Isomerization in fluorescent protein chromophores involves addition/elimination

Article abstract of DOI:10.1021/ja803416h

The green fluorescent protein (GFP) chromophore undergoes both photochemical and thermal isomerizations. Typically, the Z form is more stable and undergoes photochemical conversion to the E form followed by thermal reversion over a period of seconds or minutes. Although the mechanism of the thermal reversion has been the subject of some investigations, the surprisingly low activation energy for this process has not sparked any controversy. We now show that the chromophore is surprisingly stable in both E and Z forms and that the facile thermal reversion is the result of a novel nucleophilic addition/elimination mechanism. This observation may have implications for the intervention of such processes, as well as blinking and kindling, in fluorescent proteins. Copyright

Full text of DOI:10.1021/ja803416h

Published on Web 10/01/2008
Isomerization in Fluorescent Protein Chromophores Involves Addition/
Elimination
Jian Dong, Fardokht Abulwerdi,† Anthony Baldridge, Janusz Kowalik, Kyril M. Solntsev, and
Laren M. Tolbert*
School of Chemistry and Biochemistry, 901 Atlantic DriVe; Georgia Institute of Technology, Atlanta, Georgia
30332-0400
Received May 7, 2008; E-mail: laren.tolbert@chemistry.gatech.edu
Fluorescent proteins related to the green fluorescent protein (GFP)
are thought to undergo Z/E photoisomerization between fluorescent
and dark states. The Z form (“cis”) is the resting fluorescent form,
while the E (“trans”) form is nonfluorescent, although exceptions are
known.1
Such proteins are also characterized by “blinking”, that is, temporary
conversion to a nonfluorescent form, which has been variously
attributed to triplet formation2 or proton transfer.3 Additionally, strong
support for cis/trans isomerization is provided by the behavior of
kindling fluorescent proteins,4 in which the resting nonfluorescent form
has trans stereochemistry. Upon irradiation into the long-wavelength
band, the protein is “kindled” to the fluorescent cis form. That kindling
is the result of photoisomerization is given strong support by recent
single crystal X-ray determinations of both forms of two kindling
proteins, dronpa and mTFP0.7, which differ in the stereochemistry
about the benzylidene bond of the chromophores.5 A key unresolved
issue in the photophysics of the fluorescent proteins is whether the
cis/trans isomerization is related to the blinking phenomenon, which,
in addition to isomerization, has been ascribed to protonation and triplet
formation mentioned earlier. We have observed that, in solution, the
isolated UV-excited GFP chromophore undergoes a fast relaxation to
yield a resting state which shows considerable twisting6 at the same
Figure 1. UV-vis (top) and IR (bottom) absorption spectra of E and Z
isomers of p-MeOBDI in MeCN.
time that the excited-state is quenched.7 In the protein, however, the
hydroxybenzylidenedimethylimidazolinone (HOBDI, X ) OH)10
common view is that the protein prohibits twisting about the double
bond. Nonetheless, both calculations8 and the aforementioned kindling
following photoisomerization from the Z form and obtained a barrier
of 13.1 kcal/mol for the isomerization from an Arrhenius plot.11
behavior require that, in at least some instances, formal isomerization,
Surprisingly, little account has been taken of the observation that
that is, decay from the twisted intermediate onto the trans hypersurface,
a high level ab initio calculation from Weber, et al., produces a
must be permitted. Moreover, the quite wide variation in blinking
behavior as a function of protein structuresconditions which either
facilitate or inhibit such isomerizationssuggest that blinking and
isomerization are intimately involved.
barrier of 57 kcal/mol,12 a value more typical of double-bond
isomerization barriers for unexceptional benzylidene molecules such
as HOBDI. Alternatively, tautomerization to a zwitterionic inter-
mediate via an uncalculated transition structure, which then rotates
with a 7.3 kcal/mol12 barrier, presents another facile mechanism
Scheme 1. Isomerization in the GFP Chromophores
for isomerization. This poses a conundrum: how does one resolve
the discrepancies between highly credible experimental and theo-
retical determinations?
To develop further insight into this process, and to exclude
possible proton-transfer processes in the isomerizations, we first
examined the methyl ether of HOBDI, MeOBDI. In the process
of exploring the optimal solvents for our physical studies, we
encountered a surprise. That is, in benzene and acetonitrile, no
thermal isomerization occurred, but isomerization was readily
observed in methanol and, at slower rates, in Me2SO-d6.13 Indeed,
we were able to isolate the trans isomer of MeOBDI by silica gel
chromatography and record its NMR, ir, and electronic absorption
spectra (see Figure 1 and Supporting Information).14 Again, this
result is inconsistent with a facile thermal isomerization, and
suggests that a more complex process is intervening.
