The crystal structures of a chloride-pumping microbial rhodopsin and its proton-pumping mutant illuminate proton transfer determinants

Microbial rhodopsins are versatile and ubiquitous retinal-binding proteins that function as light-driven ion pumps, light-gated ion channels, and photosensors, with potential utility as optogenetic tools for altering membrane potential in target cells. Insights from crystal structures have been central for understanding proton, sodium, and chloride transport mechanisms of microbial rhodopsins. Two of three known groups of anion pumps, the archaeal halorhodopsins (HRs) and bacterial chloride-pumping rhodopsins, have been structurally characterized. Here we report the structure of a representative of a recently discovered third group consisting of cyanobacterial chloride and sulfate ion-pumping rhodopsins, the Mastigocladopsis repens rhodopsin (MastR). Chloride-pumping MastR contains in its ion transport pathway a unique Thr-Ser-Asp (TSD) motif, which is involved in the binding of a chloride ion. The structure reveals that the chloride-binding mode is more similar to HRs than chloride-pumping rhodopsins, but the overall structure most closely resembles bacteriorhodopsin (BR), an archaeal proton pump. The MastR structure shows a trimer arrangement reminiscent of BR-like proton pumps and shows features at the extracellular side more similar to BR than the other chloride pumps. We further solved the structure of the MastR-T74D mutant, which contains a single amino acid replacement in the TSD motif. We provide insights into why this point mutation can convert the MastR chloride pump into a proton pump but cannot in HRs. Our study points at the importance of precise coordination and exact location of the water molecule in the active center of proton pumps, which serves as a bridge for the key proton transfer.

