Structural Mechanism for Light-driven Transport by a New Type of Chloride Ion Pump, Nonlabens marinus Rhodopsin-3^,

Abstract

The light-driven inward chloride ion-pumping rhodopsin Nonlabens marinus rhodopsin-3 (NM-R3), from a marine flavobacterium, belongs to a phylogenetic lineage distinct from the halorhodopsins known as archaeal inward chloride ion-pumping rhodopsins. NM-R3 and halorhodopsin have distinct motif sequences that are important for chloride ion binding and transport. In this study, we present the crystal structure of a new type of light-driven chloride ion pump, NM-R3, at 1.58 Å resolution. The structure revealed the chloride ion translocation pathway and showed that a single chloride ion resides near the Schiff base. The overall structure, chloride ion-binding site, and translocation pathway of NM-R3 are different from those of halorhodopsin. Unexpectedly, this NM-R3 structure is similar to the crystal structure of the light-driven outward sodium ion pump, Krokinobacter eikastus rhodopsin 2. Structural and mutational analyses of NM-R3 revealed that most of the important amino acid residues for chloride ion pumping exist in the ion influx region, located on the extracellular side of NM-R3. In contrast, on the opposite side, the cytoplasmic regions of K. eikastus rhodopsin 2 were reportedly important for sodium ion pumping. These results provide new insight into ion selection mechanisms in ion pumping rhodopsins, in which the ion influx regions of both the inward and outward pumps are important for their ion selectivities.

Introduction

Microbial rhodopsins compose a large family of 7-transmembrane (TM)³ proteins containing all-trans-retinal, as the light-absorbing chromophore. The first microbial rhodopsin, bacteriorhodopsin, was found in halophilic archaea in 1971 and works as a light-driven H⁺-pump (1). Subsequently, a light-driven Cl⁻ (chloride ion)-pump halorhodopsin (HR) was also discovered in halophilic archaea (2). A while after these discoveries, microbial rhodopsins were thought to be restricted to hypersaline environments. However, because of the recent developments in molecular biology techniques, three different functional rhodopsins, a light-driven H⁺-pumping proteorhodopsin, a light-driven Na⁺ (sodium ion)-pumping rhodopsin, and a light-driven Cl⁻-pumping rhodopsin, were found in marine microorganisms (3–5). Now it is known that ion-translocating rhodopsins are widely distributed in a variety of microorganisms in all three domains of life (6–8).

The first Cl⁻-pumping rhodopsin containing the unique NTQ motif (Asn-85, Thr-89, and Gln-96 in bacteriorhodopsin numbering) was found in the marine flavobacterium Nonlabens marinus in 2014, and it was named NM-R3 (5). A phylogenetic analysis based on rhodopsin sequences revealed that NM-R3 belongs to a distinct lineage from the archaeal light-driven Cl⁻-pumping rhodopsins, such as HR (TSA motif) (5). HR was found in Haloarchaea in a hypersaline environment, whereas the eubacterium N. marinus was found in seawater (Fig. 1). Moreover, these genes belong to distinct phylogenetic lineages, suggesting that Cl⁻-pumping rhodopsins evolved independently in marine bacteria (5, 9). According to this analysis, the Na⁺-pumping rhodopsin type clade containing the unique NDQ motif, such as that containing the marine flavobacterium Krokinobacter eikastus rhodopsin 2 (KR2), is the closest relative to the Cl⁻-pumping rhodopsins (35.7% sequence identity between KR2 and NM-R3) (Fig. 2) (5, 10). In the bacteriorhodopsin (DTD motif), Asp-85 and Asp-96 work as the H⁺ acceptor and donor, respectively, and these carboxyl groups play key roles in H⁺ translocation (11, 12). In HR, Thr and Ser (corresponding to Asp-85 and Thr-89 in bacteriorhodopsin, respectively) interact with Cl⁻ (13, 14). In addition, recent studies showed that Asp-116 of KR2 (corresponding to Thr-89 in bacteriorhodopsin) works as an H⁺ acceptor, and the protonation of Asp-116 is a critical step for Na⁺ translocation (10, 15). These results suggested that the amino acid residues located at positions 85, 89, and 96 (bacteriorhodopsin numbering) play crucial roles in differentiating the specificity of the ion pumping activity.

