Mapping the ultrafast vibrational dynamics of all-trans and 13-cis retinal isomerization in Anabaena Sensory Rhodopsin

Partha Pratim Roy a, Youshitoka Kato b, Rei Abe-Yoshizumi bc, Elisa Pieri d, Nicolas Ferré d, Hideki Kandori bc and Tiago Buckup *a
aPhysikalisch-Chemisches Institut, Ruprecht-Karls Universität Heidelberg, D-69210, Heidelberg, Germany. E-mail:
bDepartment of Life Science and Applied Chemistry, Nagoya Institute of Technology, Showa-ku, Nagoya 466-8555, Japan
cOptoBioTechnology Research Center, Nagoya Institute of Technology, Showa-ku, Nagoya 466-8555, Japan
dAix-Marseille Univ, CNRS, ICR, Marseille, France

Received 28th August 2018 , Accepted 15th November 2018

First published on 15th November 2018

Discrepancies in the isomerization dynamics and quantum yields of the trans and cis retinal protonated Schiff base is a well-known issue in the context of retinal photochemistry. Anabaena Sensory Rhodopsin (ASR) is a microbial retinal protein that comprises a retinal chromophore in two ground state (GS) conformations: all-trans, 15-anti (AT) and 13-cis, 15-syn (13C). In this study, we applied impulsive vibrational spectroscopic techniques (DFWM, pump-DFWM and pump-IVS) to ASR to shed more light on how the structural changes take place in the excited state within the same protein environment. Our findings point to distinct features in the ground state structural conformations as well as to drastically different evolutions in the excited state manifold. The ground state vibrational spectra show stronger Raman activity of the C14–H out-of-plane wag (at about 805 cm−1) for the 13C isomer than that for the AT isomer, which hints at a pre-distortion of 13C in the ground state. Evolution of the Raman frequency after interaction with the actinic pulse shows a blue-shift for the C[double bond, length as m-dash]C stretching and CH3 rocking mode for both isomers. For AT, however, the blue-shift is not instantaneous as observed for the 13C isomer, rather it takes more than 200 fs to reach the maximum frequency shift. This frequency blue-shift is rationalized by a decrease in the effective conjugation length during the isomerization reaction, which further confirms a slower formation of the twisted state for the AT isomer and corroborates the presence of a barrier in the excited state trajectory previously predicted by quantum chemical calculations.


Photoinduced isomerization of the retinal protonated Schiff base (RPSB) powers many fundamental biological processes like photosynthesis, vision and gene expression by initiating the photocycle via the conversion of light energy into chemical potential.1–8 The retinal derivative, which serves as the central element for photochemically triggered biological activity, is found to be covalently bound to a lysine residue of the seventh helix of the protein membrane in all retinal pigments.9,10 Retinal is well known for its different structural isomeric forms in a variety of organisms, e.g. 11-cis in visual rhodopsin, 9-cis in iso-rhodopsin, all-trans in microbial retinal protein (MRP) and the widely varying reaction rates and the isomerization efficiencies in these proteins. For instance, the retinal isomerization in visual pigments takes place within 100 fs11–14 with a quantum yield about 65%,15 whereas it takes more than 0.6 ps to complete the isomerization with a yield <40% for MRP's.16–18 Moreover, the photoisomerization of RPSB (all-trans) in solution happens at a much slower rate (∼1 ps)19–21 and leads to the formation of a mixture of different stereoisomers with a quantum yield of a few percent for each subproduct.22 All these observations have been taken as an indicator that the electrostatic interaction between the retinal chromophore and the amino acid residues of the surrounding proteins steer the isomerization reaction.

In recent decades, there have been extensive experimental12,14,23–29 and theoretical30–34 investigations performed to explore the underlying mechanism of this ultrafast isomerization reaction. Visual rhodopsin and bacteriorhodopsin (BR) are two retinal proteins that have been most thoroughly studied in this regard. In general, the retinal isomerization is believed to be initiated by a rapid reorganization of the C–C bond length within a few tens of femtoseconds after the excitation, known as bond length alternation (BLA),32 to form the sub-picosecond reactive excited state (ES) coined as the I intermediate.35 For BR, this state decays non-exponentially to form the J intermediate35,36 during the course of internal conversion (IC), which is associated with a weak spectral evolution.37,38 Moreover, the cross-section of the emission state has been found to be constant throughout the fluorescence lifetime.38 Both of these observations have been interpreted as a non-ballistic IC. Retinal isomerization in visual rhodopsin is, however, different: drastic spectral evolution during the IC12,14,27 has been interpreted as a coherent isomerization reaction,12,14,39 where a nuclear wave packet generated by impulsive stimulated Raman scattering (ISRS) evolves in a ballistic fashion towards the ground state (GS) potential surface.

This contrast between the nature and speed of the isomerization reactions for two different classes of retinal proteins has been often rationalized by the difference in the planarity of the GS structures of RPSB bound inside the cavity of the retinal pocket.40 Structural strains on RPSB inside different protein pockets originate either due to distinct structural conformations of each isomer or due to the change in the electrostatic interaction between RPSB and the protein moiety. Structural investigations9,40,41 have shown that the 11-cis GS isomer in visual rhodopsin is present as a non-planar, pre-twisted structure, which facilitates the reaction proceeding rapidly.42 However, for BR, neither NMR studies40,43 nor the resonance Raman spectra44 indicate any evidence of such a pre-twisting for the GS all-trans isomer. Furthermore, a quantum chemical computational study42 of an artificially twisted retinal in the gas phase showed sub-100 fs dynamics, which further indicates that the pre-straining alone can lead to a rapid isomerization process even in the absence of any specific electrostatic interaction of RPSB with the surrounding.

One systematic way to point out the reason would be to investigate the RPSB of two different conformations under the same protein environment. Recently discovered Anabaena Sensory Rhodopsin (ASR) has been assisting in shedding new light on this issue.7,9,45–47 Like other MRPs, it comprises 13-cis, 15-syn (13C) and all-trans, 15-anti (AT) in GS. Being a photochromic sensor, the isomeric ratio depends on the wavelength of external illumination light.45 When it is illuminated with an orange light (∼590 nm) it forms a mixture (∼40[thin space (1/6-em)]:[thin space (1/6-em)]60) of AT and 13C isomers in light adapted (LA) photo-stationary equilibrium.46,47 On the other hand, it adopts a thermally relaxed AT form under dark-adapted (DA) conditions. Within the photocycle (Fig. 1), each of the AT and 13C isomers undergoes isomerization around C13[double bond, length as m-dash]C14 in a sub-ps timescale that results in a hot photo-intermediate J (13-cis, 15-anti and all-trans, 15-syn, respectively). Later, it forms the K-photoproduct (KAT and K13C) within 100 ps via vibrational relaxation. Subsequently, both KAT and K13C undergo further isomerization around C15[double bond, length as m-dash]N on a longer timescale (<nS) to generate the 13C and AT GS, respectively (Fig. 1). Therefore, ASR is a unique member of the rhodopsin family, which allows to compare the isomerization reaction in both directions (AT to 13C and 13C to AT) of the photocycle within the same protein environment. This is extremely advantageous for evaluating the individual effects of the GS conformation on the isomerization dynamics. Pump–probe spectroscopy has already shown the stark contrast in the reaction kinetics of AT and 13C isomers in ASR.46,47 It has been observed that the 13C isomer shows ballistic kinetics46 and the isomerization completes within 100 fs, which is very similar to visual rhodopsin.46,47 The AT isomer shows, however, about sevenfold (about 750 fs)46,47 slower kinetics, which is reminiscent of BR. This large disparity in the reaction dynamics has been qualitatively explained by the quantum chemical excited trajectory calculation, which suggests the existence of a small barrier or plateau in the excited potential energy surface of the AT isomer but not for the 13C isomer.31 The quantum yield of the photoproduct after the isomerization for each direction is also very different: it is about 2.7 times higher for the AT than that of the 13C isomer.48 Lower quantum yields with short excited state lifetimes, often observed also in other proteins, have often been interpreted as a discrepancy lacking clarification. Ultimately, the excited state lifetime and quantum yield are two fundamentally independent quantities: while the former depends on the topology near the Franck–Condon region and potential barriers in the ES surface, the latter depends on the shape of the conical intersection seam and the wave packet trajectories. A potential way to address this persisting question is the comparison of isomer-specific ultrafast structural changes in each case, which will give more insight into the molecular origin of this difference in potential energy surfaces.

image file: c8cp05469j-f1.tif
Fig. 1 Schematic representation of the photocycle of the AT and 13C isomers of ASR. After excitation, each isomer undergoes isomerization around the C13[double bond, length as m-dash]C14 bond in the sub-ps timescale to form the corresponding hot photo-product, JAT and J13C, which generate KAT and K13C within 100 ps, respectively. Subsequent isomerization around C15[double bond, length as m-dash]N occurs on the sub-ns timescale to complete the photo-cycle.

In the context of the retinal isomerization mechanism, a few vibrational modes such as torsion, C[double bond, length as m-dash]C and C–C stretching, and hydrogen-out-of-plane (HOOP) wags are well known to participate in this reaction. A detailed study of the origin and evolution of the vibrational modes in the GS and ES has been reported mainly for visual rhodopsin,23,28,49 BR24,25,50 and RPSB in solution.51–54 Low-frequency torsional modes (<400 cm−1) are well accepted as one of the key features of ES wave packet dynamics. In particular, a set of rapidly damped low-frequency coherences (100–400 cm−1) has been observed in the spectral region of stimulated emission and excited state absorption.28,50,52–57 Pump-impulsive vibrational spectroscopic studies28,53,56 on RPSB both in solution and inside the protein environment have shown that these modes are activated only after the initial excited relaxation of RPSB. Besides the low-frequency torsion, the HOOP wag (800–1000 cm−1) is the most frequently discussed vibrational mode. This out-of-plane mode, being Au/A2 symmetric, is Raman inactive for RPSB of planar geometry, which belongs to the C2v (cis) or C2h (all-trans) point group and gets Raman activated only when the RPSB adopts a distorted non-planar structure. Hence, the activity of this mode is an indicator of the non-planarity or pre-twisting of the RPSB in GS, as well as the degree of distortion during the isomerization due to the rotation around the isomerizing C[double bond, length as m-dash]C bond.23,28,58 Moreover, a number of experimental observations28,59 and quantum chemical computations32,60 have also suggested that the phase and amplitude of the HOOP mode control the stereochemical outcome of the retinal photo-isomerization. Finally, high frequency (>1000 cm−1) wave packet dynamics have been very often interpreted as GS wave packet motion,25,61 with very few reports62,63 proposing that the conformational changes during the isomerization result in the modulation of the amplitude and frequency of the high-frequency fingerprint modes.

