Research ArticleStructural Biology

Design of a light-gated proton channel based on the crystal structure of Coccomyxa rhodopsin

See allHide authors and affiliations

Science Signaling  19 Mar 2019:
Vol. 12, Issue 573, eaav4203
DOI: 10.1126/scisignal.aav4203

The light between pumps and channels

Whereas pumps move ions actively, channels move ions passively. Fudim et al. generated a high-resolution crystal structure for Coccomyxa subellipsoidea rhodopsin, a light-activated proton pump. These data enabled the authors to identify a critical interaction between Arg83 and Tyr14 in a transmembrane domain and generate a point mutant of this rhodopsin that behaved as a light-gated proton channel. These results provide greater insight into the molecular determinants that distinguish proton pumps from channels.


The light-driven proton pump Coccomyxa subellipsoidea rhodopsin (CsR) provides—because of its high expression in heterologous host cells—an opportunity to study active proton transport under controlled electrochemical conditions. In this study, solving crystal structure of CsR at 2.0-Å resolution enabled us to identify distinct features of the membrane protein that determine ion transport directivity and voltage sensitivity. A specific hydrogen bond between the highly conserved Arg83 and the nearby nonconserved tyrosine (Tyr14) guided our structure-based transformation of CsR into an operational light-gated proton channel (CySeR) that could potentially be used in optogenetic assays. Time-resolved electrophysiological and spectroscopic measurements distinguished pump currents from channel currents in a single protein and emphasized the necessity of Arg83 mobility in CsR as a dynamic extracellular barrier to prevent passive conductance. Our findings reveal that molecular constraints that distinguish pump from channel currents are structurally more confined than was generally expected. This knowledge might enable the structure-based design of novel optogenetic tools, which derive from microbial pumps and are therefore ion specific.


Microbial rhodopsins are integral membrane photoreceptor proteins that are found in numerous prokaryotes, fungi, and algae (1, 2). They act as light-activated sensors, enzymes, ion channels, or light-driven ion pumps. Light sensitivity is mediated by all-trans-retinal, covalently bound through a protonated Schiff base to a conserved Lys residue in transmembrane helix seven (TM7). The best characterized prototype of a light-driven proton pump is bacteriorhodopsin (BR) from Halobacterium salinarum (3). Active proton transport across the membrane by BR generates an electrochemical gradient that is used by the cells for adenosine triphosphate synthesis and other vital energy-requiring functions (4). Thermal relaxation occurs through several distinct intermediates (K, L, M, N, and O) (57). Structural changes along transitions between these states enable successive proton release to the extracellular medium and delayed proton uptake from the intracellular side. The stepwise but spatially noncontinuous, unidirectional proton pathway is spectroscopically well characterized (8, 9). However, molecular constraints that define the pumping power have not been identified, primarily because of poor heterologous expression of BR in host cells, which is why only few electrical studies have been carried out under controlled electrochemical conditions (1012).

Nevertheless, the desired use of microbial pumps as optogenetic tools is based on their comparably small size and ion specificity. Still, more detailed knowledge about the electrophysiological properties of light-driven ion pumps is required for optogenetic application of ion pump derivatives (1315). We have described Coccomyxa subellipsoidea rhodopsin (CsR) from the polar freshwater algae C. subellipsoidea (Trebouxiophyceae) as a versatile model to study light-driven proton pumping under controlled electrochemical load (16). Tyr57 and Arg83 (in BR Tyr57 and Arg82) in the extracellular half-channel are critical residues to prevent proton leakage in light-driven proton pumps. In particular, R83Q and Y57K substitutions result in pronounced passive proton conductance at moderate electrochemical load but with strong inward or outward rectification and destabilized protein, preventing further in vitro analysis (16). Therefore, the molecular details of the effects of these amino acid substitutions remain unclear.

In this study, we present the crystal structure of wild-type CsR (CsR-WT) at 2.0-Å resolution, which allowed us to identify the unique structural constraints that account for the functional differences between CsR and other light-driven proton pumps such as BR. Modifying the unique interaction in CsR between Arg83 and Tyr14 in TM1 altered the protein’s mode of action. In particular, replacement of Tyr14 with a negatively charged glutamate residue transformed CsR into an operational channel with almost symmetric photocurrents under physiological conditions while maintaining proton selectivity. Our study unravels molecular details that differentiate light-driven ion pumps from light-gated ion channels [channelrhodopsin (ChR)]. The results highlight the role of Arg83 as a barrier in the extracellular half-channel to prevent proton leakage and points out the involvement of carboxylates at the extracellular surface. We conclude that the structural restraints differentiating light-driven pumps from light-gated channels are less notable than anticipated.


