Observe while it happens : catching photoactive proteins in the act with non-adiabatic molecular dynamics simulations

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY-NC-ND 4.0 https://creativecommons.org/licenses/by-nc-nd/4.0/ Observe while it happens : catching photoactive proteins in the act with non-adiabatic molecular dynamics simulations © 2020 The Authors. Published by Elsevier Ltd. Published version Groenhof, Gerrit; Modi, Vaibhav; Morozov, Dmitry Groenhof, G., Modi, V., & Morozov, D. (2020). Observe while it happens : catching photoactive proteins in the act with non-adiabatic molecular dynamics simulations. Current Opinion in Structural Biology, 61, 106-112. https://doi.org/10.1016/j.sbi.2019.12.013 2020 Observe while it happens: catching photoactive proteins in the act with non-adiabatic molecular dynamics simulations Gerrit Groenhof, Vaibhav Modi and Dmitry Morozov Organisms use photo-receptors to react to light. The first step is usually the absorption of a photon by a prosthetic group embedded inside the photo-receptor, often a conjugated chromophore. The electronic changes in the chromophore induced by photo-absorption can trigger a cascade of structural or chemical transformations that culminate into a response to light. Understanding how these proteins have evolved to mediate their activation process has remained challenging because the required time and spacial resolutions are notoriously difficult to achieve experimentally. Therefore, mechanistic insights into photoreceptor activation have been predominantly obtained with computer simulations. Here we briefly outline the challenges associated with such computations and review the progress made in this field. Address Nanoscience Center and Department of Chemistry, P.O. Box 35, 40014 University of Jyva ¨skyla ¨, Finland Corresponding author: Groenhof, Gerrit ([email protected]) Current Opinion in Structural Biology 2020, 61:106–112 This review comes from a themed issue on Theory and simulation Edited by Alan E Mark and Christine Peter For a complete overview see the Issue and the Editorial Available online 9th January 2020 https://doi.org/10.1016/j.sbi.2019.12.013 0959-440X/ã 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Introduction Life on earth is powered by the sun. To harness its power, organisms have developed a wide range of mechanisms for collecting and storing solar energy. In addition, many organisms have also evolved photo-sensory proteins to detect and respond to light. While the various photo-response mechanisms may be highly diverse, they share the common feature that the first step consists of the absorption of a photon by a chromophore co-factor inside the protein matrix. The sudden electronic response of that chromophore to photo-absorption can subsequently trigger a series of chemical or conformational changes that ultimately leads to the capture of the photon’s energy into a covalent bond or to the activation of a signaling cascade. High-resolution structures of photoactive proteins have been obtained with X-ray crystallography and NMR spectroscopy, but these techniques can only capture static snapshots of what is essentially a dynamic process. While the dynamical response to photon absorption can be probed accurately with transient spectroscopy methods, these techniques are only sensitive to molecular energy levels and thus do not provide atomistic information about the photo-induced structural changes. Therefore, experimentally it remains challenging to simultaneously access the time and spacial resolutions required for understanding the activation mechanism of photoactive proteins in atomic detail. In contrast, computationally the relevant time and spacial resolution can both be accessed routinely, and molecular dynamics (MD) computer simulations have become the method of choice to advance our understanding of how these proteins have evolved to mediate their photochemistry. Indeed, since MD trajectories contain information about the time-evolution of both atomic positions and energies, simulations provide a link between static structure determination and transient spectroscopy. Here, we outline the main challenges associated with modeling the ultra-fast dynamics when photoactive proteins absorb light, and illustrate