Non-cryogenic structure of a chloride pump provides crucial clues to temperature-dependent channel transport efficiency

Non-cryogenic protein structures determined at ambient temperature may disclose significant information about protein activity. Chloride-pumping rhodopsin (ClR) exhibits a trend to hyperactivity induced by a change in the photoreaction rate because of a gradual decrease in temperature. Here, to track the structural changes that explain the differences in CIR activity resulting from these temperature changes, we used serial femtosecond crystallography (SFX) with an X-ray free electron laser (XFEL) to determine the non-cryogenic structure of ClR at a resolution of 1.85 Å, and compared this structure with a cryogenic ClR structure obtained with synchrotron X-ray crystallography. The XFEL-derived ClR structure revealed that the all-trans retinal (ATR) region and positions of two coordinated chloride ions slightly differed from those of the synchrotron-derived structure. Moreover, the XFEL structure enabled identification of one additional water molecule forming a hydrogen bond network with a chloride ion. Analysis of the channel cavity and a difference distance matrix plot (DDMP) clearly revealed additional structural differences. B-factor information obtained from the non-cryogenic structure supported a motility change on the residual main and side chains as well as of chloride and water molecules because of temperature effects. Our results indicate that non-cryogenic structures and time-resolved XFEL experiments could contribute to a better understanding of the chloride-pumping mechanism of ClR and other ion pumps.

