Light-driven anion-pumping rhodopsin with unique cytoplasmic anion-release mechanism

Abstract

Microbial rhodopsins are photoreceptive membrane proteins found in microorganisms with an all-trans-retinal chromophore. The function of many microbial rhodopsins is determined by three residues in the third transmembrane helix called motif residues. Here, we report a group of microbial rhodopsins with a novel Thr–Thr–Gly (TTG) motif. The ion-transport assay revealed that they function as light-driven inward anion pumps similar to halorhodopsins previously found in archaea and bacteria. Based on the characteristic glycine residue in their motif and light-driven anion-pumping function, these new rhodopsins are called glycylhalorhodopsins (GHRs). X-ray crystallographic analysis found large cavities on the cytoplasmic side, which are produced by the small side-chain volume of the glycine residue in the motif. The opened structure of GHR on the cytoplasmic side is related to the anion releasing process to the cytoplasm during the photoreaction compared to canonical halorhodopsin from Natronomonas pharaonis (NpHR). GHR also transports SO₄^2– and the extracellular glutamate residue plays an essential role in extracellular SO₄^2– uptake. In summary, we have identified TTG motif-containing microbial rhodopsins that display an anion-releasing mechanism.

Microbial rhodopsins are a super-family of heptahelical transmembrane photoreceptive proteins with a retinylidene Schiff base (RSB), so-called the retinal chromophore, covalently bound to the protein moiety via a Schiff base linkage (1, 2, 3, 4). The retinal chromophore isomerizes to 13-cis form upon light absorption, leading to the subsequent sequential conformational changes of protein to exert biological molecular functions. The isomerized retinal chromophore spontaneously re-isomerizes to the all-trans form, which recovers the initial state and enables to absorb next photon. This cyclic reaction of microbial rhodopsins is called photocycle. While the molecular functions of microbial rhodopsins are highly diverse, ion-transporting rhodopsins are most abundant. There are two types of ion-transporting rhodopsins: ion-pumping rhodopsins and channelrhodopsins. The channelrhodopsins passively transport various ions along the electrochemical potential difference across cell membranes, while ion-pumping rhodopsins actively transport ions in a fixed direction, capable of overcoming electrochemical potential (5).

Several ion-pumping rhodopsins with different directions and ion selectivity are known: outward and inward proton (H⁺) pumps, inward anion pump, and outward sodium (Na⁺) pump (6, 7, 8, 9). Among them, bacteriorhodopsin (BR) from halophilic archaeon, Halobacterium salinarum, is the best characterized ion-pumping rhodopsin through many biochemical, spectroscopic, and structural biological studies (6, 10). The core process of the outward H⁺ transport function of BR involves two H⁺ transfers: from the RSB to D85 (the H⁺ acceptor) in the third transmembrane helix (TM3 or helix C) on the extracellular side and D96 (the H⁺ donor) in the same helix on the cytoplasmic side back to the RSB. Owing to their functional importance, D85 and D96, along with T89, is called the “DTD motif.” BR D85 is deprotonated in the dark state, and its negative charge stabilizes the protonated state of the RSB as a counterion. D96 is known to be replaced by other amino acids in other outward H⁺-pumping rhodopsins, specifically, glutamic acid, lysine, glycine, and serine, resulting in DTE (11), DTK (12), DTG (13, 14), and DTS (15, 16) motifs, respectively. In contrast, BR D85 is completely conserved in all outward H⁺-pumping rhodopsins. This residue is substituted with non-protonatable hydrophilic residues in inward anion pumps as TSA motif in haloarchaeal halorhodopsins (HRs), NTQ motif in bacterial anion pumps (ClRs), and TTD/TSA motifs in bacterial halorhodopsins (BacHRs) (4). In all inward anion-pumping rhodopsins, the absence of the counterion residue is compensated by binding an anion near the RSB (4). This anion is transported toward the cytoplasmic side upon light absorption and then anther anion binds from the extracellular side at the end of the photocycle.

We identified a new microbial rhodopsin subgroup featuring a novel Thr–Thr–Gly (TTG) motif in the genomes of alphaproteobacteria: Sphingomonas lenta (Accession number: WP_176473163), Salinarimonas soli (Accession number: WP_149817501), and an uncultured Sphingomonadaceae bacterium (Accession number: CAA9503817). To our best knowledge, the glycine residue at the position of the third motif residue homologous to BR D96 is present only in outward H⁺-pumping rhodopsins with the DTG motif (13, 14). However, substituting the D residue of DTG motif suggests that the function of the novel rhodopsins with the TTG motif is distinct from that of the outward H⁺ pump. Hence, we studied the function of these rhodopsins by assaying the ion-transport activity of the proteins heterologously expressed in Escherichia coli cells. Then, we studied the molecular mechanism of ion transport by investigating the photoreaction and structure by spectroscopic and X-ray crystallographic analyses, respectively.