the progress made in this field with recent applications. We conclude this review with an outlook on how we expect this field to develop further. Challenges Photon absorption induces a transition from the electronic ground state (S0) into an electronic state of higher energy, usually the first singlet excited state (S1). Because the potential energy surface associated with an electronic excited state differs from that of the electronic ground state, photo-absorption can trigger an immediate dynamic response (Figure 1). However, due to the over-damped and highly dissipative molecular surroundings of proteins, part of the energy released during the initial structural relaxation is quickly lost. Furthermore, the total excited state lifetime is limited due to spontaneous emission. Therefore, to make sufficient energy available for the downstream chemistry under these constraints, it is imperative to decay rapidly from the excited state into a metastable product on the electronic ground state that is chemically or structurally different from the initial configuration (Figure 1). Such decay occurs most efficiently Available online at www.sciencedirect.com ScienceDirect Current Opinion in Structural Biology 2020, 61:106–112 www.sciencedirect.com at or near so-called conical intersections between the ground and excited state potential energy surfaces [1]. Thus, modeling photo-induced chemistry in photoactive proteins requires, in addition to a structural model, an accurate description of the excited-state electronic wave function, as well as an approach to deal with the radiationless deactivation at conical intersections. Excited-State Quantum Chemistry While quantum chemistry has matured to the point that chemical accuracy can be achieved for processes that occur in the electronic ground state, the possibilities to model excited states are still limited as reliable methods that can provide a balanced description of both ground and excited states, are available only for small molecules and at high computational expense [4]. Owing to their typical size and complexity, modeling the excited states of photo-receptors therefore requires additional approximations, ranging from using isolated molecular cluster models [5] to hybrid quantum mechanics/molecular mechanics (QM/MM) treatments [6], in which the atoms involved in the photo-chemistry are described at a suitable level of ab initio theory or time-dependent density functional theory (TDDFT), while the remainder of the system, including the rest of the protein, solvent and ions, is modeled with a computationally cheaper molecular mechanics force field [7]. Non-adiabatic molecular dynamics At a conical intersection between electronic potential energy surfaces, the adiabatic, or Born-Oppenheimer picture, in which molecular dynamics evolve on a single potential energy surface, breaks down [8]. Instead, near the intersection the electronic states are coupled and nuclear motion can induce so called non-adiabatic transitions between these electronic states. Although the break-down of the Born-Oppenheimer approximation implies that also the nuclear dynamics can no longer be described with Newton’s equations of motion, the non-adiabatic effects can be satisfactorily modeled with so-called semi-classical MD methods [9]. A popular approach to include non-adiabatic transitions in classical MD simulations is Surface Hopping. Introduced by Tully over three decades ago [10], the trajectory can hop between two potential energy surfaces. The hops are instantaneous and triggered by changes in the quantum mechanical populations of the electronic states due to their non-adiabatic coupling with nuclear degrees of freedom [9]. A promising alternative to surface hopping is the full multiple spawning (FMS) method of Martı ´nez and co-workers [11], which solves the electronic and nuclear Schro ¨dinger equations simultaneously and has been adapted to QM/MM simulations [12]. Exciton dynamics in multi-chromophoric systems Inordertocapture photonsunderthelowlightintensitiesin some habitats on earth, photosynthetic organisms evolved elaborate strategies for increasing the absorption crosssection of their photosynthesis apparatus. Thus, in addition to reaction centers, in which photons are transformed into