Chemical Physics 275 (2002) 109–121
www.elsevier.com/locate/chemphys
Red-shifted mutants of green fluorescent protein:
reversible photoconversions studied by hole-burning and
high-resolution spectroscopy
€lker a,*
T.M.H. Creemers a,1, A.J. Lock a,2, V. Subramaniam b, T.M. Jovin b, S. Vo
a
Huygens and Gorlaeus Laboratories, Center for the Study of Excited States of Molecules,
Leiden University, P.O. Box 9504, 2300 RA Leiden, Netherlands
b
Department of Molecular Biology, Max-Planck Institute for Biophysical Chemistry,
Am Fassberg 11, D-37077 G€ottingen, Germany
Received 30 July 2001; in final form 11 September 2001
Abstract
Mutants of green fluorescent protein (GFP) are usually designed to absorb and emit light as ‘‘one color’’ systems, i.e.
with a single, photostable conformation of the chromophore. We have studied two red-shifted GFP-mutants (S65T and
EYFP) by means of hole-burning and high-resolution optical spectroscopy at low temperature, and compare the data to
those previously reported for RS-GFP. We prove that these GFP-mutants are not ‘‘one color’’ systems because they can
be reversibly phototransformed from one conformation into another. The results are rationalized in terms of energylevel schemes that are similar to that previously derived by us for wild-type GFP. In these schemes, each mutant can be
interconverted by light among at least three conformations that are associated with the protonation-state of the
chromophore. The results have relevance for the study of protein–protein interactions by fluorescence resonance energy
transfer (FRET), where GFP-mutants of different colors are used as labels in donor–acceptor pairs. Furthermore, we
present a detailed mechanism that explains the ‘‘on–off’’ and ‘‘blinking’’ behavior of single GFP-molecules with the
proposed energy-level diagrams. Ó 2002 Elsevier Science B.V. All rights reserved.
1. Introduction
*
Corresponding author. Tel.: +71-527-5859; fax: +71-5275819.
E-mail address: silvia@molphys.leidenuniv.nl (S. V€olker).
1
Present address: Philips Research Laboratory, 5656 AA
Eindhoven, The Netherlands.
2
Present address: FOM Institute for Atomic and Molecular
Physics (AMOLF), 1098 SJ Amsterdam, The Netherlands.
Autofluorescent proteins, such as green (GFP)
[1–3] and red (DsRed) [4–7] fluorescent proteins are
important in molecular and cell biology as markers
for gene expression and protein localization, and
for visualization of dynamic events inside living
cells, like the study of protein–protein interactions
by fluorescence resonance energy transfer (FRET)
[8–11]. GFP from the jellyfish Aequorea victoria
was the first protein discovered to possess an
0301-0104/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 1 - 0 1 0 4 ( 0 1 ) 0 0 5 3 7 - 7
110
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
intrinsic, bright fluorescence without the need of
external cofactors [1,12]. After it was cloned in 1992
[12] and expressed in another organism in 1994 [13],
its crystal structure was determined with a resolu in 1996 [14,15]. Based on the crystal
tion of 1.9 A
structure, many mutants have been designed with
the aim of improving the optical properties of wildtype(wt)-GFP (e.g. photostability and fluorescence
quantum yield) and to expand its spectral range for
multicolor labeling [1–3,16–18]. Neither the denatured GFP nor the GFP-chromophore by itself
emits fluorescence. The DNA sequence coding for
GFP can be fused to the gene of a desired target
protein for which the expression or cellular location is of interest. The expressed GFP folds independently of the protein to which it is fused
and acts as an in vivo fluorescence label.
GFP is a small, single-chain protein consisting
of 238 amino acids organized in eleven anti-parallel -strands forming a compact, cylindrical
barrel-like structure [12,14,15]. Inside the barrel,
an a-helix contains the sequence from which the
chromophore is derived. The chromophore, a
p-hydroxybenzylideneimidazolinone [1,16,19], originates from an autocatalytic cyclization and
subsequent oxidation of three amino-acid residues
in the chain: serine (Ser)-65, tyrosine (Tyr)-66 and
glycine (Gly)-67 [16,20]. It is completely protected
from the bulk solvent and rigidly held within the
barrel, forming a fluorescent p–p system.
The absorption spectrum of wild-type GFP (wtGFP) at room temperature is characterized by two
maxima that have been associated with different
protonation-states of the chromophore: a neutral
A-form absorbing at 398 nm and an anionic or
deprotonated B-form absorbing at 478 nm
[1–3,16,21]. From previous studies at room temperature and at 77 K it was concluded that the
interconversion from the A- to the B-form should
occur through an intermediate-state I involving a
very fast excited-state proton transfer reaction,
followed by a slow solvation process [22,23]. In
[22] it has also been proposed that I is responsible
for the strong, green fluorescence at 508 nm
[2,3,21]. However, the intermediate-state I had
only been conjectured but not observed, and the
details of the photoconversion remained unclear
[22,23].
Recently, we identified the I-form of wt-GFP by
means of high-resolution spectroscopy at liquidhelium temperature [24]. In that study we also
located the 0–0 transitions of the three A-, B- and
I-forms by hole-burning and determined vibrational frequencies of their ground- and excitedstates. In addition, we unraveled the pathways of
photoconversion between these three forms. The
results were summarized in an energy-level scheme
[24]. Compared to room-temperature spectroscopy, laser spectroscopy at low temperature has
the advantage that many of the thermally induced
conversions are blocked, and the discrimination of
individual species is made easier. In addition, at
low temperature the spectra become more structured and highly resolved vibrational information
is obtained. Furthermore, energy-level schemes
derived from low-temperature experiments set
limits to the interpretation of room temperature
results.
