Mechanism of Ion Transport across Membranes

The high resolution structures now available for mitochondrial ATPase (1) and mammalian and bacterial cytochrome c oxidases (2, 3) have raised hopes that the long sought description of the proton translocation mechanism in transmembrane pumps might be near. According to the simplest version of the alternating access hypothesis (reviewed in Refs. 4–6), such transport is based on the energy-dependent cycling through two protein conformations in which the access of a single ion binding site changes from one membrane side to the other. It has been difficult to put this idea to a test because most proton pumps are large or multisubunit proteins in which the site of ion translocation is at some distance from the chemical reactions that drive it, and the chemical reactions are themselves complex. If there is a general mechanism in proton pumps, clues to it are more likely to come from a simpler system. Recent progress with bacteriorhodopsin, a small retinal protein in which the thermal reisomerization of photoisomerized retinal drives the proton transport, has yielded a step-by-step mechanism for the translocation cycle. The principles and perhaps some of the details in this mechanism, described briefly below, may prove to apply to other ion pumps. They seem relevant also to signal receptors (7).

The high resolution structures now available for mitochondrial ATPase (1) and mammalian and bacterial cytochrome c oxidases (2,3) have raised hopes that the long sought description of the proton translocation mechanism in transmembrane pumps might be near. According to the simplest version of the alternating access hypothesis (reviewed in Refs. 4 -6), such transport is based on the energy-dependent cycling through two protein conformations in which the access of a single ion binding site changes from one membrane side to the other. It has been difficult to put this idea to a test because most proton pumps are large or multisubunit proteins in which the site of ion translocation is at some distance from the chemical reactions that drive it, and the chemical reactions are themselves complex. If there is a general mechanism in proton pumps, clues to it are more likely to come from a simpler system. Recent progress with bacteriorhodopsin, a small retinal protein in which the thermal reisomerization of photoisomerized retinal drives the proton transport, has yielded a step-by-step mechanism for the translocation cycle. The principles and perhaps some of the details in this mechanism, described briefly below, may prove to apply to other ion pumps. They seem relevant also to signal receptors (7).

Structure and Photochemical Reaction Cycle
Bacteriorhodopsin is a seven-helix transmembrane protein, with an all-trans-retinal lying at a small angle to the membrane surface and linked via a protonated Schiff base to Lys 216 near the middle of helix G (Fig. 1). Most of the structure has been known at 3.5-Å resolution from two-dimensional crystals (8), and more recently the entire structure was described at 3.0 Å (9). Higher resolution from three-dimensional crystals, at 2.5 Å is now available (10). The interhelical cavity is divided by the Schiff base into extracellular and cytoplasmic "half-channels" that together describe the trajectory of the transported proton. The extracellular half-channel contains numerous charged or hydrogen-bonding residues, whereas the cytoplasmic region is simpler and mostly hydrophobic. In the first transport event after absorption of a photon, the Schiff base proton is mobilized by photoisomerization of the retinal to 13-cis,15-anti, and transferred to Asp 85 in the extracellular region, causing the release of a proton to the surface. The Schiff base is then reprotonated from Asp 96 from the cytoplasmic side. Asp 96 in turn is reprotonated from the surface. These proton transfers together add up to translocation across the membrane. They and other reactions of the retinal and the protein during the cycle have been measured by various spectroscopic methods and consist of the interconversions of the intermediate states designated as J, K, L, M, N, and O, and substates of several of these. Much effort has been expended to describe these reactions and the protein residues involved (4,(11)(12)(13)(14)(15). But to understand bacteriorhodopsin as a proton pump we must know also what determines the rates of the proton transfers and how the pK a s and the geometry of the donors and acceptors change so as to give them a cytoplasmic-to-extracellular direction.

