Ultrafast proton-coupled isomerization in the phototransformation of phytochrome

Abstract

The biological function of phytochromes is triggered by an ultrafast photoisomerization of the tetrapyrrole chromophore biliverdin between two rings denoted C and D. The mechanism by which this process induces extended structural changes of the protein is unclear. Here we report ultrafast proton-coupled photoisomerization upon excitation of the parent state (Pfr) of bacteriophytochrome Agp2. Transient deprotonation of the chromophore's pyrrole ring D or ring C into a hydrogen-bonded water cluster, revealed by a broad continuum infrared band, is triggered by electronic excitation, coherent oscillations and the sudden electric-field change in the excited state. Subsequently, a dominant fraction of the excited population relaxes back to the Pfr state, while ~35% follows the forward reaction to the photoproduct. A combination of quantum mechanics/molecular mechanics calculations and ultrafast visible and infrared spectroscopies demonstrates how proton-coupled dynamics in the excited state of Pfr leads to a restructured hydrogen-bond environment of early Lumi-F, which is interpreted as a trigger for downstream protein structural changes.

Fig. 1. Photocycle of bacterial phytochromes as derived from spectroscopic data.

In Agp2, Pfr is the stable dark state. The biliverdin chromophore (BV) is bound to a cysteine residue, with a ZZZssa and ZZEssa configuration in Pr and Pfr, respectively. In both states, the chromophore is protonated at all four pyrrole nitrogens. The curved solid and dotted arrows refer to thermal and photochemical reactions, respectively. Upon photoexcitation, isomerization at the methine bridge between rings C and D occurs on a picosecond timescale, corresponding to a ZZZssa → ZZEssa and ZZEssa → ZZZssa conversion between Pr and Pfr, respectively. The last step in the reaction cascade from Pfr to Pr is associated with a structural change of the tongue segment in the PHY domain (depicted in Fig. 2), a phytochrome-specific building block of the photosensor. States highlighted by the white and grey backgrounds represent the BV chromophore in the ZZZssa and ZZEssa states, respectively. ms, millisecond; ps, picosecond; μs, microsecond.

Fig. 2. Crystal structure of Pfr Agp2.

The whole structure (PDB 6G1Y) (upper panel) with PAS (grey), GAF (green), and PHY (purple) domains and the tongue region (blue), and a close-up view in the chromophore binding region around the BV chromophore (lower panel). The BV is covalently linked to the C13 side chain via ring A, and stabilized by various intermolecular interactions including hydrogen bonds with the side chains of Y165, Q190, D196, R211, R242, H248, Y251 and H278 as well as the water molecules W3, W4, W6 and W7. As a result, BV is embedded in a complex hydrogen-bonding network. PropB forms a salt bridge to R211; propC is hydrogen-bonded to Y165 and H278 and is connected via a hydrogen-bond water network (HBWN) with propB; water molecules W1 to W10. The electronic transition dipole moment of the chromophore (Extended Data Fig. 1) is indicated by the red arrow. D196 is hydrogen-bonded to the NH group of ring D, and its backbone C=O group is hydrogen-bonded to the NH groups of rings A, B and C, as well as to the conserved water W7. Rotation of ring D takes place upon photoexcitation. Note that the water molecules are included in the original PDB entry, but not with the same numbering.

Fig. 3. Wild-type Agp2 Pfr dynamics at pH 7.5 with excitation at 742 nm, dynamics as a function of wavenumber and pump–probe delay times.

