I. INTRODUCTION

Top Previous Next References

Voltage-gated proton channels are unique ion channels in several respects. They are called proton channels because they behave like ion channels and are highly selective for protons. Although protons exist in solution almost entirely in the form of hydronium ions, H₃O⁺, all proton-selective channels conduct protons as H⁺, rather than H₃O⁺. This is true even for water-filled pores like gramicidin. It remains a matter of some contention whether proton channels should be considered to be ion channels at all, although this designation seems more appropriate than any alternative and is becoming accepted (444). Proton channels differ from carriers and unequivocally are not pumps. Protons are unique ions with respect to their behavior in bulk solutions, their interactions with proteins, and the mechanism by which they traverse ion channels and other molecules. The unique chemical properties of protons explain why proton channels hold the records for both the largest and smallest single-channel currents. Thus there is an introductory discussion of selected aspects of proton chemistry. For a detailed discussion of the methods of pH measurement, the reader is referred to the superb review by Roos and Boron (850).

This review includes what I as a student of voltage-gated proton channels consider to be useful and relevant. Although the main focus is voltage-gated proton channels, there is substantial coverage of salient properties of a number of other proton-conducting molecules, for several reasons. First, the structure and even the molecular identity of voltage-gated proton channels is essentially unknown, whereas the structures of a number of other proton-conducting molecules are known to within a few Angstroms. Second, certain features that differentiate proton channels from other ion channels may be shared among molecules whose function involves proton translocation. Once nature discovers a solution to a design problem, this solution tends to recur (245). Proton conduction through the prototypical ion channel, gramicidin, provides a frame of reference with respect to which we interpret many results (deuterium and temperature effects, pH dependence, unitary conductance, etc.). It is possible to distinguish two broad classes of proton-permeable molecules. Some molecules couple the flux of protons to a bioenergetic or enzymatic goal, such as photosynthesis or CO₂ hydrolysis. Other molecules are simple proton channels that apparently exist for the sole purpose of mediating proton flux across membranes. In both cases, proton flux is tightly regulated, either by coupling to events central to the function of the molecule or by a gating mechanism that turns proton flux on and off at appropriate times. A premise of this review is that the molecular details of proton movement through all types of proton-conducting molecules are likely to display similarities with general applicability.

The properties common to all voltage-gated proton channels are described in detail. Then the properties and proposed functions of H⁺ channels in specific cells are discussed. There is a strong emphasis on proton channel function in phagocytes, because much more is known about function in these cells than in any other. Evidence for and against the proposal that part of the phagocyte NADPH oxidase complex functions as a proton channel (427) is summarized.

I do not expect more than a handful of people to read the entire review. For those who study any of the numerous molecules with proton pathways, I hope to present a synopsis of their molecule from the vantage point of an electrophysiologist interested in proton conduction. I feel that it is useful to have information specifically regarding proton conduction in various channels/molecules assembled in one place. Those who study "normal" ion channels and are curious about H⁺ channels will want to know their biophysical properties, which will appear esoteric and tedious to others. Phagocyte biologists will be interested mainly in the section on H⁺ channels in phagocytes. For everyone else, the review should be a resource enabling a particular bit of information to be located in the table of contents.

	II. CHEMISTRY OF PROTONS

Top Previous Next References

Protons in aqueous solution almost always exist in hydrated form as hydronium ions, H₃O⁺ (or H₃O⁺ · nH₂O, including waters of hydration), also called oxonium (605) or hydroxonium ions (1070). Protons exist as H⁺ <1% of the time during transfer from one water to another (184). The three protons in H₃O⁺ are equivalent, and each is equally likely to jump to a neighboring water molecule (84). The proton is unique among cations in being interchangeable with the protons that form water molecules. This capability is significant in light of the tiny concentration of "free protons" (H₃O⁺) in physiological solutions, ~40 nM, and the enormous total concentration of H in water, 110 M. Only one proton in a billion is part of H₃O⁺ at any moment. The average lifetime of the H₃O⁺ ion is ~1 ps in liquid water at room temperature: estimates in chronological order include 0.65 ps (84), 0.24 ps (184), 3.0 ps (287), 1.7 ps (636), 1.1 ps (11), 1.3 ps (1095), 0.95 ps (1050), and 0.5-0.79 ps (890). The proton is also unique as a monovalent cation in having no electrons, giving it a radius 10⁵ smaller than other ions, which greatly facilitates proton transfer reactions (80) and electrostatic interactions with nearby molecules (696).

The quintessential feature of water and other proton conduction pathways is the hydrogen bond (80, 84, 287, 361, 380, 469, 470, 592, 605, 799, 800, 967, 1101). Huggins appears to have originated the concept of the hydrogen bond while in the laboratory of Latimer and Rodebush. Huggins conceived the idea of a hydrogen "kernel" held between two atoms in organic compounds, which he did not publish until 1922 (468); several earlier investigators discussed interactions that in retrospect could be considered examples of hydrogen bonds (490). In 1920, Latimer and Rodebush (592) adopted this idea and applied it to water, foreseeing the existence of networks of water molecules, and used hydrogen bonding to explain the high mobility of protons in water as "a sort of Grotthuss chain effect, rather than ... a rapid motion of any one H₃O⁺ ion."

"Water ... shows tendencies both to add and give up hydrogen, which are nearly balanced. Then, in terms of the Lewis theory, a free pair of electrons on one water molecule might be able to exert sufficient force on a hydrogen held by a pair of electrons on another water molecule to bind the two molecules together. Structurally this may be represented as

View larger version (7K): [in this window] [in a new window]

Such combination need not be limited to the formation of double or triple molecules. Indeed, the liquid may be made up of large aggregates of molecules, continually breaking and reforming under the influence of thermal agitation. Such an explanation amounts to saying that the hydrogen nucleus held between 2 octets constitutes a weak `bond' ¹ " (592).

Linus Pauling coined the term hydrogen bond in a general paper on chemical bonds (798) and developed and popularized the idea in a chapter of his book, The Nature of the Chemical Bond (800).

Water molecules tend to form tetrahedral hydrogen bonded structures, at least ideally (84). In ice the tetrahedral structure exists (799) and is evidently so rigid at very low temperature (i.e., the dielectric constant drops drastically) that proton conduction is limited (188, 261, 313). In liquid water, however, the tetrahedral ideal is not achieved, and the actual coordination number decreases with increasing temperature (300, 366), which likely accounts for the greater decrease in activation energy at higher temperatures for proton transport than for other ions (319, 605, 784, 786). Water can be considered a "broken down ice structure" with continual formation and breaking of hydrogen bonds (707). Although protons in water are formally considered to exist as H₃O⁺ molecules, it has long been recognized that larger molecular groupings exist and are central to the understanding of proton conduction. As early as 1936, Huggins (470) explicitly postulated the existence of H₅O, showed that proton conduction can occur by shifts in the identities of the water molecules that comprise the cation, and suggested that the rapidity of such shifts accounts for the high mobility of protons in water. The two main larger species are the so-called "Zundel cation," two waters sharing an excess proton as H₅O (470, 1102, 1103), and the "Eigen cation," four waters sharing an excess proton as H₉O (80, 287, 1070), although a transitional H₁₃O structure has also been proposed (1049). These quasi-molecules are in a sense fictitious, in that they are idealizations that exist only transiently along with many undefined intermediate or alternative states (664, 890). Quantum molecular dynamics simulations show that a proton in water sometimes shuttles back and forth between two neighboring water molecules many times per picosecond, behavior that defines a Zundel (or Huggins) cation, but also spends time associated with a single water (which is hydrogen bonded to three first shell waters) as an Eigen cation (890, 1050). Eigen thought that the proton in H₉O was essentially delocalized (288) and shared among three of the waters surrounding the H₃O⁺ molecule; the fourth water is oriented incorrectly for rapid proton transfer (605). Ab initio molecular dynamics calculations indicate that a proton in water is affiliated with one oxygen atom as H₃O⁺ (H₉O, including the primary hydration shell) 60% of the time, and 40% of the time it is intermediate between two oxygens as H₅O (1025). Although the proton spends blocks of time as H₉O (i.e., associated with a single oxygen), these events occur within bursts of oscillations between the same pair of oxygens as though the proton remembers its former partner (1050), and hence, appearances to the contrary, was never truly delocalized.

That there is a fundamental difference between protons and other cations is clear from the fivefold higher conductivity of H⁺ in water than other cations like K⁺ (84, 217). In fact, considering its degree of hydration (based on solution density) H⁺ might be expected to have a low mobility like Li⁺ (84, 845) but has nine times higher mobility (845). It has long been appreciated that protons are conducted by a special mechanism in which they hop from one water molecule to the next, which is often called the Grotthuss mechanism, although de Grotthuss' proposal (254a) differs from current views. The Grotthuss mechanism is also called "prototropic" transfer (605), to distinguish it from ordinary "hydrodynamic" diffusion of H₃O⁺ as an intact cation. Danneel (217) suggested that a proton in an electric field might bind to one side of a water molecule and that another proton could leave the far side of the molecule, thus saving the time it would have taken to diffuse that distance. A key distinction from other ions is that during proton conduction the identity of the conducted proton changes (84). Except for Hückel's theory (467a), the equivalence of the three protons in H₃O⁺ is generally considered to be essential to the special prototropic conduction mechanism. Danneel further proposed in 1905 (217) that proton conduction by a Grotthuss mechanism requires two processes: proton hopping from one water molecule to the next, and also a reorientation of water molecules. Glasstone, Laidler, and Eyring (366) concluded that proton transfer was rate-limiting and that water rotation was rapid. Conway, Bockris, and Linton (184) concluded that the proton transfer step was rapid and proposed that the rate-determining step was the reorientation of the recipient water molecule in the electrical field of the donor H₃O⁺ (184, 448). More recent theories growing out of Eigen and co-workers' views agree that the proton transfer step is rapid, but ascribe the rate-limiting step to reorganization of the hydrogen-bonded network through which H⁺ conduction occurs (10a, 11, 221, 479, 664, 1024, 1025, 1050).

