Modulation of Light Energy Transfer from Chromophore to Protein in the Channelrhodopsin ReaChR

Abstract

The function of photoreceptors relies on efficient transfer of absorbed light energy from the chromophore to the protein to drive conformational changes that ultimately generate an output signal. In retinal-binding proteins, mainly two mechanisms exist to store the photon energy after photoisomerization: 1) conformational distortion of the prosthetic group retinal, and 2) charge separation between the protonated retinal Schiff base (RSBH⁺) and its counterion complex. Accordingly, energy transfer to the protein is achieved by chromophore relaxation and/or reduction of the charge separation in the RSBH⁺-counterion complex. Combining FTIR and UV-Vis spectroscopy along with molecular dynamics simulations, we show here for the widely used, red-activatable Volvox carteri channelrhodopsin-1 derivate ReaChR that energy storage and transfer into the protein depends on the protonation state of glutamic acid E163 (Ci1), one of the counterions of the RSBH⁺. Ci1 retains a pK_a of 7.6 so that both its protonated and deprotonated forms equilibrate at physiological conditions. Protonation of Ci1 leads to a rigid hydrogen-bonding network in the active-site region. This stabilizes the distorted conformation of the retinal after photoactivation and decelerates energy transfer into the protein by impairing the release of the strain energy. In contrast, with deprotonated Ci1 or removal of the Ci1 glutamate side chain, the hydrogen-bonded system is less rigid, and energy transfer by chromophore relaxation is accelerated. Based on the hydrogen out-of-plane (HOOP) band decay kinetics, we determined the activation energy for these processes in dependence of the Ci1 protonation state.

Significance

The function of channelrhodopsins relies on the absorption of light energy by the retinal chromophore. The energy is stored by either chromophore distortion or charge separation before it is transferred to the protein to actuate conformational changes leading to an output signal that triggers physiological response. Here, we show for the red-shifted channelrhodopsin ReaChR how subtle structural changes in the binding pocket, such as the protonation state of a specific amino acid residue, influence the biophysical properties of the chromophore-binding pocket in terms of hydrogen bonding and electrostatics and modulate the mechanism and efficiency of energy transfer to the protein. This functional pH sensitivity allows the protein to adopt is activity to different environmental pH conditions.

Introduction

Retinal-binding proteins have evolved mechanisms to store the energy of the absorbed photon in the chromophore and subsequently guide it to the protein to drive conformational rearrangements that eventually create an output signal, such as active and passive ion transport or enzyme activation. In the microbial bacteriorhodopsin (BR), ∼30% of the photon energy is stored in the early photoproduct, K (1), and in bovine rhodopsin, even >60% in the early photoproduct bathorhodopsin (2). Depending on the environment of the retinal-binding pocket, storage of the absorbed light energy is largely achieved by two mechanisms: 1) transient distortion of the prosthetic group retinal, and/or 2) charge separation in the active site between the protonated retinal Schiff base (RSBH⁺) by which the prosthetic group is covalently linked to the protein and glutamic and/or aspartic acids that act as its counterion(s) (3, 4, 5, 6). Energy transfer to the protein is accordingly accomplished by chromophore relaxation and/or reduction of charge separation in the RSBH⁺-counterion complex. In BR that possesses two counterions (D85 and D212) bound to the RSBH⁺ via a water molecule (7), energy storage is mainly achieved by charge separation of the RSBH⁺-counterion pair (∼70% of totally stored energy) (4,5), whereas in bovine rhodopsin with only one counterion (E113), the contribution of charge separation to energy storage is comparably smaller (20–50% of totally stored energy), and chromophore distortion becomes more relevant (3,6,8).

Channelrhodopsins (ChRs), as another class of retinal-binding proteins, are directly light-gated ion channels originally found in the eyespot of motile green algae (9,10) and are widely used as optogenetic tools (11). Most cation-conducting ChRs have two counterions (Cis) in form of a glutamate (Ci1) and an aspartate (Ci2) in vicinity (<4 Å distance) of the RSBH⁺, directly bound to the RSBH⁺ without a water bridge. Instead, the more recently discovered anion-conducting ChRs exhibit only one counterion in their active site (12).

Interestingly, the pK_a of Ci1 significantly differs among the cation-conducting ChRs. Low pK_a-values leading to deprotonated Ci1 at a physiological pH are found in the blue-absorbing variants Platymonas subcordiformis ChR2, Chlamydomonas reinhardtii ChR2 (CrChR2), and chimera of CrChR1 and CrChR2 (C1C2) (13, 14, 15). The blue shift of their spectrum is largely caused by the negative charge of deprotonated Ci1 that stabilizes the positive charge of the chromophore in the RSBH⁺ region. This leads to a relative lowering of the ground state energy with respect to the excited state according to the point charge model (16). Neutralization of Ci1 by mutations results in bathochromic shifts in the action or absorbance spectra (30 nm in P. subcordiformis ChR2-E106Q (15), 20 nm in CrChR2-E123T (17), and 11 nm in C1C2-E162Q (18)) because these mutations increase the displacement of the π-electrons of the chromophore in direction of the RSBH⁺ proton so that the energy difference between the electronic ground state and excited state is lowered (19). Contrarily, a high pK_a of Ci1, and consequently its protonation at physiological pH, contributes to the bathochromic shift of green- and orange-absorbing ChRs, such as C. augustae ChR1 and Chrimson. Mutations of Ci1 in these variants cause moderately blue-shifted absorption maxima (6 nm in C. augustae ChR1-E169Q (20) and 5 nm in Chrimson-E165A (21)), providing additional evidence for its protonated form.