chemical free energies through photo-induced redox chemistry, these organisms express light-harvesting proteins that self-organize into large complexes around the reaction centers [13]. Because these proteins contain one or more chromophores, the likelihood of absorbing a photon increases, while the absorbed photon is quickly channeled into the reaction center through the excitonic couplings between the chromophores. Simulating these processes on a computer would in principle require that all chromophores are included inside the QM region. However, since excited state quantum chemistry is challenging even for a single chromophore [4], calculating the excited state dynamics of a multi-chromophore system would be computationally prohibitive unless additional approximations to the ones already introduced above, are made. In the exciton model Non-adiabatic dynamics in photoactive proteins Groenhof, Modi and Morozov 107 Figure 1 Current Opinion in Structural Biology ν Schematic representation of the potential energy surfaces involved in a photo-chemical reaction. The photo-isomerization in Photoactive Yellow Protein is used as an example [2]: upon photon absorption, the covalently bound para-coumaric acid chromophore (pCA) undergoes rapid rotation around the central double bond from the trans configuration into a 90twisted configuration. In this twisted geometry, the electronic ground (S0) and excited (S1) states are nearly degenerate, and radiationless deactivation to S0occurs. After decay, the system relaxes into a new local minimum on the S0potential energy surface, associated with a strained cis chromophore that stores sufficient energy for subsequent conformational changes [3]. www.sciencedirect.com Current Opinion in Structural Biology 2020, 61:106–112 [14  , 16,17  ,18] rather than including all chromophores in a single quantum chemistry calculation, the excited state wave functions are computed for each of the chromophores separately instead. The excitonic Hamiltonian, containing the excitation energies of the chromophores on the diagonal and the dipole–dipole excitonic couplings between the chromophores as off-diagonal elements, is diagonalized to yield the multi-chromophore wave functions, energies, gradients and non-adiabatic couplings, required for semi-classical MD simulations of multichromophore complexes [15,17  ] (Figure 2). Applications While the first semi-empirical excited-state MD simulation of retinal in a Rhodopsin model dates back more than four decades [19], the field did not mature until the early 2000s, when crystal structures of several photoreceptors were available and computer technology had advanced enough for fully atomistic QM/MM simulations of Bacteriorhodopsin [20,21] and Photoactive Yellow Protein [2]. The total number of trajectories in those early studies was relatively low, but nevertheless yielded a consistent picture of the photo-isomerization process and the role of the protein environment. Since then, excited state dynamics simulations have been used to investigate retinal photo-isomerization in other proteins as well, including bovine Rhodopsin [22–24], squid Rhodopsin [25], Melanopsin [26], channel Rhodopsin [27  ], and Halorhodopsin [28,29]. Furthermore, non-adiabatic QM/MM trajectories have also been computed for chromophore photo-isomerization and excited-state proton transfer in various fluorescent proteins [30,31,32,33  ,34]. Comparison to time-resolved spectroscopy The accuracy of the QM/MM simulation models, which often rely on a minimal QM region embedded within a non-polarizable MM environment, remains a matter of concern. Because experimental data on ultra-fast structural dynamics are still very rare, the validity of the simulations can only be verified indirectly by computing and comparing spectroscopic observables. While quantum yields and excited-state lifetimes are relatively straightforward to compute and compare, it is also possible to compute the time-evolution of spectral signals, and compare to data from fluorescence up-conversion or transient absorption experiments [23]. For example, the agreement between the fluorescence decay calculated from 60 FMS trajectories of retinal photo-isomerization in Channelrhodopsin-2 [27  ] and experiment [35] suggests that despite