Isotropic signals reflect bleaching (−), stimulated emission (−), excited state (+) or product absorption (+); system response < 40 fs full-width at half-maximum (FWHM). a, Absorption difference spectra at selected delay times (colour key), pump pulse spectrum (cyan dotted line), bleaching signals following the inverted absorption spectrum (pink dashed line) from 12,000 cm⁻¹ to 17,000 cm⁻¹, ES absorption (ESA) from 18,000 to 13,500 cm⁻¹ and SE signals below ~13,000 cm⁻¹. At frequencies above 17,000 cm⁻¹ and below 12,000 cm⁻¹ we see a complete decay of the ESA and SE on a timescale of 3.5 ps. Dynamics after this timescale represent ground-state processes. The centre positions of SE1, SE2 and GS* are indicated by grey bars. Absorption signal changes in mOD, i.e. 10⁻³. b, Proposed reaction scheme: excitation to S₁ (red arrow); relaxation and 1:1 bifurcation in the ES with 50 fs to protonated (grey arrows) and deprotonated (blue arrows) chromophores; 150-fs relaxation of the protonated fraction to GS* and relaxation of the deprotonated fraction in the ES; reprotonation of the chromophores and decay to ELF (70%, black arrow with isomerization) and to GS* (30%) with 1.5 ps; GS* decay with 4 ps to S₀. The overall quantum yield (QY) of ELF generation is 35%. c, Transients at selected wavenumbers display ultrafast dynamics in the electronic ES (triangles) and the dynamics of the SE (squares and circles). Simulated dynamics (lines) are shown with decay times of 50 fs, 150 fs, 1.5 ps and 4 ps. The solvent signal (grey) reflects the system response. The 4-ps dynamics visible in c (squares) reflect cooling and relaxation of the ground state GS* to the parent Pfr state. ‘Residual’ indicates the difference between the data and simulations at 12,435 cm⁻¹ (orange line), displaying coherent oscillations up to ~2 ps. Source data

Fig. 4. Vibrational absorption difference spectra of Agp2-WT at different delay times after photoexcitation in D₂O.

a, Spectral region from 1,900 to 1,480 cm⁻¹ at pD 8.2. Two datasets are separated by a grey bar: an isotropic dataset from 1,770 to 1,480 cm⁻¹ and a scaled perpendicular polarized dataset from 1,900 to 1,770 cm⁻¹. Strong sample absorption increased the noise around 1,640 cm⁻¹. CB indicates the proton-loaded water network; propC_Pfr and propC indicate the carbonyl vibration of propC in the Pfr ground state and after photoexcitation, respectively. v(C=O)_D and v(C=O)_D* correspond to the bleaching signal and ES of carbonyl ring D, respectively. v(C=C) represents the vibrations of A–B stretching and ring B stretching. The spectral range from 1,600 to 1,520 cm⁻¹ shows mainly C=C stretching dynamics of the chromophore with strong bleaching signals at 1,596 cm⁻¹ (A–B stretching) and 1,566 cm⁻¹ (ring B stretching), accompanied by redshifted positive signals due to excited-state and hot ground-state absorption. Upper inset: transient of the presented CB averaged from 1,772 to 1,832 cm⁻¹ and the smoothed transient (black line). Lower inset: QY estimation by comparing signals at the beginning (blue squares) and after photoreaction (orange squares). Simulations for v(C=O)_D at 1,684 cm⁻¹, v(C=O)_D* at 1,666 cm⁻¹ at the beginning and v(C=O)_D at 1,684 cm⁻¹ and v(C=O)_DELF at 1,692 cm⁻¹ after photoreaction. Individual bands (dotted lines) and sum of bands (black lines). The QY of 33 ± 7% is given by the ratio of the two v(C=O)_D signals. b, Isotropic dataset from 1,795 to 1,619 cm⁻¹ at pD 9. Peak positions are indicated. Two small peaks at 1,719 and 1,702 cm⁻¹ are indicated in the first 0.5 ps. In this spectral range, signals from a protonated D196 would be expected. Interestingly, the carbonyl stretching of propC shows a frequency downshift in the ELF upon chromophore excitation, although propC is not directly connected to the chromophore’s delocalized π-electron system. However, deprotonation of ring D or ring C may cause subtle alterations of the structural and electronic properties of the propionic side chain, as also suggested by the calculated frequencies (Extended Data Fig. 7). Note that such effects may account for the RR activity of this mode (Extended Data Fig. 10). Source data

Fig. 5. Water networks in Agp2-WT.