This work is focused on the red-activatable ChR ReaChR (22), a variant of Volvox carteri ChR1 that has been successfully applied in neurologic, cardiologic, and behavioral studies (23, 24, 25, 26) and on the mutants E163T, D293N, and E130Q. The selection of these mutants is based on the availability of the electrophysiological data that allows us to relate spectroscopic results with functionality (27).

We show here, by a combination of Fourier-transform infrared (FTIR) and ultraviolet-visible (UV-Vis) spectroscopy, pH-titration experiments, and site-directed mutagenesis, that the pK_a of Ci1 is 7.6, so both protonated and deprotonated Ci1 exist at native pH conditions. The Ci1 protonation state in ReaChR modulates the mechanism of energy storage and affects the efficiency of the energy transfer from the chromophore to the protein. Our study is corroborated by molecular dynamics (MD) simulations of ReaChR in the dark state that reveal only one predominant active-site conformation for protonated Ci1 (Fig. 5 b), whereas for deprotonated Ci1, three different hydrogen-bonding networks are formed, indicating a higher structural heterogeneity of the dark state at high pH (Fig. 5, c–e). We suggest that the protonation state of Ci1 leads to different conformations of the active site in the dark state and, subsequently, to different hydrogen-bonded networks in the K state. Our spectroscopic data imply that protonated Ci1 is involved in a hydrogen-bonded system that efficiently stabilizes the distorted conformation of the chromophore. Deprotonated Ci1, on the other hand, leads to a more flexible active site that ultimately facilitates the energy flow to the protein.

Figure 5.

Hydrogen-bonding network in the active-site region. (a) ReaChR homology model with RSBH⁺, Ci1, and Ci2 shown in licorice representation. (b) For protonated Ci1, the RSBH⁺ interacts predominantly with Ci2, whereas Ci1 and Ci2 form a hydrogen bond with each other. For deprotonated Ci1, three distinct hydrogen-bonding networks can be observed in which the RSBH⁺ interacts either with Ci1 (c), Ci2 (d), or a water molecule (e). Distances are given in angstroms.

Materials and Methods

Experimental procedures

Molecular biology, expression, and purification

The mutations were created via site-directed mutagenesis (QuikChange; Agilent Technologies, Santa Clara, CA). The recombinant proteins for pH titrations and FTIR measurements of the ReaChR wild-type at pH 7.4, E163T and D293N, were expressed in HEK293T cells as previously described (28). The recombinant protein for pH titration of C1C2 and FTIR measurements of ReaChR wild-type at pH 5 and pH 9 was expressed in Pichia pastoris, as described elsewhere (29,30). The expression system and the purification conditions had negligible effects on the FTIR difference spectra (Fig. S1).

pH titrations

Recombinant proteins were equilibrated in titration buffer (10 mM citrate and 10 mM Bis-tris propane, 10 mM CAPS (pH 3.3–5.2), and 100 mM NaCl (Carl Roth, Karlsruhe, Germany) and 0.03% (w/v) n-dodecyl β-D-maltopyranoside (Glycon Biochemicals, Luckenwalde, Germany)) at 20°C, placed inside a half-microcuvette (6Q; Starna Scientific, Pfungstadt, Germany) and titrated with small volumes of 1 M NaOH (Carl Roth) under gentle agitation (Variomag Electronicrührer Micro; H+P Labortechnik, Munich, Germany). pH was monitored by a pH meter (pH-Meter 766; Knick Elektronische Messgeräte, Berlin, Germany) equipped with a microelectrode (Z451; SI Analytics, Weilheim, Germany). Before each spectral recording (Cary 300 Bio; Agilent Technologies), samples were stirred for 1–2 min to ensure complete equilibration. Dilution of the protein solutions by the added NaOH was considered. pH was plotted against λ_max, and data were fitted by the Henderson-Hasselbalch equation.

Buffer exchange and sample deuteration

Repeated buffer exchange of stock solutions (Dulbecco’s phosphate-buffered saline (pH 7.4) and 0.03% (w/v) n-dodecyl β-D-maltopyranoside) was performed using Centricons (GE Healthcare, Chalfont St. Giles, UK) to obtain wild-type samples at pH 5 (20 mM citrate and 100 mM NaCl), pH 9 (20 mM TRIS and 100 mM NaCl), pD 7.8 (Dulbecco’s phosphate-buffered saline in D₂O), or pD 9.4 (20 mM TRIS, and100 mM NaCl in D₂O). To achieve better H-D exchange rates within the channel, the samples were additionally illuminated several times during incubation in D₂O buffer.

FTIR measurements

FTIR samples were prepared on a BaF₂ window by repeated drying under a nitrogen stream and subsequent rehydration. After preparation, the sample was sealed with a second BaF₂ window. Samples were illuminated with LEDs (maximal emission wavelength ∼530 nm). Samples were equilibrated in the cryostat DN (Oxford Instruments, Abingdon, UK) at the respective temperature for at least 45 min. After measurement, the sample was heated up again to a minimum of 20°C to allow relaxation. FTIR measurements were performed using an ifs66v/s FTIR spectrometer (Bruker Optics, Karlsruhe, Germany). A 1850 cm⁻¹ optical cutoff filter was used. Spectra were recorded with 200 kHz scanner velocity and a spectral resolution of 2 cm⁻¹. The difference spectra were corrected for baseline drifts using a spline algorithm and the baseline-correction mode implemented in the OPUS 6.5 software package (Bruker Optics, Karlsruhe, Germany). The FTIR spectra used for the evaluation of the hydrogen out-of-plane (HOOP) band decay kinetics shown in Fig. 4 b were subjected to singular value decomposition and a rotational procedure as described in (31) to reduce spectral noise.