Herein, we apply (multidimensional) time-resolved vibrational spectroscopic techniques, such as degenerate four-wave mixing (DFWM),50,52 pump-degenerate four-wave mixing (pump-DFWM)53,64–66 and pump-impulsive vibrational spectroscopy (pump-IVS),28,58,65,67 to compare the vibrational dynamics of the AT and 13C isomers of ASR. Tracking the excited state evolution of the nuclear wave packet shows (i) a delayed appearance of low frequency (100–400 cm−1) modes and (ii) a blue-shift for two characteristic high-frequency modes, namely, the CH3 rocking (∼1000 cm−1) and C[double bond, length as m-dash]C stretching (∼1500 cm−1) modes. The delayed rise in the low frequency (100–400 cm−1) modes supports the indirect activation mechanism of the delocalized torsional modes by localized high frequency (>1000 cm−1) modes, such as C[double bond, length as m-dash]C stretching, via internal vibrational energy redistribution (IVR).50,53 Experimentally observed blue frequency shifts during the course of isomerization hint at a reduction in the effective π-conjugation length due to the formation of the non-planar 90° twisted state. Most remarkably, AT shows a slower frequency shift than 13C, which, according to our interpretation, indicates that AT reaches this twisted state much later than the 13C isomer. In addition, a relatively stronger amplitude of the HOOP mode in the GS of 13C compared with that of AT has been observed both in non-resonant DFWM, IVS and spontaneous Raman spectra, which indicate that the GS of the 13C isomer is more pre-twisted than the AT isomer of the RPSB in ASR.

Experimental methods and analysis

Sample preparation

The ASR sample was prepared according to the standard reported protocol.47 A purified sample was concentrated and dialyzed against a buffer solution containing 200 mM NaCl and 25 mM Tris–HCl to maintain a pH of 7.0, and 0.01% DDM (n-dodecyl-β-D-maltopyranoside, Anagrade, Anatrace) was added to help the protein to get stabilized by forming micelles. For pump–probe experiments, the concentration of DDM was reduced (still above the CMC of DDM: 0.007%) as compared with that used in the previous report,47 in order to minimize the formation of bubbles while circulating it through the flow cell during the measurement.

Time-resolved experiments

The experimental setup used for the pump-DFWM and pump-IVS study is shown in Fig. S1 and S2 (ESI), respectively. Briefly, a regeneratively amplified titanium–sapphire laser system (795 nm, 1 kHz) was used as a fundamental laser source. The output was equally divided into two home-built non-collinear optical parametric amplifiers (nc-OPA's) to generate the ultrashort actinic-pulse (AP) and degenerate pump/stokes/probe (DFWM) or push/probe (IVS) spectra (Fig. 2). In pump-DFWM and pump-IVS, the spectrum of the actinic pulse was spectrally resonant with the GS absorption (Fig. 2(a)), while the DFWM/IVS spectra overlapped with the photo-induced absorption (PIA) band of ASR (Fig. 2(b)). All pulses were compressed below 15 fs by a prism pair (Fig. S3, ESI).
image file: c8cp05469j-f2.tif
Fig. 2 (a) Ground state absorption spectra of ASR in dark (AT) and light-adapted (mixture of AT and 13C) conditions are shown by black and orange lines, respectively. The spectrum of the pure 13C isomer (blue dotted line) was derived by taking a linear combination of the dark and light-adapted absorption spectra using a known isomeric ratio. The normalized difference absorption (ΔA) spectrum at a pump–probe delay of 100 ps is shown by the magenta curve. Different characteristic bands are shown: GSB (ground state bleach) and PIA (photo-induced absorption). Green and red (solid in (a) and line in (b)) curves represent the spectrum of the actinic pulse (AP) and degenerate pump/stokes/probe (DFWM) or push/probe (IVS) spectrum, respectively. (b) The pulse sequence used in pump-DFWM (top) and pump-IVS (bottom) experiments.

The energy of the actinic pulse beam was attenuated to 100 nJ and focused on a spot with a diameter of 90 μm. The diameter of each of the DFWM/IVS beams was set to 50 μm and the energy to about 50 nJ. The delay between the pump and push pulse (T) was controlled by a mechanical delay stage. For all measurements, the probe delay (τ) was scanned using a rapid scan approach. A single transient, with a length of 2 ps and time steps of 4 fs, was acquired in about 500 ms.

The optical density (OD) of the ASR sample was around 0.7 at 545 nm (λmax). The sample was circulated through a fused silica flow cell with path length of 0.5 mm by a motor-driven pump to ensure that the illuminated volume of the sample was replenished between two consecutive laser shots. The flow rate was optimized during the measurement in order to avoid the formation of bubbles. The sample was kept in the dark overnight for dark-adapted (DA) measurements. For light adaptation, the sample was illuminated for 1 hour with 500 mW from a LED source (Luxeon LXHL-PL01) centered at 590 nm. During the measurement of light adapted (LA) samples, the transparent sample reservoir was exposed to the same LED source to maintain the LA photostationary state and it was covered with black paper during the measurement of DA. The GS absorption spectrum was measured before and after each set of time-resolved measurements to ensure that the isomeric ratio remained the same and also to check if any degradation of the sample occurred during the span of a measurement (about 20 min).

Signal analysis

The non-oscillatory contributions in each of the DFWM and IVS transients were subtracted by a bi-exponential fitting (Fig. 3(a and b)). In each case, 1 ps of the transients was fitted, leaving out the initial 80 fs to avoid the coherent artefact. A typical algorithm66 was followed to convert the remaining oscillatory signals (Fig. 3(c and d)) from the time domain to the spectral domain (Fig. 3(e and f)). The residuals were multiplied by a Gaussian window and subsequently, zero padding was done before performing a fast Fourier transformation (FFT).
image file: c8cp05469j-f3.tif
Fig. 3 Transients obtained from (a) DFWM and (b) IVS measurements with the non-resonant DFWM/IVS excitation spectrum (i.e. in the absence of the actinic pulse) at a detection wavelength of 590 nm under dark-adapted conditions. The non-oscillatory signal was subtracted in each case by a bi-exponential fitting (red line) before FFT. The corresponding oscillatory signal, obtained after fitting DFWM and IVS transients, are shown in (c) and (d) respectively. Here, a Butterworth filter was applied to the residuals to cut-off the low frequency component below 100 cm−1 in order to suppress noise. The corresponding FFT spectra obtained in DFWM and IVS experiments are shown in (e) and (f) respectively.

The signal to noise ratio in (pump-) DFWM measurements was almost one order of magnitude better than that in (pump-) IVS measurements. For (pump-) DFWM measurements, each transient was averaged 60 times, whereas it was averaged 600 times for (pump-) IVS measurement to achieve a comparable S/N ratio. This difference in the S/N ratio is because the self-heterodyne IVS signal suffers from the fluctuation of intensity of the local oscillator probe, whereas the DFWM signal, being a homodyne technique, is background free.

Fig. 3 also shows the well-known effect of the homodyne detection in (pump)-DFWM on the decay of the oscillatory signal:65 oscillatory contributions in the residuals decay faster in the DFWM signal compared with IVS signal. The intrinsic interference between population grating and vibrational coherence results in a faster decay of the oscillation and consequently makes the FFT spectra broader (compare Fig. 3(e and f)). This also causes small deviations in the spectral peak positions between DFWM and IVS FFT spectra (see e.g.Fig. 4), especially for those vibrational modes (e.g. 1100–1400 cm−1) that are very closely spaced.

image file: c8cp05469j-f4.tif
Fig. 4 The FFT spectra obtained after subtraction of the non-oscillatory components from the transients of the DFWM (a–d) and IVS (e–h) experiment probed at 590 nm (a, b, e and f) and 630 nm (c, d, g and h). The left and right columns represent the FFT spectra of DA and LA ASR, respectively. The spontaneous Raman spectra of DA and LA ASR have been shown on the bottom graphs (i) and (j), respectively. The grey shaded area (750–950 cm−1) represents the characteristic frequency region of the HOOP modes of RPSB.