Crystal structure of CsR-WT

To facilitate the heterologous expression and purification of CsR-WT, the protein was C-terminally truncated by 74 amino acids, keeping intact TM1 to TM7. Functionality of the truncated version was demonstrated by comparing electrical properties of the two CsR variants (fig. S1, A and B). CsR-WT was crystalized in lipidic cubic phase (LCP) at 20°C under dim red light. The obtained rhombohedral CsR crystals diffracted up to 2.0-Å resolution with one monomer per asymmetric unit and visibility of amino acid residues 2 to 226 (Table 1). The crystal packing of CsR-WT displayed a homotrimeric arrangement that is stabilized primarily by hydrophobic protein–lipid interactions, with an overall structural topology similar to that of other light-driven pumps, such as BR and halorhodopsin (fig. S2, A to C) (17, 18). The initial electron density map calculated without the retinal ligand was readily interpreted with a typical all-trans-retinal chromophore bound to Lys215 of TM7 through a retinal Schiff base (RSB) (Fig. 1A and fig. S3, A and B). Protonation of the RSB (RSBH+) is expected to be stabilized by Asp86 and Asp211, forming a BR-like pentameric counterion complex involving three well-ordered water molecules (Fig. 1B). Asp86 and Asp211 are kept anionic by additional hydrogen bonds to Thr90 and Tyr57/Tyr185, respectively. Contact to the extracellular side is established through Arg83 and an additional hydrogen-bonding network including six water molecules in total (Fig. 1, B to D). The highly conserved Arg83, previously described as a key proton shuttle residue in BR (19) and CsR (16), forms, in CsR, a hydrogen bond with the hydroxyl group of a nonconserved Tyr14 at the beginning of TM1 (Fig. 1, C and D). The benzene ring of Tyr14 is stabilized by an almost pure hydrophobic cage shaped by Tyr57 and Met60 of TM2, Ala204 of TM7, Tyr80 of the β-sheet–TM3 interconnection, and Ile6 of the N terminus (Fig. 1C). Moreover, the extracellular moiety of CsR-WT displays several unique features. Within the antiparallel β-sheet between TM2 and TM3, a central Arg77 engages with the carboxylic group of Glu203 and the main-chain carbonyl of Ile198, adjusting the interconnection loop between TM6 and TM7 (Fig. 1D). This stabilizing network contributes to the assembly of a potential proton release complex. The proton release complex in CsR includes two water molecules and the residues Glu193 in TM6 and Glu203 in TM7 (Fig. 1D), which are both conserved but displaced in comparison to BR (Glu194 and Glu204 in BR) (fig. S4) (20). One reason for this displacement is the lack of Arg77 in BR, where BR-Glu204 forms a hydrogen bond to the hydroxyl group of BR-Ser193 instead. The central Arg77-Glu203 interaction is also present in Acetabularia rhodopsin I and II (ARI and ARII) (21, 22); however, the second glutamate (Glu193) is replaced by Ala196 in ARI and by Ser189 in ARII. In CsR, Glu193 forms hydrogen bonds with Tyr84 and Tyr207 (Fig. 1D), which keep it facing into the water cavity, eliminating the glutamate dyad interaction observed in BR (fig. S4) (17). In BR, a similar interaction between BR-Glu194 and BR-Tyr83 forms during the M state; however, Tyr207 is conserved solely within the proteorhodopsin family and is substituted by Phe208 in BR (20, 23, 24). An overview about structural differences within the extracellular half of CsR and other microbial rhodopsins is shown in fig. S5 (A to D). Nonetheless, the crystal structure supports our previous assumption of a common proton pumping mechanism shared by CsR and BR (16), due to the similar positioning of conserved proton transport–relevant residues (fig. S4).

Table 1 Data collection and refinement statistics.

PDB, Protein Data Bank; RMSD, root mean square deviation.

View this table:
Fig. 1 Structural architecture of CsR-WT.

(A) Overall crystal structure of monomeric CsR. The transmembrane domain consists of seven α helices (TM1 to TM7). Membrane boundaries were calculated using the positioning of proteins in membrane-server (PPM) (74) and are shown as red and blue dashed lines. (B) Close-up view of the counterion complex displaying a conserved pentameric configuration of the counterions and three well-ordered water molecules. (C) Residues involved in forming a hydrophobic cage around Tyr14 (green). (D) Close-up views of the hydrogen-bond network connecting the RSB to the extracellular side, composed of the counterion complex and a highly entangled proton release complex. (E) Close-up view of the cytoplasmic half-channel. Important amino acid residues involved in forming the half-channel are depicted as sticks. The covalently linked chromophore all-trans-retinal (RET) is shown in red. Several ordered lipids are displayed as green sticks. Hydrogen bonds are depicted as dashes. Water molecules (W) are shown as purple spheres.

In contratst to the extracellular side, which displays typical features of various light-driven proton pumps, the cytoplasmic half-channel of CsR is closely related to that of BR. Within the cytoplasmic half-channel, the proton donor of the RSB, Asp97 (BR-Asp96), forms a hydrogen bond to Thr46 (BR-Thr46), which, in BR, plays a crucial role in modulating the pKa of BR-Asp96 (25). Apart from Thr46, Asp97 is almost completely surrounded by nonpolar residues in CsR (Phe28, Ile49, Leu94, Ile98, Leu174, Leu178, Phe218, and Val222; Fig. 1E). The only exception is the nonconserved His42, which is located above Asp97 at the edge of the cytoplasmic side (Fig. 1E). The cytoplasmic surface of CsR lacks all the carboxylates associated with proton uptake in BR (26), suggesting that His42 may be involved in the reprotonation of Asp97 in CsR, but this remains to be validated experimentally.

Introducing passive conductance

As we previously demonstrated (16), the extracellular half-channel of CsR contains two key residues important for maintaining proton pumping, or more accurately, for preventing passive proton leakage especially at high electrochemical load. One residue is Arg83, which forms multiple hydrogen bonds to different partners, most relevant for the stability, functionality, and overall configuration of the protein. The substitution of Arg83 with glutamine caused passive conductance but severely destabilized the recombinant protein. Furthermore, the absence of the positively charged guanidinium group altered the pKa of the primary proton acceptor Asp86, as demonstrated by a pH-induced bathochromic shift at pH 5 in the R83Q mutant (pKa, ~5.8; fig. S6, A and B), also reported for the analogous mutation in BR (27). To improve the properties of the engineered passive proton transporter, the configuration of Arg83 must be altered in a less destructive manner.

