Review
10.1586/14789450.3.5.545 © 2006 Future Drugs Ltd ISSN 1478-9450 545www.future-drugs.com
Understanding the folding of GFP
using biophysical techniques
Sophie E Jackson†, Timothy D Craggs and Jie-rong Huang
†Author for correspondence
Chemistry Department,
Lensfield Road,
Cambridge CB2 1EW, UK
Tel.: +44 1223 762 011
Fax: +44 1223 336 362
sej13@cam.ac.uk
KEYWORDS:
aggregation, denatured state,
equilibrium intermediate,
kinetic intermediate, misfolding,
oligomeric state, protein
folding selection
Green fluorescent protein (GFP) and its many variants are probably the most widely used
proteins in medical and biological research, having been extensively engineered to act
as markers of gene expression and protein localization, indicators of protein–protein
interactions and biosensors. GFP first folds, before it can undergo an autocatalytic
cyclization and oxidation reaction to form the chromophore, and in many applications the
folding efficiency of GFP is known to limit its use. Here, we review the recent literature on
protein engineering studies that have improved the folding properties of GFP. In addition,
we discuss in detail the biophysical work on the folding of GFP that is beginning to reveal
how this large and complex structure forms.
Expert Rev. Proteomics 3(5), 545–559 (2006)
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 [1,2]. Its widespread use results from
its unique spectroscopic properties, the
238-residue protein undergoing an autocata-
lytic post-translational cyclization and oxida-
tion of the polypeptide chain around residues
Ser65, Tyr66 and Gly67, to form an extended
and rigidly encapsulated conjugated π system,
the chromophore, which emits green fluores-
cence (FIGURE 1) [3]. No cofactors are necessary
for either the formation or the function of the
chromophore [4], which is embedded in the
interior of the protein surrounded by an
11-stranded β-barrel (FIGURE 2) [5,6]. GFP is
remarkable for its structural stability and high
fluorescence quantum yield, a result of the fact
that, in the native state, the chromophore is
rigid and shielded from bulk solvent. On pro-
tein denaturation, the chromophore remains
chemically intact but fluorescence is lost.
Therefore, the green fluorescence is a sensitive
probe of the state of the protein. It is clear that
GFP needs to fold efficiently in order to func-
tion in the myriad of biological assays and
experiments in which it is used, and inefficient
folding is known to limit its use in some appli-
cations. This review focuses on recent engi-
neering studies aimed at improving the folding
properties of GFP, as well as recent studies
using a range of biophysical techniques to char-
acterize the folding pathway of this complex
and important protein.
Improving the folding properties of GFP:
engineering studies
Wild-type GFP is very prone to misfolding
and aggregation when expressed in
Escherichia coli [1] and, therefore, there has
been considerable effort in engineering vari-
ants of GFP that have better folding proper-
ties. In addition to the intrinsic tendency of
GFP to misfold and aggregate, fusions of
GFP with other proteins frequently show
reduced folding yields [7]. Furthermore, circu-
lar permutants of GFP are often employed as
biosensors and these also show a strong ten-
dency to misfold and aggregate [8,9,10]. The
need for better folding fluorescent proteins,
particularly at higher temperatures, has led to
a number of studies that have isolated so-
called ‘folding’ mutants. These are mutants of
GFP that are brighter in vivo, usually as a
result of more efficient folding. However, in
some cases they may act by improving
CONTENTS
Improving the folding
properties of GFP:
engineering studies
Chromophore formation
Preventing dimerization
High-energy barriers &
slow equilibrium
Oligomeric
fluorescent proteins
Partially structured &
denatured states of GFP
Dynamics
Single-molecule unfolding
& folding
Expert commentary
Five-year view
Key issues
References
Affiliations
For reprint orders, please contact reprints@future-drugs.com

Jackson, Craggs & Huang
546 Expert Rev. Proteomics 3(5), (2006)
chromophore formation (see section on chromophore forma-
tion). In general, these mutants suppress misfolding and
aggregation, rather than accelerating folding. The results from
a number of different laboratories are discussed below.
One of the earliest studies in this area was from the Haselof
group who combined random mutagenesis and screening of
E. coli colonies for increased brightness to obtain a mutant of
GFP which showed a 35-fold increase in green fluorescence inten-
sity when expressed in E. coli and yeast [11]. In this case, two muta-
tions were identified, Val→Ala163 and Ser→Gly175, which were
found to suppress the tendency of GFP to misfold and aggregate
at 37°C [11]. In a separate study by the Kohno group, another
mutation, Ser→Pro147, which enhances green fluorescence at
elevated temperatures, was also identified [12].
Perhaps the most important early study on GFP folding muta-
tions is that by the Stemmer group, who isolated the so-called
‘cycle 3’ mutant also termed GFPuv, which is 42 times more
fluorescent than wild-type GFP and is now used extensively in a
variety of applications. This variant was obtained using fluores-
cence screening of a library of GFP mutants created by DNA
shuffling techniques [13]. It contains three mutations (Phe→Ser99,
Met→Thr153, Val→Ala163) and its structure and folding kinet-
ics have been studied [14,15]. The three mutations lie on the surface
of the protein in three different β-strands. While the side chains of
Ser99 and Thr153 are exposed, the side chain of Ala163 is buried.
It was found to fold with double exponential kinetics with rates
very similar to wild type, thus establishing that its enhanced fluo-
rescence in vivo is not a result of changes in structure or folding.
In a more detailed study of this mutant, Kuwajima and coworkers
demonstrated that although its unfolding and refolding kinetics
are very similar to the wild type, the cycle-3 mutant is much less
prone to aggregation [15]. The mutations, which all lie on the same
face of the β-barrel, reduce the overall hydrophobicity of GFP
and, thereby, suppress aggregation.
Figure 1. Mechanisms for the formation of and the structure of the p-hydroxy-benzylideneimidazolidinone chromophore of GFP.