FIGURE 1.

Unrooted maximum likelihood tree based on microbial rhodopsin amino acid sequences. Eight clades were categorized as follows: ClR (chloride ion-pumping rhodopsin); HR (halorhodopsin); NaR (sodium ion-pumping rhodopsin); SRI/SRII (sensory rhodopsin I and sensory rhodopsin II); BR (bacteriorhodopsin); ASR (Anabaena sensory rhodopsin), PR (proteorhodopsin), and XR (xanthorhodopsin). Dark yellow color indicates chloride ion-pumping rhodopsin clades. The NM-R3 gene used in this study is indicated by the red circle.

FIGURE 2.

Sequence alignment of NM-R3, KR2, halorhodopsins, and bacteriorhodopsin with known structures. Sequence alignment among representative microbial rhodopsins that function as light-driven ion pumps is shown. The amino acid sequences of NM-R3 from N. marinus rhodopsin-3 (accession number BAO55276) (5), K. eikastus light-driven sodium pump rhodopsin 2 (KR2) (PDB code 3X3B) (15), hHR (PDB code 1E12) (13), nHR (PDB code 3A7K) (14), and bacteriorhodopsin from Halobacterium salinarum R1 (BR) (PDB code 1C3W) (31) were used for comparison. The numbers shown in the right or upper column represent the positions of amino acid residues. Conserved residues among these rhodopsins are highlighted in red-filled boxes, and motif residues are highlighted in blue-filled boxes. The α-helix and β-sheet structures are depicted as yellow- and pale blue-filled boxes, respectively. The gray boxes represent missing residues in the determined structure.

Light energy is captured by the positively charged retinal, and this induces retinal isomerization from the all-trans to 13-cis conformation to drive the ion translocation (16, 17). During the light-driven ion translocation processes in bacteriorhodopsin and proteorhodopsin, two crucial carboxylic amino acid residues act as the primary Schiff-base proton acceptor and donor, respectively (11, 12, 18). In the case of HR, the binding of the negatively charged Cl⁻ stabilizes the protonated Schiff base, and retinal isomerization flips the N-H dipole, thus driving the movement of Cl⁻ toward the intracellular side of retinal (16, 19, 20). Analyses of the structure-based model of Na⁺ translocation in KR2 (light-driven Na⁺-pumping rhodopsin) suggested that the proton is transferred from the Schiff base to the neighboring aspartate and enables the passage of Na⁺ across the neutral Schiff base (15, 21).

In this study, we determined the crystal structure of a new type of light-driven chloride-pump rhodopsin (NM-R3) at 1.58 Å resolution. Moreover, this is the first report of a comparison between the structures of a light-driven inward chloride ion pump and an outward sodium ion pump. We discovered the common property that the “ion influx regions” are critical for their functions, in the evolutionarily similar but functionally different NM-R3 and KR2 pumps.

Experimental Procedures

Construction of Plasmids

A codon-optimized NM-R3 gene, originating from the marine flavobacterium N. marinus strain S1-08^T (5), was attached by overlap PCR to sequences encoding a modified histidine tag and the cleavage site for tobacco etch virus (TEV) protease at the N terminus and then subcloned into the plasmid vector pCR2.1-TOPO (5, 22).

The E5A, S60A, S91A, N92V, N98T, N98T/T102S, T102D, T102V, Q109A, Q109E, and T51P/T243S mutations were separately introduced into the NM-R3 gene subcloned into pET21a. These C-terminally histidine-tagged NM-R3 mutants were expressed in Escherichia coli C41 (DE3) cells (Lucigen). E. coli cells expressing wild-type and mutants of NM-R3 were cultured as described previously (5).