the approximations, the QM/MM simulations can capture the non-adiabatic dynamics in this system rather accurately (Figure 3a). Comparison to two-dimensional electronic spectroscopy The energy flow inside coupled multi-chromophoric systems, such as light harvesting complexes, is conveniently probed with two-dimensional electronic spectroscopy, a technique that correlates excitation and emission energies as a function of time delay between absorption and emission [37,38]. Thus, the validity of simulations of a multi-chromophore system can verified by computing the 2D electronic spectra. In Figure 3b, the calculated [14  ] and measured 2D electronic spectra of Light Harvesting complex II are compared [36]. The good agreement also here suggests that the underlying simulation model is sufficiently accurate and that the simulations can be used to guide the interpretation of the measured 2D signals [14  ]. Comparison to time-resolved serial femtosecond X-ray crystallography Time-resolved crystallography at free electron lasers has opened up a new experimental window into the regime of ultra-fast molecular dynamics [39]. Thanks to femtosecond pulses of extreme brilliance (i.e. 1012 hard X-ray 108 Theory and simulation Figure 2 (a) (b) LH2 complex Exciton evolution B850 ring B800 ring Current Opinion in Structural Biology Panel (a): The light harvesting complex II (LH2) from Rhodoblastus acidophilus, containing 27 Bacteriochlorophyll-a chromophores (18 B850, 9 B800) that are excitonically coupled (reprinted with permission from [14  ]. Copyright 2017 American Chemical Society). Panel (b): Snapshots of a non-adiabatic MD trajectory that show how the excitation, initially localized on one of the chromophores, evolves through the LH2 complex (reproduced from Ref. [15  ] with permission from the PCCP Owner Societies). Current Opinion in Structural Biology 2020, 61:106–112 www.sciencedirect.com photons per pulse [40]) it has become possible to perform pump–probe experiments with X-rays on photoreceptor micro-crystals or nano-crystals. A short optical pump pulse initiates the photo-chemistry, while a femtosecond X-ray pulse probes the diffraction after a time delay [41]. Because the probe pulse captures a single protein crystal in a random orientation, a large series of diffraction patterns needs to be collected for each pump–probe delay, indexed and merged before the changes in electron density due to photon absorption can be calculated. Recently Schmidt and co-workers have applied this timeresolved Serial Femtosecond X-ray crystallography technique (tr-SFX) to Photoactive Yellow Protein [42]. Figure 4 shows how the electron density difference in the chromophore binding pocket evolves as a function of time after photo-excitation with a 140 fs laser pulse at 450 nm. Comparing the transient structures obtained by refining extrapolated difference structure factor amplitudes, to the structures of a non-adiabatic multi-configurational QM/ MM trajectory, suggests a remarkably good agreement between simulation and experiment. Similarly, also for rsEGFP2, a photo-chromic variant of the Aequorea victoria Green Fluorescent Protein, agreement was found between semi-empirical QM/MM simulations and tr-SFX experiments [33  ]. However, due to the very high power of the optical pump laser, it remains unclear whether these early experiments probed the biologically relevant photo-isomerization mechanism. Indeed, a recent power titration of Bacteriorhodopsin suggest that with the laser powers used in those tr-SFX experiments, the initial dynamics is dominated by multi-photon absorption rather than the biologically relevant single photon process [43  ]. This finding not only raises questions about the interpretation of a preceding SFX study on Bacteriorhodopsin [44], but of all femtosecond time-resolved SFX results on photoactive Non-adiabatic dynamics in photoactive proteins Groenhof, Modi and Morozov 109 Figure 3 (a) (b) Simulated transient fluorescence spectrum of Channelrhodopsin-2 Comparison of LH2 2DES maps at 300K Simulated Experimentally measured Decay of fluorescence Excitation Wavelength (nm) Intensity Intensity Intensity Wavelength (nm) Normalized intensity Detection Wavelength (nm) Time (fs) Time (fs) 800 700 600 500 400 0500 