Results

Glycilhalorhodopsin: a new anion pumping microbial rhodopsins

We found a distinct group of microbial rhodopsin with the unprecedented TTG motif (Figs. 1A and S1). If we expressed one of them from S. soli (SsGHR) in E. coli cells and illuminated with green light (530 ± 5 nm) in NaCl solution, the pH level increased by H⁺ uptake into the cell bodies (Fig. 1B). An addition of protonophore, carbonyl cyanide m-chlorophenylhydrazone (CCCP), a protonophore, enhanced the signal intensity, indicating the pH increase was not induced by transporting H⁺ or OH⁻ but by Na⁺ release from or Cl⁻ uptake into the cell bodies, which leads the secondary H⁺ influx that is accelerated by CCCP into the cells (17). Although the activity and specificity of sodium-pumping rhodopsins depend on the size of cations in the solvent (18), SsGHR activity in NaCl and KCl solutions were similar (Fig. 1, B and C). In contrast, the activity decreased in the presence of other monovalent anions, Br⁻, I⁻, and NO₃⁻, which are larger than Cl⁻ (Fig. 1, B and C). Based on its anion-size dependence, we concluded that this new rhodopsin is a novel light-driven inward anion pump different from HRs (7, 19, 20, 21) and bacterial anion-pumping rhodopsins (17, 22, 23). Similar anion-size dependent activities were observed for other rhodopsins with the TTG motif from S. lenta (SlGHR) and uncultured Sphingomonadaceae bacterium (SbGHR) (Fig. S2). Since this is the first anion-pumping rhodopsin group with glycine in the motif residues, they were named glycilhalorhodopsin (GHR). Interestingly, all GHRs transported SO₄²⁻ (Fig. 1, B, C, and S2). This is the third example of SO₄²⁻ pumping by rhodopsins after Synechocystis halorhodopsin (SyHR) (20) and bacterial anion-pumping rhodopsin from Alteribacter aurantiacus (AaClR) (23).

Figure 1.

Glycylhalorhodopsin from Salinarimonas soli (SsGHR).A, phylogenetic relationship of microbial rhodopsins. B, light-induced pH changes by SsGHR expressed in E. coli cells without (blue) and with (red) CCCP. The time period of light illumination is shown by green bars. Representative results from three independent experiments were shown. C, initial H⁺ transport rates of SsGHR. Data are mean ± standard deviation (S.D.). Individual data points are shown by white circles. D, absorption spectra of SsGHR in the presence (purple) and absence (red) of 1 M NaCl.

To investigate the molecular properties and anion-transport mechanism of GHR, we purified the SsGHR protein from E. coli cells. The purified protein exhibited a visible absorption peak with the maximum absorption wavelength (λ_max) at 550 nm in 1 M NaCl solution (Fig. 1D, purple spectrum). The visible peak was red-shifted and the second peak appeared at 400 nm in 0 M NaCl solution (Fig. 1D, red spectrum). The blue shift of the main peak in the presence of Cl⁻ indicates Cl⁻ binding near the RSB (9). The 400-nm peak for deprotonated RSB in the absence of Cl⁻ functioning as a counterion to the protonated RSB is known for other anion-pumping rhodopsins (7, 24). To determine the dissociation constant (K_d) of Cl⁻, absorption spectra were measured at different [Cl⁻] in n-dodecyl-β-D-maltopyranoside (DDM, Fig. 2) and 1-palmitoyl-2-oleoyl-phosphatidyl-ethanolamine/1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPE/POPG) lipid membrane (Fig. S3A). The absorption change in lipids were reproduced by the Hill equation with K_d = 8.7 ± 0.2 mM (Fig. S3A). In contrast, two different K_d, 1.1 ± 0.1 and 15 ± 1 mM, with similar amplitudes were observed in DDM (Fig. 2C), indicating heterogeneity in the K_d value among SsGHR molecules in DDM.

Figure 2.

Absorption spectral changes of SsGHR at different Cl⁻concentrations.A, absorption spectral change of SsGHR at different [Cl⁻]. B, difference spectra between each [Cl⁻] and 0 M Cl⁻. C, absorption changes at 409, 539, and 613 nm in the difference spectra plotted against [Cl⁻]. The absorption changes at three different wavelengths were globally fitted by a function obtained by summing the two Hill equations with the Hill coefficient = 1 (solid lines) to estimate the K_d of Cl⁻ and the S.D. D, resonance Raman spectra in H₂O (red) and D₂O (blue). Inset: C=N stretch bands fitted with the Lorentzian function (yellow lines). Bandwidth: 12.2 ± 0.5 cm⁻¹ for H₂O and 8.5 ± 0.3 cm⁻¹ for D₂O.

To investigate the retinal configuration, the retinal chromophore was converted to retinal oxime by adding hydroxylamine, and it was analyzed by high-performance liquid chromatography (HPLC). More than 90% SsGHR bound to all-trans-retinal both in the dark- and light-adapted states, and approximately 5 to 6% 13-cis-form was observed (Fig. S4A). Hence, all-trans-retinal is considered the functional form of the chromophore and isomerizes to 13-cis-retinal upon illumination as in other microbial rhodopsins. We investigated the pH dependence of the absorption spectrum of SsGHR in 1 M NaCl (Fig. S4B). When pH was lowered from 7.0 to 0.7, small and complicated peak shifts were observed with three acid dissociation constant values (pK_a; 0.97 ± 0.02, 2.42 ± 0.09, and 4.78 ± 0.09), indicating that multiple residues away from the retinal chromophore change their protonation states at acidic pH. At alkaline pH, a large blue shift of the absorption to approximately 400 nm was observed. This blue shift represents RSB deprotonation, with pK_a = 9.69 ± 0.01. In contrast, in the absence of NaCl, the RSB pK_a decreased to 7.45 ± 0.03, indicating that the bound Cl⁻ stabilizes the protonated RSB by functioning as a counterion.