NH O
N
OH
O
O
N
H
H
OH
ser65
tyr66 gly67
H2O
O2
H2O2
H2O2
O2
Path B
H2O
Path A
N
H
N
OH
OH
O
N
N
OH
OH
O
N
N
OH
O
N
N
OH
O

Folding of green fluorescent protein
www.future-drugs.com 547
In a recent publication, the Waldo group has used DNA shuf-
fling techniques to create a library of GFP and DsRed mutants
fused to a poorly folding bait protein, in this case, bullfrog red
cell H-subunit of ferritin, an insoluble protein when expressed
at 37°C [16]. Using this ‘folding interference method’, they went
through four rounds of selection and obtained a ‘superfolder’
GFP that has six additional mutations to the parent GFP – the
cycle 3 ‘folding reporter’ GFP. A comprehensive characterization
of this superfolder GFP demonstrated that it not only has
enhanced folding properties in vivo, but shows improved toler-
ance to circular permutation, chemical denaturants and faster
folding kinetics. In addition, this superfolder GFP was mutated
to create superfolder CFP, BFP and YFP. Using the same
method, a superfolder DsRed was also isolated but not charac-
terized as extensively. Each of the single mutations in super-
folder GFP was made in the parent fold-
ing reporter GFP to assess the effects on
stability and folding. The mutations were
found to have different effects,
Ser→Arg30 and Tyr→Asn39, folded
faster than the parent, and were more sta-
ble towards chemical denaturants. A crys-
tal structure showed that a reorganization
of side chains around these mutations
resulted in increased favorable electrostatic
interactions or hydrogen bonding net-
works. In contrast, Tyr→Phe145 and
Ile→Val171 had little effect on folding
rates or stability and presumably act by
reducing misfolding and aggregation. The
final mutations Asn→Thr105 and
Ala→Val206 did not show increased fold-
ing rates or stability, however, both muta-
tions increase the β-forming propensity of
the polypeptide chain.
The results from these four studies are
summarized in FIGURE 3, which shows the
position of the folding mutations in the
GFP structure, and in TABLE 1, which
summarizes the effect of the mutations.
Chromophore formation
In order to form the mature chromo-
phore, the polypeptide backbone must
undergo four distinct processes:
folding, cyclization, oxidation and
dehydration [17,18]. Peptide cyclization is
initiated by nucleophilic attack of the
Gly67 amide nitrogen on the Ser65 carbo-
nyl carbon, forming an imidazolone ring.
Dehydration of the Ser65 carbonyl oxygen
and dehydrogenation of the Thr66 Cα–Cβ
bond produces the fully conjugated,
p-hydroxybenzylidene-imidozolidinone
chromophore (FIGURE 1).
Initial studies by Reid and Flynn established that oxidation
is the rate-limiting step in chromophore formation and con-
firmed that it is an autocatalytic process [4]. Since then, many
studies have focused on the detailed mechanism of chromo-
phore formation, and the involvement of residues outside the
chromo-tripeptide. In particular, Arg96 and Glu222 have
been found to be involved in the autocatalysis of chromophore
formation [19,20].
Two mechanisms for chromophore formation have been pro-
posed. In the first, cyclization is followed by dehydration and
oxidation. In this mechanism, the heterocycle formed after
cyclization is stabilized by dehydration, and then dehydrogena-
tion of the Tyr66 Cα-Cβ occurs to form the fully conjugated
chromophore [18,21]. In the second mechanism, it has been pro-
posed that the oxidation step precedes the dehydration step,
Figure 2. Ribbon diagram of the structure of GFP as determined by x-ray crystallography [5,6].
The p-hydroxy-benzylideneimidazolidinone chromophore is located in the central α-helix and is
inaccessible to solvent.

Jackson, Craggs & Huang
548 Expert Rev. Proteomics 3(5), (2006)
This is supported by structural studies on the Y66L variant of GFP
wherein a trapped intermediate was observed in which cyclization
had occurred, and in which the hydroxyl leaving group remained
attached to the heterocyclic ring [22]. However, the α-carbon of res-
idue 66 was shown to be trigonal planar, consistent with ring oxi-
dation by molecular oxygen. Further evidence in support of this
mechanism was obtained from kinetic studies on chromophore
formation and the concomitant production of H2O2 [23]. In this
study, the Wachter group reported time constants for three kinetic
steps. The first step, involving folding and peptide cyclization pro-
ceeded with a time constant of 1.5 min. The second step, corre-
sponding to the oxidation, which was found to be rate limiting,
proceeded with a time constant of 34 min, whilst the final step
proceeded with a time constant of 11 min. Under highly aerobic
conditions, it was proposed that the dominant path to chromo-
phore formation follows the cyclization–oxidation–dehydration
mechanism. Both mechanisms may occur in parallel, the relative
flux being dependent on oxygen concentration and the efficiency
of ring dehydration for the particular GFP variant.
In contrast to the above mechanisms, one computational DFT
study suggested that oxidation could precede cyclization [24]. Get-
zoff and coworkers have interpreted these data slightly differently
and suggested that although the cyclization reaction appears
thermodynamically unfavorable (consistent with the relative ther-
modynamic stabilities calculated by DFT [24] it still occurs first
and is then trapped by the dehydration of the ring [21].
Both computational methods [25] and x-ray crystallography [21]
have shown that the central α-helix exhibits a dramatic approxi-
mately 80° bend during chromophore formation. The resultant
strained structure is proposed to raise the energy of the precy-
clized state closer to that of the cyclized intermediate, hence
reducing the activation energy. It also serves to position the
Gly67 nitrogen (the nucleophile) and the Ser65 carbonyl oxygen
in close contact, priming the cyclization step.
A recent computational study from the Zimmer group has pro-
posed that the cycle-3 mutations and the Ser→Pro147 mutant
exhibit increased fluorescence at room temperature due to the for-
mation of a tighter turn than wild type in the precyclized protein
around residues 65–67 [26]. Thus, these mutations may improve
the rate of chromophore maturation, in addition to reducing the
overall hydrophobicity, and hence aggregation propensity.
Preventing dimerization
Wild-type GFP is known to dimerize at high concentrations [6].
In their fluorescence resonance energy transfer (FRET) study of
lipid rafts, Zacharias and coworkers measured the homoaffinity
Figure 3. Topological map of green fluorescent protein showing the elements of secondary structure and the positions of the folding mutants.