An understanding of the photophysics and
conformational interconversions occurring in GFP
and its mutants is crucial for their utilization as in
vivo labels. GFPs are also challenging from a
theoretical chemistry point of view because the
origin of the intrinsic fluorescence is still unclear
[25–27].
Particularly interesting for FRET experiments
are ‘‘one color’’ GFP-mutants (i.e. with only one,
photostable conformation), in contrast to wtGFP, which exhibits at least three conformations.
The various mutants, when used for FRET, should
absorb and emit light at different wavelengths.
Most of the successful GFP-mutations reported in
the literature involve a change of one or more
amino-acid residues within, or close to the chromophore [1,3,18,20]. For example, a mutation of
tyrosine (Tyr)-66 to histidine (His) or tryptophan
(Trp) leads to a blue shift of the spectra with respect to wt-GFP [16], whereas mutations on serine
(Ser)-65 result in a red shift [17,20].
Here we present new results on the photophysics
of two red-shifted mutants of GFP: S65T and
EYFP, and compare these to the data previously
reported by us for RS-GFP [28]. All data were
obtained by high-resolution optical spectroscopy
at liquid-helium temperature. In the S65T mutant,
Ser-65 is replaced by threonine (Thr) [1,15,17,20],
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
the latter having an extra methyl group, which results in a major absoption peak at 490 nm (Fig. 1).
This peak is shifted to the red of that of the B-form
(472 nm) of wt-GFP (Fig. 1). More than one mutation in the vicinity of the chromophore often
leads to a shift of the absorption maximum further
to the red. For example, RS-GFP has a triple
mutation at positions 64, 65 and 69 [1,3,17] and
absorbs at 495 nm (Fig. 1), whereas EYFP absorbing at 520 nm (Fig. 1) has four amino acids
replaced at positions 65, 68, 72 and 203 [1,15,29].
These red-shifted mutants, which appear to be
lacking the protonated A-form (at 400 nm), have
simpler spectra than wt-GFP (Section 3.1). The
maxima of S65T and RS-GFP (Fig. 1) are very
close to the maximum of the I-form of wt-GFP
111
(495 nm) [24]. This similarity challenged us to investigate whether these mutants are in the B-form
as proposed for S65T [18,30], or in the I-form as
suggested by our own results on wt-GFP [24].
In contrast to current views, we demonstrate
that the red-shifted mutants S65T, RS-GFP and
EYFP are not photostable and can be reversibly
phototransformed between various conformations
(Section 3.2). Our results are relevant for studies of
protein–protein interactions by means of FRET
experiments in which pairs of GFP-mutants are
fused as donors and acceptors to the proteins of
interest. The energy-level schemes that result from
our experiments (Section 3.3) yield an interpretation for the ‘‘on–off’’ and ‘‘blinking’’ behaviour
observed in single molecules of GFP-mutants at
room temperature (Section 3.4) [29,31,32].
2. Experimental
2.1. Samples
Expression plasmids for S65T and RS-GFP
were gifts of Dr. R. Rivera-Pomar (MPIbpc,
G€
ottingen, Germany), while that for EYFP was a
gift of Dr. D. Piston (Vanderbilt University,
Nashville, TN, USA). The mutants carry the following changes in the protein sequence: S65T (Ser
65!Thr), RS-GFP (Phe 64!Met, Ser 65!Gly,
Glu 69!Leu) and EYFP (Ser 65!Gly, Val
68!Leu, Ser 72!Ala and Thr 203!Tyr). The
recombinant proteins with a 6-histidine tag at the
amino terminus were expressed and purified on a
Ni-chelating resin (Ni-NTA-Agarose, Qiagen,
Hilden, Germany) using standard procedures [33].
All proteins were dissolved in 10 mM Na-phosphate buffer, pH 7, containing 50% (v/v) spectroscopic grade glycerol. Protein concentrations as
determined from the respective chromophore extinction coefficients [34] were 20 lM (S65T and
RS-GFP) or 5 lM (EYFP).
Fig. 1. Absorption spectra of wild-type GFP and the red-shifted mutants S65T, RS-GFP and EYFP at 1.6 K. The maxima
of the absorption bands are indicated. The red-shifted mutants
are characterized by a strong band around 500 nm and by the
absence of the A-band at 400 nm, present in wt-GFP (taken
from [28]).
2.2. Hole-burning and high-resolution optical spectroscopy
All spectroscopic measurements were performed at 1.6 K in Leiden. The samples were
112
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
placed in a cuvette (thickness 3 mm) and introduced into a 4 He-bath cryostat that was filled with
liquid N2 . After a few minutes, the liquid N2 was
blown out, the cryostat filled with liquid He and,
subsequently, pumped down to 1.6 K.
The samples were excited with a dye-laser
(Molectron DL 200, bandwidth 1 cm1 ) pumped
by a pulsed N2 -laser (Molectron UV 22) [24].
Continuous tunability between 355 and 535 nm
was achieved using 10 dyes. Absorption spectra
were obtained by scanning the laser and detecting
the transmission through the sample with a
photomultiplier (type EMI9658B) and an electrometer (Keithley 610CR). Excitation-, emissionand hole-burning (resolution 1 cm1 ) spectra
were detected in fluorescence at 90° with respect to
the excitation beam through a 0.85 m double
monochromator (SPEX 1402, resolution 5 cm1 )
with the same photomultiplier and electrometer.