Experimental results

Non-resonant DFWM/IVS experiments

DFWM and IVS experiments with almost non-resonant DFWM/IVS spectra (Fig. 2(a)) were carried out to capture the pure GS vibrational spectra. The FFT spectra obtained after the subtraction of non-oscillatory contributions from the transients showed the activity majorly in the high-frequency region (>1000 cm−1) for both DFWM (Fig. 4(a–d)) and IVS (Fig. 4(e–h)) experiments. A weak activity in the low-frequency region (<400 cm−1) was also observed, especially at the detection wavelength at 630 nm, where the relative amplitudes of the high frequency (>1000 cm−1) modes were lowered. This variation in the relative intensity of the low and high-frequency modes from the edge (590 nm) to the center (630 nm) detection wavelength of the DFWM/IVS spectrum is in agreement with the natural spectral dependence of CARS-based schemes.68

The comparison between the FFT spectra of DA and LA ASR shows mainly three characteristic features. Firstly, the modes that appeared at 1003 and 1530 cm−1 in the DA DFWM FFT spectrum (Fig. 4(a and c)) shifted to 1007 and 1539 cm−1 (Fig. 4(b and d)), respectively, for LA ASR. Secondly, the DFWM FFT spectrum of DA ASR showed two peaks around 1165 and 1230 cm−1, while for LA ASR they appeared at 1180 and 1300 cm−1. All these changes were also observed in IVS measurements (compare Fig. 4(e, g to f, h)). The modes around 1100–1400 cm−1 in IVS also showed qualitatively similar contrasts between DA and LA ASR as observed in DFWM. Three peaks at 1161, 1215 and 1270 cm−1 for DA ASR (Fig. 4(e and g)) changed to 1175, 1220, 1270 and 1340 cm−1 under LA conditions (Fig. 4(f and h)). Finally, the modes appearing in the region of 750–950 cm−1 (grey shaded area in Fig. 4), in particular, the mode at about 806 (DFWM)/803 (IVS) cm−1, showed relatively stronger amplitude for LA ASR than for DA ASR. This is a very important result that will be discussed later.

The observed GS fingerprint vibrational modes of RPSB have been already previously assigned,7,44,69–74 and we followed the same assignment of the main spectral peaks to specific vibrational motions (Table 1). For example, the mode around 1000–1010 cm−1 is assigned to the CH3 rocking mode and that around 1530–1540 cm−1 is assigned to the C[double bond, length as m-dash]C stretching mode. The multiple peaks in the region from 1100 to 1400 cm−1 are mostly known as the signature of the stretching vibration of different C–C bonds present in RPSB, sometimes coupled with C–C–H in-plane rocking modes. In addition, the low-frequency modes (190, 300 cm−1) are usually assigned to the delocalized torsional modes, whereas the Raman activity in the region 750–950 cm−1 (grey shaded area in Fig. 4) is well known for the HOOP wags (Table 1). In addition, a non-resonant (λexc = 785 nm) spontaneous Raman measurement (Fig. 4(i and j)) was performed to further corroborate the vibrational spectra obtained from the time-resolved experiments (DFWM and IVS). Although the non-resonant Raman results match very well with most of the frequencies and respective shifts observed with DFWM/IVS (Fig. 4(a–h)), there are three major contrasts: (i) different frequencies for modes in the spectral region between 1100 and 1400 cm−1. (ii) The complete absence of low frequency modes (<750 cm−1) in the non-resonant Raman measurements (which were active in the DFWM/IVS measurements). (iii) Different amplitudes of the 803/805 cm−1 mode.

Table 1 Comparison of the central frequencies (in cm−1) of the spectral peaks obtained from non-resonant DFWM, IVS and spontaneous Raman measurements. Assignment of the modes was done based on the literature
Dark adapted (98% AT) Light adapted (64% 13C; 36% AT) Assignment of the modes
DFWM IVS Spon. Raman DFWM IVS Spon. Raman
200, 300, 520 190, 280 210, 350, 520 190, 280 Delocalized torsion71,74
900 803, 890 802, 890 810, 900 803, 890 802, 890 Hydrogen out of plane wag7,44,73
1003 1003 1002 1007 1007 1008 CH3 rocking7,69,70,73
1165, 1230 1161, 1215, 1270 1177, 1196, 1209, 1271 1180, 1300 1185, 1202, 1305, 1338 1175, 1220, 1275, 1340 C–C stretching + C–C–H in plane rocking7,69,70,73
1530 1530 1530 1530 1539 1539 C[double bond, length as m-dash]C stretching7,69,70,73

Separation of isomer-specific contributions

While the spectral features observed for DA and LA ASR (Fig. 4) already mirror the different isomer concentrations, a quantitative extraction of the pure AT and 13C spectra was performed. In our previous report,75 we successfully showed the separation of the GS isomer-specific contributions for two individual high-frequency modes (CH3 rocking and C[double bond, length as m-dash]C stretching) by a single Gaussian fit. A Gaussian model was used instead of a Lorentzian model since the former showed a better fitting quality. Here, we expand and present a more global approach (Fig. 5) to fitting all the modes that appear in the spectral region from 950–1600 cm−1 by a series of Gaussians:
image file: c8cp05469j-t1.tif(1)
where each Gaussian (Gi) represents an individual vibrational mode. The amplitude (ai), central frequency (νi) and width (ωi) of each Gaussian was varied in this multi-Gaussian fit (eqn (1)). Since DA ASR exclusively contains the AT isomer (98%),46,47 the multi-Gaussian fit (Fig. 5(a)) gives the parameters (ai, νi, ωi) for each of the vibrational modes of AT ASR. This information (FFTAT[thin space (1/6-em)]GS) can be used to disentangle the pure spectrum of 13C GS by a constrained fit:
image file: c8cp05469j-t2.tif(2)
Here, the fitted spectra (FFTAT[thin space (1/6-em)]GS) obtained from the previous multi-Gaussian fit was kept constant and the parameters (aj, νj, ωj) of a new set of Gaussians, which represent the modes of the 13C isomer, were varied to fit the LA non-resonant DFWM FFT spectrum (Fig. 5(b)). The factor f in eqn (2) depends on the percentage (36%)46 of the AT isomer present in the LA form. Since the DFWM signal is directly proportional to the square of the sample concentration, f should be equal to the square of the fraction of AT isomers present in LA ASR: (0.36)2 = 0.13. The fitted value obtained for the factor (f = 0.14 ± 0.02) matches this value very well. The pure spectrum of the ground state of the 13C isomer is shown in Fig. 6.

image file: c8cp05469j-f5.tif
Fig. 5 Multi-Gaussian fit of (a) DA and (b) LA non-resonant DFWM FFT spectra probed at 590 nm. The series of Gaussians are represented by green curves in each graph. The black line in (b) represents the spectrum of the AT isomer, which was kept constant during this constrained fit.

image file: c8cp05469j-f6.tif
Fig. 6 The separated pure spectra of AT GS, 13C GS, KAT and K13 are represented by black, red, blue and orange lines, respectively. The black dotted lines represent the central frequencies obtained for AT GS.

The same approach can also be applied to obtain the spectrum of the K-photoproduct of each isomer (K13C and KAT, see Fig. 1). In order to obtain the DFWM signal of these photoproducts, each isomer was directly excited by an actinic pulse spectrally resonant with the respective GS absorption. The DFWM spectrum was still spectrally resonant with the photo-induced absorption (PIA) band (see Fig. 2) but was delayed at T = 100 ps after the AP. This particular delay was chosen as the formation of the vibrationally thermalized K-photoproduct (Fig. 1) and is known to take place within 100 ps.46,47 The pump-DFWM signal in this condition contains several contributions (Fig. 1). The signal of DA ASR after 100 ps, for example, contains contributions of the AT GS and of KAT, which are reasonably easy to separate in pump-DFWM and less in pump-IVS due to noise (not shown). The pump-DFWM signal of the LA ASR, however, is much more challenging: at 100 ps delay, it contained the signal of the GS of both isomers, as well as from both photoproducts. In order to extract the K13C, the spectra of AT, KAT and 13C must be used, which is inherently more susceptible to noise. With this information in mind, the pure spectra of four different stereoisomers (GSAT, GS13C, KAT and K13C) appearing in the photo-cycle are depicted in Fig. 6 (see also Fig. S5, ESI). Several differences in spectral signatures of these four species can be observed. For example, the C[double bond, length as m-dash]C stretching mode of 13C GS (1541 cm−1) gets blue-shifted as compared with that of AT GS (1530 cm−1). Also, isomerization at the C13[double bond, length as m-dash]C14 position caused the C[double bond, length as m-dash]C stretching mode to shift to the higher frequency from AT GS (1530 cm−1) to KAT (1538 cm−1), whereas it shifted to the lower frequency from 13C GS (1541 cm−1) to its corresponding sub-ns photoproduct K13C (1524 cm−1). A similar trend was also observed for the CH3 rocking mode of KAT, while for K13C, it was blue-shifted compared with that for 13C GS. The frequencies and amplitudes of modes in the spectral region between 1100 and 1300 cm−1 are very different for each isomer. Finally, it is important to note that the spectral peaks appearing in the region 1300–1450 cm−1, which have been assigned to the C–C–H in-plane rocking mode (Table 1), did not change from 13C GS (1305, 1427 cm−1) to K13C (1307, 1427 cm−1), whereas a significant change was observed from AT GS to KAT. A single weak peak at 1408 cm−1 appeared for AT GS, whereas at least two strong peaks appeared for KAT (1364, 1446 cm−1) in this region.

