Pumping mechanism of NM-R3, a light-driven bacterial chloride importer in the rhodopsin family

See allHide authors and affiliations

Science Advances  07 Feb 2020:
Vol. 6, no. 6, eaay2042
DOI: 10.1126/sciadv.aay2042


A newly identified microbial rhodopsin, NM-R3, from the marine flavobacterium Nonlabens marinus, was recently shown to drive chloride ion uptake, extending our understanding of the diversity of mechanisms for biological energy conversion. To clarify the mechanism underlying its function, we characterized the crystal structures of NM-R3 in both the dark state and early intermediate photoexcited states produced by laser pulses of different intensities and temperatures. The displacement of chloride ions at five different locations in the model reflected the detailed anion-conduction pathway, and the activity-related key residues—Cys105, Ser60, Gln224, and Phe90—were identified by mutation assays and spectroscopy. Comparisons with other proteins, including a closely related outward sodium ion pump, revealed key motifs and provided structural insights into light-driven ion transport across membranes by the NQ subfamily of rhodopsins. Unexpectedly, the response of the retinal in NM-R3 to photostimulation appears to be substantially different from that seen in bacteriorhodopsin.


Rhodopsins are integral membrane proteins holding a retinal chromophore within a hydrophobic cavity, where they form a protonated Schiff base (PSB) with an essential lysine residue in the seventh (G) helix (Fig. 1A) (1). Light-induced isomerization of the chromophore allows these proteins to act as photosensors or light-driven pumps. The structures of bacteriorhodopsin (bR), which pumps protons out, and halorhodopsin (hR), which pumps chloride ions in, are established (24). Two representative structures of hRs from Halobacterium salinarum (HsHR) and Natronomonas pharaonis (NpHR) show common structural features, including a homotrimeric conformation with a large hydrophobic cap, formed by the extracellular loop between helices B and C, which covers most of the extracellular surface of the protein (4). The Schiff base linking the protein to the chromophore is protonated during the dark state when the receptor is at rest. In bR, the photocycle is initiated by the photoisomerization of the all-trans retinal (ATR), moving a proton from the Schiff base to the acceptor side chain of Asp85 (5). The Schiff base is reprotonated by Asp96, leading to proton pumping. These two aspartate residues, together with Thr89, form a “DTD” (Asp, Thr, and Asp) motif in the C helix, which is characteristic of proton transporters in the rhodopsin family, although the threonine is not functionally required (5). In hR, a TSA (Thr, Ser, and Ala) motif replaces the DTD motif, and the specificity of the protein for either chloride ions or protons is strongly linked to whichever of the two motifs is present. The replacement of Asp85 in bR with threonine yields a chloride-transporting mutant (6, 7).

Fig. 1 Structural characterization of NM-R3 retinal chromophores after light excitation.

(A) Phylogenetic tree showing relationships among rhodopsins. (B) Difference absorption spectra showing the increasing red shift of the visible band near 600 nm with laser power. Spectra were obtained from the difference between the absorption spectra for the dark and light states, shown in fig. S1. (C) Difference electron density maps showing changes triggered near the retinal. |Fobs|light − |Fobs|dark difference Fourier electron density maps contoured at ±3.5σ. Positive and negative values are shown in blue and red, respectively.

A new subfamily called NQ (Asn and Gln) rhodopsins was first identified in 2000. NQ rhodopsins are found in marine bacteria and archaea, with Asn and Gln in place of the acceptor and donor aspartate residues of bR, respectively. Nonlabens marinus rhodopsin-3 (NM-R3) is an inward chloride pump in the NQ rhodopsin subfamily (7, 8). It shares 35% sequence identity with Krokinobacter eikastus rhodopsin 2 (KR2), a light-driven sodium exporter (913). Two crystal structures of NM-R3 in the dark state have been published, and two chloride ions are found; Cl-1 lies close to the Schiff base, and Cl-2 lies in an extracellular surface pocket (11, 12). The network of bonds between chloride ions and highly ordered water molecules in NM-R3 is different from that of other chloride-pumping rhodopsins (4, 14, 15). Overlaying NM-R3 and NpHR with secondary-structure matching (SSM) (16) shows that the proteins share 17% sequence identity, and a core region of 202 Cα atoms gives a root mean square deviation (RMSD) of 1.79 Å. The lack of conservation among functionally important residues, such as Thr218 of NpHR (17), equivalent to Met197 in NM-R3, suggests that the mechanisms of these proteins are notably different. The B-C loop of NM-R3 reveals a different conformation than that in other rhodopsin subfamilies and includes three tyrosine residues (Tyr76, Tyr78, and Tyr83), which form part of a cluster of aromatic side chains on the extracellular face of the protein. Moreover, three nearby aromatic residues (Phe15, Trp72, and Tyr83) stack together to form the “3 omega motif,” which is highly conserved among the NQ family of ion pumps but not found in bR or hR. Detailed spectroscopic studies of NM-R3 have suggested that the photocycle involves five or six distinct intermediates, as in bR (18).

