Control of retinal isomerization in bacteriorhodopsin in the high-intensity regime

Abstract

A learning algorithm was used to manipulate optical pulse shapes and optimize retinal isomerization in bacteriorhodopsin, for excitation levels up to 1.8 × 10¹⁶ photons per square centimeter. Below 1/3 the maximum excitation level, the yield was not sensitive to pulse shape. Above this level the learning algorithm found that a Fourier-transform-limited (TL) pulse maximized the 13-cis population. For this optimal pulse the yield increases linearly with intensity well beyond the saturation of the first excited state. To understand these results we performed systematic searches varying the chirp and energy of the pump pulses while monitoring the isomerization yield. The results are interpreted including the influence of 1-photon and multiphoton transitions. The population dynamics in each intermediate conformation and the final branching ratio between the all-trans and 13-cis isomers are modified by changes in the pulse energy and duration.

Bacteriorhodopsin (bR) is a photosynthetic protein found in the purple membrane of Halobacterium salinarum and capable of conversion of solar energy into chemical energy. This energy conversion is efficient (1–3) and has several possible applications (4–12). A retinal chromophore is responsible for photon absorption. After photoexcitation retinal undergoes ultrafast isomerization from the all-trans to a 13-cis configuration, accompanied by additional changes in the conformation of bR (3, 8). The initial steps of the bR photocycle (see Fig. 1) have been studied intensively (11, 13–33), but there are still unanswered questions regarding the electronic potential energy surfaces (PES) of retinal, the interaction with its surroundings in the protein, and related ultrafast vibrational coupling. A number of models have been proposed, each explaining parts of the large number of experiments (4, 13, 14, 16, 34–40). Attempts have been made to reconcile the differences between these models (4).

Fig. 1.

Scheme of the isomerization reaction of retinal in bR. H = the conformation excited vertically from the all-trans (bR₅₇₀) ground state. S_n = the higher excited state reached through absorption from the H state. I₄₆₀ is the conformer corresponding to the shallow bottom of the first excited electronic potential energy surface. The red arrow depicts the transition from I₄₆₀ to the ground all-trans conformation. The J₆₂₅ conformer is tentatively assigned to the conical intersection region (see discussion below). K₅₉₀ designates the 13-cis isomer.

We aim to understand how the optical pulse shape and intensity affect the all-trans → 13-cis yield and to explore potential pathways for producing high photoproduct yields on an ultrafast time scale. This is relevant for energy storage using bio-molecular machines (9, 32). Recently, Prokhorenko et al. showed that the isomerization yield of retinal in bR could be manipulated in a low intensity, biologically relevant regime through the use of phase and amplitude shaped optical fields (25). Modifications of as much as ±20% were observed compared with unshaped pulses capable of exciting an equal number of molecules. Yet the ultimate yields remain small as photon flux was restricted to excite ≈0.3% of the chromophores in the excitation volume. In a different experiment, Vogt et al. used much higher intensity, shorter wavelength pump pulses to excite bR and a shaped 800-nm dump pulse to study the evolution of the molecule on the excited state PES (30). They found that the excited population is transferred most effectively back to the all-trans state by means of a near-infrared “dump” pulse with a delay of 200 fs, and with the minimum bandwidth for its pulse duration (i.e., a TL pulse). This is consistent with the dynamics of the I₄₆₀ intermediate state. We investigate the pulse shape dependence of the isomerization yield at pulse excitation levels up to 2 orders of magnitude higher than previous studies (25), so that pump-dump mechanisms or multiphoton excitation can help to control the molecular dynamics. These studies provide information on excited state dynamical mechanisms that produce increased yields of photoproduct. We find that short, intense pulses increase the isomerization yield by >50% over the same energy delivered in long, low intensity pulses. The yield increases approximately linearly well beyond the saturation of the initial 1-photon transition. To investigate the mechanism we perform a number of systematic scans while monitoring signals proportional to the population of the initial conformations involved in the isomerization reaction. We find that higher excited states (S_n) play a significant role in enhancing the yield. We present a model describing the multipathway isomerization process.

Results

We report on 5 types of transient absorption experiments. The first type is white-light continuum absorption spectroscopy from 460 to 870 nm. Here, the data demonstrate that the photoproduct observed at long times (38–40 ps) is consistent with the formation of the K intermediate. The remaining 4 experiments detect narrow-band time-delayed spectral absorption to determine how intermediate state formation is controlled by the shape of the excitation pulse (see Fig. 2), A learning algorithm was used to allow multiparameter modification of the excitation pulses. Searches were performed to identify optimal pulses to enhance and minimize photoproduct formation probed at 650 nm 30–40 ps after excitation. Photoproduct formation was also analyzed using 1 parameter searches varying linear chirp and pulse intensity. Finally transient absorption kinetic traces were obtained at a range of pulse energies and chirps to probe the time evolution of the various photoproduct signals. These experiments are discussed in more detail below.