Partial Reactions of the Photocycle Observed without Illumination
Although the proton transport is the consequence of a "photocycle," mechanistic clues have been gained recently from reactions of the unphotolyzed protein that correspond to various single photocycle steps. First, the observation of a biphasic titration curve for Asp 85 suggested that its pK a is linked to the protonation of another residue (16,17) that turned out to be Glu 204 (18 -20) perhaps together with liganded water but certainly other residues near the extracellular surface, such as Glu 194 . The nature of the linkage is that protonation of Asp 85 will cause deprotonation of the Glu 204 site and vice versa. This is the kind of coupling that would cause proton release at the extracellular surface after proton transfer from the Schiff base * * This minireview will be reprinted in the 1997 Minireview Compendium, which will be available in December, 1997. shown with all-trans-retinal at 3-Å resolution (9). The proton moves from the cytoplasmic membrane surface to the extracellular in several segments (proton transfers are shown with curved green arrows). A detailed description of these events, some supported by more evidence than others, is given in the text. Briefly, in the first phase (left panel) the protonated Schiff base of the photoisomerized 13-cis,15-anti-retinal protonates the anionic Asp 85 . By electrostatic interaction (blue arrow), this causes proton transfer from Glu 204 , or water liganded to it, to Glu 194 and from there proton release to the surface. In the second phase a large scale conformation change (shown in the right panel as tilt of helix F) occurs and causes increased hydration of the cytoplasmic region and thus a decrease of the pK a of Asp 96 . This causes reprotonation of the Schiff base. Recovery of the initial protein conformation results in reprotonation of Asp 96 from the bulk and reisomerization of the retinal to all-trans. The final proton transfer from Asp 85 to Glu 204 completes the pathway of the proton across the protein. The access of the retinal Schiff base to its proton acceptor and donor is therefore determined by directed shifts of pK a s and probably also by a geometrical reorientation that connects it with first the acceptor and then the donor. (Coordinates were generously provided by Y. Kimura.) to Asp 85 (21-25) even though Asp 85 itself remains protonated until the end of the photocycle (26 -28). Conversely, once the proton is released at a pH higher than the pK a of the Glu 204 site, the pK a for Asp 85 will rise. Under physiological conditions, where the difference between the pH and the pK a for the release is large, the proton release will shift the protonation equilibrium between Asp 85 and the Schiff base toward virtually complete and unidirectional proton transfer.
The second observation, made from x-ray diffraction, was that deprotonation of the Schiff base of the D85N mutant by raising the pH in the dark caused the protein to assume an equilibrium mixture of conformations that exhibit structural changes seen otherwise only in the M photointermediate (29). In the D85N/ D96N double mutant the equilibrium contained a large amount of the M-like conformation even with the Schiff base protonated. The pK a s for the protonated Schiff base and the changes in crystallographic parameters for D85N were the same, and the isomeric composition of the retinal was indifferent to the shift of protein structure (30). As expected from this, the pK a of the Schiff base in D85N was nearly unchanged when the retinal was replaced with an analogue locked in the all-trans configuration. These observations provided a hint to the cause of the proton transfer switch in the photocycle that allows reprotonation of the Schiff base from the cytoplasmic side; if it is a result of the shift of the global protein conformation, it depends on deprotonation of the Schiff base, i.e. loss of interaction between the protonated Schiff base and its complex counterion rather than directly on the isomeric state of the retinal.
In the third study (31), mutations in the extracellular proton channel were shown to cause parallel decreases in the rate of the final photocycle step and the deprotonation of Asp 85 in pH jump experiments in the dark. The correlation of the rates over 3 orders of magnitude suggested that the loss of proton from Asp 85 to the Glu 204 site or directly to the medium depending on the conditions is the rate-limiting step in the recovery of the initial state. Other mutations, nearer the retinal, also caused the slowing of the photocycle, but for these there was no correlation with the deprotonation of Asp 85 . In these cases a perturbed interaction between the 13-methyl or 9-methyl groups of the retinal and the protein interfered with reisomerization to all-trans, as suggested earlier by other kinds of evidence (32)(33)(34).