Cell-free Protein Synthesis and Purification

Cell-free protein synthesis of NM-R3 was performed essentially according to the previously reported protocols used for Acetabularia rhodopsin I production (22–24). Briefly, the cell-free reaction mixture included 100 μm all-trans-retinal, 0.4% digitonin, and 6.67 mg/ml egg yolk phosphatidylcholine. The dialysis unit was incubated for 6 h at 30 °C, gradually cooled to 4 °C, and kept overnight in the dark. The membrane fraction was collected by ultracentrifugation and solubilized with buffer A (20 mm Hepes-NaOH (pH 7.4), containing 500 mm NaCl), including 1.0% n-dodecyl β-d-maltopyranoside (DDM).

The NM-R3 protein was affinity-purified on Ni-NTA superflow resin (Qiagen) equilibrated with buffer B (buffer A including 0.04% DDM), which was washed extensively with 15 mm imidazole in buffer B, and eluted with 400 mm imidazole in the same buffer. The affinity tag was cleaved with a final concentration of 0.1 mg/ml TEV protease. The cleaved tag and the His-tagged TEV protease were removed by passage through Ni-NTA superflow resin, and the NM-R3 protein was recovered from the flow-through fraction with buffer B and 15 mm imidazole. The protein solution was concentrated with a 50,000 molecular weight cutoff Amicon ultra filter unit, applied to a Superdex 200 10/300 column (GE Healthcare), and equilibrated in buffer B and 1 mm dithiothreitol. The resulting peak fraction from the size-exclusion chromatography was collected and concentrated again.

Crystallization

NM-R3 was crystallized by the in meso method. The purified protein solution (57 mg/ml) and monoolein (40:60 w/w) were mixed with a micro syringe-based mixing device, and a 50-nl portion of the homogenized mixture was placed on a glass plate well, using a micro-dispenser (25). The precipitant solution (0.8 μl), containing 0.1 m Tris-HCl (pH 8.0), 400–500 mm ammonium formate, and 40% polyethylene glycol 200, was overlaid on the mixture, and then the well was sealed with a glass coverslip. The glass plates were incubated at 20 °C for a week.

Data Collection, Structure Determination, and Refinement

Data were collected at BL32XU of the SPring-8 synchrotron (26). The diffraction data were processed with the XDS programs (27), and the structure was solved by molecular replacement, using the Phaser program (28) in the Phenix suite (29), with Acetabularia rhodopsin I (PDB code 5AWZ) as the search model (22). The refinement was conducted using the Phenix (29) programs, and the structure was manually rebuilt with the Coot (30) program. Data collection and refinement statistics are presented in Table 1.

TABLE 1.

X-ray data collection, phasing and refinement statistics

Statistics for the highest resolution shell are shown in parentheses.

	NM-R3
PDB code	5B2N
Data collection
Space group	C2
Cell dimensions
a, b, c (Å)	101.6, 49.8, 75.3
α, β, γ (°)	90.0, 130.4, 90.0
Wavelength (Å)	1.0
Resolution (Å)	29.28–1.58 (1.64–1.58)
Total reflections	159,709 (15,641)
Unique reflections	38,987 (3826)
Redundancy	4.1 (4.1)
Completeness (%)	99.0 (98.0)
I/σ (I)	11.57 (1.50)a
Rmerge (%)	7.70 (93.13)
Rmeas (%)	8.87 (106.8)
CC½ (%)	99.8 (69.3)
Refinement
Rwork (%)	17.58 (33.33)
Rfree (%)	19.76 (36.27)
r.m.s.d. bond lengths (Å)	0.006
r.m.s.d. bond angles (°)	0.88
Average B factor (Å2)
All	23.25
Protein	21.62
Ligand	38.48
Solvent	37.51
Ramachandran plot
Most favored regions (%)	99.0
Outliers (%)	0

^a I/σ (I) = 2.0 is 1.62 Å resolution.

Thermostability Assay

Evaluations of the thermostabilities of the wild-type and mutants of NM-R3 were performed by measuring the decrease in the absorption spectra upon temperature-induced protein aggregation. The affinity column purified protein (150 μg), in 100 μl of Hepes-NaOH (pH 7.4) buffer containing 1,000 mm NaCl, 0.04% DDM, and 400 mm imidazole, was incubated at 60 °C for 1–90 min. After heating, the solution was rapidly cooled on ice, and the protein aggregates were removed by centrifugation (10,000 × g, 3 min). The absorption spectrum of the supernatant was measured with a UV-visible spectrophotometer (V-630 Bio, Jasco).

