The interplay between chromophore and protein determines the extended excited state dynamics in a single-domain phytochrome

Significance

Bilin-binding photoreceptors are light-signaling proteins that mediate various processes from photomorphogenesis, phototaxis, chromatic acclimation, to photosynthesis. They are also promising tunable optical agents for use in optogenetics and superresolution microscopy. Using an integrated approach of crystallography, spectroscopy, and QM/MM calculations, this work examines the ultrafast dynamics of a photoactive single-domain phytochrome. Our work reveals in detail the critical role of the protein environment in defining the excited state lifetime and thereby the quantum efficiency of the bilin photoisomerization. This insight provides design principles for engineering of bilin-based photoreceptors for biotechnological and medical applications.

Abstract

Phytochromes are a diverse family of bilin-binding photoreceptors that regulate a wide range of physiological processes. Their photochemical properties make them attractive for applications in optogenetics and superresolution microscopy. Phytochromes undergo reversible photoconversion triggered by the Z ⇄ E photoisomerization about the double bond in the bilin chromophore. However, it is not fully understood at the molecular level how the protein framework facilitates the complex photoisomerization dynamics. We have studied a single-domain bilin-binding photoreceptor All2699g1 (Nostoc sp. PCC 7120) that exhibits photoconversion between the red light-absorbing (P_r) and far red-absorbing (P_fr) states just like canonical phytochromes. We present the crystal structure and examine the photoisomerization mechanism of the P_r form as well as the formation of the primary photoproduct Lumi-R using time-resolved spectroscopy and hybrid quantum mechanics/molecular mechanics simulations. We show that the unusually long excited state lifetime (broad lifetime distribution centered at ∼300 picoseconds) is due to the interactions between the isomerizing pyrrole ring D and an adjacent conserved Tyr142. The decay kinetics shows a strongly distributed character which is imposed by the nonexponential protein dynamics. Our findings offer a mechanistic insight into how the quantum efficiency of the bilin photoisomerization is tuned by the protein environment, thereby providing a structural framework for engineering bilin-based optical agents for imaging and optogenetics applications.

Phytochromes represent a large and versatile superfamily of photoreceptors identified in plants, fungi, and a wide range of bacteria (1, 2). These photoreceptors are involved in regulation of morphogenesis, photosynthesis, phototaxis, and physiological response to harmful radiation (1, 2). They have also been exploited as synthetic biology tools for optogenetics applications (3) and as fluorescence probes in biomedical imaging (4). Photoreceptors in the phytochrome superfamily bind tetrapyrrole chromophores (bilin) via a thioether linkage to a conserved cysteine residue and undergo reversible photoconversion with strong absorption shifts between a thermostable parental state and a photoproduct state. This photochromism arises from the Z ⇄ E isomerization of the C₁₅ = C₁₆ bond in the methine bridge connecting the pyrrole rings C and D of the bilin (SI Appendix, Scheme S1).

In canonical phytochromes (group I), the photosensory core module consists of the PAS-GAF-PHY array (1, 2). The PAS and GAF domains form a figure-of-eight knot, while a long extension (“tongue”) from the PHY domain interacts with the GAF domain-embedded chromophore (1, 2). In these phytochromes all three N-terminal domains (PAS-GAF-PHY) are essential for the photoconversion (5–7). Some phytochromes, such as Cph2 from Synechocystis PCC6803 (8, 9), lack the PAS domain, and therefore they are classified as knotless phytochromes (group II) (1, 2). Phytochromes in those two groups photoconvert reversibly between the red light-absorbing (P_r) state and the far red-absorbing (P_fr) state (5). Most of them adopt P_r as the dark-adapted state and P_fr as the photoproduct state. In contrast, many cyanobacteriochromes (CBCRs) are modular components of larger signaling proteins that exhibit a tandem GAF domain structure, in which one or more GAF domains covalently incorporate bilin chromophores (10). CBCRs are characterized by wide spectral diversity with absorption maxima spreading over the near-ultraviolet, visible, and near-infrared (near-IR) spectral regions (11–15). Moreover, single GAF domains from CBCRs preserve their photoconversion capability. The compact size and diverse photochemistry make photoactive GAF domains attractive as optical probes in optogenetics applications and superresolution microscopy (4). However, to tailor such modular photoreceptors for specific applications, a better understanding of their photochemistry at the molecular level is needed.