The crystal structure suggested that Tyr14 (Fig. 1, C and D) may be a candidate to focus on for further studies of the proton pathway within the extracellular half-channel, because it is distant from both the counterion complex and the proton release complex. However, we first needed to rule out that Tyr14 itself contributes substantially to proton pumping in CsR. We tested the necessity of the Tyr14-Arg83 hydrogen bond by substituting tyrosine with phenylalanine. Electrophysiological recordings in Xenopus laevis oocytes showed that the pumping characteristics of the Y14F mutant were similar to those of the WT protein (Fig. 2A). The currents remained solely outward directed at holding voltages between −125 and +75 mV, and the influence of the external pH (pHo) remained small (Fig. 2A). This result ruled out a necessity of the Tyr14-Arg83 interaction for proton pumping. Nevertheless, the Y14F mutation accelerated the early steps of the photocycle preceding deprotonation of the RSBH+ (fig. S7A), which are associated with the movement of Arg82 in BR (28, 29). Accordingly, we modified the Tyr14 position to displace Arg83 and reshape the configuration of the extracellular half-channel. We expected that introducing a positively charged lysine would alter the extracellular half-channel by a repulsive force between Arg83 and Lys14. However, at pHo 7.5, CsR-Y14K showed exclusively outward pump currents, although current amplitudes were decreased (Fig. 2A). Only at high electrochemical load (namely, a strong pHo gradient at pHo 5) and negative holding voltage did CsR-Y14K show inward photocurrents (Fig. 2A), suggesting that secondary rearrangements were required to induce passive conductance. Because pH differences affect primarily the residues at the extracellular surface, protonation of carboxylates such as Glu194, Asp196, or Glu204 is likely involved. To establish a more rigid configuration of the extracellular half-channel and therefore further characterize the molecular constraints of unidirectional proton transport, we replaced Tyr14 with the negatively charged glutamate. An additional attractive force, which may result in formation of a potential Glu14-Arg83 salt bridge within the photocycle and could alter the Arg83-Glu203 interaction (Fig. 1D), thereby creating space for additional water molecules. This would connect the counterions with the extracellular side and maintain the extracellular half in a more “open” configuration. This structure-guided design was matched by experimental observations.

Fig. 2 Electrophysiological characterization of CsR-WT and CsR variants with different amino acids at position 14 in X. laevis oocytes.

(A) Absolute photocurrent amplitudes of different CsR variants at three different holding potentials and pHo values. Number of measured oocytes at pH 5, pH 7.5, and pH 10: WT, n = 10, 34, 7; Y14F, n = 5, 6, 5; Y14Q, n = 5, 7, 5; Y14K, n = 6, 6, 6; Y14D, n = 4, 6, 3; and CySeR (Y14E), n = 4, 10, 4. Error bars represent mean SE. (B) Representative photocurrents for CsR-WT (top) and CySeR (bottom) at an external pHo of 7.5 and different holding potentials. Positive currents indicate outward-directed proton flux, whereas negative currents indicate inward-directed proton flux. (C) Reversal potential of CySeR at different pHo values. The black line represents the reversal potential of a pure proton channel according to the Nernst equation. Acidification of the protein environment shifts the reversal potential toward active transport.

CsR-Y14E showed passive photocurrents that were almost symmetric at neutral pH, and the current amplitude was graded by the electrochemical gradient (Fig. 2B). Replacing NaCl with N-methylglucamine (NMG-Cl) or Na-gluconate did not influence photocurrents, indicating that only protons were conducted (fig. S8). We named this designed light-gated proton channel “CySeR” (Coccomyxa Y replaced E rhodopsin). Nonetheless, CySeR showed a reversal potential of −30 mV under quasi-symmetric pH conditions (pHo 7.5) (Fig. 2, B and C), indicating some residual pump activity. This residual pumping was increased with extracellular acidification, whereas the protein functioned as a pure channel under alkaline conditions (Fig. 2C). These findings are consistent with a pH-dependent protonation of Glu14 (<pHo 8), leading to the neutralization of the ionic glutamate and a more WT-like configuration of the extracellular half-channel. To challenge this hypothesis, we carried out Fourier transform infrared (FTIR) spectroscopy of CsR-WT and CySeR. At pH 5, a positive band at 1720 cm−1 that appeared after illumination for CySeR, but not CsR-WT, can be attributed to the protonation of Glu14 during the photocycle and, in conjunction with this observation, suggests that Glu14 is deprotonated in darkness (fig. S9).

For further validation of the Glu14-Arg83 interaction, we tested two other mutations. Replacement of Tyr14 with aspartate caused bidirectional photocurrents (Fig. 2A), whereas replacement with glutamine preserved outward-directed currents (Fig. 2A), ruling out steric hindrance and emphasizing the necessity of an electrostatic interaction. Because the side chain of Tyr14 is embedded in a hydrophobic cage of amino acid residues from diverse structural parts (Figs. 1C and 3A), we carried out short molecular dynamic simulations (1 ns) to test the influence of the Y14E substitution on the configuration of the extracellular half-channel. The calculations (Fig. 3B) suggested that the side chain of Glu14 in CySeR is shifted compared to WT Tyr14 because of rearrangement of the hydrophilic side chain within this hydrophobic pocket (Fig. 3, C compared to D). In the molecular dynamic simulation, the supposed orientation of Glu14 in CySeR does not show hydrogen-bonding interaction with Arg83 due to the long distance of 5 Å (Fig. 3, A to D), explaining that conformational modifications as they apparently occur during the photocycle are needed to bring the residues into closer contact.

Fig. 3 Structural comparison of short molecular dynamic simulations of CsR-WT and CySeR.

(A) Molecular dynamic simulation of CsR-WT (blue ribbon representation) showing the environment of Tyr14. The side chain of Tyr14 (green sticks) is tight and cage-like embedded by hydrophobic amino acid side chains (gray sticks). The hydrogen bonding network of Tyr14-Arg83-Glu203 stabilizing this structural microenvironment, intramolecular interactions, and connection to the proton release complex are shown. (B) Superposition of CsR-WT (blue ribbons and light gray sticks, Tyr14 highlighted as green sticks) and CySeR (black ribbons and sticks, Y14E highlighted as magenta sticks), after short molecular dynamic simulations (1 ns). The simulated structure of CsR-Y14E suggests that the side chain of the substituted Glu14 is shifted, most likely by repulsion and delocalization of the negatively charged and hydrophilic side chain within this region. (C) Close-up view and surface representation of Tyr14. (D) Close-up view and surface representation of Glu14, showing the >5-Å distance between Arg83 and Glu14.