While the photoisomerization mechanism has been the subject
of several studies9 and corresponds in unexceptional ways to the
mechanisms of other arylidene chromophores, the mechanism of
the thermal reverse isomerization is more problematic. The blinking
phenomenon requires that isomerization, if involved, be thermally
reversible. Tonge has recently measured the rates of thermal
isomerization of the representative E protein chromophore p-
These waters were further muddied by a recent claim15 that an
analogous 4-methylbenzylidene derivative does not isomerize under
† REU student, Department of Chemistry and Biochemistry, University of Tulsa,
Tulsa, OK 74104.
9
14096 J. AM. CHEM. SOC. 2008, 130, 14096–14098
10.1021/ja803416h CCC: $40.75
2008 American Chemical Society

C O M M U N I C A T I O N S
Figure 2. Pathways for the E-XBDI isomerization. (a) Direct; (b) addition/elimination; (c) isomerization by tautomerization.
similar conditions, which was rationalized as the result of a strong
“push-pull” transition structure (see second resonance structure for Path
a in Figure 2) in which a donating group such as hydroxyl weakens
the double bond character. However, such a structure belies the strong
electron donation effect of the methyl group, which would be
inconsistent with a difference in rates of 3 or more orders of magnitude.
To resolve this conundrum, we embarked on the synthesis and rate
study of the effect of substituents on the thermal isomerization of cis-
XBDI, using substituents that are either donors or acceptors. We
reasoned that use of a classical Hammett plot would reveal the nature
of the isomerization and confirm or disprove the proposed mechanism.
Thus we used or synthesized BDI derivatives with decreasing para-
donating ability, HO, CH3O, CH3, H, and Cl, with the expectation
that a Hammett plot would produce a negative F value if the “push-
pull” zwitterionic nature of the transition state dominates as in Path a.
In the process, we discovered that the reported15a ethyl derivative does
not have the assigned structure, an observation we confirmed by single-
crystal X-ray diffraction. The NMR of the reported derivative is
characterized by a downfield doublet at 7.45 ppm, whereas every BDI
derivative we have made has a doublet at 8.0-8.2 ppm for the alpha
aryl proton. Because this product was obtained in 15% yield, we
conclude that this was a byproduct of the reaction.
Figure 3. Hammett σF plot for para-E-XBDI thermal isomerization in
CD3CN in the presence of 0.089 M DABCO; kH ) 1.4 10-5 s-1
.
To test this mechanism, we examined the thermal reversion in a variety
of solvents. As a test case, we again used MeOBDI, since this molecule
is isoelectronic to HOBDI without allowing for the possibility of intra-
or intermolecular proton transfer involving hydroxyl group. Unlike
the results in Me2SO-d6, we observed no isomerization in the absence
of added base in either acetonitrile or benzene. However, both exhibited
facile isomerization in the presence of primary amines, for example,
propyl amine. In the case of tertiary amines, either 1,4-
diaza[2.2.2]bicyclooctane (DABCO) or 4-(dimethylamino)pyridine
(DMAP) isomerization was observed only in acetonitrile solvent.
In our hands, all five analogs underwent ready photoisomerization
in a variety of solvents, including MeOD, Me2SO-d6, and MeCN-
d3, allowing us to study the thermal reversion by proton NMR
spectroscopy (see Supporting Information). In addition, all pho-
toisomers, including MeBDI, underwent the reverse reaction at a
convenient rate in Me2SO-d6. Ignoring HOBDI (see below), the
plot of log(kX/kH) gave a positive correlation with σ, with the best
results obtained in acetonitrile (see Figure 3), a result inconsistent
with the previous mechanistic analysis.15a,b
If the proposed mechanism were correct, the rate of isomerization
should be rate-limiting in formation of the adduct, which in term
should be proportional to the concentration of nucleophile. Again
we used DABCO with MeOBDI in acetonitrile. As anticipated for
the proposed bimolecular mechanism, isomerization was linear with
base concentration over 1 order of magnitude (see Figure S3,
Supporting Information). Repeating the Hammett plot with 0.089
M DABCO, we obtained a F-value of 0.84, consistent with
nucleophilic attack (see Figure 3). HOBDI itself isomerized without
added base in MeCN-d3, but the isomerization was inhibited by
the presence of added amine.
What, then, is the correct mechanism? The key is provided by the
positive F and by the observation that thermal isomerization occurs
only in nucleophilic solvents such as Me2SO-d6. This result is
reminiscent of that for SNAr reaction, in which substitution by an
electron-releasing group at the ipso position retards the reaction,
producing a positive F.16 Thus we conceived instead a similar
mechanism, an addition/elimination mechanism (Path b in Figure 2).