Figure 4.

Energy transfer from chromophore to protein in dependence of the Ci1 protonation state. (a) Determination of the mean activation energies for energy transfer based on the evaluation method introduced by Austin et al. (32). The data points were obtained from fits of HOOP band decay kinetics (see Materials and Methods for details) in the range from 130 to 180 K (see Fig. S8) and fitted to a linear function (red and blue line). The intercept on the ordinate multiplied by −1 provides an estimate of the mean activation energy of the energy transfer to the protein, yielding values of 26.2 ± 1.5 kJ/mol at pH 9 and 34.8 ± 5.1 kJ/mol at pH 5. (b) HOOP band decay kinetics of the wild-type at pH 5 (red), pH 7.4 (black), and pH 9 (blue); the Ci1 mutant E163T (violet); and the Ci2 mutant D293N (green) at 150 K. Data points were fitted using the empirical equation introduced by Austin et al. (32). The resulting t_0.8-values that represent the time required to reach 80% of the initial signal intensity are summarized. Fitting errors are indicated. For details on band integration, see Materials and Methods.

Evaluation of kinetic data

To characterize the pH dependence of chromophore relaxation, the K intermediate was induced by continuous illumination (15 min) at selected cryotemperatures (from 130 to 180 K), and the kinetics of HOOP band decay were tracked after light off. To determine the kinetics, HOOP bands were integrated between the following wavenumber limits: wt (wild-type; pH 5) 960–990 cm⁻¹, wt (pH 7.4) 960–989 cm⁻¹, wt (pH 9) 968–990 cm⁻¹, E163T (pH 7.4) 971–986 cm⁻¹, and D293N (pH 7.4) 960–990 cm⁻¹, depending on the respective band shapes.

Data evaluation was based on the following presumptions:

• 1)The K intermediate is selectively accumulated in the given temperature range. This is supported by corresponding UV-Vis measurements performed at physiological pH and flash photolysis UV-Vis measurements performed at pH 5 and 9 (28).
• 2)HOOP band decay after light off is due to chromophore relaxation in the ongoing photocycle that could be described as P(d(istorted))→P(r(elaxed)) and not due to a shunt reaction back to the dark state P(d)→D(r). This assumption is based on the observation that whereas the HOOP bands decay after lights off, other retinal fingerprint bands indicating depletion of the dark state remain unchanged (Fig. S2).
• 3)P(d)→P(r) is an irreversible reaction.

In contrast to these presumptions, the decay kinetics cannot be sufficiently described by

$$\text{N}\left(\text{t}\right)=\text{A×exp}\left(-\text{kt}\right)$$ (1)

Instead, the empirical equation

$$N\left(t\right)={\left(1+t/t_0\right)}^{-n},$$ (2)

with parameters t₀ and n introduced by Austin et al. (32), yielded satisfactory results. This equation takes into account that at cryostatic conditions (typically <180 K), the interconversion of conformational substates is impaired so that relaxation from K(d)₁, K(d)₂, … K(d)_n requires distinct activation energies E_A(dr)₁, E_A(dr)₂, … E_A(dr)_n. The mean activation energy of the distributed energy spectrum is given by the intercept on the ordinate of a plot of$RT\times ln\left(\frac{n}{t_0}\right)$.

Computational methods

Structural model

The lack of a ReaChR crystal structure necessitates a homology model. This model was generated using the following three ChRs as templates: C1C2 (Protein Data Bank, PDB: 3UG9 (14)), CrChR2 (PDB: 6EID (33)), and Chrimson (PDB: 5ZIH (34)). After aligning the sequences of the proteins with Clustal Omega (35), the initial model was constructed with MODELER (version 9.20) (36). The final ReaChR dimer model was built by aligning two monomers of the homology model to a C1C2 dimer that had been assembled with PISA (37). Standard protonation states were used for all amino acids, with the exception of E130 and D196, which were chosen to be neutral according to previous findings (27), and Ci1, for which the negative and neutral protonation state were simulated independently. The protonation state of Ci1 was always identical in both monomers of the same simulation. The two monomers of the ReaChR dimer model were connected via three disulfide bridges between C67 of monomers 1 and 2, C74 of monomer 1 and C76 monomer 2, and C76 of monomer 1 and C74 monomer 2, respectively.

MD simulations

CHARMM-GUI (38, 39, 40, 41) was utilized to generate a hydrated 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid membrane, in which the dimer model was embedded. The final simulation setup contained a total of ∼150,000 atoms (∼33,000 water molecules and 299 lipids) in a simulation box of size 110 × 110 × 120 Å³. Two or four charge-neutralizing chloride ions were added to the simulations with deprotonated and protonated E163, respectively.