The special prototropic conduction mechanism appears to require a hydrogen-bonded structure (361, 469, 605). Water is an ideal medium for prototropic conduction because of its propensity to form hydrogen bonds; water has a higher viscosity compared with other solvents due to hydrogen bonding (300). Proton conduction occurs essentially by means of changes in the identity of the water molecules that participate in the hydrogen-bonded network that includes the excess proton. The mechanism of proton conduction in water has been described as "structural diffusion," which was felt to reflect the delocalized nature of the solvated proton within a hydrogen-bonded network (287, 319, 1070). The concept of structural diffusion of protons in water is supported by ab initio molecular dynamics simulation (1024). Proton conduction occurs as a result of isomerization between Zundel and Eigen cations (10a, 11, 1024). The rate-determining step appears to be the breaking of a second shell hydrogen bond, which allows the replacement of one of the waters by a different one (10a, 287, 288, 664). This process has been called the "Moses mechanism," with second shell hydrogen bonds breaking in the path of the proton and reforming behind (10a, 11a), just as the Red Sea parted to allow Moses and his companions to cross (Exodus 14:21-27). At this point the modern view (10a) diverges from most earlier models in which the water molecule immediately adjacent to H₃O⁺ is required to rotate into an appropriate configuration to accept the proton (80, 184, 448, 467a). The three first shell hydrogen bonds are too strong to be easily broken (10a), whereas the second shell hydrogen bonds are expected to be of normal strength, 2.6 kcal/mol, consistent with empirical measurements of proton mobility (636, 678). In Agmon's view, the widely used traditional method of estimating the "abnormal" component of H⁺ mobility by subtracting the mobility of a "normal" cation such as Na⁺ or K⁺ from the total H⁺ mobility (319, 361, 366, 467a, 605, 628, 636, 678, 845) is incorrect. Because the H₃O⁺ ion is tightly hydrogen bonded to its first shell neighbors, it is effectively immobilized. Consequently, essentially all of the mobility of protons in solution is of the abnormal (Grotthuss type) variety (11). Another difference is that in contrast to Eigen's delocalized proton that could move freely within the H₉O complex (287, 319, 1070), in the current view the proton is mainly associated with a single oxygen or vascillates rapidly between two oxygens, and eventually transfers successfully as a result of second shell hydrogen bond rearrangement (10a, 890, 1024, 1025, 1095).

Because waters inside proton channels may be bound or constrained in some way, proton movement through water-filled channels is often considered to be more analogous to proton transport in ice than in water (732, 733). Proton conduction in ice is fundamentally different from that in liquid water (288, 552, 771, 783). The extensive hydrogen bond rearrangement that characterizes proton transfer in water cannot occur in ice (552, 771). Liquid water is mainly three-coordinated, but the ice structure enforces four-coordination. Repulsion from the fourth water pushes the H₃O⁺ closer to its neighbors, decreasing the energy barrier for proton transfer (552, 771). Historically, the question of proton conduction in ice has proven to be difficult and controversial (42, 44, 96, 157, 188, 288, 294, 380, 500, 808). Eigen and colleagues reported that the mobility of H⁺ in ice was extremely high (289), 1-2 orders of magnitude higher than in water (288), and differing "from that of conduction band electrons in metals by only about 2 orders of magnitude" (287). Subsequently, the general consensus has been that these measurements were contaminated by conduction through melted water at the surface and that the true mobility is much lower, 3 × 10⁴ to 6.4 × 10³ cm² · V¹ · s¹, typically ~10³ cm² · V¹ · s¹ (142, 157, 294, 575, 734, 782, 808, 809). The mobility of H⁺ in water is 3.6 × 10³ cm² · V¹ · s¹ (845). It is a major problem to determine the number of defects (ionic or bonding) in ice, which must be known to calculate mobility. "Pure" ice almost invariably contains enough impurities to dominate attempts to measure the mobility of ionic defects, which are present at only ~1 per 10¹³ H₂O molecules at 20°C (808). This problem can be overcome by "doping" the ice with carriers so that their concentration is known and the signal is larger and thus more accurately measurable (380). In ice studies, it is important to distinguish events at the surface from events occurring within the bulk phase, although the former can be useful in dissecting elementary processes that contribute to proton mobility (260, 356, 357, 1028, 1083).

The only ions that carry current in ice are H⁺ and OH, and both move as a consequence of proton or proton defect movement (783). Both protons and Bjerrum defects (see sect. IIID) must move for sustained current (380); movement of L defects (or protons) alone simply produces (or eliminates) polarization (782). In pure ice at moderate temperatures, the dominant charge carrier is the Bjerrum L defect (the conduction of which occurs by rotation of water molecules), and thus for DC conduction the motion of the ionic defect (H₃O⁺) is rate determining (809). Protons tend to become shallowly "trapped" by the more abundant Bjerrum L defects, but above 110 K they escape at a significant rate and are mobile until they encounter the next trap (1083). Data on H₃O⁺ "soft-landed" onto the surface of ice were interpreted to mean that at temperatures below 190 K proton conductance in ice is essentially absent (188). One danger that must be considered in such studies is that protons can be trapped at the ice surface (1028), probably because the 4-coordinated state that is enforced inside ice is less favorable than the less stringent coordination at the surface (552). Earlier studies of isotope exchange in pure and in doped ice had indicated that Bjerrum defect and proton migration occurred to a similar extent in ice in the 135-150 K range, although OH lacked mobility (181). A recent study of isotope exchange in pure ice nanocrystals at 145 K revealed clear evidence of mobility of both Bjerrum L defects and protons, based on the distinctive infrared spectra of D₂O, coupled HDO molecules, and isolated HDO (1028). Most evidence indicates that protons are mobile in ice at least down to 110 K (1083), and possibly as low as 72 K (808), that proton mobility in ice is practically temperature independent (782, 808), and that the mobility of H₃O⁺ at ~100 K is within an order of magnitude of that in liquid water (808).

The hydroxide anion (OH) also has anomalously high conductivity compared with other anions, ~198 cm² · S/eq (218, 845, 943), although not quite so extreme as H⁺ at ~350 cm² · S/eq (786, 845, 921). In addition, the activation energy for OH conductivity is higher than for H⁺ (288, 623, 636). The high mobility is believed to reflect OH migration by a Grotthuss-like mechanism in which the OH moves from one water to the next by virtue of a proton hopping in the opposite direction (80, 84, 184, 217, 469, 605, 845). Protons move via prototropic transfers between H₃O⁺ and H₂O, whereas OH migrates by prototropic transfers between H₂O and OH. The rate-determining step in OH mobility may be the same as for H⁺ mobility, the breaking of a second shell hydrogen bond (11b), although a recent proposal invokes the crucial breaking of a first-shell hydrogen bond (1026). That OH mobility is less than H⁺ mobility in spite of the similarity of mechanism has been explained in several ways. Bernal and Fowler (84) proposed that the two protons in the donor H₂O are held more tightly than the three protons in the donor H₃O⁺ molecule, thus reducing the likelihood of the former proton transfer. Conway et al. (184) felt the critical difference was the electrostatic facilitation by the extra proton in H₃O⁺ of the prerequisite and rate-limiting water rotation that precedes proton transfer. Gierer and Wirtz (361) suggested a charge mechanism: for H⁺ transfer the proton hops between neutral H₂O molecules, whereas for OH the proton hops between two residual negative charges (288, 361). Agmon (11b) proposed that contraction of the O-O bond distance adds an extra 0.5 kcal/mol to OH transfer. Onsager proposed that H⁺ mobility is higher because the additional kinetic energy of the excess proton increases the energy of H₃O⁺ and favors subsequent proton transfer, whereas in OH conduction the energy of the proton is transferred from OH to H₂O and thus does not contribute to the next transfer (J. F. Nagle, personal communication).

Eigen (287) studied proton transfer reactions extensively and formulated general rules that govern such reactions. Proton transfer reactions tend to be very rapid and are described as "diffusion controlled" because the rate of the reaction is determined by the frequency of molecular encounters resulting from diffusion (287). The rate of proton transfer in normal proton transfer reactions depends on the pK_a difference between donor and acceptor, as illustrated in Figure 1.² When pK_acceptor > pK_donor, the forward reaction is rapid and independent of the pK_a difference. Protonation of various bases occurs with a rate constant >10¹⁰ M¹ · s¹, with the electrostatically favorable recombination of H⁺ and OH clocking in at 1.4 × 10¹¹ M¹ · s¹ (287). When the forward reaction is diffusion controlled, the reverse reaction will occur at a rate that is linearly related to the pK difference (Fig. 1A). By definition, log k_f  log k_r  pK_acceptor  pK_donor = pK (290). If the reaction is asymmetrical with respect to charge (e.g., HX + Y = X + HY⁺), then the diffusion-controlled limit will be different for the forward and backward reactions (Fig. 1B). A Brönsted plot (123a) provides similar information (787). A more thorough theoretical development of the kinetics of proton transfer invokes Marcus rate theory (654), as has been applied successfully to carbonic anhydrase (931).

View larger version (16K): [in this window] [in a new window]   Fig. 1. Idealized dependence of the normalized rates of proton transfer reactions on the pKa difference between donor and acceptor molecules participating in the reaction. In A, the transfer is symmetrical with respect to charge (e.g., HX+ + Y = X + HY+), whereas in B, the reaction results in charge neutralization. The slopes of the forward reaction () and backward reaction () limit at 0 or 1 at large pK. The limiting rate constant (kmax) is 109 to 1010 M1 · s1 for a diffusion-controlled reaction. [From Eigen and Hammes (290), copyright 1963 John Wiley & Sons, Inc. Reprinted by permission of John Wiley & Sons.]

In terms of a proton conduction pathway that is composed of a series of protonation sites, proton hops may not obey the same rules as proton transfer reactions in diffusion-controlled reactions, due to steric constraints, etc. However, the general principles of the pK dependence of transfer rates are likely to apply. Continuous prototropic transfer is most efficient when the donor and acceptor are symmetrical, as in water to water transfer (605). In solvent mixtures, the solvent with higher affinity traps the proton (605). Ab initio molecular orbital method calculations indicate that in a long water wire, multiple proton transfers (hops) can occur simultaneously (i.e., energetically coupled to each other) using the energy cost associated with a single transfer event (882). An example of coherent proton tunneling has been observed directly in a network of four coupled hydrogen bonds (465).

There is long-standing debate over the suggestion that protons may diffuse laterally at the surface of the membrane at a higher rate than they diffuse in bulk solution. The question has been discussed extensively in the context of bioenergetic membranes (404, 418, 527, 530, 706, 724, 731, 820, 821, 1002, 1079). This question has arisen in several instances in which the apparent single proton channel current is larger than the maximum rate at which protons can diffuse to the channel, as predicted by simple diffusion models. To some extent, surface enhancement may be ascribed to geometric factors, i.e., the difference between diffusion in two and three dimensions (353) without specifying the mechanism by which protons would bind to the surface. A proton trapped at the membrane surface will diffuse into a proton channel if it does not first desorb, whereas a proton in three-dimensional bulk solution has a low probability of diffusing into the channel. In unbuffered solutions, surface conduction dominates; in buffered solutions, the dominant pathway depends on protonated buffer concentration and the effective size of the proton collecting antenna (353) (see below).