Fig. 2.

bR absorption spectrum (red line), pump spectrum (green line) and probe spectra (black line) for the 4 monitored wavelengths. The transition labels follow the nomenclature of Fig. 1. The J₆₂₅ and K₅₉₀ signals are monitored at the same wavelength (650 nm) at early (<2 ps) and late (>30 ps) pump-probe delays.

Learning Control Experiments.

Optical control experiments were carried out to find the optimal phase profile of the pump pulse to maximize the yield of the 13-cis photoproduct. A genetic algorithm (GA) search (41, 42) was used to find the optimal pulses that maximize the absorption signal at 650 nm at a time delay of 40 ps after excitation (see Fig. 2). This is the wavelength with largest differential absorption between the 13-cis and all-trans state. The time delay was sufficient to assure that all of the population had relaxed to the all-trans or 13-cis conformation. At pump energies >30 nJ, the GA consistently converged to a transform-limited (TL) pulse solution. Fig. 3Inset shows fitness vs. generation for a typical GA optimization. The black upper curve shows the best pulse shape in each generation of the evolutionary search whereas the gray curve shows the average performance of all pulse shapes in that generation. Both reach asymptotic values after ≈20 generations. The dashed line shows the 13-cis absorption for a TL pulse, and the optimization experiment converges on this value. At pump energies <30 nJ the search algorithm fails to improve the isomerization yield for any pulse shape. We also tried to minimize K production using the GA. The pulses retrieved from these minimizations were all long pulses (> 500 fs) and had no distinguishing amplitude features in the time domain and no distinguishing spectral phase features. This is a common result for minimizations performed by the GA, particularly when optimization is based on multiphoton absorption and peak intensity is the controlling factor.

Fig. 3.

Summary of absorption data. (Inset) Fitness as a function of generation for a typical GA run. The black line represents the fitness (intensity of the K absorption) of the best individual pulse in each generation, the gray line represents the average fitness for each generation and the thin dashed line represents the fitness of the worst pulse in each generation. The red line represents the fitness of a TL pulse. The GA almost doubles the intensity of the K absorption over that of random stretched pulses in the early generations, converging on the K absorption intensity characteristic of transform-limited pulses. (Main Figure) I₄₆₀, bR₅₇₀, J₆₂₅ and K₅₉₀ signals versus pump energy for transform-limited excitation pulses. The horizontal scale represents the estimated energy entering the focal volume where the pump and probe pulses overlap. The corresponding excitation level is given along the top axis. The arrow indicates the estimated linear saturation level of 4.8 × 10¹⁵ photons per square centimeter (25). The data are scaled to give similar linear behavior at low pump intensities, highlighting the differences in saturation at higher pump intensities. The I₄₆₀ signal is given by the stimulated emission (SE) at 850 nm (red curve) and excited state absorption (ESA) at 487 nm (black curve). The initial bR₅₇₀ bleach is measured at 570 nm 100 fs (solid green) and 200 fs (dashed green) after the coherent spike (zero pump-probe delay). The J₆₂₅ and K₅₉₀ signals are given by the transient absorption at 650 nm, at 2 ps (light blue) and 40 ps (dark blue) respectively. The K and J absorption signals continue to increase with pump-intensity despite the saturation of the bleaching signal and the I₄₆₀ population.

Intensity Scans.

To explore further the results of these optimization experiments we performed intensity scans in which the bR₅₇₀, I₄₆₀, J₆₂₅, and K₅₉₀ populations were monitored as a function of pump energy for TL excitation pulses at specific pump-probe wavelengths and delays (see Fig. 2). The results are plotted in Fig. 3. The green curves show the decrease (bleach) of the bR₅₇₀ absorption 100 fs and 200 fs after photoexcitation. This signal is proportional to the population removed from the ground state and has a sublinear dependence on the excitation energy. Previous studies reported that the population is initially excited to the H conformation and a fluorescence Stokes shift is observed as the population moves to I₄₆₀ in ≈200 fs (16, 23, 43). This time-delay in the growth of the I₄₆₀ signal is also observed in the present experiments (see section 2.4 below). The intensity-dependent difference between the bR₅₇₀ bleach at 200 fs (dashed green line) and 100 fs (solid green line) as pump energies exceed 30 nJ reflects the appearance of an absorbing species or depletion of an emissive species on an ultrafast time scale. This may indicate that a fraction of the initially excited population returns rapidly to the all-trans ground state. Alternatively the absorption may arise from the formation of vibrationally hot I₄₆₀ or another intermediate excited state species after multiphoton excitation to higher electronic states. The intensity dependence of the I₄₆₀ population was monitored via stimulated emission at 850 nm and excited state absorption at 487 nm. Both I₄₆₀ signals saturate more strongly than the all-trans bleach, demonstrating that some of the bleached population is excited from H to higher states, avoiding the I₄₆₀ state altogether.