Microbial rhodopsins are versatile and ubiquitous retinalbinding proteins that function as light-driven ion pumps, lightgated ion channels, and photosensors, with potential utility as optogenetic tools for altering membrane potential in target cells. Insights from crystal structures have been central for understanding proton, sodium, and chloride transport mechanisms of microbial rhodopsins. Two of three known groups of anion pumps, the archaeal halorhodopsins (HRs) and bacterial chloride-pumping rhodopsins, have been structurally characterized. Here we report the structure of a representative of a recently discovered third group consisting of cyanobacterial chloride and sulfate ion-pumping rhodopsins, the Mastigocladopsis repens rhodopsin (MastR). Chloride-pumping MastR contains in its ion transport pathway a unique Thr-Ser-Asp (TSD) motif, which is involved in the binding of a chloride ion. The structure reveals that the chloride-binding mode is more similar to HRs than chloride-pumping rhodopsins, but the overall structure most closely resembles bacteriorhodopsin (BR), an archaeal proton pump. The MastR structure shows a trimer arrangement reminiscent of BR-like proton pumps and shows features at the extracellular side more similar to BR than the other chloride pumps. We further solved the structure of the MastR-T74D mutant, which contains a single amino acid replacement in the TSD motif. We provide insights into why this point mutation can convert the MastR chloride pump into a proton pump but cannot in HRs. Our study points at the importance of precise coordination and exact location of the water molecule in the active center of proton pumps, which serves as a bridge for the key proton transfer.
One of the foremost challenges in understanding functional diversity of microbial rhodopsins is dissecting their complex structure-function relationships (1)(2)(3)(4)(5). Despite the distinct functional roles and different amino acid sequences, microbial rhodopsins have surprisingly similar structures, especially in their transmembrane domains frugally recycled by nature. It turns out that the functional differences among microbial rhodopsins are regulated by relatively small variations in the side chains, internal water molecules, and bound ions, which necessitate the acquisition of high-resolution structural data to probe structure-function relationships. This is particularly true for ion-pumping rhodopsins, a structurally uniform group with various ion specificities, whereas greater structural diversity has been seen in sensory rhodopsins, enzymerhodopsins, and channelrhodopsins (3,6).
Ion-pumping microbial rhodopsins are widespread among Eukarya, Bacteria, and Archaea (7-9). These microbial rhodopsins share a similar structural template consisting of seven tightly bundled transmembrane a-helices with the N and C termini located outside and inside the cell, respectively (Fig. 1). The all-trans-retinal chromophore is covalently bound through a protonated retinal Schiff base (PRSB) linkage to the e-amino group of a lysine residue in the middle of the seventh helix. For ion pumps, the absorption of a photon triggers the retinal chromophore to isomerize from all-trans-to 13-cis-configuration, which induces a photocycle involving protein structural changes to pump an ion through the membrane against a concentration gradient. Despite their similar overall architecture, ion specificities and transport vectorialities of these rhodopsins can be adjusted by fine-tuning their structures (10,11), yielding proteins able to pump H 1 outwards or inwards, Na 1 outwards, or Cl -(and some other anions) inwards.
Lately, a motif of three conserved residues on the third transmembrane helix (helix C) has been adopted as a function determinant for ion-pumping rhodopsins (12,13). Archaeal (bacteriorhodopsins or BRs) and fungal outward H 1 pumps usually have an Asp-Thr-Asp (DTD) motif, whereas bacterial outward H 1 pumps more commonly have either a DTE, DTK, or DTG signature representing the proton acceptor, the acceptor hydrogenbonding partner, and the cytoplasmic proton donor (or its nonprotonatable homolog), respectively. Light-driven retinal-binding Na 1 pumps possess an NDQ motif, in which the three residues play critical roles in binding transported Na 1 and in the accompanying critical H 1 transfers (14).
Unlike the above-mentioned subfamilies of cation pumps, anion transporters do not have a common function-determining motif and are represented by at least three divergent groups. Long-known archaeal Clpumps (halorhodopsins or HRs) have a TSA motif, whereas recently discovered eubacterial chloride-pumping rhodopsins (ClRs or NTQ rhodopsins) possess an NTQ motif; however, in both groups the first two amino acids of the function-determining motif bind a chloride anion in the dark (15)(16)(17)(18)(19). Additionally, there is a novel distinct anion-pumping cyanobacterial rhodopsin group containing both chloride and sulfate transporters, which have sequence similarity both to HRs and BRs (20)(21)(22)(23). Similar to HRs, this group possess threonine and serine as the first motif-forming residues, but it has aspartate, valine, leucine, or isoleucine as its third member (TSD/TSV/TSL/TSI), along with a large number of unique polar residues elsewhere. The first characterized member of this group is Mastigocladopsis repens rhodopsin (MastR or MrHR), which pumps chloride ions inwards through a function-determining TSD motif (20,23).
Although high-resolution structures of representative HRs (16,17) and ClRs (18,19) are available, the third (cyanobacterial) group of chloride-pumping rhodopsins has not been characterized structurally. Similar to proton-pumping rhodopsins, comparison of structures of divergent chloride-pumping rhodopsins with the same biological function (i.e. the same ion specificity) could help in defining common structural determinants of this function and its group-specific fine-tuning elements. Toward this goal, here we provide the high-resolution crystal structure of the chloride-pumping MastR, which reveals the unique fine structural elements of this novel anion transporter along with the common HR-like chloride-binding motif in the context of a predominantly BR-like structural template.
An alternative, elegant approach to understanding key structural determinants of the functions of microbial rhodopsin ion pumps is their functional conversion, in which a WT rhodopsin of chosen specificity is mutated to obtain a different functionality. Functional conversion of ion pumps has been achieved with varying success through site-directed mutagenesis, usually tar-geting the conserved three amino acids at the function-determining motif (10,11,20,24).
In this paper, we compare the X-ray crystallographic structures of chloride-pumping MastR and its proton-pumping T74D mutant (Figs. S1-S3) in the context of other chloride and proton pumps with known high-resolution structures. The structures suggest that the hydrogen bonds of the PRSB and adjacent water were both strong for MastR and MastR-T74D, supporting the hypothesis that outward proton transport requires both a strongly hydrogen-bonded water and a strongly hydrogen-bonded Schiff base. It appears that the BR-like structure of MastR, which ensures proper positioning of the water molecule interacting with the Schiff base, allows for its easy functional conversion to a proton pump, in contrast to HRs in which this is not possible, thereby revealing important structural prerequisites for outward proton pumping by microbial rhodopsins.

Results and discussion
Crystal structure of MastR We purified MastR and the mutant MastR-T74D after overexpression in Escherichia coli (Fig. S4) and crystallized the proteins using the bicelle crystallization method (25). We obtained hexagonal crystals (Fig. S5, A and B) in which trimers of three parallel monomers arrange in a hexagonal pattern (Fig. S6, C and G) to yield layers that stack directly on top of each other (Fig. S6, D and H). Within each layer, MastR packs as parallel trimers (Fig. S6C), and MastR-T74D packs as antiparallel trimers (Fig. S6G), respectively. In the stacking pattern, MastR forms stacked alternating layers (Fig. S6D), whereas MastR-T74D forms stacked parallel layers (Fig. S6H). This difference in packing results in MastR possessing an asymmetric unit with two rhodopsins connected via their extracellular loops, whereas MastR-T74D contains only a single protein in the asymmetric unit ( Fig. 2A). We determined the cryogenic X-ray crystal structures of MastR and MastR-T74D at resolutions of 2.33 and 2.50 Å, respectively (Fig. 2, Fig. S2, and Table S1).
Like other microbial rhodopsins, MastR has seven transmembrane helices (helices A-G), connected by three intracellular and three extracellular loops ( Fig. 2A). The intracellular loop connecting helices B and C (B-C loop) forms a short antiparallel b-sheet. The structure contains 45 water molecules, two chloride ions, a glucose moiety from octylglucoside detergent, and the retinal chromophore. The electron density of the chromophore revealed retinal in the all-trans-configuration (Fig. 2B), covalently bound to the e-amino group of Lys-204 on helix G via a PRSB, in agreement with prior Raman spectroscopy experiments (23). The MastR and MastR-T74D structures are almost identical with an RMSD of 0.26 Å and differ mainly in the PRSB region around the site of the mutation and the presence or absence of Clions for the WT and mutant, respectively (Fig. 2, B and C).