S30R
Y39N
A206V
M153T
S147P
Y145F
I171V
S175G
F99S
N105T
N
C
V163A

Folding of green fluorescent protein
www.future-drugs.com 549
of YFP by sedimentation equilibrium analytical ultracentrifuga-
tion and found a Kd of 0.11 mM [27]. By replacing hydrophobic
residues at the crystallographic interface of the dimer with posi-
tively charged residues (A206K, L221K or F223R), they went
on to engineer mutants in which dimerization was essentially
eliminated. The A206K mutant was so extremely monomeric
in nature that it was difficult to determine an accurate dissocia-
tion constant for a hypothetical dimer. This mutation has now
been introduced into the full range of fluorescent proteins [2].
High-energy barriers & slow equilibrium
In their study of the cycle-3 variant, Fukuda and colleagues,
established that the unfolding and refolding of GFP was slow
and, as a consequence, the unfolding equilibrium is reached
over a period of days (rather than an hour or less for smaller
proteins) and the protein appears to be very stable with respect
to chemical denaturants [15]. These results have now been cor-
roborated by a number of different groups that have also shown
that GFP unfolds and refolds very slowly compared to small,
monomeric proteins [28]. FIGURES 5 & 6 shows the rate at which
the unfolding equilibrium is reached for GFP at different tem-
peratures and pH, as measured in our laboratory [29]. Even at
37°C, a true equilibrium is reached only after several weeks
(FIGURE 6). Careful analysis of these data to two- and three-state
models reveals that there is a stable intermediate state popu-
lated under equilibrium conditions [29]. The thermodynamic
parameters obtained from the analyses shows that the inter-
mediate state is compact compared to the denatured state; how-
ever, there has still been a significant increase in the solvent
accessible surface area on unfolding of the native to the inter-
mediate state. Although the intermediate state has very little
green fluorescence (approximately 10% of the native state, con-
sistent with the access of water to the chromophore) it is still
remarkably stable with respect to the denatured state, with a
free energy of unfolding of over 10 kcal/mol at 25°C, pH 6
The data are consistent with an intermediate state in which
β-strands 7–9 have unfolded and exposed the chromophore but
in which the rest of the β-barrel structure remains intact [29].
The fact that GFP reaches an unfolding equilibrium only
very slowly is indicative of high-energy barriers for both the
folding and unfolding reactions. The rate constants for folding
and unfolding have been measured by a number of different
groups and are consistently found to be small compared with
those measured for small, monomeric proteins [28].
Regan and coworkers have used the β-barrel structure of GFP
to study the effect of pairs of interacting residues across parallel
β-strands on stability and folding [30]. Positions 17 (β1) and
122 (β4) were mutated using library cassette mutagenesis meth-
ods, and a series of mutants produced and analyzed with differ-
ent cross-strand pairs. Unfolding and folding rate constants
were measured in vitro using pH-jump experiments and, under
the experimental conditions used, unfolding half-lives in the
order of 3 minutes and refolding half lives in the order of
1–8 min were observed. In addition to the in vitro measure-
ments, the rate of maturation of wild-type and mutant GFPs
were measured in vivo. Different rates for the maturation of
GFP were observed for the mutants in vivo and in vitro a ten-
fold range in folding rates was observed but differences in the
unfolding rates were undetectable. In this case, wild type was
found to have the highest folding rate and was the most fluo-
rescent in cells. The results established that there is a correla-
tion between folding rates measured in vitro and levels of
intracellular fluorescence.
In another study, GFP has been cyclized using intein technol-
ogy and the effects on unfolding and refolding measured [31].
The cyclic variant is identical in sequence and structure to the
linear parent GFP, except that the N- and C-terminals are cova-
lently linked together through a short region of peptide. Unfold-
ing half lives of approximately 0.1–0.6 min were measured
directly in high concentrations of the chemical denaturant gua-
nidinium chloride. These data were extrapolated to extract an
unfolding half life in water in the order of 3 min consistent with
the Regan results. In this case, two phases, fast and slow, were
identified in the refolding reactions with half lives of 1–2 and
40 min, respectively. The cyclized GFP was found to be more
stable than the linear parent GFP and unfold at approximately
half the rate.
Careful detection and analysis of the refolding of GFP as
probed by green fluorescence shows that there are more than
two kinetic phases. FIGURE 5 shows the results of a double
Table 1. Structural and folding parameters for GFP
folding mutants.
Mutation Position Properties Ref.
S175G Loop
exposed
Suppresses aggregation [11]
V163A β-strand 8
buried
No effect on folding rates,
reduces hydrophobicity.
[11,13]
S147P Loop
buried
Increased maturation rate [12,26]
M153T β-strand 7
exposed
No effect on folding rates,
reduces hydrophobicity.
[13]
F99S β-strand 4
exposed
No effect on folding rates,
reduces hydrophobicity.
[13]
S30R β-strand 2
exposed
Mutation stabilizes protein,
faster folding.
[16]
Y39N Loop
exposed
Mutation stabilizes protein,
faster folding.
[16]
N105T β-strand 5
exposed
No effect on stability or folding,
increased β propensity.
[16]
Y145F Loop
buried
No effect on stability or folding,
presumably reduces aggregation.
[16]
I171V β-strand 8
buried
No effect on stability or folding,
presumably reduces aggregation.
[16]
A206V β-strand 10
buried
No effect on stability or folding,
increased β− propensity.
[16]

Folding of green fluorescent protein
www.future-drugs.com 551
pH-jump experiment performed in our laboratory that shows
at least three distinct kinetic phases. Recently, Kuwajima and
coworkers have undertaken the most comprehensive study of
the folding of GFP to date [32]. Multiple probes including the
green fluorescence, tryptophan fluorescence and far-UV CD
were employed to reveal five folding phases including a rapid-
burst phase. Half-lives for these phases range from 20 ms to
6 min under the conditions used. A complex kinetic scheme
has been proposed by the Kuwajima group based on their
results (FIGURE 6) in which there is heterogeneity in the dena-
tured state due to proline isomerisation; there are several inter-
mediate states including the rapidly formed ‘burst-phase’ inter-
mediate, in which there is a nonspecific collapse of the
polypeptide chain, and an on-pathway intermediate with mol-
ten-globule-like properties; that there are at least two slow
phases which are limited by proline isomerisation. A structure
for the second intermediate state is proposed based on their
results and those from other groups: the intermediate is known
to be compact with significant secondary structure, but it does
not show green fluorescence or rigid tertiary structure. This late
intermediate identified from kinetic experiments appears to be
similar to the equilibrium intermediate observed in our
studies [29].