The photoproduct populations at 650 nm exhibit a very different intensity dependence. The 13-cis population, K₅₉₀ (dark blue), observed 40 ps after excitation exhibits a nearly linear dependence on intensity. The J₆₂₅ absorption monitored at 2 ps (light blue) displays a similar dependence on the pump energy, although there may be a slight energy dependent mismatch between the J₆₂₅ and K₅₉₀ curves. The pulse energy dependence of the bR₅₇₀ bleach at 570 nm 30 ps after excitation is consistent with the K₅₉₀ and J₆₂₅ signals as expected for depletion of the reactant trans-retinal and production of the photoproduct. The near linear dependence of the ultimate photoproduct yield is in distinct contrast to the sublinear behavior ofthe initial bR₅₇₀, bleach and the I₄₆₀ emission. The increase in the K₅₉₀ yield can be maintained only if the branching efficiency toward this conformation increases with intensity.

Linear Chirp Scans.

The influence of pulse shape on isomerization yield was explored further by monitoring the different photoproduct signals while systematically varying the second order spectral phase (linear chirp) of the excitation pulses. Linear chirp can be a useful tool to probe the influence of pulse shape on excited state population or wavepacket dynamics in absorption to 1-photon allowed excited states and to probe the spectral resonances in resonant 2-photon absorption (44). The magnitude of the K₅₉₀ absorption signal depends on linear chirp only at energies >30 nJ whereas at low energies the chirp scans are flat within the ≈ 5% RMS noise of the measurement. Fig. 4 compares the K₅₉₀ (red squares) and J₆₂₅ (blue diamonds) absorption signals as a function of linear chirp for 80 nJ excitation pulses. Both signals are maximized by a pulse with a small negative chirp, near the limit of the temporal resolution. These data suggest that the higher excited state responsible for the enhanced production of K is primarily accessed from the Franck–Condon region of the initially excited H state (Fig. 1). Evolution of the sample away from this region reduces the overall signal from this channel as evinced in the rapid drop-off with chirp. We also observe a small asymmetry for pulses of opposing linear chirp. This asymmetry, which is seen as a steeper drop in photoproduct formation with positive chirp, could be due to the slightly asymmetric excitation spectrum. It could also reflect the influence of a pump-dump pathway reducing the excited state population. Excited state depletion is often observed with positively chirped excitation. The difference between the early (J₆₂₅) and late (K₅₉₀) signals reflects an influence of excitation pathway (1-photon vs. 2-photon) on the photoproduct relaxation dynamics.

Fig. 4.

Probe absorption change vs. chirp of the pump pulse. (Upper) The J₆₂₅ (blue diamonds) and K₅₉₀ (red squares) signals versus linear chirp at 80 nJ pulse energy. (Lower) K₅₉₀ signal versus linear chirp at 12 nJ pulse energy (green triangles). All signals are measured at 650 nm. The J₆₂₅ signal is measured at 2 ps and the K₅₉₀ signal is measured at 40 ps.

Transient Scans.

The above analysis maps the multipathway evolution of the retinal chromophore but does not explain the mechanism through which the branching ratio is controllable. To address this question we performed transient scans with several different excitation pulses while monitoring the dynamics of the conformations involved in the isomerization process. Fig. 5Top shows transient scans taken at 570 nm, the peak of the bR₅₇₀ ground state absorption. The absorption at 570 nm has contributions from the all-trans conformer and intermediates I₄₆₀, J₆₂₅ and K₅₉₀, and this complicates the interpretation of the bleach recovery. Nevertheless, the bR₅₇₀ bleach dominates the signal at this wavelength, particularly at early times, before J₆₂₅ and K₅₉₀ come into play. The decay of the bleach has 3 notable features. The first feature is an intensity dependent absorption signal at early times that is altered by the pump phase and energy. This feature is responsible for the difference in the intensity dependence 100 fs and 200 fs after excitation (see Fig. 3). At low pump excitation levels, there is an increase in the bleaching signal at early times followed by a decay, whereas at higher energies there is an enhanced bleach at early times followed by a somewhat slower decay leading to a plateau between ≈100 and 200 fs. The second feature, whose onset is marked by the vertical arrow at T − T₀ ≈ 250 fs, can be modeled as an exponential decay with a time constant T₁ ≈ 600 fs. This is responsible for most of the decay of the bleach corresponding to repopulation of the bR ground state or production of the J₆₂₅-K₅₉₀ states, which also absorb weakly at this wavelength. The third feature has an intensity dependent time constant T₂ ≈ 5–15 ps corresponding to relaxation of the photoproducts. The relative magnitude of the overall bleach recovery is dependent on pump energy.