The AMBER16 software package (42) was used to run the MD simulations with the CHARMM36 protein (43,44) and lipid (45) force field, the TIP3P water model (46), and the ion parameters of Roux and co-workers (47). Covalent bonds with hydrogen atoms were constrained using the SHAKE algorithm (48). During the first 50 ps of the equilibration phase, the system was heated from 100 to 300 K, with restraints placed on the heavy atoms. These restraints were gradually released throughout the 30 ns of equilibration. The first 750 ps of equilibration used a canonical ensemble (NVT) before switching to an isothermal-isobaric ensemble (NPT) at 1 bar. After the equilibration phase, an unconstrained production run of 300 ns was started, and the integration time step was increased from 1 to 2 fs. Five independent simulations were performed for the ReaChR model with deprotonated and protonated E163, respectively. For each of the 10 simulations, coordinate snapshots were saved every 10 ps. The last 100 ns of each simulation were used for the data analysis to guarantee converged results. The data analysis was performed with VMD (49). For the hydrogen-bonding analysis, a hydrogen-bonding criterion of 2.5 Å was used to inspect each monomer individually, similar to (50). The final average percentages were obtained by counting the snapshots with a particular hydrogen-bonding state and then dividing by the total number of hydrogen bonds measured. The heavy atom distances were obtained by measuring the distance between side-chain oxygens (for Ci1, Ci2, and E130) and/or RSBH⁺ nitrogen for each snapshot and each monomer. The distances were then divided according to the hydrogen-bonding partner of the RSBH⁺. Averages were then computed for each of these groups (see Tables S1–S6). The time courses of the heavy atom distances between the RSBH⁺ and Ci1, Ci2, and water in dependence of the Ci1 protonation state are shown in Figs. S3–S5.

Results

Influence of selected ReaChR mutants on the absorption maxima and pH dependence

To investigate the influence of the counterions Ci1 (E163) and Ci2 (D293) and of the central gate residue E130 (Fig. 1 a) on the pH dependence of the absorption spectrum in ReaChR, we performed pH titrations of ReaChR wild-type and of the E163T, D293N, and E130Q mutants in the dark (Fig. S6). The selection of these mutations is based on their electrophysiological characterization, which allows us to correlate the spectroscopic data with a functional parameter and is due to their excellent expression levels (27). For comparison, similar experiments were conducted with the blue-absorbing C1C2 wild-type in which the pH dependence is less pronounced. Fitting the titration curves yielded two pK_a-values of 5.37 ± 0.22 (pK_a(1)) and 8.07 ± 0.02 (pK_a(2)) for C1C2 and one pK_a-value for ReaChR (7.56 ± 0.03; Fig. 1 b). We assign pK_a(1) of C1C2 to Ci1 because it fits best with the value obtained by semiempirical pK_a calculations (pK_a = 5.83 (14)). The origin of pK_a(2) is not completely clear; however, an assignment to Ci2 is ruled out because a semiempirical pK_a prediction (51) and computational studies of C1C2 (52) and CrChR2 (53) suggest that Ci2 retains a low pK_a (3.21 in (14)) and is therefore deprotonated in the pH range that we have investigated.

Figure 1.

Active site of ReaChR (a), comprising the protonated retinal Schiff base (RSBH⁺) that links the retinal to the protein and its counterions Ci1 (E163) and Ci2 (D293). Additionally, the central gate residue E130 is shown. The pH titration of C1C2 and ReaChR wild-type (b) and mutants of Ci1 (E163T), Ci2 (D293N), and E130 (c). The Henderson-Hasselbalch fit of the pH-titration curve of ReaChR wild-type reveals a pK_a-value of 7.56 ± 0.03. The underlying UV-Vis spectra are shown in Fig. S6. Measuring errors are indicated.

In ReaChR, the wild-type pK_a of 7.56 in detergent is increased by the mutation of both Ci1 (E163T, 8.77 ± 0.09) and Ci2 (D293N, 9.79 ± 0.13; Fig. 1 c) and lowered when the central gate residue E130 is mutated (E130Q, 5.70 ± 0.17). Despite the fact that there is an influence of all three mutants on the titration curves, we assign the pK_a of 7.56 to Ci1, as explained in detail in the Discussion.

pH dependence of the early photocycle transitions

To elucidate the influence of the Ci1 protonation state on the ReaChR photocycle, and the energy transfer in particular, we performed FTIR measurements at pH 5 and pH 9 to obtain species with pure Ci1-COOH and Ci1-COO⁻ configurations, respectively. FTIR spectra recorded at 150 K (Fig. 2) reflect the D→K transition (photocycle scheme in Fig. S7) (28). The blue-shifted UV-Vis absorption of the dark state at pH 9 (λ_max = 503 nm), as compared with pH 5 (λ_max = 535 nm (28)), is reflected by a pH-dependent upshift of the negative ν(C=C) vibrational band from 1527 (pH 5) to 1543 cm⁻¹ (pH 9) in line with the correlation of ν(C=C) modes of the chromophore with the UV-Vis absorption maximum (54,55).

Figure 2.

FTIR steady-state spectra of ReaChR wild-type at pH 5 (Ci1-COOH, red) and pH 9 (Ci1-COO⁻, blue) at 150 K, representing the D→K transition (28). The υ(C=C) frequency upshift from 1527(−) (pH 5) to 1543(−) cm⁻¹ (pH 9) correlates with its blue-shifted UV-Vis absorption (λ_max = 535 (pH 5) and 503 nm (pH 9) (28)).

Fingerprint bands at 1232(−), 1200(−), and 1186(+) cm⁻¹ in the spectrum of the acidic sample reflect isomerization from all-trans, 15-anti (D) to 13-cis, 15-anti (K) retinal (56). The vibrational mode at 1186(+) cm⁻¹ is assigned to the υ(C₁₄-C₁₅) and υ(C₁₀-C₁₁) vibrations of 13-cis retinal (18,57, 58, 59). At pH 9, the intensities of the vibrational bands at 1201(−) and 1185(+) cm⁻¹ are reduced, and an additional negative band is observed at 1192(−) cm⁻¹. Reduction of the band intensities can be explained by the cancellation of the negative band observed at 1200(−) cm⁻¹ at pH 5 by the positive band of the 13-cis photoproduct that is probably upshifted at pH 9 with respect to the spectrum at pH 5 (in which it arises at 1186(+) cm⁻¹).

