The extremely slow-exchanging core and
acid-denatured state of green fluorescent protein
Jie-rong Huang,1,2 Shang-Te Danny Hsu,1 John Christodoulou,1,3
and Sophie E. Jackson1
1Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW,
United Kingdom
2Division of Structure Biology, Biozentrum, University of Basel, Klingelbergstrasse 50, 4056 Basel,
Switzerland
3Institute of Structural Molecular Biology, Department of Biochemistry and Molecular Biology,
University College London; and School of Crystallography, Birbeck College, United Kingdom

Received 19 May 2008; accepted 7 July 2008; published online 15 September 2008)
Green fluorescent protein „GFP… is a large protein with a complex eleven-
stranded
-barrel structure. Previous studies have shown that it has a complex
energy landscape for folding on which there are several intermediate states and
a denatured state with significant residual structure. Here, we use two different
types of H/D exchange measurement and nuclear magnetic resonance „NMR…
techniques to probe the energy landscape for folding of GFP in further detail. H/D
exchange experiments were performed over a wide range of conditions including
different concentrations of denaturant. Results show that the penetration model
dominates the exchange mechanism, consistent with the known stability and
slow unfolding kinetics of GFP. H/D exchange experiments at high pH establish
that there is an extremely slow-exchanging superstable core of amide protons in
GFP that are clustered and located in
-strands 1, 2, 4, 5, and 6. These residues
form part of a mini-
-sheet which we propose constitutes a folding nucleus.
Using a pulsed-labeling strategy, the acid-denatured state has been investigated
and the residual structure observed in earlier studies shown to locate to

-strands 1 and 3. There is some evidence that this residual structure is
stabilized by a localized hydrophobic collapse of the polypeptide chain.
[DOI: 10.2976/1.2976660]
CORRESPONDENCE
Sophie E. Jackson:
sej13@cam.ac.uk
Green fluorescent protein (GFP) from the
jellyfish Aequorea victoria is one of the most
important proteins currently used in biological
and medical research having been extensively
engineered for use as a marker of gene expres-
sion and protein localization, as an indicator of
protein-protein interactions and as a biosensor
(Tsien, 1998). Its widespread use results from
its unique spectroscopic properties, the 238-
residue protein undergoing an autocatalytic
post-translational cyclization and oxidation
of the polypeptide chain around residues
Ser65, Tyr66, and Gly67, to form an extended
and rigidly immobilized conjugated
system,
the chromophore, which emits green fluores-
cence (Zimmer, 2002). No cofactors are neces-
sary for either the formation or the function
of the chromophore (Reid and Flynn, 1997),
which is embedded in the interior of the pro-
tein surrounded by an 11-stranded -barrel
(Ormo et al., 1996; Yang et al., 1996) (Fig. 1).
In all cases, GFP needs to fold efficiently in
order to function in the myriad of biological as-
says and experiments in which it is used, and
its inefficient folding is known to limit its use
in some applications (Tsien, 1998). Over the
past 5 years, a number of studies have been
published on the folding and stability of GFP
and related fluorescent proteins, see Jackson
et al. (2006) for a recent review. Kuwajima and
co-workers have studied the complex refolding
kinetics of GFP from an acid-denatured state
and proposed a model of folding in which the
protein folds through parallel pathways and
several intermediate states (Fukuda et al.,
2000; Enoki et al., 2004). Recently, they have
HFSP Journal ART ICLE
HFSP Journal © HFSP Publishing $25.00
http://hfspj.aip.org/doi/10.2976/1.2976660

used small-angle x-ray scattering along with chromophore
and tryptophan fluorescence studies to show that the two ki-
netic intermediates observed during refolding from pH 2.0 to
7.5 are closely related to an intermediate populated under
equilibrium conditions at pH 4.0, which has molten-globule-
like properties (Enoki et al., 2006). In previous studies, we
have combined H/D exchange nuclear magnetic resonance
(NMR) techniques with fluorescence measurements of che-
mically induced denaturation to establish that there is at least
one stable and highly structured intermediate state populated
under equilibrium conditions (Huang et al., 2007). Recently,
studies on the superfolder GFP (sfGFP) showed that hyster-
esis observed in the unfolding and refolding equilibrium
curves was due to the trapping of GFP in a native-like inter-
mediate state (Andrews et al., 2007). This rough-energy
landscape of sfGFP was attributed to the novel backbone cy-
clization of GFP which is the first step in chromophore for-
mation (Andrews et al., 2007). Clearly, GFP can populate a
number of intermediate states depending upon conditions. It
is the aim of this study to gain further information on the
nature of these intermediate states and the folding energy
landscape for this large -barrel protein.
In our previous H/D exchange studies of GFP, around 40
amide protons were identified that were so highly protected
from exchange that accurate rate constants could not be ob-
tained under the conditions used (pH 7.4, 37 °C) (Huang
et al., 2007). Many of these residues are clustered on one
side of the -barrel structure of GFP located in -strands 1,
2, 4, 5, and 6, with considerably fewer located on the oppo-
site face in -strands 3, 7, 8, 9,10, and 11 (Fig. 1). Here,
we investigate the exchange behaviour of amide protons in
GFP in further detail. In addition to studying the denaturant
concentration dependence of the exchange rates of those
amide groups which have measurable exchange rates at pH
7.4, 37 °C, we measured the exchange rates of the 40 very
slow-exchanging amide protons over a wide range of con-
ditions which promote faster exchange. The results show
that there is a superstable core within the structure of GFP,
in which amide protons are protected from H/D exchange
even at high pH, high temperature, and high concentrations
of denaturant.
Denatured states containing hydrophobic clusters and/or
well-defined secondary structure have been observed for
an increasing number of proteins (Pace et al., 1992; Shortle
et al., 1992; Shortle and Ackerman, 2001; Klein-
Seetharaman et al., 2002; Robic et al., 2003; Wirmer et al.,
2004; Religa et al., 2005). Fersht and co-workers have
suggested that, under some conditions, the denatured
states of proteins may be similar to early folding intermedi-
ates (Oliveberg and Fersht, 1996; Religa et al., 2005).
Studies on the properties of denatured states have therefore
become an important area in the study of protein folding
pathways. From far-ultra violet (UV) circular dichroism
(CD), it is clear that there is residual secondary structure
in the acid-denatured state of GFP (Enoki et al., 2004) re-
sults supported by small-angle x-ray scattering studies
(Enoki et al., 2006) and 19F-NMR/photo-CIDNP experi-
ments (Khan et al., 2006). Other than the fact that residual
structure is present in the denatured state of GFP at low pH,
little is known about which regions of the protein are in-
volved. Here, we have employed a H/D exchange approach
for probing those nonexchanged protons in the acid-
denatured state that are associated with residual structure.
Ten residues are observed which are protected from ex-
change in the acid-denatured state—revealing important
information on the structure in this state. A significant
number of these residues are in regions which are identified
as forming part of the superstable core of GFP. The implica-
tions of these results for the folding mechanism of GFP are
discussed.
RESULTS AND DISCUSSION
Mechanisms of H/D exchange in proteins
There are two dominant mechanisms by which amide
protons in proteins can exchange with solvent. One involves
a global or subglobal unfolding event (Englander and
Kallenbach, 1983) and the other involves penetration of a
solvent molecule into the protein structure and into proxim-
ity of the site of the amide group undergoing exchange
(Miller and Dill, 1995; Dempsey, 2001). In the latter case,
the so-called penetration model, hydrogen exchange between
the protein and bulk solvent, is induced by the redistribution
of interior hydrogen bonds or from the random association of
pre-existing interior cavities (Nakanishi et al., 1973; Karplus
and McCammon, 1981; Wagner and Wuthrich, 1982; Otting
et al., 1991; Mayo and Baldwin, 1993). Thus, solvent
molecules enter the protein core through transiently formed
channels and cavities, which are small and exhibit rapid
Figure 1. Schematic representations for the structure of
GFPuv viewed from two opposite sides drawn by PyMol
„DeLano Scientific LLC…. The chromophore is shown in stick
mode. Each -strand is numbered from the N to the C terminus. The
very slowly H/D exchanging residues at pH 7.4, 37 °C are repre-
sented as balls
red balls: residues on -sheets; yellow balls: resi-
dues inside the barrel. The side containing 10, 7, 8, 9 has
less very slow-exchanging protons compared to the side containing
4, 5, 6, 1, 2.
ART ICLE
HFSP Journal