9
J. AM. CHEM. SOC. VOL. 130, NO. 43, 2008 14097

C O M M U N I C A T I O N S
of an internal nucleophile thus becomes an additional point of
mutation of such chromophores that may mediate their photophysi-
cal properties. Such mutations are being explored.
Acknowledgment. We thank the U.S. National Science Foun-
dation (CHE-0456892 and 0809179 for L.M.T. and K.M.S.) for
generous financial support.
Supporting Information Available: Synthetic and kinetic details,
and NMR spectra for all BDI derivatives. This material is available
free of charge via the Internet at http://pubs.acs.org.
Figure 4. Nucleophile in mTFP0.7.
The results we have observed are congruent with the addition/
elimination mechanism shown as Path b in Figure 2. Although this
is a plausible mechanism, literature precedents are fairly rare. The
isomerization of cinnamate anion has been shown to involve such
processes.17 Such mechanisms have been proposed as possible
mechanisms in biological isomerizations.18 A more immediate
question is the relevance to the chemistry of fluorescent proteins,
particularly to the blinking phenomenon. As noted earlier, if
blinking is associated with cis/trans isomerization, blinking back
“on” requires either reverse photoisomerization or a thermal process,
and this work demonstrates that the unassisted process has too high
a barrier to compete. Conversely, the addition/elimination mech-
anism, as a bimolecular diffusional process, is also relatively slow.
In a protein, however, such diffusion is irrelevant; rather, the
question is whether there is a competent nucleophile to initiate such
a mechanism. In the particular case of mTFP0.7, fluorescence
recovery through reisomerization occurs over a span of minutes.5a
Is there a nucleophile available for this process? Indeed there is!
The crystal structure of trans-mTFP0.7 is characterized by the
presence of a water molecule interposed between a glutamate and
the benzylidene carbon of the chromophore such that the glutamate
can promote addition of water to the double bond (see Figure 4).19
Finally, we consider the isomerism of HOBDI itself. This derivative,
the closest analogy to the wild-type GFP chromophore, undergoes
isomerization in the absence of a nucleophile, although its limited
solubility in nonpolar solvents precluded studies in benzene. Curiously,
the presence of DABCO depressed the isomerization rate (see Figure
3). On the one hand, if the Falk’s mechanism were valid, the methoxy
derivative MeOBDI, with similar resonance characteristics, should
undergo equally facile isomerization. On the other hand, if deproto-
nation were required, then the base should accelerate the reaction.
Furthermore, calculations are not consistent with the deprotonated form
of HOBDI undergoing faster reaction. An alternative mechanism was
intimated by Weber, and suggested by Yang, that is, tautomerization
to a quinomethane derivative (see Path c in Figure 2).9e This
mechanism is supported by a small but measurable deuterium isotope
effect (see Figure 3 and Supporting Information). The depressive effect
of DABCO thus may be the result of a reduced equilibrium concentra-
tion of tautomer. In DMSO, we observed measurable, if slow, rates
for HOBDI and MeOBDI.13 We note that the latter cannot undergo
such a mechanism, suggesting that tautomerization is highly solvent
dependent. Does this mechanism apply within the ꢀ-barrel? We can
only speculate, although the crystal structure of mTFP0.7 shows a
favorable disposition for nucleophilic attack via Path b (see Figure 4).
Addition/elimination emerges as a compelling mechanism for
isomerization of fluorescent protein chromophores. The presence
References
(1) Petersen, J.; Wilmann, P. G.; Beddoe, T.; Oakley, A. J.; Devenish, R. J.;
Prescott, M.; Rossjohn, J. J. Biol. Chem. 2003, 278, 44626–44631.
(2) (a) Habuchi, S.; o, R.; Dedecker, P.; Verheijen, W.; Mizuno, H.; Miyawaki,
A.; Hofkens, J. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 9511–9516. (b)
Jimenez-Banzo, A.; Nonell, S.; Hofkens, J.; Flors, C. Biophys. J. 2008,
94, 168–172.
(3) (a) Haupts, U.; Maiti, S.; Schwille, P.; Webb, W. W. Proc. Natl. Acad.
Sci. U.S.A. 1998, 95, 13573–13578. (b) Liu, Y.; Kim, H.-R.; Heikal, A. A.