1000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 780 800 820 840 860 880 900 780 800 820 840 860 880 900 3.0 2.0 1.0 0.0 -1.0 1.4 0.0 -0.4 860 820780 780 820 860 1000 500 0 0.0 0.5 1.0 Simulation Experiment Current Opinion in Structural Biology Panel (a): time-resolved emission (top) and fluorescence up-conversion signal at a single wavelength (bottom) of Channelrhodopsin-2 computed from 60 FMS trajectories (red) [27  ], compared to experiment (blue) [35] (reprinted with permission from [27  ]. Copyright 2019 American Chemical Society). Panel (b): Simulated [14  ] and measured [36] two-dimensional electronic spectroscopy pump-probe maps of Light Harvesting complex II (LH2) (reprinted with permission from [14  ]. Copyright 2017 American Chemical Society). www.sciencedirect.com Current Opinion in Structural Biology 2020, 61:106–112 proteins so far. Although very challenging, it is possible to simulate the dynamics after multi-photon absorption. Therefore, this issue could in principle be addressed with non-adiabatic MD simulations as well. A further limitation of SFX is that dynamics can only be probed inside crystals. Because crystal packing could restrict the conformational movements of the protein, the structural changes might be different in solution. Indeed, comparing X-ray scattering from phytochrome photoreceptor proteins in solution on the one hand, and X-ray diffraction from these proteins in crystal on the other hand, suggests that the photo-induced structural changes are much larger in solution than in the crystal [45]. However, because large-scale conformational changes typically occur on timescales that are much longer than the excited state lifetime, the effects of crystal packing are probably negligible on such ultra-fast timescales. Nevertheless, also here, MD simulations can be used to systematically investigate the differences between the conformational dynamics in crystals and in solution. Summary and outlook Nature has evolved various strategies for harvesting, converting and storing solar energy. Understanding these strategies might be the key to new solar technologies. While tr-SFX at free electron lasers will most certainly lead to new and unprecedented insights into the inner workings of Nature’s photoactive protein machinery, this technique is still in its infancy and access to free electron facilities is limited. However, the agreement between simulations and SFX experiments suggest that despite limited accuracy, non-adiabatic molecular dynamics are a good alternative to SFX. Future developments will undoubtedly lead to more accurate QM/MM models with, for example, a multi-reference rather than multiconfigurational description of the QM region, as recently demonstrated for an isolated GFP chromophore [46  ], or with a polarizable MM environment rather than static point charges [47]. Furthermore, in contrast to experimental data, trajectories can be analyzed at the atomistic level and thus provide insights beyond the reach of experiment. For example, information about the interactions between chromophore and protein environment, which is essential to understand the catalytic effect of the protein, can be obtained much easier from simulations than experiment and hence lead to new testable hypotheses. Because these insights can be crucial also for designing new systems, combining experiment and computation, in particular with a high-throughput strategy [48  ], might be a promising avenue towards new 110 Theory and simulation Figure 4 SFX QM/MM MD 269 fs 799 fs 3000 fs Current Opinion in Structural Biology Comparison between time-resolved crystallography (upper panel) and non-adiabatic MD simulation (bottom panel) of Photoactive Yellow Protein. Each panel shows a front and side view of the chrompohore pocket as a function of pump-probe delay. In the top panels, the experimental electron difference densities with respect to the unactivated resting state of the protein are shown. Red contours indicate loss of electron density whereas blue contours indicate gain. In all panels the X-ray structure of the resting state is shown in yellow, while the structures refined against the extrapolated difference structure factor amplitudes, are shown in pink and green [42]. Current Opinion in Structural Biology 2020, 