Observation of low frequency vibrational modes in a mutant of the green fluorescent protein

Coherent vibrational dynamics in a mutant of the green fluorescent protein are studied by diffractive optic ultrafast transient grating spectroscopy.

Two low frequency (<100 cm−1) modes are observed.

We propose that one at 95 cm−1 may be associated with intramolecular torsional motion in the excited state.

The possible importance of such modes in understanding the very high quantum yield of the intact GFP is discussed.

The green fluorescent protein (GFP) is a non-invasive genetically encodable fluorescence marker protein.

It has found numerous and wide-ranging applications in molecular and cell biology.1

The fluorophore in wild-type (wt) GFP has been identified as 4′-hydroxybenzylidene-imidazolinone, formed through a cyclisation and oxidation reaction of three adjacent amino acid residues, Ser65–Tyr66–Gly67.2,3

The resulting product is highly stable and strongly fluorescent.

The chromophore is protected from the environment by its location at the centre of an end capped β-barrel structure, and it is held in place by both covalent bonds and an extensive H-bonding network, involving the protein backbone and adjacent water molecules.4,5

The stability and the high fluorescence yield of the chromophore are consequences of its interactions with the surrounding protein structure.

When the protein is denatured the fluorescence yield decreases by at least three orders of magnitude.6,7

In addition, synthetic analogs of the GFP chromophore are essentially nonfluorescent in solution at room temperature.7

We investigated the radiationless decay of the GFP chromophore by ultrafast polarisation spectroscopy8–11 and time resolved fluorescence.12,13

The radiationless decay mechanism is ultrafast internal conversion, due to intramolecular motion about the bridging bond of the chromophore in its excited state.8–11

The isomerization coordinate was shown to be near barrierless and, from the weak dependence of lifetime on medium viscosity, volume conserving.11–13

Recent theoretical calculations are in broad agreement with this mechanism, and provide additional information on the nature of the isomerization coordinate and excited electronic states.14,15

The outstanding challenge in developing a complete picture of the photodynamics of GFP is in understanding the nature of the protein–chromophore interactions which result in such a dramatic suppression of radiationless decay.

It was proposed that in the excited state two coordinates are involved in coupling the ground and excited states,13,15 a two-state two-mode model.16

One co-ordinate involves ultrafast motion out of the Franck–Condon region along a high frequency stretching coordinate.

This stretching motion is coupled with motion along a low frequency torsional coordinate, which brings ground and excited states close in energy.

This suggests that important information on the photodynamics can be obtained from vibrational spectroscopy.

There have been a number of studies of the vibrational spectroscopy of GFP and its isolated chromophore,17–21 but these have focussed on modes above 400 cm−1.

For example resonance Raman spectroscopy reveals a strong mode around 1560 cm−1, which may correspond to the initial stretching coordinate taking the chromophore out of the Franck–Condon region.17,19

However, both theory and experiment suggest that low frequency torsional modes are important in understanding the radiationless decay mechanism.13,15

These present a particular challenge to conventional Raman spectroscopy, because of experimental problems associated with scattered light, and fundamental problems involving thermal populations.

Thus the important low frequency region of the GFP vibrational spectrum has not been investigated in any detail.

While low frequencies are problematic for frequency domain spectroscopy they are quite readily observed by time domain ultrafast vibrational coherence spectroscopy.22,23

In this method a short coherent light pulse, the pump, excites the S0 → S1 transition of a solute.

The pulse impulsively excites coherent intramolecular motion whenever the pulse width is smaller than c/2πν, where ν is the frequency of an intramolecular mode.

Coherent motion in the sample may be either directly excited on the S1 surface, or excited by impulsive stimulated resonance Raman scattering in S0.22,23

The resultant coherent vibrational motion in the sample modulates its optical properties, which are probed by a time delayed probe pulse.

There exist two previous reports of coherent vibrational dynamics in GFP.

Boxer and co-workers reported ultrafast fluorescence of wtGFP, with 100 fs time resolution, and noted the appearance of a weak oscillatory signal in the first picoseconds;24 they did not extract a vibrational frequency.

More recently Cinelli et al. reported ultrafast pump–probe spectroscopy with 10 fs time resolution on the EGFP mutant of GFP.25

They found a strong oscillation damped on a 1 ps time scale, at the relatively high frequency of 497 cm−1.

This was assigned to vibrational dynamics in the excited state.

In the following we describe a transient grating experiment with 60 fs time resolution, which is significantly lower than that of Cinelli et al., so we cannot access modes with frequencies >400 cm−1.

However, the signal-to-noise attainable with the background free TG measurement is very high.

Thus we have been able to record coherent vibrational dynamics of lower amplitude and lower frequency than was accessible in earlier experiments.

To optimise the signal-to-noise ratio we used a diffractive optics based TG technique, which amplifies the TG signal while reducing jitter.

Our experiment is based on the designs of Miller,26 Fleming27 and their co-workers (Fig. 1).

Briefly the ca. 500 nm 1 μJ pulses from a non-collinear optical parametric amplifier (NOPA), pumped by a 1 kHz amplified Ti:sapphire laser (Clark MXR), were divided at a beam splitter into pump and time-delayed probe beams.

These pulses had a width of 60 fs at the sample position.

The linearly polarised pump and probe pulses were combined with a 300 mm lens on the surface of an 80 groove mm−1 transmission grating (Edmund Optics), which split each into two replicas.