Chloride ion-binding sites
The MastR structure further shows two chloride ion-binding sites: one adjacent the PRSB and the other forming a crystal contact between the B-C loops of two stacked proteins. The Shown is a seven-transmembrane helix (denoted A-G) structure of microbial rhodopsin with all-trans-retinal chromophore bound covalently to a conserved lysine residue (cyan) on helix G. Three amino acids on helix C serve as a functional indicator on whether the protein pumps H 1 or Na 1 outward (orange and yellow arrows, respectively) or Clinward (green arrow). The motif derives from key residues in the BR proton-pumping photocycle, with Asp-85 (acceptor of Schiff base proton), Thr-89, and Asp-96 (proton donor for Schiff base) forming the DTD motif. For pumping of H 1 , Na 1 , and Cl -, different motifs are employed as indicated. Cyanobacterial chloride pump MastR possesses a TSD motif, which was mutated to DSD in this study to convert the MastR chloride pump into a proton pump.
primary Clion adjacent the PRSB is well-ordered between a single water molecule and Thr-74 and Ser-78 of the TSD motif as shown by the electron density in Fig. 2B. The Clion is the counterion of the positively charged PRSB and connected to it via a bridging water molecule. This chloride ion-binding site can also be occupied by bromide and iodide but not fluoride, as shown by UV-visible absorption ( Fig. S7A) (20). The T74A mutation was previously shown to strongly reduce chloride affinity (from K D of 2 to 85 mM) but not impede Clion transport, whereas the S78A mutation decreased the pK a and stability of the PRSB (23), and the S78T mutation decreased chloride affinity dramatically (almost 200-fold) (26). This agrees with the structure, which confirms that both Ser-78 and Thr-74 are involved in Clion binding, similar to HR (17).
The second chloride ion-binding site is formed by the extracellular sides of two MastR protomers ( Fig. 2A). A Clion is found to bridge two B-C loop b-sheets by interacting with the side-chain nitrogen of Asn-60 and backbone nitrogen of Val-61 of both proteins. This well-ordered Clion appears to be a crystal-stabilizing contact and may not be a functionally relevant Clion as concluded from bromide-replacement experiments (Fig. S8) and the MastR-T74D structure. Because MastR-T74D packs as stacked parallel layers, the B-C loops do not get into contact, and the lack of electron density for a Clion near the Key residues have been labeled with the mutated residue highlighted in red. In MastR, a Clion (green sphere) is part of an extended hydrogen-bonding network including Thr-74, Ser-78, a single water molecule (red sphere), and the PRSB. In the proton-pumping T74D mutant, the Clion is lacking, which rearranges the hydrogen-bonding network. 2F c 2 F o electron density maps are contoured at 0.8 s. The direction of ion transport is indicated.
B-C loop indicates low chloride affinity when no crystal contact between B-C loops occurs (Fig. S6H).