One general way that surface conduction could enhance proton fluxes through a channel is by the "antenna effect" (400, 867). Rather than requiring a proton to diffuse directly to the channel entrance, the entire membrane surface, by virtue of its many negatively charged groups, might collect protons, which then travel in the plane of the membrane surface to the channel. Detailed experimental and computational studies have been done on this question (155, 353, 400, 653, 867). Protonation reactions are often extremely rapid and limited only by diffusion, with rate constants typically 1-6 × 10¹⁰ M¹ · s¹ (287, 290, 400, 653, 867). One of the most rapid reactions known is the recombination of H⁺ and OH with a rate constant 1.4 × 10¹¹ M¹ · s¹ (287). However, occasionally higher rate constants are observed. An anomalously high protonation rate measured for a site on a Ca²⁺ channel, 4 × 10¹¹ M¹ · s¹, was explained by proposing the site to be negatively charged and located in the channel vestibule, which would funnel the electric field lines and enhance the electrostatic attraction (823). If two negatively charged groups (e.g., at the surface of a membrane) are close enough together that their Coulomb cages overlap, the "virtual second-order" rate constant governing the transfer of a proton from one group to the other can be 10¹² M¹ · s¹ or greater (400), with the current record being 6 × 10¹² M¹ · s¹ (867). The probability that a proton that is bound to a site with 1 charge at the interface between membrane and aqueous solution will transfer to a neighboring site, also with 1 charge, rather than entering bulk phase, calculated with the Debye-Smoluchowski equation, is close to 100% for a 12-Å separation, decreasing with distance to ~40% for a 60-Å separation (867). It seems clear that rapid proton transfer in the plane of the membrane is possible.

On the other hand, the extent to which rapid surface conduction might play a significant role must be established in each specific situation. In a study on proton transfer rates between superficial amino acid groups on tuna cytochrome c oxidase, all of the virtual second-order rate constants were <10⁹ except for one that was as large as 10¹¹, which was between groups within 10 Å of each other (652). A cluster of three carboxylates on bacteriorhodopsin acts as a proton-collecting antenna, each with a high protonation rate of 5.8 × 10¹⁰ M¹ · s¹, but the dimensions of the antenna are smaller than those of the molecule. Long-range proton migration occurs along a protein monolayer, but depends critically on molecular packing, and is abolished at low or high protein densities (331). Molecular dynamics simulation indicates that proton transport near the surface of a dipalmitoylphosphatidylcholine membrane is inhibited rather than enhanced (953). Finally, de Godoy and Cukierman (253a) explored the effects of bilayer composition on H⁺ currents through gramicidin channels. The limiting H⁺ conductance at low pH was the same in bilayers formed from protonatable phospholipids that presumably should be capable of mediating lateral H⁺ conduction and bilayers formed from covalently modified phospholipids that cannot be protonated. Furthermore, differences in the H⁺ conductance at higher pH were fully accounted for by electrostatically induced changes in local H⁺ concentration near the membrane, providing no evidence of significant lateral H⁺ conduction (253a). In summary, it appears that rapid proton transfer at the membrane surface may occur under specialized conditions but cannot be assumed to occur generally.

The usual way to control pH is with buffered solutions. Because the control of pH is never perfect, recognizing systematic sources of error is useful. Voltage-gated proton channels appear to be perfectly selective for protons over all other ions besides deuterium, as discussed in section VE, and hence act as local pH meters (237). Selectivity is evaluated by measuring the reversal potential (V_rev) in solutions of various pH, and comparing the result with the Nernst potential for H⁺ (E_H)

(1)

Although reasonable agreement between the measured V_rev and E_H is often obtainable, the agreement is rarely perfect. If we tentatively accept the conclusion that voltage-gated proton channels are perfectly H⁺ selective (see sect. VE), then any deviation of V_rev from E_H indicates that the true pH differs from the nominal pH. The primary cause of this deviation in patch-clamp experiments is imbalance between the rate that proton equivalents cross the cell membrane and the rate the buffer from the pipette replenishes the cytoplasmic compartment. The intracellular compartment is a large unstirred volume, and proton efflux such as that occurring during H⁺ currents will deplete protonated buffer from the cell. For example, a 10-µm-diameter cell has a volume of 524 fl, and if it is filled with a pipette solution that has 100 mM buffer at its pK_a, the entire cell will contain 1.6 × 10¹⁰ protonated buffer molecules. During a modest sustained outward H⁺ current of 100-pA amplitude, 6.25 × 10⁸ H⁺ leave the cell each second, deprotonating 4% of the total protonated buffer. Even at intracellular pH (pH_i) 6 there are only 315,000 free protons in the entire cell, all of which would be consumed during 0.5 ms of H⁺ current. Thus, essentially the entire H⁺ current is carried by protons that immediately previously were bound to buffer molecules. Replenishment of buffer occurs by diffusion from the pipette solution and requires the diffusion of these rather large molecules through a small <1-µm-diameter pipette tip into the cell.

Calculations based on Pusch and Neher's empirical determination of diffusion rates (827) predict a time constant of 19 s for the equilibration of 250-Da buffer molecules from a pipette with 5-M tip resistance into a 15-µm-diameter cell. This time constant is proportional to cell volume (776). The rate of equilibration of pH_i will be slower than that for simple buffer diffusion, due to the effective slowing of H⁺ diffusion by fixed (immobile) intracellular buffers (514). Direct estimates of the time constant of equilibration of pH_i in HL-60 cells and macrophages of unspecified size were 11 s (258) and 58 s or 97 s (519), respectively, representing at least qualitative agreement.

The presence and action of any membrane transporter that moves proton equivalents across the cell membrane will alter V_rev. Thus, when Na⁺ is present only in the external solution and pH_i is low, the inward Na⁺ gradient and outward H⁺ gradient both conspire to activate Na⁺/H⁺ antiport. H⁺ extrusion by the antiporter is rapid enough to raise pH_i substantially (i.e., by 0.5 unit or more) in alveolar epithelial cells studied in whole cell patch-clamp configuration, in spite of the presence of 119 mM buffer in the pipette solution (237). H⁺ is extruded by the antiporter faster than the supply is replenished by diffusion of protonated buffer from the pipette. Geometrical factors influence this balance, with smaller cells or larger pipette openings attenuating the change in pH_i due to antiport activity. Thus manifestations of Na⁺/H⁺ antiport were less pronounced in human neutrophils (237) or murine microglia (546) than in the larger rat alveolar epithelial cells, but obviously differences in the expression of Na⁺/H⁺ antiport molecules could also play a role. Any other mechanism that results in net movement of H⁺ equivalents across the membrane will alter pH_i. Several mechanisms of membrane H⁺ flux are discussed in section IIIA, of which the shuttle mechanism in particular could cause attenuation of the pH gradient across the membrane (see sect. IIIA3).

A systematic deviation arises when V_rev is measured by the conventional tail current protocol. A depolarizing prepulse activates the H⁺ conductance (g_H) and then the voltage is repolarized to various levels, and the direction of the tail current (the decaying current waveform that reflects the progressive closing of H⁺ channels) is observed. The necessity to activate a substantial g_H during the prepulse to elicit an interpretable tail current, combined with the extremely slow activation kinetics of voltage-gated proton channels in mammalian cells, inevitably causes significant depletion of intracellular protonated buffer during the prepulse. If a comparable H⁺ current is elicited during the prepulse in solutions of varying pH, the error will be a relatively constant addition of a few millivolts to the measured V_rev. This systematic error may explain why the vast majority of V_rev measurements in the literature are more positive than E_H. On the other hand, V_rev measurements that encompass negative pH [pH_i > extracellular pH (pH_o)] indicate deviation in the opposite direction in this range (166, 519, 886), suggesting that an element of dissipation of any pH gradient may also play a role. As a result, measurement of the change in V_rev at several pH rather than the absolute V_rev often provides a cleaner estimate, which explains the fondness that many experimentalists have for this way of expressing their data. Direct measurements of V_rev using prepulses that elicit smaller or larger currents have been shown to raise pH_i and hence shift V_rev positively roughly in proportion to the integral of the outward H⁺ current during the prepulse (70, 232, 372, 473, 519, 709), although this effect is not apparent in large cells (134). It is important to recognize that the deviation of V_rev from E_H is not an error, but instead accurately reflects the effects of the pulse protocol on pH_i. We consider voltage-gated proton channels to be perfect pH meters (see sect. VE).

An expedient way to estimate V_rev is to activate the g_H and then ramp the membrane voltage "downward" from positive to negative (372). If enough channels open at positive voltages and the ramp is rapid enough that the channels remain open, then V_rev can be taken as the zero current voltage, although any leak conductance and capacity current must be either negligibly small or corrected. The problem remains that it is first necessary to activate the g_H to observe V_rev, so this approach does not avoid the problem of depletion. Another clever way to estimate V_rev is simply to interpolate between the H⁺ current at the end of a depolarizing pulse and that at the start of the subsequent tail current (473). One required assumption is that the instantaneous current-voltage relationship be approximately linear. This method is useful in certain situations, particularly if one suspects that significant depletion has occurred. The advantage is that both required data points are obtained by applying a single pulse, and they are measured at nearly the same time. Again, this approach does not avoid the effects of depletion. In fact, its originators used this approach to demonstrate that H⁺ efflux during large depolarizing pulses alkalinized the cytoplasm significantly.

H⁺ currents increase pH_i in proportion to the amount of H⁺ extruded. For small currents, the change in pH_i may be negligible, but for large currents, depletion of protonated buffer will noticeably increase pH_i. These effects are less pronounced in large cells (134) because they reflect the area-to-volume ratio. Restoration of pH_i is determined by the geometrical factors already discussed, and typically requires tens of seconds up to several minutes. A useful rule of thumb is that because voltage-gated proton channels do not inactivate, when the H⁺ current peaks and then droops during a sustained depolarization, this always reflects an increase in pH_i. Experimentally, this phenomenon can be annoying, but it is simply a manifestation of the ability of the H⁺ conductance to do its job, namely, to extrude acid at a rate adequate to alkalinize the cytoplasm rapidly.

Perhaps not surprisingly, variations in extracellular buffer from 1 to 100 mM had very little effect on voltage-gated proton currents (241). The bath solution represents an effectively infinite sink for protons. The situation for intracellular buffer is more complicated. Several whole cell patch-clamp studies in which pH_i was determined have revealed that including 5-10 mM buffer in the pipette solution does not control pH_i as well as higher buffer concentrations, e.g., 100-120 mM (232, 258, 519, 574). In addition, the time course of the H⁺ current during a single depolarizing pulse was shown to depend strongly on "internal" buffer concentration in excised inside-out patches of membrane (241). The initial turn on of H⁺ current was similar, but the longer the pulse, the more the current with 1 mM buffer drooped relative to that with 10 mM buffer. Nevertheless, decreasing internal buffer from 100 to 1 mM attenuated the H⁺ current by only ~50%; thus this effect is attributable to H⁺ current-associated pH changes, rather than a limitation of the conductance of the channel by buffer (241) (cf. sect. VJ).