Mechanism of Transport
These recent results contributed some missing pieces in the puzzle and helped to refine the proton transport mechanism that follows. The active site consists of the protonated Schiff base and the anionic Asp 85 and bound water (35-39) that stabilizes the buried charges. Photoisomerization of the retinal from all-trans to 13-cis,15-anti changes this geometry (40,41) and redistributes the -electron system along the retinal chain (42,43), causing the pK a difference between the Schiff base and Asp 85 to narrow from about 5 to Ͻ1 (44). Fig. 1 shows schematically the proton transfer steps in the early and late phases of the transport cycle that ensues. As Asp 85 becomes protonated, linkage of the pK a s of Asp 85 and the Glu 204 site (Fig. 1, blue  arrow in left panel) causes the pK a of the Glu 204 site to drop from 9 to about 5. As a result, Glu 204 protonates Glu 194 , and the latter releases the proton to the surface. 1 At neutral (physiological) pH release of the proton to the extracellular surface is strongly favored. This in turn causes the pK a of Asp 85 to further rise (17,19,46), and its protonation equilibrium with the Schiff base shifts toward nearly full proton transfer (24). This tran-sition, identified as the M 1 3 M 2 reaction that was deduced from the kinetics of absorbance changes (47), completes the first phase of the transport cycle (Fig. 1, left panel). Reprotonation of the Schiff base from Asp 85 is now no longer possible.
When the proton transfer converts the charged Schiff base-Asp 85 complex to a neutral pair, stabilization of the initial structure of the protein is removed, and an alternative conformation is assumed (30). Difference maps from neutron, x-ray, and electron diffraction of the M state show density increases at the cytoplasmic segments of helices B and G (48 -54). Blue shift of the absorption maximum of the chromophore with unprotonated Schiff base at this time in the D96N mutant but not in the wild type (55) suggested that the Schiff base enters into hydrogen bonding, most likely through bridging water molecules (8,56) with Asp 96 . Thus, after proton transfer back from Asp 85 is made energetically unfavorable by the elevated pK a of this residue, the connection of Schiff base is switched to Asp 96 . The existence of such an accessibility change, based on geometrical rearrangement at the Schiff base, is suggested also by the observation that under some conditions transport can be demonstrated even when both Asp 85 and Asp 96 are replaced with neutral residues (57,58). The change in local geometry initiates the second phase of proton translocation (Fig. 1,  right panel).
At this time in the photocycle the pK a of the Schiff base is about 8 (59), i.e. much lower than its initial value of Ͼ13 (60), but the pK a of Asp 96 is presumed to be still at least as high as 11, as in the unphotolyzed protein (61). The latter pK a will decrease to about 7 so as to make Asp 96 a proton donor (62). The reason appears to be a further protein conformation change in the M intermediate (observed when this state is stabilized in the D96N mutant and termed M N ) that consists of the outward tilt of the cytoplasmic end of helix F (48 -54). The effects of dehydration by osmotic agents (63) and hydrostatic pressure (64) suggested that it is the increased hydration of the cytoplasmic region that lowers the pK a of Asp 96 and makes this residue a proton donor to the Schiff base. The observed consequences of chemical modifications of engineered cysteines and cross-linking at the cytoplasmic end of helix F, and not elsewhere, attributed the increased hydration to the cleft that is formed upon tilt of this helix (65). Reversal of the tilt of helix F restores the initial high pK a of Asp 96 (66), and this residue is then reprotonated from the cytoplasmic surface. The increase of the lateral cross-section of the protein in the bilayer during this process (50) may be the cause of the cooperativity in the two-dimensional lattice of the purple membrane upon photoexcitation (67).
Reisomerization of the retinal to a twisted all-trans configuration (not shown in Fig. 1) is followed by proton transfer from Asp 85 to the Glu 204 site and relaxation of the twisted retinal chain (31), completing the photocycle. Thus, the recovery of the initial proton affinities of Asp 85 (pK a about 2.5) and the Glu 204 site (pK a about 9) at the end of the photocycle is the reason for the irreversibility of this reaction. It is because the pK a of Asp 85 returns at this time to its initial very low value that proton transport can proceed against a high transmembrane proton potential, and sufficient protonmotive force is created to drive the light-driven synthesis of ATP (68,69).

Role for Bound Water in Proton Transfers
Much indirect evidence indicates that there is specifically bound water near the Schiff base and Asp 85 , and its hydrogen bonding undergoes changes throughout the photocycle. Water near the Schiff base and Asp 85 was suggested to mediate the initial proton transfer (41,70). Hydrogen-bonded water molecules that connect Asp 85 to Glu 204 in structures calculated from molecular dynamics (71, 72) may play the major role (19) in the linked pK a s of Asp 85 and the Glu 204 site. The changed access of the Schiff base to Asp 96 has been suggested to be through reorganized water molecules at the active site (73). Water molecules between the Schiff base and Asp 96 were postulated to mediate proton transfer across the 12-Å distance in the interior of the protein (8,9). The effects of osmotic agents on this reaction (63) suggested that first an increased, then a decreased, amount of bound water near Asp 96 associated with the formation and the closing of a cleft at the cytoplasmic surface is what regulates the proton affinity of this residue as a proton donor to the Schiff base and then an acceptor from the cytoplasmic surface. Finally, in the last step of the sequence proton is lost from Asp 85 to Glu 204 most likely through the mediation of the hydrogen-bonded water chain that connects them. Indeed, the large D 2 O isotope effect (5-10-fold) associated with both proton release and the deprotonation of Asp 85 that takes place later (31) strongly suggest that bound water plays an important role in these processes.