Oligomerization of MastR
Trimers of MastR are formed by interaction of helix B of one protomer with helices D9 and E9 of another protomer (Fig. 3A). The interface is stabilized by three polar contacts: (i) the backbone oxygen of Leu-44 with the hydroxyl group of Tyr-106, (ii) a water-bridged hydrogen bond between Thr-37 and Ser-99, and (iii) a water-bridged hydrogen bond between Glu-30 and the backbone nitrogen of Leu-94. Main hydrophobic interactions at the intermonomer interface are provided by Phe-41 on helix B, Tyr-106 on helix D9, and Trp-125 on helix E9 (Fig. 3A).
The oligomeric structure of microbial rhodopsins in solution can also be determined by visible CD spectroscopy because the exciton coupling of the retinal chromophores gives rise to distinct visible CD curves (27). Monomeric rhodopsins do not possess exciton coupling and only display a single positive peak. Trimeric rhodopsins exhibit bilobed shape spectra with a positive peak at the short-wavelength side of the l max and a negative peak on the long-wavelength side of l max . In pentameric rhodopsins, the different mutual orientation of the retinals gives rise to an inverted bilobe shape, with the negative and positive peak on the short-and long-wavelength sides of l max , respectively. We measured the visible CD spectra of detergent-solubilized MastR, which showed a positive peak at ;510 nm and a negative peak at ;560 nm relative to its l max of 540 nm, indicating that trimers also form in n-dodecyl b-D-maltopyranoside (DDM) micelles (Fig. S9).
The trimeric assembly found in the MastR crystal is commonly observed in crystals and purple membranes of archaeal rhodopsins such as BR (28,29), HR (30), and archaerhodopsin (31). A BR-like trimer was also found for Anabaena sensory rhodopsin (ASR) by solid-state NMR (32). Interestingly, the interactions at the MastR intermonomer interface appear to be very strong, like for the ASR trimer (32) and are stronger than for BR because ASR and MastR remain as a trimer in detergent (Fig. S9), whereas BR dissociates into monomers (33). This might explain why in similar bicelle crystallization conditions MastR formed a trimeric assembly, whereas BR crystallized as monomer (25).
The BR-like quaternary structure of MastR is in contrast to pentameric assemblies found for eubacterial ClRs such as chloride-pumping Fulvimarina rhodopsin (FR) (34), for which atomic force microscopy revealed a pentamer (27), as well as eubacterial sodium NDQ pumps (35) such as Krokinobacter eikastus rhodopsin 2 (KR2) (36) and eubacterial proton pumps Gloeobacter rhodopsin (GR) (37) and proteorhodopsin (PR) (38). The crystal structures of the eubacterial Nonlabens marinus ClR (NmClR/NM-R3) (18,19) and sodium pumps (KR2) (39,40) have as a common feature an extended helix B and a 3omega motif, which result in a flip of the B-C loop in the direction of helices A and B (Fig. 3B). It has been suggested that the flipped orientation of the B-C loop supports the specific pentameric assembly in eubacterial pumps (37). MastR does not have these pentamer-inducing structural features and exists as a trimer accordingly.
When the structures of known chloride pumps are compared with MastR and BR, major differences are noticeable at the extracellular side where the Clion entrance is located (Fig. 4). Although the transmembrane regions of BR, HRs, and MastR all look very similar (but quite different from ClR), the length of the B-C loops and size of the b-sheets varies largely, leading to different shapes of the chloride entrance. An ion entrance pore is developed best for MastR as indicated by the solvent-accessible inlet shown in Fig. 4. This is due to MastR having the shortest B-C loop, even shorter than the proton pump BR with which MastR shows best overall structural homology on the extracellular side. In contrast, HRs have long B-C loops that cover most of the extracellular surface, and only small solventaccessible areas are found that serve as inlets for ions. Natronomonas pharaonis halorhodopsin (NpHR) has the longest B-C loop, which together with an additional N-terminal helix running parallel the membrane forms a hydrophobic cap that was proposed to prevent a rapid exchange of charged ions between the active center and the extracellular medium (16). A histidine residue in the middle of HsHR's long B-C loop is critical for the anion-pumping mechanism (41). The B-C loop thus appears to be an element that chloride pumps use to regulate the entrance of anions.