Non-cryogenic protein structures determined at ambient temperature may disclose significant information about protein activity. Chloride-pumping rhodopsin (ClR) exhibits a trend to hyperactivity induced by a change in the photoreaction rate because of a gradual decrease in temperature. Here, to track the structural changes that explain the differences in CIR activity resulting from these temperature changes, we used serial femtosecond crystallography (SFX) with an X-ray free electron laser (XFEL) to determine the non-cryogenic structure of ClR at a resolution of 1.85 Å, and compared this structure with a cryogenic ClR structure obtained with synchrotron X-ray crystallography. The XFEL-derived ClR structure revealed that the alltrans retinal (ATR) region and positions of two coordinated chloride ions slightly differed from those of the synchrotronderived structure. Moreover, the XFEL structure enabled identification of one additional water molecule forming a hydrogen bond network with a chloride ion. Analysis of the channel cavity and a difference distance matrix plot (DDMP) clearly revealed additional structural differences. B-factor information obtained from the non-cryogenic structure supported a motility change on the residual main and side chains as well as of chloride and water molecules because of temperature effects. Our results indicate that non-cryogenic structures and time-resolved XFEL experiments could contribute to a better understanding of the chloride-pumping mechanism of ClR and other ion pumps.
The crystallization and diffraction technologies for membrane proteins have become more advanced, resulting in the accumulation of membrane protein structure information (1)(2)(3)(4)(5)(6)(7). In particular, crystallization of membrane proteins in the lipidic cubic phase (LCP) 4 is a critical method for solving the structures of membrane proteins such as G protein-coupled receptors (GPCRs) (8,9). However, it remains difficult to form appropriately sized, well-ordered crystals suitable for obtaining a complete diffraction dataset at synchrotron facilities. Recently, with the development of LCP injectors (10), membrane protein structure determination has become possible from microcrystals at room temperature using serial femtosecond crystallography (SFX) with X-ray free electron lasers (XFELs) developed by Chapman and colleagues (11)(12)(13)(14). Structural determination of membrane proteins using XFELs showed that XFEL-derived structures are comparable to synchrotron-derived structures (15)(16)(17)(18)(19). However, understanding of the structural differences between structures determined using SFX at XFEL facilities at room temperature and those determined using the conventional macromolecular crystallography method (MX) at synchrotrons in a cryogenic state remains limited, along with the implications for molecular functions. Rhodopsin is an excellent target that could serve as a good model system for the technical development of structural methods such as time-resolved SFX for examining light-activated membrane proteins grown in an LCP environment (20 -26). In microbial light-driven ion-pumping rhodopsin, the chromophore all-trans retinal (ATR) of rhodopsin binds covalently to a lysine residue through a Schiff base linkage. Absorption of light induces a conformational change in the retinal from all-trans to 13-cis, leading to overall structural changes in the protein, thereby activating ion transport and a photocycle that traverses several key intermediate states coming back to the dark state (21,22,(27)(28)(29). The motif of the protonated Schiff base (PSB) changes according to the unique residue of the proton acceptor and proton donor in rhodopsin, which induces a difference in the ion transport pathway and pumping mechanism. Several studies have reported protonand chloride-pumping rhodopsins (ClRs) with various PSB motifs, including bacteriorhodopsin (bR) with DTD motifs, thermophilic bacteria rhodopsin (TR) with DTE motifs, and halorhodopsin (HR) with TSA motifs (23, 30 -33).
In recent years, XFEL-derived structures for proton-pumping rhodopsins have been reported and compared with those determined at synchrotrons, and the proton-transporting mechanism was clearly identified from the nanosecond to millisecond time scale with retinal photoisomerization using timeresolved SFX experiments (21,22,34). However, the structural information of ClRs close to the native condition (room temperature) and the correct ion-transporting mechanism using time-resolved SFX methods remain unknown. Proton-and chloride-pumping rhodopsin have different directions of ion uptake and exhibit completely different photocycle mechanisms. In particular, most ClRs lack the photochemical intermediate "M" state in the photocycle (33,35). Therefore, it is very important to clarify the three-dimensional structure and accurate ion transport pathway of ClR at room temperature using XFEL.
Recently, a ClR with a novel PSB (NTQ) motif exhibiting unique ion-transporting activities and pathways has been reported in the marine bacterium Nonlabens marina S1-08(T) (26, 36 -38). Growth of this strain is modulated by environmental temperature, with a growth temperature range of 10 -30°C; the optimal temperature for growth is 20 -25°C, and no growth occurs below 5°C or above 37°C (39). In early studies investigating the photoelectric signal of bacteriorhodopsin, it had been suggested that the rate constant could be changed according to the pH or temperature change resulting in the modulation of rhodopsin's functions (40). In the present study, we also found that the channel functions of ClR are modulated by temperature in a similar manner. Therefore, it would be of interest to know the correlation between temperature and function of ClR, and in this respect, ClR could serve as a model protein to obtain structural dependence on temperature.
Here, we report the first room temperature crystal structure of ClR containing the NTQ motif in the chloride ion transfer pathway using XFEL. The structure of ClR at room temperature, determined at a resolution of 1.85 Å, reveals tremendous structural details. We analyzed structural differences by comparing two structures at both room temperature and the cryogenic temperature. Although the overall structure determined by XFEL was comparable to that determined by synchrotron X-ray crystallography, the chloride ion-binding sites (one at the active center of the pump and the other near the loop on the cytoplasmic side) were found to be in distinct positions showing different hydrogen bond networks. Interestingly, one additional water molecule forming a hydrogen bond network with a chloride ion (Cl Ϫ -I) near PSB was clearly observed in the XFELderived structure, but not in the structure determined at the cryogenic temperature.
Our findings will be essential in gaining a better understanding of the structural differences and mechanical properties attributed to temperature during data collection by both XFEL and synchrotron analyses, providing hints into the temperature dependence of hydrogen bonding in ClR and its correlation with chloride-pumping function. The room temperature structure of ClR thus sets the foundation for further time-resolved SFX studies on chloride ion transportation pathways and their underlying molecular mechanism.

