The spectroscopic results for the two double mutants reveal a mechanism for ESPT which differs greatly from that of wtGFP. Specifically the ground state A → A* spectrum is perturbed, the proton transfer is ultrafast (probably sub-picosecond, –) and insensitive to deuteration () and no distinct carbonyl mode can be assigned to the formation of the protonated acceptor. Rather than proton transfer through a network to a distant receptor, as occurs in the wild type protein, the present results suggest an interpretation based on the existence of a low barrier H-bond between D148 and the chromophore in which the proton participates in a polar covalent bond with both oxygens ().33
This arrangement allows barrierless ESPT upon photoexcitation within a strongly coupled chromophore D148 complex, as illustrated in and discussed further below. Recent work by Remington and co-workers on the structure of S65T/H148D GFP provides concrete evidence of the presence of such a low barrier H-bond in these mutants. This group has resolved the structure to 1.5 Å, revealing an extremely short (≤ 2.4Å) H-bond connecting the chromophore to D148 (S.J. Remington, personal communication).
Comparison of the proposed H+ transfer routes in wtGFP and S65T/H148D upon excitation.
Figure 11 A representation of the proposed potential energy surfaces for wtGFP (A) and S65T/H148D GFP (B). In wtGFP the phenolic OH bond is shorter and the acceptor remote. In S65T/H148D the donor and acceptor are close yielding the low barrier and flat potential (more ...)
The interaction between the chromophore and D148 is evident in the TRIR spectrum, where the formation of a new H-bond is suggested by the perturbation of the phenol mode at 1600 cm−1 (). We interpret the observed broadening as indicative of either static or dynamic disorder, due to small changes in geometry around the shared proton, leading to significant shifts in the frequency of the phenol localized mode. The chromophore D148 interaction is also evident in the red shifted electronic absorption spectrum of the A → A* transition. Raman and TRIR data suggest that ground state modes not associated with the phenol ring are unperturbed by D148. Thus the spectral shift mainly arises from stabilization of the chromophore excited electronic state. This is interpreted as arising from the formation of an extended H-bond leading to a localization of charge on the phenolic oxygen, such that the phenol develops a more quinoidal structure than is found in S65T or wtGFP.
Another major difference in the TRIR spectra between S65T/H148D and wtGFP is the absence of an excited state absorption at 1560 cm−1 in the former. A band is observed at this position in S65T, as well as in wtGFP, and is assigned to the C=O absorption of A* and protein modes perturbed by interaction with it in both S65T and wtGFP. The absence of this band in S65T/H148D is consistent with the rapid formation of I* from A* within the time resolution of the TRIR measurements.
In both H148D mutants a temporal evolution in the protonation state of D148 was anticipated in TRIR, based on previous observations in wtGFP.19
However, this is not observed. No transient C=O stretch mode due to the formation of aspartic acid was seen (expected above 1716 cm−1
; losses of the asymmetric (1576 cm−1
) and symmetric (1410 cm−1
) stretch modes of the ionized aspartate were also not observed, even though I* is formed promptly and exists for several hundred picoseconds. The absence of any change assignable to the protonation state of the carbonyl of D148 is unexpected (unlike the weak or undetectable contribution of I* which has been noted for other mutants10
). It suggests that the spectra of the initial ionic and final neutral states of D148 are not very different. We propose that the presence of a low barrier H-bond in the ground state can explain the lack of evolution in our TRIR data. The polar covalent nature of the short H-bond between the phenolic proton and D148 implies the presence in the ground state of the neutral-like aspartic acid. Upon excitation, D148 does not change protonation state, but merely binds the proton more tightly. Thus no large shifts are observed in D148 vibrational modes. The proton-phenol oxygen bond however does change upon excitation, converting from a polar covalent to a hydrogen bond in an ultrafast barrierless process (). This allows for the instantaneous appearance of the anionic chromophore and green fluorescence. This is very different from the situation in wtGFP, where the proton acceptor, E222, converts from the ionized to the acid form upon chromophore excitation, with the conversion observed in the TRIR spectra.