Pump-DFWM experiments in the sub-ps timescale

In order to follow the isomerization reaction of both 13C and AT isomers in the excited state, pump-DFWM measurements were performed by varying the actinic pulse delay (T) up to about 1 ps (Fig. 7). The Fourier-transform spectra for DA and LA are displayed in Fig. 7(a and b), respectively. In both situations, two common changes can be observed after the arrival of the actinic pulse (T > 0): (i) amplification of the high frequency modes (>1000 cm−1) and (ii) the appearance of the strong low frequency modes (100–400 cm−1), in particular for LA ASR (Fig. 7(a)), which were very weak in the GS (Fig. 4). Moreover, a closer look at the high frequency (>1000 cm−1) modes (Fig. 7(c) and Fig. S6, ESI) showed a significant spectral shift with the actinic pulse delay. The major spectral shifts of DA and LA ASR happened within 800 and 200 fs, respectively, which matched the respective excited state lifetimes46,47 of the AT (750 fs) and 13C (120 fs) isomers. The frequency evolution for each of the characteristic vibrational modes is described in detail in the following.
image file: c8cp05469j-f7.tif
Fig. 7 The evolution of pump-DFWM FFT spectra probed at 590 nm for (a) DA and (b) LA ASR with the actinic pulse delay. (c) The frequency shift of the C[double bond, length as m-dash]C stretching (left) and CH3 rocking (right) modes for DA ASR with different actinic pulse delays. The black dotted lines represent the central frequencies obtained for AT GS (Fig. 6). The frequency shifts of the corresponding modes in LA ASR are shown in Fig. S6 of ESI.
C[double bond, length as m-dash]C stretching and CH3 rocking mode. Fig. 7(c) depicts the evolution of the C[double bond, length as m-dash]C stretching and CH3 rocking modes for DA ASR. Here, a negative time delay means the actinic pulse comes after DFWM interaction, which should basically give the GS vibrational spectra. Hence, the central frequencies of C[double bond, length as m-dash]C stretching and CH3 rocking modes at T = −100 fs, which appeared at 1530 and 1002 cm−1, matched the GS frequencies displayed above (Fig. 4). At the initial positive T-delays (until 200 fs), each of these modes showed a rapid blue-shift and subsequently, a slower red-shift was observed (see DA in Fig. 7(c) and Fig. S7(a), (c), ESI). Although the FFT spectra at T > 0 contained the major contribution of the transient species being resonantly probed at the corresponding time, a minor contribution of GS coherence, due to the non-resonant Raman interaction, still contaminated the signal. As a consequence, pump-DFWM FFT spectra at any positive actinic pulse delay (T) was not the pure spectra of the transient species formed within the corresponding time (T) after the initiation of the reaction by actinic pulse. Since the pure GS spectra of the AT isomer was already known (Fig. 5 and 6), it was possible to extract the pure spectra of the transient species present at different actinic pulse delays by the aforementioned constrained multi-Gaussian fitting (Fig. 5 and Fig. S5, ESI).

By applying this method, the frequencies of the C[double bond, length as m-dash]C stretching and CH3 rocking modes for AT were obtained and are shown in Fig. 8(a and c), respectively. AT shows a clear strong blue-shift of 21 and 18 cm−1 for the C[double bond, length as m-dash]C stretching and CH3 rocking modes, respectively, at T = 200 fs as compared with that of GS species (T < 0). At longer delays (T > 200 fs), a red-shift to 1538 and 1005 cm−1 occurred. These two latter frequencies are the central frequencies of the C[double bond, length as m-dash]C stretching and CH3 rocking modes, respectively, for previously extracted pure KAT (Fig. 6). The separation of all contributions to the transient LA ASR signal is more challenging than for DA ASR. Here, two ground state species were excited (AT and 13C), followed by two excited states and two photoproducts, which led to much larger frequency uncertainties and ambiguous results. Therefore, the frequency shifts in the C[double bond, length as m-dash]C stretching and CH3 rocking modes for the LA ASR are shown without extraction (Fig. 8(b and d), respectively). A similar trend was observed as for AT, i.e., an initial blue-shift took place followed by a red-shift. However, the blue-shift was much smaller and the entire dynamics was much faster in the LA than for AT (compare e.g.Fig. 8(a and b)). The maximum of the blue-shift appeared at a much earlier actinic pulse delay (about T = 40 fs) compared with the AT form (about T = 200 fs).

image file: c8cp05469j-f8.tif
Fig. 8 The excited state evolution of C[double bond, length as m-dash]C stretching (a and b) and CH3 rocking modes (c and d) obtained for the pure AT isomer (a and c) and LA (b and d) from ASR. The central frequencies of pure GS (AT and 13C) and K-photo intermediates (KAT and K13C) are shown by black, red, blue and orange dotted lines, respectively, in each graph. Probe detection wavelength was 590 nm.
Fingerprint (1100–1400 cm−1) region. As observed for the GS spectra (Fig. 4 and 6), the fingerprint region around 1100–1400 cm−1 (Fig. 9) showed a congested spectrum after the actinic excitation, particularly for AT ASR. The double peak feature (1165 and 1230 cm−1) before time zero merged into a single major peak (1193 cm−1) along with a minor peak (1330 cm−1) at T = 100 fs for AT ASR. Afterwards, the major peak showed a red-shift till 200 fs, which was followed by a slow monotonous blue shift until the formation (200 to 800 fs) of the hot-intermediate, commonly known as the J-species in literature. Subsequently, it underwent a small red-shift during vibrational relaxation on a much longer timescale (1 to 100 ps) to form the thermally relaxed photoproduct (KAT). This is in contrast to the signal of LA ASR, where the major peak at ∼1181 cm−1 underwent a blue-shift initially (T = 40 fs) to 1187 cm−1 and did not shift afterwards within the excited state lifetime (<120 fs) of the 13C isomer. It only showed a small (1187 to 1183 cm−1) red-shift during the photoproduct vibrational relaxation (∼1 ps to 100 ps), similar to the AT isomer.
image file: c8cp05469j-f9.tif
Fig. 9 The evolution of fingerprint (1100–1400 cm−1) modes in the pump-DFWM FFT spectra for (a) AT and (b) LA ASR at different actinic pulse delays (T) probed at 590 nm. The vertical lines in (a) and (b) represent the corresponding central frequencies of AT and 13C GS, respectively.
Low frequency (<400 cm−1) torsion and HOOP modes. As shown in Fig. 4, the low-frequency modes below 400 cm−1 were absent or very weak in GS. However, a delayed activation of these low-frequency modes (Fig. 7(a and b)) was observed after the excitation with the AP. This is in contrast to the activity of the high-frequency modes, whose amplitude increased as soon as the AP arrived (Fig. S8, ESI). This contrasting activity of low frequency (<400 cm−1) modes was further evidenced in Fig. 10 at several actinic pulse T-delays. A significant relative amplification of the low-frequency modes (190, 300 cm−1), compared with the high frequency (>1000 cm−1) modes, was observed at T > 0 for DA and LA ASR. It is important to note that LA ASR showed a much stronger low-frequency activity than DA ASR.
image file: c8cp05469j-f10.tif
Fig. 10 The relative amplification of low frequency modes probed at 590 nm before (black, T < 0) and after (red, T > 0) the arrival of the actinic pulse for DA (a and c) and LA (b and d) ASR.

The evolution of the amplitude of the low-frequency modes was different from other modes. While the high-frequency modes (e.g. C[double bond, length as m-dash]C stretch) showed an instantaneous increase in the activity at very early actinic pulse T-delays, low-frequency (e.g. 190 cm−1) activity showed a delayed rise for LA ASR (Fig. 7). The FFT amplitude of the low-frequency (190 cm−1) modes took about 100 fs to reach the maxima, whereas that of the C[double bond, length as m-dash]C stretching mode reached the maxima within the experimental time resolution after the actinic excitation (T = 20 fs). In addition, the exponential rise time (70 ± 20 fs) of low-frequency (190 cm−1) modes matched the exponential decay time (60 ± 40 fs) of the C[double bond, length as m-dash]C stretching mode (Fig. S8, ESI). Moreover, no significant evolution of HOOP modes (800–1000 cm−1) was observed in our measurements. This will be further discussed and investigated later.


Ground state Raman activity

Strong Raman activity of the vibrational modes mainly in the high frequency (>1000 cm−1) region of the spectra has been observed in non-resonant experiments (Fig. 4). The convolution of the GS absorption spectrum with the non-resonant DFWM/IVS excitation spectrum (Fig. 2) gives a FWHM of about 500 cm−1 and, therefore, there is a negligible induction probability of a vibrational coherence above 500 cm−1 in the ES potential surface. Hence, all modes with frequencies well over 500 cm−1 in these non-resonant measurements, including HOOP, CH3-rocking, C–C and C[double bond, length as m-dash]C modes, are assigned to the GS manifold. This is further supported by the spontaneous Raman spectra (Fig. 4(i and j)), which show a good correlation with the FFT spectra obtained from time-resolved experiments for frequencies well over 500 cm−1. However, the weak but detectable Raman activity observed in the low-frequency region (200 and 300 cm−1) in both DFWM and IVS signal, does not appear in any spontaneous Raman spectra. If this low-frequency activity originates from GS, it should definitely be visible in the spontaneous Raman measurements. Therefore, we assigned all low frequency (<400 cm−1) activity detected with DFWM/IVS to the excited state manifold. A detailed discussion about the origin and activation mechanism of these modes will be presented in next section.

The extraction of the pure GS spectra of AT and 13C isomers (Fig. 6) enabled us to make a quantitative comparison. The central frequencies of the C[double bond, length as m-dash]C stretching and CH3 rocking modes were shifted by about 11 cm−1 (1530 vs. 1541 cm−1) and 4 cm−1 (1002 vs. 1006 cm−1), respectively, between AT to 13C ASR. This is in good agreement with the reported values for ASR.46 A similar increase in high frequency modes from the all-trans to cis isomer has also been reported recently for chanello-rhodopsin58 and visual rhodopsin,49 which further corroborates the separation of the pure spectra of ASR. Furthermore, the C–C stretching region shows multiple distinguishable peaks between AT and 13C ASR (Table 1). Two peaks at around 1165 and 1230 cm−1, observed for AT ASR (Fig. 6), are quite common for the retinal chromophores in the AT conformation in other proteins (Table 2). On the other hand, the appearance of a distinguishable mode above 1300 cm−1 for 13C ASR is a general marker of the cis form (Table 2).49,58 In addition, the C–C stretching mode, which appears at 1180 cm−1 in the case of 13C ASR, has been attributed as an indicator for the formation of the 13C-isomer of BR.76