Hyperactivity and anion-binding affinity of CIR for anion transport at low temperatures
The Gram-negative, orange-pigmented, and rod-shaped bacteria Nonlabens marina S1-08(T) were isolated from seawater of the western north Pacific Ocean (30°11Ј N, 145°05Ј E; depth 0 m). It is well-known that Nonlabens marina S1-08(T) ceases cell growth at temperatures below 10°C, unlike other marine bacteria (39), and the optimal growth temperature is between 20°C and 25°C (Table S1). Thus, to determine how temperature affects the anion transport activity of the ClR derived from Nonlabens marina S1-08(T), we measured lightinduced pH changes for different monovalent salts, sodium chloride, and sodium bromide at five different temperatures, Ϫ20°C, Ϫ10°C, 4°C, 10°C, and 25°C. As the temperature gradually decreased, the chloride ion transport rate increased and was almost 2-fold faster at Ϫ20°C than at 25°C in the absence of a protonophore, and the same trend was not observed for bromide ions (Fig. 1, A and B). Likewise, in the presence of the protonophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP), the pH change was much more drastic during chloride and bromide ion transportation at lower temperatures. The enhanced activity observed in the presence of CCCP completely disappeared following the addition of tetraphenylphosphonium ion (TPP ϩ ), a hydrophobic cationic reagent that disrupts membrane potential. This implied that the physiological function of ClR in pumping activity can be sensitively modulated by temperature changes. In this regard, abnormal functions, such as ClR hyperactivity involved in phototransduction biochemical reactions, could also play a role in cell death or suspended growth of Nonlabens marina S1-08(T) by altering the electrochemical balance of the bacterial cell.
The binding affinity of ClR for different anions was measured by UV-visible absorption spectrometry to determine whether the hyperactivity was caused by a change in the binding affinity with the anion because of the temperature difference. The maximum wavelength with and without 1 M sodium chloride was 533 nm and 555 nm, at both room temperature (25°C) and 4°C, respectively (Fig. 2, A and B). Likewise, the maximum wavelength with and without 1 M sodium bromide was 537 nm and 555 nm at room temperature (25°C), and 533 nm and 555 nm at 4°C, respectively (Fig. S1, A and B). A similar blue shift was observed with bromide and chloride ions in the maximum absorption peak upon titration at both 25°C and 4°C. Anion titration allowed for the determination of dissociation constants (K d ) from the absorbance changes at 581 nm (Fig. 2, C and D and Fig. S1, C and D). The apparent K d values were estimated using the Hill equation and are summarized in Table  S2. Surprisingly, the binding affinity of chloride and bromide

Non-cryogenic structure of ClR derived from XFEL
ions were approximately twice as strong at 4°C than at room temperature (Fig. 2E). These binding-affinity results were consistent with the anion transport activity results (Table S2).

XFEL-and synchrotron-based structures demonstrate subtle differences in two chloride positions and hydrogen-bonding networks
The ClR structures were determined by XFEL and synchrotron analysis to investigate structural changes that can explain functional differences of ClR at different temperatures. Crystal structures were collected at a resolution of 1.85 Å and 1.75 Å at room temperature (294 K) by XFEL and at cryogenic temperature (93 K) by the synchrotron, respectively. Overall, the two structures showed good agreement with no significant changes in the overall topology. The root mean square deviation (RMSD) of the 253 C␣ atoms of a single molecule in the asymmetric unit was calculated to be 0.128 Å with similar overall residual B-factors (Å 2 ) except for some structural points (Fig.  S2). Specifically, helices A and B and part of helix G showed lower B-factor values, especially, Lys 235 , which formed a cova- ClR-expressing E. coli cells were light-induced by a green-light laser-on (540 nm) for 5 min and were dark-adapted before and after green-light laser-on for 5 min each.