An important question with regard to the mechanism is whether or not A* fluoresces prior to proton transfer. While the narrowing at the blue edge of the emission spectrum might indicate the presence of short-lived A* fluorescence, the absence of an isoemissive point argues against a simple two state model. In addition the lack of an observable isotope effects argues against a distinct A* state feeding the anionic form in an activated process. Finally there is no corresponding 10 ps rise-time of the 508 nm emission, unlike that seen for wtGFP.16, 17
We propose therefore, on the basis of both the undetectable A* emission in gated fluorescence with 4ps time resolution and the absence of A* related modes in TRIR with 400 fs time resolution, that A* does not fluoresce directly but is quenched on a sub picosecond time scale with a major fate being ESPT. A consequence of the ultrafast proton transfer is the formation of a vibrationally hot excited state population in the initial I* state. This is proposed to be the source of the higher energy emission, which relaxes as the excited state cools on the picosecond timescale through vibrational energy transfer to the protein matrix. A time scale of 10 ps is typical for vibrational cooling.
The distinction between absorption and excitation spectra () suggests a further level of complication in the photophysics of S65T/H148D GFP. The enhanced emission from the directly excited B state is explicable either because A* has an additional decay route competing with ultrafast ESPT or because the vibrationally hot I* state can transform into a nonradiative conformation which is not accessible from B*. In either case it suggests a new degree of control over the fluorescence emission of GFP mutants.
These ideas are summarized in a representation of the reaction coordinate associated with the proton transfer reaction shown in . To a certain extent the picture used here is borrowed from extensive investigations of excited state intramolecular proton transfer, which has been studied in the condensed and gas phase for a number of systems.35–37
These ideas can be extended to intermolecular proton transfer in the case of proteins because the structure of the protein holds the reaction partners in place, whereas in solution they would be free to diffuse apart. Significantly, it is this feature of S65T/H148D GFP photophysics which makes it suitable as a model system to study low barrier or barrierless proton transfer reactions in proteins. In intramolecular systems the proton transfer may occur on a sub-picosecond timescale and be insensitive to hydrogen/deuterium exchange,38
a situation mirrored for the intermolecular (in the sense of chromophore to D148) proton transfer reactions that occur in the GFP double mutants.
In the proposed reaction coordinates for wtGFP and double mutant proton transfer are contrasted. For wtGFP the multi-step proton relay reaction is represented by a single effective barrier between the chromophore donor and the E222 acceptor, since the precise shape of the reaction coordinate is not known. The key factor is that there is a significant barrier to proton transfer in the ground state such that the A form is the most stable. In S65T/H148D the distance between the O atoms in the donor chromophore and the acceptor is far shorter than any of the distances between O atoms along the H-bond network in wtGFP. Even if the donor acceptor proton affinity are unchanged this has the effect of producing a lower barrier, stretching both the OH bond of the chromophore and probably modifying the C=O of the acceptor. The lower barrier will in turn imply a greater range of distances are accessible to the OH bond at a given temperature, possibly contributing to the observed broadening of the chromophore phenyl mode. In the figure the pKa for the donor and acceptor are assumed unchanged, so the proton nevertheless remains localized on the chromophore, such that it remains in the A state. It should be emphasized that is a simplified one dimensional representation of what is in reality a multidimensional surface. Indeed theory and experiment for the intramolecular case show that the probability of proton transfer is likely to be modified by vibrational modes which modulate the distance between the O atoms involved.35
Upon electronic excitation the pKa of the chromophore is reduced.39, 40
In the case of wtGFP the effect is to make the anionic I* form more stable. However, the proton transfer proceeds from A* to I* through a barrier, leading to the observed picosecond lifetime of the A* and its sensitivity to isotopic exchange. The effect of the same pKa change for the S65T/H148D is to make the proton transfer barrierless, such that the I* state is reached on a sub-picosecond time scale after excitation of A*, by essentially a translation of the proton between the two O atoms. This mechanism is consistent with the absence of deuterium isotope effect and the prompt rise of the 508 nm emission observed. It is our proposal that the similarity of the initial and final structures, suggested by the absence of a product state contribution in the TRIR spectra, is also consistent with the strong low barrier H bond formed in the ground electronic state.
The sub-picosecond proton transfer means that reaction occurs faster than the excess energy can be dissipated, so the initial I* state is formed vibrationally hot. The hot initial state may either transform to a non-radiative state (accounting for the relative inefficiency of emission from A* excitation compared to direct B* excitation) or cool on a 10 ps timescale. Vibrational cooling results in the narrowing of the emission spectrum with time and the observation of a fast component on the blue edge of the time resolved fluorescence spectra.
Although somewhat speculative this model is consistent with the range of experimental observations reported. Of course the potential energy surfaces displayed need to be supported by detailed high level quantum chemical calculations. Even if such surfaces become available (in itself a challenging objective) the subsequent calculation of the ultrafast rate coefficients will require non equilibrium dynamics calculations.