Concerning the thermal diastereomerization of the green fluorescent protein chromophore

Article abstract of DOI:10.1007/s00706-005-0418-4

Two model compounds for the green fluorescent protein chromophore were prepared. One of them incorporates the natural 4-hydroxybenzylidene group of the natural tyrosin derived chromophore, the other one bears a methyl group instead of the hydroxy group. W

Full text of DOI:10.1007/s00706-005-0418-4

Monatshefte f€ur Chemie 137, 163–168 (2006)
DOI 10.1007/s00706-005-0418-4
Concerning the Thermal
Diastereomerization of the Green
Fluorescent Protein Chromophore
ꢀ
Beate Hager, Bettina Schwarzinger, and Heinz Falk
Institute of Organic Chemistry, Johannes Kepler University Linz, 4040 Linz, Austria
Received September 28, 2005; accepted October 11, 2005
Published online December 13, 2005 # Springer-Verlag 2005
Summary. Two model compounds for the green fluorescent protein chromophore were prepared. One
of them incorporates the natural 4-hydroxybenzylidene group of the natural tyrosin derived chromo-
phore, the other one bears a methyl group instead of the hydroxy group. Whereas the photochemically
prepared (E)-diastereomer of the first compound very effectively reverted thermally (room tempera-
ture) to the thermodynamically stable (Z)-diastereomer, the (E)-diastereomer of the second derivative
proved to be stable even at elevated temperatures for more than a day. This finding can be rationalized
by constructing the appropriate resonance structures showing that only in the first case an effective
delocalization enables partial single bond character of the benzylidene double bond. From the stand-
point of chemical etiology, only Nature’s choice of the tyrosin derived chromophore of the green
fluorescent protein provides an efficient radiationless thermal relaxation channel for the unwanted
photo-diastereomerization product formed after excitation besides the dominating fluorescence channel
of its chromophore.
Keywords. Green fluorescent protein; Benzylideneimidazolinones; Thermal barrier; Chemical etiol-
ogy; Radiationless relaxation.
Introduction
The greenish bioluminescence of the jellyfish Aequorea victoria originates from
the closely associated calcium(II) binding protein aquorin and the green fluorescent
protein (GFP) [1, 2]. Because of its importance as a cloneable reporter of gene
expression, GFP has been intensively studied over the past decade and its applica-
tions, structural details, and photophysical behavior have been thoroughly reviewed
[3]. The fluorescence of GFP is emitted from its (4-hydroxybenzylidene)imidaz-
olinone chromophore I, which is autocatalytically generated by a posttranslational
intramolecular cyclization of the S65-Y66-E67 pre-protein sequence followed by
dehydrogenation of the tyrosine (Y) moiety [4]. Interestingly enough, the strong
ꢀ
Corresponding author. E-mail: heinz.falk@jku.at

164
B. Hager et al.
fluorescence of this unit is lost upon denaturation of the protein and is also absent
in model compounds of the chromophore, a feature that is attributed to the result-
ing flexibilization of the chromophore [5].
Formula 1
Numerous X-ray studies of the protein and its mutant analogs as well as of
model compounds have revealed that the most stable form of the benzylidenic
substructure I is the (Z)-diastereomer [3]. A detailed study [6] has demonstrated
that the free energy difference between the (Z)-diastereomer and its photoaccessi-
ble (E)-diastereomer is in the order of 5 kJ ꢁ molꢂ1 depending only slightly on the
ionization of the chromophore. The thermal activation barrier between the two
diastereomers has been estimated to amount to about 50 kJ ꢁ molꢂ1, again depend-
ing only marginally on the ionization of the chromophore [6]. This latter value is
considerably smaller than the values usually found for comparable diastereomer-
ization reactions at exocyclic double bonds of heterocycles, e.g. in the dipyrri-
nones, which are in the order of 100 kJ ꢁ molꢂ1 [7]. This rather large difference
led us to the question if the low activation barrier encountered is an intrinsic
property of the benzylideneimidazolinone chromophoric system or is due to the
special choice of Nature of the tyrosine amino acid moiety. Thus, in this study we
compare the thermal diastereomerization possibility of the model compound pair
(E)=(Z)-1, which is close in constitution to the natural GFP chromophore (Z)-I,
with the model system (E)=(Z)-2, which is devoid of the tyrosine derived 4-hydroxy-
phenyl group.
Formula 2
Results and Discussions
The (Z)-diastereomers of 1 and 2 were prepared by condensation of 4-hydroxy-
and 4-methylbenzaldehyde with N-acetylglycin followed by an Erlenmeyer-