Non-cryogenic structure of ClR derived from XFEL
lent bond with retinal and exhibited the lowest B-factor value in the ClR structure at room temperature. Moreover, we also identified several important structural differences through a detailed comparison of these two structures.
First, we investigated the near-retinal region, which is the main trigger for ion-transporting activity. Interior residues surrounding the retinal chromophore were revealed with highquality electron density maps, especially with a clear retinal ␤-ionone ring (Fig. 3, A and B). The covalent bonding angle between the retinal C15 and Lys 235 residue in the two structures differed by about 3° (Fig. 3, C and D). This difference could be related to the dynamic properties that cause structural changes in response to temperature changes. Moreover, the chloride ion (Cl Ϫ -I), located in the active core of the XFEL-derived structure, moved toward the PSB region, resulting in a shorter distance between Cl Ϫ -I and the PSB by ϳ0.3 Å compared with that of the synchrotron-derived structure (Fig. 3, C and D). Ion pumps such as ClR stabilize chloride and water molecules around the PSB region by interacting with proton acceptors or donor residues for ion transport, and the distance between them significantly differed between the two structures ( Fig. S3, A, B, D, and E). Disruption of hydrogen bonds is essentially a  Table S2.

Non-cryogenic structure of ClR derived from XFEL
temperature-dependent process (41). The rearrangement of water molecules and chloride ions leads to a change in the hydrogen network. Important information gained from the XFEL structure was the discovery of new hydrogen bonding networks with an additional water molecule (W614), having B-factor of 26.4 Å 2 and near Cl Ϫ -I (Fig. 3, A and B). ClR was previously thought to have unique hydrogen bond networks with two water molecules distinct from other chloride pumps such as HR, which have three well-resolved water molecules mediating the hydrogen bond network between Cl Ϫ -I-PSB and Asp 239 -Arg 108 (in HR from Halobacterium salinarum) (42). However, the ClR structure derived from XFEL confirms that it also has three well-resolved water molecules, W507, W508, and W614, and forms similar hydrogen bonding networks between Cl Ϫ -I-PSB and Asp 231 -Arg 95 , like other chloride pumps (Fig. S3A). The B-factors of W507 and Cl Ϫ -I were ϳ16.3 Å 2 and 20.8 Å 2 in the XFEL-derived ClR structure, compared with the corresponding values of 24.4 Å 2 and 29.2 Å 2 in the synchrotron-derived structure, indicating more stable dynamics in this region.
As a result, the XFEL-derived ClR structure determined at room temperature showed much stronger water-mediated hydrogen bonds and considerably more stable dynamic properties near the PSB and Cl Ϫ -I. In particular, chloride ions exhibited increased interactions with W507 and the PSB in the XFEL-derived structure. The hydrogen bond between Arg 95 and the water molecule W508 was similar, whereas the hydrogen bond between Asp 231 and the water molecule W507 was very different because of the presence of the additional water molecule W614 (Fig. S3, A, B, D, and E).
Second, structural rearrangement was detected around the chloride ion (Cl Ϫ -II) in the cytoplasmic region (Fig. 3D). The location of Cl Ϫ -II changed by ϳ0.5 Å, resulting in a longer distance between Cl Ϫ -II and Lys 46 amide nitrogen in the XFEL-derived structure by ϳ0.2 Å compared with that in the synchrotron-derived structure. The B-factors of Cl Ϫ -II in the two structures were quite comparable: Both are about 40 Å 2 .
Additionally, one water molecule in the cytoplasmic region, W515, showed different positions and B-factor values between the two structures. W515 forms a hydrogen bond between Lys 46 and Ile 112 by 3.0 and 2.7 Å in the XFELderived structure, although it shows a much weaker hydrogen bond between W515 and Lys 46 in the synchrotron-derived structure (Fig. S3, C and F). In the XFEL-derived structure, the position of W515 shifted by 1.4 Å toward the backbone carbonyl oxygen of Lys 46 and the B-factor was increased from 27.8 Å 2 to 47.1 Å 2 (Fig. 3, E and F). The information of hydrogen bonding distance and B-factors from the structure determined at room temperature using the XFEL method provides a structural explanation for the abnormal activities of ClR at low temperature.