Putative chloride ion transport pathway
The putative chloride pathway is shown in Fig. 5. In the ground state MastR structure, there is an open tunnel leading into a water-filled cavity. The tunnel is located between the B-C and F-G loops and is likely to serve as the chloride entrance. This tunnel is pinched by Glu-192 and Glu-182, which may act as a regulatory gateway, similar to the proton release pathway regulated by the homologous Glu-204 and Glu-194 in BR (42). Similar to Glu-194 of BR, Glu-182 of MastR seems to be electrostatically coupled to the Schiff base region, because its replacement strongly affects chloride affinity (26). Its transient protonation in the photocycle may be required for chloride uptake. Additionally, a mutation of Glu-192 was shown to shift the N/O state equilibrium toward the N-state and slows down the photocycle (23), which suggests slower chloride uptake from the extracellular environment. This suggested entrance is in accord with the T193A and T197A mutant phenotypes (23), which show no photocycle changes, because these residues are not near the entrance cavity and are located between helices A and G.
In the dark, the Clion, stabilized by residues Thr-74 and Ser-78, together with a single water molecule, fills a small cavity close to the PRSB. After retinal photoisomerization, this Clion should move toward the cytoplasmic side, which has two small cavities, with only a single water molecule in each. We propose that the chloride is more likely to move toward the slightly larger cytoplasmic cavity, which is surrounded by polar residues His-166, Ser-203, and Trp-170, which are good candidates for binding chloride. This is in contrast with the other smaller cavity, which only possesses Cys-43 and Asn-39 as potential chloride interaction partners. It was previously shown that the H166A mutation results in much slower decay of the L2 and N states, which slows down the overall photocycle turn-over by a factor of .10 (23). Thus, His-166 is potentially part of the chloride release pathway, similar to the homologous conserved threonine of HRs (43). The His-166 forms a stacking interaction with Trp-170, which may also contribute to slower dynamics in the H166A mutant, because the homologous Trp residue is known to play a key role in isomerization-induced conformational changes in other microbial rhodopsins (44).
Next to the other cytoplasmic cavity, we observe a hydrogenbonding network comprising Asp-85, Asn-39, and Ser-211, which connect helix C, B, and G, respectively. A particularly strong interhelical hydrogen bond is observed between Asp-85 and Asn-39 (2.7 Å distance between the heavy atoms), which is consistent with the interaction of these residues suggested by FTIR results obtained earlier (23). In the same work, it was observed that Asp-85 deprotonates in response to the chloride translocation and must be reprotonated in the last step for the photocycle to complete. The requirement for Asp-85 reprotonation suggests that it may be used to prevent the backflow of chloride. From the structure, it seems likely that the protonation state of Asp-85 is regulated by its interaction with Asn-39 and Ser-211, which may change as a result of changing interhelical distances originating from light-induced conformational changes (such as helical tilts characteristic for rhodopsins).
Interestingly, FTIR revealed that there is an accompanying long-living perturbation of a buried cysteine (hydrogen bond weakening), which was not assigned conclusively. There are three cysteines in MastR: Cys-43, Cys-77, and Cys-130 (Fig. S10).
Cys-77 is located close to the function-determining motif forming Thr-74 and Ser-78 and may form a weak intrahelical bond with the carbonyl oxygen of Val-73 (distance between the heavy atoms is 3.6 Å). This bond may be further weakened upon structural perturbations accompanying Clion translocation toward the cytoplasmic side. Cys-130 is located next to the b-ionone ring of retinal and does not appear to be involved in hydrogen bonding. In view of the discussed transient deprotonation of the neighboring Asp-85-Asn-39 complex, it is also tempting to ascribe the cysteine perturbation found by FTIR to Cys-43, which shares the water cavity with Asn-39.
Finally, we propose that the Clion is released through the pathway delineated by Ser-211 and Thr-214. Although these residues are not essential for transport, the photocycles of S211A and T214A show an accumulation of the O state and Overlay of MastR (pink) with BR (PDB code 1c3w; RMSD, 0.72 Å), HsHR (PDB code 5ahy; RMSD, 1.08 Å), NpHR (PDB code 3a7k; RMSD, 1.05 Å), and NmClR (PDB code 5g28; RMSD, 7.15 Å) reveals that MastR has the best overlap with proton-pumping BR. Clions (green spheres) are only shown for the labeled pump. Cavities (red) and the extracellular solvent-accessible inlet (yellow) are traced out with spheres calculated by Hollow (65). MastR has short intracellular loops, especially the B-C loop (red for MastR and blue for others), which creates a solvent-accessible pore that permeates deep into the extracellular side. On the intracellular side, MastR has the fewest and smallest cavities of the proteins compared.
EDITORS' PICK: Structural insight into functional conversion of a Clpump slightly faster photocycle turnover. Moreover, Ser-211 is conserved among cyanobacterial ion pumps but is nonpolar in BR and HRs (23). Along the Ser-211 and Thr-214 trajectory are residues Thr-217 and Ser-155, which may assist in Clion transportation. A clear understanding of the chloride transport and pathway will, however, require structural information on the photocycle intermediates.

Retinal-binding pocket of MastR and MastR-T74D
The most significant structural differences between MastR and MastR-T74D occur adjacent to the retinal chromophore as shown in Figs. 2 and 6. In the MastR structure, a Clion is located between the side-chain hydroxyl groups of Thr-74 (d Cl-O = 3.1 Å) and Ser-78 (d Cl-O = 3.0 Å) of the TSD motif. This Clion is part of an extended hydrogen-bonding network in which the Clion is connected to the Thr-74 and Ser-78 side chains, as well as via an adjacent water molecule (d Cl-O = 3.1 Å) to the PRSB (d PRSB-H2O = 2.8 Å) and the carboxyl group of Asp-200. The Asp-200 and Arg-71 side chains further form an electrostatic interaction (Fig. 6A).
The MastR-T74D structure, in contrast, contains no Clion in the retinal-binding pocket. The carboxylate side chain of Asp-74 fills the space where the Clion is located in MastR and takes over the function of the Clion in the extended hydrogen-bonding network. The electrostatic interaction between Asp-200 and Arg-71, however, is weakened, because the Arg-71 guanidinium group flipped away from Asp-200 (Fig. 6C). This is in line with the observed positions for the homologous Arg side chains in other proton (Fig. 6D) and chloride pumps (Fig. 6B), in which it points toward the extracellular surface for the former and toward retinal for the latter. The different electrostatic environments of the PRSB in WT and mutant result in different absorption spectra with maxima of 522 nm for MastR-T74D and 540 nm for MastR, respectively (Fig. S4C) (20).