fluctuations of interior atoms (on the order of tenths to sev-
eral angstroms). In this case, the exchange process is highly
localized and, as such, amide protons in adjacent residues
can exchange independently rather than cooperatively. In
contrast, chemical exchange in the “local unfolding model”
stemming from subglobal or global unfolding events are a
cooperative process according to their position within ele-
ments of secondary structure. This is characterized by the ex-
change rates of a group of amide protons having a similar
dependence on denaturant concentration, i.e., similar m
values and values of GHX, the exchange free energy; this
generates the so-called isotherms (Bai et al., 1995) shown in
Fig. 2(A). In fact, both models are observed and can operate
simultaneously (Miller and Dill, 1995). Data can be fit to a
combined model using Eq. (1) (Yan et al., 2002)
G = − RT ln


exp

− Gpt
RT

+ exp

mD − GHX
0
RT


.
1
The first term in the parenthesis is the contribution from the
penetration-induced exchange (where Gpt is the energy re-
quired for penetration) and the second term the contribution
from global/subglobal unfolding events (where GHX
0 is the
H/D exchange energy of the global/subglobal event in the
absence of denaturant, and [D] is the concentration of dena-
turant). If the first term is small compared to the second, then
the equation simplifies to
G = GHX
0 − mD . 2
This produces a straight line with slope of minus m in a plot
of GHX versus [D], e.g., line (a) in Fig. 2(A). However, if
the penetration term dominates the exchange mechanism,
the slope of this straight line decreases, Fig. 2(A).
In order to distinguish between the penetration model and
the subglobal/global unfolding model, several groups have
investigated the dependence of amide exchange rates on de-
naturant concentration. This method has also been used to
provide information on the cooperativity of unfolding events
for proteins which have amide groups which exchange
through subglobal unfolding events (Mayo and Baldwin,
1993; Bai et al., 1995; Chamberlain et al., 1996; Yan et al.,
2002; Bhutani and Udgaonkar, 2003).
H/D exchange rates for GFP at pH 7.4 in the presence
of denaturant
In order to establish the exchange mechanism of the slow-
exchanging amide protons in GFP whose exchange rates
could be determined at pH 7.4, 37 °C (Huang et al., 2007), a
series of experiments were undertaken at different concen-
trations of chemical denaturant. 1H– 15N heteronuclear
NMR spectroscopy was used to determine the H/D exchange
rates at pH 7.4 in six different concentrations of GdmCl
ranging from 0 to 1.0 M. Samples were pre-equilibrated
in GdmCl solutions at 37 °C for 8 weeks prior to the mea-
surements in order that the samples had reached equilibrium
(Huang et al., 2007). The values of GHX calculated from
the measured exchange rate constants as a function of de-
naturant concentration are shown in Fig. 2(B) and Sup-
plementary Material S1. Figure 2(B) shows typical data for
just a few amide protons. However, none of the amide ex-
change rates showed any dependence on denaturant concen-
tration, Supplementary Materials S1, suggesting that the sol-
vent penetration model dominates under these conditions
(Woodward et al., 1982). In order to test this, values of the
free energy GHX as a function of denaturant concentration
were calculated using a range of penetration energies Gpt,
from 4 to 16 kcal mol−1, and previously determined values
for the global unfolding energy, GHX
0 , 16 kcal mol−1 and
m value 6.5 kcal mol−1 M−1 obtained from fluorescence
equilibrium measurements (Huang et al., 2007) [Fig. 2(A)].
Even when Gpt is as high as 10 kcal mol
−1, the penetration
mechanism still dominates the H/D exchange processes for
Figure 2. Comparison of calculated and observed exchange energies.
A Estimation of GHX versus denaturant concentration for
different exchange mechanisms. All curves are generated by Eq.
1, with m=6.5 kcal mol−1 M−1, GHX0 =16 kcal mol−1, and T=310 K. Lines

a–
h are generated using different values of Gpt from 16 to 4 kcal mol−1.
B Measurable H/D exchange peaks at pH 7.4, 37 °C as a
function of denaturant concentration. Only residues on -strand 1 are shown, the rest are shown in supplementary material. V12
squares,
L18
triangles, D19
reverse triangles, G20
diamonds, and D21
circles.
HFSP Journal
Extremely slow-exchanging core and acid- . . . | Huang et al.