Channel cavities of ClR
For more in-depth structural comparisons, the volume sizes and side chain orientations of the participating residues were investigated for four core cavities buried in the seven transmembrane domains of ClRs. The volume of individual cavities showed quantitative dependence on the number of atoms (or residues) and water lining the cavity. The cavity volumes (Å 3 ) in the two ClR structures from the XFEL and the synchrotron are marked in Fig. 4, A and B. Overall, cavities 1 and 3 in the XFELderived ClR structure were slightly larger (87.79 and 2.76 Å 3 versus 86.05 and 2.63 Å 3 ), whereas cavities 2 and 4 were smaller than that in the synchrotron-derived structure (11.02 and 0.17 Å 3 versus 13.90 and 0.29 Å 3 ).
In addition, although the structures were generally similar, the hydrogen bonds between the residues forming the cavity and the surrounding water differed. In cavity 1, Cl Ϫ -I and five water molecules form a hydrogen bond network, and the side chain orientation of participating residues were very similar in the two structures, with some notable variations in the distances between certain residues, as explained in the previous section ( Fig. 4C and Fig. S3). In cavity 2, different side chain packing for Ile 64 , Lys 65 , Arg 95 , Asn 98 , Trp 99 , Thr 102 , Asp 231 , and Lys 235 was observed in the hydrogen bond network, with Cl Ϫ -I

Non-cryogenic structure of ClR derived from XFEL
and water molecules depending on the presence of an additional water molecule, W614 (Fig. 4D).
Interestingly, the water molecule W509 did not share the same position, and its B-factor value differed between the two resolved structures (34.2 versus 41.4 Å 2 ). In cavity 3, Asn 3 and Asn 92 showed very different side chain orientations, and the position of one nearby water molecule, W501, was not observed in the XFEL-derived structure ( Fig. 4E and Fig. S4, A and B). In cavity 4, Met 58 , Lys 106 , Gln 109 , and Lys 235 showed slight differences with respect to side chain orientation (Fig. 4F). Thus, the high-resolution ClR structures measured by both XFEL and synchrotron allow accurate comparison of the cavity and side chain orientations.

Temperature-dependent conformational change of ClR
Because the overall conformational changes were quite subtle, we constructed a difference distance matrix plot (DDMP) to detect regional structural rearrangements that might be caused by the temperature differences (Fig. 5) (55). DDMP analysis demonstrates dramatic internal distance differences in several regions based on the distances among C␣ atoms. According to the DDMP results, structural perturbation resulting from the

Non-cryogenic structure of ClR derived from XFEL
cryo-cooling conditions occurred during data collection. Specifically, major structural perturbations occurred in the B-C loop, the C-D loop, a part of helix C in the extracellular side, a part of helices E and G in the intracellular side, and the C-terminal helix (Fig. 5, A and B). The poly-chains Val 79 -Leu 85 and Gln 86 -Thr 89 , located in the B-C loop, moved in opposite directions, and this may have affected the -stacking interaction that formed the 3motif of ClR (Fig. 5C). The extracellular section of helix C moved toward the interior of the protein, with the same orientation as poly-chain Val 79 -Leu 85 in the B-C loop (Fig.  5C, upper panel). The intracellular portion of helix E is another region that showed the widest conformational change (Fig. 5, A  and B) (55).
Interestingly, C-terminal helix movement was clearly observed in the direction opposite to helix G movement (Fig. 5C,  lower panel). These motions were reported to be important for maintaining the function and stability of ClR in previous studies (42) and are similar to the related motion in the function of bR (34). As a result, the structural orientation of the C-D loop, a part of helix E, and C-terminal helix moved toward the exterior of a protein, whereas helix G moved in the opposite direction (Fig. 5C, lower panel).
The detailed structural information of the XFEL-derived ClR provided the basis for revealing the ion channel mechanism of the light-induced photocycle of ClR at room temperature.