The HOOP vibrations <1000 cm⁻¹ reflect chromophore distortions and indicate that the retinal in the K intermediate is in a strained conformation after photoisomerization (60, 61, 62). At pH 9, the intensity of the HOOP band around 980(+) cm⁻¹ is reduced as compared with pH 5, indicating a less distorted chromophore. This observation, along with the alterations in the retinal fingerprint region, indicates a substantial impact of the Ci1 protonation state on the relaxation process of the chromophore after conical intersection on the early S₀ energy landscape and on the subsequent photoreaction of ReaChR, which is further investigated in the following sections.

Influence of counterion complex mutations on the early photocycle transition

On the basis of the difference spectra recorded at pH 5 and pH 9 (see Fig. 2), we investigated the influence of the residues Ci1 and Ci2 on the formation of the early photocycle intermediates of ReaChR. Therefore, we recorded FTIR steady-state difference spectra of the wild-type in H₂O and D₂O and of the Ci1 mutant E163T and the Ci2 mutant D293N at pH 7.4 and 150 K for comparison (Fig. 3). While we observe no significant differences between the wild-type and D293N, Ci1 mutation leads to an upshift of the retinal band from 1232(−) to 1236(−) cm⁻¹ and to a reduction of the band pattern at 1201(−) and 1186(+) cm⁻¹, similar to the spectrum of the Ci1-COO⁻ configuration at pH 9 (see Fig. 2).

Figure 3.

Retinal fingerprint region in FTIR steady-state spectra of ReaChR wild-type at pH 7.4 in H₂O (gray) and D₂O (black), the Ci1 mutant E163T (violet), and the Ci2 mutant D293N (green) at 150 K. The spectra represent the D→K transition (28).

The HOOP band at pH 7.4 that reflects distortions of the chromophore after photoisomerization (60,61) has a maximum at 968(+) cm⁻¹ and a shoulder at 974(+) cm⁻¹. In D₂O, the band at 968 cm⁻¹ is downshifted to 953 cm⁻¹, and a D₂O-insensitive band at 963 cm⁻¹ becomes visible. The band at 968(+) cm⁻¹ is assigned to the C₁₅-HOOP mode because it is sensitive to the H-D exchange because of coupling to N-H or N-D vibrations of the RSBH(D)⁺ (63), as observed for other microbial rhodopsins (18,64). In the Ci1 mutant E163T, this band is upshifted by 12 to 980(+) cm⁻¹ and is significantly reduced, similar to the Ci1-COO⁻ spectrum recorded at pH 9 (see Fig. 2). The spectral similarities between the wild-type obtained at pH 9 and the E163T mutant obtained at pH 7.4 (reduced band intensities at 1201(−) and 1186/85(+) cm⁻¹ and in the HOOP band region <1000 cm⁻¹) show that in the early photoproducts stabilized at 150 K, both deprotonation and neutralization of Ci1 induce similar structural rearrangements of the active site.

At both conditions, the HOOP band intensities, and thus chromophore distortion, are reduced as compared with the wild-type with protonated Ci1.

Because the chromophore distortion represents a mechanism of energy storage of the absorbed light energy (65, 66, 67) and its relaxation reflects energy transfer to the protein in the ongoing photoreaction, we analyzed this relaxation process in more detail.

Chromophore relaxation dynamics

To get deeper insight into the dynamics of the energy transfer from the chromophore to the protein in dependence of the Ci1 protonation state, we determined the activation energy of this process based on the HOOP band decay kinetics in the temperature range from 130 to 180 K (see Fig. S8). Because the decay kinetics could not be described by a single exponential function, they were fitted by the empirical equation (1 + t/t₀)⁻ⁿ (with parameters t₀ and n) introduced by Austin et al. (32). This equation takes into account that at temperatures <180–220 K, the dark state comprises conformational substates with slightly different activation energies (32,68) (see Materials and Methods for details). A plot of RT × ln(n/t₀) versus temperature (Fig. 4 a) was fitted to a linear function, and the intercept of the ordinate represents the mean activation energy (E_A) for chromophore relaxation (inset). In this way, we obtained a mean activation energy of 34.8 ± 5.1 kJ/mol for the chromophore relaxation at pH 5 and 26.2 ± 1.5 kJ/mol at pH 9.

Next, we compared the decay kinetics of the HOOP bands that are indicative for chromophore relaxation of the ReaChR wild-type at pH 5, pH 7.4, and pH 9 and of the Ci1 and Ci2 mutants E163T and D293N, respectively (Fig. 4 b). All measurements were performed at 150 K. The decay is the slowest at pH 5 (t_0.8 = 4522 ± 73 s) when Ci1 is protonated and the fastest at pH 9 (t_0.8 = 29 ± 2 s) when it is deprotonated. Consequently, an intermediate value (t_0.8 = 299 ± 16 s) was found at pH 7.4 because of the presence of both Ci1-COOH and Ci1-COO⁻. Similar to pH 9, the decay kinetics is accelerated in the Ci1 mutant E163T (t_0.8 = 61 ± 7 s). These findings hint at a stabilizing effect of protonated Ci1 on the distorted chromophore conformation that is reduced by both its deprotonation and mutation. Ci2 neutralization (D293N) leads to delayed decay kinetics of the HOOP band as compared with the wild-type at pH 7.4 (t_0.8 = 3508 ± 90 s) and resembles the value of the wild-type at pH 5.