a, Absorbance difference spectra of Agp2-WT in H₂O (red) and in D₂O (blue) at 0.5 ps (averaged from 0.4 to 0.6 ps) after photoexcitation at 765 nm, compared with calculated absorption spectra (scaled) of two different proton-loaded HBWNs in a confinement in H₂O (orange) and D₂O (green) by ab initio Born–Oppenheimer MD simulations (Extended Data Fig. 6): static HBWN between propB and propC; transient HBWN between propC and W6. b, Structural snapshot from MD simulations up to 10 ns performed on Agp2-WT with an enolic ground state. A metastable transient water network (blue area) is found. Hydrogens of water molecules and hydrogen bonds are shown. Potential transient waters TW1 and TW2 were identified in the crystal structure (PDB 6G1Z) with a lower electron density compared to other water molecules published in the crystal structure. The potential transient waters presented here were not published in the PAiRFP2-Pfr crystal structure (PDB 6G1Z), because their electron density is lower compared to the conventional limit for interpretable electron densities of water molecules. The lower electron density in the X-ray structure could reflect a lower residence time and higher flexibility, that is, transient waters, in the ground-state structure, supported by the metastable geometry in MD simulations. The transient water network extends from the CO group of ring D via TW1, Y165, TW2, H278 to propC and propB. Blue and cyan arrows indicate possible proton transfer pathways from ring D to TW1 and from ring C to H248, respectively. Source data

Fig. 6. pH-dependent dynamics.

a, pH dependence of the photoreaction of Agp2-WT at different pH values in H₂O. All datasets were scaled at 635 nm and 1 ps. Transients are at 12,435 cm⁻¹ for pH values of 6.6–8.4 (graduated line colours from black to green). Inset: pH dependence averaged from transient absorption amplitudes at fixed delay times and spectral positions. The pH dependence indicates a pK_a value of 7.2 ± 0.3. b, Difference spectra at selected delay times reflecting the electronic dynamics of Agp2-H278Q in H₂O in the visible spectral range at pH 6.2 and 7.8. Inset: transients at 11,000 cm⁻¹ for pH 6.2 (orange) and pH 7.8 (green). On lowering the pH from 7.8 to 6.2, the photoreaction of Agp2-H278Q is substantially slowed to ~100 ps, similar to the results for Agp2-H278A (Extended Data Fig. 9). This is reflected by the SE around 11,000 cm⁻¹ at 150-ps delay time at pH 6.2. Moreover, the spectral features in Agp2-WT reflecting SE2 and GS* are missing, supporting that these features are indicative for the ultrafast photoreaction in Agp2-WT. With increasing pH starting at pH 6.2, the fraction of WT-like dynamics in Agp2-H278Q increases at the expense of the slow dynamics on the hundreds of picoseconds timescale (Extended Data Fig. 9). Source data

Extended Data Fig. 1. QM/MM approach for electronic transitions.