In addition to buffers, application of an NH gradient has proven to be a useful way to control pH_i in patch-clamped cells (242, 248, 387) (see also sect. IIID). Control over pH_i is excellent and rapid when the NH gradient is symmetrical, becoming less effective for large NH (hence pH) gradients (248, 387). An advantage of this technique is that pH_i can be changed in a cell simply by altering the bathing solution.

Several issues related to buffers are relevant to the study of proton channels. Experimental control of pH requires adequate buffering, as just discussed in section IIE. Buffering power (or buffering capacity) is defined as dB/dpH (1036), i.e., the concentration of strong base required to change the pH of a solution by one unit. A more rigorous discussion of this and other definitions can be found elsewhere (849, 850). The reported buffering power of the cytoplasm in mammalian cells ranges from 18 to 77 mmol · pH¹ · liter¹ (850). The measured buffering power of most cells increases substantially at lower pH, typically three- to fivefold between pH_i 7.5 and pH_i 6.5 (24, 41, 92, 324, 603, 630, 840, 850, 1067). A similar observation has been made for the Golgi (153). The buffering power is maximal at the pK_a of the buffer (425, 1064), where it is (ln10)[B]/4 ~ 0.58[B], where [B] is the total buffer concentration (559, 849, 1036). Thus a cytoplasmic buffering power of 58 mmol · pH¹ · liter¹ would reflect the presence of the equivalent of at least 100 mM simple buffer in cytoplasm. To control pH experimentally, many investigators use solutions with 100 mM exogenous buffer near its pK_a. Under normal conditions, this is adequate to prevent pH changes large enough to alter H⁺ currents noticeably (240) (but see cautionary tales in sect. IIE).

When a cell is dialyzed with a pipette solution containing inadequate buffer, intrinsic cytoplasmic buffers override the attempts of the pipette solution to control pH_i. The larger the cell, the more difficult is the control of pH_i. Byerly and Moody (135) compared the rate of equilibration of pipette solutions containing K⁺ or highly buffered H⁺ with cytoplasm in large neurons (90-120 µm in diameter) studied with suction pipettes one-third the cell diameter. The effective equilibration of H⁺ even with high buffer concentrations (50-100 mM) was three to five times slower than that of K⁺, and with 20 mM buffer, little control over pH_i was achieved (135). Similarly, the effective diffusion coefficient of H⁺ in cytoplasm is five times slower than that of mobile buffers (15). In small cells studied with patch pipettes containing pH 5.5 solutions, pH_i deduced from the V_rev of H⁺ currents was ~5.7 for 119 mM MES buffer and ~6.3 for 5 mM MES (232). A pipette solution with 1 mM buffer appeared to have essentially no effect on pH_i (240).

Buffers have variable tendencies to chelate metal ions (805). Because we could not find much information on this property for normal pH buffers beyond the initial description of the Good buffers (370), we measured the binding constants of several buffers for Zn²⁺, Cd²⁺, Ni²⁺, and Ca²⁺ (163). Certain buffers bind Zn²⁺ avidly, including tricine and N-(2-acetamido)-2-iminodiacetic acid (ADA). The latter has been used to establish free Zn²⁺ concentrations in the nanomolar range (22, 792).

	III. MECHANISMS OF PROTON PERMEATION THROUGH MEMBRANES

Top Previous Next References

The Born self-energy cost of an ion permeating a pure lipid bilayer is prohibitive (794), ~58.6 kcal/mol for the H₃O⁺ (243). Therefore, a mechanism subtler than brute force is required to translocate protons across membranes. Four mechanisms that have been proposed to explain proton permeation through biological membranes are as follows: transient water wires (sect. IIIA2), weak base or acid shuttles (sect. IIIA3), phospholipid flip-flop (sect. IIIA4), and specific proteins (channels, carriers, and pumps) whose function is to transport protons. High "intrinsic" proton permeability must be explained by one of these mechanisms. As will become apparent however, the proton permeability of cell membranes that contain voltage-gated proton channels is several orders of magnitude higher than the highest estimate for simple phospholipid bilayers. In most cells with H⁺ channels, any proton permeability of the membrane itself is negligible in comparison (242).

It has been maintained widely and for some time that membrane proton permeability (P_H) is anomalous in two respects. First, P_H is many orders of magnitude higher (10⁴ to 10² cm/s) than the permeability of other cations (10¹² to 10¹⁰ cm/s) (227, 228, 390, 755, 797). Second, the proton conductance (G_H) is practically independent of pH (226, 395, 396, 755). These observations have been challenged on various counts, and some of the complications will be mentioned here.

P_H is difficult to measure, and reported values vary over many orders of magnitude, ranging from <10⁹ to 10¹ cm/s (153, 396, 585, 688, 755, 764, 766, 797, 804). Although various studies report no (124), moderate (585), or strong (i.e., up to ~100-fold) (228, 390, 396, 755, 764, 804, 1033) dependence of P_H on the composition of the membrane, this dependence does not come close to resolving the vast disparity in reported values. The idea that P_H is anomalously high was challenged by Nozaki and Tanford (766), who measured P_H 10⁹ cm/s in phospholipid vesicles and estimated the true value to be 5 × 10¹² cm/s. Deamer and Nichols (227) argued that these measurements were limited by development of a diffusion potential. Diffusion potentials can be avoided by allowing counterion flux (114). The finding that several cells have undetectably small P_H (185, 1054) suggests that proton permeability is not a general property of cell membranes.

Another source of variability may be differences between conductance and permeability measurements. Radioactive tracers reveal unidirectional flux, whereas electrical currents reflect only net flux, i.e., the difference between the unidirectional fluxes. For example, at E_H there is no net H⁺ current, but there still can be large bidirectional fluxes. Hence, permeability estimates based on fluxes may be higher than electrical estimates made near E_H. On the other hand, if H⁺ current is measured during a large driving voltage, fluxes will be practically unidirectional, so the two estimates should be reasonably consistent.

It has been suggested that both the high apparent P_H and the pH independence of G_H might be the result of proton accumulation near the negatively charged phospholipid head groups at the membrane-solution interface (342). In this view, P_H is high because its calculation assumes the bulk solution concentration and neglects the possibility that the local concentration of protons at the membrane surface may be proportionally much higher than other cations, due to the closer approach of H₃O⁺ than a hydrated cation to the negatively charged membrane. It has been known at least since 1937 that negative surface charges tend to lower the surface pH, by up to 2 pH units in physiological solutions (215, 378, 988). Numerous studies indicate that negative surface charges can concentrate protons and other cations near membranes, resulting in higher conductance than expected from bulk concentrations (32, 214, 531, 716). Higher P_H is measured in negatively charged phospholipid membranes (764). Furthermore, because the negative charges at the surface are essentially fully screened by protons, the local proton concentration is relatively independent of bulk pH, and thus the apparent insensitivity of proton flux to bulk pH is also explained (342).

A fundamental difficulty with measuring P_H is that in the physiological pH range, the [H⁺] is up to 10⁶ smaller than that of other cations. Because the calculation of P_H effectively normalizes the measured flux according to the nominal [H⁺], any error is magnified, and the error is amplified at higher pH. At least in electrical measurements, most errors tend to increase the apparent P_H. In alveolar epithelial cells studied by voltage clamp in solutions lacking small ions, P_H < 10⁴ cm/s, even assuming that the entire leak is carried by H⁺ (242). In fact, the "leak" current was insensitive to pH and the leak reversal potential did not change in a direction consistent with H⁺ selectivity, thus P_H 10⁴ cm/s by direct electrical measurement and any proton permeability was too small to detect (242). Similar observations were made in myelinated nerve (440). Also consistent with a low P_H, large changes in apical pH_o do not change pH_i in alveolar epithelial monolayers (510). From the viewpoint of a cell trying to maintain homeostasis, any proton leak is undesirable. In light of the >10⁴ increase in P_H that occurs when the cell membrane is depolarized and H⁺ channels open, the background level of proton leak is negligible for most purposes.

It is questionable whether the traditional permeability coefficient P_H is useful for H⁺ flux through either membranes or most channels. The Goldman-Hodgkin-Katz (GHK) model (368, 444, 456) assumes that permeation is a simple process that occurs at a rate proportional to the rate that the permeant ion species encounters the membrane, which in turn is proportional to the bulk concentration. P_H is thus predicted to be a constant that is independent of pH, and lowering the pH by one unit should increase the H⁺ flux (or g_H) 10-fold. In fact, deviations from this prediction are more the rule than the exception. To the extent that simple membrane H⁺ conductance is independent of [H⁺] (226, 395, 396, 755), the parameter P_H, far from being constant, increases 10-fold/unit increase in pH. The P_H of Golgi membranes increases 3.4-fold/unit increase in pH (153). P_H calculated in alveolar epithelial cells during maximal activation of H⁺ currents increases ~5-fold/unit increase in pH (166, 242). This type of behavior demonstrates that these systems do not operate within the assumptions built into the GHK permeability equations, and hence, permeability calculations have little meaning. In contrast, for gramicidin P_H is constant over a wide pH range; i.e., the single-channel H⁺ conductance increases 10-fold/unit decrease in pH (Fig. 13). This counter-example suggests that the pH dependence of P_H in other systems does not reflect something peculiar about the diffusion of protons to membranes, at least at pH <5. Instead, it more likely indicates that a rate-limiting step in the permeability process is slower than the diffusional approach of protons to the membrane. In the case of voltage-gated proton channels, permeation through the channels is thought to be rate determining (166, 234, 238-240, 242-245). The GHK equations provide a valuable frame of reference by predicting the behavior of a simple system. However, in the frequently occurring situations in which P_H depends strongly on pH, the parameter P_H is not a meaningful way to evaluate or compare proton fluxes.

A transient water wire might occur if, due to thermal fluctuations, a chain of water molecules happened to align across the membrane (225, 228, 755). Although fatty acid monolayers and cell membranes present a significant barrier that slows water diffusion by ~10⁴ (34, 147), water can permeate most cell membranes, and several waters might follow the same path once a trailblazer has led the way. A hydrogen-bonded chain of water molecules intercalated between membrane phospholipids might be imagined to conduct protons. A membrane-spanning chain would need to be ~20 water molecules long, and the Born energy cost of forcing a proton into the bilayer might be reduced by virtue of partial hydration by nearby waters (730). The proton flux could be independent of pH if the rate-determining step were the breaking of hydrogen bonds between neutral waters, which might initiate the turning step of the hop-turn mechanism (730) (see sect. IIID). A recent modification of this idea is the translocation of protons by small clusters of water molecules in the membrane (405).