In a paper just published online, the Kuwajima group have
reported an equilibrium intermediate that is populated at
pH 4 [33]. This intermediate was shown to be the same as one
of the intermediates detected during refolding and through a
combination of fluorescence and small-angle x-ray scattering
(SAXS) experiments shown to have properties similar to a
molten-globule state.
Oligomeric fluorescent proteins
The folding behaviour of other GFP-like fluorescent proteins
(FPs) has also been investigated. DsRed is a FP from the coral
Discosoma and is a homotetramer of β-barrels whose fluores-
cence is red shifted compared to GFP. The acid denaturation
of DsRed was measured and partial renaturation achieved on
alkalization [34]. Several distinct states of the protein were
Figure 5. Kinetics of folding of GFP. Refolding kinetics of GFP monitored by green fluorescence in a stopped-flow apparatus. Refolding is initiated by a rapid
pH jump from pH 1.5 to 6.3. Craggs & Jackson, [UNPUBLISHED RESULTS]. Inset shows the faster phases observed between 0 – 25 s. Data fit to a triple
exponential process.
GFP: Green fluorescent protein.
0.5
0.0
-0.5
1.0
-1.5
-2.0
0 100 200 300 400 500 600
Fl
uo
re
sc
en
ce
Time (s)
0.5
0.0
-0.5
-1.0
-1.5
-2.0
0 5 10 15 20 25
Time (s)
Fl
uo
re
sc
en
ce

Jackson, Craggs & Huang
552 Expert Rev. Proteomics 3(5), (2006)
found during the unfolding and refolding processes corre-
sponding to different oligomeric states – monomer, dimer,
trimer and tetramer.
This work was followed up by a comparative study of the
folding and stability of five different FPs with varying oligo-
meric states and spectral properties [35]. Folding and unfolding
kinetics were measured in addition to stability measurements
using a wide range of probes. For all five proteins, the kinetics
were found to be slow in agreement with other studies and a
quasi-equilibrium produced. Despite some limitations, this
study clearly showed that there is a wide range in stabilities of
FPs, in general the higher-order oligomers being more stable.
However, this study also demonstrated that FPs with the same
oligomeric state can have very different stabilities.
Partially structured & denatured states of GFP
A pressure-induced unfolding study of red-shifted GFP
(rsGFP) has combined different optical probes of the native
state of the protein – fluorescence and absorbance measure-
ments probing tertiary structure whilst FT-IR was employed to
monitor changes in secondary structure [36]. Although very sta-
ble to high pressures at lower temperatures, and unfolding irre-
versibly at high temperatures and high pressures, conditions
were found where rsGFP was reversibly unfolded by pressure.
Two transitions were revealed by the different probes, the first
at about 4 kbar where changes associated with the tertiary
structure were observed and attributed to penetration of water
into the β-can structure, particularly in the region of the
chromophore. This creates a ‘swollen pretransitional’ state,
which has relatively small changes in tertiary structure and
which retains its secondary structure. The second transition,
observed at higher pressures (~8 kbar),
represents a global unfolding of the pro-
tein with a loss of secondary structure of
the β-barrel. Interestingly, the helical struc-
ture of GFP seems to be maintained in the
denatured state under these conditions.
In addition to the studies described
above, there is further evidence for resid-
ual structure in the denatured state of
GFP. In their analysis of the acid denatura-
tion and refolding of GFP, the Kuwajima
group found that there is significant sec-
ondary structure in the acid-unfolded state
as shown by far-UV circular dichroism
and this residual structure was shown to
be sensitive to salt [32]. We have investi-
gated this residual structure further using
19F-NMR spectroscopy in combination
with photochemically induced dynamic
nuclear polarisation (CIDNP) tech-
niques. The 19F spectrum of 19F-tyrosine-
labeled GFP is shown in FIGURE 8 along
with the full assignment. The assignment
was made by combining the results of
19F-NMR spectra of single Tyr→Phe mutations, relaxation
data and results from photo-CIDNP experiments [37]. The
19F-NMR spectrum of denatured GFP is shown in FIGURE 8 and
clearly shows two peaks, the larger corresponding to the
19F-labeled tyrosines which are all in a chemically identical
environment in the denatured state, the smaller peak corre-
sponding to the single 19F-labeled tyrosine, which has a chemi-
cally distinct environment as a result of its position within the
chromophore. Although there is little evidence for residual
structure from the 19F spectrum of the denatured state, the
results from photo-CIDNP experiments provide support for
the far-UV CD results. The photo-CIDNP experiments report
on the solvent exposure of the tyrosine side chains and FIGURE 8
shows the results for the native state of GFP. Four peaks are
observed corresponding to the four solvent exposed tyrosines.
Interestingly, a strong correlation is found between the solvent
accessibility of the highest occupied molecular orbital (HOMO)
of a given tyrosine, and its photo-CIDNP signal rather than the
solvent accessible surface area (SASA) of the residue. The SASA
data suggest that five, rather than four fluorotyrosine residues
should be polarizable in GFP (i.e., Tyr39, Tyr151, Tyr182,
Tyr200 and Tyr143). Although Tyr143 has a very similar overall
solvent accessibility to Tyr200, the HOMO accessibilities differ
by an order of magnitude. Inspection of the crystal structure
shows that the part of Tyr143 that protrudes into the solvent is
an unreactive CβH2-CαH-NH fragment that bears little
HOMO electron density, whereas it is the reactive aromatic side
chain that is exposed for Tyr200 [5,6].