GFP and the denaturant dependency only becomes evident
when the concentration of denaturant is higher than 1.0 M
GdmCl. These simulated curves, which have been generated
using reasonable estimated parameters, agree with our ex-
perimental results which show that, at pH 7.4, GHX are in-
dependent of denaturant concentration.
These results are somewhat unusual [the exchange rates
of most proteins studied to date show some dependence on
denaturant concentrations (Takei et al., 2002; Yan et al.,
2002; Krishna et al., 2004)], and we attribute this to the fact
that GFP is a very stable protein, i.e., it has a large free en-
ergy of unfolding (Dietz and Rief, 2004; Huang et al., 2007).
It is interesting to note that the values ofGHX obtained from
the exchange experiments are all lower than those obtained
from chemical denaturation unfolding curves monitored by
fluorescence and, therefore, exchange is unlikely to be
caused by a global unfolding event. As GHX does not de-
pend upon denaturant concentration, it is also unlikely that
exchange is caused by a subglobal unfolding event, although
a very localized unfolding event where there is only a very
small change in solvent accessible surface area cannot be
ruled out completely. In other words, for GFP, it is likely that
the penetration energy is sufficiently small compared to the
subglobal or global unfolding energies that, under these con-
ditions, the penetration model dominates the exchange
mechanism.
A superstable core of GFP characterized at extreme
conditions
To further examine the 40 extremely slow-exchanging amide
groups identified in our earlier work, experiments were car-
ried out at a higher pH (8.5) in the presence of 0–1.6 M Gd-
mCl in order to accelerate the exchange processes. In all
cases, the exchange profiles show no dependence on denatur-
ant concentration. Three groups of amide protons were iden-
tified under these new conditions: (i) those that exchange in
four weeks (grey peaks in Fig. 3); (ii) those that exchange in
12 weeks (Leu60, Val93, Asn121, Met218, Leu201, and
Leu220) (peak labels underlined in Fig. 3); and (iii) those
that show no exchange during the period of the experiment,
includes residues Ile14, Leu15, Tyr92, Tyr106, Val112,
Lys113, and Asn120 [bold labels in Fig. 3; their representa-
tive exchange profiles measured in 0.5 M GdmCl are shown
in Supplementary Material S2.
Although the seven residues which show most protection
from exchange do not exchange sufficiently fast for an accu-
rate exchange rate to be measured, it is possible to estimate
an upper limit of kex10
−8 s−1 based on the experimental
data (see Supplementary Material S2). It is well established
from studies on other systems, that the highest energy for
H/D exchange is likely to correspond to the global unfolding
energy if isotopic effects and proline isomerization are
considered (Bai et al. 1994; Huyghues-Despointes et al.,
1999). Assuming the seven slowest-exchanging residues are
at the EX2 limit, their corresponding GHX would be in the
range of 14–18 kcal mol−1 (Table I) based on the estimated
exchange rates. These high values indicate that the structure
is very stable, which is consistent with previous studies
on GFP (Huang et al., 2007). A direct comparison of GHX
with the free energy of unfolding obtained from equilibrium
measurements is not possible as the two experiments
were carried out under different conditions. However, if it is
assumed that the stability of GFP does not vary much be-
tween pH 7.4 and 8.5, and there is some experimental evi-
dence to support this (Alkaabi et al., 2005), the free energy
of unfolding at 1.6 M GdmCl can be estimated to be approxi-
mately 6 kcal mol−1 [Fig. 2(A)]. As the values for GHX
measured are considerably higher than this, then this sug-
gests that there is considerable residual structure in the dena-
tured state of GFP under native conditions. If, on the other
hand, residues exchange at the EX1 limit, then kex=kop, i.e.,
the opening rate constant is around 10−8–10−10 s−1. This is
within the range of values obtained for the unfolding rate
constant at pH 7.4 in 1.6 M GdmCl calculated from extrapo-
lation of the unfolding rate constants measured by stopped-
flow spectroscopy at high concentrations of denaturant
(Huang et al., 2007). It is unlikely the high levels of amide
protection are due to aggregation as no change in chemical
shift or line broadening is observed during the course of
the experiment. Thus, the seven slowest exchanging residues
in GFP (Fig. 4) all show extremely strong protection. These
results are consistent with previous studies which show that
GFP has a high stability and slow unfolding rates (Huang
Figure 3. Comparison of the residual peaks in the HSQC spec-
trum of GFP measured under different conditions. The peaks
obtained at pH 7.4, 37 °C after 1 month of exchange are shown in
grey. The 15 peaks that exchange very slowly are labeled: the six
that exchanged after 12 weeks are underlined, while the seven that
exchange extremely slowly are shown in bold.
ART ICLE
HFSP Journal