Low-temperature trapping has been used to obtain a crystallographic model of an early intermediate in the bR photocycle (19). More recently, the x-ray free electron laser (XFEL) was used to obtain a time series of models in femtosecond to millisecond time scales (20, 21). Although the retinal-binding pocket is highly conserved, Asp85 in bR is replaced with Asn98 in NM-R3, which lies approximately 3 Å further from the retinal, creating the Cl-1 binding site, where a chloride ion can interact directly with the PSB. No other internal chloride ion binding sites have been suggested by spectroscopy (22). In the ground state of bR, Arg82 lies 3.8 Å from Asp212 (23); in NM-R3, the equivalent side chains of Arg95 and Asp231 form a hydrogen bond 2.7 Å long. NM-R3 also has Ser234 in place of Ala215, which forms a hydrogen bond with the conserved side chain of Tyr204 alongside the chromophore, suggesting that the retinal-binding pocket of NM-R3 is substantially less flexible than that of bR. Since the light-induced structural changes and ion conduction pathway of NM-R3 remain much less well characterized than those of bR, we studied the photoexcitation of crystalline NM-R3, at different temperatures and using different excitation intensities, to elucidate the structural mechanism of the chloride pump and help understand the differences between the various ion pumps in the rhodopsin family.


Structural determination of early intermediate photoexcited states of NM-R3 by laser pulses of different intensities and temperatures

Using single-crystal microspectrophotometry, we exposed single crystals of NM-R3 at 95 K to light from a continuous wave (CW) green laser (532 nm) with 3% (CW95K-ND3), 10% (CW95K-ND10), or 30% (CW95K-ND30) neutral density (ND) filters; other crystals at 95 K (pulse-95K) or 140 K (pulse-140K) were irradiated with pulsed laser light. For each crystal, a stable light-induced red shift of the visible absorption band near 550 nm was observed, and the spectral changes increased as the laser intensity increased (Fig. 1B and fig. S1). The changes were rapid and fully reversible. Moreover, similar rate constants were observed for NM-R3 in solution, indicating that the crystal packing does not prevent rapid stimulation by light or reversion to the dark state. In both the crystal and solution conditions, accumulation and relaxation kinetics exhibited an initial overshoot and undershoot, respectively.

X-ray datasets were collected in the dark and after irradiation (Fig. 1C, fig. S2, and table S1). Comparing the 263 Cα atoms of residues 2 to 264, the highest RMSD between dark-state models was 0.074 Å. A comparison of the light- and dark-state models for each crystal revealed a maximum Cα RMSD of only 0.064 Å for CW95K-ND10, with a maximum displacement of 0.14 Å at Lys119, at the protein surface and remote from crystal contacts. Even with the most intense photoexcitation (CW95K-ND30), the model showed small atomic shifts, and Cl-1 moved only by 0.2 Å. However, difference maps could be calculated with low background noise. The R-merge between the dark and photoexcited datasets was around 5% throughout, for every crystal, and showed no increase for images taken toward the end of data collection.