The photo-CIDNP spectrum of the acid denatured state of
GFP shows two peaks consistent with the 19F spectrum of the
acid denatured state. However, in this case and in contrast to
Figure 6. Kinetics of folding of GFP. Kinetic scheme for the folding of green fluorescent protein as
proposed by Kuwajima and coworkers [32]. Dc and Dt are the denatured states with cis and trans proline
isomers, N is the native state, and I1c, I2c, I2’c, I1t, I2t, I2’t are the different intermediate states populated.
D
c
Dt I1t
I2t
I1c
I2c
I2’c
I2’t
N

Folding of green fluorescent protein
www.future-drugs.com 553
the results on the native state, the signs of the two peaks are
opposite. This provides strong evidence that the denatured state
is heterogeneous, containing subensembles with significantly
different rotational correlation times [37].
Dynamics
The dynamics of GFP have been measured using a variety of
experimental approaches. The dynamics associated with the
chromophore have been probed by fluorescence correlation spec-
troscopy (FCS) [38]. In this technique, time-resolved fluctuations
in fluorescence are used to report on the dynamic and thermody-
namic processes that affect the fluorescence
of the protein (in this case, enhanced GFP,
eGFP). Protonation of the hydroxyl group
of Tyr66 is shown to induce large changes
in absorption and emission spectra with a
pKa of 5.8. The autocorrelation function of
fluorescence emission shows contributions
from two chemical relaxation processes,
one pH dependent and the other pH
independent. The FCS data provide infor-
mation on the dynamics and equilibrium
properties of the protonation process.
GFP and its variants, like all known flu-
orescent proteins exhibit complex photo-
physical and photochemical behavior.
This interesting area falls outside the
scope of this review but the interested
reader is directed to [39] as a good starting
point for further reading.
15N-NMR relaxation measurements have
been used to study the dynamics of GFP on
a ps–ns timescale and have shown that
most of the β-barrel backbone is rigid on
these timescales [40]. H/D exchange tech-
niques were also employed to study confor-
mational dynamics on a μs–ms timescale.
The rates of exchange were found to vary
enormously and were assigned to four
classes – fast, intermediate, slow and very
slow. The slowest exchanging amide pro-
tons did not show significant levels of
exchange over the time course of the experi-
ment. These studies identified a region
comprising of β-strands 7, 8 and 10 that
show increased rates of exchange compared
to the rest of the protein and indicate that
this region has a higher degree of flexibility
in agreement with molecular dynamic sim-
ulations. The spectra of a mutant of GFP
(His-Gly148 located on β-strand 7 and
known to affect the chromophore) showed
interesting additional peaks showing that
this mutant is in slow exchange between
two conformations.
Single-molecule unfolding & folding
Single-molecule force spectroscopy has been used to investi-
gate the mechanical unfolding of GFP [41]. Here, GFP is sand-
wiched into a multidomain construct with either Ig8 or
DdFLN and the unfolding of the chimeric protein studied
with pulling Atomic Force Microscopy techniques. The results
suggest that GFP mechanically unfolds via two intermediate
states, the first is characterized by the detachment of the seven-
residue N-terminal α-helix to form a kinetically stable but
thermodynamically unstable state that retains the β-barrel
structure. The second metastable intermediate state has one
Figure 7. Results from the 19F NMR studies on GFP [37]. The structure of GFP is shown with the side
chains of the ten tyrosine residues shown in red. The chromophore is shown in space filling mode in the
centre of the β-barrel.

Jackson, Craggs & Huang
554 Expert Rev. Proteomics 3(5), (2006)
complete β-strand detached from the barrel. A schematic of
the free energy surface of GFP and the mechanical unfolding
pathway are shown in FIGURE 9.
Owing to its unique spectroscopic properties and high
quantum yield, GFP and its variants have been the subject of
many single-molecule experiments using optical fluorescence
techniques. In general, Yellow Fluorescent Proteins (YFPs),
created by the substitution of Thr203 to an aromatic amino
acid, have been used due to their optical
properties. They have been shown to
exhibit interesting photophysical proper-
ties, including on/off blinking [42] and
flickering [43]. The folding of one GFP
variant, the GFPmut2 construct, has
been investigated at the single-molecule
level by encapsulation in wet nanoporous
silica gels [44].
In collaboration with the group of
David Klenerman, our own group has
recently undertaken single-molecule
studies of variants of GFP under non-
equilibrium conditions where we can
monitor both unfolding and folding
reactions. A confocal microscope is used
in conjunction with novel nanopipette
technology [45] to observe both equilib-
rium behaviour and unfolding kinetics
of the YFP, Citrine [46] labeled with an
acceptor dye, Alexa 647, by single-pair
FRET (sp-FRET) and dual-colour sin-
gle molecule fluorescence coincidence
spectroscopy (sm-FCS). The citrine
mutant was chosen for its increased
photostability compared with other
YFPs [2].
Initial single-molecule kinetic unfold-
ing studies were conducted by injection
of native, labeled citrine into varying
concentrations of guanidinium chloride
(GdmCl) contained in a one millilitre
sample chamber. Using diffusion sp-
FRET, histograms of FRET efficiencies
were generated for each 14 min interval
over the time course of the reaction
(FIGURE 10). Two populations were
observed at each time point, with
FRET efficiencies of approximately
0.65 corresponding to folded, labeled
Citrine, and 0.00, made up of unla-
beled citrine and GdmCl impurities.
The unfolding rate constants were
obtained by plotting the number of
acceptor events against time and fitting
to a single exponential decay (FIGURE 11).
Alternatively, unfolding rate constants
were obtained by plotting the change in the Gaussian fit to
the labeled and unlabeled FRET peaks. The labeled and
unlabeled peaks decreased with the same rate suggesting that
the attachment of the dye has not significantly affected the
unfolding rate or the stability of the protein. The unfolding
rate was also monitored by sm-FCS, observing the decrease
in the number of coincident events (fluorescence above the
background count in both the Alexa and the Citrine channel
Figure 8. Results from the 19F NMR studies on GFP [37]. (A) 19F-NMR spectrum of labeled wild-type
GFP with the assignments shown. (B) Photo-CIDNP spectrum of labeled wild-type GFP denatured in 6 M
GdmCl. (C) Photo-CIDNP spectrum of labeled wildtype GFP denatured at pH 2.9 and
(D) Photo-CIDNP spectrum of labeled wild-type GFP in native buffer conditions.