3. Results and discussion
In this paper, we present a detailed hole-burning and high-resolution spectroscopy study at low
temperature of the red-shifted GFP-mutants S65T
and EYFP and a comparison to RS-GFP [28]. The
results indicate that these three mutants act similarly in many respects. In order to stress the similarities and differences among these mutants, the
spectra of S65T and EYFP are presented in conjunction with those of RS-GFP, even though most
of the data for the latter have been reported elsewhere [28]. We compare the results to those previously obtained by us for wt-GFP [24].
3.1. Absorption spectra at 1.6 K
Fig. 1 shows the absorption spectra at 1.6 K of
wt-GFP and the three red-shifted mutants. They
are significantly more structured than at room
temperature [22–24]. The two strong bands observed in the spectrum of wt-GFP at 407 and 472
nm are attributed (see above) to the neutral or
protonated A-form and the anionic or deprotonated B-form of the GFP-chromophore. By combining room- and low temperature spectroscopy
with hole-burning, we recently discovered the
I-form of wt-GFP as the intermediate and located
its 0–0 transition at 495 1 nm [24]. We also
found that in thermodynamic equilibrium (i.e., in a
previously non-illuminated sample) the intermediate I-form is present at room temperature but
absent at low temperature. This is illustrated in
Fig. 1 (top), where only the A- and the B-forms are
present at 1.6 K. The I-form, however, can be
photoinduced at low temperature by exciting either the A- or the B-form with a laser. In the same
study, we identified the 0–0 transition of A at
434 1 nm and that of B at 477 1 nm and found
that the photoreactions between A and I, and between I and B are reversible, i.e. A $ I $ B. Since
we did not observe a direct interconversion between A and B, we concluded that I is a real intermediate species [24].
All three red-shifted GFP-mutants have a
strong absorption maximum at their red edge and
weaker bands towards the blue (Fig. 1). The
spectra appear to lack the A-form (at 400 nm)
characteristic for wt-GFP. As noted above, it is
interesting that S65T and RS-GFP have their
maxima in the same spectral region as the photoinduced I-form of wt-GFP (495 nm) [24]. This
similarity suggests that the mutants are in a conformation corresponding to the I-form of wt-GFP
and not to the B-form, as claimed in a study of
S65T [18,30] (see below). The absorption spectrum
of S65T at 1.6 K shows relatively strong bands on
the blue side of its main peak (490 nm) as compared to the spectra of EYFP and RS-GFP. The
latter two are similar, although shifted with respect
to each other (see Fig. 1). We prove here that the
bands in the spectrum of S65T correspond mainly
to other conformations of the chromophore and,
to a lesser extent, to vibrational bands of the
conformation absorbing at 490 nm.
3.2. Excitation-, emission- and hole-burning spectra
3.2.1. S65T
A series of excitation- and fluorescence spectra
at 1.6 K of a previously non-illuminated S65T
sample are depicted in Fig. 2. The positions and
shapes of the spectra depend markedly on the
wavelengths of excitation ðkexc Þ and detection ðkem ),
indicating the presence of more than one form or
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
113
(a)
(b)
(c)
Fig. 2. S65T. Left: Excitation- and emission spectra at 1.6 K. Their intersections, which depend on the excitation- and detection
wavelength, determine the rough positions of the 0–0 transitions of the three forms, I (a), B (b) and A (c). The approximate vibrational
frequencies (within 50–100 cm1 ) corresponding to the bands shown in the spectra (averaged over the ground- and excited-states) for
the three forms are: (a) I-form: 205, 815 and 1520 cm1 , (b) B-form: 260 and 1500 cm1 , (c) A-form: 270 and 1780 cm1 . Right:
Excitation spectra before (––––) and after (- - - - - -) burning into the 0–0 transition of a given form. (a) Hole burnt into I0–0 at 495 1
nm; the I-form photoconverts into A and B. (b) Hole burnt into B0–0 at 478 1 nm; B transforms into I. (c) Hole burnt into A0–0 at
429 1 nm; A transforms into I. Thus, A $ I $ B (see also Fig. 6(a)).
114
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
species in S65T. By varying kexc and kem , three pairs
of spectra are observed (left-hand column of Fig. 2)
displaying three intersections which we ascribe to
the 0–0 transition regions of three forms (I, B and
A, from top to bottom). This behavior is similar to
that previously observed for wt-GFP [24]. All three
forms are already populated at T ¼ 1:6 K, because
they fluoresce independently when excited individually (Fig. 2, left). From the wavelengths of the
intersections we obtain a rough estimate of the
spectral positions of the 0–0 transitions, which by
hole-burning we could accurately locate: I0–0 at
495 1 nm, B0–0 at 478 1 nm and A0–0 at 429 1
nm. The burnt holes are narrow (Fig. 2, right), with
a width that is not given by the homogeneous
linewidth [35] but equal to twice the laser bandwidth of 1 cm1 . In general, holes burnt in vibronic bands are difficult to detect because they are
very broad (a factor of 103 compared to holes in
0–0 transitions) due to the short (picoseconds)
vibronic-state lifetimes [35].