et al., 2007), but also with a model in which there is signifi-
cant residual structure in this region of the protein under na-
tive conditions.
The acid-denatured state of GFP has residual structure
A variety of NMR methods have been used to obtain infor-
mation on the structure of denatured states of proteins
(Dyson and Wright, 2004). Typically, chemical shifts, cou-
pling constants, 15N-relaxation rates, amide-proton tempera-
ture coefficients, as well as NOE data have all been used as
probes of structure in the denatured ensemble. These experi-
ments have revealed that there can be both native-like and
non-native secondary structure present, the exact structure
depending upon the protein and experimental conditions.
For example, both native and non-native secondary
structural preferences were observed in the urea-denatured
state of barstar (Bhavesh et al., 2004), in the GdmCl-
denatured state of HIV-1 protease (Bhavesh et al., 2001;
2003; Chatterjee et al., 2005), and in the acid-denatured state
of hUBF HMG box 1 (Zhang et al., 2005). In contrast,
-type preferences were seen in denatured apomyoglobin
which is a highly helical protein (Mohana-Borges et al.,
2004). As GFP is a large protein and an assignment of the
denatured state is not possible without extensive and expen-
Table I. Estimation of the GHX for the extremely slow-exchanging amide proton in GFP using different limiting exchange rates.
Residues kint s−1
GHX kcal mol−1
(If kex=10−8 s−1) (If kex=10−9 s−1) (If kex=10−10 s−1)
I14 68.4 14.0 15.4 16.8
L15 98.9 14.2 15.6 17.0
Y92 507 15.2 16.6 18.0
Y106 717 15.4 16.8 18.3
V112 90.2 14.1 15.6 17.0
K113 422 15.1 16.5 17.9
N120 78.6 14.1 15.5 16.9
Figure 4. The superstable core
of GFP. The three-dimensional
structure of GFP showing the posi-
tions of all the extremely slow ex-
changing residues
red balls. Note
that they are all located on one side
of GFP, and connect 4, 5, 6,
1, and 2. The topology diagram

bottom left showing the hydrogen
bonding by the seven very slow-
exchanging amide protons.
HFSP Journal
Extremely slow-exchanging core and acid- . . . | Huang et al.

sive labeling strategies, an alternative approach to those dis-
cussed above was employed here. H/D exchange of amide
protons in the denatured state was measured by refolding the
protein and probing the native state with conventional NMR
experiments.
A pulse-labeling strategy was used to carry out a H/D ex-
change study on the acid-denatured state of GFP. Denatured
samples at pH 2.9 were left in D2O for 5, 20, 40, or 80 min
and then refolded and 15N, 1H correlation spectra of the na-
tive state acquired. A one-dimensional proton spectrum of
the native state of a sample of GFP which had never been
unfolded was compared with samples which had been la-
beled in the acid-denatured state and subsequently refolded,
in order to ensure that refolding was complete and success-
ful. After labeling with D2O in the acid-denatured state, a
significant number of crosspeaks were observed to disappear
from the amide proton region of the native spectrum, how-
ever, the aliphatic and aromatic regions of the spectra re-
mained identical (see Supplementary Material S3). This
shows that the protein refolded to the correct native state
after denaturation.As expected, a large number of amide pro-
tons exchange rapidly in the denatured state. However, a
number of peaks showed significant remaining intensity after
5 min of denaturation under acid conditions and some of
these peaks remained even after prolonged exchange times
such as 20, 40, and 80 min (Supplementary Material S4). If
there were no protection against exchange in the denatured
state, the residue with the slowest intrinsic exchange rate at
pH 2.9, 25 °C would have less than 5% of the initial inten-
sity after 80 min of exchange and therefore would be unde-
tectable (inset of Fig. 5). The fact that ten crosspeaks re-
mained clearly visible after 80 min of incubation at pH 2.9
in D2O indicates that there is a significant degree of protec-
tion of these amide groups as a result of residual structure
(Fig. 5). The results clearly show that there is significant re-
sidual structure in GFP at low pH, although these types of
experiments do not directly reveal whether the structure is
native or non-native.
If the residual structure in acid-denatured GFP is native-
like, then structural mapping of the highly protected residues
show that they form a well-defined cluster spanning across
-strands 1, 2, 3, 6, and 11 (see Fig. 6).Although the residues
that show protection are located only in -strands 1 and 3, in
the native state they hydrogen bond with groups in -strands
2, 6, and 11. For example, the amide proton of Leu44 is pro-
tected in the denatured state, and its hydrogen bond acceptor
is the carbonyl oxygen of Leu220 on -strand 11. However,
the amide group of Leu220 is not protected in the acid-
denatured state. Although this may suggest that the residual
structure is not native-like, it is possible that in a highly dy-
namic and flexible state, such as a denatured state, only one
of the hydrogen-bonding partners is protected.An alternative
explanation is that the protection of amide groups in
-strands 1 and 3 in the acid-denatured state is a result of
non-native residual structure. It is interesting to note that
eight out of the ten protected amides are associated with resi-
dues which have hydrophobic sidechains and that these are
all located relatively close in sequence in the region spanning
from residue 12 to 47 on -strands 1 and 3. Thus, residual
structure in the acid-denatured state of GFPmay be driven by
hydrophobic collapse which results in non-native contacts,
and non-native hydrogen bonding. This would explain why
the native hydrogen-bonding partners of residues 12–18 and
44–47 do not show protection. Such hydrophobic collapse in
a denatured state has been observed for other proteins
(Bhavesh et al., 2003; Iimura et al., 2007), for a recent re-
view see reference (Bowler, 2007).
Whether the residual structure in the acid-denatured
state of GFP is native or non-native, it is clearly present in
the N-terminal region of the protein and this has some po-
Figure 5. Comparison of acid-denatured-refolded spectrum
„black… and native state spectrum „green…. The sample was incu-
bated under acid-denaturing conditions for 80 min before refolding.
The inset is the estimated exchanging profile of the amide proton
with the slowest intrinsic rate constant 8.6710−4 s−1 under the
conditions used.
Figure 6. Three-dimensional representation of the residues
with protected amide protons under acid denaturing condi-
tions. Yellow balls: residues with protected hydrogens in the acid-
denatured state. Trp57 is shown in red and the chromophore in
yellow.
ART ICLE
HFSP Journal