61:106–112 www.sciencedirect.com discoveries in photobiology, as well as future technologies for harnessing the power of the sun. Conflict of interest statement Nothing declared. Acknowledgements This work has been done as part of the BioExcel CoE (www.bioexcel.eu), a project funded by the European Union contracts H2020-INFRAEDI-02- 2018-823830, H2020-EINFRA-2015-1-675728. We thank Vladimir Mironov for critically reading the manuscript. References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as:  of special interest  of outstanding interest 1. Bernardi F, Olivucci M, Robb MA: Potential energy surface crossings in organic photochemistry. Chem Soc Rev 1996, 25:321. 2. Groenhof G, Bouxin-Cademartory M, Hess B, De Visser SP, Berendsen HJC, Olivucci M, Mark AE, Robb MA: Photoactivation of the photoactive yellow protein: Why photon absorption triggers a trans-to-cis lsomerization of the chromophore in the protein. J Am Chem Soc 2004, 126:4228-4233. 3. Gamiz-Hernandez AP, Kaila VRI: Conversion of light-energy into molecular strain in the photocycle of the photoactive yellow protein. Phys Chem Chem Phys 2016, 18:2802-2809. 4. Lischka H, Nachtigallova D, Aquino AJ, Szalay PG, Plasser F, Machado FB, Barbatti M: Multireference approaches for excited states of molecules. Chem Rev 2018, 118:7293-7361. 5. Suomivuori C, Fliegl H, Starikov EB, Balaban TS, Kaila VRI, Sundholm D: Absorption shifts of diastereotopically ligated chlorophyll dimers of photosystem I. Phys Chem Chem Phys 2019, 21:6851-6858. 6. Warshel A, Levitt M: Theoretical studies of enzymatic reactions: dielectric, electrostatic and steric stabilization of carbonium ion in the reaction of lysozyme. J Mol Biol 1976, 103:227-249. 7. Modi V, Donnini S, Groenhof G, Morozov D: Protonation of the biliverdin ixa chromophore in the red and far-red photoactive states of a bacteriophytochrome. J Phys Chem B 2019, 123:2325-2334. 8. Ko ¨ppel H, Domcke W, Cederbaum LS: Multimode molecular dynamics beyond the Born–Oppenheimer approximation. Adv Chem Phys 1984, 57:59-246. 9. Crespo-Otero R, Barbatti M: Recent advances and perspectives on nonadiabatic mixed quantum-classical dynamics. Chem Rev 2018, 118:7026-7068. 10. Tully JC: Nonadiabatic molecular dynamics. Int J Quant Chem 1991, 40:299-309. 11. Ben-Nun M, Quenneville J, Martı ´nez TJ: Ab initio multiple spawning: Photochemistry from first principles quantum molecular dynamics. J Phys Chem A 2000, 104:5161-5175. 12. Curchod BF, Martı ´nez TJ: Ab initio nonadiabatic quantum molecular dynamics. Chem Rev 2018, 118:3305-3336. 13. Jang SJ, Mennucci B: Delocalized excitons in natural lightharvesting complexes. Rev Mod Phys 2018, 90:035003. 14.  Segatta F, Cupellini L, Jurinovich S, Mukamel S, Dapor M, Taioli S, Garavelli M, Mennucci B: A quantum chemical interpretation of two-dimensional electronic spectroscopy of light-harvesting complexes. J Am Chem Soc 2017, 139:7558-7567 A QM/MM model of Light Harvesting Complex II, using a TD-DFT description for the chromophores embedded inside a polarizable protein environment, is used in combination with classical molecular dynamics sampling to calculate site energies and inter-chromophore couplings. These parameters are subsequently used to simulate the flow of excitation energy in this complex and compute the two-dimensional electronic spectroscopy signatures. 15.  Sisto A, Stross C, van der Kamp MW, O’Connor M, McIntosh- Smith S, Johnson GT, Hohenstein EG, Manby FR, Glowacki DR, Martı ´nez TJ: Atomistic non-adiabatic dynamics of the lh2 complex with a GPU-accelerated ab initio exciton model. Phys Chem Chem Phys 2017, 19:14924-14936 An exciton model of Light Harvesting Complex II, using TD-DFT for the chromophores and an implicit representation of the proteins, is used to perform non-adiabatic dynamics simulations of energy transfer between the chromophores. The rate at which the excitation transfers from B800 ring to B850 ring are in a good agreement with experimental results. 16. Sisto A, Glowacki DR, Martı ´nez TJ: Ab initio nonadiabatic dynamics of multichromophore complexes: a scalable graphical-processing-unit-accelerated exciton framework. Acc Chem Res 2014, 47:2857-2866. 17.  