Functional conversion of chloride pumps
A distinctive feature of MastR is that a change of the TSD motif to DSD (T74D mutation) converts the chloride pump into proton pump (20). However, functional conversion of archaeal HsHR and NpHR chloride pumps into proton pumps by altering their TSA motif failed. The introduction of single or multiple mutations to generate various putative proton pumps with an aspartate as proton acceptor yielding DSA, DSD, or DTD motifs were unsuccessful (10,45,46), although the DTD motif is characteristic for BR proton pump. On the other hand, conversion of BR in the opposite direction, i.e. H 1 to Clion pump, by changing the motif from DTD to TSD (D85T mutant) was successful, even though the obtained chloride pump was fairly inefficient (24,47). Similarly, for eubacterial pumps, conversion was only possible in one direction. For FR, an NTQmotif chloride pump, mutation of NTQ to DTE converted FR to a proton pump, but opposite conversion of GR, a DTE-motif proton pump, into a chloride pump by mutation of DTE to NTQ failed (11). The possibility of functional conversion implied that the pumps have a common fundamental transport mechanism and shed light on the evolutionary conserved residues of proton and chloride pumps across all domains of life (9,48,49).
A structural comparison of proton and chloride pumps can provide insights into why some conversions are possible but others are not (Fig. 6). The extended hydrogen-bonding network linked to the PRSB is at the heart of the pump, because upon retinal isomerization, the N-H dipole of the PRSB will flip in chloride pumps, and for proton-pumping rhodopsins the PRSB proton will even dissociate for proton transfer to the proton acceptor. The geometry of the extended hydrogen-bonding network defines the starting point in the pumping cycle and is thus a determinant of pump function.
When we compare the hydrogen-bonding networks of proton pumps (Fig. 6, C and D), we find that in all cases, a central water molecule connects the PRSB and the carboxyl side chains of the proton acceptor (Asp-85 in BR) and a conserved aspartate (Asp-212 in BR (7)). Characteristic is that these hydrogen bonds are strong, having distances between the heavy atoms shorter than or equal to 2.9 Å. In some proton pumps the hydrogen-bonding network is extended by additional water molecule(s) and polar side chains including the second amino acid of the function-determining motif. For the chloride pumps the extended hydrogen-bonding network is similar but in addition contains a Clion between the first two amino acids of the function-determining motif, which is bridged via a water molecule to the PRSB (Fig. 6, A and B). It is striking that the hydrogen bonds are weaker, especially the bond between the central water and PRSB, which is ;3.5 Å for archaeal HRs and eubacterial NmClR. For MastR the hydrogen-bonding network is tighter, more similar to proton pumps, which might (together with the overall structural similarity between MastR and the proton pump BR) be an explanation for why the Clto H 1 ion pump conversion was successful. Strong hydrogen bonding of the PRSB water was concluded from FTIR studies to be a prerequisite for proton pumping and is consistent with the structural data on the PRSB hydrogen-bonding networks in Fig. 6  (10).
The position of the PRSB water in the hydrogen-bonding network seems to be crucial for a proton pump, because it constrains proper geometry between the aspartate proton acceptor and the PRSB, which affects the pK a of the acceptor group. A proper pK a is important for a proton transfer event in later stages of the photocycle. By comparing the structures, we find that Ala-53 in BR is an important determinant to achieve the correct geometry of the PRSB hydrogen-bonding network. This residue is a serine in the non-outward proton transporting families including HRs, Xenorhodopsins (such as ASR) (50), and Schizorhodopsins (51) (Fig. S3). 25 years ago, it was proposed  6xl3). B, chloride pumps HsHR (PDB code 5ahy), NpHR (PDB code 3ak7), and NmClR (PDB code 5g28). C, proton pump MastR-T74D mutant (PDB code 6wp8). D, proton pumps BR (PDB code 1c3w), Exiguobacterium sibiricum rhodopsin (PDB code 4hyj), and GR (PDB code 6nwd). Retinal and residues of the hydrogen-bonded PRSB network are shown as a stick model. Water is shown as red spheres, and chloride is shown as green spheres. Hydrogen bonding distances are indicated in angstrom (Å), with distances 3 Å shown in blue and those .3 Å shown in pink.
EDITORS' PICK: Structural insight into functional conversion of a Clpump that the short side chain of Ala-53 is crucial to allow changes of the PRSB-Counterion complex geometry during the photocycle to enable proton transfer to the acceptor (52). Indeed, decreasing the alignment of PRSB and Asp-85 in BR by the A53V mutation almost abolished formation of the M state and resulted in disappearance of the hydroxyl stretch band of the strongly bound water in the L state. The A53S mutation completely abolished the M state. 7 The available structures of various proton and chloride pumps and our newly added MastR structures confirm the importance of the geometry in the PRSB hydrogen-bonding network. It appears that the serine side chain in HRs may displace the Schiff base water away from the Schiff base, preventing their successful conversion to proton pumps.
It is important to note that there must be additional determinants for the successful conversion of pump function. In an attempt to convert NpHR to a proton pump, Muroda et al. (10) generated two mutants with six or ten replacements to yield sequence conservation patterns including BR's DTD motif and the Ala-53 equivalent. A functional proton pump, however, could not be generated, pointing at the additional fine-tuning required to achieve proper conformational changes and proton affinities in the ion-transporting photocycle.
In an extensive pump interconversion study comprising proton, chloride, and sodium pumps (11), the authors describe several examples of asymmetric functional conversion in which conversion was only possible when mutation reversed the evolutionary amino acid changes in the function-determining motif. In case of converting the FR chloride pump into a proton pump and increasing the proton pump characteristic accumulation of the M state, the S255F mutation outside the PRSB hydrogen-bonding network was required in addition. It has been proposed that MastR evolved from a proton pump (20,23). The high structural similarity of MastR to BR and the conservation of key functional residues (Fig. S3) explains why additional mutations were not necessary for the generation of an effective proton pump.