tential significance with regard to the folding of the large
-barrel protein. Such residual structure is likely to reduce
the initial conformational search the polypeptide chain has
to undertake during the first steps of folding and this may
lead to faster and more efficient folding. For GFP, this is
supported by results from folding studies which have
shown that the refolding efficiency and rate from the acid
denatured state (Fukuda et al., 2000; Enoki et al., 2004),
which we and others have shown contains significant re-
sidual structure, is considerably higher than from the chemi-
cally denatured state produced by high concentrations of Gd-
mCl which appears to have little or no residual structure
(Khan et al., 2006).
Kuwajima and co-workers have shown that a burst-phase
intermediate is transiently populated during the refolding of
GFP from the acid-denatured state. A rapid increase in tryp-
tophan fluorescence was observed and this was attributed to
the formation of an early intermediate in which the single
tryptophan residue (Trp57) was partially buried (Enoki et al.,
2004). This observation can be explained in more detail by
our results: in the acid-denatured state, the whole polypep-
tide is flexible and without fixed tertiary structure including
Trp57, except for the region containing the residual structure
identified in this study (yellow balls in Fig. 6). Trp57 (red
stick in Fig. 6) is very close in sequence and space to the
residual structure in the N-terminal region of the protein, and
it is therefore likely that it interacts rapidly with the hydro-
phobic cluster formed by residues 12–18 and 44–47 imme-
diately after refolding is initiated, thereby protecting the aro-
matic side chain from bulk solvent and increasing the
fluorescence yield (Enoki et al., 2004; 2006).The results pre-
sented here are consistent with this model and provide a
mechanism for Trp57 burial. After this step, the on-pathway
intermediate is formed as discussed in our previous studies
(Huang et al., 2007) and by Enoki et al. (2004).
Implications for the folding mechanism of GFP
Using two different types of H/D exchange measurement by
NMR, we have probed the energy landscape for folding of
GFP in detail and have gleaned important details on the
mechanism of amide exchange, the regions of the protein
which are highly protected from exchange, as well as identi-
fying residual structure in the acid-denatured state.Although
in the past there has been some debate about the validity of
using H/D exchange results measured under equilibrium
conditions to inform on folding pathways (Clarke et al.,
1997; Clarke and Itzhaki, 1998), recent work has established
that the results from H/D exchange experiments measured
under such conditions are similar to those obtained in a ki-
netic pulsed H/D exchange experiment (Englander et al.,
2007). Together, our results provide additional and valuable
information relevant to the folding mechanism of GFP.
It is clear from a large number of studies on small pro-
teins that residual structure in denatured states can play an
important role in the early states of folding (Oliveberg and
Fersht, 1996; Religa et al., 2005). From studies of GFP, it is
known that there is considerable residual structure under
acid-denaturing conditions as shown by far-UV CD (Enoki
et al., 2004), small-angle x-ray scattering (Enoki et al., 2006)
and 19F-NMR photo-CINDP (Khan et al., 2006) experi-
ments, however, it was not know which regions of the protein
were involved in the residual structure. Here, we have iden-
tified the residues associated with this residual structure and
shown that they locate to -strands 1 and 3. As many of the
residues involved in the residual structure have hydrophobic
sidechains, in addition to the fact that the native-state
hydrogen-bonding partners of these residues do not show
protection in the acid-denatured state, it is likely that this re-
sidual structure is driven by a localized hydrophobic collapse
or clustering and that non-native interactions are involved.
We propose that the first step in folding either from an ex-
tended, random-coil-like state, or from the ribosome, is the
rapid formation of this residual structure in the extreme
N-terminus of the protein. Formation of this structure may
result in the partial burial of Trp57 and correspond to the
early intermediate identified by Kuwajima and co-workers
(Enoki et al., 2004).
The identification and characterization of a superstable
core of GFP, in which amide protons show an extremely high
level of protection from exchange, reveals a partially struc-
tured state of the protein that is transiently populated under
native conditions. The superstable core involves residues in
-strands 1, 4, 5, and 6 which together form a four-stranded
mini -sheet structure. In this case, and in contrast to the
results in the acid-denatured state, most of the native
hydrogen-bonding partners are protected to the same degree,
strongly suggesting that native-like secondary structure is
formed in this region. In fact, the protection of these amides
is so strong that it is possible that these residues form re-
sidual structure in the denatured state which is transiently
populated under native conditions. This structure may be
formed cooperatively although this cannot be stated with cer-
tainty as it was not possible to measure the exchange rates for
these amide groups. From our results, however, we propose
that this region is either structured in the denatured state
populated under native conditions, or constitutes a nucleus
for folding formed very early on the folding pathway, and
subsequent to its formation the rest of the -barrel structure
forms over several steps.
From the H/D exchange rates described in this study, in
conjunction with our previous studies and those of others, we
propose a model where, after -strands 1, 4, 5, and 6 have
formed some native-like structure, then this structure is first
consolidated by a strengthening of -strands 4 and 6, with
some interactions with residues in the central -helix and
residues in adjacent strands 10 and 11 forming (residues
Leu60, Gln94, Asn121, Leu201, Met218, and Leu220 all
show levels of protection intermediate between the super-
HFSP Journal
Extremely slow-exchanging core and acid- . . . | Huang et al.