QM/MM approach for electronic transitions. Excitation energies for the S₀ → S₁ transitions were computed using RI-CC2/cc-pVDZ for different models with protonated and deprotonated pyrrole rings by transferring the proton to the bulk or to adjacent amino acid side chains (see (f)). The model with a fully protonated BV chromophore is consistent with the Pfr ground state and has an S₀ → S₁ excitation energy of 15645 cm⁻¹, 2347 cm⁻¹ higher than the experimentally measured absorption maximum. Steady state fluorescence spectra of Agp2-WT are not available due to its ultrafast excited state deactivation. Calculated transition energies differ slightly from the experimental ones, but relative transition energies differences are more precise. Thus, we compare spectral shifts. We transferred the proton from ring B, C, and D to bulk water, to H248 or D196, or changed the protonation state of H248 from ε to δ (see (f)). (a) and (b): Difference in electrostatic potential between first excited state and ground state of BV in Pfr. Upon excitation the electron density increases at ring B (blue color). Negative values (blue) indicate a decrease when going from the ground to the first excited state, while positive values (red) indicate an increase. The electrostatic potential is mapped onto the electron density at an isovalue of 0.02. (a) Difference in electrostatic potential with only BV (excluding propionates) inside the QM region. (b) Same as in (a) but with Y251. For both QM regions the potential increases near the D ring (bottom right) and decreases near the B ring (upper left). c) and d): Density differences between first excited state and ground state of BV in Pfr. Negative values (blue), positive values (red, the density difference is visualised for an isovalue of ±0.001). (c) Transition density with transition dipole moment (TDM) vector for the S₀ → S₁ transition. The TDM(x,y,z)=(4.72, 1.60, 0.43) a.u. (d) Electron density difference (EDD) between first excited state and ground state of BV in Pfr. The EDD was computed with CC2/cc-pVDZ and shows how the electron density shifts upon excitation from the ground (blue) to the first excited state (red). Changes in the EDD are mostly located on the B and C ring of BV. Contribution of Y251 to the excitation is negligible. There are also other tyrosines, for example Y165, but the electronic TDM of these tyrosines are located perpendicular to the TDM of BV, excluding an efficient coupling. (e) Table: Excitation from the ground state (S₀) to the first excited state (S₁) for S₀- and S₁-optimised structures (absorption and emission, respectively), as well as S₀ and S₁ transition dipole moments (TDM).

Extended Data Fig. 2. Polarization resolved data and analysis.

Polarization resolved data and analysis. Agp2-WT at pH 7.5, excited at 740 nm. (a) Absorption difference spectra; parallel (solid lines) and perpendicular polarized (dashed lines) with respect to the pump pulse polarization; around 13700 cm⁻¹ parallel and perpendicular signals deviate for small delay times and at 200 fs, but are nearly identical at 80 fs; at 13000 cm⁻¹ parallel and perpendicular signals show similar ratios (~2) for parallel and perpendicular polarization for all presented delay times; around 17500 cm⁻¹ parallel and perpendicular polarizations show nearly identical signals. (b) Target model from Fig. 3b used to explain the observed dynamics. Here, we introduced two apparent states S₁₁ and S₁₂ for clarification. Two yields, QY₁ and QY₂ had to be estimated. The formation of the deprotonated chromophore in the electronic excited state S₁₁ (blue line) is given by QY₁ = 0.5, and the formation of the protonated photoproduct ELF is given by QY₂. To do the estimation, we looked at the polarization resolved signals after subtraction of the bleaching contribution (d). Next, we assumed that the excited state absorption in the S₁₁ represents two electronic contributions, a protonated one (black line) and a deprotonated one (blue line); the S₁₂ electronic state consists of a single electronic transition (deprotonated); the anisotropy of S₁₁ and S₁₂ is very similar around 15000 cm⁻¹ indicating similar transition dipole moments (see c and d). The modelling was best given for QY₁ ~ 0.5 and QY₂ ~ 0.7, resulting in a total early Lumi-F yield of around 0.35. The value is in good agreement with the estimation from our IR data comparing the initial and final amplitude of the ring D v(C = O) mode (Fig. 4a lower inset). (c) Species associated difference spectra (SADS) from the target mode in (c): Left panel: isotropic SADS spectra; Right panel: polarization resolved SADS, parallel (thicker lines) and perpendicular polarization (thinner lines), with the same color code for both panels. Nomenclature used from (b). (d) Polarization resolved SADS spectra subtracted by the bleaching contribution for the target model in (b), and the polarization resolved bleaching contribution upon photoexcitation; S₁: SADS 50 fs - bleach, S₁₁: SADS 150 fs - bleach, S₁₂: SADS 1.5 ps – bleach; GS*: SADS 4 ps – bleach, and ELF: SADS const - bleach. The fraction of the bleaching contribution follows the target model and quantum yields. The spectral region with influence from scattering light (increasing with decreasing signal strength) is marked by a grey rectangle. Source data

Extended Data Fig. 3. Dynamics in the visible range.