The four beams were focussed onto the sample using a spherical mirror (radius 50 cm), via a folding mirror.

Beams 3 and 4 (pump) write a holographic grating in the sample.

The decay of the grating is monitored by the intensity of the signal scattered from the time delayed probe pulse 1 in the direction 2.

The intensity is measured by a photodiode placed behind an analysing polarizer.

The intensity of beam 3, and thus the formation of the grating, was modulated by a mechanical chopper.

The photodiode output was measured by a lock-in amplifier referenced to the chopper frequency.

The data presented below were measured in the homodyne configuration (with beam 2 blocked) and so report the square of the sample's third order nonlinear response function.

A 100 mg ml−1 solution of the EGFP mutant (S65T/F64L) of GFP (Clontech, used as received) was contained in a 1 mm pathlength cell.

In EGFP the anionic form of the chromophore dominates, which has a strong transition at 490 nm, nicely matching the output of the NOPA.

The absorbance was approximately unity.

The high absorbance and absence of a measurable signal from the solvent suggest that the TG signal arises from sample dichroism.

Before each TG measurement the pulsewidth from the NOPA at the sample position was carefully optimised by adjusting a dispersive prism pair.

It should be noted that without such optimisation the coherent vibrational dynamics described below could not be reliably observed.

Others have shown that the observation of vibrational coherences is dependent on the degree of chirp in the pulses.23

The average of 50 TG measurements is shown in Fig. 2a.

Data were measured at 510 nm, on the low energy side of the S0 → S1 spectrum.

The rise of the TG signal is limited by the laser pulsewidth, and reaches a maximum at t = 0.

The signal decays to a near constant level in <200 fs.

Superimposed on this is an oscillatory component.

The oscillatory part of the response is assigned to coherent vibrational motion in EGFP.

The non-oscillatory response can be fit to ultrafast (ca. 100 fs) and ca. 10 ps exponential relaxation times, plus a dominant constant term.

The ‘constant’ is ascribed to the nanosecond excited state lifetime.

The remaining ultrafast and picosecond relaxation times are assigned to intramolecular vibrational redistribution and vibrational cooling in the excited state respectively.28

Since we are mainly interested in the coherent vibrational dynamics we subtracted these exponential components from the experimental data to isolate the residual oscillatory response (Fig. 2b).

The dominant component in the residual data is an oscillation with a period of 370 ± 20 fs, or 95 cm−1, which persists for at least 2 ps.

The amplitude of the oscillation is not regular, but the data are well reproduced when a second component of approximately 35 cm−1 is included (the lower frequency is not accurately determined, as data are available for at most two cycles).

The fit with these two components is shown as the solid line in Fig. 2b.

The same frequencies were extracted from measurements at 497 nm.

The frequencies reported in Fig. 2 are lower than any previously observed for GFP.

They are also lower than any frequency calculated for the ground state, though Tozzini et al. found deformation modes around 200 cm−1 in their density functional theory and Car–Parrinello MD simulations.20

In a CASSCF calculation of the excited state of the GFP chromophore Olivucci and co-workers reported three modes lying between 100 and 160 cm−1.15

These were associated with bending, torsion and pyramidalization in the bridging bonds.

These are the modes which appear to be involved in promoting ultrafast internal conversion in the chromophore.

It is difficult to definitively assign the frequencies observed in Fig. 2 to either the ground or excited state, since both may be excited in vibrational coherence spectroscopy.

However, the higher frequency oscillation is slow (370 fs) compared to the 60 fs pulsewidth.

This would favour an assignment to a vibrational mode in the excited state.23

The poorly resolved low frequency mode may be in either the ground or excited state.

The frequency seems too low to be associated with an intramolecular mode, and may reflect orientational (librational) motion of the chromophore, or some other slow dynamics, within the protein matrix.

If, as we suggest, the higher frequency oscillation reflects the excitation of coherent intramolecular torsional motion in the excited state, then two factors are of note.

First, the vibrational coherence survives the initial relaxation out of the Franck–Condon excited state.

This relaxation is responsible for a 2000 cm−1 Stokes loss in <100 fs in the free chromophore in solution,13 but the Stokes shift is greatly reduced in the intact protein.

Second the torsional mode in the intact protein is damped on the picosecond timescale, whereas in the free chromophore it promotes sub-picosecond internal conversion.

One possible mechanism for the suppression of radiationless decay in the protein is that protein–chromophore interactions lead to a modification of the torsional potential such that it is more harmonic and of higher frequency than in the free chromophore, and therefore couples less effectively with the ground state surface.

Alternatively, the protein–chromophore interaction may modify the vibrational coordinate responsible for relaxation out of the Franck–Condon excited state in such a way that it cannot access that region of the potential energy surface from which ultrafast internal conversion occurs.

Such explanations require further experimental and theoretical study.

A TG study of additional mutants of GFP is planned.

In conclusion, we have used a sensitive diffractive optics based TG method to investigate the low frequency modes of a GFP mutant.

A mode at 95 cm−1 is detected, which is lower in frequency than any previously reported mode in GFP.

This is tentatively assigned to excitation of torsional motion on the excited state surface.

The possible implications of this observation for understanding the enhancement of the fluorescence yield of the chromophore by the protein matrix are discussed.