If the assignments of the three I-, B- and Aforms in S65T are correct, we expect the photoconversions I ! A and I ! B to occur. This is
indeed observed (Fig. 2(a), right): when burning a
hole into I0–0 , both the red wing and the vibronic
bands of I decrease in intensity. Simultaneously,
the intensities of the spectral regions assigned to
the A- and B-forms increase slightly. If a hole is
subsequently burnt into B0–0 (Fig. 2(b), right), the
vibronic bands of B decrease and the excitation
spectrum of I increases (not shown), proving that
B is converted into I and vice versa. Similarly,
when a hole is burnt into A0–0 (Fig. 2(c), right), the
intensity of the total A-band slightly decreases and
the intensity of I increases (not shown). We conclude that the I- and A-forms, and the I- and Bforms reversibly photoconvert, i.e. I $ A and
I $ B. These reactions occur through radiationless pathways (probably the triplet-state) during
decay from the excited-state of one form to the
ground-state of the other, for example, I ! A and
A ! I (see also Fig. 6(a)). Photoconversions do
not seem to take place in the excited singlet-state
because all three forms fluoresce when excited individually (Figs. 2(a)–(c), left).
We conclude that the I-form is the predominantly populated conformation in S65T at low
temperature. This is in contrast to the result obtained for wt-GFP, where the A- and the B-forms
(Fig. 1, top) are the species populated at 1.6 K,
not the I-form. This has significant consequences
when trying to relate the conformations with the
structures reported in the literature [14,15]. By
similarity of wavelengths and spectral shapes, we
associate the structure reported for S65T absorbing at 490 nm [15,30] with the I-form
(I0–0 ¼ 495 1 nm) and not with the B-form
(B0–0 ¼ 478 1 nm) of wt-GFP as suggested in
[18,30]. The spectral position of A0–0 at 429 1
nm is also close to that found for A0–0 in wtGFP (434 nm) [24]. The approximate vibrational
frequencies corresponding to the bands observed
in the excitation- and emission spectra of Figs.
2(a)–(c) (left column) are mentioned in the caption. By means of ultra-high-resolution experiments we recently observed more than 30 sharp
vibrational lines for each of the mutants studied.
An analysis of the corresponding frequencies is in
progress.
3.2.2. RS-GFP
As already mentioned, most of the results on
RS-GFP have been reported elsewhere [28] but,
for completeness sake and for comparison to the
results obtained on other GFPs, we recapitulate
them here. In Fig. 3, a series of emission spectra
taken at various excitation wavelengths is shown
for a previously non-illuminated RS-GFP sample.
In contrast to S65T, the spectra are all identical
regardless of kexc , proving that only one form is
present in RS-GFP at 1.6 K before burning. We
tentatively assign this spectrum to the I-form. This
assignment is supported by (1) the emission- and
excitation spectra of Fig. 4(a), which show mirror
symmetry about their intersection at 499 nm, and
(2) the fact that no other intersection was observed
while exciting and detecting at various wavelengths. By comparison to wt-GFP [24] and S65T
(see above), the wavelength of the intersection and
the similarities of the shape and spectral positions
of the emission bands strongly suggest that RSGFP exists here in the I-form. Because it further
proved possible to burn a narrow hole at 499 1
nm (Fig. 4(b), right), we ascribe this wavelength to
the 0–0 transition of I.
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
Fig. 3. RS-GFP. Emission spectra at 1.6 K taken at various
excitation wavelengths between 390 and 486 nm. The spectra
are very similar, suggesting that only one form (I, see text) is
present in a previously non-illuminated RS-GFP sample at low
temperature.
By burning into the I-form of RS-GFP, not only
a 0–0 hole appears but, simultaneously, the absorption band at 495 nm (I-form) decreases and
the region between 460–475 nm increases (Fig.
4(b)), indicating that a second conformation has
been photoinduced. Since a hole could subsequently
be burnt at 476 1 nm into the new photoproduct
(Fig. 4(b)), and by similarity to wt-GFP [24], we
associate this hole with the 0–0 transition of the Bform. Thus, the reaction I ! B occurs through a
radiationless pathway (probably via the tripletstate). Conversely, when the newly formed B is excited, fluorescence is observed from I but not from
from B (not shown), implying that the reaction
B ! I is barrierless. The short excited-state lifetime of 1:1 ns obtained at room temperature may
be related to this process [36]. By burning a hole into
B0–0 , the intensity of the vibronic region of I at 495
nm increases (not shown), from which we conclude that B can be reversibly photoconverted into I.
115
By burning into the I-form of RS-GFP, both
the B-form (Fig. 4(b)) and the A-form (Figs. 4(c)
and (d)) are produced. The excitation spectrum in
Fig. 4(c) (Spectrum 1) shows the absorption region
of the expected A-form while detecting at 450 nm
(the emission region of A). A hole can be burnt on
the red wing of A at A0–0 ¼ 434 1 nm (Fig. 4(c),
right). Simultaneously, the entire A-band decreases (Spectrum 2) and the I-form increases (not
shown). When subsequently burning into I, the Aband in turn increases (Spectrum 3) proving that
the photoreaction I $ A is reversible. This is
confirmed by looking at the fluorescence spectrum
of the A-region excited at 395 nm, before and after
burning into I (Fig. 4(d)). The region between 440
and 475 nm, where the A-form is expected to emit,
shows no fluorescence signal before burning into I,
but exhibits three bands at 439, 449 and 473 nm
after burning. We conclude that the radiationless
reaction I ! A takes place and that A emits light
after having been produced and excited (see Fig.
6(b)).
Finally, after having burnt the I-form, the excitation spectrum of A (absorption region at
350–450 nm) does not increase when detected in
the emission region of I (Fig. 4(b)), indicating that
the excited-state reaction A ! I does not occur,
in contrast to what is observed in wt-GFP. Thus,
the reaction A ! I proceeds radiationlessly,
probably through the triplet-state.