The photocycle dynamics of SsGHR

Next, we investigated the photocycle of SsGHR by measuring the transient absorption change of the retinal chromophore after the excitation by a nanosecond 532-nm laser pulse (Fig. 3, A and B). At t = 80 μs after the excitation, a blue-shifted absorption of product species was observed at 475 nm (Fig. 3A, red line). Then, another product species with red-shifted absorption at 621 nm was observed in the millisecond time region, along with the decay of the blue-shifted species. Finally, the red-shifted absorption decayed simultaneously with the recovery of the initial state bleach. Because the conversion from blue- to red-shifted species is similar to that from the L- to O-intermediates observed for archeal and cyanobacterial HRs (25, 26), we use the same nomenclature hereafter. To obtain a detailed time profile with high time-resolution, the transient absorption change at probe wavelengths specific for the L (475 nm) and the O (621 nm) absorption and the initial state bleach (565 nm) was monitored using a photomultiplier tube (Fig. 3B). Accordingly, fast-decaying red-shifted component similar to the K intermediate of HRs was observed at t = 1 to 5 μs. All time-traces were well reproduced by the sum of five exponential functions (Fig. 3B, yellow lines), and a photocycle model involving five intermediate species was constructed (Fig. 3D). Based on this photocycle model, the absolute spectra of intermediate species observed after t = 80 μs in Figure 3A were calculated (Fig. 3C). Although it is difficult to obtain the transient absorption spectrum in the time region faster than 80 μs due to the limitation of the time-resolution of our ICCD system, we observed significant delays at 565 and 621 nm, along with the rise at 475 nm with identical time constants, 2.1 and 57 μs. This result indicates that the K intermediate having a red-shifted absorption spectrum compared to the L intermediate exists in microsecond region. Based on this result, we considered that this process represents the gradual conversion from the K to L in a double exponential manner (K₁ → K₂ $\rightleftarrows$ L₁ → L₂). The second and third spectra showed double peaks, whereas the first spectra have only a single peak of the L₂ intermediate (Fig. 3C). The double peaked spectra indicate that two species with different absorption wavelengths are quasi-equilibrated. While the first double-peaked spectra (Fig. 3C, green spectrum) represents the L₃ $\rightleftarrows$ O₁ quasi-equilibrium, the wavelength (544 nm) of the blue-side peak of the second one (cyan spectrum in Fig. 3C), which is different from that of the L, indicates quasi-equilibrium between O₂ and another blue-shifted species, N. Interestingly, the ratio between the blue- and red-side peaks changed at different [Cl⁻] (Fig. 3E and F), indicating that both the L₃ $\rightleftarrows$ O₁ and the O₂ $\rightleftarrows$ N quasi-equilibria are caused by the blue-shifted L- or N-intermediate production by the binding of a Cl⁻ ion to the O intermediates. By fitting the relative absorbance of each peak at different [Cl⁻] by the Hill equation, the estimated K_d of Cl⁻ were 350 ± 20 and 500 ± 300 mM for the L₃ $\rightleftarrows$ O₁ and O₂ $\rightleftarrows$ N quasi-equilibria, respectively, which were more than 40 times higher than the value of the initial state. If the Cl⁻ ion bound near the RSB is transported inwardly by the alternative access mechanism, one Cl⁻ is released to the cytoplasmic side solvent and another Cl⁻ is taken up from the extracellular side during the L₃ $\rightleftarrows$ O₁ and O₂ $\rightleftarrows$ N quasi-equilibria, respectively, as known for archeal and cyanobacterial HRs (7, 25, 26).

Figure 3.

Photocycle of SsGHR and its Cl⁻-concentration dependence.A, transient absorption spectra of SsGHR, t = 80 μs to 10 ms. The protein was reconstituted in POPE/POPG lipid vesicles in 1 M NaCl and 20 mM HEPES–NaOH (pH 7.0). B, time-evolutions at 475, 565, and 621 nm. The transient absorption changes were global-fitted with a multiexpotential function (yellow lines). C, the calculated absorption spectra of three kinetically-defined intermediate states. D, the photocycle model of SsGHR. Time constants are mean ± S.D. E, [Cl⁻] dependence of the absorption spectrum of the L₃/O₁ intermediate (left) and the absorbance at 515 and 610 nm (right). F, [Cl⁻] dependence of the absorption spectrum of the O₂/N intermediate (left) and the absorbance at 500 and 605 nm (right).

Previously, a H⁺ release was observed during the photocycle of MrHR, and it was suggested that D200, homologous to SsGHR D216, probably deprotonates during the L decay (26). Hence, we investigated H⁺ movement between the SsGHR and solvent using pyranine, a pH-indicator dye (Fig. S5A). However, no absorption change of pyranine was observed, indicating no H⁺ release or uptake occur during the photocycle, which is different from MrHR (26). Additionally, since the reaction rates of the photocycle were not significantly different between H₂O and D₂O solvents, a rate-limiting internal H⁺ transfer event does not occur (Fig. S5B). The resonance Raman spectra of SsGHR in 1 M NaCl are shown in Figure 2D. It showed a strong band of C–C stretching vibration at 1201 cm⁻¹. Because this is identical to the band of BR with the all-trans-retinal chromophore, the resonance Raman spectrum as well as the HPLC analysis (Fig. S4A) indicates SsGHR binds to all-trnas-retinal chromophore in the dark. The C=N str. band of the RSB appeared at 1636 cm⁻¹ in H₂O and shifted to 1623 cm⁻¹ in D₂O. The downshift in D₂O is caused by the decoupling of the C=N str. with N–D rocking vibrations upon the H/D exchange, indicating the RSB is protonated in the initial state. The RSB forms the stronger hydrogen bond, the larger the C=N str. band shift between H₂O and D₂O (27). The 13-cm⁻¹ shift of the C=N str. band of SsGHR is similar to that of NpHR (10.2 cm⁻¹) (28), HsHR (12 cm⁻¹) (29), SyHR (15 cm⁻¹) (20), and MrHR (16 cm⁻¹) (20) and weaker than that of TaHeR (17 cm⁻¹) (30), GtACR1 (18 cm⁻¹) (31), BR (19 cm⁻¹) (32), BPR (21 cm⁻¹) (33), NpSRII (21 cm⁻¹) (34), KR2 (22 cm⁻¹) (35), GPR (23 cm⁻¹) (33), CaChR1 (23 cm⁻¹) (36), and C1C2 (31 cm⁻¹) (37). Hence, the RSB in all anion-pumping rhodopsins form a weaker hydrogen bond compared to other functional microbial rhodopsins. Additionally, the C=N str. band peak width in D₂O (8.5 ± 0.3 cm⁻¹) was narrower than that in H₂O (12.2 ± 0.5 cm⁻¹) (Fig. 2D). C=N stretching band broadening indicates that the hydrogen bonding partner of the RSB is not a Cl⁻ ion but a water molecule, consistent with the crystal structure shown below (32).