Similar to the spectra observed by microspectrophotometry with single crystals, electron density difference maps (light minus dark for each crystal) showed greater and more widespread changes with higher temperatures or laser intensities (Fig. 1C and fig. S3A). At 140 K, although changes were found throughout the protein, the local structural change near the retinal was smaller than that in CW95K-ND30, suggesting that photoexcitation effects had diffused throughout the entire structure. The temperature factors of the backbone were similar in all datasets and were unaffected by laser exposure (fig. S3B) so that features of the difference density maps obtained in this study cannot arise from simple thermal effects during the experiments.

Detailed structural differences of NM-R3 by photoexcitation

Figure 2 shows a close-up series of the difference density maps obtained with continuous laser excitation protocols, focusing on the Schiff base and adjacent residues. Contoured at 5σ, the CW95K-ND3 difference map showed features close to the PSB, and a few other features appeared at 3.5σ (Fig. 2A). Positive and negative density indicated strongly concerted movements of the Schiff base, by about 0.2 Å, toward Thr102 and away from Asp231. Strong density appears on either side of Nε and Cδ of the Lys235 side-chain atoms but much less so for the retinal atoms. Despite the strong side-chain movement, the main-chain atoms of Lys235 are unmoved. The retinal itself overlays closely with that of bR, with the exception of the C3 pucker, but there is no direct evidence for retinal isomerization in the NM-R3 electron density maps (at the level of excitation achieved). Unlike bR, no difference density is seen near the retinal C13═C14 bond (Fig. 1C and fig. S4). Paired positive and negative difference density features appear over the retinal C20 methyl group, which undergoes a structural shift comparable with that in bR, but the retinal maintains the all-trans conformation (21). Separate refinement of the model against data collected from the photoexcited crystal, and calculation of an FlightFdark difference map using independent phases for the light and dark states, confirms that the largest structural changes at the chromophore are restricted to the lysine side chain and retinal C15, which form the Schiff base. The photoresponse of the retinal of NM-R3 in the early intermediate state is therefore markedly different from that of bR, both in which atoms move strongly and the direction of movement (21).

Fig. 2 Atomic movements induced by photoexcitation.

|Fobs|light – |Fobs|dark difference Fourier electron density map contoured at ±3.5σ in (A) CW95K-ND3, (B) CW95K-ND10, and (C) CW95K-ND30. For each map, three panels show different views of the retinal, and the central panel shows unexpected density peaks that appear consistently in each map. Red arrows suggest the movement of chloride ions following photoexcitation. Positive and negative values are shown in blue and red, respectively. The stick model represents the dark state of NM-R3. Additional views of the electron density maps are available in movies S1 and S2.

Several internal cavities (IC1 to IC3) have been identified in NM-R3 that contain water molecules and might help conduct chloride ions (12). IC1 holds Wat11 in our models, and IC3 holds Wat16, but neither water molecule is displaced in the difference maps. Wat4 forms hydrogen bonds with Trp99, Arg95, and Asp231. It is weakly disordered at lower excitation levels and showed slight movement toward Arg95 in the CW95K-ND30 map (Fig. 2C). Wat10 and Wat37 were found within a pocket surrounded by Leu21, Leu22, Leu65, Arg95, Thr228, and Asp231; both showed negative difference densities with photoexcitation, but the temperature factors (18 to 25 Å2) did not increase significantly. Water molecules do not therefore account for the strong positive density peaks that appear in hydrophobic pockets consistently across datasets (Fig. 2 and fig. S5), and these peaks may be interpreted as chloride movement from the extracellular to the cytoplasmic face of the protein.

Cl-1 is shifted away from the PSB toward Asn98. Negative density with no corresponding positive density is found at the side chain of Asp231, indicating a loss of order rather than a directed movement. With increasing laser intensity and temperature, these changes increased, and features more distant from the chromophore became apparent (Fig. 2 and fig. S5). The side chains of Cys55, Met58, and Cys105 become disordered, and Trp201 (equivalent to Trp182 in bR) becomes more mobile with stronger excitation. Contoured at 3.5σ, Ser234 starts to flip the rotamer, breaking its hydrogen bond with Tyr204. Near the Cl-1 binding site, two positive difference electron density peaks appear, indicating the outgoing chloride ion (Cl-1B) and incoming chloride ion (Cl-4), but only the first of these peaks is evident in the CW95K-ND3 map (Fig. 2A). The intensity of both peaks becomes much higher with stronger excitation (Fig. 2, B and C). Two additional positive difference density peaks are found further from the retinal and represent the positions of chloride ions Cl-5 and Cl-6. Cl-5 appears between Ser60 and Cys105, and Cl-6 appears on the far side of Cys105, in a pocket formed with Ala125 and Ile129 (Fig. 2B). These features are found in all the maps, including pulse-95K and pulse-140K, with the Cl-6 position becoming especially strong in the CW95K-ND30 map (Fig. 2C).