Measurement of Light-induced Cl⁻ Transport Activity in E. coli

The Cl⁻ transport activity of the NM-R3 mutants was studied by monitoring the light-induced pH change. E. coli cells expressing wild-type NM-R3 or its mutants were cultured as described previously (5). The rhodopsin-expressing cells were collected by centrifugation (4,300 × g, 3 min), washed three times, and then resuspended in 100 mm NaCl for measurement. The cell suspension (6 ml) was placed in darkness and then illuminated using a 300-watt xenon lamp (MAX-303, Asahi Spectra) with a bandpass filter (520 nm; MX0520, Asahi Spectra) for 3 min. The light-induced pH changes were measured with a pH meter. All measurements were performed at 4 °C.

Results and Discussion

Overall Structure and Comparison with Those of Bacteriorhodopsin, Halorhodopsins, and KR2

We synthesized the full-length NM-R3 protein, using an E. coli cell-free protein production system (23, 24). We obtained over 25 mg of NM-R3 from the 9-ml reaction mixture. Its absorbance maximum wavelength was 532 nm (Fig. 3a), which is almost identical to that of NM-R3 overexpressed in E. coli (5).

FIGURE 3.

Absorption spectra, overall structure of NM-R3, and structural comparison between NM-R3 and KR2. a, absorption spectra of wild-type and mutant NM-R3 proteins. The blue, red, pink, purple, green, and black dotted lines represent the wild-type, T102V mutant, T102D mutant, N92V mutant, S60A mutant, and N98T/T102S mutant NM-R3, respectively. b, crystal structure of the NM-R3 monomer, viewed parallel to the membrane. The acyl chain and the retinal are depicted by orange and light blue stick models, respectively. Water molecules and the chloride ion are depicted by red and light green spheres, respectively. c, structural comparison between NM-R3 (yellow) and KR2 (pink). Overall architectures viewed parallel to the membrane.

The NM-R3 proteins were crystallized by the in meso method, and the resultant crystals diffracted x-rays to 1.58 Å resolution and contained one NM-R3 protomer in the asymmetric unit. The NM-R3 monomer is composed of short N- and C-terminal helices and 7-TMs (TM1 to TM7) connected by three cytoplasmic loops (ICL1–3), two extracellular loops (ECL2–3), and one antiparallel β-sheet at ECL1 (Fig. 3b). Within the NM-R3 monomer, one Cl⁻, 12 lipids, and numerous water molecules were clearly observed. The electron densities around the retinal revealed that the chromophore is in the all-trans conformation and is covalently attached to Lys-235 on TM7.

We compared the NM-R3 structure with those of bacteriorhodopsin (PDB code 1C3W, 21% sequence identity) (31) and two light-driven Cl⁻ pump HRs. The structural comparison with bacteriorhodopsin showed a root-mean-square deviation (r.m.s.d.) of 2.6 Å. The r.m.s.d. values between the NM-R3 structure and the two HR structures, HR from Halobacterium salinarum (hHR, PDB code 1E12, 20.6% sequence identity) and HR from Natronomonas pharaonis (nHR, PDB code 3A7K, 19.9% sequence identity), were 4.1 and 3.9 Å, respectively (13, 14). Although the structures of the TM regions and the active sites, as well as the positions of the protonated Schiff bases, were similar between NM-R3 and bacteriorhodopsin or HRs, the loop and short N- and C-terminal helix regions adopt quite different structures.

K. eikastus rhodopsin 2 (KR2), a 7-TM protein containing retinal as a chromophore, is a light-driven Na⁺-pumping rhodopsin (10). Surprisingly, despite the opposite transport directions and the opposite charges of the ion molecules transported by NM-R3 and KR2, the structure of NM-R3 is very similar to that of KR2 (r.m.s.d., 1.0 Å) (Fig. 2 and 3c), especially in the N-terminal regions and the cytoplasmic sides. In contrast, the structures of TM3 and TM7 are different (Fig. 3c). We propose that the similarity between the structures of NM-R3 and KR2 is due to their close evolutionary relationship, with one probably evolutionarily derived from the other.

