors.
Stereospecific assignment of methyl groups
Meaningful comparison of the methyl groups in the ligand-
free and -bound states of the NCBD requires stereospecific
assignments, which can be determined from the sign of the
methyl crosspeaks in a13C CT-HSQC spectrum recorded on
a sample containing 10%13C labeling as described previ-
complex with the activation domain of ACTR (Fig. 2). In
cases where the methyl peaks are overlapped in the13C
CT-HSQC spectrum, the chemical shifts were determined
using a C(CO)NH spectrum recorded on a uniformly
13C-labeled sample (32), which correlates the13C chemical
shifts to the well-dispersed amide protons. In these cases,
the stereospecific assignments are based on the geminal
neighbor in the [10%13C]-NCBD CT-HSQC. Using this
method, stereospecific assignments were obtained for all
geminal methyl groups of free NCBD and in complex
13C CT-HSQC spectra were recorded on
13C]-NCBD and [10%
c-Angle dynamics from chemical shifts
Aliphatic side chains primarily occupy the three possible
staggered conformations for each c-angle and may populate
either a single rotamer or a distribution of staggered confor-
mations.13C chemical shifts report on the rotamer distribu-
tion of the protein side chain due to the g-gauche effect
(33,34). The g-gauche effect decreases the chemical shifts
time HSQC spectra of 10%-13C NCBD in the
free form (left) and in complex with ACTR (right).
Stereospecific assignments are obtained from the
signsof the peaks that are shown as black (positive)
and gray (negative). The Ile-d1 peaks (top panel)
are recorded on a uniformly labeled sample.
Methyl region of the
Biophysical Journal 102(7) 1627–1635
The Core of NCBD Is Well Ordered1629
of nuclei that are gauche to another heavy atom by ~5 ppm.
This effect has been particularly useful for determining the
side-chain conformations of leucine (22,35,36), isoleucine
(37), and valine residues (38) from their chemical shifts.
Leucine and isoleucine side chains mainly occupy two
c2-angles, which simplify the interpretation of the
methyl chemical shifts. For leucine, the populations of the
staggered conformations of the c2-angle depend linearly
on the chemical shift difference between the two methyl
groups, d(Cd1) - d(Cd2) (35), and for isoleucine it depends
directly on d(Cd1) (37). Fig. 3 A shows the populations of
the trans rotamer of the leucine and isoleucine residues of
unbound NCBD and in complex with NCBD. Valine resi-
dues populate all three c1-angles and are thus more difficult
to interpret quantitatively. Instead, we report the chemical
shift difference between the two geminal methyl groups of
each valine residue, which may serve as a qualitative probe
of the side-chain dynamics (Fig. 3 B). In both the free state
and in the complex, NCBD does not have any leucine and
isoleucine side chains with unique rotamers. Disordered
proteins that do not form hydrophobic cores have side-chain
distributions close to their random coil values (22), which is
signified by the horizontal lines in Fig. 3, A and B. Depend-
ing on the dynamics, the side-chain rotamer distributions in
a mobile hydrophobic core would approach the values of the
random coil. In both states, NCBD has methyl rotamer
distributions distinct from random coil. Comparison of the
free and the ACTR-bound states of NCBD reveals that 9
out of 12 methyl-bearing side chains are closer to the
random coil values in the free form than in the complex.
This suggests that the side chains in the hydrophobic core
undergo slightly more rotameric averaging in the unbound
c-Angle dynamics from coupling constants
The rotamer distribution of the side chains can also be
measured using coupling constants. For aliphatic side
c2-angle and can thus be directly compared to thevalues ob-
tained from methyl chemical shifts of leucine and isoleucine
residues.3JCdCavalues of ~4 Hz suggest that the Cdis trans
to the Caand values of <1 Hz suggest that the Cdis gauche
relative to the Ca. The coupling constants were determined
using the long-range C-C experiment as described previ-
ously (25). The overlap particularly in the leucine methyl
region makes this approach challenging. Still, the3JCdCa
could be determined for 16 methyl groups in the free state
and for 15 in the complex with ACTR. Because the3JCdCa
coupling constant and the methyl chemical shifts report on
the same dihedral angle, their internal consistency can be
checked by correlating the chemical shift difference to the
coupling constants as described previously (35). The corre-
lation in Fig. 3 C shows that the coupling constants gener-
ally agree with the conclusion from the chemical shifts,
except for residue L2071. Both methyl chemical shifts of
this residue are low suggesting that they are both gauche
to the Caand thus populate the rare gauche-conformation.