Table 2 Comparison of the selected fingerprint vibrational modes for all-trans and cis isomers found for ASR in this report and those reported for retinal protonated Schiff base (RPSB) in solution and different proteins: bacteriorhodopsin, visual rhodopsin and chanello-rhodopsin
Sample CH3 rocking [cm−1] C–C stretch + C–C–H in plane [cm−1] C[double bond, length as m-dash]C stretch [cm−1]
AT-RPSB in methanol52 1010 1160, 1205 1565
Bacteriorhodopsin (AT)50 1008 1165, 1210 1530
Visual rhodopsin (AT)49 1167, 1322 1541
Visual rhodopsin (11-cis)49 1173, 1275, 1313, 1363 1550
Chanello-rhodopsin (AT)58 1011 1161, 1208, 1281 1531
Chanello-rhodopsin (13-cis)58 1017 1157, 1196, 1301, 1369 1545
ASR (AT)This[thin space (1/6-em)]report 1002 1164, 1229 1530
ASR (13-cis)This[thin space (1/6-em)]report 1006 1094, 1180, 1305 1541

The activity of HOOP modes (800–1000 cm−1) in the GS (grey shaded region in Fig. 4) has been interpreted as an indirect indicator of a distorted non-planar structure of RPSB.71,77 The relative amplitude of the HOOP modes, particularly at about 805 cm−1, is higher in the LA (major component is 13C) ASR compared with DA (98% AT) ASR. This observation is consistent in each (DA vs. LA) of the spectra measured by three different spectroscopic techniques (DFWM, IVS and spontaneous Raman). All these results together hint at the presence of a non-planar structure for the 13C isomer, although it is much less distorted than the 11-cis isomer in visual rhodopsin.44,69,71,77 A detailed assignment of H-wag modes is, however, necessary in order to specify the region of distortion in the long retinal chain. The complete assignment of H-wag modes for ASR has not yet been reported but it is well known for BR from the resonant Raman study by Smith et al.70 In that report, a mode at 800 cm−1 was assigned to the C14–H out-of-plane wag and showed strong amplitude for the 13C isomer, whereas it almost disappeared for the AT isomer, similar to ASR isomers in our study. Later, solid-state NMR data78,79 indicated a difference in the structural rigidity along the C13[double bond, length as m-dash]C14–C15 moiety between AT and 13C isomers inside the retinal pocket of BR. Following the same line, our observation can be interpreted as both isomers being pre-twisted around or close to the C13[double bond, length as m-dash]C14 bond of the retinal chromophore but the 13C isomer is significantly more distorted than AT.

The exact position of this distortion and the differences between the two isomers was further numerically investigated by applying a classical (harmonic) atomistic force field and performing molecular dynamics simulations of ASR embedded in a membrane model (see ESI). The geometry of AT and 13C GS inside the retinal pocket shows that both isomers are equally distorted around the C13[double bond, length as m-dash]C14 bond, i.e. the dihedral angle of C12–C13–C14–C15 is found to be +191.6° and 11.7° (Table S1, ESI) for the AT and 13C isomers, respectively. This is in good agreement with the value (∼13°) previously reported by QM/MM geometry optimization calculations.31 These new calculations, however, show that the isomers significantly differ around the neighboring C14–C15 bond (dihedral C13–C14–C15–N). While AT is nearly planar (177.1°) around the C14–C15 bond, 13C is twisted almost 10° (189.5°). This relatively larger pre-twist around the C14–C15 bond for the 13C isomer corroborates very well the experimental observation of higher amplitude of C14–H wag as discussed above in the previous paragraph.

Finally, a brief comparison of the vibrational signatures of the GS isomers with its corresponding K-photoproducts (Fig. 6) helped to further rationalize the frequency shifts in terms of conformational differences. The conformational changes altered the delocalization of δ and π-electron density for the macro-molecules like RPSB, which explains the frequency shift of the CH3 rocking, C–C and C[double bond, length as m-dash]C stretching and C–C–H in-plane rocking modes. The changes in the frequency and amplitude of the C–C–H in-plane rock (1300–1400 cm−1) modes (Fig. 6) are important as these are reported to be affected by the torsion around the active C[double bond, length as m-dash]C bond.62 Although the frequency of this mode changes from AT GS (1408 cm−1) to KAT (1364, 1446 cm−1), it remains almost the same for 13C GS (1305, 1427 cm−1) and for K13C (1307, 1427 cm−1), indicating a very similar strain around the C13[double bond, length as m-dash]C14 bond for the latter two isomers. These results also corroborated the FTIR results which illustrated that isomerization caused a larger rotation around the active C13[double bond, length as m-dash]C14 bond in the AT as compared with that in the 13C isomer.73

Excited state evolution of high-frequency modes (≥1000 cm−1)

The pump-DFWM measurements captured the sub-ps frequency shifts of high-frequency modes. In general, the C[double bond, length as m-dash]C stretching and CH3 rocking modes showed a blue-shift and subsequent red-shift for both AT and LA (major 13C) ASR. Since the DFWM spectra used for this measurement cover the region (580–720 nm) where both excited state absorption and stimulated emission of ASR overlaps,46,47 there are two possible origins of the observed frequency shifts: (i) the excited state evolution of the nuclear wave packet; (ii) the vibrational relaxation of the GS wave packet, generated by a stimulated emission pumping (SEP) process.24,38,66 For the latter, we would expect a red-shift just after the excitation for an anharmonic ground state potential.24 This is because the GS wave packet, generated via SEP, in the higher lying vibrational states is lower in frequency than a relaxed GS wave packet, generated via non-resonant ISRS in the lower lying vibrational states. On the contrary, we observed a blue-shift as compared to the GS, excluding that possibility and hence, the frequency shift was interpreted as mostly due to the excited state wave packet motion. In addition, the timescale of the frequency shifts is significantly different for AT and 13C isomers. It matches very well the excited state lifetime of each isomer, which further indicates that the observed frequency shifts originate due to the transient evolution of the ES species.

Sub-ps transient frequency shifts in the fingerprint region have been frequently interpreted as the conformational changes associated with the isomerization process.62,63 Here, the blue shift of the C[double bond, length as m-dash]C stretch (Fig. 8) is interpreted as the reduction in conjugation length due to the rotation around the C13[double bond, length as m-dash]C14 bond during the isomerization. After the excitation by the actinic pulse, the RPSB is promoted to the Franck–Condon point of the excited state where it is still in the same geometry as it is in the ground state (Fig. 11(a)). In this geometry, an extended π-conjugation is present along the retinal chain by the pz-orbitals from C5 up to the protonated Schiff base N atom. As the isomerization reaction progresses, the RPSB starts to adopt a more non-planar twisted structure (Fig. 11(b)) towards the photoproduct geometry. During this evolution, the C14–C15–N π-conjugated moiety becomes almost perpendicular to the remaining π-conjugated system (C5 to C13) at the 90° twisted form (AT*) and the conjugation between two moieties, mentioned before, gets completely disrupted. Hence, the effective conjugation length gets reduced during evolution from the FC towards the 90° twisted form. It is well known for linear conjugated polyenes that the reduction of effective conjugation length causes a frequency blue-shift of the C[double bond, length as m-dash]C stretching mode.80,81 Thus, a decrease in the effective conjugation length explains the blue-shift of C[double bond, length as m-dash]C stretching mode observed here for retinal in ASR. The blue-shift of the CH3 rocking mode can also be interpreted in a similar way. The disruption of the C13[double bond, length as m-dash]C14 bond during isomerization reduces the partial positive charge over the C13 atom induced by the positively charged Schiff base N-atom. The partial positive charge over C13 tends to pull the electron density from the CH3 moiety (bonded to C13) towards it but the lack of inductive effect in the twisted state pushes the electron density towards the CH3 moiety. This possibly causes the blue-shift of the rocking mode of CH3. This also corroborates the picture of change in electron density during the isomerization of BR, recently captured by the transient femtosecond X-ray spectroscopy.29 Thus, following our argument, the slower frequency shift observed for the AT isomer (Fig. 8(a)) as compared to the 13C isomer (∼LA, Fig. 8(b)) can be interpreted as it takes longer for the AT isomer to rotate around the C13[double bond, length as m-dash]C14 bond to form the 90° twisted state compared to 13C. This also corroborates the QM/MM calculations,31 which predict a barrier in the excited trajectory of AT and, therefore, a slower formation of the respective twisted transition state. Furthermore, the subsequent slow red-shift for both isomers is a signature of an increase in π-conjugation as the RPSB structurally changes from the 90° twisted species to the relatively more planar initial GS or photoproduct.

image file: c8cp05469j-f11.tif
Fig. 11 The reduction of the effective π-conjugation length from (a) the Franck–Condon excited state to (b) the twisted transition state (far Franck–Condon region). In (a), there is an extended π-conjugation from the C5-atom to the N-atom of the Schiff base, which is reduced in (b).

The frequency shift of the fingerprint modes in the region from 1100–1400 cm−1 are relatively more complicated than that observed for the C[double bond, length as m-dash]C stretching mode. This is because different C–C stretching modes (C8–C9, C10–C11, C12–C13, etc.) of retinal exhibit distinct closely spaced peaks that overlap in that region. In addition, C–C–H in-plane rocking modes are also sometimes coupled with the C–C stretching, which makes the spectral evolution of this fingerprint mode even more challenging to rationalize. However, the most intense peak below 1200 cm−1, which is known to be mostly uncoupled from C–C–H rocking modes, follows a trend that can also account for the change in the effective π-conjugation length. It has also been observed for linear conjugated polyenes80,81 that the small reduction in the effective conjugation length causes a blue-shift of the mode below 1200 cm−1 but further reduction of the effective conjugation length eventually results in a red-shift. This matches very well with the frequency shift of AT ASR (Fig. 9), where it undergoes an initial blue-shift until 100 fs and then red-shifts until 200 fs. This is the delay that has been interpreted above as the time to reach the 90° twisted state for AT ASR. Afterwards, the effective conjugation length was expected to increase again due to the slow formation of relatively planar KAT, thus causing a blue-shift of the finger print mode. This effect is much weaker for the 13C (∼LA) isomer, which results in no observable shift. This once again hints at a smaller rotation around C13[double bond, length as m-dash]C14 in the 13C isomer as compared with that in the AT isomer in ASR.