3.2.3. EYFP
Although the most red-shifted mutant EYFP at
1.6 K shows a single strong absorption maximum
at 520 nm (Fig. 1), it is already present in three
conformations. This is revealed by the emissionand excitation spectra of Figs. 5(a) and (b) that
show bands that have to be attributed to all three
A-, B- and I-forms. When exciting at 400 nm, the
region in which the A-form is expected to absorb,
we observe very weak emission bands at 450, 462
and 476 nm and much stronger ones between 500
and 550 nm (Fig. 5(a), bottom). The intensity of
the emission bands at 507 and 540 nm decrease
when kexc is tuned towards the red, indicating that
these bands, by analogy to wt-GFP [24], belong to
the I-form that is populated via the excited A form (see also Fig. 6(c)). For kexc on the low-en-
116
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
(a)
(c)
(d)
(b)
Fig. 4. RS-GFP. (a) Excitation- (kdet ¼ 520 nm) and emission- (kexc ¼ 395 nm) spectra at 1.6 K of a previously non-illuminated
sample (before burning). Only the I-form is present. Its 0–0 transition at 499 nm is roughly determined by the intersection of the
spectra. The approximate vibrational frequencies corresponding to the bands of the I-form are: 160, 780 and 1510 cm1 (see also the
caption of Fig. 2) (taken from [28]). (b) Excitation spectrum before (––––) and after (- - - - - -) burning a hole into I0–0 at 499 1 nm
(right). After burning, the spectrum of the I-form decreases, whereas the vibronic bands of the B-form (between 450 and 475 nm)
increase; thus I ! B. By subsequently burning into B0–0 at 476 1 nm, a hole is produced (left) and the spectrum of I increases
again (not shown); thus, I $ B. (c) Excitation spectra of A (kdet ¼ 450 nm): (1) by burning into I0–0 at 499 1 nm, not only the Bform (Fig. 4(b)) but also the A-form is produced; thus, I ! A; (2) by subsequently burning into A0–0 at 434 1 nm, a hole is
produced (right) and the intensity of the whole A-spectrum decreases while the spectrum of I increases (not shown); thus, A ! I;
(3) after subsequently burning into the I-form, the spectrum of the A-form increases again, proving that A $ I. (d) Fluorescence
spectrum excited in the absorption region of the A-form ðkexc ¼ 395 nm), before and after burning into I0–0 at 499 1 nm. The
vibrational bands of I (in the region 500–550 nm) decrease after burning and new bands between 440 and 480 nm appear. We assign
them to the A-form, proving again that I ! A. By burning into A, these bands decrease (not shown); thus, A ! I. The
approximate vibrational frequencies corresponding to the bands of the A-form are: 265, 770 and 1920 cm1 (see also captions of
Figs. 2 and 4(a)).
ergy side of A0–0 , the A-form is no longer excited.
In addition to some weak emission from I originating by direct excitation of I, only strong bands
of the B-form at 527 and 547 nm remain (Fig.
5(a)). Thus, the reaction A ! I ! I takes place
in EYFP in a similar fashion as in wt-GFP. Simultaneously, by exciting A , a very weak emission from A ! A is observed (see also Fig. 6(c)).
Direct excitation of the I-form at 490 nm yields
only a weak I ! I emission, whereas the reactions
I ! B and I ! A occur radiationlessly, probably
through the triplet-state. Excitation into B only
(500 nm < kexc 6 524 nm) leads to a B ! B
emission and to a radiationless reaction from
B ! I (see Fig. 6(c)).
The excitation spectra of the A-, I- and B-forms
of EYFP selectively detected at 470, 510 and 530
nm, respectively (Fig. 5(b)) show vibronic bands
extending from 400 to 520 nm. Holes can be
burnt into the three forms, with A0–0 at
446 1 nm; I0–0 at 496 3 nm and B0–0 at 524 1
nm. We note that the 0–0 transitions of the A- and
B-forms of EYFP are red-shifted by about 10 and
almost 50 nm with respect to the 0–0 transitions of
wt-GFP [24], S65T and RS-GFP (see above and
Fig. 6). This strong red shift is probably related to
the T203Y mutation in EYFP, which is responsible for the p–p stacking between the phenol rings
of Tyr-203 and the chromophore (Tyr-66) [37]. We
assign the most red-shifted form of EYFP to the
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
B-form because it is photoconverted into I, and
not into A.
3.3. Energy-level diagrams
The results obtained for the three mutants are
summarized in the energy-level schemes of Fig. 6.
These schemes exhibit a general pattern similar to
that deduced for wt-GFP [24], but differ in the
details. They show the 0–0 transitions of the A-, Iand B-forms and the pathways of photoconversion
between them. Reversible phototransformations
occur between A and I, and between B and I, but
not directly between A and B.
(a)
117
For S65T (Fig. 6(a)), we found that the I-form
is the most populated. It is represented by the
strongest band absorbing at 490 nm (Fig. 1).
Since the A- and B-forms are also present at 1.6 K,
but with considerably less intensity, we assume
that their ground-state levels lie a few tens of cm1
above that of the I-form. This is contrary to what
occurs in wt-GFP, where the I-form is not populated at 1.6 K and can only be photoinduced from
the A- and/or from B-forms. Excited-state photoconversions in S65T seem to occur neither between
A and I , nor between B and I ; the barriers
between these levels are probably high (a few
1000 cm1 ). However, the photoconversions
(b)
Fig. 5. EYFP. (a) Emission spectra at various excitation wavelengths between 399 and 515 nm. All three forms A ; I and B fluoresce
when selectively excited. In addition, by exciting A at kexc 6 446 nm, not only A fluoresces (weakly), but I fluoresces strongly at 507
nm and 540 nm (see the two lower spectra). By exciting I and B (446 nm < kexc 6 496 nm), only weak fluorescence from I is observed, whereas B emits strongly. The vibrational bands of the ground-states of the three A-, I- and B-forms are marked. (b) Excitation
spectra detected in vibrational bands of the B-, I- and A-forms (top to bottom). The three forms are present at 1.6 K in a previously
non-illuminated sample (before burning), with the most red-absorbing B-form carrying the largest population. The arrows indicate the
positions at which holes (right) were burnt. The spectral positions of vibronic bands of the excited-states of the three forms are marked.