Chloride conduction pathway of NM-R3

Contouring the difference electron density maps at 3.2σ shows a set of positive peaks within the body of the protein that apparently map out the migration path for substrate chloride ions moving from the extracellular face to the cytoplasmic face (Fig. 3). Additional views of the movement of the chloride ions are shown in fig. S4. In the dark state, chloride ion Cl-1 is found very close to the chromophore but lying toward the extracellular face. On photoexcitation, this chloride ion moves slightly to position Cl-1B, or further to the two positions (Cl-5 and Cl-6) near Cys105, closer to the cytoplasmic face of the protein. An additional chloride ion, Cl-3, is found in a pocket near Phe90, near the extracellular face. This peak of positive density appears most strongly in the CW95K-ND3 map and is reduced at higher temperature or laser power (Fig. 3). The position Cl-4 appears to indicate a chloride ion arriving near the PSB region from the extracellular face via Cl-3. In contrast to Cl-3, the height of this peak is significantly increased at higher laser power (Fig. 3 and Table 1).

Fig. 3 Chloride ion conduction path from the extracellular to cytoplasmic face.

|Fobs|light − |Fobs|dark difference Fourier electron density maps contoured at ±3.2σ covering a ribbon diagram of the entire protein model. Residues surrounding each chloride ion are represented by stick models. Positive and negative values are shown in blue and red, respectively. Two chloride ions (Cl-1 and Cl-2) are present in the refined dark model, and four additional ions (Cl-3, Cl-4, Cl-5, and Cl-6) were identified from the difference density maps. The Cl-1 ion moves slightly to site Cl-1B. A chloride ion at the extracellular face, Cl-3, was observed only in CW95K-ND3.

Table 1 Summary of the intensity quantification of the difference Fourier electron density map for key water molecules, chloride ions, and residues.

Values given in units of σ (equal to the root mean square electron density of the unit cell). Values below ±3.0σ were set to zero, as being of too low significance. Cells are colored according to the σ value, being darker green when the magnitude of the measured difference is stronger. Residue nomenclature: SC, side chain; BB, back bone; (+), positive density feature; (−), negative density feature.

View this table:

Since the movements are so small, difference distance matrix plots are helpful to understand the overall structural changes (at the backbone level) of NM-R3 on photoexcitation (fig. S6). The changes seen in CW95K-ND30 and pulse-140K are very similar. Residues near the extracellular region (Ala75 to Ala77, Leu85 to Leu88 in the B-C loop, Ser91 to Trp99 in the C helix, and Thr138 to Ser160 in the D and E helices) move toward the interior of the protein. Residues near the cytoplasmic region (Leu106 to Leu116 in the C helix, Ile49 to Thr51 in the B helix, and Gln249 to Ala253 in the G helix) move outward (fig. S6).

Functional studies on residual mutants related to the chloride conduction pathway

Following the observation of density near internal residues, it was tested whether replacing these side chains might affect chloride transport activity. Pumping activity was greater for the C105A mutant than the wild type, while C105F exhibited weak activity, and S60F was nearly inactive. The Cl-3 site sits between Phe90 and Gln224. Replacing either residue with tryptophan strongly reduced chloride transport, and the Q224W mutant was essentially inactive (Fig. 4). These results provide direct evidence that the residues identified by our crystallographic analysis have functional roles.

Fig. 4 Mutational analyses of functional residues in the chloride ion conduction pathway.