R80, T83, and T87 are essential for the anion binding and transport of SsGHR

To obtain structural insights into the binding pocket of SsGHR, an X-ray crystallographic analysis was conducted. The SsGHR structure was obtained with a 2.63 Å resolution, and it formed a trimeric structure as other HRs and BR (Fig. 4A). Though the transmembrane part of SsGHR overlapped well with that of NpHR, MrHR, and other microbial rhodopsins (Figs. 4B and S6A), the characteristic C-terminus extending parallel to the lipid-bilayer plane over the cytoplasmic side of TM1 was observed only for SsGHR (Fig. 4B). A Cl⁻ ion forms hydrogen bonds with T83, T87, and a water molecule in the RSB region of SsGHR (Fig. 4C). This water molecule also forms hydrogen bonds with the RSB and D216. D216 is homologous to BR D212 (Fig. S1) that is a counterion stabilizing the protonated RSB (38). The hydrogen bond between the RSB and the water molecule is consistent with the C=N str. band broadening in the Raman spectrum in H₂O compared with in D₂O (vide supra). The hydrogen bonding network of the Cl⁻-binding site is similar to that in other HRs and NM-R3 (Figs. 4C and S6B). The Cl⁻ ion is present at a position similar to that of the counterion aspartate of outward H⁺ pumps (BR D85 and PspR D73, Fig. S6B). The orientation of R80, which is highly conserved in many microbial rhodopsins (4), differed from that of the homologous arginine in other Cl⁻ pumps, and did not interact with the water molecule and/or D216 only in SsGHR (Fig. S6B). The positive charge of R80, however, would stabilize the Cl⁻ binding in SsGHR by the long-range Coulomb interaction.

Figure 4.

X-ray crystallographic structure of SsGHR.A, the SsGHR trimer viewed from the direction parallel to the lipid bilayer (left) and from the cytoplasmic side with halorhodopsin from Natronomonas pharaonis (NpHR; PDB ID: 3A7K), MrHR (PDB ID: 6XL3), and bacteriorhodopsin (BR; PDB ID: 1M0L) (right). The approximate position of the membrane surface is shown by gray rectangles. B, the overlaid structures of SsGHR (green) and NpHR (purple) (left) and SsGHR (green) and MrHR (orange) (right). C, the RSB region in SsGHR (left), in NpHR (middle), and in MrHR (right). Cl⁻ ions and water molecules are shown as magenta and cyan spheres, respectively. Hydrogen bonding distances are indicated in angstrom (Å), with orange dashed lines. The motif residues are highlighted by blue characters.

To clarify the roles of the interaction of R80, T83, and T87 with Cl⁻, the pump activities of R80Q, T83A, and T87A mutants were measured (Fig. 5A). Moreover, the Cl⁻-transport activity of T87A mutant is approximately one-third that of the wildtype (WT), whereas no transport was observed for R80Q and T83A mutants, indicating that R80 and T83 are essential for Cl⁻ transport. To understand the reason for the decreased activities, these mutants were purified, and their Cl⁻-binding activity and photocycle were investigated (Fig. 5, B and C). Interestingly, Cl⁻ binding in R80Q and T83A mutants was weak at 100 mM NaCl with K_d > 1 M. Their photocycles did not show the N- and O-intermediates and are similar to that of the WT without Cl⁻ binding (Fig. S7), consistent with no anion-transport activity at 100 mM NaCl. In contrast, T87A exhibited substantial Cl⁻ binding in the initial state at 100 mM NaCl with K_d > 500 mM (Fig. 5B) and the N- and O-intermediate accumulation (Fig. 5C) consistent with its substantial Cl⁻ pumping activity.

Figure 5.

Ion-transport activity, Cl⁻binding, and photocycle of SsGHR mutants.A, light-induced pH changes by SsGHR mutants expressed in E. coli cells without (blue) and with (red) CCCP (top, representative results from three independent experiments were shown), and initial H⁺ transport rates of SsGHR mutants (bottom, mean ± S.D., n = 3 independent experiments). Individual data points are shown by white circles. The time period of light illumination is shown by green bars. B, absorption spectral change at different [Cl⁻] (left), difference spectra between each [Cl⁻] and 0 M (middle), and absorption changes at representative wavelengths in the difference spectra plotted of SsGHR mutants against [Cl⁻] (right). (R80Q: top, T83A: middle, T87A: bottom). C, transient absorption spectra (left), time-evolutions at representative probe wavelengths (second from the left), the calculated absorption spectra of kinetically-defined intermediate states (second from the right), and the photocycle models (right) of SsGHR mutants (R80Q: top, T83A: middle, T87A: bottom). Time constants are mean ± S.D.