The approximate vibrational frequencies corresponding to the bands of the three forms are: 130, 780 and 1560 cm1 for the B-form,
400 and 1570 cm1 for the I-form and 200, 780 and 1430 cm1 for the A-form (see also captions of Figs. 2, 4(a) and (d)).
118
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
(a)
(b)
A ! I; I ! A; I ! B and B ! I, still occur
radiationlessly (presumably through the tripletstate).
The ground-state levels of the A- and B-forms
of RS-GFP at 1.6 K are not populated in a previously non-illuminated sample, but they become
populated as a consequence of steady illumination
(burning) of the I-form. Thus, they have to be at
least 100 cm1 higher in energy than the groundstate level of the I-form (see Fig. 6(b)). Judging
from the fluorescence spectra, however, these levels are populated at room temperature [36]. After
B is formed radiationlessly from I ðI ! BÞ, it can
be excited directly leading to the reaction
B ! I ! I. Since no emission is observed from
B to B, we conclude that there is no barrier between B and I . In contrast, by exciting A, the
reaction A ! I does not take place, but emission
from A to A is observed. The barrier between A
and I , thus, has to be high (probably a few
1000 cm1 ). The transformations A ! I; I ! A
and I ! B in RS-GFP occur radiationlessly.
In EYFP (Fig. 6(c)), the three A-, I- and Bforms are already present at 1.6 K. The groundstate level of the B-form is now the lowest-lying
level so that the B-form is the most populated one,
in contrast to the situation in S65T and RS-GFP.
By excitation into A, only very weak fluorescence
is observed from A , but strong fluorescence from
I , as in wt-GFP [24]. Thus, the process from A to
I involves a very small energy barrier. Since excitation into I does not lead to emission from B ,
the barrier between I and B has to be high (at
least a few 1000 cm1 ). When exciting into B, there
is no reaction between B and I either. We conclude that I and B reversibly photoconvert through
radiationless pathways.
(c)
Fig. 6. Energy-level diagrams of S65T (a) RS-GFP (b) and
EYFP (c). At least three forms, A, B and I, reversibly photoconvert, as indicated by the diagonal arrows. In S65T all forms
are populated at 1.6 K, with the I-form constituting the largest
fraction (lowest ground-state level). In a non-illuminated sample of RS-GFP, only the I-form is populated at low temperature, whereas the A- and B-forms are photoinduced by burning
into I. In EYFP, all three forms are populated at 1.6 K, with the
B-form having the largest population (lowest ground-state
level) (taken from [28]).
3.4. Single GFP-molecules: interpretation of their
‘‘on–off’’ and ‘‘blinking’’ behavior
Based on the energy-level schemes of Fig. 6, we
propose a tentative explanation for the ‘‘blinking’’
and ‘‘on–off’’ behaviour observed in single mutant
GFP-molecules at room temperature [29,31,32].
The experiments of [29,31] were performed on a
mutant of GFP similar to our EYFP, whereas
those of [32] were done on S65T. By continuous
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
illumination with a laser at 488 nm, the individual
molecules showed fluorescence (‘‘on’’-state) during
intervals of hundreds of microseconds to several
seconds (‘‘blinking’’) until a dark, long-lived
(‘‘off’’)-state was reached after a few minutes. In
the case of EYFP, the ‘‘on’’-state could be recovered from the ‘‘off’’-state by irradiation with a
lamp at 405 nm [29]. For both EYFP and S65T it
was reported that the ‘‘on’’-periods were dependent on laser excitation intensity, becoming shorter
at higher intensities, whereas the ‘‘off’’-periods
were independent of light intensity at 488 nm.
Let us assume that the energy-level scheme for
the EYFP mutant of [29,31] is similar to that of
our EYFP (Fig. 6(c)). Illumination at 488 nm
would then excite both the I- and the B-form. If a
single molecule, for example in the B-form, is excited, it will fluoresce (‘‘on’’) intermittently
(‘‘blinking’’) from the B -state. ‘‘Blinking’’ in
GFPs is most probably due to intersystem-crossing cycles to the triplet-state, where the molecule
remains microseconds to milliseconds before it
returns to the ground-state and is excited again
[38,39]. Eventually, the molecule will be photoconverted into the I-form. While in the I-form, the
molecule can either be excited into I at 488 nm
and emit light (from an ‘‘on’’-state with ‘‘blinking’’) or be photoconverted into the A- or B-form.
If the molecule returns to B, it will be excited again
at 488 nm and emit light (from an ‘‘on’’-state with
‘‘blinking’’). If it proceeds to A, the molecule will
not be excited at 488 nm because the energy of the
0–0 transition of A (446 nm) exceeds that of the
laser. The A-form thus becomes a dark (‘‘off’’)state until the molecule goes either thermally over
the ground-state barrier to the I-state (see Fig.