J. Phys. Chem. B 2006, 110, 24138–24146.
(4) (a) Chudakov, D. M.; Feofanov, A. V.; Mudrik, N. N.; Lukyanov, S.;
Lukyanov, K. A. J. Biol. Chem. 2003, 278, 7215–7219. (b) Quillin, M. L.;
Anstrom, D. M.; Shu, X.; O’Leary, S.; Kallio, K.; Chudakov, D. M.;
Remington, S. J. Biochemistry 2005, 44, 5774–5787. (c) Henderson, J. N.;
Remington, S. J. Physiology 2006, 21, 162–170.
(5) (a) Henderson, J. N.; Ai, H.-W.; Campbell, R. E.; Remington, S. J. Proc.
Natl. Acad. Sci. U.S.A. 2007, 104, 6672–6677. (b) Andresen, M.; Stiel,
A. C.; Trowitzsch, S.; Weber, G.; Eggeling, C.; Wahl, M. C.; Hell, S. W.;
Jakobs, S. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 13005–13009.
(6) Usman, A.; Mohammed, O. F.; Nibbering, E. T. J.; Dong, J.; Solntsev,
K. M.; Tolbert, L. M. J. Am. Chem. Soc. 2005, 127, 11214–11215.
(7) Solntsev, K. M.; Poizat, O.; Dong, J.; Rehault, J.; Lou, Y.; Burda, C.;
Tolbert, L. M. J. Phys. Chem. B 2008, 112, 2700–2711.
(8) Maddalo, S. L.; Zimmer, M. Photochem. Photobiol. 2006, 82, 367–372.
(9) (a) Toniolo, A.; Olsen, S.; Manohar, L.; Mart´ınez, T. J. Faraday Discuss.
2004, 127, 149–163. (b) Martin, M. E.; Negri, F.; Olivucci, M. J. Am.
Chem. Soc. 2004, 126, 5452–5464. (c) Gepshtein, R.; Huppert, D.; Agmon,
N. J. Phys. Chem. B 2006, 110, 4434–4442. (d) Altoe, P.; Bernardi, F.;
Garavelli, M.; Orlandi, G.; Negri, F. J. Am. Chem. Soc. 2005, 127, 3952–
3963. (e) Yang, J.-S.; Huang, G.-J.; Liu, Y.-H.; Peng, S.-M. Chem.
Commun. 2008, 1344–1346.
(10) To avoid confusion, we use HBDI to refer to the parent (X ) H), although
this term has been used widely to denote “HOBDI” previously.
(11) (a) Vengris, M.; van Stokkum, I. H. M.; He, X.; Bell, A. F.; Tonge, P. J.;
van Grondelle, R.; Larsen, D. S. Springer Ser. Chem. Phys. 2005, 79, 610–
612. (b) Vengris, M.; van Stokkum, I. H. M.; He, X.; Bell, A. F.; Tonge,
P. J.; van Grondelle, R.; Larsen, D. S. J. Phys. Chem. A 2004, 108, 4587–
4598.
(12) Weber, W.; Helms, V.; McCammon, J. A.; Langhoff, P. W. Proc. Natl.
Acad. Sci. U.S.A. 1999, 96, 6177–6182.
(13) Yang et al.9e report that no isomerization of HOBDI is observed in DMSO
in one day. However, under our conditions, we observed measurable
isomerizations over 2-3 days.
(14) While this paper was under review, a separate isolation of a similar
chromophore was reported: Voliani, V.; Bizzarri, R.; Nifosi, R.; Abbruzzetti,
S.; Grandi, E.; Viappiani, C.; Beltram, F. J. Phys. Chem. B 2008, 112,
10714–10722.
(15) (a) Hager, B.; Schwarzinger, B.; Falk, H. Monatsh. Chem. 2006, 137, 163–
168. Note that this report is of the N-ethyl variant. See also: (b) Fendler,
K.; Hager, B.; Falk, H. Monatsh. Chem. 2007, 138, 859–862.
(16) Um, L.-H.; Lee; J.-Y.; Fujio, M.; Tsuno, Y. Org. Biomol. Chem. 2006, 4,
2979–2985.
(17) (a) Reed, G. A.; Dimmel, D. R.; Malcolm, E. W. J. Org. Chem. 1993, 58,
6364–6371. (b) Feit, B. A.; Pazhenchevsky, R.; Pazhenchevsky, B. J. Org.
Chem. 1976, 41, 3246–3250.
(18) Cis-trans Isomerization in Biochemistry; Dugave, C., Ed.; Wiley: New York,
2006.
(19) We thank Prof. S. J. Remington for drawing our attention to this fact and
providing the figure.
JA803416H
9
14098 J. AM. CHEM. SOC. VOL. 130, NO. 43, 2008