Conclusion
The crystal structures of MastR and MastR-T74D gave insight into the ion transport pathway and the mechanism of this new group of chloride ion pumps with TSD motif. MastR resembles and most likely evolved from an archaeal proton pump, explaining why the T74D mutation enabled functional conversion from Clto H 1 pump. The distinct hydrogenbonding network including a strongly bound water molecule between the PRSB and Asp-74 is a prerequisite for the successful conversion. Proton transfer requires the correct pK a of proton acceptor Asp-74, which is achieved by a suitable geometry of the hydrogen-bonding network. In the future, it will be interesting to study MastR T74D photocycle intermediates to determine the proton acceptor pK a and understand details of the ion transport steps. In addition, because of MastR's slow photocycle (23) (Fig. S11) and crystallization in bicelles, this chloride pump should be suitable for time-resolved serial synchrotron X-ray crystallography using fixed-target crystal delivery systems (53,54).

Experimental procedures
Gene construction, protein expression, and purification The plasmid for expression of MastR was described previously (23). The gene encoding MastR was cloned into pET21a (1) vector via NdeI-XhoI restriction sites, which added a Cterminal hexahistidine tag. The MastR-T74D plasmid was prepared from the MastR plasmid using a QuikChange Lightning site-directed mutagenesis kit (Agilent). The mutation was confirmed by DNA sequencing (ACGT Corp, Toronto, Canada).
For purification, the bright red MastR-containing solution was filtered (0.22 mM, Millipore), equilibrated with 20 mM of imidazole, and loaded onto a HisTrap TM 1-ml nickel-nitrilotriacetic acid column (GE Healthcare). The column was washed with 10 column volumes (CV) of equilibration buffer (0.1% (w/v) DDM, 20 mM imidazole, buffer A). The DDM detergent was reduced to 0.05% (10 CV of 0.05% DDM, 20 mM imidazole, buffer A), and then exchanged to n-octyl b-D-glucopyranoside (OG) (20 CV of 2% OG, 20 mM imidazole, buffer A). The protein was eluted with 30 CV of a linear imidazole gradient (20-750 mM imidazole, buffer A, 1% OG). Elution was monitored at 280 and 537 nm (or 522 nm for MastR-T74D). Consecutive fractions of MastR in OG with sufficient purity (A 537nm /A 280nm . 0.6, eluted at ;400-450 mM imidazole) were combined and concentrated to 10 mg/ml using a 30-kDa concentrator (Amicon). Gel filtration was performed on 1 ml of concentrated protein loaded on a Superdex TM 200 10/300 GL column (GE Healthcare) using buffer B (10 mM MES, pH 6.5, 300 mM NaCl, 0.8% OG) at a rate of 0.3 ml min 21 with 0.4-ml fractions collected. Consecutive fractions with A 537nm /A 280nm . 0.6 were combined and prepared for crystallization. The obtained protein had a purity of .95% as analyzed by Coomassie-stained SDS-PAGE (Fig. S4, A and B).