(A) The anion-pumping activity of wild-type (WT) and mutant NM-R3 (C105A, C105F, S60F, Q224W, and F90W). Light-induced pH changes were measured using Escherichia coli cells expressing NM-R3 suspended in a solution containing 100 mM NaCl in the absence (gray solid lines) or presence (black solid lines) of the protonophore CCCP (30 μM) or in the presence of both 30 μM CCCP and 50 mM TPP+ (black broken lines). The broken red line in each panel indicates the maximum activity level achieved by WT protein with photo-stimulation. div., division. (B) Positive difference electron density of CW95K-ND30 contoured at ±3.5σ (top) and CW95K-ND3 contoured at ±3.2σ (bottom), showing the path of chloride ions through the protein.

Ser234, equivalent to an alanine in bR or HsHR (figs. S7 and S8), is associated with color tuning and ion transport (24, 25). This serine is conserved in rhodopsins from cyanobacteria and marine bacteria but is replaced by asparagine or threonine in rhodopsin family members such as sensory rhodopsins. Mutation experiments showed that placing either of these residues at position 234 maintains NM-R3 activity and do not cause any red shift in the visible spectrum. The S234A mutant, however, has decreased activity, red shift of the principal visible absorption band, and weaker chloride ion binding (fig. S8). The discovery that Ser234 breaks its hydrogen bond with Tyr204 on photoexcitation (Fig. 2 and fig. S5) suggests that this residue may explain some of the differences in retinal conformation between bR and NM-R3 in early intermediate states.


Since its discovery almost 50 years ago, bR has been used as a model system, both as a protein fold and as a light-driven pump. Dozens of structural analyses have been reported, and numerous studies have attempted to relate the structural changes caused by photoexcitation to the directed transport of hydrogen ions (26). The large number of crystallographic structural models allows them to be clustered, showing clear grouping by crystallization conditions and space group, but this diversity of crystal forms has not prevented comparison of diffraction data from similar crystals in resting and excited states (27). By illuminating crystals at very low temperatures (100 or 110 K), several groups trapped early intermediate K states in the photocycle. In each case, the overall protein conformational change measured by global Cα shifts was very low. X-ray data for one study were released publicly, showing significant difference density only around the retinal, the water molecule hydrogen bonded to the chromophore becoming disordered, and Lys216 moving sideways. Controversy arose once it was realized that the disorder of Asp85 may be caused by radiation damage, since excessive exposure to x-rays can remove the carboxyl groups (28). Asp85 and the conserved Asp212 are the most sensitive residues, although the loss of these groups gives only negative difference density with no corresponding positive density. Care is therefore required to minimize x-ray exposure, especially if the same crystal is used to collect datasets in both the ground and excited states to reduce noise. bR is also known to show a color change on x-ray exposure, and difference Fourier maps show paired positive and negative peaks at the retinal and nearby water molecule. Matsui and colleagues (28) concluded that the changes due to x-ray exposure are significantly smaller than those of the K state in the photocycle. However, it is important to consider these effects in crystallographic studies of other rhodopsins, especially in regard to the earliest changes at the chromophore on photoexcitation. In this study, we used low temperature to block the photocycle of the chloride transporter NM-R3 at an early intermediate. To separate the effects of photoisomerization from those of x-ray damage, we used five separate crystals to determine independent difference Fourier maps showing the changes induced by different levels of photoexcitation. In this way, the effects of x-ray damage, which must be comparable for each dataset, may be distinguished from the effects of visible light.

Recently, structural changes in bR have been investigated through the use of XFELs to create time-resolved movies through a series of femtosecond snapshots (20), allowing comparison with previous crystallographic studies of trapped intermediates (29). XFEL allows the structural dynamics of retinal chromophore isomerization in bR to be studied at subpicosecond time scales (21), but the function of proton pumps remains extremely difficult to resolve because the hydrogen ion itself does not strongly scatter x-rays. Synchrotron radiation has now been used to study changes in bR over time scales from femtoseconds to milliseconds (29), but chloride ion transporters offer an enormous advantage over bR because the substrate is highly electron dense. Since proton and chloride pumps move their respective substrate ions in opposite directions and also have transporting mechanisms with different photocycles, it is necessary to study the structure of the light-induced active form of the chloride ion pump to understand its function. Previous studies of NpHR, which requires a homotrimeric conformation for activity, have documented light-induced structural changes under various conditions but did not detail the movement of chloride ions at high resolution (30). It was, however, found that removal of chloride from a site next to the chromophore leads to a large deformation of the C helix and the B-C loop (31). As with NM-R3, this chloride site overlays Asp85 of bR.