Finally, we investigated the SO₄²⁻ transport of SsGHR. We found that E198 is present near the extracellular cavity of SsGHR and forms an ionic lock with R80 (Fig. 6A). Since mutations of homologous glutamate confer a SO₄²⁻ transport ability to MrHR (39, 40), we constructed SsGHR E198Q, A, and T mutants (Fig. 6, B and C). In contrast to SsGHR E198Q whose Cl⁻ and SO₄²⁻ transport activities were significantly lower than that of the WT, only the SO₄²⁻ transport acitivity decreased for SsGHR E198A and E198T, while they transport Cl⁻ efficiently as the WT. If we compare the relative transport activity of SO₄²⁻ against Cl⁻ transport activity, all E198 mutants showed smaller values than the that of WT (Fig. 6D), indicating that SO₄²⁻ selectivity was reduced by mutating E198, which is different from the previous finding on MrHR.

Figure 6.

Cl⁻and SO₄²⁻transport activity of SsGHR mutants.A, the structue around SsGHR E198. A Cl⁻ ion is shown as a magenta sphere. Hydrogen bonding distances are indicated in Å with orange dashed lines. B, light-induced pH changes by SsGHR mutants expressed in E. coli cells without (blue) and with (red) CCCP (representative results from three independent experiments were shown). The time period of light illumination is shown by green bars. C, initial H⁺ transport rates (mean ± S.D., n = 3 independent experiments) of SsGHR mutants. D, the ratio of SO₄²⁻ pump activities (mean ± S.D., n = 3 independent experiments) to the respective Cl⁻ pump activities of the WT and mutants. Individual data points are shown by white circles in C and D.

Discussion

In this study, GHR, a novel microbial rhodopsin with a TTG motif, was clarified to function as a light-driven anion pump. While GHR transports Cl⁻ and Br⁻, larger I⁻ and NO₃⁻ are hardly transported (Fig. 1B). Interestingly, GHR exhibited significant SO₄²⁻ transport, which was known only for SyHR (20) and recently reported AaClR (23). The K_d of anions in SsGHR is similar to those in other HRs but lower than those in bacterial anion pumps (Table S1).

The Cl⁻ release process from SsGHR during the photocycle is also different from that of the best characterized NpHR (Fig. 3). Though the equilibrium between the L and O of NpHR is external [Cl⁻]-independent (25), the ratio between the L and O of SsGHR was significantly alterred by changing [Cl⁻] (Fig. 3E). In these L intermediates appearing at t = 250 μs after the photoexcitation, the Cl⁻ ion is located at the cytoplasmic side of the retinal chromophore as suggested for NpHR (41) and NM-R3 (42). Hence, the [Cl⁻]-dependent spectrum of L₃ $\rightleftarrows$ O₁ of SsGHR indicates that the Cl⁻ ion moves in and out between the binding site near the RSB in the protein body and the cytoplasmic solvent phase in this equilibrium (Fig. 7A, left). In contrast, the Cl⁻ ion is not exchanged between the binding site and the cytoplasmic solvent in the L₃ $\rightleftarrows$ O₁ equilibrium in NpHR (25) and NM-R3 (42). In the X-ray crystallographic structure of the N intermediate (which is thought to be equivalent to the spectroscopically identified L₃ (7, 25)), a water chain is formed between the cytoplasmic Cl⁻ binding site near I134 in TM3 and K215 in TM6 (Fig. S8A) (43). The Cl⁻ ion goes back and forth between these two site in NpHR and then releases to the cytoplasmic solvent during the conversion from the L₃ $\rightleftarrows$ O₁ to the N $\rightleftarrows$ O₂ (Fig. 7A, right). Hence, the cytoplasmic Cl⁻ releasing mechanism of SsGHR is different from that of the best characterized NpHR. Several cavities are present in the cytoplasmic side of the SsGHR structure (Fig. 7B), but no such cavity is presesnt in NpHR and MrHR. Particularly, the largest cavity includes many hydrophilic residues, Q43, Q47, T98, N178, and Y223, and a water molecule, while G93, the third residue of the TTG motif, also faces this cavity (Fig. S8B, left). This unique hydrophilic cavity would be related with the direct Cl⁻ exchange between the Cl⁻-binding site near the retinal chromophore and the cytoplasmic solvent during Cl⁻ release, and the small volume of G93 due to the lack of its side chain is essential to keep the Cl⁻-release pathway opened. G84, the third residue of the DTG motif in PspR, faces a similar cavity, which forms the H⁺ uptake pathway from the cytoplasmic solvent to the RSB (Fig. S8B, right) (44). Hence, the cytoplasmic glycine residue has possibly evolved for the direct ion communication between the retinal-binding site and the cytoplasmic solvent in both Cl⁻ and H⁺ pumps. Interestingly, Cl⁻ interacts with a backbone carbonyl of glycine residue in CLC anion channel (45), indicating a similar interaction is formed between G93 and Cl⁻ in the L₃ intermediate of SsGHR.

Figure 7.

Summary.A, schematic model of the ion-transport mechanism of SsGHR (left) and NpHR (right) B, cavities (yellow) in SsGHR (green), NpHR (purple), MrHR (orange), and PspR (red). SsGHR G93 facing the cytoplasmic cavity is shown in a stick model.