As the coupling constants report on the same information
as the chemical shifts and the latter can be measured more
precisely, we did not interpret the coupling constants beyond
noting that they support the conclusions from the methyl
3JCdCais particularly useful as it probes the
The ps-ns timescale motions of methyl groups can be quan-
titatively probed using the deuterium relaxation rates for the
CH2D isotopomer of methyl groups (26,39). The relaxation
rates of five coherences were recorded on both NCBD in the
unbound state and in complex with ACTR. Relaxation rates
were determined by fitting of a single exponential decay to
the peak intensity as a function of relaxation time. 21 and 24
methyl groups yielded signals of sufficient quality to allow
further analysis of the free and ACTR-bound states, respec-
tively (Fig. 4 and Fig. S1 and Tables S1 and S2 in the Sup-
porting Material). To determine the overall correlation time
of NCBD, R1and R2relaxation rates were determined for
15N nuclei in the unbound state and in complex with
chains in free NCBD and in the complex with ACTR. (A) Populations of
the trans conformer of the c2-angle for isoleucine and leucine residues
derived from13C methyl chemical shifts. Horizontal lines correspond to
the random coil values of Leu (thin) and Ile (bold) adopted from (22).
(B)13C chemical shift differences between the geminal methyl groups of
valine side chains. (C) Correlation between rotamer distributions deter-
mined from chemical shifts and3JCaCbcouplings. The * corresponds to
residue L2071 in the free state that predominantly populated the rare
gauche-rotamer and thus is an outlier.
Comparison of the rotamer distribution of aliphatic side
Biophysical Journal 102(7) 1627–1635
1630Kjaergaard et al.
ACTR (Fig. S2). In the complex, the15N relaxation rates are
similar for most residues and accordingly a correlation time
of 9.3 ns is estimated based on the ratio of the truncated
averaged values of the R1and R2rates as described previ-
ously (29). This value is consistent with a complex with
a molecular mass of 14.5 kDa. To interpret the relaxation
rates in terms of order parameters for the complex, the spec-
tral densities J(0), J(uD), and J(2uD) were extracted from
the measured relaxation rates and fitted to the two-param-
eter spectral density function (LS2) proposed by Lipari
and Szabo (40,41) by constraining the rotational correlation
time to the value obtained from15N relaxation data (Fig. 4,
B–C, and Table S3). All data sets are well fit by the LS2
model and the fit fall within the error bars of the spectral
densities. With the current data set, fits to more complex
models for the spectral densities would lead to overfitting
of the data. If some local ns motions of the methyl groups
are present, it would mean that trwould overestimate the
correlation time for the ns motions, which would lead to
underestimation of S2axis. In such a case the reported order
parameters are lower limits (27). However, as J(uD) z
J(2uD) within error for all methyl groups the correlation
times for the ns motions cannot be much shorter than the
9.3 ns determined for tr. Free NCBD has highly variable
R2rates (Fig. S2) and at low temperatures most signals
are broadened beyond detection suggesting ms-ms dynamics
throughout the sequence, as also reported previously (13).
Consequently, the correlation time of free NCBD cannot
be reliably determined from15N relaxation rates. Therefore,
it was necessary to record deuterium relaxation data at two
static magnetic fields to sample the spectral density function
at more values. Five spectral densities were extracted from
the relaxation rates and fitted to a three-parameter spectral
density function, LS3 (27), where the local ns dynamics
and the overall rotation of the molecule are combined into
an effective correlation time for each methyl group. All
methyl groups of free NCBD fit LS3 with htCeffi ¼ 4.8 5
0.8 ns (Table S4). The average tCeffis in the expected range
for the rotational correlation time for a protein with the size
of NCBD (6.6 kDa). The longest tCefffrom the fits (up to
6.3 ns) are longer than expected for the overall tumbling
of NCBD and may thus be overestimated. For these residues
the fitted S2axisare thus lower limits as discussed previously.