Menger MF, Plasser F, Mennucci B, Gonza ´lez L: Surface hopping within an exciton picture. an electrostatic embedding scheme. J Chem Theory Comput 2018, 14:6139-6148 A multi-scale (QM/MM) approach is developed for non-adiabatic exciton dynamics. Direct electrostatic interactions between chromophores and protein environment are included in the calculation of the site energies and excitonic couplings. 18. Meneghin E, Volpato A, Cupellini L, Bolzonello L, Jurinovich S, Mascoli V, Carbonera D, Mennucci B, Collini E: Coherence in carotenoid-to-chlorophyll energy transfer. Nat Commun 2018, 9:3160. 19. Warshel A: Bicycle-pedal model for 1st step in vision process. Nature 1976, 260:679-683. 20. Warshel A, Chu Z: Nature of the surface crossing process in bacteriorhodopsin: computer simulations of the quantum dynamics of the primary photochemical event. J Phys Chem B 2001, 105:9857-9871. 21. Hayashi S, Tajkhorshid E, Schulten K: Molecular dynamics simulation of bacteriorhodopsin’s photoisomerization using ab initio forces for the excited chromophore. Biophys J 2003, 85:1440-1449. 22. Hayashi S, Taikhorshid E, Schulten K: Photochemical reaction dynamics of the primary event of vision studied by means of a hybrid molecular simulation. Biophys J 2009, 96:403-416. 23. Polli D, Altoe `P, Weingart O, Spillane KM, Manzoni C, Brida D, Tomasello G, Orlandi G, Kukura P, Mathies RA et al.: Conical intersection dynamics of the primary photoisomerization event in vision. Nature 2010, 467:440. 24. Schnedermann C, Yang X, Liebel M, Spillane KM, Lugtenburg J, Ferna ´ndez I, Valentini A, Schapiro I, Olivucci M, Kukura P, Mathies RA: Evidence for a vibrational phase-dependent isotope effect on the photochemistry of vision. Nat Chem 2018, 10:449-455. 25. El-Tahawy MMT, Nenov A, Weingart O, Olivucci M, Garavelli M: Relationship between excited state lifetime and isomerization quantum yield in animal rhodopsins: beyond the onedimensional Landau–Zener model. J Chem Phys Lett 2018, 9:3315-3322. 26. Rinaldi S, Melaccio F, Gozem S, Fanelli F, Olivucci M: Comparison of the isomerization mechanisms of human melanopsin and invertebrate and vertebrate rhodopsins. Proc Natl Acad Sci U S A 2014, 111:1714-1719. 27.  Liang R, Liu F, Martı ´nez TJ: Nonadiabatic photodynamics of retinal protonated Schiff base in Channelrhodopsin 2. J Phys Chem Lett 2019 Non-adiabatic QM/MM molecular dynamics in combination with MM molecular dynamics are used to investigate the photo-activation of Channelrhodopsin 2. From the QM/MM trajectories the time-resolved fluorescence signal is computed and compared to experiment, suggesting good agreement. 28. Punwong C, Martı ´nez TJ, Hannongbua S: Direct qm/mm simulation of photoexcitation dynamics in bacteriorhodopsin and halorhodopsin. Chem Phys Lett 2014, 610–611:213-218. Non-adiabatic dynamics in photoactive proteins Groenhof, Modi and Morozov 111 www.sciencedirect.com Current Opinion in Structural Biology 2020, 61:106–112 29. Punwong C, Hannongbua S, Martı ´nez TJ: Electrostatic influence on photoisomerization in bacteriorhodopsin and halorhodopsin. J Phys Chem B 2019, 123:4850-4857. 30. Li X, Chung LW, Mizuno H, Miyawaki A, Morokuma K: Primary events of photodynamics in reversible photoswitching fluorescent protein dronpa. J Phys Chem Lett 2010, 1:3328-3333. 31. Morozov D, Groenhof G: Hydrogen bond fluctuations control photochromism in a reversibly photo-switchable fluorescent protein. Angew Chem Int Ed 2016, 55:576-578. 32. Park JW, Rhee YM: Electric field keeps chromophore planar and produces high yield fluorescence in green fluorescent protein. J Am Chem Soc 2016, 138:13619-13629. 33.  Coquelle N, Sliwa M, Woodhouse J, Schiro `G, Adam V, Aquila A, Barends TR, Boutet S, Byrdin M, Carbajo S et al.: Chromophore twisting in the excited state of a photoswitchable fluorescent protein captured by time-resolved serial femtosecond crystallography. Nat Chem 2018, 10:31 Time-resolved SFX combined with excited-state MD simulations identify a twisted chromophore intermediate in the off-to-on switching mechanism of the photochromic rs-EGFP2 protein. 34. Armengol P, Spo ¨rkel L, Gelabert R, Moreno M, Thiel W, Lluch JM: Ultrafast action chemistry in slow motion: atomistic description of the excitation and fluorescence processes in an archetypal fluorescent protein. Phys Chem Chem Phys 2018, 20:11067-11080. 