The side chain of Tyr-255, residing between TM7 and the short C-terminal helix, interacts with the backbone oxygen atom of Gln-41, the side chain of Arg-48, and the side chain of Asp-74 of the next molecule in the crystal. As a result, the C-terminal α-helix of NM-R3 is kinked at Ala-253–Tyr-255 and oriented in a different direction from those of bacteriorhodopsin, HR, and KR2. The differences in their ion-pumping mechanisms may be a consequence of the subtle differences in these structures, as described below.

Chloride Ion-binding Site

The NM-R3 protein crystal was grown in the presence of excess Cl⁻, and a well defined Cl⁻-binding site was observed inside the crystal (Fig. 4a). The chloride atom at the binding site was initially modeled by referring to the HR structure. This electron density is consistent with the size of a chloride atom and has similar coordination to that in the HRs. The protonated Schiff base points toward the extracellular side. The Cl⁻ forms hydrogen bonds with the Schiff base (3.1 Å) and a water molecule (3.2 Å), located near the side chain groups of residues Trp-99 (3.6 Å) and Thr-102 (3.7 Å) (Fig. 4a). Trp-99 is conserved in the microbial-type rhodopsins (Fig. 2). The sole water molecule near the Cl⁻ is fixed by hydrogen bonds with Arg-95 and Asp-231 (Fig. 4a).

FIGURE 4.

Structural comparison of the chloride ion binding regions and putative ion translocation pathway between NM-R3 and halorhodopsins or KR2. Structures of the chloride ion binding regions of NM-R3 (a), hHR (PDB code 1E12) (13) (b), nHR (PDB code 3A7K) (14) (c), and corresponding region of KR2 (PDB code 3X3C) (15) (d) are shown. Numbers indicate the distance (Å) between two atoms connected by dashed lines. Chloride ions and water molecules are depicted by light green and red spheres, respectively.

The crucial amino acid residues for Cl⁻ binding in HRs are not conserved in NM-R3. The Thr and Ser residues that interact with Cl⁻ in HRs (Fig. 4, b and c) (13, 14) are replaced by Asn-98 and Thr-102 in NM-R3, respectively (Fig. 4a). Despite these differences in the amino acid residues interacting with the Cl⁻, the Cl⁻ positions in relation to the protonated Schiff base are similar in the NM-R3 and HR structures. In nHR, previous mutation studies revealed that Ser-130 (corresponding to Thr-102 in NM-R3) is more critical than Thr-126 (corresponding to Asn-98 in NM-R3) for Cl⁻ transport (32–34). In addition, it was proposed that Asp-116 in KR2 (corresponding to Thr-102 in NM-R3) plays an important role in Na⁺ pumping, as estimated from the structure and mutant experiments (Fig. 4d) (15, 21). Therefore, Thr-102 of NM-R3 may play a critical role in the ion pump activity.

To identify the important amino acid residues for the ion pump activity around the Cl⁻, we introduced mutations (S60A, N98T, N98T/T102S, T102D, and T102V) near the Cl⁻ (Fig. 4a). The T102D and N98T/T102S mutations of NM-R3 change the motif from NTQ to NDQ (KR2 type) and TSA (HR type), respectively. Ser-60 interacts with Asn-98 by hydrogen bonding (Fig. 4a). All of the mutants were expressed at sufficient levels for pump measurements in E. coli. The S60A, N98T, and N98T/T102S mutants retained partial Cl⁻ pump activity.

The N98T and S60A mutations had a small effect on the Cl⁻ pumping activity. Because of the loss of the hydrogen bond between Asn-98 and Ser-60 in the S60A mutant, the side chain of Asn-98 is predicted to be more flexible. These mutant experiments, as well as the fact that the distance between Asn-98 and the Cl⁻ exceeds 4.0 Å, suggested that Asn-98 of NM-R3 was not important for Cl⁻ binding.