The order parameters for unbound NCBD and NCBD in
complex with ACTR are compared in Fig. 4 C. The methyl
groups near the C-terminus are more flexible in the free
form than in the complex, which is in agreement with the
region forms a part of helix 3 in the complex (9), but is
largely disordered in the unbound state (14). In the core
region containing the three helices, the dynamics is reduced
in the complex as often observed upon ligand binding.
Unlike the backbone, methyl groups have a high variation
in the intrinsic dynamics between side chains due to the
large difference in chain lengths, and the variation in order
parameters shown Fig. 4 C is both a result of chain length
and local side-chain packing. To isolate the contribution
from side-chain packing, we subtracted the average order
parameter for each type of methyl group. The average
values were obtained from a database consisting of methyl
relaxation studies of 18 well-folded proteins used for predic-
tion of methyl order parameters (30). Additionally, this
approach allows us to compare the dynamics of the hydro-
phobic core of the free state of NCBD with that of folded
15N backbone experiments (13). This
NCBD probed by deuterium relaxation. (A) Repre-
sentative relaxation decays from the V2087 g2
methyl group of the five deuterium coherences re-
corded on free NCBD at 750 MHz. Additionally,
three relaxation decays were measured for the
free form at 500 MHz. (B) Spectral densities
were extracted from the relaxation rates and two
parameters (ACTR complex) and three parameters
(free NCBD) model-free spectral density functions
were fitted to extract methyl order parameters. (C)
Comparison of methyl order parameters for free
NCBD and the complex with ACTR. (D) Methyl
order parameters relative to average order parame-
ters for each residuetype determined from a library
of 18 well-folded proteins.
Hydrophobiccore dynamics of
Biophysical Journal 102(7) 1627–1635
The Core of NCBD Is Well Ordered1631
proteins. In the core region of NCBD, the residue-type cor-
rected order parameters of unbound NCBD is scattered
around 0, indicating that in the ps-ns timescale the hydro-
phobic core of the protein is approximately as dynamic as
the core of an average folded protein. Fig. 5 shows the
difference from the average order parameters plotted onto
the structures of NCBD in the complex with ACTR and in
the conformation dominating the free state. Without a bound
ligand, NCBD has a cluster of methyl groups in the inter-
faces between the three helices that have average or above
average rigidity. These residues correspond to the small
hydrophobic core that we have recently proposed to be
reasonably well ordered based on ring-current effects (14).
In contrast, methyl groups located on the outer surface of
the protein are more dynamic than average. Exceptions
from this pattern are the two methyl groups of A2099 and
A2100 that are exposed and rigid. Alanine methyl groups,
however, are closely related to the dynamics of the back-
bone (42) and thus the rigidity of these nuclei do not repre-
sent side-chain packing. In the complex with ACTR, the
interface between helix 1 and 2 is more rigid than in the
free state. The hinge region and outer surface of helix 1,
in contrast, seemingly becomes more dynamic in the
Order parameters for methyl groups are not uniformly
distributed, but tend to cluster in three classes (43). The first
class is centered around S2axis~0.35 and results from rota-
meric interconversion; the second class is centered around
S2axis ~0.6 and results from motions within a rotameric
well and motions of connecting bonds; the third class is
centered around S2axis ~0.8 and results from restricted
motions within a rotameric well (43,44). The distribution
of order parameters in NCBD is shown in Fig. S3. The
distributions of S2axisfor both free and complexed NCBD
are multimodal and roughly fall into three classes. This
analysis supports the notion that free NCBD has a well-
ordered core with several nonalanine methyl groups in the
most rigid class. Upon ligand binding a shift in the distribu-
tion from the group centered around 0.6 to the group around
0.8 is observed, suggesting that the motions of the methyls
on average become more restricted by restricting the
motions of bonds adjacent to the methyls that are them-
selves restricted to motions within a rotameric well even
in the free state. This is consistent with the observation
that the methyl chemical shifts generally move away from
their random coil positions upon ligand binding.