The primary photoconversion kinetics varies significantly among phytochromes and CBCRs. In canonical phytochromes (e.g., PhyA, Cph1, Agp1), the forward (P_r → P_fr) photoisomerization commences with an ultrafast (subpicosecond [sub-ps]) departure of the excited bilin from the Franck-Condon (FC) region. This is followed by conformational dynamics (∼2- to 5-ps lifetimes) of the chromophore on the excited state (ES), involving the onset of ring D rotation. The process is completed (∼30- to 50-ps lifetimes) with the formation of the primary, redshifted intermediate (Lumi-R) after overcoming a barrier on the S₁ ES surface (16–22). Recent studies on photoisomerization in the P_r state of Cph1 suggested ground state (GS) heterogeneity with two different GS populations of the bilin (23): 1) a faster (∼5-ps lifetime) nonreactive population and 2) a slower (∼30- to 50-ps lifetime), reactive population with a pretwisted ring D geometry (22, 24). In contrast, results from two-dimensional spectroscopy support a homogeneous model (25). In some bacteriophytochromes (e.g., RpBphP3, DrBph) including bathy phytochromes (Agp2, PaBphP), longer ES lifetimes (100–300 ps) were attributed to ES proton transfer (ESPT) and reorganization of the hydrogen-bonding network of the chromophore (26–30). Compared to canonical phytochromes, the early ES dynamics of the P_r state in CBCRs is generally slowed down, extending to hundreds of picoseconds (31–34). It was proposed that a GS heterogeneity is responsible for different P_r* reactive and nonreactive populations (31–33). Alternatively, the P_r* decay was described by a single very broad lifetime distribution at ∼400 ps (34).

In this work, we present the crystal structure of a photoactive single GAF domain obtained from a knotless phytochrome in the P_r state and investigate the primary forward photoconversion (P_r → Lumi-R) dynamics using ultrafast spectroscopy and hybrid QM/MM simulations. This single-domain phytochrome system offers a unique link between the CBCR and phytochrome subfamilies, allowing us to study how the photoisomerization dynamics is modulated by the protein environment. Our results provide a mechanistic framework for engineering phytochrome photoreceptors for biotechnological applications.

Results and Discussion

Crystal Structure of All2699g1 Reveals a Single-Domain Phytochrome Photoreceptor.

All2699 from Nostoc sp. PCC7120 consists of three tandem GAF domains at the N terminus followed by a histidine kinase (SI Appendix, Fig. S1A). The first bilin-binding GAF domain of All2699, denoted All2699g1, binds phycocyanobilin (PCB) and photoconverts reversibly between the P_r (λ_max: 637 nm) and P_fr (λ_max: 689 nm) states (35). We have determined the crystal structure of All2699g1 (PDB ID 6OZA) using the molecular replacement method (SI Appendix, Methods and Table S1). Although All2699g1 was first categorized as a CBCR due to its ability to undergo full P_r ⇄ P_fr photoconversion in a single bilin-binding GAF domain (35, 36), All2699g1 exhibits all of the structural elements characteristic of a phytochrome, specifically the presence of a pyrrole water and hydrogen bonding of the pyrrole nitrogen atoms to the backbone carbonyl of the conserved Asp counterion (Fig. 1). In addition, All2699g1 features two extra β-strands to the central 2–1–5–4–3 β-sheet in the GAF core, while a much shorter linker is found in CBCRs (Fig. 1A and SI Appendix, Fig. S2B). This region topologically corresponds to a large loop insertion found in canonical phytochromes where the figure-of-eight knot structure between the PAS and GAF domains is formed (SI Appendix, Fig. S2). Further evidence is provided by sequence alignment (SI Appendix, Fig. S5) and a homology model (36) that features a tongue-like extension from the second GAF domain of All2699, similar to the PHY domain in phytochromes.