Although both NM-R3 and HR have Cl⁻ pumping activity, the decrease in the Cl⁻ pumping activity by the N98T/T102S mutation (HR type, TSA motif) is probably induced by differences in their structures. However, the Asp mutation at Thr-102 (Na⁺-pumping rhodopsin type, NDQ motif) of NM-R3 could not convert it into the Na⁺-pumping rhodopsin-like outward Na⁺ pump (data not shown). Therefore, converting the ion selectivity must require mutations not only at the motif region of NM-R3 but also at other regions, despite the close evolutionary relatedness of Cl⁻-pumping and Na⁺-pumping rhodopsins. The T102V mutation, which makes this position less polar, induced a large spectral blue shift, whereas the other mutants changed the spectra only slightly (Fig. 3a). In addition, the T102V mutant had minimal Cl⁻ pump activity and slightly reduced thermostability (Fig. 5, a and b). Accordingly, the determinants of the Cl⁻/Na⁺ binding ability were not restricted to Thr-102.

FIGURE 5.

Light-induced chloride ion transport activities, and thermostability assays of NM-R3. a, chloride pump activities of wild-type and mutant NM-R3 proteins in E. coli cells. Each division of the vertical scale represents 0.25 pH change. b, thermostability assays of detergent-solubilized wild-type and mutant NM-R3 proteins. The absorption spectra of each sample was measured, and its λ_max value was plotted. Error bars represent the standard deviation of replicate samples (n = 3). Each experiment was performed independently.

Structure of the Cl⁻ Ion-releasing Region Near the Cytoplasmic Side of NM-R3 and Comparison with That of KR2

The putative ion translocation pathway from retinal to the ion-releasing region near the cytoplasmic side of NM-R3 is composed of two water molecules and seven side chains (Ala-50, Thr-51, Ser-54, Gln-109, Trp-201, Ser-234, and Thr-243) (Fig. 6a). Among them, the side chain of Gln-109 is stabilized by hydrogen bonds with two water molecules and the side chain of Ser-54, which is highly conserved among the putative Cl⁻-pumping rhodopsins and Na⁺-pumping rhodopsins (data not shown) (Fig. 6, a and b). Additionally, Asp-96 in bacteriorhodopsin (corresponding to Gln-109 in NM-R3) is known to function as the proton donor to the Schiff base in the late intermediate (16, 35). However, the Q109A and Q109E mutations of NM-R3 did not affect the Cl⁻ pumping activity (Fig. 5a), although the Na⁺ pumping activity of the Q123A mutant of KR2 (Fig. 6b, corresponding to Gln-109 in NM-R3) was significantly decreased (10, 15). Therefore, the role of Gln-109 in NM-R3 is different from those of the corresponding residues in KR2 and bacteriorhodopsin, and it does not appear to be important for the Cl⁻ transport function in NM-R3.

FIGURE 6.

Structural comparison of the ion translocation pathway between NM-R3 and KR2. Regions near Gln-109 of NM-R3 (a) and Gln-123 of KR2 (b) in the putative ion translocation pathway are shown. Close-up views of the intracellular regions in NM-R3 (c) and KR2 (d) are shown. e, region near the putative chloride ion influx region of NM-R3 (yellow) and the superimposed structure of KR2 (3X3C, pink). The blue electron density map indicates the unknown molecule interacting with NM-R3. The electron density map of the unknown molecule is shown with an F_o − F_c map contoured at 3.0 σ. The black and pink characters indicate the amino acid residues of NM-R3 and KR2, respectively. Numbers indicate the distance (Å) between two atoms connected by dashed lines. Chloride ions and water molecules are depicted by light green and red spheres, respectively.