Denaturation of NCBD using urea has previously demon-
strated that the protein has a folded core that unfolds coop-
eratively (14). The denaturation was monitored using far
ultraviolet circular dichroism spectroscopy that probes the
average loss of helicity with increasingly denaturing condi-
tions, but tells little about the hydrophobic core and the
behavior of individual residues. To probe the change in
the side-chain packing in the hydrophobic core during un-
folding, we recorded NMR spectra in increasing concentra-
tions of urea. The peaks in all spectra are in fast exchange
and can be followed throughout the titration range from
native conditions to highly denaturing conditions. Six types
of chemical shifts were measured for each urea concentra-
tion, resulting in between 4 and 8 chemical shifts for each
residue. The backbone13C nuclei, Ca, and C0, are sensitive
to the secondary structure; the amide15NHand1HNnuclei
are reporting on both the secondary and tertiary structure
and the methyl13C and1H are sensitive to the hydrophobic
packing. Fig. 6 and Fig. S4 show the chemical shift change
a sigmoidal chemical shift change occurs between 0 and
4 M urea, which is consistent with the data obtained by
circular dichroism (14). The nuclei from each of the three
helices were fit globally to a two-state transition assuming
a shared m-value and denaturation midpoint. Despite their
poor definition by the data the pretransition baselines have
small slopes of the same magnitude as the posttransition
baselines showing that the baselines do not absorbvariations
in chemical shift between each titration curve (Fig. S5) as
proposed previously for BBL (45). The global model fits
the data and results in m-values of (4.4, 4.3, and 4.1 kJ/
molM) and a denaturation midpoint of (2.37, 2.36, and
2.32 M) for helix 1, 2, and 3, respectively. This corresponds
S2axis -S2axis,v plotted on the structure of NCBD in the complex with
ACTR and in the dominating form of the free state. Methyl groups are
shown as spheres and color coded by S2axis-S2axis,vfrom red (?0.3) to
blue (0.3) as indicated by the scale.
Differencesfrom average methylorder parameters,
Biophysical Journal 102(7) 1627–1635
1632Kjaergaard et al.
to a thermodynamic stability of 9.9 5 0.3 kJ/mol, which is
equivalent to ~2% of the protein being in the unfolded state.
The stability is slightly higher than the one found using
circular dichroism spectroscopy, most likely due to the
different solvent conditions used in this study.
If each urea unfolding curve is fitted individually, we
observe a large spread in thermodynamic parameters, e.g.,
the1HNcurves have unfolding midpoints between 1.8 and
2.8 M (Fig. S6). Diversity in the unfolding midpoints have
previously been interpreted as evidence of downhill folding
(46), and in such a case global fitting of the unfolding curves
would be invalid. It has been suggested that NCBD folds
downhill (15), and it is important to clarify whether this
diversity represents differential stability of different parts
of the molecule. The spread in unfolding midpoints,
however, differ between the nuclei. The parameters from
individual fits of Cacurves are thus distributed over
a much narrower range than1HN(Fig. S6). The uncertainties
of the fits are substantial, which can be attributed to the
poorly restrained pretransition baselines due to the low
stability of the domain. The error bars of the data points
thus overlap in the central region of the plot. To further
test whether the dispersion in fitting parameters represents
differential stability or fitting uncertainty, we correlated
the fitting parameters from the multiple nuclei within the
same residue. As is shown for1HNand Cain Fig. 7 and
by the pairwise correlation coefficients for all other nuclei
in Table S1, there is no correlation between these parame-
ters. This shows that the spread in the fitting parameters is
a consequence of fitting uncertainty and does not represent
differential stability. Of importance, the correlation of
multiple nuclei within each residue is a rigorous test for
whether dispersion in the structural stability is true. From
the current data set with unusually many data points per
residue, we thus conclude that the unfolding of NCBD is
well described by a barrier-limited two-state model.