Fig. 1.

Crystal structure of All2699g1. (A) Ribbon representation of the structure highlights the conserved GAF core domain shown in rainbow colors from the N (blue) to C terminus (red) except for the two extra β-strands (gray). (B) All2699g1 shows the transfacial Asp87O-pW-His139Nε coordination highly conserved among phytochromes. Red dashed lines mark the hydrogen bonds and ionic interactions between the bilin chromophore and the protein moiety. (C and D) Side-by-side comparison between All2699g1 (in gray and cyan) (C) and a representative CBCR AnPixJ (PDB ID/3W2Z) (in gray and pink) (D). Red shaded circles mark the acidic side chain of the highly conserved Asp, which points away from the chromophore in All2699g1 while forming hydrogen bonds with the pyrrole nitrogen atoms in AnPixJ.

The P_r state of the All2699g1-PCB crystal was confirmed by the crystal color and single-crystal absorption spectrum (for comparison, see Fig. 2B and SI Appendix, Fig. S1D). The simulated annealing omit map shows a ZZZssa chromophore covalently attached to Cys138 via a thioether linkage at the C₃¹ position (Fig. 1 and SI Appendix, Scheme S1) in an R configuration (SI Appendix, Fig. S3). At the β-face (i.e., below the rings B–C coplane) of the bilin, All2699g1 features the DIP motif (residues 87–89, SI Appendix, Fig. S5) which is highly conserved among phytochromes (Fig. 1 B and C and SI Appendix, Fig. S6A). Specifically, the main chain carbonyl of Asp87 forms hydrogen bonds with the pyrrole nitrogen atoms of rings A/B/C. This disposition contrasts with those found in CBCRs where the acidic side chain of Asp mediates hydrogen bonds with the pyrrole nitrogens (SI Appendix, Fig. S6B). Furthermore, All2699g1 features a highly conserved His139 at the α-face (i.e., above the rings B–C coplane), which interacts with the pyrrole rings via a pyrrole water molecule, denoted pW. Such trans-facial coordination, i.e., Asp87O-pW-His139Nε, is characteristic for all phytochromes from plant, cyanobacteria, and nonphotosynthetic bacteria (Fig. 1B). Additional similarities are found near the propionate groups, where the ring B propionate forms salt bridges with Arg133, while the ring C propionate adopts a recoiled conformation forming hydrogen bonds with His139 and a water molecule that bridges to ring D (Fig. 1 B and C and SI Appendix, Fig. S6A). In contrast, the ring B propionate recoils in CBCRs while an extended ring C propionate interacts with a conserved Arg residue (corresponding to Arg102 in AnPixJ) from the GAF-β3 strand (Fig. 1D and SI Appendix, Fig. S6B). Based on these crystallography data and structural comparisons, we categorize All2699 as a knotless phytochrome. Therefore, All2699g1 represents a phytochrome-like single-domain photoreceptor that is distinct from CBCRs.

Fig. 2.

Photoconversion of All2699g1 in solution. (A) Absorption difference spectrum for the P_r → P_fr photoconversion. (B) Steady state absorption and fluorescence spectra of the P_r and P_fr forms and computed P_r spectrum. The Pr spectrum was computed using the resolution-of-identity algebraic-diagrammatic construction through second order [(RI-ADC(2)]. The pure P_fr spectrum was obtained by subtraction of the pure P_r spectrum (35%, upper limit contribution) from the spectrum of the photostationary state at 590 nm irradiation (PSS₅₉₀) and then multiplying by a factor of 1.5 to yield the spectrum for a fully converted system. The extinction coefficient of P_r is ∼79,000 M⁻¹ cm⁻¹, as determined previously (35). The extinction coefficient of P_fr was determined to be ∼70,000 M⁻¹ cm⁻¹ from that of ^15ZP_r and using the isosbestic point at 660 nm. (C) CD spectra of the P_r and P_fr, and computed P_r CD spectrum [RI-ADC (2)]. The CD spectrum of pure P_fr was obtained from the PSS₅₉₀ CD as described in B.