CIDNP: Chemically induced dynamic nuclear polarisation; GFP: Green fluorescent protein.
-52 -54 -56 -58 -60 ppm
-52 -54 -56 -58 -60 ppm
-52 -54 -56 -58 -60 ppm
A
B
C
D
92
106 92 66 74
145 151
39
200
182
143
143

Folding of green fluorescent protein
www.future-drugs.com 555
in the same 1 ms bin). The unfolding rate constants from
the single-molecule experiments agreed well with results
from bulk solution study.
Expert commentary & five-year view
There is no doubt about the importance of GFP in current
biological and medical research: a quick literature search on
publications using GFP comes up with more than 9000 hits
from the year 2000 onwards. Its widespread use results from
the inherent, unique spectroscopic properties of GFP, in addi-
tion to the comprehensive engineering that has been per-
formed on the protein to modify and optimize optical, chemi-
cal and physical properties. An exhaustive review on the
protein is just not possible and, in this article, we have focused
on the folding properties of the protein. Folding is an essential
step in the maturation and use of GFP, but one that is some-
times limited by the competing reactions of misfolding and
aggregation. Described in the review are the best results
achieved so far to improve the folding properties of the pro-
tein. Remarkably, it has been 10 years since the publication of
the cycle 3 mutant in 1996 [13], and only recently has a new
GFP with improved folding been reported [16]. It is unclear
whether this is because it is inherently difficult to generate bet-
ter folding variants of GFP, or whether selection procedures are
limiting in the process.
The recent interest and work done in characterizing the
folding pathway of GFP in vitro, leads us into an exciting new
area – that of rationally designing mutants to aid in the folding
of GFP. Once a stable core or folding nucleus has been identi-
fied, then mutations can be designed which specifically stabi-
lize this nucleus and increase folding rates. As well as under-
standing more about the folding pathway, in particular,
characterizing the intermediate states in addition to the rate-
limiting transition state barriers, we also require a better
understanding of how and why this protein misfolds and
aggregates. Mutations can then be designed which not only aid
folding but also suppress aggregation. Combinations of both
types of mutation should be particularly effective in creating a
new ultrafolding GFP.
Information resources
Recent review articles
For anyone confused as to which of the many available fluores-
cent proteins to use in their experiments, a recent review by
Shaner, Steinbach and Tsien, is an excellent guide [2]. The review
discusses the different factors to be considered when choosing an
FP such as spectral properties, brightness, expression, toxicity,
photostability, oligomerisation and sensitivity to environmental
effects, in addition to summarizing this data for a large number
of FPs. A slightly earlier review from the Tsien laboratory also
contains a section on the design and construction of FP-based
fluorescent reporting systems [47]. In addition, this review
describes applications of FPs as ‘passive’ markers of protein
expression and localisation, and as ‘active’ indicators of small-
molecule messenger dynamics, enzyme activation and pro-
tein–protein interactions. For a more specific review on using
mutants of GFP to monitor protein conformations and interac-
tions by fluorescence resonance energy transfer see review by
Miyawaki and Tsien [48]. It must be stressed that there are now
numerous reviews on the applications of GFP in biology. The
three reviews cited are just a starting point for interested readers.
In addition to review articles, there is a great deal of informa-
tion also available on GFP on both academic and commercial
websites. Highly recommended is Roger Tsien’s website which
contains a complete list of publications from the lab, as well as
images, movies, discussion documents and links.
 www.tsienlab.ucsd.edu/
For those interested in single-molecule studies of GFP, the
website of William Moerner is recommended:
Figure 9. Free-energy surface for the mechanical unfolding of GFP by atomic force microscopy. Taken from [23].
Fr
ee
e
ne
rg
y
[k
B
I]
A B
22
<14
23 20–25
>3.7
3.2 0.28 6.5
0.55 69.3
End-to-end distance (nm)
GFP GFPΔα GFPΔαΔβ
Reaction coordinate
Fr
e
e
e
n
e
rg
y

Jackson, Craggs & Huang
556 Expert Rev. Proteomics 3(5), (2006)
 www.stanford.edu/group/moerner/
 Information on commercially available FPs can be obtained
from the Clontech website, in the Living Colours Fluorescent
Proteins section.
www.clontech.com/clontech/
Recommended books on GFP include:
 GFP, Properties, Applications and Protocols 2nd Ed. Chalfie
and Kain, John Wiley & Sons Inc
 Glowing Genes: A Revolution in Biotechnology (2005) Marc
Zimmer, Prometheus Books
 Aglow in the Dark (2005) Vincent Pieribone & David Gru-
ber, Belknapp Press of Harvard University Press, Cambridge,
Massachusetts, USA and London, England
Figure 10. Typical single molecule histograms of YFP unfolding. 50 pM YFP in 4 M GdmCl. Data were acquired at 14 min intervals over 280 min at 25°C. The
left- and right-hand lines are lines through the centre of the two Gaussian peaks.
0
280 min
0
196 min
0
112 min
0
56 min
0
28 min
0
14 min
400
200
500
250
800
400
1200
600
1500
750
2000
1000
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
FRET
N
um
be
r o
f m
ol
ec
ul
es

Folding of green fluorescent protein
www.future-drugs.com 557
Figure 11. A single molecule unfolding time course by monitoring the number of acceptor events (>25 counts) of 50 pM YFP in 4M GdmCl.
Data were fitted to a first-order equation (line of best fit).
Key issues
 Protein engineering techniques and selection methods used to generate variants of GFP with improved folding properties.
 Mechanism of chromophore formation.
 Mutations that reduce the tendency of GFP to dimerize.
 Complex kinetic mechanism for folding involving multiple intermediate states and parallel pathways.
 Partially structured states of GFP and residual structure in the denatured state.
 Chromophore and backbone dynamics
 Single-molecule folding and unfolding studies.