In the context of the frequency-shift for the excited state modes, it is also important to note that the experimental time resolution cannot resolve the very fast red-shift taking place as a result of the BLA. This red-shift must take place according to numerical calculations in timescales around 20 fs.32 The fact that an initial blue-shift is observed in all measurements at delays where BLA is temporally overlapping possibly means that the pure blue-shift is potentially much bigger than the observed one.

Excited state evolution of low frequency (<400 cm−1) and HOOP modes

A central result of pump-DFWM measurements was the observation regarding the strong activity of the low-frequency modes around 190 and 300 cm−1 at positive actinic pulse delay, which are absent in the GS (T < 0). Very similar low-frequency peaks (160, 210 and 300 cm−1) have also been observed previously for ASR in the transient absorption experiments57 by probing in the near infrared region (950–1450 nm), which is far away from the GS absorption and only covered by the stimulated emission. Furthermore, few strong Raman modes below 400 cm−1 have been observed in the excited state of RPSB in solution52 by Kraack et al. and also observed more recently for rhodopsin28 and chanello-rhodopsin58 by Schnedermann et al. Moreover, most importantly, the low-frequency modes (e.g. ∼190 cm−1) in the pump-DFWM transients showed faster dephasing than the high-frequency modes (e.g. C[double bond, length as m-dash]C stretching) (Fig. S10, ESI). All these observations together clearly suggest that the low-frequency Raman activities originated from the ES manifold.

Another central result is the lack of HOOP activity in the excited state in our pump-DFWM (and pump-IVS) measurements, which is surprising. There are two plausible explanations for this observation. The surrounding protein environment in ASR, which is different from rhodopsin and BR, where the HOOP activity is stronger,23,28 can significantly reduce the Raman transition probability of a specific mode by modifying the retinal pocket. Secondly, the activity of the HOOP mode of BR has been observed to be strongly dependent on the center wavelength of DFWM spectra, used to create the vibrational coherence.50 The Raman activity of this mode was found to be mainly enhanced for blue detuned excitation. In contrast, the DFWM excitation spectra in our study were red detuned compared with the ground state absorption (Fig. 2).

To clarify the lack of HOOP activity as well as to find out the activation mechanism of the low-frequency (<400 cm−1) ES modes, we carried out two additional sets of DFWM measurements with two different excitation spectra (Fig. 12). Spectrum 1 is the same as used before (Fig. 2), which is almost non-resonant to GS absorption and, hence, only capable of exciting GS modes and also possibly the ES modes below 500 cm−1 due to a small overlap with GS. Spectrum 2, however, being completely resonant to the GS absorption, can directly excite all GS as well as ES vibrational modes. Two major changes (Fig. 12) were observed for both DA and LA ASR by tuning the DFWM spectrum, namely (i) amplification of low-frequency and HOOP modes and the (ii) frequency-shift of high-frequency modes.

image file: c8cp05469j-f12.tif
Fig. 12 DFWM FFT spectra obtained by using near-resonant (1) and resonant (2) excitation spectra for DA (black) and LA (grey) ASR detected at different probing wavelengths, shown by vertical grey lines in (a) and (b).
(i) Relative amplification of low frequency (<400 cm−1) and HOOP modes. The FFT spectra for the non-resonant spectrum (Fig. 12(a)), detected at 590 and 630 nm, show the activity mostly above 1000 cm−1 along with a few weak low-frequency modes (∼190, 300 and 510 cm−1), as observed before (Fig. 4). In contrast, a strong relative amplification of the FFT amplitude of low-frequency modes (<400 cm−1) with respect to high-frequency modes (>1000 cm−1), was observed for the FFT spectra in the resonant DFWM experiment (Fig. 12(b)) as compared to the non-resonant DFWM measurement (Fig. 12(a)). On the one hand, these low-frequency (<400 cm−1) modes have been observed to be completely absent in the non-resonant steady-state Raman spectrum (Fig. 4(i and j)) and therefore, these modes have been attributed to the ES manifold. This assignment nicely matches the weak Raman activity of these low-frequency modes in the non-resonant DFWM/IVS FFT spectrum (Fig. 12(a)). On the other hand, the same set of low frequency (<400 cm−1) modes is relatively enhanced in resonant DFWM measurements, in comparison to the high frequency (>1000 cm−1) vibrational modes. One obvious explanation for this observation would be that the low-frequency modes are much more Franck–Condon active than the high-frequency modes. However, an alternative interpretation can also be drawn if one considers different activation mechanisms for the high and low-frequency modes as was reported in theoretical studies on the retinal chromophore model30 as well as experimental studies for BR50 and RPSB in solution.52,53 In these studies, the high-frequency (>1000 cm−1) modes were reported to be Frank–Condon active, whereas the low-frequency (<400 cm−1) torsional modes were considered to be impulsively excited by the high-frequency modes via internal vibrational energy redistribution (IVR). This is further supported by our observation of the delayed rise in the low-frequency amplitude as compared to the high-frequency modes in transient pump-DFWM spectra (Fig. S8, ESI).

A second question remained regarding the reason behind the passiveness of the HOOP mode Raman activity in our pump-DFWM experiments. DFWM experiments with two different spectra (Fig. 12) showed that the mode at 805 cm−1, previously assigned as a HOOP mode, showed a significantly strong amplitude in the resonant DFWM experiment (Fig. 12(a)), which was very weak in the GS (Fig. 12(b)). This shows that the HOOP modes are induced only when probed with blue-detuned DFWM spectra, not with red-detuned DFWM spectra, as used in our pump-DFWM experiment. This observation is very similar to that with BR50 and thus seems to be an intrinsic molecular property of retinal.

(ii) Frequency shift of the high frequency (>1000 cm−1) modes. The high frequency mode, e.g. the C[double bond, length as m-dash]C stretch (1530 cm−1 for DA and 1540 cm−1 for LA) in the non-resonant DFWM FFT spectrum (Fig. 12(a)) gets blue-shifted in the resonant DFWM FFT spectrum (Fig. 12(b)) (1538 cm−1 for DA and 1543 cm−1 for LA). A similar trend was also observed for CH3 rocking modes: 1002 cm−1 shifted to 1008 cm−1 for DA while 1007 shifted to 1010 cm−1 for LA. This is expected since non-resonant DFWM FFT spectra contain only GS modes, whereas resonant DFWM FFT spectra contain the mixture of GS and ES modes in the high-frequency region. As we have shown above, both C[double bond, length as m-dash]C stretching and CH3 rocking modes show a frequency blue-shift in the ES as compared to the GS (Fig. 8).

Pre-twisting and isomerization dynamics

In the context of retinal photochemistry, pre-twisting has been claimed to be one of the potential elements that can accelerate the isomerization process. In particular, the pre-straining inside the protein pocket has been repeatedly shown for visual rhodopsin by X-ray crystallography,9 NMR,40 resonant Raman studies44,69,71,77 and theoretical simulations,42,82 which account for its fast ballistic IC dynamics. Although non-resonant Raman spectra (Fig. 4), as reported in this study, address a relatively more pre-twisted geometry of the 13C isomer in ASR, no significant difference in distortion around the C13[double bond, length as m-dash]C14 between the AT and 13C isomers has been pointed out by the X-ray crystallographic study of ASR with 2.0 Å resolution.83 Very similar observations were made for BR where the 13C isomer demonstrated kinetics that was 3 times faster84 as compared to AT but the structural data indicated that both isomers were nearly planar without any indication of protein-catalyzed strain as in visual rhodopsin. However, indirect evidence of weak pre-straining for the 13C isomer was found in the NMR78 and resonance Raman70 studies of BR, although it seemed to be negligible as compared to the 11-cis isomer of visual rhodopsin. More recently, a closer inspection of the X-ray data depicted a twist around the C14–C15 bond in the 13C isomer but not in the AT isomer of BR.10 Similarly, our classical GS dynamics simulation (Table S1, ESI) showed no difference in the distortion around the C13[double bond, length as m-dash]C14 bond for the AT and 13C isomers; rather it indicated the presence of a twist (∼10°) around the C14–C15 bond for the 13C isomer, whereas the AT isomer appeared to be almost planar. Thus, this twist around the C14–C15 bond, which is adjacent to the active C13[double bond, length as m-dash]C14, may play a central role in the faster dynamics of 13C in ASR as claimed for 11-cis isomer in visual rhodopsin.85

This difference in distortion must originate from differences in the packing of the RPSB inside the retinal pocket. It fits the observation of faster dephasing (Fig. S9, ESI) of GS coherence of AT as compared to the 13C isomer, which can be interpreted as the stronger coupling of the AT isomer to the surrounding than 13C. A recent femtosecond X-ray study29 of BR has depicted the importance of the specific electrostatic interactions between protein and RPSB to guide the isomerization in a certain direction. Hence, this difference in electrostatic interaction between the AT and 13C isomers with the protein surrounding may lead to different trajectories during the reaction. It corroborates the FTIR study,72 which showed that the sub-ps isomerization caused a stronger disruption of the H-bond between the surrounding water molecule and the protonated N-atom of the Schiff base in the case of the AT isomer as compared to 13C. This H-bond has been previously46 suspected to be responsible for hindering the rotation of the protonated Schiff base and thus slowing down the isomerization around the C13[double bond, length as m-dash]C14 bond. This also matches our observations for AT ASR of a slower frequency shift in the fingerprint modes (C[double bond, length as m-dash]C and C–C stretching, CH3 rocking), which are sensitive to localized structural changes and have been interpreted as the delayed formation of the twisted transition state for AT as compared to the 13C isomer.