Dynamics in the visible range. (a) Decay associated spectra of Agp2-WT at pH 7.5 excited at 740 nm (40 nm FWHM, < 40 fs time-resolution). Left panel shows the isotropic decay associated spectra. The right panels show the same components polarization resolved. Thicker lines present parallel polarization of pump and probe, thinner lines the perpendicular polarized case. (b) Transients of the same dataset; data (colored) and resulting model (black) for selected frequencies. Left panel shows isotropic data and simulations, while the right panel shows the polarization resolved data. Thick lines are for parallel polarization, thin lines for perpendicular polarization. Used decay constants are 50 fs, 150 fs, 1.5 ps, 4 ps and a constant component. For the very early delay times small deviations between modelled transients and data are visible. Introducing an additional time constant around 100 fs would have resulted in reduced deviations, but also in an ill-conditioned model. (c) Comparison of the solvent signal (red) with the sample signal (blue) at different spectral positions and different polarization directions with respect to pump pulse polarization. Source data

Extended Data Fig. 4. Coherent oscillations in Agp2-WT at pH 7.5.

Coherent oscillations in Agp2-WT at pH 7.5. (a) Left: Residuals after subtraction of the exponential model. The coherent oscillations are clearly visible. The shown oscillations show a relative phase-shift of π; middle: The resulting power-spectrum. The main features are two peaks at ~304 and ~340 cm⁻¹; right: The power of the two frequencies at 299 cm⁻¹ (red line) and 342 cm⁻¹ (black line) as a function of the probed visible frequency. (b) Top: Residuals at 14200 cm⁻¹ after applying a running mean. Middle: STFT of the residuals. The color indicates the Fourier-amplitude (not power). The STFT used a 6 points per segment with 5 overlapping points and applied a Hann-window. Clearly visible is the high average amplitude from 150 to 450 cm⁻¹ as a function of time. Bottom: Most of the amplitude decays within several hundred femtoseconds. The remaining amplitude decays on a timescale of about 1 ps. The black line presents a two-exponential fit of the amplitude, with the first point ignored. (c) Upper left: Power-spectrum at 13450 cm⁻¹ in H₂O (pH 7.5) and D₂O (pD 7.8). Lower left: Calculated Raman-spectra for deuterated and protonated pyrrole rings and propC. The calculated spectrum reproduces the data rather well and lets us assign the two main peaks to specific normal modes at 307 cm⁻¹ and 342 cm⁻¹ (right panel). Both involve ring D rocking and out-of-plane bending of the methine-bridge between ring C and D. Source data

Extended Data Fig. 5. Polarization resolved and isotropic vibrational dynamics.

Polarization resolved and isotropic vibrational dynamics. (a) Polarization resolved decay associated spectra of Agp2-WT at pD 8.2; parallel polarization (full symbols), perpendicular polarization (open symbols). The dataset was fitted with fixed decay constants of 150 fs, 1.5 ps, and 4 ps for both datasets simultaneously. The 150 fs components are not shown, because of strong mixing with the non-linear artefact during the system response. Comparing the DAS signals for 1.5 ps and 4 ps in the spectral range from 1600 cm⁻¹ to 1480 cm⁻¹ show similar or slightly red-shifted negative and positive signals. This supports the assignment of a hot ground state to the 4 ps dynamics. The origin of the CB can be further investigated by its polarization resolved signals. After photoselection by our linear polarized pump pulse we compared the absorption signals of parallel (A_par) and perpendicular (A_per) polarized probe pulses with respect to the pump pulse polarization. Taking the dichroic ratio D = A_par/A_per from 1780 to 1751 cm⁻¹ and simulating the data with a horizontal line, we found D = 1.7 ± 0.3 corresponding to a relative angle between the TDM of the CB band and the S₀ → S₁ TDM (see Fig. 2) of 30° to 43° or (34 ± 10)°. The expected vibrational TDM for a CB is almost completely polarized along the direction of maximal hydrogen bonded water network extension, supported by our ab initio Born-Oppenheimer MD simulations (see Extended Data Fig. 6). (b) Polarization resolved absorption difference spectra for different pump-probe delay times of Agp2-WT at pD 8.2 with delay times averaged at 600 fs (500 to 700 fs), 1200 fs (900 to 1300 fs), and 6100 fs (5500 to 6800 fs); an adjacent averaging filter (3 points) was used to smooth spectra. The spectra from 1660 cm⁻¹ to 1780 cm⁻¹ and 1490 to 1620 cm⁻¹ are from two different experiments. The dotted black and grey lines indicate the level of the CB for the absorption difference spectra at 600 fs for parallel and perpendicular polarization. From 1780 to 1751 cm⁻¹ we found D = 1.9 ± 0.3 corresponding to a relative angle of 26° to 38°. (c) and (d) Isotropic polarized transients at selected wavenumbers (open and solid symbols) together with the simulated data (solid lines) from the DAS presented in (a). After about 10 ps the photoreaction is completed and the first photoproduct early Lumi-F (ELF) is formed. The marker band for ELF is presented at 1695 cm⁻¹. Note, the perturbed free induction decay (PFID) is responsible for non-zero signals before time zero. Proper simulations of PFID signals are very demanding for a multitude of contributing bands in adjacent spectral regions. Source data