Pump and channel photochemistry

To expose mechanistic differences between CsR and CySeR, laser-induced photocurrents were compared with transient absorption changes. For spectroscopic measurements, the truncated versions of CsR and CySeR were expressed in human embryonic kidney (HEK) 293T cells and purified in 0.02% n-dodecyl-β-d-maltoside (DDM). Absorption of purified CsR showed only minor pH dependence, with dark state absorption peaks at 536 nm for pH 5.0, 535 nm for pH 7.4, and 532 nm for pH 9.0 (Fig. 4A). Substitution of Tyr14 with glutamate caused small deviations in electrostatic interactions at the RSB counterion complex, observable as small hypsochromic shifts under neutral and alkaline conditions (531 nm for pH 7.4 and pH 9.0) (Fig. 4B), whereas absorption under acidic conditions shifted to 538 nm. The observed hypsochromic shift may indicate displacement of the counterions or the positively charged guanidinium group of Arg83 in CySeR, which would be in line with the observed shift in CsR-R83Q (fig. S4A). To monitor time-dependent photochemical processes, the samples were excited with green laser pulses (10 ns and 530 nm), and we recorded the spectral evolution of photointermediates from 100 ns until 10 s after the flash. CsR-WT exhibited the photochemistry typical of BR-like proton pumps (Fig. 4C) (30, 31). Three photointermediates were distinguishable and are referred to in chronological order as K, M, and O, in accordance with BR nomenclature. The first observed photointermediate preceding deprotonation of the RSBH+, the so-called K state, was red shifted compared to the dark state, peaking at about 580 nm (Fig. 4C, bottom panel, and fig. S10). Deprotonation of RSBH+ occurred within ~1.8 μs (table S1), causing a strong blue shift observable in the accumulation of M intermediate at about 406 nm (Fig. 4C and fig. S9). The deprotonation of the RSBH+ was also the first detectable electrogenic step, also previously shown for BR (32). By performing electrical measurements in single turnover experiments, we were able to correlate the outward-directed photocurrents (displacement current) with M state formation (Fig. 4C, top panel). CySeR showed a similar primary outward-directed charge transition (Fig. 4D, top panel), as well as similar photochemical properties at pH 7.4 (Fig. 4D, bottom panel) with comparable time constants for early photocycle transitions (table S1). This result was unforeseen because an established electrostatic interaction between Glu14 and Arg83 was expected to influence the primary conformational changes, including the reorientation of Arg83, as mentioned above and described for BR (29). Early photocycle reactions were impaired in CySeR only at pH 9.0, slowing down the transition from K to M by about twofold compared to CsR-WT (table S1 and fig. S7B).

Fig. 4 Functional comparison of CsR-WT and CySeR.

(A and B) Absorption spectra of purified CsR-WT (A) and CySeR (B) at pH 5 (green), pH 7.4 (red), and pH 9 (blue). Maximum absorption of the chromophore is displayed. (C and D) Averaged electrical recordings of CsR-WT (C; n = 6 cells, mean ± SE) and CySeR (D; n = 6 cells, mean ± SE) in HEK293 cells in 110 mM NaCl (pHo/i 7.2) under different holding potentials in single turnover experiments (top). Averaged laser flash photolysis data of CsR-WT (C; n = 15 measurements) and CySeR (D; n = 15 measurements) in 20 mM tris-HCl (pH 7.4), 150 mM NaCl, and 0.03% DDM (bottom). Distinguishable photocycle intermediates are marked with the nomenclature of analogous BR intermediates.

The main contribution to the early charge transition is the translocation of the proton from RSBH+ to the primary acceptor Asp86, as described for both BR (33) and the Guillardia theta cation channelrhodopsin 2 (34). In BR, proton release to the extracellular side occurs within 100 μs after illumination (35). However, we do not expect proton release in the microsecond range in CsR because only 3% of the overall charge displacement occurred during this early time window (Fig. 5, A and B). More pronounced electrogenic events occurred later, corresponding to reprotonation of the RSB from the internal bulk phase, observed as M decay (τM1, ~1 ms; τM2, ~4.4 ms) (table S1) (33, 36) and formation of O intermediates upon reisomerization of the chromophore and return to the dark state accompanied by structural rearrangements (τO, ~233 ms) (Figs. 4C and 5, A and B). During the O state, both the RSBH+ and the proton acceptor are protonated; therefore, O state absorption appears red shifted relative to the dark state (Fig. 4C, bottom panel, and fig. S10). Furthermore, electrophysiological recordings showed that only the last electrogenic step was strongly voltage dependent (Fig. 5B).

Fig. 5 Kinetic analysis of current components in CsR-WT and CySeR.

(A) Charge transfer within the photocycle of CsR-WT for each of the calculated time constants averaged over n = 6 cells. Total transported charge represented by a black curve. (B) Voltage dependence of the determined time constants for CsR-WT. (C) Voltage dependence of the time constants (left) and amplitudes (right) for CySeR. Error bars represent the SD.

Hence, the slower proton transfer of these last steps of the photocycle accounts for the low current amplitude at negative voltages (Fig. 2, A and B). Moreover, within the time frame of O intermediate accumulation, differences between CsR-WT and CySeR were observed already at pH 7.4, which were less pronounced in CySeR because of an accelerated transition to dark state (Fig. 4D, bottom panel, and table S1). In support of these findings, electrophysiological recordings showed an accelerated decay of the photocurrent within the latter time frame, especially at negative holding potentials (Fig. 5C, left panel). A voltage-independent time constant in the millisecond range exclusively represented inward-directed currents, whereas outward-directed currents were described by two distinct time constants (Fig. 5C, right panel).


