Characterization of a Cyanobacterial Chloride-pumping Rhodopsin and Its Conversion into a Proton Pump*

Light-driven ion-pumping rhodopsins are widely distributed in microorganisms and are now classified into the categories of outward H+ and Na+ pumps and an inward Cl− pump. These different types share a common protein architecture and utilize the photoisomerization of the same chromophore, retinal, to evoke photoreactions. Despite these similarities, successful pump-to-pump conversion had been confined to only the H+ pump bacteriorhodopsin, which was converted to a Cl− pump in 1995 by a single amino acid replacement. In this study we report the first success of the reverse conversion from a Cl− pump to a H+ pump. A novel microbial rhodopsin (MrHR) from the cyanobacterium Mastigocladopsis repens functions as a Cl− pump and belongs to a cluster that is far distant from the known Cl− pumps. With a single amino acid replacement, MrHR is converted to a H+ pump in which dissociable residues function almost completely in the H+ relay reactions. MrHR most likely evolved from a H+ pump, but it has not yet been highly optimized into a mature Cl− pump.

represented by visual pigments in the eyes. The other is microbial rhodopsin, which shows divergent functions in unicellular microorganisms, such as light-driven ion pumps and light sensors for phototaxis and the regulation of gene expression. Different from animal rhodopsins, the microbial sensors represent a minority of the microbial rhodopsins. However, microbial sensors have individualities in their signaltransduction modes, which include interactions with other membrane proteins, interactions with cytoplasmic components, and light-gated ion channel activity.
Microbial rhodopsins were originally discovered in highly halophilic archaea in the early 1970's -1980's (3). Since 1999, their homologues have begun to be identified in various microorganisms inhabiting a broad range of environments (4)(5)(6). It is now clear that in the microbial world, ion-pumping rhodopsins are abundant and widely distributed. In microorganisms, these ion pumps are probably the most convenient system for light energy utilization. The first and second ion pumps to be discovered were bacteriorhodopsin (BR) (7) and halorhodopsin (HR) (8) from halophilic archaea, which are an outward H + pump and an inward Clpump, respectively. In 2000, an outward H + pump named proteorhodopsin (PR) was discovered in a marine proteobacterium (9). Later, the PR homologues were identified in eubacteria living in the oceans worldwide (10). Recently, two groups of ion pumps, the outward Na + pump (NaR) and inward Clpump (ClR), were also discovered in eubacteria (11,12). Their representatives are Krokinobacter eikastus rhodopsin 2 (KR2) (13) and Fulvimarina pelagi rhodopsin (FR) (14), respectively. The phylogenetic position of these ion pumps is shown in Fig 1A. In contrast to their functional diversity, these ion pumps have essentially the same structural folds composed of seven helices and the chromophore retinal, which binds to a conserved lysine residue via a protonated Schiff base (PSB; deprotonated Schiff base is abbreviated as SB). In the dark, most ionpumping rhodopsins dominantly contain retinal with the all-trans configuration, while some minorities such as BR and HR from archaeon Halobacterium salinarum (HsHR) can also accommodate 13-cis retinal (2). Regardless of this difference in the dark states, only the photoisomerization from all-trans to 13-cis can trigger the conformational changes for respective ion pumping functions. Thus, ion-pumping rhodopsins appear to share a common transport machinery, where the transportable ions are probably determined by essential residues at appropriate positions. This scenario was partially demonstrated in 1995, when the H + pump BR was converted to an inward Clpump by the single amino acid replacement of Asp85 with Thr, the corresponding amino acid in HR (15). However, the success of pump-to-pump conversion was confined to this BR case. The reverse conversion, that is, from a Clpump to a H + pump, has not been achieved (16)(17)(18) even after ten mutations of HR (18). Here, we report a new class composed of an inward Clpump and its conversion to an outward H + pump. Functional characterization was performed with a microbial rhodopsin from the cyanobacterium Mastigocladopsis repens, which was isolated from soil at Punta de la Mora, Tarragona in Spain. This microbial rhodopsin is designated as M. repens HR (MrHR) due to its similarities with HR. By a single amino acid replacement, MrHR begins to pump H + outwardly. Thus, this is the first successful conversion from an inward Clpump to an outward H + pump.