Concerning the Thermal Diastereomerization
165
Fig. 1. Thermal relaxation of (E)-1 (25ꢃC) and (E)-2 (25–50ꢃC) in CD3OD
Pl€ochl synthesis [8]. Then, the resulting azlactones 3 and 4 were converted to the
target azlactames (Z)-1 and (Z)-2 by refluxing with ethylamine according to the
procedure of Kojima et al. [9]. The (Z)-configuration of these compounds was
inferred by comparison with similar model compounds of proven configurations
[6]. Photodiastereomerizations of the (Z)-diastereomers dissolved in methanol into
their (E)-diastereomers were achieved by irradiation with the light of a high-
1
pressure mercury lamp. The H NMR resonances were assigned straightforwardly
and the thermal fate of the initial (E)-diastereomers could be easily followed mon-
itoring the nicely separated CH-, CH3-, and CH2CH3-signal intensities in CD3OD
solution.
In these experiments it turned out that the signals of (E)-1 decayed rapidly into
those of (Z)-1 even at room temperature following a first order reaction (Fig. 1). In
contrast to this finding, (E)-2 remained stable up to the boiling point of methanol
for several hours to even a day (Fig. 1). The equilibration between (Z)- and (E)-1
led to a free enthalpy difference of about 5 kJ ꢁ molꢂ1 consistent with findings on
other model systems [6] and calculations [3]. Due to the limited precision of the
NMR measurements we refrained from estimating an activation energy. But to
determine this value for 1 would, of course, be unnecessary because it had been
estimated already for a similar molecule [6]. The important and significant result of
the experiment contained in Fig. 1 is that (E)-1 easily reverts to (Z)-1 whereas (E)-2
does not. This result clearly demonstrated that the 4-hydroxy group of 1 is a
conditio sine qua non for the very easy thermal equilibration providing an efficient
thermal relaxation channel of the system.
These findings, unexpected at first sight, might be properly rationalized by
constructing the appropriate resonance structures of the conjugated systems of 1
and 2 taking the hydroxy group of 1 (in its different ionization states) into account
in the classical manner of organic chemistry as drawn in Fig. 2. Accordingly,
resonance structures involving the benzylidene partial structure can be constructed
exclusively in the case of the 4-hydroxy group (dissociated or undissociated makes

166
B. Hager et al.
Fig. 2. Resonance structures of (Z)-1 and (Z)-2
no difference) in 1. Thus, 1 can be addressed as a vinylogous carboxylic acid or
vinylogous carboxylate (II) as shown in Fig. 2 yielding partial single bond char-
acter and pronounced delocalizability of the benzylidenic double bond. From this
result it becomes immediately clear why the ionization state of the 4-hydroxy
group is not important [6] for the diasteromerization process: the phenolic –O–H
and –OðꢂÞ groups behave similarly in providing conjugation in this push–pull
electronic system, and thus relay partial single bond character to the benzylidene
double bond. This effect is non-existent in the 4-methyl analogue 2, which behaves
as a benzylidene-imidazolidinone, i.e. like a ‘‘normal’’ heterocyclic compound
bearing an exocyclic double bond like the dipyrrinones [7].
The result of the present investigation reminds one vividly of the very smooth thermal diaste-
reomerization encountered at the exocyclic double bond in position 5 of the 2,3-dihydrobilin-1,19-
diones (the chromophore of e.g. phytochrome) as compared to the normal thermal stability of this
double bond in the bilin-1,19-diones (as in e.g. biliverdin) [10]. Only the partially saturated system of
the 2,3-dihydrobilin-1,19-diones compares to the easily delocalizable and accordingly thermally easily
diastereomerizable bond system of an N-acylenamine.
In conclusion, in the sense of chemical etiology Nature has been very wise to
select the tyrosine unit for constructing the chromophore of GFP instead choosing
e.g. a phenylalanine moiety. Obviously, as derived above only the tyrosine derived
chromophore is amenable to display the properties of a vinylogous carboxylic acid
or vinylogous carboxylate partial structure needed for an efficient thermal relaxa-
tion channel and the necessary flexibility to adapt to the needs of the functional
unit of the native GFP. In particular, the easy thermal diastereomerization allows
for an efficient radiationless thermal relaxation of the unwanted photo-diastereo-
merization product ((E)-I) that might be formed concomitantly after excitation
((Z)-I ! (E)-I) besides the dominating fluorescence channel of the chromophore.
Experimental
Solvents were of p.a. quality unless stated otherwise. Photodiastereomerizations of (Z)-1 and (Z)-2
were executed by means of a Hanau TQ 150Z2 UV-lamp directly in the NMR-tube in MeOH-d4
solutions for 1 h. NMR spectra were recorded on a Bruker Avance DPX 200 MHz spectrometer and
DRX 500 MHz spectrometer using a TXI cryoprobe with z-gradient coil. 2D NMR experiments were