SsGHR exhibits SO₄²⁻ transport that was known only for SyHR (20) and AaClR (23). The mutation of E198 located at the entrance of the extracellular cavity (Fig. 6A) reduced the SO₄²⁻ selectivity against Cl⁻. Hence, E198 plays an important role in SO₄²⁻ uptake (Fig. 6B and C). Interestingly, MrHR has a homologous glutamate (E182). But, mutation to an alanine or a threonine (MrHR E182A and E182T) confers substantial SO₄²⁻ transport ability to MrHR (39, 40), indicating that MrHR E182 inhibits SO₄²⁻ uptake in contrast to SsGHR E198. SyHR has a threonine (T183) at the same position and SyHR T183E mutant has low SO₄²⁻ selectivity (40). Hence, glutamate at this position can facilitate SO₄²⁻ uptake only in SsGHR. Although the reason for the different roles between SsGHR E198 and MrHR E182 is unclear, we propose different interactions with the arginine residues (SsGHR: R80, MrHR: R71). Whereas E182 does not interact with R71 in MrHR, E198 forms a salt bridge with R80 (Fig. S9), leading to a characteristic flipping of R80 observed only in SsGHR among anion-pumping rhodopsins (Fig. S6B). R80 flipping would change the anion uptake process from the extracellular side and possibly further open the extracellular cavity to accommodate a large sulfate SO₄²⁻ ion. On the other hand, since the distance between R80 and the Cl⁻ (9.3 Å) is shorter than that between E198 and the Cl⁻ (13.5 Å), the positive charge of R80 is still capable of contributing to the stabilization of the Cl⁻ binding.

All GHRs reported in this study were found in soil bacterial species. Hence, the characteristic TTG motif and large cytoplasmic cavities would be related to a non-aquatic environment. Although the physiological significance of GHR is unknown, anion-pumping rhodopsins are proposed to play a role in the regulation of cytoplasmic osmotic pressure. Specifically, when the soil where S. soli resides dries out, the Cl⁻ transport by SsGHR may help prevent water efflux. The physiological relevance of SsGHR in S. soli will be clarified by investigating the light-dependent phenotype of native bacterial cells. Our study revealed the mechanism of Cl⁻ release and SO₄²⁻ transport in SsGHR is distinct from that in other HRs. The detail of their structural dynamics and the importance of the cytoplasmic cavity including G94 for the Cl⁻ release can be revealed by time-resolved serial femtosecond X-ray crystallography (46) and cryo-electron microscopic observation for the cryo-trapped photointermediates (47). The structural analysis of the photointermediates will also provide insights into the specific interactions between the protein and SO₄²⁻, which make the SO₄²⁻ transport more efficient compared to the transport of I⁻/NO₃⁻ ions.

Experimental procedures

Phylogenetic analysis of microbial rhodopsins

For the phylogenetic analysis of microbial rhodopsins, the amino acid sequences of 47 microbial rhodopsins (Fig. 1) were aligned using ClustalW (48). The evolutionary history was inferred using the Neighbor-Joining method (49). The optimal tree with the sum of branch length = 23.48112077 is shown. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Poisson correction method (50) and are in the units of the number of amino acid substitutions per site. The analysis involved 47 amino acid sequences. All ambiguous positions were removed for each sequence pair. There were a total of 1260 positions in the final dataset. Evolutionary analyses were conducted in MEGA X (51).

DNA plasmid construction for GHR expression

The genes coding GHRs and optimized for E. coli expression with specific codons were synthesized by Genscript. Subsequently, it was cloned into the pET21a (+) vector (Novagen, Merck KGaA) at the NdeI–XhoI site. The plasmid was transformed into E. coli C43 (DE3) strain cells (Lucigen). For mutagenesis, the standard protocol of the QuikChange site-directed mutagenesis method (Agilent Technologies) was followed using the primer sequences listed in Table S2.

Ion-transport activity assay

GHR-expressing E. coli cells were collected through centrifugation (4800g, 2 min, 20 °C; CF15RF, Eppendorf Himac Technologies) and then washed with unbuffered 100 mM salt solutions (NaCl, KCl, NaBr, NaI, NaNO₃, and Na₂SO₄). The cells were subsequently equilibrated three times with rotational mixing in unbuffered 100 mM respective salt solutions at room temperature for 10 min each time. After equilibration, the cells were suspended in 7.5 ml unbuffered 100 mM respective salt solutions, and the optical density at 660 nm (OD₆₆₀) was adjusted to 2. Since the Schiff-base linkage between the retinal and rhodopsin is stable, we did not add all-trans-retinal in the salt solutions used for the ion-transport activity assay. The cell suspension was then placed in the dark in a glass cell at 20 °C and exposed to λ = 530 ± 5 nm light from a 300 W xenon light source (MAX-350, Asahi Spectra) through a bandpass filter (HQBP530-VIS, Asahi Spectra) and a heat-absorbing filter (HAF-50S-50H, SIGMAKOKI). Light-induced pH changes were measured using a pH electrode (9618S-10D, HORIBA). The same measurements were repeated after adding 10 μM carbonyl cyanide m-chlorophenylhydrazone (CCCP) to quantitatively compare the ion-transport activity. To evaluate the expressed protein amount, the near-UV absorption of retinal oxime generated by the hydrolysis reaction between the RSB in the proteins and hydroxylamine was measured. The process involved washing E. coli cells expressing rhodopsins with a solution containing 133 mM NaCl and 66.5 mM Na₂HPO₄ (pH 7). The washed cells were then treated with 1 mM lysozyme and a small amount of DNase I for 1 h, followed by sonication to disrupt them. To solubilize rhodopsins, 3% DDM (ULTROL Grade; Merck KGaA) was added and the samples were stirred overnight at 4 °C. The rhodopsins were bleached with 500 mM hydroxylamine and illuminated with visible light (λ > 500 nm) from the output of a 300 W xenon lamp through a long-pass filter (Y-52, AGC Techno Glass) and a heat-absorbing filter. The absorbance changes were measured using a UV–vis spectrometer (V-750, JASCO) to determine the amount of rhodopsin expressed in E. coli cells. The relative ion-transport activities were normalized by the absorbance of expressed rhodopsins at 530 nm.