We compared the NM-R3 structure with that of KR2, around the cytoplasmic region. Differences in hydrogen bonding networks were found in Ala-50, Thr-51, and Thr-243 in NM-R3, and the corresponding amino acid residues Ser-60, Asn-61, and Gly-263 in KR2 (Fig. 6, c and d). These amino acid residues were predicted to contribute to the formation of the ion selective filter. Indeed, the double mutant N61P/G263W (15) and the single mutant G263F (21) of KR2 could pump both Na⁺ and K⁺. To determine the role of this region, we measured the pump activity of the double mutant T51P/T243S of NM-R3. This mutant showed slightly decreased Cl⁻ pumping activity (Fig. 5a), suggesting that the ion-releasing region is not involved in anion selectivity or anion binding activity.

Structure of the Cl⁻ Ion Influx Region Near the Extracellular Side of NM-R3 and Comparison with That of KR2

Both NM-R3 and KR2 have unique features with regard to the short N-terminal helix located in the extracellular region, and other microbial-type rhodopsins lack this short helix. The E11A mutant of KR2 (corresponding to Glu-5 in NM-R3 within the short N-terminal helix) was unstable but only slightly decreased its Na⁺ pumping activity (10). The thermostability assay of the E5A mutant of NM-R3 revealed that it was unstable (Fig. 5b), and the ion pumping activity of the E5A mutant in E. coli cells was decreased (Fig. 5a) relative to that of the wild-type protein. These results indicated that the N-terminal helix in NM-R3 also plays an important role in structural stability, similar to that of KR2, although the importance of this region in ion pumping activity is different from that of KR2.

An unknown electron density was detected at the ion-influx region, formed by the side chains of eight amino acids in NM-R3 (Asn-3, Glu-5, Ser-91, Asn-92, Tyr-140, Gln-143, Glu-146, and Arg-223) (Fig. 6e, blue electron density). The mass of the unknown electron density corresponded to a C4 or C5 carbon chain, as deduced from its size, but we could not fit any possible molecules derived from the cell-free reaction mixture, lipids, detergents, or buffers, etc. to the electron density with certainty. We also could not identify the material derived from the electron density by a liquid chromatography coupled to quadrupole time-of-flight mass spectrometry analysis (data not shown). There is a structural difference at ECL1 between NM-R3 and KR2, especially in the side chains of Ser-91–Asn-92 (corresponding to Asn-105–Asn-106 in KR2), probably due to this unknown molecule (Fig. 6e). To determine the role of these amino acid residues, we measured the activities of NM-R3 proteins with the S91A or N92V mutation, which render this region less polar. These mutants lost ion translocation activity and exhibited reduced thermostability (Fig. 5, a and b). These results suggested that the polar amino acid residues within this region are important for both ion pumping activity and protein stability. Therefore, we speculate that this extracellular region forms the ion translocation pathway. The unknown molecule must be identified in future work to clarify the ion translocation pathway.

Evolutionarily Similar NM-R3 and KR2 Differ in Pumping Specificity and Ion Selection Mechanism

In conclusion, the structure of the light-driven inward Cl⁻ pump NM-R3 is very similar to that of the light-driven outward Na⁺ pump KR2. The mutational analyses revealed that most of the important amino acid residues for Cl⁻ pumping exist at the ion influx region, located on the extracellular side of the retinal binding pocket of NM-R3. In contrast, the mutations of residues at the Na⁺-releasing region on the extracellular side of KR2 had minimal effects on the Na⁺ pumping (10, 15). In addition, the cytoplasmic regions of NM-R3 do not seem to be involved in Cl⁻ pumping, although the corresponding regions of KR2 were reported to be important for Na⁺ pumping. The NM-R3 results provide new insights into ion selection mechanisms in ion pumping rhodopsins, in which the ion influx regions of both the inward and outward pumps are important for their ion selectivities.

Author Contributions

T. H., S. Yoshizawa, T. K. S., W. I., and M. S. designed the research and wrote most of the paper. T. H. performed the protein purification, the thermostability assays, the protein crystallography. S. Yoshizawa and Y. N. assayed the Cl⁻-pumping activities. N. O. designed and constructed the expression plasmids. M. H. provided the lipidic mesophase crystallization techniques. E. F. D., K. K., and S. Yokoyama contributed to the project organization and edited the manuscript. All authors commented on the manuscript.