There are some difficulties with the transient water wire proposal. Although water permeability varies 27-fold in different synthetic membranes (309), and P_H varies ~100-fold in different membranes, there is no correlation between P_H and water permeability (396). Molecular dynamics simulations indicate that the free energy barrier to formation of a water wire in a membrane is 108 kJ/mol, and thus the likelihood of a membrane-spanning pore forming is very low (658). The lifetime of such a water wire was <10 ps in this study (long enough to transport no more than one proton) and averaged 36 ps in a later simulation study (1038). The H⁺ flux calculated for this mechanism could be made to agree with experimental estimates only by assuming that a proton permeates instantaneously and that the entry rate of protons into the water wire is 10⁸ faster than provided by diffusion (658). Furthermore, simulations of H⁺ permeation through optimal water wires indicate that ~100 ps is required for H⁺ to permeate a 30-Å channel (120), which is longer than the predicted lifetimes of the transient water wires (658, 1038). The mean interval between H⁺ permeation events through gramicidin during the largest H⁺ currents recorded through any ion channel (2.2 × 10⁹ H⁺/s in gramicidin at +160 mV and 5 M HCl) (207) is 455 ps, which may or may not represent the maximum conduction rate (see sect. IVA4). A spontaneous water wire would have to be narrow and transient, because otherwise other ions might permeate (730), violating the observation that P_H is 10⁶ greater than that of other ions (755). Paula et al. (797) reported that P_H decreased from ~10² to ~10⁴ cm/s as the bilayer thickness was increased from 20 to 38 Å, and concluded that protons were conducted via transient water wires in thin membranes and by a solubility-diffusion mechanism in thicker membranes. As pointed out by Deamer (225), if P_H measured in biological membranes was found to be lower than in model (5) membranes, then the latter would be poor models, because biological membranes may have a variety of additional transport mechanisms that would, if anything, increase H⁺ flux. If water wires conduct protons across ordinary cell membranes, then they do so at a rate that is negligibly low compared with the proton fluxes that occur when voltage-gated proton channels are active (242).

Protons can cross membranes via weak acids or weak bases that act as proton carriers (106, 169, 671). It has been suggested that contaminant weak acids might account for the high P_H reported in phospholipid bilayer membranes (396). The weak acid mechanism has long been recognized (486) and is illustrated in Figure 2. When a weak acid is added to the extracellular solution, the protonated form (HA) will be present at a concentration determined by its pK_a and the pH as described by the Henderson-Hasselbalch equation (415, 425). The protonated form can permeate the membrane far more readily than the anionic form (A), and thus the predominant result will be entry of HA down its gradient into the cell. Once inside, HA will dissociate into A and H⁺, to an extent determined by pH_i. The net result is that protons have been transported into the cell and released there, thus increasing pH_o and decreasing pH_i. The addition of a weak base will have the opposite effect. Again, the neutral form is far more permeant, but when B, a weak base, enters the cell, it leaves its proton behind, lowering pH_o, and once inside the cell it will tend to bind H⁺ thus increasing pH_i. The neutral form of the acid or base will continue to diffuse across the membrane until its concentration is the same inside and outside the cell.

View larger version (11K): [in this window] [in a new window]   Fig. 2. Diagram illustrating the effects on local pH when weak acids (A) or weak bases (B) are present. The neutral form of each molecule typically is many orders of magnitude more permeant than the charged form. If the acid or base is present on one side of the membrane, the neutral form will permeate and change the pH on both sides of the membrane. The protonated weak acid, HA, carries its proton across the membrane and then may dissociate inside the cell, lowering intracellular pH (pHi) and increasing extracellular pH (pHo). These pH changes will be buffered, and the extent of the change will depend on geometrical considerations. The deprotonated weak base will permeate, in effect leaving a proton behind, and will tend to pick up a proton inside the cell, increasing pHi and lowering pHo.

A corollary to this mechanism is that weak acids and bases tend to equilibrate across membranes according to the pH on each side, which is important for determining intracellular drug concentrations (e.g., Refs. 233, 443, 744). This mechanism has been exploited as a way to estimate the pH inside cells or organelles (e.g., Refs. 152, 703, 1045). Another application of this phenomenon is the NH prepulse technique (850), which is a standard method to study pH_i recovery from an acid load. This principle has been exploited to regulate pH_i in cells under whole cell voltage clamp (242, 248, 387). One can establish a known NH (or triethylammonium⁺, for example) gradient by including a known concentration in the pipette solution, and then adjusting the NH in the bathing solution. Ideally, the NH gradient will establish an equivalent H⁺ gradient. For example, 5 mM NH in the bath and 50 mM NH in the pipette (and thus in the cell) will lower pH_i by 1 unit relative to pH_o.

Because of their exquisite sensitivity to pH, voltage-gated proton channels are effective pH meters that can be used to report pH changes (237). Adding NH to the bath produces intracellular alkalinization, which greatly diminishes H⁺ currents (473). Conversely, addition of sodium lactate or sodium acetate to the external solution rapidly and effectively acidifies the cytoplasm, enhancing voltage-gated proton currents (473, 710).

As a practical consideration, if one wants strict control over pH_i, one must worry about the presence of weak acids or bases in the solutions. Obviously, small molecules with pK_a near ambient pH (e.g., HCO, NH, etc.) are perilous, but even larger molecules with pK_a >2 units from ambient may produce significant changes in pH_i by the proton shuttle mechanism. For example, N-methyl-D-glucamine (NMDG), a commonly used large "impermeant" cation with pK_a 9.6, can cause significant shunting of the pH gradient by the shuttle mechanism (938). Whether it does so quickly enough to affect H⁺ currents in a patch-clamped cell has not been reported, but deviations of V_rev from E_H appear somewhat greater in a study using NMDG solutions (232) than tetramethylammonium⁺ solutions in the same cells (166). Tetrabutylammonium⁺ is sufficiently lipophilic to permeate cell membranes (233) and has been shown to enhance proton flux (764).

Another mechanism that might allow net proton flux across a membrane is phospholipid flip-flop (396). This is a subset of the weak-acid mechanism just discussed, but does not require any molecules exogenous to the membrane. Membrane phospholipids might transport protons, acting effectively as carriers. The negatively charged phosphate groups may become protonated, neutralizing their charge, and then the molecule could flip-flop across the membrane, releasing the proton on the other side. Long-chain fatty acids can also transport protons across membranes by this mechanism (397). Biological long-chain fatty acids may transport protons across the membrane by a weak acid mechanism, although their slow intrinsic flip-flop rate makes them relatively inefficient (397). Although it seems likely that this mechanism can occur under some conditions (397, 547), it has been argued that it does occur only at relatively high concentrations of fatty acids, such as 300 µM oleic acid (327).

Direct evidence of gating is provided by current fluctuations (see sect. VG) as well as direct single-channel currents (168). Gating is a defining characteristic of ion channels.

The result of time- and voltage-dependent gating is that over a span of ~40 mV, the membrane permeability to H⁺ increases reversibly by at least three to four orders of magnitude (242). Again, it is difficult to imagine how a simple membrane, even if perforated by "water wires," could manifest such a remarkable transition.

As a result of voltage-dependent gating, the steady-state g_H rectifies strongly. Outward H⁺ flux is at least three to four orders of magnitude greater than inward H⁺ flux, which is undetectably small. How could this kind of rectification (asymmetry of flux) be accomplished across a simple membrane?

The H⁺ current through voltage-gated proton channels is 1.9 times larger than D⁺ currents (242), substantially larger than the isotope effect for simple membrane permeation (226, 804) (see sect. VI).

H⁺ currents are inhibited by Zn²⁺ and other polyvalent cations, selectively, with high affinity, and in a complex manner, by metal ions binding to a site with an apparent pK_a 6.2-7.0 (163). Such effects are readily explained by interaction with a membrane protein but difficult to explain otherwise.

If protons permeated the membrane itself, H⁺ flux ought to be governed by simple diffusion theory, i.e., the Fick equation applied to membranes (368). H⁺ flux should be proportional to concentration, and nonsaturable, limited only by diffusion of buffer across the unstirred layer near the membrane (399, 527). As will be discussed further in sections IVO and VF, over a range of 4 pH units, the g_H,max increases only ~2-fold/unit decrease in pH.

Substances such as phloretin or sodium dodecyl sulfate that alter the internal dipole potential of membranes and thereby profoundly affect ion conductances mediated by carriers (29, 955) have no effect on proton currents (V. Cherny and T. DeCoursey, unpublished observations). One can imagine that ions inside a channel protein might be shielded, but if protons permeated the bilayer itself, one would expect sensitivity to the internal dipole potential.

In conclusion, proton channels do exist and are almost certainly membrane proteins.

Ion channels are membrane proteins that provide a low-resistance pathway across cell membranes. That voltage-gated proton channels facilitate H⁺ efflux across membrane is evident from the observation that opening H⁺ channels by depolarization of the membrane potential increases P_H by >4 orders of magnitude (242).

Voltage-gated proton channels are entirely passive. An open H⁺ channel permits passive H⁺ conduction down the electrochemical gradient. H⁺ channels cannot be considered to be "pumps" in any sense of the word. Removal of ATP prevents neither H⁺ channel opening nor H⁺ conduction. ATP has subtle effects (574, 710), but these are unrelated to phosphorylation (710) (Cherny and DeCoursey, unpublished data) (see sect. VIB3).

In distinction from many carriers, symporters, and antiporters, no co-ion or counterion is required (238).

A fundamental distinction between carriers and channels is that carriers must undergo a conformational change during each ion translocation cycle. The argument becomes semantic at this point. If H⁺ conduction occurs by a HBC mechanism (see sect. IIID), the turning step of the hop-turn mechanism arguably might be considered a conformational change. However, ions probably interact with normal ion channels during permeation, and it is possible that conformational changes in the protein (induced by the presence of the ion) must occur before conduction can proceed. Second, the rearrangement of hydrogen bonds required during the turning step may be subtle and hardly qualifies as a conformational change. Finally, the distinction of carriers from channels based on the conformational change criterion is invoked to explain the lower turnover rate of carriers. In fact, the participation of a protonatable residue at the entrance to several proton channels has been shown to increase the efficiency of proton conduction (see sect. IVN). In summary, the term channel is appropriate.

