In absorption, emission and excitation spectra for blGFP and S65T at pH 5.5 are shown. For blGFP the A form dominates both the absorption spectrum and the emission following 385 nm excitation. The Stokes loss (judged from the absorption and emission band maxima) is 65 nm. Excitation at 500 nm results in extremely weak green B* emission. The excitation spectrum suggests that this arises in part from direct excitation of a low population of the ground state B form chromophore. The A form excitation also contributes emission at this wavelength, however this is due to the broad emission profile of A* overlapping the very weak green emission, rather than ESPT. Significantly the Stokes loss for the B form is much smaller than for the A form. From these data it may be concluded that the three mutations have stabilised the chromophore in the neutral form and suppressed ESPT to the extent that it does not compete significantly with the radiative and non-radiative decay of A*. For S65T at low pH the picture is more complicated. Both A and B forms of the chromophore contribute to the absorption and emission spectra. A is dominant in absorption but B* dominates the fluorescence spectrum. However, the excitation spectra recorded in the respective A* and B* emission bands suggest that the enhanced B* component arises from the higher relative fluorescence yield of B*, rather than a fast A* → B* ESPT reaction. Once again the Stokes loss for the A form is larger than for the B form.
Figure 2 Absorption (blue), emission (green) and excitation (red) spectra of (a) blGFP at pH 6.5 and (b) S65T GFP at pH 5.5. Emission and excitation spectra of the neutral form are shown as solid lines, the anionic form as dashed lines. For both mutants, the 508 (more ...)
This picture is supported by the time resolved fluorescence measurements ( and ). In time gated fluorescence spectra recorded as a function of time after excitation are shown. For blGFP, time gated spectra decrease in intensity with increasing time but the spectral profile remains constant, suggestive of emission from a single excited state, the energy of which is not shifting on the picosecond timescale. The spectral profile is very broad, extending beyond 600 nm (as was also observed in the time gated A* emission from wtGFP(36
)). The apparent very weak vibronic structure in the A* emission probably arises from interference effects in the collection optics and is not believed to be a feature of the molecule. In general vibronic structure in the neutral GFP chromophore is only observed at low temperatures.(37
Figure 3 Time resolved fluorescence spectra (a) blGFP 3 ps (red) 10 ps (green) 50 ps (blue) 100 ps (orange) 300 ps (violet) after excitation. (b) S65T at pH 5.5 2 ps (red) 10 ps (green) 20 ps (blue) 50 ps (orange) 150 ps (violet) 400 ps (Indigo) after excitation. (more ...)
Time dependent fluorescence intensity (440 – 460 nm) for blGFP (squares) and S65T pH 5.5 (circles). Empty symbols for deuterated solvent, filled for H2O.
The broad asymmetric spectrum and large Stokes loss all point to a significant difference between absorbing and emitting state geometries for the neutral chromophore in blGFP. However, the spectral profile hardly evolves at all in the 2 – 500 ps time window observed. Thus, if the large Stokes shift reflects a time dependent geometry change in the chromophore (or its environment) following excitation it must occur on an ultrafast, probably sub-picosecond, time scale. We return to this point in the discussion of the TRIR data.