stable core residues and those identified in an earlier study
which we have attributed to a late folding intermediate
(Huang et al., 2007). At the final stage, considerable native
secondary structure is made between -strands 1, 4, 5, 6 and
to a lesser extent the N-terminal region of the central  helix
with additional structure emerging and some protection seen
in all the other  strands except -strand 9. This constitutes
the formation of the late folding intermediate which has been
shown to be highly structured by fluorescence, H/D ex-
change and other studies, with partial unfolding in the region
of -strands 7-10 (Helms et al., 1999; Seifert et al., 2003;
Enoki et al., 2006; Andrews et al., 2007; Huang et al., 2007).
At some point, either in the last transition state or during the
energetically downhill process post transition state, these last
three to four strands become completely structured and the
native state is achieved.
MATERIAL AND METHODS
Protein expression and reagents
Truncated GFPuv expression and purification is described
elsewhere (Huang et al., 2007). Ultrapure guanidinium chlo-
ride (GdmCl) was purchased from ICN Biomedicals, Inc.
D2O (99.9%) and
15NH4Cl were from Cambridge Isotope
Laboratories, Inc. All other chemicals were of analytical
grade and purchased form Sigma, BDH, or Melford Labora-
tories. Millipore-filtered, double-de-ionized water was used
throughout.
H/D exchange experiments
Protein samples were pre-equilibrated at 37 °C in various
concentrations of GdmCl. Deuterated GdmCl solutions at
the required pH, buffer, and concentration were prepared in
D2O and were used for initiating H/D exchange by means of
a NAP-5 Column (Amersham Bioscience). This buffer ex-
change step took about 5 min before the first heteronuclear
single-quantum coherence (HSQC) spectrum was acquired,
longer than in previous studies where lyophilized protein was
dissolved directly into D2O (Huang et al., 2007). However,
in these studies, buffer exchange using the NAP-5 column
was necessary to prevent the aggregation of GFP during lyo-
philization in the presence of GdmCl. As the aim of these
experiments was to focus on the very slow-exchanging
amide protons, the increased dead time was not an issue. The
final concentration of GFP was about 600 µM, and all ex-
periments were performed at 37 °C and the samples were
stored at 37 °C between measurements.
All NMR spectra were acquired on a Bruker AVANCE
500 or 700 MHz spectrometer equipped with 5 mm 1H, 13C,
and 15N triple-resonance cryogenic probe heads. Phase-
sensitive HSQC using Echo/Antiecho-TPPI gradient selec-
tion with decoupling during acquisition (Palmer et al., 1991;
Grzesiek and Bax, 1993; Schleucher et al., 1994) was used
to record the two-dimensional 1H– 15N spectra. All HSQC
spectra were acquired with 1024 t2 and 128 t1 complex
points. All spectra were acquired at 37 °C. Acquired data
were processed by NMRPipe (Delaglio et al., 1995) and
analyzed with NMRView (Johnson and Blevins, 1994). In-
tensities of peaks were exported from NMRView and ana-
lyzed by PRISM (GraphPad, Inc.).
Analysis of H/D exchange data and the EX1/EX2 limit
The mechanism of H/D exchange is well studied and re-
viewed (Linderstrom-Lang, 1958; Hvidt, 1966; Huyghues-
Despointes et al., 1999; Englander, 2000; Dempsey, 2001;
Ferraro et al., 2004; Krishna et al., 2004), and can be de-
scribed by the following two-step model:
NHclosed

kcl
kop
NHopen⇀
kint
NDopen, 3
where kop and kcl are the opening (unfolding) and closing
(folding) rate constants, respectively. The intrinsic rate con-
stant for exchange kint depends on the residue type and vari-
ous conditions (pH, temperature, neighboring amino acids,
and isotope effects) and can be estimated on the basis of
model compound data (Bai et al., 1993). The exchange rate
constant kex determined from the model is
kex =
kop · kint
kcl + kint
, 4
which has two limits, so-called EX1 and EX2 (Dempsey,
2001; Krishna et al., 2004).At the EX2 limit, kclkint,GHX
can be calculated from the measured exchange rate constants
using the following equation:
GHX = − RT ln Kop = RT ln
kint
kex
, 5
where Kop is the equilibrium opening constant and is equal to
kop/kcl.At the EX1 limit, kintkcl, the exchange rate constant
is simply equal to the opening rate constant
kex = kop. 6
The web-based program SPHERE (Zhang and Roder; Bai
et al., 1993; Connelly et al., 1993) was used for calculating
kint, and pDs of 7.8 and 8.9 (0.4 pH units above the measured
pH in D2O) were used.
The acid-denatured sample preparation and NMR
spectra acquisition
1 mL of 600 µM GFP in PBS, pH 7.4 was diluted ten times
into predeuterated PBS, pH 2.9. The denatured sample was
left at 25 °C (incubated in water bath) for varying lengths of
time. During this time, unprotected amide protons ex-
changed with solvent deuterons, but amide hydrogens which
are involved in residual structure remain protected from ex-
change. The denatured solution was then diluted ten times
into deuterated PBS buffer, pH 7.4. In this step, protein
refolded and the hydrogen-bonded protons in the native
structure were protected against exchange. The sample was
ART ICLE
HFSP Journal