Voltage-gated proton channels exhibit gating: reproducible time- and voltage-dependent activation and deactivation of H⁺ current. Excess current fluctuations that reflect stochastic opening and closing transitions, i.e., gating, have been observed (168, 236, 720). Demonstration of the existence of gating is often presented as "proving" an ion channel mechanism. Whether carriers might exhibit behavior interpretable as gating is unclear. By this criterion, voltage-gated proton channels are ion channels.

As the defining property of voltage-gated channels, including proton channels, gating is a major feature that distinguishes channels from other types of transporters. A channel without gating is simply a pernicious hole in a cell membrane. In contrast, the activity of carriers (porters and pumps) is mainly regulated by substrate availability, and secondarily by biochemical modulation. Carriers can perform their physiological functions without a clear requirement for gating. Specifically, porters and pumps have no correlate of the full open state of ion channels, in that at no time in their reaction cycle is there a continuous pathway for the ion across the membrane. The open state enables channels to have high turnover rates, whereas carriers must undergo conformational changes during each transport cycle.

Voltage-dependent gating must be distinguished from voltage-sensitive flux. Any process that results in net charge translocation across a cell membrane must in principle be voltage sensitive. The ion flux will depend on the driving force (596), which includes the electrical potential difference (voltage) across the membrane. Simple diffusion of ions across membranes is voltage sensitive, and so must be ionic flux through porters and pumps whose stoichiometry of ion movement is unbalanced, so that net charge translocation occurs. Well-known examples include the Na⁺-K⁺ pump (833), the Na⁺/Ca²⁺ exchanger (540), a Na⁺/HCO cotransporter (848), the H⁺-dependent glucose transporter (947), and many H⁺/amino acid transporters (103, 875). The ion transport rate varies with voltage because each cycle of the carrier delivers net charge across the membrane's electric field. Even if the charge-transferring steps are not rate limiting, the overall process must still be voltage sensitive because voltage will affect the probability that the transporter exists in states immediately adjacent to the rate-limiting step (596). However, the voltage sensitivity may not be very obvious for a particular measurement. For example, the current generated by the H⁺-ATPase in Neurospora changed less than twofold over 300 mV (377). The translocation of electrons across the membrane by NADPH oxidase is nearly voltage independent over a 150-mV range (252). In a model of pump currents, Hansen et al. (410) showed that the current-voltage relationship could be flat or nearly so over a wide voltage range, but steep at other voltages. Voltage gating, in contrast, implies a discontinuous process: a clear difference in the mode of operation of the transporter protein at different voltages. In the case of voltage-gated ion channels, the probability of being open or closed (conducting or not) depends on membrane voltage. Channel gating may reflect a conformational change in the protein or, in some cases, occlusion of the conducting pathway. For all voltage-gated channels, gating is stochastic: the probability of being open or closed depends on voltage. The current through an open ion channel is voltage sensitive, generally increasing as the voltage is increased relative to the reversal potential.

It can be argued that carriers and pumps must function to a variable degree of effectiveness and that this is equivalent to the gating of ion channels; that is, carriers may also exist in states of low functional probability, which are incapable of reacting with the substrate. This circumstance is obvious when a noncompetitive inhibitor is present, but can in principle occur under less well-defined conditions, for which the term lazy-state behavior has been coined (411), corresponding to the "closed" ion channel. If we could look at individual carriers, as we can at individual channel molecules, we should see these noncycling intervals (C. L. Slayman, personal communication). However, thus far it has been impossible to measure transport through individual carrier molecules (whose maximal currents would be in the attoampere range), so direct demonstration of this phenomenon is lacking. It has been proposed for the F_o proton channel of H⁺-ATPase (1046), that the interaction between Trp²⁴¹ and His²⁴⁵ comprises a "gate." Protonation of His²⁴⁵ at low pH allows interaction with Trp²⁴¹, which by conformational changes or pK_a shifts, as speculated, allows protons to enter the channel and access the crucial Asp⁶¹ (see sect. IVF). The term gate has also been applied to bacteriorhodopsin (the best understood "active" transporter) in a similar sense, to describe the conformational change in the Schiff base that causes proton flux to be unidirectional (969). In both of these cases, however, the distinction from channels remains, because an open H⁺ channel allows continuous H⁺ flux across the membrane down its electrochemical gradient, which does not occur during the normal functioning of bacteriorhodopsin, F-ATPases, or any other carrier-type protein.

The only conceivable alternative descriptor, carrier, is inappropriate (see sect. IIIC). Ion carriers (uniporters) must be voltage sensitive, because at least one form (ion-bound or ion-unbound) must carry net charge across the membrane. As a result, a large applied voltage may trap the carrier at one side of the membrane, and hence transport will not be sustained. The resulting transient current has been described for mutant forms of voltage-gated K⁺ channels (R365H and R368H) in which His shuttles protons across the membrane (960, 961). The elegant analysis of possible outcomes of histidine scanning studies of the voltage sensor of K⁺ channels by Starace and Bezanilla (960) distinguishes between carriers and channels. A protonatable His acting as a carrier binds a proton at one membrane surface, moves during voltage-dependent gating to a new position in which the protonated His is exposed to the other membrane face, and then releases the proton. The result is sustained H⁺ current that is maximal near voltages where P_open is 0.5, i.e., where the probability of gating transitions is maximal (Fig. 3C). In contrast, mutants in which His becomes accessible simultaneously to both membrane surfaces act as H⁺ channels, providing a continuous pathway for protons to cross the membrane. This proton channel turns out to be gated because only in one conformation, whose probability of occurrence is voltage dependent, is the His accessible to both sides of the membrane. In this case, the g_H has a normal sigmoid voltage dependence like other voltage-gated channels. As shown in Figure 3, A and B, the H⁺ current increases monotonically with voltage over a range of 400 mV in native voltage-gated proton channels. This behavior is channel-like.

View larger version (22K): [in this window] [in a new window]   Fig. 3. Absence of saturation of voltage-gated H+ currents contrasted with nonmonotonic voltage dependence of carrier-mediated H+ currents. A: H+ currents are illustrated for pulses from a holding potential (Vhold) of 60 mV, in 20-mV increments up to +380 mV, at pHo = pHi = 7.0 in a human eosinophil studied in permeabilized-patch configuration, as generally described in Ref. 246. The pulse duration was reduced at larger depolarizations to avoid depletion of cytoplasmic protonated buffer, which nevertheless occurred during some pulses as evident from the droop of the current. The records with heavier lines were recorded at a faster time base (calibration bar on left). This cell had been stimulated with PMA (phorbol ester) and then treated with diphenylene iodonium. B: H+ current-voltage relationship from the data in A indicates no hint of saturation (V. V. Cherny and T. E. DeCoursey, unpublished data). C: nonmonotonic current-voltage relationships at three pHi in a K+ channel mutant (R365H) that acts as a proton carrier. The current disappears at large positive or negative voltages because a form of the carrier is pinned at one side of the membrane. [From Starace et al. (961). Copyright 1997, with permission from Elsevier Science.]

One objection to the term channel is based on the miniscule single-channel conductance. Traditionally (444, 596), channels, carriers, and pumps are characterized as having distinctive maximum turnover rates: 10⁵-10⁸, 10²-10⁴, and 10¹-10³ s¹, respectively. Although it is reasonable to argue that finding a turnover rate much higher than the typical range suggests an erroneous classification, the same logic does not apply to a smaller-than-typical turnover rate. If a putative carrier translocated 10⁷ ions/s, one might suspect that it was in fact a channel. However, if a channel conducts only 10⁴ ions/s, this just means it is a channel with a low conductance. In the case of H⁺ channels, the permeant ion normally is present at concentrations <10⁷ M. The H⁺ conductance of the gramicidin channel is the largest of any ion channel at very low pH (see sect. IVA), but extrapolated to pH 7 (see sect. IVP) is smaller than that estimated for voltage-gated proton channels.

Proton permeation through a narrow channel or through a protein is generally considered to occur by a mechanism different from the permeation of other cations, just as proton conductance in bulk water differs from that of other cations. Myers and Haydon (723) explained the anomalously large proton conductance through the gramicidin channel by a Grotthuss mechanism of protons hopping across the row of water molecules inside the channel. The gramicidin channel is known to be a narrow pore occupied by a dozen or so water molecules in single file (611). Protons can also permeate channels that do not contain a continuous row of water molecules.

Lars Onsager explicitly proposed what has become known as the hydrogen-bonded chain (HBC) mechanism in 1967 (778, 779). He proposed that ions (778), including protons (779-781), might cross biological membranes through networks of hydrogen bonds formed between side chains of amino acids in membrane proteins. This mechanism was abandoned as a means of cation permeation except for the special case of protons (276, 471, 733, 734). The unique properties of protons make it possible to devise a pathway through a membrane protein that is not a water-filled pore like traditional ion channels (276, 733, 900). Nagle and Morowitz (733) considered in detail the properties and nature of proton conduction via a HBC. The HBC may comprise water molecules, side groups of amino acids capable of forming hydrogen bonds, or a combination of the two. Amino acids suggested as potential HBC elements include Ser, Thr, Tyr, Glu, Asp, Gln, Asn, Lys, Arg, and His (734). Zundel (1102) has measured large proton polarizability, which he considers to indicate facilitation of proton transfer, in Tyr-Arg, Cys-Lys, Tyr-Lys, Glu-His and Asp-His hydrogen bonds. Conduction across a HBC occurs by migration of defects or faults. Bjerrum (96) described two classes of defects in ice: orientational and ionic. Two main types of orientational faults can occur as a result of rotation of one water molecule through 120° (Fig. 4). A Bjerrum D (doppelt = double) defect occurs when two neighboring water molecules are oriented with two protons between them. A Bjerrum L (leer = vacant or empty) defect occurs when the oxygens of two adjacent waters point toward each other. These orientational defects can propagate through the ice crystal (Fig. 4C). Two types of ionic defects occur in ice when H₃O⁺ and OH are formed. These ionic defects migrate by means of proton jumps. Various other defects in ice have been proposed (490).

View larger version (8K): [in this window] [in a new window]   Fig. 4. Formation of Bjerrum orientational faults in ice. A: normal orientation of water molecules with hydrogen bonds (implicit) between each oxygen and a hydrogen of a neighboring molecule. B: rotation of one water molecule results in two types of orientational faults: the D defect where two hydrogens point toward each other, and the L defect where there is no hydrogen between two oxygens. C: illustrates the infrequent occasion when these orientational defects separate and then migrate through the ice crystal. [Redrawn from Bjerrum (96).]