EXPERIMENTAL PROCEDURES
Phylogenetic Analysis-The 57 amino acid sequences were aligned using MUSCLE. All sequence data were obtained from the NCBI database. The phylogenetic tree was constructed by using the maximum likelihood method, with bootstrap percentages based on 1,000 replications. Initial trees for the heuristic search were obtained by applying the neighbor-joining method (19). Evolutionary distances were calculated with the JTT matrix-based method (20). The branch lengths denote the number of amino acid substitutions. All analyses were performed using MEGA6.
Gene preparation, protein expression and purification-Escherichia coli strain DH5α was used for DNA manipulation. An MrHR gene (NCBI accession ID: WP_017314391) with codons optimized for E. coli expression was chemically synthesized (Funakoshi, Japan) and inserted into the NdeI/XhoI site of the pET-21c(+) vector (Merck Japan, Tokyo, Japan). This plasmid results in MrHR with additional amino acids in the C-terminus (-LEHHHHHH). The T74D mutation was introduced using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA). The DNA sequences were confirmed by a standard method using an automated DNA sequencer (model 3100, Applied Biosystems, Foster City, CA). MrHR and the T74D mutant were expressed and purified from E. coli BL21(DE3) cells. The procedures were essentially the same as those previously described (21). The cells were grown at 37 ºC in 2×YT medium supplemented with 50 μg/mL ampicillin. At the late exponential growth phase, the expression was induced by the addition of 1 mM isopropyl-β-D-thiogalactpyranoside in the presence of 10 μM all-trans retinal. After 3 h of induction, pink-colored cells were harvested by centrifugation (6400 × g, 8 min at 4 °C) and washed once with buffer (50 mM Tris-HCl, pH 8.0) containing 5 mM MgCl 2 . Then, the cells were broken with a French press (Ohtake, Tokyo, Japan) (100 MPa × 4 times). After removing the undisrupted cells by centrifugation (5600 × g, 10 min at 4 °C), the supernatant was ultracentrifuged (178,000 × g, 90 min at 4 °C). The collected cell membrane fraction was suspended with the same buffer containing 300 mM NaCl and 5 mM imidazole and was then solubilized with 1.5% ndodecyl β-D-maltopyranoside (DDM) (Dojindo Lab, Kumamoto, Japan) at 4 °C overnight. After removal of the insoluble fraction by ultracentrifugation (178,000 × g, 60 min at 4 °C), the solubilized MrHR proteins were subjected to Ni-NTA agarose (Qiagen, Hilden, Germany). The unbound proteins were removed by washing the column with 10 column volumes of wash buffer (50 mM sodium phosphate, pH 7.5) containing 400 mM NaCl, 50 mM imidazole and 0.05% DDM. The bound protein was eluted with buffer (50 mM Tris-HCl, pH 7.0) containing 300 mM NaCl, 500 mM imidazole and 0.1% DDM. The yield of MrHR was 15 mg from 1 L culture. The concentration was determined from the absorbance at 537 nm under an assumed extinction coefficient of 40,000 M -1 cm -1 . The purified samples were replaced with an appropriate buffer solution by passage over Sephadex G-25 in a PD-10 column (Amersham Bioscience, Uppsala, Sweden).
Ion-pump activity measurements-The MrHR activity was measured in E. coli suspensions using a conventional pH electrode method (22), which detects the pH changes by the pump activity of H + itself or passive H + transfer in response to the membrane potential created by the pump activity for another ion. The E. coli suspensions were prepared as follows. The cells expressing MrHR were harvested at 3,600 g for 5 min at 4 ºC and washed twice with an unbuffered solution containing 200 mM salt (NaCl, Choline-Cl, NaBr, NaI or NaNO 3 ). They were resuspended in the same salt solution and gently shaken overnight at 4 ºC in the presence of 10 μM carbonyl cyanide m-chlorophenylhydrazone (CCCP). Then, the cells were washed twice with the same salt solution without CCCP and finally suspended at an A 660 of 0.5. This cell density was approximately 5% of the corresponding value for the original culture medium. As deduced from the purification yield, the original culture medium contains at least 15 μg/mL MrHR. Thus, the cell suspensions for the activity measurements contain at least 0.75 μg/mL of MrHR. For the activation of MrHR, 530 ± 17.5 nm green LED (LXHL-LM5C, Philips Lumileds Lighting Co., San Jose, CA) was used.