The time gated emission for S65T at low pH reveals two distinct emission bands (). The first is blue shifted with a broad spectrum, very similar to that observed for blGFP, and is thus assigned to the A* state; this emission has a much faster decay time in S65T. The second emission is most clearly resolved at long times and has a maximum at 508 nm, a narrow emission spectrum and a relatively long fluorescence lifetime. These features are characteristic of emission from the B* (anionic) form of the chromophore. This spectrum may have two origins – direct excitation or ESPT from the A* state. Direct excitation certainly plays a significant role as the 508 nm peak can be seen in the earliest time spectra. Indeed, simple subtraction of the gated spectrum of the long lived emission recorded at 400 ps (where the A* emission has decayed away) from all the earlier spectra yields an emission spectrum characteristic of the isolated A* state (inset to ). This result suggests that ESPT is not a significant process in S65T at low pH, but rather that B* is directly excited and dominates the emission at later times because of its intrinsically longer lifetime. This conclusion is consistent with the excitation spectra () and results for related mutants studied by Boxer and co-workers.(11
The kinetics of the A* fluorescence are shown in . Data were obtained from the time gated spectra integrated between 440 nm and 460 nm, thus avoiding any contribution from the long lived B* emission in S65T GFP. The A* decay of S65T GFP at pH 5.5 is non single exponential but is dominated by a fast (sub 100 ps) component, consistent with the low fluorescence quantum yield of this sample. Similar results have been obtained for other mutants in which the ESPT has been disrupted, including S65T/H148E GFP(39
), the deGFP mutants studied by Boxer et al. at low pH(11
) and Y66F GFP(1
) which lacks the phenolic hydroxyl group. Thus, in the absence of ESPT the A* state has a sub 100 ps lifetime, which may be compared on the one hand with the sub ps lifetime observed for the model compound (HBDI) in a range of solvents(20
) and on the other with the > 1 ns lifetime and high fluorescence yield of the anionic form of the chromophore. These data suggest that while the protein stabilises the A* state compared to HBDI in solution it does not do so to the extent that it has a significant fluorescence quantum yield, but only so that it has a sufficiently long lifetime to allow the ESPT in wtGFP to occur with a high yield. The most significant protein stabilisation of chromophore excited state occurs for the highly fluorescent B* (or I*) excited state, giving it its characteristic nanosecond lifetime and high quantum yield. Evidently the protein matrix is well adapted to suppress radiationless decay in the anionic form of the chromophore.
The minor longer lived component in the A* emission has an identical emission spectrum to the short lived component, and a shorter lifetime than the B* emission. Thus, it is assigned to a subpopulation of directly excited A* states which are stabilised compared to the dominant population. Equivalently, the result suggests a minority protein environment which stabilises A* in S65T GFP to a greater extent than the majority structure. The suggestion that changes in protein structure can stabilise the A* state is supported by the blGFP A* decay, which occurs on a much longer time scale than for S65T (). Thus, these data show that radiationless decay of the A* form (which most likely arises from internal conversion following excited state structural reorganisation(41
)) can be suppressed by mutagenesis. However, the origin of the stabilisation is not definitively determined. Both the T203V and E222Q mutations result in more neutral groups adjacent to the chromophore. T203V is known to favour the stabilisation of the neutral A ground state.(13
) These changes in charge state in the chromophore environment may also contribute to stabilisation of the excited A* state in the blGFP mutant.
More mechanistic detail on the processes underlying GFP photodynamics can be obtained from TRIR spectroscopy.(8
) In the TRIR spectra for blGFP, S65T GFP at pH 5.5 and the neutral model chromophore HBDI are presented. TRIR spectra of the model compound and wtGFP have been discussed in great detail elsewhere.(35
) The three strong bleaches seen for the model compound can be assigned to vibrational modes of the ground state chromophore mainly comprising (with decreasing wavenumber) the C=O stretch, the C=C stretch and a phenyl ring localised mode (with weaker phenyl and C=N modes at higher frequency). These three ground state modes are all evident in the wtGFP spectrum(35
) and in the two mutants presented in , although the intensity distribution is significantly modified through interaction with the protein; specifically the relative intensity of the carbonyl and C=C modes to the phenyl mode are increased for the A state in the protein.
Figure 5 (a) TRIR Spectra for HBDI in DMSO 2 ps (red) 4 ps (green) 6 ps (blue) 10ps (orange) 30 ps (violet) after excitation (b) S65T at pD 5.5 2 ps (red) 6 ps (green) 10 ps (blue) 30 ps (orange) 100 ps (violet) after excitation (c) blGFP 2 ps (red) 4 ps (green) (more ...)