Crystallization and harvesting
24% (w/v) bicelles composed of 2.8:1 1,2-dimyristoyl-sn-glycero-3-phosphocholine:CHAPSO were prepared in advance as previously described (55) and were stored at -20°C. Prior to crystallization, the bicelles were thawed at room temperature and then kept on ice. To crystallize MastR, the protein was first concentrated to ;15 mg/ml assuming e 537 = 40,000 M 21 cm 21 (20). The protein and bicelles were combined in a 2:1 ratio (resulting in 10 mg/ml protein and 8% (w/v) bicelle) and incubated on ice for 2 h. The precipitant solution was prepared (pH 4.6, 3.6 M sodium phosphate monobasic monohydrate (Hampton Research), 180 mM 1,6-hexanediol, 3.5% triethylene glycol). Hanging-drop vapor-diffusion crystallization experiments were set up on standard pregreased 24-well crystallization trays. The 6 ml of hanging drop (4 ml of MastR-bicelle mixture, 1.5 ml of precipitant solution, 0.5 ml of 1% OG) was mixed 10 times on a thick siliconized cover slide and held over 0.5 ml of precipitant solution in the reservoir. Crystal trays were stored at 34°C and left undisturbed for 3 days. The crystals appeared after 2-3 days and reached a maximum size of 50-150 mM in 4-6 days. The crystallization of MastR-T74D is nearly identical, with the only difference being the pH of the crystallization buffer (pH 4.0).
To prevent dehydration while harvesting crystals, 6 ml of the precipitant solution was immediately pipetted onto the sample, followed by 6 ml of precipitant solution containing 12.5% ethylene glycol as cryoprotectant. The glass slide containing the hanging drop was placed on an ice pack to make the drop more fluid. After 90 s of incubation, single crystals were harvested using MicroLoop LD TM 50-200 mM (MiTeGen) and then flash frozen in liquid nitrogen.

Bromide soaking experiment
After MastR crystals had grown, the glass cover was removed, and 6 ml of 0.1 M sodium bromide in precipitant solution was added to the hanging drop. The cover was quickly resealed and then stored (10 min, 20°C). The crystals were then soaked in cryoprotectant (5 min, 6 ml of precipitant solution with 12.5% ethylene glycol) and then flash frozen with liquid nitrogen.

Data collection and analysis
X-ray diffraction experiments were carried out on Beamline 23-ID-B of the Advanced Photon Source (APS) at Argonne National Laboratory (Lemont, IL, USA). The data were collected at 100 K using a 1.0332 Å, 20 3 20-mm X-ray beam that was attenuated by a factor of 5. Reflections were collected every 0.2°(0.4 s) for a total of 60°using a Dectris Eiger X 16M detector operated in continuous, shutterless data collection mode at a distance of 250 mm. For MastR-T74D, data from a single crystal was employed. For MastR, a single, large, hexagonal crystal was measured in three separate spots, and the resulting data were merged in data processing. Diffraction data were processed using XDS (56) and scaled using AIMLESS (57) from the CCP4i program package (58). The MastR structure was solved by molecular replacement with Balbes (59) using archaerhodopsin-1 (60) as a search model (PDB entry 1UAZ). The MastR-T74D structure was solved by molecular replacement using MastR as the starting model. The model was manually built in Coot (61), and iterative refinement was done with the phenix.refine routine of Phenix (62). Refinement statistics are summarized in Table S1. Figures with three-dimensional structures were prepared using PyMOL (The Pymol Molecular Graphics System, Version 1.8.0.5, Schrödinger, LLC).

CD spectroscopy
CD spectra were recorded as described previously (63). In brief, the spectra were collected at 20°C with a 10-mm-pathlength quartz microcell on a model no. J-810 spectropolarimeter (JASCO, Easton, MD, USA). A spectral range of 700-400 nm with a scanning speed of 50 nm/min were used. Four independent accumulations were averaged.

Flash photolysis
Flash-photolysis spectroscopy was performed as described previously, using a custom-built single-wavelength spectrometer (64). Briefly, 7-ns pulses of the second harmonic of an Nd-YAG laser at 532 nm (Continuum Minilite II) initiated the photocycle. Absorption changes of the monochromatic light (Oriel QTH source and two monochromators) were recorded with an Oriel photomultiplier, an amplifier with a 350 MHz bandwidth, and a Gage AD converter (CompuScope 12100-64M).

Root-mean square deviation calculation
The root-mean-square deviation (RMSD) was calculated using PyMOL. A sequence alignment was performed, and the C a of aligned residues were superimposed. Five rounds of refinement were implemented, and structural outliers were discarded.

Data availability
The refined models have been deposited in the Protein Data Bank under codes 6xl3 (MastR) and 6wp8 (MastR-T74D). All other data that support the findings of this study are available from the corresponding author upon reasonable request.