HPLC analysis-The retinal configurations of MrHR were examined in both of the dark/light-adapted states. For dark adaptation, the MrHR samples were kept in the dark for 1 week in 10 mM MOPS, pH 6.5, containing 100 mM NaCl and 0.05% DDM. For light adaptation, the samples were irradiated for 2 min by green LED light as described above. MrHR undergoes a prolonged photocycle as described below. To avoid the contamination of retinal in the photolyzed state, the retinal oxime extraction was carried out after 1 min incubation in the dark. The extraction and the following HPLC analysis was performed as previously described (22).
Absorption spectra measurements and Flash photolysis-UV-visible spectra of MrHR samples were measured with a UV-1800 spectrometer (Shimadzu, Kyoto, Japan). Flashinduced absorbance changes were obtained in the 5 μs to 10 s time range on a single wavelength kinetic system. For the photoexcitation, the second harmonic (7 ns, 532 nm) of a Q-switched Nd:YAG laser (Surelite I-10, Continuum, Santa Clara, CA) was used. The details have been previously described (21). To improve the S/N ratio, 30 laser pulses were used at each measurement wavelength. All measurements were performed at 25 ºC.

RESULTS AND DISCUSSION
Mastigocladopsis repens, isolated from the surface of soft powdery calcareous soil, is a blue-green alga belonging to the cyanobacterial order Stigonematales and was morphologically characterized according to the filaments it forms, which have many branches (23). In 2013, the existence of a gene encoding a microbial rhodopsin was revealed by whole genome sequencing of M. repens (24). We named this microbial rhodopsin "M. repens halorhodopsin (MrHR)" after its photochemical properties, described below. At present, six other homologues of MrHR are found in the NCBI database. Figure 1A shows the phylogenetic tree of microbial rhodopsins, which revealed that MrHR homologues constitute a new clade that is distinct from the known microbial rhodopsins. The six homologues are also encoded in cyanobacteria and share 63-89% amino acid identities with MrHR. Meanwhile, the host strains belong to different cyanobacterial orders (Chroococcales and Nostocales) from M. repens, reflecting morphological differences. The phylogenetic relationships of MrHR homologues are shown in Fig 1B together with the names, orders and sources of their host strains. In addition to M. repens, all other genomes were sequenced (24)(25)(26)(27)(28). Cyanobacteria are a genetically diverse group and are widespread in fresh water, marine, terrestrial and extreme environments such as hot springs and salt lakes. However, out of the seven strains of MrHR homologues, six were isolated from terrestrial environments, such as soil and rock scrapings, and biofilms on stone monuments and building facades ( Fig 1B) (23,29,30). Correspondingly, the desiccation resistance was experimentally confirmed for several strains (31). Two strains also contain the gene for Anabaena sensory rhodopsin (ASR), a putative sensor for chromatic adaptation that was originally discovered in the cyanobacterium Anabaena sp. PCC7120 (32). The other five strains do not contain other microbial rhodopsin genes.
Microbial rhodopsins are largely categorized into ion pumps and sensors. All ion pumps are solely encoded in their respective operons in the genomes, whereas all sensors are encoded to be adjacent to the genes encoding the cognate transducers within the same operons. For MrHR homologues, their genes are solely encoded, implying that they function as ion pumps. Previous studies on ion pumps (12, 13) have indicated that three amino acids are responsible for determining the transportable ions (Fig 2A and B). For H + pumps, a H + from PSB is translocated during the photoreaction. This H + is initially transferred to D85 in BR (D97 in PR), and SB subsequently accepts a H + from D96 in BR (E108 in PR) (Fig 2A). These residues are often referred to as the H + acceptor and donor, respectively. For Clpumps, the acceptor is replaced with a 'T' in HR and 'N' in FR (Fig 2B), which enables the binding of substrate Clas the counter ion of the PSB. Furthermore, for HR and KR2, another residue corresponding to T89 in BR is known to play a crucial role (13,21): These residues are 'S' in HR and 'D' in KR2, respectively (Fig 2B). Thus, three residues (D85, T89, and D96 in BR) are now designated as the "motif": These are DTD and DTE for BR and PR, TSA for HR, and NTQ and NDQ for FR and KR2 (Fig 2B). For MrHR, the motif is TSD (Fig  2A and B), which is close to TSA in HR. This implies that MrHR may pump Clinwardly, despite a low amino acid identity of 22% for HR and 12% for FR. The identities for the H + pumps are also low: 29% for BR and 17% for PR. However, the donor in the H + pumps is conserved as 'D' in MrHR, similar to BR(D) or PR(E). Furthermore, MrHR conserves E194 and E204 in BR, which consists of the H + releasing complex to the extracellular medium ( Fig 2B). Thus, MrHR conserves the residues that are characteristic of both the Clpump (HR) and the H + pump (BR and PR).