Protein expression and purification

E. coli cells harboring the SsGHR-cloned plasmids were cultured in 2 × YT medium containing 50 μg/ml ampicillin. The expression of C-terminal 6 × His-tagged proteins was induced by 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) in the presence of 10 μM all-trans-retinal (Toronto Research Chemicals) for 4 h at 37 °C. The harvested cells were sonicated (Ultrasonic Homogeniser VP-300N; TAITEC) for disruption in buffer containing 50 mM Tris–HCl (pH 8.0) and 5 mM MgCl₂. The membrane fraction was collected by ultracentrifugation (CP80NX, Eppendorf Himac Technologies) at 142,000g for 1 h. The proteins were solubilized in a buffer containing 50 mM MES–NaOH (pH 6.5), 300 mM NaCl, 5 mM imidazole, 5 mM MgCl₂, and 1% DDM (ULTROL Grade; Merck KGaA). The solubilized proteins were separated from the insoluble fractions by ultracentrifugation at 142,000g for 1 h. The proteins were purified using a Co-NTA affinity column (HiTrap TALON crude; Cytiva). The resin was washed with a buffer containing 50 mM MES–NaOH (pH 6.5), 300 mM NaCl, 50 mM imidazole, 5 mM MgCl₂, and 0.1% DDM. The proteins were eluted in a buffer containing 50 mM Tris–HCl (pH 7.0), 300 mM NaCl, 500 mM imidazole, 5 mM MgCl₂, and 0.1% DDM. The eluted proteins were dialyzed in a buffer containing 20 mM HEPES–NaOH (pH 7.0), 100 mM NaCl, and 0.05% DDM to remove imidazole. The sample for crystallization was applied to a size exclusion chromatography system (ÄKTA go, Cytiva) and a Superdex 200 10/300 Gl gel filtration column (GE Healthcare) equilibrated in a buffer containing 20 mM HEPES–NaOH (pH 7.0), 100 mM NaCl, and 0.05% DDM. The sample of peak fractions were concentrated to 50 mg/ml using a centrifugal filter device (Millipore, 50 kDa molecular weight cutoff).

HPLC analysis of the retinal chromophore

The HPLC analysis of retinal configuration was conducted using purified SsGHR in a 6-mix buffer (trisodium citrate, MES, HEPES, MOPS, CHES, CAPS (10 mM each, pH 7.0), 1 M NaCl, and 0.05% DDM). The samples were adjusted to an optical density (OD) of 1.73 at λ_max and kept at 4 °C for 3 days to ensure dark adaptation. HPLC equipped with a silica column (particle size 3 μm, 150 × 6.0 mm; Pack SIL, YMC), a pump (PU-4580, JASCO), and a UV–vis detector (UV-4570, JASCO) was used for the analysis. The solvent containing 15% ethyl acetate and 0.15% ethanol in hexane was flowed at 1.0 ml min⁻¹. The sample was denatured by adding 495 μl methanol to 5 μl sample. Retinal oxime formed by the hydrolysis reaction with 33 μl 2 M hydroxylamine was extracted with hexane and 200 μl solution was injected into the HPLC system. For analyzing light-adapted SsGHR, the sample solution was illuminated with λ > 500 nm (Y-52; AGC Techno Glass) for 1 min, incubated for 2 min in dark, and then denaturation of protein and hydrolysis reaction of retinal chromophore were carried out. The molar composition of the retinal isomers in the sample was calculated using the molar extinction coefficient at 360 nm for each isomer (all-trans-15-syn: 54,900 M⁻¹ cm⁻¹; all-trans-15-anti: 51,600 M⁻¹ cm⁻¹; 13-cis-15-syn: 49,000 M⁻¹ cm⁻¹; 13-cis-15-anti: 52,100 M⁻¹ cm⁻¹; 11-cis-15-syn: 35,000 M⁻¹ cm⁻¹; and 11-cis-15-anti: 29,600 M⁻¹ cm⁻¹ (52, 53)).

pH titration

To determine the pK_a of the Schiff base, the pH dependence of the absorption spectra of SsGHR was measured. The protein concentration was adjusted to approximately OD = 0.5 at λ_max and solubilized in a 6-mix buffer (trisodium citrate, MES, HEPES, MOPS, CHES, CAPS (10 mM each, pH 7.0) and 0.05% DDM) with and without 1 M NaCl. Absorption spectra of the sample were measured at different pH values by adjusting the pH using HCl and NaOH for the Cl⁻-bound form and gluconic acid (HOCH₂(CHOH)₄COOH) and NaOH for the Cl⁻-unbound form. The spectra were measured using a UV–vis spectrometer (V-750, JASCO, Japan) at intervals of 0.3 to 0.6 pH values. pH-dependent absorption changes at representative wavelengths were global-fitted by the Henderson–Hasselbalch equation (1) to determine the pK_a.

$$\Delta A=base+\sum _{i=1}^n{\Delta A}_{\mathit{max},i}\frac{1}{\begin{array}{c}1+{10}^{\left(\text{pH}-p{K}_{a,i}\right)}\end{array}}$$ (1)

where ΔA: absorption change, ΔA_max,i: maximum absorption change of the i-th component, pK_a,i: acidity constant of the i-th component (i = 1–n).

Anion titration

To determine the K_d of anions in SsGHR, the anion dependence of the absorption spectra of SsGHR was measured. The protein concentration was adjusted to approximately OD = 0.5 at λ_max and solubilized in an anion-free buffer (20 mM HEPES–NaOH (pH 7.0)). Absorption spectra of the sample were measured at different anion concentrations by adding a buffer with NaCl, NaBr, or NaI. The spectra were measured using a UV–vis spectrometer (V-750, JASCO). Salt concentration-dependent absorption changes at representative wavelengths were global-fitted by the Hill equation (2) with the Hill coefficient n = 1 to estimate the K_d of anion X⁻.

$$\Delta A=\sum _{i=1}^n{\Delta A}_{\mathit{max},i}\frac{\left[{\mathrm{X}}^{-}\right]}{\begin{array}{c}{K}_{\mathrm{d},i}+\left[{\mathrm{X}}^{-}\right]\end{array}}$$ (2)

where ΔA: absorption change, ΔA_max: maximum absorption change of the i-th component, K_d: dissociation constant of the i-th component, [X⁻]: anion concentration (i = 1–n).