Photochromism of the Red-/Far Red-Absorbing All2699g1.

In both solution and crystals, the dark-adapted P_r state of All2699g1 has an absorption maximum at 637 nm (Q band) and a fluorescence maximum at 666 nm independent of the excitation wavelength (Fig. 2B and SI Appendix, Fig. S1 C and D) (35). The photoproduct P_fr state shows an absorption maximum at 689 nm, and its fluorescence is not detectable in steady state experiments, possibly due to a very short lifetime of the P_fr* state. Under steady state irradiation (590 nm), the extent of P_r → P_fr conversion is ∼65%, limited by a significant spectral overlap between the two forms. The photoisomerization efficiencies of both the forward (P_r → P_fr) and the reverse (P_fr → P_r) reactions were determined to be ∼8% (SI Appendix, Methods).

We computed the absorption spectrum of the P_r state of All2699g1 (Fig. 2B) by performing QM/MM calculations with a molecular dynamics (MD) sampling protocol (SI Appendix, Methods) based on the crystal structure (Fig. 1). The lowest energy maxima at 654 nm (1.89 eV) and 367 nm (3.38 eV), corresponding to the Q and Soret bands, respectively, show a slight redshift (0.05 eV and 0.13 eV) compared to the experimental counterparts. The same simulation protocol also yielded good agreement between the calculated and observed absorption spectra for the P_r state of the CBCR Slr1393g3 (37).

Circular dichroism (CD) spectroscopy shows that the P_r and P_fr states of All2699g1 switch their sign of the CD signal corresponding to the Q band (35). The P_r state shows a negative, while the P_fr state shows a positive CD signal above 500 nm (Fig. 2C). For bilin-binding proteins, the signs of the Soret and Q bands in the CD spectrum are indicative of the orientation of the peripheral rings A and D with respect to the coplane of rings B and C (38). The calculated CD spectrum in the P_r state (Fig. 2C) shows a negative sign of the Q band and a positive sign of the Soret band (intensity ratio of 4:3), which is in good agreement with the experimental CD data (intensity ratio 1:1). This result is consistent with the facial disposition of rings A and D in the All2699g1 structure (Fig. 1B). This behavior is analogous to what is observed for plant and cyanobacterial phytochromes, but differs from bacteriophytochromes and many CBCRs where the Q band in the CD spectrum does not switch its sign upon conversion into the photoproduct state and remains negative (31, 38–40).

Ultrafast Dynamics of P_r* and the Formation of Lumi-R.

To examine the early dynamics in the P_r → P_fr transformation, we conducted femtosecond (fs) transient absorption (TA) experiments (Fig. 3 and SI Appendix, Fig. S12; see SI Appendix for details on the setup). The results did not show any significant kinetic isotope effect (SI Appendix, Fig. S15) suggesting that ESPT, such as the one found in some bacteriophytochromes (26), is not involved in the P_r* kinetics of All2699g1. Further, no major excitation wavelength dependence was detected which indicates homogeneity of the absorption band (SI Appendix, Figs. S12 and S13). Therefore, we focus on the 635-nm excitation dataset (Fig. 3). The TA data show a broad positive signal in the spectral range below ∼575 nm due to ES absorption (ESA). Above 575 nm, the TA data are dominated by the negative signals of the GS bleach (GSB) and the stimulated emission (SE). These spectral features are very similar to those found in the forward dynamics of red/green CBCRs (31, 34). The ESA, GSB, and SE signals appear to decay simultaneously on the 0.1- to 1-ns timescale forming a very clear isosbestic point signature at ∼575 nm (Fig. 3A and SI Appendix, Fig. S12). Their decay uncovers a new product absorption signal in the 650- to 700-nm range (Fig. 3A) that can be straightforwardly assigned to the primary photoproduct Lumi-R.