This work has investigated the mechanistic origin of the huge dynamical differences observed in the isomerization of AT and 13C isomers in Anabaena Sensory Rhodopsin. In this regard, the evolution of the GS as well as the ES structural changes of each isomer has been followed by applying DFWM, IVS, pump-DFWM and pump-IVS spectroscopy techniques. The experiments were able to unveil three major structural and dynamical differences in the isomerization of each isomer: (i) HOOP activity in the GS is stronger for the 13C ASR as compared to AT ASR. (ii) Large (up to 20 cm−1) and delayed transient frequency blue-shifts were observed for the C[double bond, length as m-dash]C stretching and CH3 rocking modes in the ES of AT ASR. (iii) There was a delayed Raman activity increase in low-frequency modes (<400 cm−1).

These experimental findings depict very different isomerization scenarios for each isomer. The stronger HOOP activity at about 805 cm−1 in the GS spectra of LA ASR indicates that the 13C isomer is already more pre-twisted in the GS than the AT ASR isomer inside the retinal pocket. Analogous to BR and supported by theoretical calculations, we have assigned this mode to the C14–H wag. Our results point to a distortion located at around the C14–C15 bond of the 13C isomer, which is neighboring to the isomerizing C13[double bond, length as m-dash]C14 bond. The evolution of frequency shifts of high-frequency modes and of the C[double bond, length as m-dash]C stretching and CH3 rocking modes, in particular, depicts a much slower formation of the twisted configuration for AT ASR as compared to 13C ASR. Finally, the delayed increase in amplitude in the transient nonlinear Raman spectra as well as the stronger relative amplification of the low-frequency modes from non-resonant to resonant DFWM experiments indicate that these modes are potentially activated by the C[double bond, length as m-dash]C bond via the IVR mechanism, similar to previous observations in BR.50

These discoveries have profound implications in understanding the mechanism of the primary events in retinal proteins. AT and 13C isomers of RPSB in ASR show several dynamic and spectral features that are known for other respective isomers of the RPSB in other retinal proteins, in particular, BR. Perhaps a major point is the very distinct evolution of high-frequency modes in the excited states of 13C and AT ASR. While C[double bond, length as m-dash]C stretching and CH3 modes, for example, are not reactive coordinates per se, they certainly reflect the structural changes taking place at localized positions along the retinal during the isomerization. The delayed decrease in the conjugated double bond length observed for AT ASR in the excited state as compared to 13C ASR surely follows the previous proposal about the presence of a barrier in the excited state manifold of AT ASR. Nevertheless, our findings suggest that the longer dynamics observed for AT ASR may result from more than one factor, namely a barrier in the excited state and the lack of a pre-distortion (compared to 13C) in the ground state. Finally, although the HOOP Raman activity in 13C ASR is much lower than for 11-cis in visual rhodopsin, the excited state lifetimes are not very different. This further reinforces that it is not a single effect that is playing a role in determining the excited state lifetime, as has been advocated in the past. We expect that a time-resolved vibrational spectroscopy study of specific point mutations of ASR will be able to pinpoint the roles of barriers and pre-distortions and unveil the extent of this effect.

Conflicts of interest

There are no conflicts to declare.


This work was supported by DFG grant BU 2696/2-1 and ANR-DFG “071 – FEMTO-ASR: Anabaena Sensory Rhodopsin: Ein biologisches Modellsystem, zur Untersuchung der Mechanismen von photochemischen Reaktionen durch konische Durschneidungen”.