MD simulations of the dark-state structure with protonated and deprotonated Ci1

To further explore the experimental assignment of the Ci1 protonation state and to demonstrate possible effects of the Ci1 protonation state on the hydrogen-bonding patterns in the dark state, we performed MD simulations. The simulations were conducted with protonated and deprotonated Ci1, as shown in Figs. 5 and S3–S5. For protonated Ci1, we obtained only one dominant hydrogen-bonding pattern (probability 94%; Fig. 5 b) in which Ci1 forms a hydrogen bond to Ci2 (3.1 Å) that is directly linked to the RSBH⁺ (2.7 Å) via a salt bridge. The central gate residue E130 is hydrogen-bonded to Ci2 via a water molecule. In contrast, for deprotonated Ci1, we obtained three hydrogen-bonding networks, indicating a larger heterogeneity of the dark state when Ci1 is charged. The highest probability (59%; Fig. 5 c) is found for the pattern in which deprotonated Ci1 is linked to the RSBH⁺ via a salt bridge (2.7 Å). In this structure, no interaction between E130 and the counterion complex was observed, whereas the hydrogen bond between Ci1 and Ci2 is weak and mediated by a water molecule (4.1 Å). The structure with 25% probability (Fig. 5 d) exhibits a salt bridge between Ci2 and the RSBH⁺ (2.8 Å) as well as a hydrogen bond between Ci1 and Ci2 (3.8 Å) that involves a water molecule. In the structure with 16% probability (Fig. 5 e), no direct salt bridge is formed between either of the counterions and the RSBH⁺, but instead, they are indirectly linked via a water molecule similar to BR (7). Ci1 and Ci2 are connected via a water molecule. In the latter two structures (Fig. 5, d and e), E130 is linked to the counterion complex via a water-mediated hydrogen bond.

Discussion

Ci1 (E163) accounts for the pH dependence of the ReaChR absorption maximum

The pH-titration experiments of ReaChR wild-type, the counterion mutants of Ci1 (E163T) and Ci2 (D293N), and the central gate mutant E130Q presented in Fig. 1 reflect a complex electrostatic interplay of these residues in the binding pocket and raise the question which of these residues account for the apparent pK_a of 7.6. Based on spectroscopic and electrophysiological experiments, we consider Ci2 as unlikely mainly for two reasons: 1) The L→M transition, i.e., formation of the conducting state, involves deprotonation of the RSBH⁺ and consequently requires a proton acceptor. However, the kinetics of this transition is hardly influenced by the E163T mutation as compared with the wild-type (18.8 vs. 10.8 μs (27)) but is significantly altered by the D293N mutation (245 vs. 10.8 μs (27)). This strongly indicates that Ci2 is the acceptor of the RSBH⁺ proton and must therefore be deprotonated in the dark. Ci2 mutation prevents its function as a proton acceptor and explains the severe influence of this mutation on the kinetics of M-state formation. 2) In ReaChR, formation of the conducting pore is initiated by formation of the M-state, which requires deprotonated Ci2 as acceptor of the RSBH⁺ proton. Although the photocurrents of E163T largely resemble wild-type-like photocurrents, the D293N mutation significantly decreases the photocurrents to 8% (27), which, again, argues for deprotonated Ci2 in the wild-type receptor. Therefore, we conclude that Ci2 remains deprotonated in the pH range investigated and strongly favor Ci1 as the residue that accounts for the pK_a of 7.6.

Finally, we have to explain why both counterion residues exhibit a residual pH dependence and cause an upshift of the pK_a, as indicated in Fig. 1 c. Because Ci2 remains deprotonated in the investigated pH range, the residual pH dependence of the Ci1 mutant (E163T) can only be explained by an additional titratable group that is not part of the counterion complex. There is strong evidence that the central gate residue E130 is involved for the following reasons: 1) E130 is located close to the counterion complex and is most likely part of a hydrogen-bonded network involving E130, Ci1, and Ci2 (Figs. 1 and 5). 2) Previous FTIR experiments have shown that E130 deprotonates in the ReaChR photocycle at a physiological pH and must therefore have a high pK_a in the dark state. 3) E130 has a strong influence on the electrostatics of the binding pocket, indicated by a significantly blue-shifted absorption maximum of the E130Q mutant as compared with the wild-type (λ_max = 513 nm vs. λ_max = 527 nm (27)). Likewise, the pH dependence of the D293N mutant reflects titration of E130 as well because this alteration strongly lowers the pK_a of Ci1 so that it remains deprotonated in the range between pH 5 and 9 and forms a salt bridge to the RSBH⁺. This assumption is supported by the only moderate spectral shifts in the spectra of the counterion mutants (λ_max(E163T) = 530 nm and λ_max(D293N) = 528 nm (27)) as compared with the wild-type (λ_max = 527 nm (22)). Accordingly, the pK_a increase of both counterion mutants (8.8 for E163T and 9.8 for D293N as compared with the wild-type; see Fig. 1 c) reflects a more efficient stabilization of protonated E130 in the dark state because of these alterations. E130Q mutation reflecting titration of Ci1 leads to a pK_a downshift to 5.7. This indicates an impaired stabilization of protonated Ci1 due to the E130Q exchange. The strong increase of the pK_a-value of Ci1 in ReaChR as compared with C1C2 can be explained by a number of variations in the retinal-binding pocket and the active site (Fig. S9).