Laser flash photolysis

The laser flash photolysis system has been previously described (18, 54). The purified SsGHR WT and mutants were reconstituted into a mixture of 1-palmitoyl-2-oleoyl-phosphatidyl-ethanolamine (POPE, Avanti Polar Lipids) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG, sodium salt, Avanti Polar Lipids) (molar ratio = 3:1) with a protein-to-lipid molar ratio of 1:50 in 100 mM NaCl and 20 mM HEPES–NaOH (pH 7.0), and DDM was removed by BioBeads (SM-2 Bio-Red). The buffer of the sample solution was exchanged with 20 mM HEPES–NaOH (pH 7.0) buffer with 0.125, 0.5, 1, and 4 M NaCl to determine the [Cl⁻] dependence. The protein solution was adjusted to have an absorption of approximately 0.8 at the excitation wavelength (λ = 532 nm), with a total protein concentration of approximately 0.15 mg mL⁻¹. A second harmonic generation beam of a nanosecond-pulsed Nd:YAG laser (λ = 532 nm, 5.7 mJ cm⁻², 2–1.25 Hz) (INDI40, Spectra-Physics) was used to excite the sample. The transient absorption spectra were obtained by monitoring white-light intensity change from a Xe arc lamp (L9289-01, Hamamatsu Photonics) passed through the sample using an ICCD linear array detector (C8808-01, Hamamatsu). To increase the signal-to-noise (S/N) ratio, 60 to 100 spectra were averaged, and the singular-value-decomposition (SVD) analysis was applied. To measure the time-evolution of transient absorption change at specific wavelengths, the intensity change of the output of an Xe arc lamp (L9289-01, Hamamatsu Photonics, Japan), which was monochromated by a monochrometer (S-10, SOMA OPTICS, Japan) and passed through the sample solution, after photoexcitation was monitored using a photomultiplier tube (R10699, Hamamatsu Photonics, Japan) that was equipped with a notch filter (532 nm, bandwidth = 17 nm) (Semrock) to remove the scattered pump pulse. To increase the signal-to-noise (S/N) ratio, 100 to 200 signals were averaged. The signals were global-fitted with a multi-exponential function to determine the lifetimes of each photo-intermediate.

Raman spectroscopy

We conducted resonance Raman spectroscopy as described previously with some modifications (37). Purified SsGHR was solubilized to a concentration of 150 μM in a solvent containing 100 mM NaCl, 20 mM HEPES–NaOH (pH 7.0), 0.05% DDM. Then, 25 μl of sample solution was placed in a quartz cell with a 0.5 mm optical path length. A 561 nm diode-pumped solid state (DPSS) laser (MLL-U-561, Changchun New Industries Optoelectronics Tech) was used as a probe laser. CW laser light was focused on the sample through a cylindrical lens and an objective lens (NA 0.8). The intensity and spot size at the sample surface were 10 μW and 200 μm × 1 μm, respectively. During the measurements, the sample was swept at 1 cm/s using a stage. Rayleigh scattering light from the sample was removed by an edge filter at 561 nm (BLP02-561R-25, Semrock). The signal penetrating the filter was coupled with a 32-cm monochromator (IsoPlane 320, Teledyne Princeton Instruments) and a liquid-nitrogen-cooled charge coupled device (CCD) camera (Pylon:100B_eXcelon, Teledyne Princeton Instruments). The resolution was 2.5 cm⁻¹ and the measurements were performed at 23 °C.

SsGHR crystallization

Concentrated SsGHR (57 mg mL⁻¹ protein) was mixed with monoolein (Nu-Chek Prep, Elysian) at 2:3 (v/v) protein:lipid ratio using the lipidic cubic phase (LCP) method at 23 °C. The mixed sample was dispensed on glass sandwich plates in a 50-nL drop and overlaid with 800 nl reservoir solution comprising 0.1 M Li₂SO₄, 0.1 M NaCl, 0.1 M HEPES (pH 8.0), and 40% PEG-200 by Mosquito LCP (TTP Labtech Ltd, Melbourn, Hertfordshire). The crystals were harvested directly from the LCP bolus, flash-cooled, and stored in liquid nitrogen.

X-ray diffraction data collection and structure determination

X-ray diffraction data were collected from a single crystal at a cryogenic temperature (100 K) on BL-1A beamline (λ = 1.0520 Å) at the Photon Factory (Tsukuba, Japan). The collected data were processed to 2.63 Å using the XDS software (55). The structure was solved by molecular replacement with Phaser (56) as a search model for cyanobacterial chloride importer (PDB ID code: 6K6K). The atomic model was built using Coot (57) and iteratively refined using the Phenix (58). Translation/Libration/Screw (TLS) refinement was performed in the late stages of refinement. The refined structures were validated using the RAMPAGE software (59). The statistics of the final refined model are given in Table S3.

Data availability

Data supporting the findings are available from the corresponding authors upon reasonable request. The atomic coordinates and structure factors of SsGHR were deposited in the Protein Data Bank under the accession code 8XX8 (https://doi.org/10.2210/pdb8xx8/pdb).

Supporting information

This article includes supporting information (17, 19, 20, 42, 60).

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Wolfgang Peti