left at 25 °C for 5 min for complete refolding to take place,
then concentrated at 4 °C to minimize further H/D ex-
change. The concentration step from a diluted sample of
100 mL to 600 µL took around 4 h using Vivaspin centri-
cons. The NMR spectra of concentrated samples were ac-
quired at 37 °C with Bruker AVANCE 500 or 700 MHz
spectrometers. The band-selective optimized flip-angle
short-transient HMQC pulse sequence (Schanda and
Brutscher, 2005) was applied in order to minimize the ex-
perimental time.
REFERENCES
Alkaabi, KM, Yafea, A, and Ashraf, SS (2005). “Effect of pH on thermal-
and chemical-induced denaturation of GFP.” Appl. Biochem.
Biotechnol. 126, 149–156.
Andrews, BT, Schoenfish, AR, Roy, M, Waldo, G, and Jennings, PA
(2007). “The rough energy landscape of superfolder GFP is
linked to the chromophore.” J. Mol. Biol. 373, 476–490.
Bai, Y, Milne, JS, Mayne, L, and Englander, SW (1993). “Primary
structure effects on peptide group hydrogen exchange.”
Proteins 17, 75–86.
Bai, Y, Milne, JS, Mayne, L, and Englander, SW (1994). “Protein stability
parameters measured by hydrogen exchange.” Proteins 20, 4–14.
Bai, Y, Sosnick, TR, Mayne, L, and Englander, SW (1995). “Protein
folding intermediates: native-state hydrogen exchange.” Science
269, 192–197.
Bhavesh, NS, Juneja, J, Udgaonkar, JB, and Hosur, RV (2004). “Native
and nonnative conformational preferences in the urea-unfolded
state of barstar.” Protein Sci. 13, 3085–3091.
Bhavesh, NS, Panchal, SC, Mittal, R, and Hosur, RV (2001). “NMR
identification of local structural preferences in HIV-1 protease
tethered heterodimer in 6 M guanidine hydrochloride.” FEBS Lett.
509, 218–224.
Bhavesh, NS, Sinha, R, Mohan, PM, and Hosur, RV (2003). “NMR
elucidation of early folding hierarchy in HIV-1 protease.”
J. Biol. Chem. 278, 19980–19985.
Bhutani, N, and Udgaonkar, JB (2003). “Folding subdomains of
thioredoxin characterized by native-state hydrogen exchange.”
Protein Sci. 12, 1719–1731.
Bowler, BE (2007). “Thermodynamics of protein denatured states.”
Mol. Biosyst. 3, 88–99.
Chamberlain, AK, Handel, TM, and Marqusee, S (1996). “Detection of
rare partially folded molecules in equilibrium with the native
conformation of RNaseH.” Nat. Struct. Biol. 3, 782–787.
Chatterjee, A, Mridula, P, Mishra, RK, Mittal, R, and Hosur, RV (2005).
“Folding regulates autoprocessing of HIV-1 protease precursor.”
J. Biol. Chem. 280, 11369–11378.
Clarke, J, and Itzhaki, LS (1998). “Hydrogen exchange and protein
folding.” Curr. Opin. Struct. Biol. 8, 112–118.
Clarke, J, Itzhaki, LS, and Fersht, AR (1997). “Hydrogen exchange at
equilibrium: a short cut for analysing protein-folding pathways?”
Trends Biochem. Sci. 22, 284–287.
Connelly, GP, Bai, Y, Jeng, MF, and Englander, SW (1993). “Isotope
effects in peptide group hydrogen exchange.” Proteins 17,
87–92.
Delaglio, F, Grzesiek, S, Vuister, GW, Zhu, G, Pfeifer, J, and Bax, A
(1995). “NMRPipe: a multidimensional spectral processing
system based on UNIX pipes.” J. Biomol. NMR 6, 277–293.
Dempsey, CE (2001). “Hydrogen exchange in peptides and proteins using
NMR-spectroscopy.” Prog. Nucl. Magn. Reson. Spectrosc. 39,
135–170.
Dietz, H, and Rief, M (2004). “Exploring the energy landscape of GFP by
single-molecule mechanical experiments.” Proc. Natl. Acad. Sci.
U.S.A. 101, 16192–16197.
Dyson, HJ, and Wright, PE (2004). “Unfolded proteins and protein folding
studied by NMR.” Chem. Rev. (Washington, D.C.) 104, 3607–3622.
Englander, SW (2000). “Protein folding intermediates and
pathways studied by hydrogen exchange.” Annu. Rev. Biophys. Biomol.
Struct. 29, 213–238.
Englander, SW, and Kallenbach, NR (1983). “Hydrogen-exchange and
structural dynamics of proteins and nucleic-acids.” Q. Rev.
Biophys. 16, 521–655.
Englander, SW, Mayne, L, and Krishna, MMG (2007). “Protein folding
and misfolding: mechanism and principles.” Q. Rev. Biophys. 40,
287–326.
Enoki, S, Maki, K, Inobe, T, Takahashi, K, Kamagata, K, Oroguchi, T,
Nakatani, H, Tomoyori, K, and Kuwajima, K (2006). “The
equilibrium unfolding intermediate observed at pH 4 and its
relationship with the kinetic folding intermediates in green fluorescent
protein.” J. Mol. Biol. 361, 969–982.
Enoki, S, Saeki, K, Maki, K, and Kuwajima, K (2004). “Acid denaturation
and refolding of green fluorescent protein.” Biophys. J. 43,
14238–14248.
See EPAPS Document No. E-HJFOA5-2-001807 for supplemental
material. This document can be reached through a direct link
in the online article’s HTML reference Section or via the EPAPS
homepage (http://www.aip.org/pubservs/epaps.html).
Ferraro, DM, Lazo, ND, and Robertson, AD (2004). “EX1 hydrogen
exchange and protein folding.” Biophys. J. 43, 587–594.
Fukuda, H, Arai, M, and Kuwajima, K (2000). “Folding of green
fluorescent protein and the cycle3 mutant.” Biophys. J. 39,
12025–12032.
Grzesiek, S, and Bax, A (1993). “Amino acid type determination in the
sequential assignment procedure of uniformly 13C/15N-enriched
proteins.” J. Biomol. NMR 3, 185–204.
Helms, V, Straatsma, TP, and McCammon, JA (1999). “Internal dy-
namics of green fluorescent protein.” J. Phys. Chem. B 103,
3263–3269.
Huang, JR, Craggs, TD, Christodoulou, J, and Jackson, SE (2007). “Stable
intermediate states and high energy barriers in the unfolding of
GFP.” J. Mol. Biol. 370, 356–371.
Huyghues-Despointes, BM, Scholtz, JM, and Pace, CN (1999). “Protein
conformational stabilities can be determined from hydrogen
exchange rates.” Nat. Struct. Biol. 6, 910–912.
Hvidt, A, and Nielsen, SO (1966). “Hydrogen exchange in proteins.”
Adv. Protein Chem. 21, 287–386.
Iimura, S, et al. (2007). “Characterization of the denatured structure of
pyrrolidone carboxyl peptidase from a hyperthermophile under
nondenaturing conditions: role of the C-terminal alpha-helix of the
protein in folding and stability.” Biophys. J. 46, 3664–3672.
Jackson, SE, Craggs, TD, and Huang, JR (2006). “Understanding the
folding of GFP using biophysical techniques.” Expert Rev.
Proteomics 3, 545–559.
Johnson, BA, and Blevins, RA (1994). “NMR view—a computer-program
for the visualization and analysis of NMR data.” J. Biomol. NMR 4,
603–614.
Karplus, M, and McCammon, JA (1981). “The internal dynamics of
globular proteins.” Crit. Rev. Biochem. 9, 293–349.
Khan, F, Kuprov, I, Craggs, TD, Hore, PJ, and Jackson, SE (2006). “(19)F
NMR studies of the native and denatured states of green fluorescent
protein.” J. Am. Chem. Soc. 128, 10729–10737.
Klein-Seetharaman, J, et al. (2002). “Long-range interactions within a
nonnative protein.” Science 295, 1719–1722.
Krishna, MM, Hoang, L, Lin, Y, and Englander, SW (2004). “Hydrogen
exchange methods to study protein folding.” Methods 34, 51–64.
Linderstrom-Lang, K (1958). “Deuterium exchange and protein structure.”
In Symposium of Protein Structure. Neuberger, A (ed.), pp 23–24,
Methuen, London.
Mayo, SL, and Baldwin, RL (1993). “Guanidinium chloride induction of
partial unfolding in amide proton exchange in RNase A.” Science
262, 873–876.
Miller, DW, and Dill, KA (1995). “A statistical mechanical model for
hydrogen exchange in globular proteins.” Methods 4,
1860–1873.
Mohana-Borges, R, Goto, NK, Kroon, GJ, Dyson, HJ, and Wright, PE
(2004). “Structural characterization of unfolded states of
apomyoglobin using residual dipolar couplings.” J. Mol. Biol. 340,
1131–1142.
HFSP Journal
Extremely slow-exchanging core and acid- . . . | Huang et al.