Concerning the Thermal Diastereomerization
167
performed using standard pulse sequences as provided by the manufacturer. Typical 90ꢃ hard pulse
durations were 8.2 ꢀs (1H) and 16.6ꢀs (13C), 90ꢃ pulses in decoupling experiments were set to 67ꢀs.
HSQC and HMBC experiments were optimized for coupling constants of 145 Hz for single quantum
correlations and 10Hz for multi-bond correlations. The NOESY mixing time was set to 400 ms. Mass
spectra were recorded on a Hewlett Packard 5989 quadrupole instrument.
(Z)-1-Ethyl-4-(4-hydroxybenzylidene)-2-methyl-1H-imidazol-5(4H)-one ((Z)-1, C13H14N2O2)
To a solution of 0.3g 3 (1.2mmol) dissolved in 15cm3 EtOH containing 0.3g K2CO3 (2.2mmol),
0.15cm3 C2H5NH2 (70% in H2O) (1.9mmol) were added and the resulting mixture was heated to
reflux for 4 h. After cooling the K2CO3 was filtered off, the solution was concentrated in vacuo, and the
crude product was purified by silica gel chromatography using CHCl3=MeOH (10=1, v=v) as the
developing solvent to give 84 mg (Z)-1 (30% yield). TLC: Rf ¼ 0.6 (CHCl3:MeOH¼ 10:1); 1H
NMR (500 MHz, MeOD-d3, 30ꢃC): ꢁ ¼ 1.23 (t, J ¼ 7.14 Hz, –CH2–CH3), 2.38 (s, –CH3), 3.68 (q,
J ¼ 7.14Hz, –CH2–), 6.84 (d, J ¼ 8.51 Hz, –H30), 7.00 (s, –CH), 8.00 (d, J ¼ 8.51Hz, –H20) ppm;
NOESY (MeOD-d3): –H20 ! –CH and –H30, –H30 ! –H20, –CH2– ! –CH2–CH3, –CH2–CH3 !
13
–CH2–; C NMR (125 MHz, MeOD-d3, 30ꢃC): ꢁ ¼ 14.6 (–CH2–CH3), 15.3 (–CH3), 36.4 (–CH2–),
116.8 (C30), 126.9 (–CH), 129.2 (C10), 135.6 (C20), 137.1 (C4), 161.5 (C40), 163.0 (C2), 172.2 (C5)
ppm; HMBC (MeOD-d3): CH2–CH3 ! C2, C5, and –CH2–, –CH3 ! C2, C10 (weak), and C4 (weak),
–CH2– ! C5, C2, and –CH2–CH3, H30 ! C40, C20, –CH, and C30, –CH ! C5, C20, and C30 (weak),
H20 ! C40, C20, C10, –CH, and C30; HSQC data were according to structure; ESI-MS (MeOHþ 1 vol-
% HCOOH, ꢂ ꢄ 1 mgcmꢂ3, positive ion mode): m=z ¼ 231 ([M ꢂ H]þ).
(E)-1-Ethyl-4-(4-hydroxybenzylidene)-2-methyl-1H-imidazol-5(4H)-one ((E)-1, C13H14N2O2)
1H NMR (200 MHz, MeOD-d3, 30ꢃC): ꢁ ¼ 1.24 (t, J ¼ 7.14 Hz, –CH2–CH3), 2.35 (s, –CH3), 3.68 (q,
J ¼ 7.14Hz, –CH2–), 6.84 (d, J ¼ 8.56 Hz, –H30), 7.15 (s, –CH), 8.25 (d, J ¼ 8.56Hz, –H20) ppm.
(Z)-1-Ethyl-4-(4-methylbenzylidene)-2-methyl-1H-imidazol-5(4H)-one ((Z)-2, C14H16N2O)
Compound (Z)-2 was prepared from 4 according to (Z)-1 in 15% yield. 1H NMR (500MHz, MeOD-d3,
30ꢃC): ꢁ ¼ 1.33 (t, J ¼ 7.14Hz, –CH2–CH3), 2.11 (s, –CH3), 2.36 (s, ar–CH3), 4.26 (q, J ¼ 7.14 Hz,
–CH2–), 7.23 (d, J ¼ 7.41 Hz, –H30), 7.39 (s, –CH), 7.49 (d, J ¼ 7.41Hz, –H20) ppm; NOESY
(MeOD-d3): –H30 ! ar–CH3 and –H20, –CH2– ! –CH2–CH3; 13C NMR (125MHz, MeOD-d3,
30ꢃC): ꢁ ¼ 14.5 (–CH2–CH3), 21.4 (ar–CH3), 22.4 (–CH3), 62.5 (–CH2–), 126.1 (C4), 130.4 (C30),
131.0 (C20), 132.0 (C10), 135.4 (–CH), 141.3 (C40), 166.8 (C5), 173.2 (C2) ppm; HMBC (MeOD-d3):
CH3 ! C2, ar–CH3 ! C40 and C30, H30 ! C40, C10, and –CH, –CH ! C5 and C20, H20 ! C40, –CH,
C30, and C10 (weak); HSQC data were according to structure; ESI-MS (MeOH þ 1 vol-% HCOOH,
ꢂ ꢄ 1 mg cmꢂ3, positive ion mode): m=z ¼ 229 ([M ꢂ H] þ). This diastereomer proved to be config-
urationally stable in H2O solution up to 100ꢃC.
(E)-1-Ethyl-4-(4-methylbenzylidene)-2-methyl-1H-imidazol-5(4H)-one ((E)-2, C14H16N2O)
1H NMR characterization was only possible for a mixture of the (Z)-2:(E)-2 diastereomers (ratio 3:1),
which resulted in a few overlaps, in particular in the aromatic region. 1H NMR (500 MHz, MeOD-d3,
30ꢃC): ꢁ ¼ 1.51 (t, J ¼ 7.14Hz, –CH2–CH3), 2.05 (s, –CH3), 2.33 (s, ar–CH3), 4.15 (q, J ¼ 7.14 Hz,
–CH2–), 7.05–7.30 (m, 5H, no exact assignment possible due to an overlap with the signals of (Z)-2)
ppm. This diastereomer proved to be configurationally stable in H2O solution up to 100ꢃC, but reverted
partially to (Z)-2 under this condition.
(Z)-4-(4-Acetoxybenzylidene)-2-methyloxazol-5(4H)-one (3, C13H11NO4)
N-Acetylglycine, 4.6g (40 mmol), 5.6 g 4-hydroxybenzaldehyde (46 mmol), and 2.6 g anhydrous
CH3COONa (30 mmol) were added to 20cm3 (CH3CO)2O and the resulting mixture was heated to
reflux for 2 h. The solution was cooled with ice and then poured into 100 cm3 ice-H2O. The precipitate