Here, we have demonstrated that the protein domain NCBD
has a hydrophobic core as well ordered as the hydrophobic
core of an average compact folded protein and almost as
rigid as the ligand-bound state of the same protein. Further-
more, we have shown that the hydrophobic core unfolds
cooperatively in a two-state transition. Together with our
previous results that show that the dominant structure of
NCBD in its ligand-free state under our experimental condi-
tions resembles its structure in complex with the ligand
ACTR (14), these results show that the behavior of NCBD
is remarkably similar to a compact folded protein. This
behavior may appear contradictory to the classification of
NCBD as a molten globule. In several other biophysical
experiments NCBD behaves as an archetypical molten
globule (9,13,18). The NMR signals of NCBD are poorly
dispersed and are broadened significantly, indicative of
exchange between at least two conformational states (13).
This conformational exchange contributes significantly to
the transverse relaxation of1H,13C, and15N. In contrast,
the quadrupolar relaxation of deuterium that we measure
here is so fast that it is only minimally sensitive to confor-
mational exchange on the 10?3–10?6timescale (47). Our
data are therefore consistent with previous relaxation data
and suggest that NCBD forms well ordered and slowly inter-
changing structures. The definition of a molten globule has
always been vague. It is thus unclear whether a protein that
experiences slow conformational exchange, but has a well-
defined hydrophobic packing at least in the major state, is
a molten globule or not. Ultimately, this is a semantic
discussion that depends on how inclusive the term molten
globule should be. Such a discussion should not, however,
derail the more important discussion of how the dynamic
behavior of NCBD affects the mechanism of ligand binding.
ally fitted unfolding curves measured on HNand Caof the same residue.
Correlation of unfolding midpoints obtained from individu-
residues from each of the three helices. The chemical shifts were measured
for NCBD from15N-HSQC, 2D HNCOCA, 2D HNCO and constant-time
13C HSQC experiments allowing us to specifically probe helicity (15N,
Ca, and C0), hydrophobic packing (methyl1H and13C), and tertiary struc-
ture (methyl1H, amide1H, and15N).
Urea denaturation of NCBD monitored by NMR for selected
Biophysical Journal 102(7) 1627–1635
The Core of NCBD Is Well Ordered1633
NCBD binds several different ligands and adopts
different conformations in the resulting complexes. Such
behavior is also observed for other IDPs (48) and it is
commonly attributed to the inherent flexibility of the disor-
dered polypeptide chain. How important disorder is for the
structural malleability of NCBD is, however, unclear.
Based on coarse-grained molecular dynamics simulations
and DSC, it was argued that the high flexibility of
NCBD and thus its promiscuity in ligand binding is a
consequence of a downhill folding mechanism for NCBD
(15). It has, however, been demonstrated that downhill
folding cannot be proven by DSC alone (49) calling the
alleged downhill folding of NCBD into question. Accord-
ingly, the well-ordered hydrophobic core of NCBD that
we have presented here is in direct conflict with such
a folding behavior that predicts rapid interconversion in
a wide energy basin. In support of our results from the
relaxation measurements on the methyl groups in NCBD,
the urea induced unfolding of NCBD followed by chemical
shift changes of nuclei in the entire protein is explained by
a single cooperative transition. Our experimental data make
it highly unlikely that NCBD follows a barrierless downhill
Collectively, our data suggest that structural plasticity of
NCBD is not due to an extraordinarily flexible structure as is
usually emphasized for other intrinsically disordered
proteins. In contrast, NCBD has evolved to use a different
mechanism where at least the dominant fold is well defined
but only marginally stable. The marginal thermodynamic
stability is important because it allows a fully disordered
state to be populated. Recently, this observation was sup-
ported by hydrogen exchange measurements showing that
the exchange behavior of free NCBD is dominated by the
fast exchange of a minor conformation in a highly disor-
dered state (50). This highly disordered state may thus
fold into alternative conformations by an induced fit mech-
anism. Alternatively, several proteins have recently been
observed to have two coexisting well-defined conformations
in solution (6–8). Fold-switching proteins invariably have
low free energies of folding (8) because an exceedingly
stable conformation would require alternative folds to
have similar stabilities to be appreciably populated, which
is likely to be evolutionarily unattainable. The ligand-free
state of NCBD may thus contain alternatively folded
conformers, which could initiate the binding reaction with
other ligands, e.g., IRF-3 in a mechanism resembling
conformational selection.Eventhough the dominant confor-
mation is as well ordered as NCBD, a protein may still adapt
to different ligands through both induced fit and conforma-
tional selection type mechanisms.