Extended Data Fig. 6. Ab initio Born-Oppenheimer MD simulations.

Ab initio Born-Oppenheimer MD simulations. (a) A hydrogen bonded water network (HBWN) taken from the ground-state structure of Agp2 of two water molecules in between a pair of carboxylic acid molecules. One carboxylic molecule is deprotonated. The total charge is -1e. The pink cylinder illustrates the confining cylindrical potential of strength k acting on the water oxygen nuclei. The carbon atoms are constraint at distance R_CC, as suggested by the crystal structure of Agp2. (b, c) IR spectra along different axes for different confining strength k in k_BT/Ų for H₂O (b) and D₂O (c). All systems show a much higher IR intensity along the X axis than in the YZ plane. The signals in the YZ plane are mostly independent of the confining strengths k. For both, H₂O and D₂O, clear bands reside at 500, 800 and around 1000 to 1200 cm⁻¹ that are associated with the modulated C-O vibrations. In addition, a continuum band appears between 2500 cm⁻¹ and 3500 cm⁻¹ for H₂O and between 1500 cm⁻¹ and 2500 cm⁻¹ for D₂O. The relatively stronger signal along the X axis on the other hand heavily depends on the confining strength k. Excess proton exchange between two carboxylic acid molecules occurs at the ps time scale. These model simulations reveal the effect of confinement on water-mediated proton transfer dynamics and its spectral signature, specifically the appearance of a continuum band in the 1700 cm⁻¹ - 2000 cm⁻¹ range for both H₂O and D₂O at strong confinement. (d) A transient HBWN, including one side of the chromophore with ring C and ring D, side chains H278 and Y165, that are each truncated at the ring, three water molecules TW1, TW2 and W6 (see Fig. 2) and an excess proton, the total charge amounts to +1e. The pink circles illustrate the location of the excess proton during the course of a non-equilibrium simulation trajectory, where the excess proton is initially placed at the water molecule TW2 near ring C. (e, f) IR spectra, averaged over three non-eq. simulations for H₂O (e), five non-eq. simulations for D₂O (f) and all spatial dimensions. In the reference simulations the excess proton is located near the nitrogen atom of ring D. The comparison clearly shows that the non-eq. simulations exhibit a continuum band between 1700 cm⁻¹ to 3000 cm⁻¹ for H₂O and between 1800 cm⁻¹ and 2200 cm⁻¹ for D₂O, which are not visible in the reference simulations. In summary, the ab initio MD simulations suggest the ‘static’ HBWN in (a) and the transient HBWN in (d) as plausible candidates that produce spectra in agreement with the experimentally observed transient IR spectra. Source data

Extended Data Fig. 7. Computation of IR spectra.