  1. J. K. Lanyi, J. Struct. Biol., 1998, 124, 164–178 CrossRef CAS PubMed.
  2. J. Sasaki and J. L. Spudlich, Biochim. Biophys. Acta, 2000, 1460, 230–239 CrossRef CAS.
  3. N. Kamo, K. Shimono, M. Iwamoto and Y. Sudo, Biochemistry, 2001, 66, 1277–1282 CAS.
  4. L. O. Essen, Curr. Opin. Struct. Biol., 2002, 12, 516–522 CrossRef CAS PubMed.
  5. G. Negel, D. Ollig, M. Fuhrmann, S. Kateriya, A. M. Musti, E. Bamberg and P. Hegemann, Science, 2002, 296, 2395–2398 CrossRef PubMed.
  6. R. Berera, R. v. Grondelle and J. T. M. Kennis, Photosynth. Res., 2009, 101, 105–118 CrossRef CAS PubMed.
  7. A. Kawanabe and H. Kandori, Sensors, 2009, 9, 9741–9804 CrossRef CAS PubMed.
  8. O. P. Ernst, D. T. Lodowski, M. Elstner, P. Hegemann, L. S. Brown and H. Kandori, Chem. Rev., 2014, 114, 126–163 CrossRef CAS PubMed.
  9. T. Okada, M. Sugihara, A.-N. Bondar, M. Elstner, P. Entel and V. Buss, J. Mol. Biol., 2004, 342, 571–583 CrossRef CAS PubMed.
  10. T. Nishikawa, M. Murakami and T. Kouyama, J. Mol. Biol., 2005, 352, 319–328 CrossRef CAS PubMed.
  11. R. W. Schoenlein, L. A. Peteanu, R. A. Mathies and C. V. Shank, Science, 1991, 254, 412–415 CrossRef CAS PubMed.
  12. Q. Wang, R. W. Schoenlein, L. A. Peteanu, R. A. Mathies and C. V. Shank, Science, 1994, 266, 422–424 CrossRef CAS PubMed.
  13. O. A. Smitienko, M. N. Mozgovaya, I. V. Shelaev, F. E. Gostev, T. B. Feldman, V. A. Nadtochenko, O. M. Sarkisov and M. A. Ostrovsky, Biochemistry, 2009, 75, 34–45 Search PubMed.
  14. D. Polli, P. Altoe, O. Weingart, K. M. Spillane, C. Manzoni, D. Brida, G. Tomasello, G. Orlandi, P. Kukura, R. A. Mathies, M. Garavelli and G. Cerullo, Nature, 2010, 467, 440–443 CrossRef CAS PubMed.
  15. J. E. Kim, M. J. Tauber and R. A. Mathies, Biochemistry, 2001, 40, 13774–13778 CrossRef CAS PubMed.
  16. R. A. Mathies, C. H. B. Cruz, W. T. Polard and C. V. Shank, Science, 1988, 240, 777–779 CrossRef CAS PubMed.
  17. H. Kandori, K. Yoshihara, H. Tomioka and H. Sasabe, J. Phys. Chem., 1992, 96, 6066–6072 CrossRef CAS.
  18. H. Kandori, H. Tomioka and H. Sasabe, J. Phys. Chem. A, 2002, 106, 2091–2095 CrossRef CAS.
  19. H. Kandori, Y. Katsuta, M. Ito and H. Sasabe, J. Am. Chem. Soc., 1995, 116, 2669–2670 CrossRef.
  20. P. Hamm, M. Zurek, T. R. Sschinger, H. Patzelt, D. Oesterhelt and W. Zinth, Chem. Phys. Lett., 1996, 263, 613–621 CrossRef CAS.
  21. S. L. Logunov, L. Song and M. A. El-Sayed, J. Phys. Chem., 1996, 100, 18586–18591 CrossRef CAS.
  22. K. A. Freedman and R. S. Becker, J. Am. Chem. Soc., 1986, 108, 1245–1251 CrossRef CAS.
  23. P. Kukura, D. W. McCamant, S. Yoon, D. B. Wandschneider and R. A. Mathies, Science, 2005, 310, 1006–1009 CrossRef CAS PubMed.
  24. D. W. McCamant, P. Kukura and R. A. Mathies, J. Phys. Chem. B, 2005, 109, 10449–10457 CrossRef CAS PubMed.
  25. A. Kahan, O. Nahmias, N. Friedman, M. Sheves and S. Ruhman, J. Am. Chem. Soc., 2007, 129, 537–546 CrossRef CAS PubMed.
  26. A. Wand, I. Gdor, J. Zhu, M. Sheves and S. Ruhman, Annu. Rev. Phys. Chem., 2013, 64, 437–458 CrossRef CAS PubMed.
  27. D. Polli, I. Rivalta, A. Nenov, O. Weingart, M. Garavelli and G. Cerullo, Photochem. Photobiol. Sci., 2015, 14, 213–228 RSC.
  28. C. Schnedermann, M. Liebel and P. Kukura, J. Am. Chem. Soc., 2015, 137, 2886–2891 CrossRef CAS PubMed.
  29. P. Nogly, T. Weinert, D. James, S. Carbajo, D. Ozerov, A. Furrer, D. Gashi, V. Borin, P. Skopintsev, K. Jaeger, K. Nass, P. Bath, R. Bossman, J. Koglin, M. Seaberg, T. Lane, D. Kekilli, S. Brünle, T. Tanaka, W. Wu, C. Milne, T. White, A. Barty, U. Weirstall, V. Panneels, E. Nango, S. Iwata, M. Hunter, I. Schapiro, G. Schertler, R. Neutze and J. Sandfuss, Science, 2018, 361, 1–7 Search PubMed.
  30. A. Cembran, F. Bernardi, M. Olivucci and M. Garavelli, J. Am. Chem. Soc., 2003, 125, 12509–12519 CrossRef PubMed.
  31. A. Strambi, B. Durbeej, N. Ferre and M. Olivucci, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 21322–21326 CrossRef CAS PubMed.
  32. I. Schapiro, M. N. Ryazantsev, L. M. Frutos, N. Ferre and M. Olivucci, J. Am. Chem. Soc., 2011, 133, 3354–3364 CrossRef CAS PubMed.
  33. I. Schapiro and S. Ruhman, Biochim. Biophys. Acta, Bioenerg., 2014, 1837, 589–597 CrossRef CAS PubMed.
  34. M. Manathunga, X. Yang, Y. Orozco-Gonzalez and M. Olivucci, J. Phys. Chem. Lett., 2017, 8, 5222–5227 CrossRef CAS PubMed.
  35. A. V. Sharkov, A. V. Pakulev, S. V. Chekalin and A. Y. A. Matveetz, Biochim. Biophys. Acta, 1985, 808, 94–102 CrossRef CAS.
  36. M. C. Nuss, W. Zinth, W. Kaiser, E. Kölling and D. Oesterhelt, Chem. Phys. Lett., 1985, 117, 1–7 CrossRef CAS.
  37. F. Gai, K. C. Hasson, J. C. McDonald and P. A. Anfinrud, Science, 1998, 279, 1886–1891 CrossRef CAS PubMed.
  38. S. Ruhman, B. Hou, N. Friedman, M. Ottolenghi and M. Sheves, J. Am. Chem. Soc., 2002, 124, 8854–8858 CrossRef CAS PubMed.
  39. R. Mathies, Nat. Chem., 2015, 7, 945–947 CrossRef CAS PubMed.
  40. M. F. Brown, M. P. Heyn, C. Job, S. Kim, S. Moltke, K. Nakanishi, A. A. Nevzorov, A. V. Struts, G. F. J. Salgado and I. Wallat, Biochim. Biophys. Acta, 2007, 1768, 2979–3000 CrossRef CAS PubMed.
  41. G. G. Kochendoerfer, P. J. E. Verdegem, I. v. d. Hoef, J. Lugtenburg and R. A. Mathies, Biochemistry, 1996, 35, 16230–16240 CrossRef CAS PubMed.
  42. A. Cembran, R. González-Luque, L. Serrano-Andrés, M. Merchán and M. Garavelli, Theor. Chem. Acc., 2007, 118, 173–183 Search PubMed.
  43. S. Moltke, A. A. Nevzorov, N. Sakai, I. Wallat, C. Job, K. Nakanishi, M. P. Heyn and M. F. Brown, Biochemistry, 1998, 37, 11821–11835 CrossRef CAS PubMed.
  44. G. Eyring, B. Curry, A. Broek, J. Lugtenburg and R. Mathies, Biochemistry, 1982, 21, 384–393 CrossRef CAS PubMed.
  45. A. Kawanabe, Y. Furutani, K.-H. Jung and H. Kandori, J. Am. Chem. Soc., 2007, 129, 8644–8649 CrossRef CAS PubMed.
  46. A. Wand, R. Rozin, T. Eliash, K.-H. Jung, M. Sheves and S. Ruhman, J. Am. Chem. Soc., 2011, 133, 20922–20932 CrossRef CAS PubMed.
  47. A. Cheminal, J. Leonard, K. S. Young, J. Kwang-Hwan, K. Hideki and H. Stefan, Phys. Chem. Chem. Phys., 2015, 17, 25429–25439 RSC.
  48. Y. Wada, A. Kawanabe, Y. Furutani, H. Kandori and H. Ohtani, Chem. Phys. Lett., 2008, 453, 105–108 CrossRef CAS.
  49. P. J. M. Johnson, A. Halpin, T. Morizumi, V. I. Prokhorenko, O. P. Ernst and R. J. D. Miller, Nat. Chem., 2015, 7, 980–986 CrossRef CAS PubMed.
  50. J. P. Kraack, T. Buckup, N. Hampp and M. Motzkus, ChemPhysChem, 2011, 12, 1851–1859 CrossRef CAS PubMed.
  51. G. Zgrablic, S. Haacke and M. Chergui, Chem. Phys., 2007, 338, 168–174 CrossRef CAS.
  52. J. P. Kraack, T. Buckup and M. Motzkus, Phys. Chem. Chem. Phys., 2011, 13, 21402–21410 RSC.
  53. J. P. Kraack, T. Buckup and M. Motzkus, Phys. Chem. Chem. Phys., 2012, 14, 13979–13988 RSC.
  54. J. P. Kraack, T. Buckup and M. Motzkus, J. Phys. Chem. Lett., 2013, 4, 383–387 CrossRef CAS PubMed.
  55. T. Sovdat, G. Bassolino, M. Liebel, C. Schnedermann, S. P. Fletcher and P. Kukura, J. Am. Chem. Soc., 2012, 134, 8318–8320 CrossRef CAS PubMed.
  56. T. Wende, M. Liebel, C. Schnedermann, R. J. Pethick and P. Kukura, J. Phys. Chem. A, 2014, 118, 9976–9984 CrossRef CAS PubMed.
  57. A. Wand, B. Loevsky, N. Friedman, M. Sheves and S. Ruhman, J. Phys. Chem. B, 2013, 117, 4670–4679 CrossRef CAS PubMed.
  58. C. Schnedermann, V. Muders, D. Ehrenberg, R. Schlesinger, P. Kukura and J. Heberle, J. Am. Chem. Soc., 2016, 138, 4757–4762 CrossRef CAS PubMed.
  59. C. Schnedermann, X. Yang, M. Liebel, K. M. Spillane, J. Lugtenburg, I. Fernández, A. Valentini, I. Schapiro, M. Olivucci, P. Kukura and R. A. Mathies, Nat. Chem., 2018, 10, 449–455 CrossRef CAS PubMed.
  60. I. Schapiro, J. Phys. Chem. A, 2016, 120, 3353–3365 CrossRef CAS PubMed.
  61. S. L. Dexheimer, Q. Wang, L. A. Peteanu, W. T. Pollard, R. A. Mathies and C. V. Shank, Chem. Phys. Lett., 1992, 188, 61–66 CrossRef CAS.
  62. T. Kobayashi, T. Saito and H. Ohtani, Nature, 2001, 414, 531–534 CrossRef CAS PubMed.
  63. A. Yabushita and T. Kobayashi, Biophys. J., 2009, 96, 1447–1461 CrossRef CAS PubMed.
  64. M. Motzkus, S. Pedersen and A. H. Zewail, J. Phys. Chem. A, 1996, 100, 5620–5633 CrossRef CAS.
  65. J. P. Kraack, A. Wand, T. Buckup, M. Motzkus and S. Ruhman, Phys. Chem. Chem. Phys., 2013, 15, 14487–14501 RSC.
  66. T. Buckup and M. Motzkus, Annu. Rev. Phys. Chem., 2015, 65, 39–57 CrossRef PubMed.
  67. S. Fujiyoshi, S. Takeuchi and T. Tahara, J. Phys. Chem. A, 2003, 107, 494–500 CrossRef CAS.
  68. J. P. Kraack, M. Motzkus and T. Buckup, J. Chem. Phys., 2011, 135, 224505 CrossRef PubMed.
  69. I. Palings, J. A. Pardoen, E. v. d. Berg, C. Winkel, J. Lugtenburg and R. A. Mathies, Biochemistry, 1987, 26, 2544–2556 CrossRef CAS PubMed.
  70. S. Smith, J. A. Pardoen, J. Lugtenburg and R. A. Mathies, J. Phys. Chem., 1987, 91, 804–819 CrossRef CAS.
  71. S. W. Lin, M. Groesbeek, I. v. d. Hoef, P. Verdegem, J. Lugtenburg and R. A. Mathies, J. Phys. Chem. B, 1998, 102, 2787–2806 CrossRef CAS.
  72. Y. Furutani, A. Kawanabe, K.-H. Jung and H. Kandori, Biochemistry, 2005, 44, 12287–12296 CrossRef CAS PubMed.
  73. A. Kawanabe, Y. Furutani, K.-H. Jung and H. Kandori, Biochemistry, 2006, 45, 4362–4370 CrossRef CAS PubMed.
  74. J. E. Kim and R. A. Mathies, J. Phys. Chem. A, 2002, 106, 8508–8515 CrossRef CAS PubMed.
  75. D. Agathangelou, Y. Orozco-Gonzalez, M. d. C. Marin, P. P. Roy, J. Brazard, H. Kandori, K.-H. Jung, N. Ferre, T. Buckup, M. Olivucci and S. Haacke, Faraday Discuss., 2018, 207, 55–75 RSC.
  76. S. Shim, J. Dasgupta and R. A. Mathies, J. Am. Chem. Soc., 2009, 131, 7592–7597 CrossRef CAS PubMed.
  77. I. Palings, E. M. v. d. Berg, J. Lugtenburg and R. A. Mathies, Biochemistry, 1989, 28, 1498–1507 CrossRef CAS PubMed.
  78. H. J. M. d. Groot, G. S. Harbison, J. Herzfeld and R. G. Griffin, Biochemistry, 1989, 28, 3346–3353 CrossRef PubMed.
  79. S. o. Smith, H. J. M. d. Groot, R. Gebhard, I. M. L. Courtin, J. Lugtenburg, J. Herzfeld and R. G. Griffin, Biochemistry, 1989, 28, 8897–8904 CrossRef CAS PubMed.
  80. H. E. Schaffer, R. R. Chance, K. K. R. J. Silbey and R. R. Schrock, J. Chem. Phys., 1991, 94, 4161–4170 CrossRef CAS.
  81. T. Kupka, A. Buczek, M. A. Broda, M. Stachów and P. Tarnowski, J. Mol. Model., 2016, 22, 101–111 CrossRef PubMed.
  82. M. Sugihara, J. Hufen and V. Buss, Biochemistry, 2006, 45, 801–810 CrossRef CAS PubMed.
  83. L. Vogeley, O. A. Sineshchekov, V. D. Trivedi, J. Sasaki, J. L. Spudich and H. Luecke, Science, 2004, 306, 1390–1393 CrossRef CAS PubMed.
  84. A. Wand, N. Friedman, M. Sheves and S. Ruhman, J. Phys. Chem. B, 2012, 116, 10444–10452 CrossRef CAS PubMed.
  85. G. Tomasello, G. Olaso-Gonzalez, P. Altoe, M. Stenta, L. Serrano-Andres, M. Mercha, G. Orlandi, A. Bottoni and M. Garavelli, J. Am. Chem. Soc., 2009, 131, 5172–5186 CrossRef CAS PubMed.


Electronic supplementary information (ESI) available. See DOI: 10.1039/c8cp05469j

This journal is © the Owner Societies 2018