Fig. 5.

Transient scans taken at 570 nm (Top), 650 nm (Middle) and 850 nm (Bottom). The monitored signals are, in order: the bR₅₇₀ (all-trans) ground state bleach the J₆₂₅ absorption and the I₄₆₀ stimulated emission. The arrow in Top indicates the onset of the exponential recovery of the bleach. The negative chirp pulse has a linear chirp rate of −0.35 × 10⁵ fs². In Middle, the traces have been normalized by the amplitude of the corresponding I₄₆₀ trace, generated in identical excitation conditions.

Fig. 5 Middle shows the intensity-dependent dynamics of the transient absorption signal at 650 nm, where the J₆₂₅ → K₅₉₀ transition is monitored. At early times this signal also contains contributions due to stimulated emission as the fluorescence undergoes a dynamic stokes shift (43). At low excitation intensity this stimulated emission feature is clearly observed, decaying on a ≈150 fs time scale as the fluorescence spectrum shifts to the red. The observed rise times of the absorption due to the J₆₂₅ intermediate is within 10% of 400 fs for all pump energies used. The ensuing decay of the 650-nm signal is caused by the relaxation toward the K₅₉₀ conformer. For excitation with 80 nJ pulses the relaxation is completed within 30–40 ps. For lower energies the relaxation is completed within 10–20 ps. Transient scans with lengths up to 150 ps show that no changes are observed after 40 ps for all traces. Transient absorption spectra at 38–40 ps were consistent with the formation of J-K with no other species contributing to the difference spectrum between 460 nm and 870 nm. The 3 traces shown in Fig. 5 are normalized using the amplitude of the corresponding I₄₆₀ signal generated under identical excitation conditions. The relative intensity of the J₆₂₅ signal obtained with the highest intensity excitation is significantly larger than that obtained with low excitation intensity. This supports the conclusion that some of the excited population reaches the photoproduct channel without passing through I₄₆₀, in agreement with the saturation of the I₄₆₀ curve in Fig. 3.

Fig. 5 Bottom shows the time dynamics of the stimulated emission of the I₄₆₀ state, measured at 850 nm. Unlike the measurements performed at other probe wavelengths, which are affected by spectral congestion from multiple intermediates, the 850-nm optical signal observed at low excitation intensity can be assigned to I₄₆₀ stimulated emission alone. As the intensity is increased a fast negative transient appears in the data where the pump and probe pulses overlap in time. This transient could arise from a Raman contribution to the signal or a short-lived stimulated emission from the state excited after 2-photon absorption. The fast transient is followed by the rise and decay of the stimulated emission signal. The stimulated emission signal attributed to I₄₆₀ is not well described by a sum of exponentials. The signal reaches a maximum 200 fs after T₀ with a rise that is distinctly nonexponential. The decay of the signal is described moderately well by an exponential decay with a 1/e time from the peak of the signal of ≈600 fs with TL 80 nJ excitation, 680 fs with a chirped 80 nJ excitation (linear chirp rate of −0.35 × 10⁵ fs²) and 750 fs with TL 37 nJ excitation, although the overall signal decay at this wavelength is better described as multiexponential. The change in the decay of I₄₆₀ may represent changes in the relative amplitudes of different components that are not well defined in the short 2-ps time window available. The I₄₆₀ decay and bR₅₇₀ bleach recovery occur on similar time scales confirming that the I₄₆₀ population is the dominant source for the bR₅₇₀ bleach recovery under all excitation conditions.