Nakanishi, M, Tsuboi, M, and Ikegami, A (1973). “Fluctuation of the
lysozyme structure. II. Effects of temperature and binding of
inhibitors.” J. Mol. Biol. 75, 673–682.
Oliveberg, M, and Fersht, AR (1996). “Thermodynamics of transient
conformations in the folding pathway of barnase: reorganization
of the folding intermediate at low pH.” Biophys. J. 35, 2738–2749.
Ormo, M, Cubitt, AB, Kallio, K, Gross, LA, Tsien, RY, and Remington,
SJ (1996). “Crystal structure of the Aequorea victoria green
fluorescent protein.” Science 273, 1392–1395.
Otting, G, Liepinsh, E, and Wuthrich, K (1991). “Protein hydration in
aqueous solution.” Science 254, 974–980.
Pace, CN, Laurents, DV, and Erickson, RE (1992). “Urea denaturation of
barnase: pH dependence and characterization of the unfolded state.”
Biophys. J. 31, 2728–2734.
Palmer, AG, Cavanagh, J, Wright, PE, and Rance, M (1991). “Sensitivity
improvement in proton-detected 2-dimensional heteronuclear
correlation NMR-spectroscopy.” J. Mater. Res. 93, 151–170.
Reid, BG, and Flynn, GC (1997). “Chromophore formation in green
fluorescent protein.” Biophys. J. 36, 6786–6791.
Religa, TL, Markson, JS, Mayor, U, Freund, SM, and Fersht, AR (2005).
“Solution structure of a protein denatured state and folding
intermediate.” Nature (London) 437, 1053–1056.
Robic, S, Guzman-Casado, M, Sanchez-Ruiz, JM, and Marqusee, S
(2003). “Role of residual structure in the unfolded state of a
thermophilic protein.” Proc. Natl. Acad. Sci. U.S.A. 100, 11345–11349.
Schanda, P, and Brutscher, B (2005). “Very fast two-dimensional
NMR spectroscopy for real-time investigation of dynamic events in
proteins on the time scale of seconds.” J. Am. Chem. Soc. 127,
8014–8015.
Schleucher, J, Schwendinger, M, Sattler, M, Schmidt, P, Schedletzky, O,
Glaser, SJ, Sorensen, OW, and Griesinger, C (1994). “A general
enhancement scheme in heteronuclear multidimensional
NMR employing pulsed field gradients.” J. Biomol. NMR 4,
301–306.
Seifert, MH, Georgescu, J, Ksiazek, D, Smialowski, P, Rehm, T, Steipe,
B, and Holak, TA (2003). “Backbone dynamics of green
fluorescent protein and the effect of histidine 148 substitution.”
Biophys. J. 42, 2500–2512.
Shortle, D, and Ackerman, MS (2001). “Persistence of native-like
topology in a denatured protein in 8 M urea.” Science 293,
487–489.
Shortle, D, Chan, HS, and Dill, KA (1992). “Modeling the effects of
mutations on the denatured states of proteins.” Protein Sci. 1,
201–215.
Takei, J, Pei, W, Vu, D, and Bai, Y (2002). “Populating partially unfolded
forms by hydrogen exchange-directed protein engineering.”
Biophys. J. 41, 12308–12312.
Tsien, RY (1998). “The green fluorescent protein.” Annu. Rev. Biochem.
67, 509–544.
Wagner, G, and Wuthrich, K (1982). “Amide protein exchange and surface
conformation of the basic pancreatic trypsin inhibitor in solution.”
J. Mol. Biol. 160, 343–361.
Wirmer, J, Schlorb, C, Klein-Seetharaman, J, Hirano, R, Ueda, T, Imoto,
T, and Schwalbe, H (2004). “Modulation of compactness and long-
range interactions of unfolded lysozyme by single point
mutations.” Angew. Chem., Int. Ed. 43, 5780–5785.
Woodward, C, Simon, I, and Tuchsen, E (1982). “Hydrogen exchange and
the dynamic structure of proteins.” Mol. Cell. Biochem. 48,
135–160.
Yan, S, Kennedy, SD, and Koide, S (2002). “Thermodynamic and kinetic
exploration of the energy landscape of Borrelia burgdorferi OspA
by native-state hydrogen exchange.” J. Mol. Biol. 323, 363–375.
Yang, F, Moss, LG, and Phillips, GN, Jr. (1996). “The molecular structure
of green fluorescent protein.” Nat. Biotechnol. 14, 1246–1251.
Zhang, X, Xu, Y, Zhang, J, Wu, J, and Shi, Y (2005). “Structural and
dynamic characterization of the acid-unfolded state of hUBF
HMG box 1 provides clues for the early events in protein folding.”
Biophys. J. 44, 8117–8125.
Zhang, Y-Z, and Roder, H. “SPHERE: a server program for hydrogen
exchange rate estimation.”
Zimmer, M (2002). “Green fluorescent protein (GFP): applications,
structure, and related photophysical behavior.” Chem. Rev.
(Washington, D.C.) 102, 759–781.
ART ICLE
HFSP Journal