168
B. Hager et al.: Concerning the Thermal Diastereomerization
1
was collected by filtration and washed with small amounts of EtOH giving 5.8g 3 (59%). H NMR
(200 MHz, MeOD-d3, 30ꢃC): ꢁ ¼ 2.30 (s, –CO–CH3), 2.40 (s, –CH3), 7.14 (s, –CH), 7.21 (d,
J ¼ 8.56Hz, –H30), 8.19 (d, J ¼ 8.56Hz, –H20) ppm.
(Z)-4-(4-Methylbenzylidene)-2-methyloxazol-5(4H)-one (4, C12H11NO2)
1
Compound 4 was prepared according to 3 in 20% yield. H NMR (500MHz, DMSO-d6, 30ꢃC):
ꢁ ¼ 2.36 (s, –CO–CH3), 2.39 (s, –CH3), 7.18 (s, –CH), 7.31 (d, J ¼ 7.68 Hz, –H30), 8.08 (d, J ¼ 7.68Hz,
–H20) ppm.
Acknowledgements
The cryogenic 500 MHz probe used was purchased from FWF project P15380 (project leader:
Prof. Dr. N. M€uller). We are grateful to DI. M. Waser for recording the 500 MHz NMR-spectra, to
Prof. Dr. C. Klampfl for his support in recording of MS and to V. Romanova for assistance in
preparative work.
References
[1] Shimomura O (1979) FEBS Lett 104: 220
[2] Charbonneau H, Walsh KA, McCann RO, Prendergast FG, Cormier MJ, Vanaman TC (1985)
Biochem 24: 6762
[3] Zimmer M (2002) Chem Rev 102: 759
[4] Heim R, Cubitt AB, Tsien RY (1995) Nature 373: 663; Niwa H, Inouye S, Hirano T, Matsuno T,
Kojima S, Kubota M, Ohashi M, Tsuji FI (1996) Proc Natl Acad Sci 93: 13617
[5] Ward WW, Bokman SH (1982) Biochem 21: 4535; Webber NM, Litvinenko KL, Meech SR
(2001) J Phys Chem B 105: 8036
[6] He X, Bell AF, Tonge PJ (2003) FEBS Lett 549: 35
ꢀ
[7] Falk H, Grubmayr K, Hofer O, Neufingerl F, Ribo JM (1976) Monatsh Chem 107: 831
€
[8] Plochl J (1883) Ber Dtsch Chem Ges 16: 2815
[9] Kojima S, Ohkawa H, Hirano T, Maki S, Niwa H, Ohashi M, Inouye S, Tsuji FI (1998)
Tetrahedron Lett 39: 5239
[10] Falk H (1989) The Chemistry of Linear Oligopyrroles and Bile Pigments, Springer Wien
New York, p 521