6(c)) or is illuminated at a wavelength k < 446 nm
(the 0–0 transition of A), for example at 405 nm as
in [29]. If the thermal reaction occurs, the length of
the ‘‘off’’-period should be independent of light
intensity as indeed observed, but should depend on
temperature (still to be demonstrated). However,
when the A-form is excited at 405 nm into A it
will be photoconverted into I , from where emission (from an ‘‘on’’-state with ‘‘blinking’’) occurs.
Once back in the I-form, the molecule will be excited again into I at 488 nm and the ‘‘blinking’’
and ‘‘on–off’’ cycle resumes.
119
We interpret the behavior of single S65T-molecules [32] in a similar way as for EYFP. For example, if a molecule is in the I-form, it will be
excited into I at 488 nm and will emit light intermittently from I (‘‘on’’-state and ‘‘blinking’’ due
to intersystem-crossing cycles to the triplet-state)
until it is photoconverted into A or B (see Fig.
6(a)). While in the A- or B-form, the molecule will
not be excited at 488 nm because the 0–0 transition
energies of these forms are larger than the energy of
the laser. It will thus remain in the ground-state of
the A- or B-form (dark, ‘‘off’’-states for excitation
at 488 nm) until it crosses thermally over the energy
barrier back into the I-form. The length of this
‘‘off’’-period should therefore not depend on light
intensity as indeed observed, but should only depend on temperature (still to be verified). Once
back in the I-form, the molecule can be excited
again into I at 488 nm and recommence the
‘‘blinking’’ and ‘‘on–off’’ cycle.
4. Conclusions
We have proven here by means of low temperature laser spectroscopy and hole-burning that the
red-shifted GFP-mutants S65T and EYFP, analogously to RS-GFP [28], are not ‘‘one color’’,
photostable proteins, as assumed in the literature.
They photoconvert reversibly among at least three
(protonated and deprotonated) forms, A, I and B,
in a manner similar to that previously demonstrated by us for wt-type GFP.
In addition to identifying the 0–0 transitions of
the various forms within a specific mutant, we
have determined the pathways of interconversion
and interpreted the results in terms of energy-level
schemes. Such reversible photoreactions between
different forms of a GFP molecule probably represent a more general phenomenon also occurring
in other GFP-mutants, as well as in similar types
of autofluorescent proteins such as the red fluorescent protein DsRed [40].
As mentioned earlier [24,28], our results have
important consequences for applications of GFPmutants in molecular and cell biology, such as the
study of protein–protein interactions by fluorescence resonance energy transfer (FRET). In this
120
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
technique a pair of differently absorbing GFPmutants, one as donor and the other as acceptor,
are used. A change in color of the fluorescence is
generally interpreted as a sign for energy transfer,
and thus for an interaction between different proteins. We have demonstrated, however, that a
change in color may also originate from a photoinduced reaction among conformations within a
given GFP-mutant.
The localization of the 0–0 transitions by holeburning will allow us now to determine their
homogeneous linewidths and get information on
dynamical processes, like excited-state proton
transfer reactions and chromophore–protein interactions at low temperature. The energy-level
diagrams, furthermore, provide detailed explanations of the ‘‘on–off’’ and ‘‘blinking’’ behavior of
specific single mutant GFP-molecules at room
temperature.
Acknowledgements
T.M.H.C., A.J.L. and S.V. would like to thank
J.H. van der Waals for insightful discussions and
for his continuous interest in this work. S.V. further
thanks F. K€
onz for editing the figures. V.S. was a
recipient of a Human Frontiers Science Program
long-term fellowship. We thank R. Rivera-Pomar
and D. Piston for gifts of GFP plasmids. The
spectroscopy experiments, carried out in Leiden,
were financially supported by the Netherlands
Foundation for Physical Research (FOM) and the
Council for Chemical Research of the Netherlands
Organization for Scientific Research (NWO-CW).
References
[1] R.Y. Tsien, Annu. Rev. Biochem. 67 (1998) 509, and
references therein.
[2] A.B. Cubitt, R. Heim, S.R. Adams, A.E. Boyd, L.A.
Gross, R.Y. Tsien, Trends Biochem. Sci. 20 (1995) 448.
[3] R. Heim, R.Y. Tsien, Curr. Biol. 6 (1996) 178.
[4] M.V. Matz, A.F. Fradkov, Y.A. Labas, A.P. Savitsky,
A.G. Zaraisky, M.L. Markelov, S.A. Lukyanov, Nat.
Biotechnol. 17 (1999) 969.
[5] S. Jakobs, V. Subramaniam, A. Sch€
onle, T.M. Jovin, S.W.
Hell, FEBS Lett. 479 (2000) 131.
[6] L.A. Gross, G.S. Baird, R.C. Hoffman, K.K. Baldridge,
T.Y. Tsien, Proc. Natl. Acad. Sci. USA 97 (2000) 11990,
and references therein.
[7] M.A. Wall, M. Socolich, R. Ranganathan, Nat. Sruct.
Biol. 7 (2000) 1133;
D. Yarbrough, R.M. Wachter, K. Kallio, M.V. Matz, S.J.
Remington, Proc. Natl. Acad. Sci. USA 98 (2001) 462.
[8] A. Miyawaki, J. Llopis, R. Heim, J.M. McCaffery, J.A.
Adams, M. Ikura, R.Y. Tsien, Nature 388 (1997) 882.
[9] J.F. Presley, N.A. Cole, T.A. Schroer, K. Hirschborn, J.M.
Zaal, J. Lippincott-Schwartz, Nature 389 (1997) 81.