Fig. 3.

Ultrafast ES dynamics of the P_r state. (A) TA data from the forward, ^15ZP_r → ^15EP_fr, dynamics of All2699g1 after 635 nm excitation. (B) Corresponding lifetime density map (LDM) obtained from the lifetime-distribution analysis of the TA data.

The lifetime-distribution analysis (see SI Appendix and ref. 41 for explanation of the methodology) reveals further details of the early ES dynamics of P_r (Fig. 3B and SI Appendix, Fig. S13). The pair of negative- and positive-amplitude lifetime distributions at ∼100 fs in the 600- to 750-nm range account for the redshift of the SE signal, and thus we assign them to the departure of the ES wavepacket from the FC region. The next lifetime distributions (∼500 fs and 2- to 8-ps ranges) are located in the overlap region between the steep edges of the GSB and the SE signals (600–700 nm), which makes them very sensitive to any spectral shift dynamics (Fig. 3B). In addition, on this scale we do not observe any substantial P_r* decay (no lifetime distribution with positive amplitude in the ESA spectral range below 600 nm). Therefore, we attribute these features to dynamics on the ES potential energy surface (PES) possibly correlated with conformational changes in the chromophore.

The LDM is dominated by the lifetime distributions above 100 ps. The spectrally broad (<600 nm) lifetime distribution with positive amplitude that stretches from ∼50 ps to ∼1.1 ns describes the P_r* decay. Above 600 nm, a very pronounced, oblong, lifetime distribution with negative amplitude is present in the 650- to 700-nm region, which is overlaid onto a weaker but spectrally much broader distribution (600–750 nm). The two negative-amplitude lifetime distributions match the combined spectral shape of the GSB and the SE. Therefore, they can be assigned to the P_r* decay and the associated recovery of the GSB. The more pronounced, negative-amplitude lifetime distribution (650–750 nm) also describes the rise of the primary photoproduct Lumi-R, which gives this component its stretched shape (in lifetime). The negative and the positive distributions >1 ns represent the nondecaying (on this timescale) GSB and Lumi-R signals.

The near-IR TA data show a broad positive absorption difference signal that is not contaminated by GSB or SE (SI Appendix, Fig. S14A). The dynamics of this signal is described by three lifetime components (SI Appendix, Fig. S14B). The two short lifetimes (100 fs and 2.2 ps) have derivative-like decay-associated spectra (DAS) and thus, similarly to the shorter lifetime distributions from the data in the visible range (Fig. 3A and SI Appendix, Fig. S13), represent spectral shift dynamics. On the other hand, the long lifetime component (270 ps) has an all-positive DAS that matches very well the center of the visible ESA decay lifetime distribution (Fig. 3B) and can thus also be assigned to the P_r* decay. Consequently, the near-IR data reaffirm that no significant P_r* decay occurs on the sub-50-ps timescale.

Distributed Character of the P_r* Photoisomerization Kinetics.

We performed combined data analysis of the experimental datasets using a sequential model (Global Target Analysis) (41, 42). This analysis yields the so-called evolution-associated difference spectra (EADS). Five states were necessary to obtain a satisfactory fit quality (Fig. 4A). The first three EADS show identical amplitudes in the ESA spectral range, thereby confirming that no major ES decay occurs with lifetimes shorter than 50 ps (Fig. 4A). This result contrasts with studies on some canonical phytochromes (Cph1, PhyA, Agp1) and some CBCRs where shorter decay components have been associated with P_r* relaxation (16–22, 31). However, in All2699g1 and in Slr1393g3 (34), the ∼3-ps component is due to a shift dynamics as evidenced by the EADS of S2 and S3 showing identical amplitude with shifted spectral position (Fig. 4A and SI Appendix, Fig. S16 B, D, and F). Similar picosecond components were also observed in some bacteriophytochromes (e.g., RpBphP3 and Agp2) (26, 29).