In this study, the crystal structure of CsR revealed unique structural features that enabled us to design the passive proton-conducting variant CySeR, for which we characterized electrical and photochemical properties. The altered functionality of CySeR can be explained by the structural rearrangement of key amino acid residues of CsR-WT. The composition of the extracellular half-channel and, in particular, the entangled residues of the potential proton release complex, share molecular features of various light-driven proton pumps. In CsR, both conserved glutamates (Glu193 and Glu203) comprising the proton release group of BR are present; however, the dyad of glutamates described for BR is not present in the dark state of CsR. The proton release complex structure in CsR resembles the M state configuration of BR, where this dyad breaks down because of the movement of Arg82 toward Glu204 (20). One reason for this difference is the altered orientation of Arg83 in CsR, which basally interacts with Glu203 in the unphotolyzed state. Another reason is the interaction between Glu203 and Arg77. This specific interaction of an analogous glutamate with an arginine of the extracellular side is also present in Acetabularia rhodopsins ARI and ARII (21, 22). The pHo sensitivity of the photocurrents is low for ARI, with only a small current increase at pHo 10 (37). For ARII, an unknown residue involved in proton release (pKa, 10.5) has been identified on the extracellular side (38) and is likely the corresponding arginine (Arg72 in ARII). We therefore conclude that the Glu203-Arg77 interaction of CsR-WT is responsible for the low sensitivity to pHo. Proton pumps of the proteorhodopsin family generally lack a proton release complex and show increased pHo dependence (39, 40).

The proton release complex of CsR shows some similarity to those of ARI and ARII; however, the Acetabularia rhodopsins lack the additional Glu193 present in CsR. Although this glutamate is associated with fast proton release in BR (41), we did not expect proton release in CsR to occur immediately after deprotonation of the RSB because of the small amount of charge translocation within this time range. A delayed proton release might be—as in ARII—caused by the configuration of Glu203. However, because the contributing residues are, in general, conserved, the overall pumping mechanism of CsR is likely comparable to that of BR. On the other hand, CsR lacks cytoplasmic carboxylates thought to facilitate reprotonation of the proton donor Asp96 in BR (42), even more Asp38 and Glu161 are replaced by the positively charged Arg40 and Arg163, respectively. Therefore, the nonconserved His42 of CsR may act as a proton shuttle, as described in the human carbonic anhydrase II and the influenza A virus proton channel (M2) (43, 44), and maintain reprotonation of Asp97 from the cytoplasmic side.

Electrical recordings showed that kinetics of the late electrogenic processes account for the high voltage sensitivity of CsR under continuous illumination. Thus, voltage dependence is equivalent to a voltage-dependent modulation of these kinetics. These late electrogenic processes include reprotonation of the RSB from the cytoplasmic proton donor Asp97 and deprotonation of the primary proton acceptor Asp86 toward the proton release complex. A similar voltage-dependent effect was observed for reprotonation in BR (11), where modulation of the cytoplasmic half-channel altered voltage dependence of the protein (12). Because voltage, or more precisely the local voltage drop, modulates the pKa gap between the proton donor (Asp97) and the RSB, electrostatic alterations within the cytoplasmic half-channel can reduce the voltage dependence of the protein, as shown for the T46 N mutation (16).

Deprotonation of Asp86, which is associated with O decay, is extremely slow in CsR. The 233-ms time constant of the O to dark state transition determined at neutral pH is about seven times slower than what is generally accepted for BR (45). Studies of BR led to the conclusion that the pKa of BR-Asp85 is partly controlled by Arg82 (46). Correspondingly, the slow O decay in CsR can be associated with differences in the configuration of Arg83. However, this is not a disadvantage as long as proton capture is rate limiting.

A unique interaction observed in the crystal structure of CsR is the hydrogen bond between Arg83 and Tyr14, which guided us toward the design of the light-gated proton channel CySeR. We reasoned that substitution of Tyr14 with glutamate would increase the attractive force between these two residues, resulting in a different configuration of Arg83. However, at neutral pH, CySeR showed residual pump activity, resulting in a reversal potential of −30 mV. In addition, flash-induced electrophysiological recordings showed that the initial outward-directed current was still present at pH 7.4. It was only at pH ≥8 that CySeR truly behaved like a proton channel rather than a leaky proton pump, with substantial differences in its photocycle properties. This observation may be explained by the pH-dependent protonation of Glu14, which would resemble a more WT-like configuration. The fact that this protonation occurs already at pH 5 suggests that the pKa of Glu14 is elevated compared to glutamic acid in solution caused by the hydrophobic cage (Figs. 1C and 3, A and C). Moreover, the elevated pKa of Glu14 and the 5 Å distance to Arg83 exclude an initial interaction of these residues (Fig. 3, B and D). Furthermore, the inward-directed currents at pH 5 (Fig. 2A) suggest that the Arg83-Glu14 interaction, necessary for passive conductance, must occur during the photocycle and requires deprotonation of Glu14 as seen in the FTIR difference spectra. Passive conductance overlaps with late photocycle steps (Fig. 4D), which are accompanied by structural rearrangements leading to the formation of a transient water channel on the cytoplasmic side (4749). Therefore, we conclude that the interplay between Glu14 and Arg83 must precede the opening of the cytoplasmic half-channel.

Glu14 is located in close proximity to an extracellular glutamate in ChRs (Glu101 in ChR2), which is crucial for efficient proton conductance in the red light–sensitive, proton-selective ChR variant Chrimson (50) and is responsible for residual proton leakage in the engineered anion-conducting ChR variant ChloC (51). However, it remains unclear whether the proton pathway relies on distortion of the extracellular pore or it is simply formed by arginine reorientation and the subsequent formation of the Arg-Glu salt bridge. Therefore, the precise mechanism of passive proton conductance cannot be completely elucidated. An Arg-Glu salt bridge may also establish an alternative proton pathway, as observed in the voltage-gated proton channel HV1, where an interaction between the carboxylic acid Asp112 with a highly conserved arginine constitutes the selectivity filter (52).