In this study, we expressed MrHR in the cell membrane of Escherichia coli and investigated the ion-pumping activity using the light-induced pH changes of the cell suspensions (Fig 3). In a NaCl solution, light-induced alkalization was observed. This pH change was not abolished by the addition of the protonophore CCCP, which eliminates the electrochemical gradient of the proton (33). This means that the alkalization was not caused by the active proton transport but was caused by the passive proton influx in response to the interior negative membrane potential, which should be created by outward Na + or inward Cltranslocation. In Choline-Cl, alkalization was also observed. Because choline is a large organic cation, microbial rhodopsin cannot transport it. This indicates that MrHR functions as a light-driven inward Clpump. The pump activity decreased in NaBr and disappeared in NaI and NaNO 3 . Thus, MrHR does not pump Na + but does pump smaller anions. The transportable anions are severely restricted to only Cland Br -, different from HR and FR, which can even transport NO 3 -(14,33). The retinal isomer composition of MrHR was examined by HPLC analysis (Fig. 4). As described above, BR and HsHR in the dark states can accommodate both all-trans and 13-cis retinals and show so-called "light-dark adaptation" (2). In their light-adapted states after continuous illumination, the unphotolyzed states predominantly contain all-trans retinal. Upon dark adaptation, their 13-cis contents increase approximately 50%. For MrHR (Fig 4), the isomer composition does not depend upon the dark/light adaptation and is predominantly alltrans, similar to most ion-pumping rhodopsins. Thus, the all-trans retinal is responsible for the ion-pumping function of MrHR.
Anion binding near the PSB was monitored by the color change of retinal using purified MrHR (Fig 5A). For other Clpumps, Clbinding causes a blueshift (14,34), except for HsHR (22). In contrast, a large redshift occurs for MrHR by its binding of Cl -, suggesting the differences in the Clbinding site from other Clpumps. Similar redshifts were also observed for Brand even for Iand NO 3 -. The dissociation constants (K d ) were determined from the absorbance changes at 550 nm (Fig 5B), and the results are summarized in Table 1. The K d values are close to those of HR, except for NO 3 -(171 mM): HR binds NO 3 more strongly (K d ~ 16 mM) (35), whereas FR binds these ions more weakly (K d = 40 -130 mM) (14). These results indicate that MrHR can bind Iand NO 3 but cannot transport them. These larger ions may be impossible to move toward a cytoplasmic half channel over the PSB region. Compared with HR and FR, the ion translocation pathway might be narrow and act as an ion-selective filter.
We also characterized the photocycle of MrHR by a flash photolysis method. Timedependent absorbance changes at selected wavelengths are shown in Fig 5C,  The slow photocycle of MrHR might reflect some differences in the physiological role from other Cl --pumping rhodopsins. Although the details are not fully resolved, the Clpumps are believed to play roles in light-driven ATP production and/or maintaining the cellular osmotic balance during the volume increase of the growing cell (38)(39)(40)(41). All host strains of MrHR homologues contain a photosynthetic apparatus. Thus, the slow photocycle of MrHR seems to contribute little to cellular ATP production under illumination. Instead, MrHR might relate to a regulation of osmotic pressure. Unlike aquatic cells harboring other Clpumps, most host strains of MrHR homologues inhabit terrestrial and non-aquatic environments, where the cells are occasionally exposed to drought stress. Thus, the cells might utilize MrHR homologues to survive in the non-aquatic habitats. Under desiccated conditions, the internal salt concentration should be increased to preserve the intracellular water content. Due to the interior negative membrane potential, cations could be passively transported inside. However, anions need an active transport system, such as MrHR. Thus, a simple scenario might be that MrHR elevates the intracellular osmotic pressure, which in turn preserves the intracellular water content under drought conditions. For further discussion, we must await future investigations.