Fig. 4.

Modeling of the ultrafast P_r kinetics. Analysis of the experimental data from the P_r → P_fr dynamics of All2699g1 after 635 nm excitation using a sequential kinetic scheme. The kinetic model fitting results in the so-called EADS. (A) EADS from fitting a sequential scheme with five states. (B) EADS from fitting a sequential scheme with four states, with the third state (S3) being modeled with a stretched exponent with β = 0.82.

The first detectable decay of the EADS is associated with the 50-ps lifetime component, while the major decay occurs with the ∼300-ps component (SI Appendix, Fig. S16). The EADS of these two components are spectrally identical. Similar lifetime components have been observed in some CBCRs and assigned to reactive P_r* populations in the frame of a heterogeneous kinetic model. However, the lifetime-distribution analysis (Fig. 3 and SI Appendix, Fig. S13), the identical EADS (SI Appendix, Fig. S16 B, D, and F), and the relatively small contribution of the 50-ps component (SI Appendix, Fig. S16 A, C, and E) raise the question whether the observed ES decay in All2699g1 has a distributed rather than distinctly heterogeneous kinetics. Therefore, we performed combined analysis with a four-state sequential model, where the third state was modeled with a stretched exponential decay (43, 44). This analysis yielded equally good results (Fig. 4B and SI Appendix, Figs. S17 and S18). Moreover, the use of a stretched exponential is in line with the remarkably broad lifetime distribution characterizing the P_r* decay (Fig. 3B). Most importantly, the description of the extended ES kinetics by a broad lifetime distribution and a stretched exponential bring the question about the role of the dynamics of the protein environment in the observed kinetics.

The use of stretched exponentials as a phenomenological description of distributed kinetics occuring in constrained environments and during protein dynamics is well documented (45–47). The distributed kinetics can arise from static heterogeneity or from dynamic changes in the environment. However, the long ES lifetime of P_r, which indicates the presence of a significant energetic barrier on the ES PES, rather suggests that the observed distributed relaxation kinetics is associated with dynamic reorganization of the protein. Therefore, based on this association, we propose a mechanistic picture for the steps involved in the P_r* photoisomerization. In essence, after the initial FC relaxation (∼100 fs), the excited bilin undergoes conformational changes, which are reflected in the 2- to 5-ps minor spectral shift dynamics. These conformational changes of the chromophore trigger larger-scale protein motions, involving the nearby amino acid residues. We suggest that these changes are necessary to accommodate the rotation of ring D. In effect, the excited chromophore can proceed along the isomerization reaction coordinate and reach a conical intersection with the GS. Under such conditions, the photoisomerization dynamics of the bilin is modulated by the protein dynamics, thereby exhibiting the observed distributed kinetics. This proposed mechanism not only explains the observed distributed kinetics in All2699g1 and in the CBCR Slr1393g3 (34), but also accounts for the relatively slow relaxation kinetics of the embedded chromophore as compared to isolated bilins (48–51) or bilins embedded in other phytochromes (e.g., PhyA, Cph1, Agp1) (16–20, 22). We note here that while GS heterogeneity possibly plays a role in the dynamics of CBCRs and phytochromes with sub-50-ps lifetime components [e.g., in some CBCRs (31) or Cph1 (22, 24)], this effect is diminished for systems with longer lifetime components, for which the protein dynamics takes a leading role in controlling the relaxation kinetics.

Mechanism of the Photoisomerization in P_r*.