Discussion
A number of functionally important membrane protein structures have been solved using crystals grown with the LCP method. However, several problems remain unsolved in order for membrane protein crystallization to obtain crystals suitable for synchrotron X-ray diffractions. Recently, SFX using XFELs has been developed and has emerged as a powerful method for the structural determination of macromolecules, especially membrane proteins such as G protein-coupled receptors. The experimental environment for XFEL is different from that used for synchrotron-based crystallography which requires cryogenic cooling protection, and this difference could influence the interaction of ligands or ions that are related to membrane protein function. The structure and dynamics for some membrane proteins depend on temperature and pressure.
Studies have indicated that XFEL-derived structures do not suffer from perturbations because of cryogenic cooling or radiation damage from synchrotrons (34,43). Here, we directly compared structures of ClR at cryogenic and room temperatures and revealed subtle structural differences that could explain the hyperactivity of ClR at low temperatures. The DDMP analysis depicted regional conformation changes for temperature-sensitive residues. The high-resolution data (1.85 Å and 1.75 Å of the structures from the XFEL and synchrotron, respectively) allowed us to accurately compare hydrogen bonding networks mediated by chloride ions and water molecules through direct structure comparison. The overall unit cell volume of microcrystals in the XFEL experiment was larger (by about 3.2%) than that of the crystals used in the synchrotron crystallography experiment, indicating that the cryogenic cooling slightly dehydrates the crystals and increases the crystal packing. In a previous bR study, the electron density map around the C2-C4 atoms of the retinal ␤-ionone ring was found to be partially depleted in the XFEL structure, indicating the possibility of multiple conformers at ambient temperature (34). In the present study, the ClR structure derived from XFEL showed a very clear electron density map of the retinal ␤-ionone ring, indicating that the structure was thermodynamically stable at physiological temperature. These findings suggest that

Non-cryogenic structure of ClR derived from XFEL
bR and ClR may have different dynamic motions of the retinal, leading to a cis-trans transition even at the same temperature conditions (Fig. S4, C and D).
Here, we also identified the hydrogen bond networks mediated by the additional water molecule, W614, near the Cl Ϫ -I and PSB regions. In general, the energy associated with hydrogen bonds is 6 -30 kJ/mol, and even a slight change in the angle and distance between the relevant atoms causes a drastic change in the energy barrier (44). For this reason, hydrogen bonding is a key element in maintaining the binding strength and ion specificity in a transporter protein, which in turn affects the ion transport efficiency. These differences were also observed at the Cl Ϫ -II position in the cytoplasmic region. The weak hydrogen-bonding network near Cl Ϫ -I in ClR would eventually lead to rapid chloride ion diffusion into the cytoplasm direction at a reduced energy cost. In this respect, the non-cryogenic structure determined at ambient temperature is helpful to better understand protein functions at the optimum temperature of their activity.
Considering the structural changes observed in the cavity and DDMP analyses, temperature-dependent structural movement of ClR can be inferred. At ambient temperatures, the ClR structure has a relatively wider intracellular portion and narrower extracellular portion compared with its structure at low temperatures. In this study, we concluded that the ClR structure derived from the XFEL method provided a structural foundation in understanding its chloride ion-pumping activity at physiological conditions.

Microcrystal preparation for XFEL and synchrotron analysis
ClR microcrystals appeared in 30% PEG dimethyl ether 500, 0.1 M sodium chloride, and 0.1 M MES (pH 6.0) buffer at a final concentration of 50 mg/ml for conventional synchrotron experiments (42). For the LCP-XFEL approach, a high density of microcrystals is required for efficient data collection. We followed optimized protocols (45) to obtain samples for a complete dataset collection. The purified protein was mixed with monoolein (1-oleoyl-rac-glycerol) at a 1:1.5 molar ratio (w/w) using a syringe lipid mixer (Hamilton, Reno, NV). After formation of a clear lipidic cubic phase, ϳ5 l of the protein-laden LCP sample was injected using a 100-l syringe filled with 55-60 l of precipitant solution. After sealing the needle stopper, the syringe was wrapped in a moist tissue to maintain the humidity and placed in a 20°C incubator for 1 week (Fig. S5B). Five additional syringes were prepared to produce a large volume of the LCP crystal sample. Microcrystals of ClR were grown in the syringes, and red-colored crystals were identified using a microscope. For the synchrotron experiment, ClR microcrystals were grown on LCP plates (Fig. S5C).