The general features of HBC conduction are illustrated in the diagram in Figure 5. Proton conduction occurs in two obligate steps, called the "hop-turn" mechanism. The hopping step reflects the movement of the ionic defect, whereas the turning step reflects the propagation of a Bjerrum L fault from the right (distal) side of the channel back to the left side. In Figure 5A, a proton enters the HBC from the left, and through a series of jumps, all of the protons in the chain advance, and the terminal proton exits into the solution at the distal end of the channel. The proton that exits is not the same one that entered, but the net result is that one proton disappears from the proximal solution and one proton emerges into the distal solution. A distinctive feature of HBC conduction illustrated in Figure 5B is that after the "hopping" step depicted in Figure 5A, the chain is oriented differently than before, such that another proton cannot enter the chain from the left. First, it is necessary to reorient the entire chain, which in the example shown is accomplished by rotation of each hydroxyl group. Presumably, the hopping and "turning" steps of this hop-turn mechanism occur sequentially. A consequence of the hop-turn mechanism is that an empty proton channel has a "memory" of the last proton to permeate, which persists until the turning step is complete. Another consequence is that in the absence of a membrane potential or other orienting factor, an approaching proton has only a 50% chance that the HBC will be oriented correctly.

View larger version (13K): [in this window] [in a new window]   Fig. 5. Diagram illustrating the hydrogen bonded chain (HBC) mechanism for proton conduction. In this example, the HBC is formed by hydroxyl groups (e.g., from serine residues). A: a proton enters the chain on the left side and then as a result of the series of proton hops indicated by the arrows, a proton exits the chain on the right. B: after the hopping process is complete, the orientation of the HBC is different, and another proton cannot enter the chain from the left. First, it is necessary for the chain to reorient, in this example by rotation of all of the hydroxyl groups as indicated by the arrows. This turning process must be complete before another proton can enter the chain in the same direction. [From Nagle and Morowitz (733), copyright 1978 National Academy of Sciences, USA.] C: proton conduction along a simple water wire. [From Hammes-Schiffer (408), copyright 1998 American Chemical Society.]

An intriguing aspect of the HBC conduction mechanism is that the net translocation of one proton across the membrane does not result in the translocation of one full elementary charge. Part of the charge is translocated during the turning step (781). In ice, the hopping step translocates 0.64e, and the remaining 0.36e is translocated during the turning step (782, 885); in gramicidin the corresponding values are 0.69e and 0.31e (901). The reorientation of hydrogen bonds within the HBC results in a net charge movement within the membrane, acting as a capacitive load. Both processes are favored by an appropriate electrical driving force (i.e., positive on the side from which proton flux originates). It would be intriguing to devise an experiment to demonstrate that H⁺ conduction occurs in these two steps.

For various reasons, it generally has been believed that in HBC conduction the hopping step is faster than the turning step, by an order of magnitude or more (733, 734, 810, 814, 816, 818, 819, 900). With a few notable early exceptions (84, 366), the rate-determining step in H⁺ conduction in water is considered to be the reorientation of water molecules rather than proton hopping (184, 448, 535, 818). If the turning step were also rate determining for H⁺ current through voltage-gated proton channels, then it would be more reasonable to consider H⁺ conduction in terms of the voltage-driven HBC reorientation rather than proton hopping.

Although the idea that protons permeate channels via a HBC mechanism has become widely accepted, some have questioned whether the concept is overused. Citing the example of ATP synthases that are driven by translocation of Na⁺ instead of H⁺, Boyer (112) suggested that the hydronium ion, H₃O⁺, may be the transported species. The HBC concept was proposed before any proton channel structure was known. Over the past two decades, specific proposed proton pathways have generally progressed from being mostly amino acids side groups (e.g., Refs. 189, 392, 483, 969) to including more and more water molecules (803). Many amino acids can be demonstrated by mutagenesis to play important roles in creating proton pathways, but it is very difficult to distinguish whether a particular amino acid comprises a direct element in the HBC or simply is required to preserve structure or to constrain water molecules in an appropriate position (458, 633, 693, 765). It may be that nontitratable amino acids contribute to proton conduction not as direct elements in the HBC but rather by providing the correct microenvironment for water molecules that actually conduct the protons. It seems clear that titratable amino acids with low enough pK_a not to hold protons too tightly (Asp, Glu, and His) do relay protons in several channels, including M₂, mutant K⁺ channels, H⁺-ATPase, carbonic anhydrase, the bacterial reaction center, and bacteriorhodopsin (Table 2). P. A. Loach (personal communication) has suggested that a prototypical transmembrane proton channel may consist of water microchannels, pockets that contain a small number of water molecules. Protons or solvated hydroxide ions could move freely within a microchannel by a hop-turn mechanism, principally via water molecules stabilized by ligands (e.g., carbonyl groups) from the protein. Gating in this context is viewed as the establishment or breaking of proton pathways that interconnect neighboring microchannels, and the microchannels with the aqueous interfaces. Single, titratable acid-base groups, such as the side chains of Asp and Glu, may play a critical role in interfacing and gating between microchannels. In addition, the water content may change drastically during conformational changes that occur in the protein, thus providing another mechanism for "gating," i.e., completing or disrupting the continuity of the proton pathway.

Because of the importance of proton pathways in various proteins, much effort has gone into studying simple systems. A favorite model ion channel that has been studied exhaustively is the gramicidin channel, in which H⁺ is conducted along a simple single-file water wire (see sect. IVA). In addition to experimental studies, molecular dynamics simulations have been done on proton transfer along water wires inside gramicidin or imaginary channels, or even in simple chains of water molecules (Fig. 5C). In general, the more detailed and exhaustive the calculation, the simpler the system is in terms of numbers of molecules, by computational necessity. Several themes emerge from these studies. The (reversible) transfer of a proton from one water to the next is very rapid (270, 868), with a mean rate of 1.2 ps¹ (817). Proton transfer is activationless in a simple water wire, suggesting that propagation of the bonding defect is rate determining (814, 818). The hydrogen bond length and angle are critically important (882); linear or angular deformation of the hydrogen bond results in higher energy barriers to proton transfer both in water wires (882) and other types of HBC (883, 948). The longer the hydrogen bond, the higher the energy barrier to proton transfer (884). This feature is relevant to the finding that electrostriction (shortening of the distance between waters) can occur in ion-occupied channels (271). Quantum effects play a significant role (231, 408, 817). The intriguing idea of concerted transfer, i.e., nearly simultaneous hopping of multiple protons along the chain (as elements in the net transfer of a single proton) as a semicollective process has been discussed and may occur to some extent (231, 816-819, 868, 882, 890, 948). Concerted transfer is facilitated when the hydrogen bonds are equivalent (948). Further results of molecular dynamics calculations specific to gramicidin channels are discussed below (see sect. IVA).

Nagle and Morowitz (733) considered the possibility that water inside a confined geometry such as a narrow ion channel might be constrained, and that proton conduction in such a water-filled pore might behave more like proton conduction in ice than in liquid water. This prediction has been supported strongly by an assortment of molecular dynamics calculations, which conclude that water inside confined spaces like narrow ion channels diffuses or reorients much more slowly than in bulk water, i.e., the water inside channels is to some extent "frozen" (118, 120, 173, 174, 295, 317, 402, 414, 509, 726, 834, 856, 877). Obviously, structural details matter: in large-diameter pores with smooth nonpolar walls, calculated water self-diffusion can actually increase beyond bulk values due to a molecular-level capillary action-like effect related to a paucity of hydrogen bonds near the walls (414), but nevertheless, structures like real biological channels exhibit distinctly reduced water mobility. In various microcavities and confined structures, both the apparent dielectric constant of water and the diffusion coefficient of protons increased with the diameter of the confined space (117). A complication arising from the analogy with ice is that we do not understand proton conduction in ice very well (see sect. IIB). Eigen (287) suggested that the very rapid proton conduction in ice reflected proton delocalization due to the rigidly ordered (by hydrogen bonds) ice structure; ironically, present proponents of the idea that proton movement in ice is greatly inhibited at low temperature use a similar argument, namely, that the rigid structure prevents the water rotation required for HBC conduction (188). In any case, the general consensus seems to be that any restriction of water mobility will tend to reduce proton mobility. Bernal and Fowler (84) considered the low dielectric constant of ice to indicate that most of the water molecules were not free to rotate, consequently decreasing proton conductance. Proton diffusion inside the phoE channel is reduced to 50% its value in bulk water (402); a similar reduction has been calculated for gramicidin channels (14, 207). A cleft in lactose permease exhibits reduced water activity and a threefold reduction in the H⁺ diffusion coefficient (726). Despite these general considerations, there is evidence in several systems that amino acids that line proton channels may function by constraining or orienting water molecules to facilitate proton transfer (458, 633, 693, 765). In fact, hydrogen bonding between the waters inside gramicidin and the channel wall facilitates proton transfer (819), and the diffusion coefficient for protons inside gramicidin channels, calculated from molecular dynamics simulations, is 40 times larger than that in bulk water (901). Finally, a recent molecular dynamics study of proton conduction in smooth, cylindrical, hydrophobic, water-filled pores indicated that although the water diffusion constant decreased with pore radius, proton mobility increased sharply at 2-Å radius compared with larger pores, providing evidence of water-wire behavior when waters are in single file (120). Ultimately, although kinetic competence is a prerequisite, the speed of proton transfer may be less critical biologically than the establishment of conditions that allow it to take place in a controlled and predictable manner.

	IV. CLASSES OF PROTON-PERMEABLE ION CHANNELS

Top Previous Next References

A number of molecules that conduct protons during their normal operation, or have at one time or another been thought to do so, are discussed here briefly, roughly in order of increasing complexity. The list is arbitrary and not meant to be complete. Many other molecules could have been included, such as Photosystem II used by green plants during photosynthesis (266, 1014), proton-translocating transhydrogenases (95, 865), fumarate reductase (583), the flavoprotein p-hydroxybenzoate hydroxylase (897), ferridoxin I from Azotobacter vinelandii in which electron transfer is controlled by coupling with the more discriminating process of proton transfer (156), etc. A number of proton-conducting molecules were reviewed recently in a special issue of Biochimica et Biophysica Acta (Vol. 1458, No. 1, 12 May, 2000).