To investigate the molecular mechanism of photoisomerization and the origin of the extended ES lifetime of the P_r state of All2699g1, we calculated a relaxed scan along the torsion of the C₁₅ = C₁₆ bond of the PCB chromophore (Fig. 5A). The relaxed scan was started from the FC point in S₁. The corresponding GS geometry is pretwisted around the C₁₅ = C₁₆ double bond (C₁₄–C₁₅ = C₁₆–C₁₇ dihedral angle of −166°), in line with the crystal structure (Fig. 1B) and the CD spectrum (Fig. 2C) of the P_r state. There are two possible directions of rotation for ring D, clockwise and counterclockwise. During the rotation in both senses the C₁₄–C₁₅ bond is shortened and the C₁₅ = C₁₆ bond is elongated (Fig. 5B). Such changes in the bond lengths, which are well known from retinal proteins (52), are due to a change of the electronic structure in the ES. However, in contrast to the inversion of single and double bond lengths observed in protonated retinal Schiff base, we find that the C₁₅ = C₁₆ double bond of the PCB chromophore is shorter than the C₁₄–C₁₅ single bond during the rotation in both the clockwise and the counterclockwise direction. Inversion of the bond length alternation was achieved after ∼30° of counterclockwise rotation (Fig. 5B). The clockwise rotation requires high distortion (∼70° from the FC point or ∼55° from planarity) to reach bond lengths nearly equal to those in the case of the counterclockwise rotation. The counterclockwise rotation appears to be energetically more favorable with the ES energy exceeding that of the FC starting only at −135° of the torsion angle. In comparison, the clockwise rotation reaches this point already at 155°, which is much closer to a planar structure. However, in both directions of rotation there is an ES energy barrier that prevents the completion of the isomerization. To pass the barrier, a large rearrangement of the protein is required, which cannot be addressed by a relaxed scan. Nevertheless, we deduce that the counterclockwise rotation is favored due to the pretwist of ring D in the GS and a smaller barrier on the ES. The progress in rotation is associated with a decrease in the S₀-S₁ energy gap, but a crossing between the GS and the ES was not encountered. This finding is in line with the 2- to 8-ps lifetime-distribution component found in our ultrafast data and assigned to conformational changes of the bilin (see above).

Fig. 5.

ES relaxed scan. (A) Relaxed scan along the torsion angle C₁₄–C₁₅ = C₁₆–C₁₇ of the PCB chromophore in the P_r form with and without the protein environment. (B) Change in the bond lengths of the methine bridge between rings C and D of the bilin. (C) Change in the dihedral angle of C₁₃–C₁₄–C₁₅ = C₁₆ as a function of the change in the dihedral angle C₁₄–C₁₅ = C₁₆–C₁₇. The red arrow indicates the S₀ → S₁ excitation. The vertical dashed gray line indicates the last point before the hydrogen-bonding network with ring D disrupts.

To estimate the role of the protein environment in the isomerization process we recalculated the optimized points of the scan without the protein (empty circles in Fig. 5A). These are single point calculations where the protein is removed but the chromophore remains in the conformation induced by the protein. The resulting energy profiles show that the ES barrier does not disappear after removal of the protein environment, which is in agreement with previous computational studies in the gas phase (51, 53). In the absence of the protein the isomerization path shows a slightly lower barrier in the counterclockwise sense with a nearly constant offset. In the clockwise direction, the barrier decreases with the progress in rotation. Overall, a larger lowering in the S₁ energy is observed for the clockwise rotation compared to the counterclockwise. Hence, there is a difference in the interaction of ring D with the protein for the two senses of rotation.