The additional complexity in the TRIR spectra on moving from HBDI to wtGFP can largely be explained by the contribution of protein modes (particularly those associated with E222) modified by the proton relay reaction, which leads to a number of new time dependent bands.(8
) No such additional kinetics can occur in blGFP and S65T GFP at pH 5.5 where the proton relay reaction has been blocked, and yet the TRIR of these mutants have a number of features which are absent in neutral HBDI. In particular a broad bleach with a complex profile is observed around 1630 – 1650 cm−1
and a broad transient absorption is observed below 1580 cm−1
. By comparison with the spectrum for HBDI much of this bleach and transient absorption should be assigned to protein modes, although chromophore C=C and C=O* both contribute in this wavenumber range.(39
) Evidently excitation to the A* state causes a number of protein modes to shift to lower frequency.
In contrast to wtGFP the time dependence of the TRIR spectra for the two mutants is relatively straightforward with a prompt (pulse limited) simultaneous appearance of both bleach and transient modes of both the chromophore and the protein, followed by recovery of the ground state on the time scale expected from the time resolved fluorescence (). This is as expected for mutants in which the ESPT reaction is blocked, so that the excited A* state decays back to the ground state by fluorescence and radiationless internal conversion.
The assignment of the new transients in the blGFP and S65T at pH 5.5 TRIR has been further investigated through a study of the anisotropy of the TRIR spectrum of blGFP. The parallel and perpendicular polarisation resolved data recorded 4 ps after excitation are shown in , along with their ratio. In the absence of molecular reorientation (which can be assumed to be negligible on the 4 ps time scale) the ratio is sensitive to the angle between the electronic transition dipole of the chromophore and the direction of the transition moment of the vibration probed.(44
) It can take values between 3 and 0.5 for an angle varying between 0° and 90°. From these data the assignment of the highest wavenumber bleach and the intense polarisation sensitive 1600 cm−1
mode to the chromophore carbonyl and phenyl ring mode respectively can be made unambiguously through a comparison with the data in ref. (35)
. The band near 1640 cm−1
with an anisotropy reaching 2 probably has a strong component of the C=C stretch. The sharp polarisation sensitive transient absorption on the high frequency side of the broad band around 1560 cm−1
may be associated with the phenyl 1 mode in the excited state (again by comparison with anisotropy data for HBDI and Figure 6 ref.(35)
The other bleach modes around 1620 to 1660 cm−1
and the majority of the transient absorption near 1560 cm−1
are assigned to overlapping contributions of protein modes perturbed by their interaction with the A* excited state and chromophore localised modes. Both the bandshape and wavenumber dependence of the anisotropy are complex for these bands, suggesting the presence of more than one vibrational mode. These modes are evidently a feature associated particularly with the A* state in mutants where the proton wire has been disrupted (we have made essentially identical observations in S65T/H148E GFP, which also does not support ESPT at low pH(39
)). In contrast, the array of shifted modes (i.e. the transients around 1560 cm−1
) are not observed in either wtGFP or T203V GFP (both of which do undergo ESPT) even at the earliest times after excitation, prior to significant proton transfer, when the A* state dominates. In these ESPT active mutants the alterations in protein modes not associated with the protonation of E222 are largely restricted to small shifts around 1630 – 1690 cm−1
) Thus it may be concluded that the large shift in protein modes revealed in is associated with a perturbation by the A* state of residues involved in the disruption of the proton wire. These modes fall broadly in the frequency range expected for amide bands. The perturbation is mainly a frequency downshift, which suggests a weakening of the bonds on A* excitation. This may be explained by a change in H-bond interaction strength between the chromophore and its environment immediately following electronic excitation, although in the absence of a crystal structure or isotopic substitution in residues adjacent to the chromophore a more definite assignment is not possible
A significant mutation in disruption of the proton wire is S65T, as was found in the ESPT inactive mutants studied by Remington, Boxer and co-workers.(11
) It does not however necessarily follow that the perturbed modes are directly associated with modes of T65. For example they are also seen in blGFP (which has the S65G mutation) but not in S65T/H148D GFP, in which very rapid proton transfer occurs (probably from the chromophore directly to D148(14
)). A more probable assignment of these protein modes is to a group of residues interacting with the chromophore through the extensive H-bonding network in GFP, which has been perturbed during disruption of the ESPT chain and the subsequent changes in protonation state. A more definitive assignment of these protein modes would require measurements on isotope edited mutants.