35. Verhoefen M-K, Bamann C, Blo ¨cher R, Fo ¨rster U, Bamberg E, Wachtveitl J: The photocycle of channelrhodopsin-2: ultrafast reaction dynamics and subsequent reaction steps. ChemPhysChem 2010, 11:3113-3121. 36. Ferretti M, Hendrikx R, Romero E, Southall J, Cogdell RJ, Novoderezhkin VI, Scholes GD, van Grondelle R: Dark states in the light-harvesting complex 2 revealed by two-dimensional electronic spectroscopy. Sci Rep 2016, 6:20834. 37. Hybl JD, Albrecht AW, Gallagher Faeder SM, Jonas DM: Twodimensional electronic spectroscopy. Chem Phys Lett 1998, 297:307-313. 38. Zhang WM, Chernyak V, Mukamel S: Multidimensional femtosecond correlation spectroscopies of electronic and vibrational excitons. J Chem Phys 1999, 110:5011-5028. 39. Hunter J et al.: Femtosecond X-ray protein nanocrystallography. Nature 2011, 470:73-77. 40. Emma P, Akre R, Arthur J, Bionta R, Bostedt C, Bozek J, Brachmann A, Bucksbaum P, Coffee R,Decker F-J, Ding Y, Dowell D, EdstromS,FisherA,FrischJ,GilevichS,HastingsJ,HaysG,Hering P, Huang Z, Iverson R, Loos H, Messerschmidt M, Miahnahri A, Moeller S, Nuhn H-D, Pile G, Ratner D, Rzepiela J, Schultz D, Smith T, Stefan P, Tompkins H, Turner J, Welch J, White W, Wu J, Yocky G, Galayda J: First lasing and operation of an A ˚ngstrom-wavelength free-electron laser. Nat Photon 2010, 4:641-647. 41. Barends TRM, Foucar L, Ardevol A, Nass K, Aquila A, Botha S, Doak RB, Falahati K, Hartmann E, Hilpert M, Heinz M, Hoffmann MC, Ko ¨finger J, Koglin JE, Kovacsova G, Liang M, Milathianaki D, Lemke HT, Reinstein J, Roome CM, Shoeman RL, Williams GJ, Burghardt I, Hummer G, Boutet S, Schlichting I: Direct observation of ultrafast collective motions in CO myoglobin upon ligand dissociation. Science 2015, 350:445- 450. 42. Pande K, Hutchison CD, Groenhof G, Aquila A, Robinson JS, Tenboer J, Basu S, Boutet S, DePonte DP, Liang M et al.: Femtosecond structural dynamics drives the trans/cis isomerization in photoactive yellow protein. Science 2016, 352:725-729. 43.  Kovacs GN, Colletier J-P, Gru ¨nbein M, Yang Y, Stensitzki T, Batyuk A, Carbajo S, Doak R, Ehrenberg D, Foucar L et al.: Three-dimensional view of ultrafast dynamics in photoexcited bacteriorhodopsin. Nat Commun 2019, 10:3576 SFX study into the photo-isomerization of retinal in bacteriorhodopsin is combined with a power titration to systematically investigate the effect of the pump laser power. The results of the titration experiments suggest that with the pump laser power used in the SFX experiment, there is significant multi-photon absorption. Quantum chemistry calculations on a cluster model of the retinal pocket are used to identify the nature of the multi-photon absorptions. 44. Nogly P, Weinert T, James D, Carbajo S, Ozerov D, Furrer A, Gashi D, Borin V, Skopintsev P, Jaeger K, Nass K, Ba ˚th P, Bosman R, Koglin J, Seaberg M, Lane T, Kekilli D, Bru ¨nle S, Tanaka T, Wu W, Milne C, White T, Barty A, Weierstall U, Panneels V, Nango E, Iwata S, Hunter M, Schapiro I, Schertler G, Neutze R, Standfuss J: Retinal isomerization in bacteriorhodopsin captured by a femtosecond X-ray laser. Science 2018, 361:eaat0094. 45. Takala H, Bjo ¨rling A, Berntsson O, Lehtivuori H, Niebling S, Hoernke M, Kosheleva I, Henning R, Menzel A, Ihalainen JA, Westenhoff S: Signal amplification and transduction in phytochrome photosensors. Nature 2014, 509:245-248. 46.  Park JW, Shiozaki T: On-the-fly caspt2 surface-hopping dynamics. J Chem Theory Comput 2017, 13:3676-3683 One of the first applications of non-adiabatic QM/MM dynamics simulations with a multi-reference QM method (CASPT2) on a photo-biological system, in this case the GFP chromophore in water. Gradients and nonadiabatic coupling vectors are computed analytically. 47. Scheurer M, Herbst MF, Reinholdt P, Olsen JMH, Dreuw A, Kongsted J: Polarizable embedding combined with the algebraic diagrammatic. J Chem Theory Comput 2018, 14:4870- 4883. 48.  Pedraza-Gonza ´lez L, Vico LD, del Carmen Marı ´n M, Fanelli F, Olivucci M: a-arm: automatic rhodopsin modeling with chromophore cavity generation, ionization state selection, and external counterion placement. J Chem Theory Comput 2019, 15:3134-3152 Automatic workflow for setting up and performing QM/MM simulations of rhodopsins. 112 Theory and simulation Current Opinion in Structural Biology 2020, 61:106–112 www.sciencedirect.com