At the FC point, the C₁₅ = C₁₆ double bond has a dihedral angle of −166°. This deviation from planarity is due to the involvement of ring D in a hydrogen-bonding network with surrounding amino acid residues and water molecules. The pyrrole nitrogen hydrogen-bonds to a water molecule (1.80 Å) forming a bridge to the Tyr142 side chain and the propionate at ring C, and the pyrrole carbonyl group hydrogen-bonds with the ε-NH group of His169 (2.01 Å) (Fig. 6). Clockwise rotation of ring D would disrupt this hydrogen-bonding network. To prevent rupture of this network, the methine single bond C₁₄–C₁₅ compensates the motion by rotating counterclockwise out of plane with respect to the B-C coplane (Fig. 5C and Movie S1). In addition, Tyr142, His169, and the bridging water molecule follow the displacement of ring D to maintain the hydrogen bonding. Nevertheless, further clockwise rotation of the double bond beyond 135° results in the breaking of the hydrogen-bonding network. As a consequence, the single bond planarizes (Fig. 6A and Movie S1) and changes by ∼50° in the course of the relaxation.

Fig. 6.

Molecular rearrangements along the relaxed scan. Counterclockwise (A) and clockwise (B) rotation of ring D with the amino acid residues in the vicinity. The initial structure is shown in light blue, while the final point in each sense of direction is shown in gray (solid).

In contrast, rotation in the counterclockwise direction retains the hydrogen-bonding network until the last obtained point of the relaxed scan (−105°). The rotation of ring D is accommodated by a minor displacement of the Tyr142 residue (the distance between the pyrrole moiety of ring D and ε-CH1 of Tyr142 decreases from 2.66 Å at the FC point to 2.21 Å at a torsion angle of −105°), while the His169 remains in its position (Movie S2). The torsion of the double bond occurs along with an out-of-plane movement of the methine bridge, where the bridge moves up relative to the B-C coplane. Further rotation beyond −105° is hindered by Tyr142 which would have to move away from the binding pocket (Fig. 6B and Movie S2). However, convergence of the relaxed scan beyond −105° was not obtained, suggesting that large motions of Tyr142 and other surrounding residues are required. This necessity of larger changes in the protein supports the model derived on the basis of the analysis of the time-resolved data and explains the long ES lifetime (∼300 ps, broad lifetime distribution).

Conclusion

By combining X-ray crystallography, ultrafast spectroscopy, and QM/MM calculations, this work provides a mechanistic insight into the photoisomerization of the bilin chromophore in the single-domain phytochrome All2699g1. A coherent picture emerges, revealing that the early picosecond dynamics of P_r* is associated with conformational changes in the chromophore, while the long ES lifetime is attributed to a large barrier for the ring D rotation in a confined environment. Relaxed scans along the rotational trajectory about the C₁₅ = C₁₆ bond further show asymmetric barriers for the clockwise and counterclockwise directions of rotation arising from specific protein–chromophore interactions. The heights of these energetic barriers imply that a significant reorganization of the chromophore environment is required for the isomerization to proceed. This protein reorganization and the associated reduction in the energetic barrier cannot occur spontaneously but are rather initiated by the early conformational changes in the excited chromophore as indicated by the 2- to 8-ps lifetime. Once activated, the protein moiety reorganizes to accommodate the rotation of ring D on the longer timescale as evidenced by the extended P_r* lifetime (broad distribution around 300 ps). In other words, the photoisomerization dynamics is governed by an intricate interplay between the excited chromophore and the protein, in which the chromophore acts as a trigger, while the protein moiety controls the reaction dynamics. In effect, the protein dictates the lifetime and the ensuing quantum yield of photoisomerization. This mechanistic insight provides a framework for design and engineering of novel photoreceptors for optogenetics and fluorescence probes for superresolution microscopy.

Materials and Methods

Detailed description of the materials and methods is provided in the SI Appendix. This includes information on the sample preparation, crystallization, and structure determination; the spectroscopic and data analysis methods; as well as details on the QM/MM computations.

Data Availability.

The coordinates and structure factor amplitudes of the All2699g1-PEB and All2699g1-PCB structures have been deposited to the Protein Data Bank (https://www.rcsb.org) under the accession numbers 6OZA and 6OZB, respectively.

Note Added in Proof.

After the acceptance of this work, a manuscript on the same protein was published (54).