Together, previous results indicate that a reversible side-chain orientation of Arg83 is necessary to maintain active proton transport and prevent proton backflow, which is now supported by our observations. These findings contradict conclusions drawn from experiments with archaerhodopsin-3, for which it has been argued that the functional determinant distinguishing active and passive ion transport is located in close proximity to the chromophore, but not on the cytoplasmic and extracellular transport channels (53). We conclude that the mutations introduced in in close proximity to the β-ionone ring of archaerhodopsin-3 (53) had primarily the effect of slowing down the pump cycle as described for corresponding mutations in BR (54). Second, we believe that the mutations in archaerhodopsin-3 in close proximity to the RSB (53) could have altered the configuration of the active site with impact on the entire protein structure, making conclusions drawn from the described channel mutant of archaerhodopsin-3 untenable. In relation to earlier studies (16, 53), our findings reveal specific and more detailed insights into the pump-channel dichotomy, which may guide further structure-based engineering approaches.


Molecular cloning, protein expression, and purification

For heterologous expression in Sf21 cells, CsR (amino acids 1 to 230) with a hexa-histidine tag at the C terminus was cloned into the pFastBac1 vector using BamHI and HindIII restriction sites. Site-directed mutagenesis was performed using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies) according to the manufacturer’s instructions. Expression of CsR in Sf21 cells was achieved by using the Bac-to-Bac Baculovirus Expression System (Invitrogen). Baculovirus-infected cells were cultured in Sf-900 II culture medium (Invitrogen) at 27°C and 100 rpm for 24 hours and then supplemented with 7.5 μM all-trans-retinal (Sigma-Aldrich). Cells were harvested 24 hours later by centrifugation (3500 rpm, 10 min, 4°C; Beckman JLA 8.100 rotor). The pellets were resuspended in a buffer containing 150 mM NaCl, 20 mM tris-HCl (pH 8), 20 mM imidazole, cOmplete protease inhibitor cocktail (Roche), and DNaseI (Roche). For solubilization, 1% DDM was added, and the mixture was stirred for 8 hours at 4°C. The insoluble fraction was removed by ultracentrifugation (50,000 rpm, 30 min; Beckmann Ti45 rotor). The supernatant was loaded on a HisTrap FF Crude 5-ml column (GE Healthcare) and then immediately eluted with 500 mM imidazole and desalted with buffer containing 150 mM NaCl, 20 mM tris-HCl (pH 8), and 0.03% DDM on a HiPrep 26/10 column (GE Healthcare). Fractions containing CsR were pooled, concentrated, and further purified by size-exclusion chromatography with the same buffer on a HiLoad 16/600 Superdex 200 pg column (GE Healthcare).

Lipidic cubic crystallization

The CsR sample (35 mg/ml) was mixed with 1-(9Z-octadecenoyl)-rac-glycerol (monoolein, X-LCP-101; Jena Bioscience) containing 10% cholesterol (C8667, Sigma-Aldrich) in a 1:1 (w/w) protein solution to lipid ratio using the Art Robbins LCP Mixer Station with two ARI LCP dispensing syringes for 1 hour at room temperature under dim red light. The protein-lipid mixture (50 nl) was delivered through an LCP dispensing robot (Gryphon, Art Robbins Instruments) to 96-well glass sandwich plates and overlaid with 1-μl precipitant solution. Crystals for data collection were grown in 42% (v/v) polyethylene glycol 400, 100 mM MES (pH 6.5), and 150 mM sodium acetate. The crystals reached full size within 10 to 14 days at 20°C and were picked from the mesophase and flash-frozen in liquid nitrogen without additional cryoprotectant.

Data collection and structure analysis

Diffraction data of more than 100 crystals were collected at 100 K using synchrotron x-ray sources at the European Synchrotron Radiation Facility (ESRF; Grenoble, France) and Berliner Elektronenspeicherring-Gesellschaft für Synchrotronstrahlung (BESSY II, Berlin, Germany). The best diffraction data for the highest resolution of CsR-WT with all-trans-retinal were collected under red safe light at the synchrotron ESRF (Grenoble, France) beamline ID23-2 (55) with a Pilatus 6 M detector at λ = 0.8729 Å. All images were indexed, integrated, and scaled using the XDS program package (56) and the CCP4 programs SCALA (57, 58) and AIMLESS (59). All CsR-WT crystals belonged to the rhombohedral space group H3 (for CsR-WT: a = b = 78.08 Å, c = 143.97 Å, α = β = 90°, γ = 120°). Table 1 summarizes the statistics for crystallographic data collection and structural refinement. Karplus and Diederichs (60, 61) concluded that the extension of resolution is beneficial up to a statistically correlation coefficient CC1/2 as small as 0.1, when the Rmerge reaches above 300% and the I/σ(I) ratio is close to 0.3. After these from the crystallographic community-accepted suggestions, a CC1/2 of 0.3 is acceptable and should be the value for resolution cutoff. Therefore, we used a resolution cutoff at 2.00 Å (CC1/2 = 0.357) for the CsR model presented here. The values of Rwork/Rfree in the outer range are 37.1 and 42.6%, respectively.

Initial phases for CsR were obtained by the conventional molecular replacement protocol (rotation, translation, and rigid-body fitting) using the proton-pumping rhodopsin ARII from Acetabularia acetabulum (PDB entry 3AM6) (21) as an initial search model. After excluding retinal from the initial search model, molecular replacement was performed using the program Phaser (62). A simulated annealing procedure with the resulting model was performed using a slow-cooling protocol and a maximum likelihood target function, energy minimization, and B-factor refinement by the program PHENIX (63) in the resolution range of 32.91 to 2.00 Å for CsR-WT in its all-trans-retinal state. After the first round of refinements for the dataset, the all-trans-retinal in the ligand-binding pocket was visible in the electron density of both the σA-weighted 2mFo-DFc map, as well as the σA-weighted simulated annealing omitted density map for all-trans-retinal chromophore. All crystallographic structures were modeled with TLS refinement (64) using anisotropic temperature factors for all atoms. Restrained individual B-factors were refined, and the crystal structure was finalized by the CCP4 program REFMAC5 (64) and other programs of the CCP4 suite (57). The final model has agreement factors Rfree and Rcryst of 22.5 and 19.2%, respectively, for CsR-WT. Manual rebuilding of the crystal structure models and electron-density interpretation were performed after each refinement cycle using the program COOT (65). Structure validation was performed with the programs PHENIX (63), SFCHECK (66), WHAT_CHECK (67), and RAMPAGE (68). Potential hydrogen bonds and van der Waals contacts were analyzed using the programs HBPLUS (69) and LIGPLOT 1.45+ (70). All crystal structure superpositions of backbone α-carbon traces were performed using the CCP4 program LSQKAB (71). All molecular graphics in this work were created with PyMOL (version 0.99; The PyMOL Molecular Graphic system, Schrödinger, LLC).