Six figures and five tables are available at http://www.biophysj.org/
We thank Peter E. Wright (The Scripps Research Institute) for sharing the
coexpression plasmid for NCBD and Mikael Akke for access to the Varian
VnmrS 500 MHz spectrometer at Lund University.
This work was supported by the EliteForsk programme (to M.K.), The John
and Birthe Meyer Foundation,the Carlsberg Foundationgrant No. 2008-01-
0368, the Danish Natural Research Council grant No. 272-08-0500 (to
F.M.P), and by the Lundbeck Foundation (to K.T.).
1. Teilum, K., J. G. Olsen, and B. B. Kragelund. 2009. Functional aspects
of protein flexibility. Cell. Mol. Life Sci. 66:2231–2247.
2. Dyson, H. J., and P. E. Wright. 2005. Intrinsically unstructured proteins
and their functions. Nat. Rev. Mol. Cell Biol. 6:197–208.
3. Wright, P. E., and H. J. Dyson. 1999. Intrinsically unstructured
proteins: re-assessing the protein structure-function paradigm. J. Mol.
4. Uversky, V. N. 2002. Natively unfolded proteins: a point where biology
waits for physics. Protein Sci. 11:739–756.
5. Wright, P. E., and H. J. Dyson. 2009. Linking folding and binding.
Curr. Opin. Struct. Biol. 19:31–38.
6. Gambin, Y., A. Schug, ., A. A. Deniz. 2009. Direct single-molecule
observation of a protein living in two opposed native structures.
Proc. Natl. Acad. Sci. USA. 106:10153–10158.
7. Alexander, P. A., Y. He, ., P. N. Bryan. 2009. A minimal sequence
code for switching protein structure and function. Proc. Natl. Acad.
Sci. USA. 106:21149–21154.
8. Bryan, P. N., and J. Orban. 2010. Proteins that switch folds. Curr. Opin.
Struct. Biol. 20:482–488.
9. Demarest, S. J., M. Martinez-Yamout, ., P. E. Wright. 2002. Mutual
synergistic folding in recruitment of CBP/p300 by p160 nuclear
receptor coactivators. Nature. 415:549–553.
10. Qin, B. Y., C. Liu, ., K. Lin. 2005. Crystal structure of IRF-3 in
complex with CBP. Structure. 13:1269–1277.
11. Waters, L., B. G. Yue, ., D. M. Heery. 2006. Structural diversity in
p160/CREB-binding protein coactivator complexes. J. Biol. Chem.
12. Lee, C. W., M. A. Martinez-Yamout, ., P. E. Wright. 2010. Structure
of the p53 transactivation domain in complex with the nuclear receptor
coactivator binding domain of CREB binding protein. Biochemistry.
13. Ebert, M. O., S. H. Bae, ., P. E. Wright. 2008. NMR relaxation study
of the complex formed between CBP and the activation domain of the
nuclear hormone receptor coactivator ACTR. Biochemistry. 47:1299–
14. Kjaergaard, M., K. Teilum, and F. M. Poulsen. 2010. Conformational
selection in the molten globule state of the nuclear coactivator binding
domain of CBP. Proc. Natl. Acad. Sci. USA. 107:12535–12540.