Nature of the Active Site
The Schiff base and its acceptor appear to constitute the single active site in this proton pump. The function of the rest of the protein may be simply to alternately deliver to and remove a proton from this site. If this is true, we should distinguish between an active site specific for the initial and critical ion transfer and a nonspecific matrix that responds to it passively by redirecting its access to either membrane surface.
There is support for this view from an observed change in the ion specificity of bacteriorhodopsin. Upon replacing Asp 85 with a threonine, the equivalent residue in the light-driven chloride pump halorhodopsin (74), bacteriorhodopsin acquired the spectroscopic properties and the chloride transport activity (that proceeds in the opposite direction from the proton transport) of halorhodopsin (75). Although the chloride transport mechanism in this recombinant bacteriorhodopsin and in halorhodopsin (76 -79) is not yet well understood, it must involve binding of chloride at the active site and its translocation to the cytoplasmic domain. Evidently, the alternating pathways of access in this protein allow the movement of not only protons but also chloride.
Halorhodopsin, in turn, can be made to transport protons. This will occur in a two-photon reaction in which the deprotonated 13-cis and protonated all-trans-retinal Schiff bases are forcibly interconverted, with an accompanying protonation and reprotonation that results in net translocation of charge (80). Analogously to the mutational change of specificity of bacteriorhodopsin, halorhodopsin also becomes a proton pump when an artificial proton acceptor, the azide anion, occupies a site near the Schiff base where otherwise the transported chloride is bound (81). Thus, when provisions are made for a proton acceptor to the Schiff base, the access change and the conduction pathways for chloride will also accommodate protons.
The hypothesis that the functioning of bacteriorhodopsin depends less on proton conduction pathways to and from the surface than on the initial proton transfer at the active site is supported by the remarkable fact that of the many hundreds of site-specific mutations studied (12,14,15) none inactivate transport entirely. Those that have the greatest effect prevent proton transfer from the Schiff base to Asp 85 , e.g. by replacing the aspartate with a neutral residue or by keeping it permanently protonated. However, even in the former case where transport of the kind in the wild type is inactivated, proton transport is observed under some conditions (57,58). Two other mutations that are less disruptive are also worthy of note. In the first, E204Q, proton release to the extracellular surface upon protonation of Asp 85 is abolished. Although proton release is delayed until the end of the photocycle when the low pK a of Asp 85 is reestablished, the direction of the proton transfers is preserved (18,82). In the second, D96N, reprotonation of the Schiff base and the subsequent proton uptake from the cytoplasmic surface are hindered. Protonation of the Schiff base is directly from the surface but again with the correct directionality (62,83,84). Thus, the proton conduction pathways to both membrane surfaces can be seriously perturbed without loss of transport activity. On the other hand, interference with these pathways does affect the turnover rate of the pump, and the rate of proton transport at high light intensities will be decreased. At physiological pH the slowing of the photocycle in D96N amounts to 2-3 orders of magnitude, as the rate-limiting step becomes the inefficient capture of a proton at the cytoplasmic surface (63). Replacement of residues at the cytoplasmic surface, such as Asp 38 , that might also have roles in proton conduction causes a severalfold slowing of the photocycle (45). Similarly, at the extracellular surface, the E204Q mutation causes decrease of the turnover by 1-2 orders of magnitude, as the loss of the proton of Asp 85 becomes rate-limiting (18,31). Avoiding such problems must be the rationale for the evolution of the proton transfer pathways described above. Indeed, the arrangement of acidic residues in the structure suggests that multiple pathways may exist for both release and uptake of protons at the two surfaces (9).

Conclusions
Thus, the transport in bacteriorhodopsin is explained by an alternating access mechanism. The change of the access of the occluded proton binding site occurs in sequential steps. In the first and last steps the direction of the transfer of proton between the Schiff base and the two membrane surfaces is determined by the magnitude and sign of a changing ⌬pK a relative to proton acceptor and donor groups. Although this would ensure, by itself, the direction of the transport, there may be in addition a change of the local geometry to switch the connectivity of the Schiff base. The lesson for other pumps may be that the local ion transfer reactions, which deliver the transported ion to the active site and remove it, are more crucial and more stringently regulated than the conduction pathways between the binding site and the two membrane surfaces.