Ultraviolet-visible absorption spectroscopy

Absorption spectra of purified protein were recorded at 293 K with a Cary 300 Bio spectrophotometer (Varian Inc.). For pH-dependent measurements, concentrated CsR samples were diluted in 20 mM MES (pH 5.0), 150 mM NaCl, and 0.03% DDM or in 20 mM tris (pH 7.4 or pH 9), 150 mM NaCl, and 0.03% DDM. We carried out pH titration measurements of CsR-Arg83Gln in buffer mixed from 200 mM Na2HPO4 and 100 mM citric acid stock solution. Buffers were supplemented with 150 mM NaCl and 0.03% DDM. Flash photolysis experiments were performed on an LKS.60 system (Applied Photophysics Ltd.) aligned to a tunable Rainbow OPO/Nd:YAG laser (Brilliant B, Quantel). Samples were excited by flashes of green laser light (535 nm) of ∼10 ns adjusted to a power of ~5 to 10 mJ per shot. A 150 W Xenon Short Arc XBO lamp (Osram) probed the resulting absorption changes. An Andor iStar ICCD camera (Andor Technology Ltd.) recorded the spectra at 41 time points between 100 ns and 10 s. After every measurement, the sample was kept in the dark for 15 s to ensure recovery to dark state. Time-resolved spectroscopic data were analyzed as previously described (72), and absolute spectra of CsR-WT were calculated as described by Krause et al. (73).

Electrophysiological recordings in X. laevis oocytes

The cloning of CsR, site-directed mutagenesis, mRNA synthesis, and extraction of oocytes from female X. laevis were carried out as previously described (16, 37). Truncated CsR was produced by mutagenesis of the Arg231 codon to a stop codon. Oocytes were injected with 30 to 35 ng of capped mRNA and incubated at 18°C in oocyte Ringer solution supplemented with 5 μM all-trans-retinal (Sigma-Aldrich). The oocytes were measured 4 days after injection. The two-electrode voltage-clamp technique was performed using a Turbo TEC-10X amplifier, without transient compensation (npi electronic). Continuous light was provided by a XBO75/2 75 W Xenon lamp (Carl Zeiss GmbH). The light passed through a 550-nm broadband filter (Optics Balzers AG) with an intensity of 10 mW/mm2. Microelectrode resistance was maintained between 0.8 and 2.0 megohm. Data acquisition and light triggering were controlled using Axon pCLAMP 9.0 software via the Digidata 1322A interface (Molecular Devices). The currents were recorded at 10 kHz and filtered to 1 kHz using built-in circuits. The photocurrent traces were baseline-corrected and filtered to 300 Hz (Gaussian low-pass cutoff filter using pCLAMP). The extracellular buffer solution was composed of 100 mM NaCl, 1 mM MgCl2, 0.1 mM CaCl2, and 5 mM MOPS (pH 7.5). For buffer solutions of pH 5 and 10, Mops was replaced by 5 mM Na-citrate or 5 mM glycine, respectively. To test cation selectivity and reduced intracellular pH, NaCl was replaced with 100 mM Na-gluconate or 100 mM NMG-Cl.

Preparation of HEK293 cells

All constructs were expressed under the control of a cytomegalovirus promotor. HEK293 cells were cultured in Dulbecco’s modified Eagle’s medium with stable glutamine (Biochrom) supplemented with 10% (v/v) fetal bovine serum (FBS Superior; Biochrom), 1 μM all-trans-retinal, and penicillin/streptomycin (100 μg/ml) (Biochrom, Berlin, Germany). Cells were seeded on poly-lysine–coated glass coverslips (1 × 105 cells/ml) and transiently transfected using the FuGENE HD Transfection Reagent (Promega) 28 to 48 hours before measurements.

Patch-clamp experiments in HEK293 cells

For patch-clamp recordings under single turnover conditions, the coding sequence of CsR was cloned into the pEGFP-N1 vector. Patch pipettes were prepared from borosilicate glass capillaries (G150F-3, Warner Instruments) using a P-1000 micropipette puller (Sutter Instruments) and fire polished for a final pipette resistance of 1.5 to 3.0 megohm.

Single fluorescent cells were identified using an Axiovert 100 inverted microscope (Carl Zeiss). Monochromatic light for continuous illumination was provided by a Polychrome V monochromator (TILL Photonics) and temporally controlled by a VS25 and VCM-D1 shutter system (Vincent Associates). For single turnover experiments, nanosecond laser pulses were generated by an Opolette Nd:YAG laser/OPO system (OPOTEK) and selected using a LS6ZM2 shutter system (Vincent Associates). Both light sources were coupled into the microscope and delivered to the sample using a 90/10 beamsplitter (Chroma). Light intensities were measured after passing through all of the optics using either a P-9710 optometer (Gigahertz-Optik) or a calibrated S470C thermal power sensor and a PM100D power and energy meter (Thorlabs) for continuous illumination or single laser pulses, respectively. Average light intensity at continuous illumination was 2.47 mW × mm−2, and average laser pulse energy adjusted by the built-in motorized variable attenuator was 11.3 ± 1.4 μJ.