[10] A. Miyawaki, O. Griesbeck, R. Heim, R.Y. Tsien, Proc.
Natl. Acad. Sci. USA 96 (1999) 2135.
[11] K.K. Jensen, L. Martini, T.W. Schwartz, Biochemistry 40
(2001) 938.
[12] D.C. Prasher, V.K. Eckenrode, W.W. Ward, F.G. Prendergast, M.J. Cormier, Gene 111 (1992) 229.
[13] M. Chalfie, Y. Tu, G. Euskirchen, W.W. Ward, D.C.
Prasher, Science 263 (1994) 802.
[14] F. Yang, L.G. Moss, G.N. Philips Jr., Nat. Biotechnol. 4
(1996) 1246.
[15] M. Orm€
o, A.B. Cubitt, K. Kallio, L.A. Gross, R.Y. Tsien,
S.J. Remington, Science 273 (1996) 1392.
[16] R. Heim, D.C. Prasher, R.Y. Tsien, Proc. Natl. Acad. Sci.
USA 91 (1994) 12501.
[17] S. Delagrave, R.E. Hawtin, C.M. Silva, M.M. Yang, D.C.
Youvan, Biotechnololgy 13 (1995) 151.
[18] G.J. Palm, A. Zdanov, G.A. Gaitanaris, R. Stauber, G.N.
Pavlakis, A. Wlodawer, Nat. Struct. Biol. 4 (1997) 361.
[19] H. Niwa, S. Inouye, R. Hirano, T. Matsuno, S. Kojima,
M. Kubota, M. Ohashi, F.I. Tsuji, Proc. Natl. Acad. Sci.
USA 93 (1996) 13617.
[20] R. Heim, A.B. Cubitt, R.Y. Tsien, Nature 373 (1995)
663.
[21] W.W. Ward, H.J. Prentice, A.F. Roth, C.W. Cody, S.C.
Reeves, Photochem. Photobiol. 35 (1982) 803.
[22] M. Chattoraj, B.A. King, G.U. Bublitz, S.G. Boxer, Natl.
Acad. Sci. USA 93 (1996) 8362.
[23] H. Lossau, A. Kummer, R. Heinecke, F. P€
ollinger-Dammer, C. Kompa, G. Bieser, T. Jonsson, C.M. Silva, M.M.
Yang, D.C. Youvan, M.E. Michel-Beyerle, Chem. Phys.
213 (1996) 1.
[24] T.M.H. Creemers, A.J. Lock, V. Subramaniam, T.M.
Jovin, S. V€
olker, Nat. Struct. Biol. 6 (1999) 557.
[25] A.A. Voityuk, M.E. Michel-Beyerle, N. R€
osch, Chem.
Phys. Lett. 272 (1997) 162;
A.A. Voityuk, A.D. Kummer, M.E. Michel-Beyerle,
N. R€
osch, Chem. Phys. 269 (2001) 83.
[26] W. Weber, V. Helm, J.A. McCammon, P.W. Langhoff,
Proc. Natl. Acad. Sci. USA 96 (1999) 6177.
[27] H.Y. Yoo, J.A. Boatz, V. Helm, J.A. McCammon, P.W.
Langhoff, J. Phys. Chem. B 105 (2001) 2850.
[28] T.M.H. Creemers, A.J. Lock, V. Subramaniam, T.M.
Jovin, S. V€
olker, Proc. Natl. Acad. Sci. USA 97 (2000)
2974.
[29] R.M. Dickson, A.B. Cubitt, R.Y. Tsien, W.E. Moerner,
Nature 388 (1997) 355.
T.M.H. Creemers et al. / Chemical Physics 275 (2002) 109–121
[30] K. Brejc, T. Sixma, P.A. Kitts, S.R. Kain, R.Y. Tsien, M.
Orm€
o, S.J. Remington, Proc. Natl. Acad. Sci. USA 94
(1997) 2306.
[31] E.J.G. Peterman, S. Brasselet, W.E. Moerner, J. Phys.
Chem. A 103 (1999) 10553.
[32] M. Garcia-Parajo, G.M.J. Segers-Nolten, J.-A. Veerman,
J. Greve, N.F. van Hulst, Proc. Natl. Acad. Sci. USA 97
(2000) 7237.
[33] R. Janknecht, G. de Martijnoff, J. Lou, R.A. Hipskind, A.
Nordheim, G. Stunnenberg, Proc. Natl. Acad. Sci. USA 88
(1991) 8972.
[34] G.H. Patterson, S.M. Knobel, W.D. Sharif, S.R. Kain,
D.W. Piston, Biophys. J. 73 (1997) 2782.
121
[35] S. V€
olker, in: J. F€
unfschilling (Ed.), Relaxation Processes
in Molecular Excited States, Kluwer, Dordrecht, 1989, and
references therein;
S. V€
olker, Annu. Rev. Phys. Chem. 40 (1989) 499–530.
[36] A. Volkmer, V. Subramaniam, D.J.S. Birch, T.M. Jovin,
Biophys. J. 78 (2000) 1589.
[37] R.M. Wachter, M.-A. Elsliger, K. Kallio, G.T. Hanson,
S.J. Remington, Structure 6 (1998) 1267.
[38] T. Basche, S. Kummer, C. Br€auchle, Nature 373 (1995) 132.
[39] A.C.J. Brouwer, E.J.J. Groenen, J. Schmidt, Phys. Rev.
Lett. 80 (1998) 3944.
[40] J. Gallus, S. Bonsma, F. K€
onz, V. Subramaniam, T.M.
Jovin, S. V€
olker, manuscript in preparation.