In 2007, Gmelin and colleagues (32) described the x-ray structure of the L1 intermediate of hR, obtained by brief photostimulation at room temperature of a crystal subsequently cryo-cooled, giving a blue shift of the absorption band close to 580 nm. Since separate crystals were needed for each dataset, the difference maps had additional noise, and bromide was substituted for chloride to observe movements at two known binding sites; one halide ion close to the retinal was found to move only 0.3 Å, and the other, 14 Å closer to the external face of the protein, did not move measurably at all. Negative density close to the retinal was interpreted as isomerization, but no corresponding positive density was observed. Our study also shows the chloride ion close to the Schiff base moving very little on constant photostimulation within NM-R3 crystals held at cryo-temperatures throughout, but other differences are noteworthy. In our study, we found density peaks within early intermediates of photoexcited NM-R3 that increased with stronger excitation and may be interpreted as waypoints on the path taken by chloride ions (movies S1 and S2). The changes at the chromophore are notably different, as well as from those observed in previous studies of bR (1921), particularly the direction of movement and the fact that the lysine side chain moves more than the retinal itself. By comparing crystals subjected to different intensities of laser but an equal x-ray dose, this study provides definitive proof that the increased disorder of Asp231 is not due to x-ray exposure and can be attributed to the effects of visible light. Since the light-minus-dark difference density maps depend only on the phases of the dark-state model (in which the retinal is clearly in the all-trans configuration), they are free of any bias that might be introduced by modeling the chromophore in an isomeric form in the photoexcited state. The absence of any significant difference density, despite the high resolution and quality of the diffraction data, shows that the retinal atoms move much less than might be expected. Retinal isomerization on light exposure is a much studied topic. Photoactivation of ATR free in solution produces a mixture of 9-cis, 11-cis, and 13-cis isomers, with a total yield of about 25%, but studies of bR show clearly that the rigid retinal pocket controls the isomerization, giving a much higher yield of about 67% of the 13-cis isomer (33, 34). It is known that mutation of Asp85 or Asp212 in bR greatly reduces the rate and products of the isomerization (35, 36), but the pocket of NM-R3 is substantially different, as described above. The hydrogen bond network formed by the side chains of Ser234, Tyr204, Asp231, and Arg95 creates a rigid surface to one side of the PSB, with the chloride ion on the other, only 3.5 Å away from the nitrogen atom. The chloride ion is enclosed within the protein by a hydrogen bond between Ser60 and Asn98, equivalent to Ile52 and Asp85 in bR. Excitation of the retinal in bR involves charge transfer away from the PSB, which, in the case of NM-R3, would help the chloride ion move away, and sensory rhodopsins are known in which this effect rather than retinal isomerization triggers the protein (37). Further work is needed, in particular time-resolved studies by XFEL, to understand how energy captured by the chromophore is directed toward ion transport and the nature of later intermediates in the photocycle of the anion pump.


Cloning, expression, and purification

The pET21b vector carrying the NM-R3 gene derived from Nonlabens marinus was transformed into BL21-CodonPlus Escherichia coli (DE3; Agilent Technologies, Santa Clara, CA, USA). The cells were grown in a high-salt Luria-Bertani medium at 37°C. When cells reached an optical density at 600 nm (OD600) of greater than 1.0, 50 μM ATR (Sigma-Aldrich, St. Louis, MO, USA) and 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) were added. NM-R3 was expressed for 6 to 8 hours at 30°C and harvested by ultracentrifugation at 3500g for 20 min at 4°C. Harvested cells were suspended in lysis buffer containing 50 mM tris-HCl (pH 7.0) and 150 mM NaCl and sonicated. The membrane fraction was isolated by ultracentrifugation at 370,000g for 40 min at 4°C and solubilized by incubating in resuspension buffer containing 50 mM tris-HCl (pH 7.0), 150 mM NaCl, 1% n-dodecyl-β-d-maltoside (DDM), and 0.2% cholesteryl hemisuccinate (CHS) for 2 hours at 4°C. Solubilized NM-R3 was purified by a TALON affinity column and then size exclusion chromatography (Superdex 200 16/60; GE Healthcare, Little Chalfont, UK) equilibrated with 20 mM Hepes (pH 7.5), 150 mM NaCl, 0.05% DDM, and 0.01% CHS. The eight NM-R3 mutants—C105A, C105F, S60F, Q224W, F90W, S234A, S234T, and S234N—were generated by site-directed mutagenesis and purified in the same manner.