These data lead to the question of how intensity influences the I₄₆₀ population dynamics. Studies have shown that I₄₆₀ can be also reached through the all-trans → H → S_n → I₄₆₀ pathway and the corresponding I₄₆₀ population has a faster decay (17, 30, 43). Transient absorption scans taken at 487 nm showed features at T − T₀ < 200 fs that did not match well the cross correlation measurements in the neat buffer solution. The mismatch could be due to fast relaxation of some of the S_n population to I₄₆₀. This component of the I₄₆₀ population will correspond to hotter vibrational levels and will easily bypass the small barrier to the right (see Fig. 1). It is not clear from the available data whether the influence of pulse parameters on the I₄₆₀ population is simply described by modifications in the peak intensity of the pulse and the relative importance of 1-photon and multiphoton pathways or whether more complicated pathways could play a role. This is a subject worth careful investigation. An alternative way to create a hot I₄₆₀ population in a controlled fashion is through stimulated Raman scattering between S_n and H. This mechanism is not mentioned in the literature, but it is consistent with the intensity-dependent effects observed here.

Discussion and Conclusions

The model emerging from the data (see Fig. 1) demonstrates the presence of at least 2 pathways for the formation of the 13-cis K₅₉₀ conformation. With low intensity excitation and phase-only control the GA was unable to identify pulse shapes capable of modifying the formation of the photoproduct. This is in direct contrast to the control observed in the study in ref. 25. The contrast in controllability may relate to the differences in available bandwidth or peak wavelength, but highlights the need for additional work to characterize the mechanisms for control.

In the present study high-intensity excitation opens pathways for one or more additional excited states S_n, reached through multiphoton excitation. The GA consistently found TL pulses to optimize the photoproduct formation whereas anti-optimization minimizing photoproduct formation led to long pulses without any outstanding features. It is interesting that the GA fails to identify pathways characterized by more complex interaction at the highest intensities where multiple resonant multiphoton processes could be accessed. From the higher excited states some portion of the population couples back to I₄₆₀. Potential coupling mechanisms include nonradiative internal conversion and stimulated Raman scattering between H and S_n. A common consequence of these 2 pathways is that the population will be vibrationally hot when reaching I₄₆₀, accounting for the more rapid decay of this state at high intensity. It is also apparent that much of the S_n population bypasses the I₄₆₀ region altogether, as suggested particularly by the intensity scans (Fig. 3) and by the normalized J₆₂₅ transient scans (Fig. 5 Middle). This population could convert to the 13-cis conformation through the same conical intersection region accessed after linear excitation, or through another conical intersection. The isomerization due to this process must be exceedingly efficient, because it competes favorably with the already efficient process that proceeds through the initial 1-photon excitation channel. The dominant multiphoton pathway results in the formation of the 13-cis photoproduct with unit or near unit efficiency.

Quantum coherence does not appear to play a significant role in the experiments described in this work. Weak vibrational coherences with a period of a few hundred wave numbers are observed in some of the transient scans for chirped pulse excitation, but not for the optimal, TL pulse. The related analysis will be presented elsewhere. Refs. 15, 21, and 45 report high frequency oscillations that persist for >1 ps. The resolution needed to observe them is beyond the capabilities of our experimental setup. The model presented above does not require coherent wave packet dynamics.

The experiments presented in this work report control of the isomerization efficiency of retinal in bR in an intensity regime, which is not found in nature. The optimal solution is simple: the pulse of highest intensity/shortest duration maximizes 13-cis isomerization yield (37). However, the postexcitation molecular dynamics are complex and clearly suggest that the population of each conformer along the multipath trajectory is influenced by the pulse energy and duration. At lower intensities, the yield does not depend on phase in our measurements. The findings presented in this work show that bacteriorhodopsin carries out its function efficiently over a range of light level conditions far beyond those found in nature. Solar radiation provides broad spectral excitation very different from the ideal characteristics of the laser radiation used in these control experiments. Yet the isomerization channel identified here may have biological relevance. Biological systems are vulnerable to photodamage after UV excitation. The isomerization channel observed in these experiments rapidly and efficiently channels high energy excitation into a biologically useful channel. It would be interesting to compare isomerization of retinal in different environments after multiphoton or direct UV excitation to determine if the protein environment tunes or controls this isomerization channel.

Materials and Methods

Tunable noncollinear optical parametric amplifiers and programmable acousto-optic pulse shapers were used in this work, together with standard white light ultrafast probe techniques (46, 47). The details are discussed in SI Text. The bR preparation was also similar to previous photoabsorption experiments on this molecule (25, 30, 45). Care was taken to limit the total bR exposure to laser radiation to avoid the effects of permanent photo-induced damage, although the intensity-dependent nature of this study necessitated higher photon doses than some previous studies. The integrated photon dose was kept to ≈5 or fewer absorbed photons per molecule.