Chromophore relaxation depends on the Ci1 protonation state

Conversion of absorbed light energy into structural rearrangements that eventually generates an output signal is decisive for the function of photoreceptors. Here, we show for the ChR variant ReaChR that the mechanism of light energy storage by the chromophore and subsequent transfer of the stored energy from the chromophore to the protein is modulated by the protonation state of Ci1 (E163, pK_a = 7.6) located in the active site. Protonated Ci1 stabilizes the distorted chromophore conformation and delays energy transfer by chromophore relaxation, whereas its deprotonated form accelerates energy transfer. Consequently, both the decay kinetics of the distorted retinal conformation and, accordingly, the activation energy for the relaxation process (E_A = 26.2 ± 1.5 kJ/mol at pH 9 vs. E_A = 34.8 ± 5.1 kJ/mol at pH 5) depend on the Ci1 protonation state (see Fig. 4). This is in line with our recent findings from transient UV-Vis spectroscopy that revealed a significant deceleration of the decay kinetics of the early K intermediate with protonated Ci1 at pH 5 (τ_off ≈ 1.72 μs) as compared with the deprotonated configuration at pH 9 (τ_off ≈ 391 ns) and the Ci1 mutant E163T at pH 7.4 (τ_off ≈ 455 ns) at single-turnover conditions (27,28). Because chromophore distortion around the C=C bonds is one of the decisive factors for the spectral red shift of K as compared with the dark state, decay of K that is accompanied with a blue shift of the absorption spectrum reflects the release of the strain energy.

In the following, we discuss the influence of the different Ci1 protonation states and Ci1 mutations on the chromophore relaxation and, thus, the dynamics of energy transfer into the protein.

Protonated Ci1 is involved in a hydrogen-bonded system that stabilizes the distorted chromophore

Both deprotonation of Ci1 at pH 9 and its mutation to a threonine lead to a comparable reduction of the chromophore distortion after photoactivation as well as to an accelerated release of the strain energy as compared with its protonated form at pH 5 (see Figs. 2, 3, and 4). This indicates that the accelerated energy transfer to the protein cannot be explained simply by the formation of the negative charge (Ci1-COO⁻) of deprotonated Ci1 at pH 9 because no negative charge is formed with the threonine mutant. Instead, we propose that both deprotonation and neutralization of Ci1 by E163T mutation interrupt or weaken a hydrogen-bonded system involving protonated Ci1 that stabilizes the distorted chromophore conformation in the K intermediate more efficiently than with deprotonated Ci1 (Fig. 6). A direct proof for a hydrogen bond change of protonated Ci1 comes from the band pattern at 1705(−) and 1713(+) cm⁻¹ in the FTIR spectra recorded at 80 K and pH 5, which is neither observed at pH 9 nor in the E163T mutant at pH 5 (Fig. S10). The position of the ν(C=O) stretch vibration at 1705(−) cm⁻¹ reveals that this hydrogen bonding is rather strong in the dark and is slightly weakened in the early K intermediate (69). Based on our MD simulations, we suggest that the 1705(−)-and-1713(+)-cm⁻¹ doublet reflects a change of the Ci1-Ci2 hydrogen bond that is formed with protonated Ci1 in the dark state (Fig. 5 b). Ci2, which is deprotonated in the dark, is linked to the RSBH⁺ via a salt bridge that might become strengthened because of the Ci1-Ci2 interaction, whereas no direct interaction between Ci1 and the RSBH⁺ is observed (Fig. 5 b). Interestingly, a similar band pattern is not observed in the FTIR spectra at 150 K (see Figs. 2 and S10). This suggests the existence of two K intermediates with similar UV-Vis absorption maxima (λ_max = 584 nm at 80 K and λ_max = 590 nm in the wild-type at pH 7.4): an early K (80 K) intermediate in which configuration and bond length of the hydrogen bond between Ci1 and the Ci2 are altered as compared with the dark state and a late K intermediate (150 K) in which the configuration of this hydrogen bond is more similar to the dark state. Therefore, we suggest that the salt bridge between Ci2 and RSBH⁺ still persists in the K intermediate and stabilizes the distorted chromophore (Fig. 6).

Figure 6.

Model for the effect of the Ci1 protonation state on energy transfer. The all-trans, 15-anti RSBH⁺ dark states of the active sites with Ci1-COOH (a) and Ci1-COO⁻ (b) are shown. With Ci1-COOH, Ci1 forms a hydrogen bond to Ci2, and Ci2 is linked to the RSBH⁺ via a salt bridge (see also, Fig. 5b). In the K state, this rigid hydrogen-bonding pattern largely persists, thus stabilizing the distorted chromophore conformation, which impedes transfer of the stored energy to the protein. For the dark state with Ci1-COO⁻ (b), we consider only the most probable active-site configuration with a salt bridge between Ci1 and the RSBH⁺ and a weaker, water-mediated interaction between Ci1 and Ci2 (see Fig. 5c). A hydrogen bond change from Ci1-RSBH⁺ to Ci2-RSBH⁺ was described to occur during the D→K transition of CrChR2 (73). This possibly also holds true for the ReaChR active site in the presence of Ci1-COO⁻, resulting in a higher flexibility and, consequently, a more efficient release of the strain energy from the chromophore to the protein.