Microcrystals suitable for conventional synchrotron experiments appeared in 0.15 M sodium chloride, 0.15 M calcium chloride, 0.1 M MES (pH 6.0), and 30% polyethylene glycol dimethyl ether 500 using NM-R3 at 45 mg/ml. Crystals grown in the lipidic cubic phase were harvested after 10 days. Crystals grew in space group C2 and contained one monomer in the asymmetric unit (table S1).

Crystal illumination and structure determination

The crystals were illuminated by laser light under the same conditions as those used to measure each crystal absorption spectrum. The diffraction data were collected at beamline NW12A with a PILATUS3 S2M detector at the Photon Factory, KEK (Tsukuba, Japan). X-ray datasets were collected from each of the five separate crystals, i.e., in the dark (resting state) and treated with one of the irradiation protocols. The crystals diffracted to a resolution of at least 1.8 and 2.1 Å in the dark and photoexcited states, respectively. Total x-ray exposure time was under 10 min in the dark and light states at 95 or 140 K. All data were processed and scaled using HKL-2000 package (38) and CrysAlisPro (Rigaku Oxford Diffraction, 2016). Initial models of the near-isomorphous light- and dark-state crystals were obtained by molecular replacement with PHASER (39) using the previously solved NM-R3 structure [Protein Data Bank (PDB) 5G28]. The models were refined with a simulated-annealing protocol using a bulk solvent correction and further refined using Coot and PHENIX (40, 41). All data collection and refinement statistics are shown in table S1.

Calculation of difference Fourier density maps

Difference Fourier maps were generated using the CCP4 suite (42) and phenix.fobs_minus_fobs_map in PHENIX (41). Structure factor amplitudes were obtained using TRUNCATE, combined using CAD, and scaled using SCALEIT (43, 44). Last, difference Fourier electron density maps (|Fobs|light − |Fobs|dark)·exp.[iΦcalc] were calculated in fast Fourier transform using phases calculated from the refined dark-state structure (45).

Microspectrophotometry for single crystals

Light-induced absorption spectral changes in single crystals of NM-R3 were measured using a Nikon Eclipse TE2000-U inverted-type microscope equipped with a metal oxide semiconductor–type photodiode array (Unisoku PK-120). The NM-R3 single crystal on a cryo-loop was cooled to 95 or 140 K under a flow of cold nitrogen gas, and its absorption spectrum was measured before and during irradiation, either with a CW 532-nm laser diode module attenuated by 3, 10, or 30% ND filters (with final power densities of 1.1, 3.8, and 11 mW mm−2, respectively) or with a 532-nm 10-kHz pulsed laser (FDSS532-Q2, CryLas, Berlin) with a 1% ND filter (with a final power density of ~0.47 mW mm−2). A 532-nm notch filter (TECHSPEC OD6/25 mm, Edmund Optics) was used to minimize the effects of light scattering.

Measurement of NM-R3 activity

BL21-CodonPlus (DE3) E. coli cells expressing NM-R3 (wild type or mutant) were incubated at 37°C in high-salt Luria-Bertani medium supplemented with ampicillin (100 mg/ml). At an OD600 of 1.0, the overexpression of NM-R3 and mutants was induced by the addition of 1 mM IPTG and 50 μM ATR. Incubation was continued for 4 hours at 37°C. The cells were collected by centrifugation at 4000g for 5 min, washed four times with 100 mM NaCl buffer, and dissolved in the desired solvents, including CCCP (carbonyl cyanide 3-chlorophenylhydrazone)and TPP+ (tetraphenylphosphonium chloride), until OD600 = 8.0. Cell suspensions were incubated in the dark for 2 hours before proton ion flux changes in the buffer were measured using an F-72G pH electrode (Horiba, Kyoto, Japan) under a 520-nm xenon light source (Elpisbio, Seoul, South Korea) at 25°C.