N
um
be
r o
f A
le
x
a-
64
7
ev
en
ts
6000
Time (x103 s)
5000
4000
3000
2000
1000
0
2 4 6 8 10 12 14 16 180
References
Papers of special note have been highlighted as:
 of interest
 of considerable interest
1 Tsien RY. The green fluorescent protein.
Ann. Rev. Biochem. 67, 509–544 (1998).
2 Shaner NC, Steinbach PA, Tsien RY.
A guide to choosing fluorescent proteins.
Nat. Methods 2, 905–916 (2005).
• Very useful practical guide to selecting
fluorescent proteins for use in a wide range
of applications.
3 Zimmer M. Green fluorescent protein
(GFP): applications structure, and related
photophysical behavior. Chem. Rev. 102,
759–781 (2002).
4 Reid BG, Flynn GC. Chromophore
formation in green fluorescent protein.
Biochemistry 36, 6786–6791 (1997).
5 Ormo M, Cubitt AB, Kallio K, Gross LA,
Tsien RY, Remington SJ. Crystal structure
of the Aequorea victoria green fluorescent
protein. Science 273, 1392–1395 (1996).
6 Yang F, Moss L, Phillips G. The molecular
structure of green fluorescent protein.
Nat. Biotech. 14, 1246–1251 (1996).
7 Waldo GS, Standish BM, Berendzen J,
Terwillger TC. Rapid protein folding assay
using green fluorescent protein.
Nat. Biotechnol. 17, 691–695 (1999).

Jackson, Craggs & Huang
558 Expert Rev. Proteomics 3(5), (2006)
8 Baird GS, Zacharias DA, Tsien RY. Circular
permutation and receptor insertion within
green fluorescent proteins. Proc. Natl
Acad. Sci. USA 96, 11241–11246 (1999).
9 Topell S, Hennecke J, Glockshuber R.
Circularly permuted variants of the green
fluorescent protein. FEBS Lett. 457,
283–289 (1999).
10 Topell S, Glockshuber R. Circular
permutation of the green fluorescent
protein. Methods Mol. Biol. 183, 31–48
(2002).
11 Siemering K. R, Golbik R, Sever R,
Haseloff J. Mutations that suppress the
thermosensitivity of green fluorescent
protein. Current Biol. 6, 1653–1663 (1996).
12 Kimata Y, Iwaki M, Lim CR, Kohno K.
A novel mutation which enhances the
fluorescence of GFP at high temperatures.
Biochem. Biophys. Res. Comm. 232, 69–73
(1997).
13 Crameri A, Whitehorn EA, Tate E, Stemmer
WP. C. Improved green fluorescent protein
by molecular evolution using DNA
shuffling. Nat. Biotech. 14, 315–319 (1996).
14 Battistutta R, Negro A, Zanotti G. Crystal
structure and refolding properties of the
mutant F99S/M153T/V163A of the Green
Fluorescent Protein. Proteins: Struct. Func.
Gen. 41, 429–437 (2000).
 The structure and folding properties of the
widely used cycle 3 variant.
15 Fukuda H, Arai M, Kuwajima K. Folding of
green fluorescent protein and the cycle 3
mutant. Biochemistry 39, 12025–12032
(2000).
16 Pedelacq JD, Cabantous S, Tran T,
Terwilliger TC, Waldo GS. Engineering and
characterization of a superfolder green
fluorescent protein. Nat. Biotech. 24, 79–88
(2006).
 Describes very recent studies on the
selection of a superfolding variant of GFP
using a folding interference method
for selection.
17 Heim R, Prasher DC, Tsien RY. Wavelength
mutations and posttranslational
autoxidation of green fluorescent protein.
Proc. Natl Acad. Sci. USA 91, 12501–12504
(1994).
18 Cubitt AB, Heim R, Adams SR, Boyd AE,
Gross LA, Tsien RY. Understanding,
improving and using green fluorescent
proteins. Trends Biochem. Sci. 20, 448–455
(1995).
19 Wood TI, Barondeau DP, Hitomi C,
Kassmann CJ, Tainer JA, Getzoff ED.
Defining the role of arginine 96 in Green
fluorescent protein fluorophore biosynthesis.
Biochemistry 44, 16211–16220 (2005).
20 Sniegowski JA, Phail ME, Wachter RM.
Maturation efficiency, trypsin sensitivity, and
optical properties of Arg96, Glu222, and
Gly67 variants of green fluorescent protein.
Biochem. Biophys. Res. Comm. 332, 657–663
(2005).
21 Barondeau DP, Putnam CD, Kassmann CJ,
Tainer JA, Getzoff ED. Mechanism and
energetics of green fluorescent protein
chromophore synthesis revealed by trapped
intermediate structures. Proc. Natl Acad. Sci.
USA 100, 12111–12116 (2003).
 Here, mutations which substantially slow
the rate but not the yield of post-
translationally modified protein are used
and the structure of the trapped
precyclisation intermediate and oxidised
post-translational state reported.
22 Rosenow MA, Huffman HA, Phail ME,
Wachter RM. The crystal structure of the
Y66L variant of green fluorescent protein
Supports a cyclization-oxidation-dehydration
mechanism for chromophore maturation.
Biochemistry 43, 4464–4472 (2004).
23 Zhang L, Patel HN, Lappe JW, Wachter RM.
Reaction Progress of Chromophore
Biogenesis in Green Fluorescent Protein.
J. Am. Chem. Soc. 128, 4766–4772 (2006).
 This study uses the kinetics of hydrogen
peroxide evolution during the maturation of
GFP in vitro to inform on the mechanism of
formation. Under highly aerobic conditions,
the dominant pathway is found to involve a
cyclisation then oxidation followed
by dehydration.
24 Siegbahn PEM, Wirstam M, Zimmer M.
Theoretical study of the mechanism of
peptide ring formation in green fluorescent
protein. Int. J. Quantum Chem. 81, 169–186
(2001).
25 Branchini BR, Nemser AR, Zimmer M.
A computational analysis of the unique
protein-induced tight turn that results in
posttranslational chromophore formation in
green fluorescent protein. J. Am. Chem. Soc.