In contrast to the dark state with protonated Ci1 for which we obtain only one structure by MD simulations with relevant probability (94%; Fig. 5 b), we observe three dark-state structures with deprotonated Ci1 (Fig. 5, c–e), indicating that isomerization, and thus K state formation, starts from a structural ensemble. Therefore, we conclude that also the K state might be structurally inhomogeneous and that in none of the possible K state structures, a hydrogen-bonded system exists that stabilizes the distorted chromophore conformation as efficiently as with protonated Ci1. A structural inhomogeneity of the dark state was also observed in QM/MM (quantum mechanics/molecular mechanics) calculations for CrChR2 (70) and for the chimeric ChR C1C2 (71,72). Contrarily, Ardevol and Hummer found a homogenous active site, i.e., only one conformation for CrChR2 in which both counterions are deprotonated and Ci1 forms a salt bridge to the RSBH⁺ (73), similar to the structure shown in Fig. 5 c. Interestingly, the authors describe a switch in hydrogen bonding of the RSBH⁺ from Ci1 to Ci2 in the transition from the dark state to the K state analog P500 (73). A similar switch might also occur during D to K transition of ReaChR with deprotonated Ci1, accounting for an enhanced flexibility that leads to a less efficient stabilization of the distorted chromophore and thus to an accelerated energy transfer to the protein in the K→L transition. In the hydrogen-bonded system with protonated Ci1, however, such a change in hydrogen bonding of the RSBH⁺ in the transition from D to K is unlikely because Ci2 remains the only charged counterion in the active site that is able to form a strong salt bridge to the RSBH⁺. Therefore, we suggest that in the K state, protonated Ci1 is involved in a more rigid hydrogen-bonded system that stabilizes the distorted chromophore conformation and decelerates energy transfer to the protein. This assumption is also in agreement with a theoretical study by Warshel and Barboy that demonstrated that the rigidity of the chromophore environment and the strain energy stored in the chromophore are correlated (65). A reorientation of hydrogen-bonding networks early after recovery to the S₀ ground state reaction coordinate and, thus, an unusually fast protein response was also observed for CrChR2 in the FTIR spectra of the K analog, P500, recorded at 80 K (74,75), and by time-resolved FTIR spectroscopy in the time range from femtoseconds to nanoseconds (76).

The Ci1 protonation state modulates the mechanism of energy storage

The reduced HOOP band intensities in the K-D FTIR difference spectra of deprotonated or mutated Ci1 as compared with protonated Ci1 (see Figs. 2 and 3) show that the Ci1 protonation state does not only influence the dynamics of chromophore relaxation but also largely affects the extent of retinal distortion that is more pronounced with protonated Ci1. Because chromophore distortion of the C=C bonds in the polyene chain typically results in bathochromic shifts of the UV-Vis absorption maximum (77), the most significant bathochromic shift should be expected at pH 5, at which the distorted conformation is efficiently stabilized. However, for the transition from the D to the K state, we observe a more pronounced bathochromic shift (from 502 to 588 nm) at pH 9 as compared with pH 5 (from 524 to 603 nm), although the chromophore is less distorted (28). This implies that the bathochromic shift caused by charge separation between Ci1 and the RSBH⁺ that arises in the active site because of deprotonation of Ci1 compensates the effect of the reduced chromophore distortion at pH 9. We therefore suggest that with deprotonated Ci1, charge separation might become the more dominant effect for energy storage as compared with the situation with protonated Ci1 in which the energy is predominantly stored in the distorted chromophore conformation.

A link between the electrostatic interactions in the active site and the mechanism of energy storage was also observed in previous studies on BR and bovine rhodopsin. In BR, similar to ReaChR, two negatively charged counterions are present in the active site at alkaline pH, and energy storage is predominantly achieved by charge separation (4,5). However, in BR, the interaction between the counterions and the RSBH⁺ is mediated by a water molecule (7). We also found a comparable structure in ReaChR with deprotonated Ci1, albeit with the lowest probability (see Fig. 5 e), whereas in the structures with higher probabilities, no water molecule is found between the counterions and the RSBH⁺ (see Fig. 5, c and d). Instead, in bovine rhodopsin with only one negatively charged counterion in the active site, the relative contribution of chromophore distortion to energy storage is enhanced (3,6,8). In ReaChR, both the protonated and the deprotonated form of Ci1 exist under physiological conditions due to its pK_a of 7.6. Thus, we suggest that in ReaChR, the mechanisms for the storage of light energy can be adapted to the environmental pH conditions: at low pH with protonated Ci1, the contribution of chromophore distortion to energy storage in the early K state is relatively strong because of the tight hydrogen bonding in the active site, whereas at higher pH, hydrogen bonding is weakened, and energy storage is predominantly achieved by charge separation of the RSBH⁺ and its counterions. This ensures efficient transfer of the absorbed light energy into the cell at changing pH conditions. However, owing to the observation that E163T destabilizes the chromophore as deprotonated Ci1, other factors apart from the Ci1 protonation state must contribute to the mechanism of energy storage because, in this mutant, no additional negative charge is present in the active site. One possible explanation is the intrusion of additional water molecules and/or anions such as Cl⁻ in the active site that compensate the neutralizing effect of the Ci1 mutation.

Author Contributions

The manuscript was written through contributions from all the authors. All authors have given approval to the final version of the manuscript. J.C.D.K. designed and conducted the FTIR measurements. B.S.K. expressed and purified the protein from HEK293T cells and performed the pH titrations. S.A. performed the homology modeling, MD simulations, and the theoretical data analysis. E.R. provided the evaluation software. I.S. contributed to analyzing and interpreting the computational results. E.R., P.H., and F.J.B. contributed to interpreting the experiments. J.C.D.K. and F.J.B. wrote the manuscript with further contribution from all the authors.

Editor: Philip Biggin.

Supporting Citations

References 78, 79, 80 appear in the Supporting Material.