15. Naganathan, A. N., and M. Orozco. 2011. The native ensemble and
folding of a protein molten-globule: functional consequence of down-
hill folding. J. Am. Chem. Soc. 133:12154–12161.
16. Burger, V. M., A. Ramanathan, ., C. S. Chennubhotla. 2012. Quasi-
anharmonic analysis reveals intermediate States in the nuclear co-acti-
vator receptor binding domain ensemble. Pac. Symp. Biocomput.
17. Ganguly, D., W. Zhang, and J. Chen. 2011. Synergistic folding of two
intrinsically disordered proteins: searching for conformational selec-
tion. Mol. Biosyst. 8:198–209.
18. Demarest, S. J., S. Deechongkit, ., P. E. Wright. 2004. Packing, spec-
ificity, and mutability at the binding interface between the p160 coac-
tivator and CREB-binding protein. Protein Sci. 13:203–210.
19. Dobson, C. M. 1994. Protein folding. Solid evidence for molten glob-
ules. Curr. Biol. 4:636–640.
Biophysical Journal 102(7) 1627–1635
1634 Kjaergaard et al.
20. Ptitsyn, O. B. 1995. Molten globule and protein folding. Adv. Protein Download full-text
21. Eliezer, D., J. Yao, ., P. E. Wright. 1998. Structuraland dynamic char-
for protein folding. Nat. Struct. Biol. 5:148–155.
22. Kjaergaard, M., V. Ie? smantavi? cius, and F. M. Poulsen. 2011. The inter-
play between transient a-helix formation and side chain rotamer distri-
butions in disordered proteins probed by methyl chemical shifts.
Protein Sci. 20:2023–2034.
23. Delaglio, F., S. Grzesiek, ., A. Bax. 1995. NMRPipe: a multidimen-
sional spectral processing system based on UNIX pipes. J. Biomol.
24. Vranken, W. F., W. Boucher, ., E. D. Laue. 2005. The CCPN data
model for NMR spectroscopy: development of a software pipeline.
25. Bax, A., D. Max, and D. Zax. 1992. Measurement of long-range
13C-13C J couplings in a 20-kDaprotein-peptide complex. J. Am.
Chem. Soc. 114:6923–6925.
26. Millet, O., D. R. Muhandiram, ., L. E. Kay. 2002. Deuterium spin
probes of side-chain dynamics in proteins. 1. Measurement of five
relaxation rates per deuteron in(13)C-labeled and fractionally(2)H-en-
riched proteins in solution. J. Am. Chem. Soc. 124:6439–6448.
27. Skrynnikov, N. R., O. Millet, and L. E. Kay. 2002. Deuterium spin
probes of side-chain dynamics in proteins. 2. Spectral density mapping
and identification of nanosecond time-scale side-chain motions. J. Am.
Chem. Soc. 124:6449–6460.
28. Farrow, N. A., R. Muhandiram, ., L. E. Kay. 1994. Backbone
dynamics of a free and phosphopeptide-complexed Src homology 2
domain studied by 15N NMR relaxation. Biochemistry. 33:5984–6003.
29. Fushman, D., R. Weisemann, ., H. Ru ¨terjans. 1994. Backbone
dynamics of ribonuclease T1 and its complex with 20GMP studied by
two-dimensional heteronuclear NMR spectroscopy. J. Biomol. NMR.
30. Carbonell, P., and A. del Sol. 2009. Methyl side-chain dynamics
prediction based on protein structure. Bioinformatics. 25:2552–2558.
31. Tugarinov, V.,and L.E. Kay.2004.Stereospecific NMRassignments of
prochiral methyls, rotameric states and dynamics of valine residues in
malate synthase G. J. Am. Chem. Soc. 126:9827–9836.
32. Grzesiek, S., J. Anglister, and A. Bax. 1993. Correlation of backbone
amide and aliphatic side-chain resonances in C-13/N-15-enriched
proteins by isotropic mixing of C-13 magnetization. J. Magn.
Reson. B. 101:114–119.