During whole-cell patch-clamp recordings, membrane resistance was generally higher than 1 gigohm, and access resistance was below 10 megohm. Pipette capacity, series resistance, and cell capacity compensation were applied. Recorded signals were filtered at 2 or 100 kHz using an Axopatch 200B amplifier (Molecular Devices) and digitized using a Digidata 1440A digitizer (Molecular Devices) at a sampling rate of 250 kHz. A 140 mM NaCl agar bridge was used for the reference bath electrode. Extracellular bath solutions contained 110 mM NaCl, 1 mM KCl, 1 mM CsCl, 2 mM CaCl2, 2 mM MgCl2, and 10 mM Hepes at pHe 7.2 (with glucose added up to 310 mOsm). The intracellular pipette solution contained 110 mM NaCl, 1 mM KCl, 1 mM CsCl, 2 mM CaCl2, 2 mM MgCl2, 10 mM EGTA, and 10 mM Hepes at pHi 7.2 (glucose added up to 290 mOsm). All electrical recordings were controlled by pCLAMP software (Molecular Devices).

The recordings were baseline-corrected and time-shifted after measurements to align the rising edges of the Q-switch signals of the activating laser pulses using Clampfit 10.4 software (Molecular Devices). The recordings were then binned to 50 logarithmically spaced data points per temporal decade, normalized to peak photocurrents at +30 mV, and averaged for up to 10 individual repetitions for each cell and voltage condition with a custom MATLAB script (MathWorks). Kinetic analysis was performed in Origin 9.1 (OriginLab). To determine statistical significance, each measurement was repeated multiple times with different biological replicates in at least two independent experiments. The number of biological replicates for each measurement is provided in the figure legends.

FTIR measurements

FTIR samples were prepared with 4 μl of concentrated protein solution. The solution was pipetted onto a BaF2 window (diameter, 15 mm) and sealed with another BaF2 window. Sample thickness was ensured by placing a 3-μm spacer between the windows. Measurements were performed with a Vertex 80v FTIR spectrometer (Bruker Optics) equipped with a liquid N2-cooled mercury cadmium telluride (MCT) detector (Kolmar Technologies) operated in rapid scan mode with a data acquisition rate of 280 kHz and a spectral resolution of 2 cm−1. An optical cutoff filter at 1850 cm−1 within the beamline was used. The temperature was stabilized at 22°C. After an equilibration time of at least 60 min, the illumination was performed with a set of green-light light-emitting diodes with an emission maximum of 538 nm. The samples were illuminated until a stable equilibrium was reached. The resulting difference spectra were corrected for baseline drifts via a spline function. Singular value decomposition was performed to reduce the noise. Steady state is defined by the saturation kinetics of the most pronounced amide I band (~1660 cm−1).


Fig. S1. Current/voltage (I/V) curve comparison of full-length and truncated CsR-WT in X. laevis oocytes.

Fig. S2. Overall crystal packing of CsR-WT.

Fig. S3. Quality of the electron density of the chromophore retinal, Arg83, and Tyr14 in CsR-WT.

Fig. S4. Structural superposition of CsR-WT and BR.

Fig. S5. Close-up views of the structural arrangement around the conserved Arg in various microbial rhodopsins.

Fig. S6. pH titration of Asp86 in CsR-R83Q.

Fig. S7. Single wavelength traces of photochemical Y14F and Y14E.

Fig. S8. Current/voltage (I/V) curves of stationary photocurrents at different extracellular salt conditions.

Fig. S9. Steady-state FTIR measurements of CsR-WT and CySeR.

Fig. S10. Calculated absolute spectra of the photocycle intermediates of CsR-WT.

Table S1. Time constants of photocycle transitions in CsR-WT and CySeR.

Reference (75)


Acknowledgments: We thank A. Koch for assistance in purifying reagents and M. Reh for technical support and oocyte preparation. We are grateful to M. Weiss and the scientific staff of the BESSY-MX/Helmholtz Zentrum Berlin für Materialien und Energie at beamlines BL14.1, BL14.2, and BL14.3 operated by the Joint Berlin MX-Laboratory at the BESSY II electron storage ring (Berlin-Adlershof, Germany) and the scientific staff of the European Synchrotron Radiation Facility (ESRF, Grenoble, France) at beamlines ID23-1, ID29, ID30A-1, ID30A-3, ID30B, and ID23-2 for continuous support. Funding: R.F. was supported by the Louis-Jeantet Foundation. P.H. is a Hertie professor for Neuroscience, supported by the Hertie Foundation. This work was supported by the ESRF (beamtime BAG to P.S.). This work was funded by grants from the Deutsche Forschungsgemeinschaft (DFG; German Research Foundation): SFB1078-B1/B2 to P.H., SFB1078-B5 to F.B., SFB1078-B6 to P.S., SFB740-B6 to P.S., DFG Cluster of Excellence “Unifying Concepts in Catalysis” (Research Field E4 to P.H. and P.S.), and the DFG under Germany’s Excellence Strategy—EXC 2008/1 (UniSysCat)—390540038 (Research Unit E to P.H. and P.S.). Author contributions: R.F. and P.H. designed the study. P.S. planned and designed the structural work of the study. R.F. expressed and purified the protein and performed spectroscopic measurements. R.F. and M.S. set up the crystallization screens. M.S. and A.S. collected diffraction data. A.S. helped with data collection and structure finalization. P.S. solved the crystal structure. R.F., A.V., and P.S. planned and designed CsR mutants. A.V. performed electrophysiological recordings in X. laevis oocytes. J.V. performed electrophysiological recordings in HEK293 cells. P.F. performed FTIR measurements. F.B. supervised FTIR measurements. G.K. performed MD simulations. P.S. and P.H. supervised experiments and helped analyze the data. R.F., P.S., and P.H. wrote the manuscript, with helpful contributions from all authors. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: The atomic coordinates and structure factors have been deposited in the PDB under accession code 6GYH. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Stay Connected to Science Signaling

Editor's Blog

Navigate This Article