120, 1–6 (1998).
26 Baffour-Awuah NY, Fedeles F, Zimmer M.
Structural features responsible for GFPuv and
S147P-GFP’s improved fluorescence. Chem.
Phys. 310, 25–31 (2005).
27 Zacharias DA, Violin JD, Newton AC, Tsien
RY. Partitioning of lipid-modified
monomeric gfps into membrane
microdomains of live cells. Science 296,
913–916 (2002).
28 Jackson SE. How do small single-domain
proteins fold? Fold. Des. 3, 81–91 (1998).
29 Huang J-R, Craggs TD, Christodoulou J,
Jackson SE. Stable intermediate states are
formed in the equilibrium unfolding of
GFP. Protein Science submitted (2006).
30 Merkel JS, Regan L. Modulating Protein
Folding rates in vivo and in vitro by
side-chain interactions between the parallel
strands of green fluorescent protein.
J. Biol. Chem. 275, 29200–29206 (2000).
31 Iwai H, Lingel A, Pluckthun A. Cyclic green
fluorescent protein produced in vivo using
an artificially split PI-PfuI intein from
Pyrococcus furiosus. J. Biol. Chem. 276,
16548–16554 (2001).
32 Enoki S, Saeki K, Maki K, Kuwajima K.
Acid denaturation and refolding of green
fluorescent protein. Biochemistry 43,
14238–14248 (2004).
 This study was the first to report detailed
kinetic experiments on the folding of GFP
and detect multiple intermediate states and
a parallel pathway resulting from
proline isomerisation.
33 Enoki S, Maki K, inobe T, Takahashi K
et al. The equilibrium unfolding
intermediate observed at pH 4 and its
relationship with the kinetic folding
intermediates in green fluorescent protein.
J. Mol. Biol. 361(5), 969–982 (2006).
34 Vrzheshch PV, Akovbian NA, Varfolomeyev
SD, Verkhusha VV. Denaturation and
partial renaturation of a tightly tetramized
DsRed protein under mildly acidic
conditions. FEBS Lett. 487, 203–208
(2000).
35 Stepanenko OV, Verkhusha VV, Kazakov VI
et al. Comparative studies on the structure
and stability of fluorescent proteins EGFP,
zFP506, mRFP1, "dimer2", and DsRed1.
Biochemistry 43, 14913–14923 (2004).
36 Herberhold H, Marchal S, Lange R,
Scheyhing CH, Vogel RF, Winter R.
Characterization of the pressure -induced
intermediate and unfolded state of red-
shifted green fluorescent protein - a static
and kinetic FTIR, UV/VIS and fluorescence
spectroscopy study. J. Mol. Biol. 330,
1153–1164 (2003).
37 Khan F, Kuprov I, Craggs TD, Hore PJ,
Jackson SE. 19F NMR studies of the native
and denatured states of green fluorescent
protein. J. Am. Chem. Soc. 128,
10729–10737 (2006).
38 Haupt U, Maiti S, Schwille P, Webb WW.
Dynamics of fluorescence fluctuations in
green fluorescent protein observed by
fluorescence correlation spectroscopy.
Proc. Natl Acad. Sci. USA 95, 13573–13578
(1998).
39 Bonsma S, Purchase R, Jezowski S, Gallus J,
Konz F, Volke RS. Green and red fluorescent
proteins: photo- and thermally induced
dynamics probed by site-selective
spectroscopy and hole burning.
Chem. Phys. Chem. 6, 838–849 (2005).

Folding of green fluorescent protein
www.future-drugs.com 559
40 Seifert MH. J, Georgescu J, Ksiazek D et al.
Backbone dynamics of green fluorescent
protein and the effect of histidine 148
substitution. Biochemistry 42, 2500–2512
(2003).
41 Dietz H, Rief M. Exploring the energy
landscape of GFP by single-molecule
mechanical experiments. Proc. Natl Acad.
Sci. USA 101, 16192–16197 (2004).
42 Dickson RM, Cubitt AB, Tsien RY,
Moerner WE. On/off blinking and
switching behaviour of single molecules of
green fluorescent protein. Nature 388,
355–258 (1997).
43 Schwille P, Kummer S, Heikal AA,
Moerner WE, Webb WW. Dynamics of
fluorescence fluctuations in green fluorescent
protein observed by fluorescence correlation
spectroscopy. Proc. Natl Acad. Sci. USA 97,
151–156 (2000).
44 Cannone F, Bologna S, Campanini B,
Diaspro A, Bettati S, Mozzarelli A,
Chirico G. Tracking unfolding and refolding
of single GFPmut2 molecules. Biophys. J. 89,
2033–2045 (2005).
45 White SS, Balasubramanian S,
Klenerman D, Ying LM. A Simple
nanomixer for single molecule kinetics.
Angewante Chem. (2006) (In Press).
46 Heikal AA, Hess ST, Baird GS, Tsien RY,
Webb WW. Molecular spectroscopy and
dynamics of intrinsically fluorescent
proteins: Coral red (dsRed) and yellow
(Citrine). Proc. Natl Acad. Sci. USA
97, 11996–12001 (2000).
47 Zhang J, Campbell RE, Ting A, Tsien RY.
Creating new fluorescent probes for cell
biology. Nat. Rev. Mol. Biol. 3, 906–918
(2002).
48 Miyawaki A, Tsien RY. Monitoring protein
conformations and interactions by
fluorescence resonance energy transfer
between mutants of green fluorescent
protein. Methods Enzymol. 327 472–500
(2000).
Affiliations
 Sophie E Jackson
Chemistry Department, Lensfield Road,
Cambridge CB2 1EW, UK
Tel.: +44 1223 762 011
Fax: +44 1223 336 362
sej13@cam.ac.uk
 Timothy D Craggs
Chemistry Department, Lensfield Road,
Cambridge CB2 1EW, UK
Tel.: +44 1223 767 042
Fax: +44 1223 336 362
 Jie-rong Huang
Chemistry Department, Lensfield Road,
Cambridge CB2 1EW, UK
Tel.: +44 1223 336 357
Fax: +44 1223 336 362