33. Tonelli, A. E., and F. C. Schilling. 1981. C-13 Nmr chemical-shifts and
the microstructure of polymers. Acc. Chem. Res. 14:233–238.
34. London, R. E., B. D. Wingad, and G. A. Mueller. 2008. Dependence of
amino acid side chain 13C shifts on dihedral angle: application to
conformational analysis. J. Am. Chem. Soc. 130:11097–11105.
35. Mulder, F. A. A. 2009. Leucine side-chain conformation and dynamics
in proteins from 13C NMR chemical shifts. Chem. Bio. Chem. 10:
36. Hansen, D. F., P. Neudecker, ., L. E. Kay. 2010. Determination of Leu
side-chain conformations in excited protein states by NMR relaxation
dispersion. J. Am. Chem. Soc. 132:42–43.
37. Hansen, D. F., P. Neudecker, and L. E. Kay. 2010. Determination of
isoleucine side-chain conformations in ground and excited states of
proteins from chemical shifts. J. Am. Chem. Soc. 132:7589–7591.
38. Hansen, D. F., and L. E. Kay. 2011. Determining valine side-chain ro-
tamer conformations in proteins from methyl 13C chemical shifts:
application to the 360 kDa half-proteasome. J. Am. Chem. Soc.
39. Muhandiram, D. R., T. Yamazaki, ., L. E. Kay. 1995. Measurement of
H-2 T-1 and T-1p relaxation-times in uniformly C-13-labeled and frac-
tionally H-2-labeled proteins in solution. J. Am. Chem. Soc.
40. Lipari, G., and A. Szabo. 1982. Model-free approach to the interpreta-
tion of nuclear magnetic-resonance relaxation in macromolecules. 1.
Theory and range of validity. J. Am. Chem. Soc. 104:4546–4559.
41. Lipari, G., and A. Szabo. 1982. Model-free approach to the interpreta-
tion of nuclear magnetic-resonance relaxation in macromolecules. 2.
Analysis of experimental results. J. Am. Chem. Soc. 104:4559–4570.
42. Mittermaier, A., L. E. Kay, and J. D. Forman-Kay. 1999. Analysis of
deuterium relaxation-derived methyl axis order parameters and corre-
lation with local structure. J. Biomol. NMR. 13:181–185.
43. Igumenova, T. I., K. K. Frederick, and A. J. Wand. 2006. Characteriza-
tion of the fast dynamics of protein amino acid side chains using NMR
relaxation in solution. Chem. Rev. 106:1672–1699.
44. Best, R. B., J. Clarke, and M. Karplus. 2004. The origin of protein side-
chain order parameter distributions. J. Am. Chem. Soc. 126:7734–7735.
45. Sadqi, M., D. Fushman, and V. Munoz. 2007. Structural biology: anal-
ysis of ‘downhill’ protein folding; analysis of protein-folding coopera-
tivity. Nature. 445:E17–E18.
46. Sadqi, M., D. Fushman, and V. Mun ˜oz. 2006. Atom-by-atom analysis
of global downhill protein folding. Nature. 442:317–321.
47. Choy, W. Y., and L. E. Kay. 2003. Modelselection for the interpretation
of protein side chain methyl dynamics. J. Biomol. NMR. 25:325–333.
48. Oldfield, C. J., J. Meng, ., A. K. Dunker. 2008. Flexible nets: disorder
andinduced fitin the associationsof p53 and 14-3-3with theirpartners.
BMC Genomics. 9(Suppl 1):S1.
49. Farber, P., H. Darmawan, ., A. Mittermaier. 2010. Analyzing protein
folding cooperativity by differential scanning calorimetry and NMR
spectroscopy. J. Am. Chem. Soc. 132:6214–6222.
50. Keppel, T. R., B. A. Howard, and D. D. Weis. 2011. Mapping unstruc-
with amide H/D exchange mass spectrometry. Biochemistry. 50:8722–
Biophysical Journal 102(7) 1627–1635
The Core of NCBD Is Well Ordered1635