Computation of IR spectra. Selected QM/MM calculated (*) and experimental vibrational frequencies of vibrational modes diagnostic of the chromophore geometry for the Pfr models as well as Lumi-F model in H₂O and D₂O. Frequencies are given in cm⁻¹. (*) The crystal structure of Agp2 in the Pfr state (PDB:6G1Y) was used as template for generating the initial structure for the Pfr and Lumi-F models. Unresolved regions in the crystal were restored by three-dimensional homology modelling using SWISS MODEL and hydrogens were added to the crystallographic structure according to predictions based on Karlsberg2 + . His248 and His278 were modeled as charge neutral each with a proton at Ne position. The Lumi-F model was generated based on the Pfr model by simply rotating the pyrrole ring D around 180 degrees. The chromophore binding site of the energy minimized- and thermally equilibrated solvated protein models were geometry optimized at QM/MM level. Accordingly, the biliverdin chromophore, the side chains of Cys13 and the pyrrole water were treated quantum mechanically at the B3LYP/6-31 G* level of theory while the protein matrix, solvent water and ions were described molecular mechanically using CHARMM36 force field. Relaxation during the minimization was allowed only for atoms located within a 20 Å-radius sphere centered at N22 of the BV cofactor. The charge-shifted scheme in combination with the electrostatic embedding approach was used to couple the QM and the MM region. The QM/MM optimized geometries were further used as input for subsequent frequency calculation of exclusively the QM fragment. These computations were performed at B3LYP/6-31 G* level of theory using GAUSSIAN09 following the same protocol as described previously in ref. . Scaling of force constants, normal mode analysis as well as correction of the QM Hessian matrix were performed using the programs developed in our group.

Extended Data Fig. 8. pH dependent measurements of Agp2-WT Pfr.

pH dependent measurements of Agp2-WT Pfr. (a) Isotropic transient spectra at selected time-points of Agp2-WT excited at 750 nm at different pH values. All datasets were normalized to their signal at 635 nm at 1 ps. At lower pH, the solubility goes down and therefore increased scattering of the pump-pulse is observed around 13300 cm⁻¹. While the data at higher pH values look all rather similar, the dynamic below pH 7.0 is distinctly slower. (b) Normalized absorption spectra of the samples measured at different pH values. The spectral difference in the Pfr state between pH 10 and pH 6.6 are small and follow the shape of the Pfr absorption spectrum. Thus, the pK_a of the Pfr chromophore is expected to be higher than 9. (c) pH dependence of the ELF-Pfr difference spectra averaged from 30 ps to 150 ps after photoexcitation; isotropic polarization. The different samples were measured directly after each other under the same experimental conditions. The difference spectra show contributions of early Lumi-F (positive signals) and of the Pfr bleaching contribution (negative signals); between 740 to 780 nm scattering of the pump pulse distorts the spectra (strongest at low pH values); since the Pfr shows negligible pH dependence (b) the pH dependence has to be induced by a pH dependence of the ELF absorption spectrum. The difference of pH 8.4 and pH 6.6 exhibits a sign change from 740 to 800 nm, displaying a red-shift of the ELF absorption to higher pH values. Thus, we see a pH dependence of the ELF absorption spectrum, supporting hydrogen bonding of ELF. (d) Direct comparison of the transients at 804 nm (12435 cm⁻¹) in H₂O at pH 7.5, in D₂O at pD 7.5, in D₂O at pD 7.8, and in D₂O at pD 9.0. The transients were scaled to support the matching dynamics. This comparison shows that there is no or negligible change in dynamics upon H/D exchange. Coherent oscillations are pronounced at pH 7.5 and pD 7.8 due to a shorter system response of <40 fs. At pD 7.5 and pD 9.0 the system response was about 80 fs. Source data

Extended Data Fig. 9. Agp2 variants.