Experimental setup at Linac Coherent Light Source, and data collection from XFEL and synchrotron
The experiment was performed at the Coherent X-ray Imaging (CXI) end station (46) at the Linac Coherent Light Source (LCLS) of the SLAC National Accelerator Laboratory. Prior to LCP-XFEL data collection, ClR crystals grown in syringes were incubated at 25°C for 1 h. Approximately 40 l of the LCP sample was transferred into a new syringe after removing the precipitant solution, and ϳ10 l of 9.9 MAG was applied to the LCP sample and homogenized. The sample was transferred into an LCP injector through an LCP loading needle. Data were collected at 120 Hz using a photon energy of 9.5 keV ( ϭ 1.3 Å) and a pulse duration of ϳ60 fs. Because of evaporative cooling, the sample temperature at the X-ray interaction region (ϳ100 m downstream of the nozzle exit) was estimated to be slightly below room temperature. The samples were injected using an LCP injector with a 75-m diameter nozzle at an average flow rate of 2 l/min. A Cornell-SLAC Pixel Array Detector (CSPAD) was used to collect the diffraction data. The panels of the CSPAD were tiled to form a two-dimensional detector that was placed at 77.5 mm from the interaction plane, allowing for diffraction data collection up to a 1.33-Å resolution at the given wavelength. About 420,465 diffraction patterns were collected in ϳ1 h from 120 l of the sample. The Cheetah program (47) was used to preprocess and filter the raw data to select diffraction patterns (actual "hits" of crystals from a liquid background). A pattern was classified as a "hit" if it contained at least 20 peaks with a signal-to-noise ratio larger than 6.0. A total of 22,094 patterns were identified as hits for an overall hit rate of 5.26%. The CrystFEL software suite (48) was used to index the diffraction patterns and merge the intensity to form the threedimensional diffraction volume. After refinement of the experimental conditions, including detector geometry, 10,805 patterns were indexed using the indexamajig module of the CrystFEL suite, and the unit cell parameters were determined for data merging. Finally, all reflections, including those predicted based on the observed peaks of indexed patterns, were merged into a dataset with a Monte Carlo approach implemented in CrystFEL. For conventional synchrotron data, we collected diffraction data from ClR crystals on the BL17A beamline at the Photon Factory (Tsukuba, Japan) using a 0.03 ϫ 0.01 mm microbeam. The diffraction data were processed with XDS software (49). We determined the cryogenic structure at Photon factory as described in a previous report (42). All parameters for data collection, model refinement, and statisti-Non-cryogenic structure of ClR derived from XFEL cal analysis of both the XFEL and synchrotron data are summarized in Table 1.