Ultraviolet-visible spectroscopy

Final samples were isolated in 10 mM Mops (pH 6.5) and 0.05% DDM using a Superdex 200 size exclusion column (GE Healthcare). The absorption spectra were scanned for samples (wild type, S234A, S234T, and S234N) using a V-650 spectrophotometer (JASCO, Oklahoma City, OK, USA) at 25°C. Anion titration experiments were performed by adding chloride ion, and SDs were calculated from triplicate experiments. Anion binding affinities were calculated using the Hill equation to analyze absorption changes at a wavelength of 581 nm, except in the case of S234A, for which data were measured at 596 nm. The dissociation constants are listed in fig. S8.

Difference distance matrix

The difference distance matrix of the light and dark forms of NM-R3 was calculated using the Differences Distance Matrix Program (Center for Structural Biology, Yale University, New Haven, CT, USA). In the difference distance map, the changes in distance between Cα-Cα atoms were indicated by the brightness of the colors, and color saturation levels were set to represent Cα shifts of 0.5 Å closer (red) and 0.5 Å away (blue) (46).

Intensity quantification of difference Fourier map

Difference Fourier electron density maps (|Fobs|light − |Fobs|dark)·exp.[iΦcalc] were calculated for each crystal, and amino acid residues, water molecules, and chloride ions that show strong features in at least one of these maps are listed in Table 1. The root mean square electron density σ was determined by Coot, and values with a magnitude greater than ±3.0σ are indicated to one decimal place. Values below the threshold (3.0σ) are marked as zero (20).


Supplementary material for this article is available at

Fig. S1. Absorption spectra of an NM-R3 crystal before and after laser irradiation.

Fig. S2. Comparison of retinal conformations before and after laser irradiation.

Fig. S3. Entire difference Fourier electron density map and average temperature factors of NM-R3 crystal structure in the dark and light state under different laser irradiation conditions.

Fig. S4. Stereoscopic views of the difference electron density maps near the retinal.

Fig. S5. Pathway for chloride ion transfer from extracellular to cytoplasmic face.

Fig. S6. Difference distance matrix plots showing differences between the dark and light states of NM-R3.

Fig. S7. Comparison of the NM-R3, HsBR, and HsHR retinal-binding pockets.

Fig. S8. Mutation analysis of Ser234, an important residue for NM-R3 function.

Table S1. Data collection and refinement statistics.

Movie S1. Detailed view of changes around the chromophore of NM-R3.

Movie S2. Difference electron density map showing changes at the chromophore of NM-R3.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank all staff members at the NW12A beamline station at the Photon Factory for assistance with data collection. Funding: This work was supported by the Research Program (NRF-20171A2B2008483, 2019M3A9F6021810 to W.L. and NRF-20161A6A3A04010213 to J.-H.Y.) through the National Research Foundation of Korea and the Platform Project for Supporting Drug Discovery and Life Science Research (BINDS) from AMED under grant number JP18am0101076 (to S.-Y.P.), JSPS KAKENHI grant number JP19H05779 (to S.-Y.P.), and Takeda Science Foundation (to S.-Y.P.). The work was supported in part by Brain Korea 21 (BK21) PLUS program. J.-H.P. and Wonbin Lee are fellowship awardees of the BK21 PLUS program. Author contributions: J.-H.Y., S.-Y.P., and Weontae Lee conceived the study and designed the experiments. J.-H.Y., J.-H.P., Z.J., and Wonbin Lee prepared the crystal samples and performed biochemical assays. M.O., J.-H.Y., and N.I. performed data collection and structure refinement. A.S.-T., N.S., M.O., and S.-Y.P. set up laser conditions and performed all spectral studies. M.O., J.-H.Y., J.R.H.T., Weontae Lee, and S.-Y.P. wrote the manuscript with feedback from all authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. All structures have been deposited in PDB, and PDB codes are presented in table S1. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article