Agp2 variants: (a-d): Vis-pump IR-probe data; (a), and (b) scaled isotropic absorption difference spectra at averaged delay times around 0.7 ps (averaged from 0.4 to 1.0 ps) and 4.5 ps (averaged from 3.0 to 5.9 ps), normalized to the bleaching signal at 1689 cm⁻¹ at 0.7 ps delay time: (a) the contribution of the CB at 1790 cm⁻¹ is strongly reduced in Agp2-H278A at pD 8.2 in comparison to Agp2-H278Q at pD 8.2 and Agp2-WT. Moreover, the shoulder at 1727 cm⁻¹ indicating a down-shift of propC C = O stretching mode due to deprotonation of ring C or ring D, as well as the rising signal at 1695 cm⁻¹ reflecting ELF formation are absent in Agp2-H278A (b) The absorbance difference spectra show nearly identical dynamics, except for the variants H278A at pD 8.2, and Y165F at pD 7.8. They deviate in the spectral shape around the CB, at the photoproduct marker band around 1700 cm⁻¹, and in the C = C stretching region around 1580 cm⁻¹, and 1530 cm⁻¹, supporting a photoreaction with no or strongly reduced proton transfer upon altering the hydrogen-bonded network around Agp2-H278. ELF formation reflected by the positive marker band at 1695 cm^—1 is not observed within hundreds of ps. (c) and (d) Vis-pump IR-probe transients averaged from 1800 to 1751 cm⁻¹ of the CB taken from the dynamics presented in (a) and (b): (c) and (d) the decay of the CB signal in the first 2 ps is clearly visible except for H278A and Y165F at pD 7.8. A direct comparison of the dynamics of Y165F and H278A shows a substantially prolonged excited state lifetime. (e) Vis-pump Vis-probe transients at different pH values of Agp2-H278Q; isotropic transients of selected wavenumbers representing excited state absorption dynamics at 17820 cm⁻¹; SE and GS* formation at 12435 cm⁻¹. The measurements were performed one after another under the same experimental conditions. At low pH the contribution of sub-picosecond dynamics is very small. With pH the contribution of the fast sub-picosecond dynamics increases. The dynamics can be fitted with a sum of two models, the WT model (depicted in Fig. 3b) and an additional model with decay times on the tens to hundreds of picoseconds. (f) Vis-pump Vis-probe transients of H278A at pH 7.8; the excited state at 600 nm and the SE at 800 nm is still visible at 150 ps, reflecting dramatically prolonged excited state lifetimes. Source data

Extended Data Fig. 10. Vibrational spectra of Agp2-WT.

Vibrational spectra of Agp2-WT. RR spectra of Agp2-WT, recorded at 90 K, in the Pfr (a, d, e) and Lumi-F state (c, f, g), cryogenically trapped at 140 K, together with the static FTIR difference spectrum “Lumi-F (positive bands) minus Pfr (negative bands)”, obtained at 90 K (b). Black and red lines represent the measurements in H₂O and D₂O, respectively. Blue lines refer to spectra obtained after H/D exchange in the dark, which is restricted to the rings A, B, and C. In the IR difference spectrum (b), Lumi-F shows two positive peaks at 1718 and 1707 cm⁻¹ in the ring D C = O stretching region. In Pfr, these modes are associated with a ca. 10-cm^—1 H/D downshift such that the 1707 cm⁻¹ signal may be related to that at 1695 cm⁻¹ in the ultrafast experiments in D₂O. The major 1718 cm⁻¹ band, also identified in the RR spectrum of Lumi-F (c), shows no H/D sensitivity ruling out significant hydrogen bonding interactions and supporting its assignment to a C = O (D) conformation in Lumi-F at a later time that is not captured by the ultrafast IR experiment. This conclusion is consistent with the view that the cryogenically trapped photoproduct represents a late Lumi-F in contrast to the transient IR experiments probing the early Lumi-F. The RR spectra (d, e, f, g) mainly display modes governed by the methine bridge stretchings. The bands in this region also reflect the sequential H/D exchange that occurs instantaneously in the dark at rings A, B, and C but at ring D only after a photocycle. This effect supports the vibrational assignments and allows concluding that in late Lumi−F (f, g) the photoisomerisation-induced structural changes comprise the entire chromophore. This conclusion is in agreement with a previous crystallographic study. Source data