Structure determination and refinement
Crystals grew in space group C2, with a ϭ 103.36 Å, b ϭ 50.09 Å, c ϭ 69.40 Å, ␣ ϭ 90.00°, ␤ ϭ 109.65°, and ␥ ϭ 90.00°for the XFEL-derived structure, and a ϭ 102.76 Å, b ϭ 49.40 Å, c ϭ 69.33 Å, ␣ ϭ 90.00°, ␤ ϭ 109.85°, and ␥ ϭ 90.00°for the synchrotron-derived structure, with one molecule in the asymmetric unit. Initial models of the near-isomorphous crystal data were obtained by molecular replacement using PHASER (50) with a previously solved structure as a search model (PDB ID: 5G28). The solutions of ClR were successfully obtained with a final translation function Z-score of 19.0. The models were refined with the simulated annealing protocol using bulk-solvent correction with data between a resolution of 50.0 Å and 3.0 Å. Electron density was interpreted and traced using COOT (51), and the model was refined with PHENIX (52) and auto-BUSTER (53). After several rounds of alternate refinement by autoBUSTER in the absence and presence of ATR, the structure of ClR containing ATR was further refined, resulting in R work / R free ϭ 0.24/0.28 over a resolution range of 18.52-1.85 Å. Likewise, the synchrotron-derived structure was refined, resulting in R work /R free ϭ 0.19/0.22 over a resolution range of 23.63-1.75 Å. Solvent molecules were placed at positions where spherical electron density peaks were found above 1.3 in the ͉2F o ϪDF c ͉ map and above 3.0 in the ͉F o -F c ͉ map, and where stereochemically reasonable hydrogen bonds could form. Validation of the final model was carried out using PROCHECK (54). During structure refinement, we tried to reduce the resolution gap and the artifact by applying the same cutoff value between the XFEL-and synchrotron-derived structure, especially for selection of chloride ions and water molecules.

Data availability
Coordinates and structure factors have been deposited in the Protein Data Bank under accession codes 5ZTL (XFEL-derived ClR structure) and 5ZTK (synchrotron-derived ClR structure). Remaining data are available from the corresponding authors on reasonable request.

Difference distance matrix plot analysis
The pairwise distance map was computed between all possible amino acid residue pairs in a three-dimensional structure and represented with a two-dimensional matrix. We used the DDMP (Center for Structural Biology, Yale University, New Haven, CT) program to analyze the protein contact map. In the difference distance map, the changes in distance between C␣-C␣ atoms were reflected by the brightness of the colors, and color saturation levels were set to represent C␣ shifts of 0.5 Å closer (red) and 0.5 Å away (blue) (55).

Cavity calculation
The PDB coordinates of ClR structures were submitted to the CASTp Web server (http://sts.bioe.uic.edu/castp/ calculation.html) 5 (56) for putative substrate-binding pocket and retinal cavity calculation, and identification as a probe radius of 1.4 Å. Each pocket was assigned a unique identification number, roughly corresponding in order of decreasing volume. After selection of the specific pocket, graphical representations of ClR were made using PyMol.

Measurement of pumping activities of ClR
E. coli BL21-CodonPlus (DE3) cells, expressing ClR, were incubated at 37°C in YT media supplemented with 100 mg/ml ampicillin. When the A 600 nm was greater than 1.0, the expression of ClR was induced by adding 1 mM isopropyl-␤-D-thiogalactopyranoside and 50 M ATR. After additional cultivation for 4 h at 37°C, the cells were collected by centrifugation at 4000 ϫ g for 5 min, washed three times with 100 mM NaCl or NaBr including 50% glycerol, and suspended in the desired solvents for measurement by adjusting the A 600 nm value to 8.0. Cell suspensions were placed in the dark for more than 2 h, under control of desired temperature from Ϫ20°C to 25°C, and then, proton ion flux changes in the buffer were measured using a pH electrode F-72G (Horiba, Kyoto, Japan) at five different temperatures, Ϫ20°C, Ϫ10°C, 4°C, 10°C, and 25°C, respectively. The light source used was a 200 milliwatt 520 nm Xeon lamp (Elpisbio, Seoul, South Korea).

UV-visible spectroscopy
ClR was purified and buffer exchange was conducted using 10 mM MOPS (pH 6.5) and 0.05% DDM on a Superdex 200 size-exclusion column (GE Healthcare). The absorption spectra were scanned for the protein samples using a V-650 spectrophotometer (Jasco, Easton, MD) at two different temperatures, 4°C and 25°C. Anion titration experiments were performed by adding chloride and bromide ions and standard deviations were calculated from triplicate experiments. Anionbinding affinities were calculated from absorption changes at a wavelength of 581 nm using the Hill equation. The binding affinities are listed in Table S2.