MrHR conserves the H + donor residue in the H + pump, which is an essential residue in the H + pump. Thus, we attempted the conversion of the MrHR to a H + pump; a mutant T74D was made in which the Asp residue was introduced as a possible proton acceptor. The λ max of T74D is located at 522 nm ( Fig 6A) and did not change upon the addition of Cl -. This indicates that there is no chloride binding near the PSB. The HPLC analyses showed that the retinal isomer composition is predominantly all-trans ( > 95%) in both dark and light-adapted states, similar to wild-type MrHR. Next, we examined the ionpumping activity using the pH change of the E. coli suspension. The results are shown in Fig 6B, which indicates the opposite pH change to the wild-type MrHR. Even in the NaCl solution, light-induced acidification was observed, showing proton extrusion. This pH change was decreased by the addition of CCCP, indicating that T74D actively pumps protons outward. Thus, a successful conversion from a Clpump to a H + pump was effected by the single amino acid replacement. Corresponding to this conversion, T74D undergoes a totally different photocycle from that of wild-type MrHR (Fig 6C). Specifically, the M-formation (400 nm) was observed, indicating that the introduced Asp residue acts as a H + acceptor from PSB. For HR, a T-to-D mutation was never observed to induce M-formation (16)(17)(18). For natural H + pumps in the dark, SB is protonated while the acceptor is deprotonated because the SB has a larger pKa than the acceptor. M-formation requires the inversion of the pKa values: The pKa of the acceptor must become larger than that of SB. T74D accomplishes these pKa changes. Following M decay, two intermediates appear sequentially, corresponding to N (470 nm) and O (590 nm) in BR. For H + pumps, the donor facilitates both M-N and N-O transitions because the M-N transition reflects proton movement from the donor to SB, and the subsequent N-O transition reflects proton uptake from the cytoplasmic medium by the donor. Thus, the donor-lacking mutants of H + pumps undergo substantially slower transitions after M formation (42,43). Interestingly, both transitions in T74D are fast compared with natural H + pumps, indicating that the donor functions very well. Moreover, M, N and O appear sequentially without quasi-equilibrium. This reflects the strict "accessibility switch" of the donor, which communicates with SB during the M-N transition; then, the accessibility switches to the cytoplasmic medium to facilitate the N-O transition. In a H + pump, this switch contributes to the one-way (irreversible) H + transport (42,43). These facts indicate that T74D contains the structural factors as well as the essential residues for an effective H + pump.
The overall photocycle of T74D resembles that of natural H + pumps but is prolonged due to the last intermediate (520 nm), which resembles MrHR' in the wild-type photocycle. A similar intermediate was also observed in PR (44), but MrHR' in T74D has a substantially long lifetime (~3 s). Another difference from a natural H + pump is the formation of an unknown intermediate (hereafter called X) at approximately 440 nm. X is formed almost simultaneously with the M-decay, but it decays independently of N and O. Thus, X probably forms from M by a branching process. SB reprotonates during the M-decay. For the M-N transition, the donor provides this proton. For the M-X transition, the proton is probably provided through another pathway, implying no contribution of X to the H + pumping activity. In natural H + pumps, the antecedent deprotonation of SB triggers the accessibility switching of SB from the acceptor to donor, which enables the exclusive H + transfer from the donor to SB (42,43). In T74D, this switching might not be stringently controlled.
MrHR functions as an inward Clpump but undergoes a prolonged photocycle. The donor residue is not essential for Cl --pumping activity, which was confirmed by the alanine replacement mutant (data not shown). In contrast, in T74D, both the introduced acceptor and the conserved donor function are comparable with those in natural H + pumps, which suggests that MrHR evolved from a H + pump, but the residues and the structure have not been optimized into a mature Clpump. The mutant T74D has reduced H + pumping activity as compared to natural H + pumps because of the slow decay of the last intermediate and the formation of the X intermediate. These weaknesses are probably related to the lost mechanisms, which are important for a H + pump but not for a Clpump. Thus, comparable studies of MrHR, Clpumps and H + pumps might provide insight into how sophisticated Clpumps evolved from H + pumps. Their NCBI accession IDs are indicated in the parentheses. The strains labeled with closed circles also contain the ASR gene. Other symbols indicate the cyanobacterial orders and the sources of the host strains. In both trees, nodes with less than 50% bootstrap percentages are collapsed.      Dissociation constants (K d ), absorption maxima (λ max ) and their shifts (Δλ max ) by the anion bindings a K d s were determined by fitting analyses using Hill equation with n = 1. The best-fit curves are shown in Fig 5B. b λ max s were the determined values at the following anion concentrations: Cl -, 0.5 M; Br -, 0.5 M; I -, 0.5 M; NO 3 -, 2 M. At these concentrations, the λ max shifts were almost saturated. c Δλ max s denote the λ max shifts from the anion-free state (506 nm).