A.  Gramicidin

1.  Gramicidin is a water-filled pore

The gramicidin channel, a peptide antibiotic composed of 15 amino acid residues formed by Bacillus brevis, has been studied extensively as a simple prototypical ion channel. Gramicidin is in some ways reminiscent of a synthetic channel discussed in section IVC in that it is a small polypeptide with no formal charges, yet it is cation selective and can conduct protons. The gramicidin channel is a dimer formed when two hemi-channels assemble together head-to-head in the membrane (310, 1029). It is readily incorporated into artificial phospholipid bilayer membranes, permitting control over the solutions on both sides of the membrane, and it tolerates extreme voltages and ionic conditions that would destroy ordinary biological ion channels in situ. There is strong experimental evidence that the gramicidin channel is a water-filled pore. Streaming potential and electro-osmotic flux measurements, both of which reflect the enforced concerted motion of water molecules and ions in a long, single-file pore, indicate that gramicidin channels contain ~12 water molecules (310, 611). Molecular dynamics simulations are consistent with the presence of ~7-10 water molecules inside the pore (173, 316, 509, 816, 819, 856, 859, 944) and suggest that the number of waters may vary with permeant ion species (271, 944). Electro-osmosis measurements detect the flux of water pushed through the channel by ions as they permeate driven by voltage, whereas streaming potential measurements detect the electrical consequences of ions pushed through the channel by water molecules as they are driven to permeate by an osmotic gradient. Movement of H⁺ through gramicidin did not generate a streaming potential (611), indicating that protons permeate without the need to move water through the channel, precisely as expected for a water-wire mechanism (723).

Gramicidin channels can conduct protons at a higher rate than any other ion channel conducts any other ion species, >2 × 10⁹ H⁺/s (207). This statement refers to normal ion-selective channels through which ions permeate in single file; poorly selective wide-pore channels such as the mitochondrial voltage-dependent anion channel (VDAC) that are permeable to molecules up to 1,000 Da may have higher conductances (318). The proton permeability of gramicidin channels (calculated from bi-ionic reversal potential measurements and the GHK voltage equation) is 43-55 times that of Na⁺ (723). Similarly, the proton conductance is 14-fold higher than the Na⁺ conductance when the conductances are normalized according to permeant ion concentrations (453). The proton conductance of a ethylenediamine analog of gramicidin was 19-25 times that of the next most permeant ion, NH, and 150-200 times that of Na⁺ (1085). The H⁺ conductance of dioxolane-linked gramicidin dimers was >40 times that of K⁺ (959). The explanation for the relatively high proton conductance is that protons permeate a water-filled gramicidin channel by a Grotthuss-type mechanism, hopping across the water chain without displacing the water molecules (723). One would expect that the presence of a normal cation might interfere with H⁺ conduction by preventing Grotthuss-type H⁺ hopping through the channel; however, the dwell time of Na⁺ is evidently so brief (~10 ns) that such interactions were barely detectable (421). Indirect confirmation that protons permeate gramicidin by a Grotthuss-like mechanism rather than as hydronium ions was provided by the effects of changing the dipole potential either by fluorination of Trp residues or by mutational replacement of Trp with Phe at the entrance to the pore. Both interventions affected H⁺ conductance and the conductance of other monovalent cations in opposite directions (132, 810). Similarly, agents like phloretin that change the internal dipole potential of membranes (29) reportedly also affect proton and alkali cation conductances through gramicidin channels in opposite directions (847), although other investigators see no effect (253a).

Examination of the concentration dependence of H⁺ conduction through gramicidin channels (Fig. 13) indicates several distinct regions (244). At pH >2, the unitary H⁺ current is directly proportional to [H⁺], which has been interpreted as indicating that the rate-determining step is the diffusional approach of H⁺ to the channel mouth (14, 230, 610). The implication is that each proton permeates independently of other protons and that permeation is so rapid that each proton enters the channel long after the previous one has left. Between pH 2 and 1 the slope of the [H⁺] versus H⁺ conductance relationship decreases, but at lower pH, the slope again approaches unity (292, 374). The "shoulder" region was ascribed to multiple occupancy of the channel by protons (292), as can be predicted to occur in rate theory simulations of ion permeation (445). Some subsequent studies have supported the existence of a shoulder region in native gramicidin A (14, 374), gramicidin B (374), and in the RR stereoisomer of covalently linked gramicidin dimers (207). However, no shoulder was evident in gramicidin A between pH 2.6 and 0.5 (421), and in gramicidin M, the H⁺ conductance was proportional to [H⁺] up to pH <0 (374). Finally, in dioxolane-linked SS dimers, the [H⁺] versus H⁺ conductance relationship was linear over a wide concentration range (1-2,000 mM H⁺), with slope 0.75, a finding seemingly inconsistent with diffusion being rate determining (207). H⁺ conduction through gramicidin can be simulated in a combined molecular dynamics and diffusion model by assuming single proton occupancy, but only at pH >1.7 (901). Above this concentration (i.e., at lower pH), the model conductance decreases, reminiscent of the decrease at high permeant ion concentration predicted by some Eyring-type models for multiply-occupied single-file channels (445, 874). However, the actual H⁺ current continues to increase at lower pH (14, 208, 292). It is generally accepted that normal ion channels can be occupied by multiple permeant ions, and this conclusion is supported for the bacterial KcsA channel by X-ray crystallographic evidence (268, 496). Nevertheless, in general, the entrance of a second ion would tend to be hindered by the presence of an ion in the pore, due to both electrostatic repulsion and the unfavorable orientation of waters near each ion (487, 609). The difficulty of forcing two protons into the same relatively short water wire might be even greater than for ordinary ions. The simultaneous presence of two protons in a water wire would create a Bjerrum D defect that would interrupt the HBC and block conduction, and the defect would have to be conducted through the channel (244). H⁺ conduction in a doubly occupied channel would likely be slower than in a singly occupied channel, because it would be limited by the rate of defect permeation. The proportionality between [H⁺] and current in gramicidin at pH <1 could reflect this slower defect permeability. Alternatively (244), the "shoulder region" might reflect 1) a shift in the rate-determining step from proton entry to water reorientation, 2) saturation of titratable groups on the membrane exterior that contribute to the supply of protons to the channel so that lateral conduction cannot occur at very low pH, or 3) a shift from a Grotthuss-type to a hydrodynamic conduction mechanism (molecular H₃O⁺ permeation).

The H⁺ conductances of both gramicidin (14) and two synthetic channels, LSLLLSL and LSLBLSL (254), saturate at pH <0 (Fig. 13). One possibility is that this saturation reflects the upper limit of the rate that H⁺ can permeate the channel (14). In a simple Michaelis-Menten (684) single binding site model of permeation (186, 595), occupancy of the site by protons would increase with concentration until it approached 100%, at which point no further increase could occur. However, because the bulk conductivity drops precipitously in the same extremely concentrated HCl solutions (+ in Fig. 13), we proposed that the apparent saturation reflects bulk diffusion limitation rather than an upper limit of H⁺ flux through these channels (244). In this view, water-filled channels could conduct >2 × 10⁹ H⁺/s if the proton supply were adequate.

It has been noted that single-channel current-voltage relationships can be sublinear, or saturating, at low permeant ion concentrations, whereas the current-voltage relationship is superlinear at higher permeant ion concentrations (27, 30, 453). Similar observations have been made for H⁺ permeation through gramicidin (14, 292, 810, 847). A widely held view is that sublinear behavior reflects diffusion-limited entry of ions, whereas superlinear behavior reflects permeation, and more specifically, ion exit from the pore as being rate-determining (27, 28, 30, 132, 292, 374, 810, 847, 901). The evidence presented by Andersen (27, 28) that the saturation phenomenon reflects a diffusion limited approach is substantial. However, the conventional interpretation of superlinearity is less well established. At high [HCl], many studies describe superlinearity (14, 207, 292, 374, 847). However, in phosphatidylethanolamine-phosphatidylcholine (PEPC) bilayers, the gramicidin single-channel current-voltage relationship was sublinear up to 400 mV even at 7 M HCl (208). The RR stereoisomer of covalently linked gramicidin dimers exhibits a sigmoid current-voltage characteristic, being superlinear up to 200 mV, but sublinear at higher voltage (36, 831). The interpretation of these data would become more complicated if one accepted the suggestion that proton conduction through gramicidin in fact occurs by OH flux in the opposite direction (847), an idea that is difficult to reconcile with the near proportionality between g_H and [H⁺] over ~5 pH units (Fig. 13).

If the rate-determining step in permeation is the diffusional approach of protons to the channel entrance, then permeation must be relatively rapid. To the extent that a shoulder occurs, the rate-limiting step must shift to some other process at lower pH. Akeson and Deamer (14) reported that the current ratio in H₂O/D₂O was 1.34 at 10 mM HCl, dropped to 1.20 at 1 M HCl, and returned to 1.35 at 5 M HCl, concluding that the rate-determining step at these concentrations was access, permeation through the channel, and exit, respectively. However, Chernyshev et al. (172) found that the H₂O/D₂O ratio was 1.22 at 50 mM HCl, increasing to 1.27 at 1 M HCl and to 1.36 at 5 M HCl. Furthermore, in the SS or RR covalently-linked dimers of gramicidin, the current ratios at all three [HCl] were 1.31-1.37, very near the bulk solution conductivity ratios of 1.32-1.35. They concluded that the rate-limiting step was in the channel or at the channel/solution interface at all concentrations. That the two dimers whose g_H displays very different dependence on [H⁺] in this range (207) have isotope effects indistinguishable from that for bulk solution suggests that the rate-determining process is similar to that for bulk diffusion.

Several investigators have attempted to calculate the apparent mobility of protons inside gramicidin (or other) channels. These empirical mobility calculations assume that the mean transit time is simply the inverse of the H⁺ flux rate. The pore is assumed to be a smooth cylinder with particular dimensions, and it is assumed that there are no interactions between permeating ion and the channel, reminiscent of Pomès and Roux's "greasy pore" (818). As a general rule, ion diffusion inside channels is slower than in bulk solution, which is not too surprising in light of the several steps that must occur (dehydration, interaction with the channel walls, rehydration, etc.). In fact, it is remarkable that the mobility of ions inside channels is as high as it is! The conductance of most ion channels is within an order of magnitude of the conductance expected for unrestricted diffusion through a cylinder 3 Å in diameter and only 5 Å long (444). Part of the modest reduction of cation mobility inside channels is attributable to the relative immobilization of water in the pore. In contrast, because protons permeate by hopping from one water to the next without requiring water permeation, their mobility should be impeded less by water immobilization. Obviously, the calculated mobility of protons inside gramicidin channels will be reduced for measurements at higher [H⁺] than the shoulder region. The mobility of protons inside the channel was calculated to be 28% of the value in bulk solution (14), which coincidentally is identical to one estimate of the relative mobility of protons in ice (294). However, this calculation was based on measurements at 4 M HCl, well beyond the "shoulder" region, at a concentration high enough that bulk HCl conductivity is beginning to decrease significantly (206, 207, 267, 606) (+ in Fig. 13). The proton mobility calculated by Cukierman (207) was identical inside the SS gramicidin dimer and in bulk solution at pH 4, and decreased progressively at higher concentrations. Furthermore, the mobility of protons in gramicidin A at 6 M HCl was 75% of the mobility measured in bulk solution at the same concentration (208). The diffusivity of protons in gramicidin at pH 5.3 was estimated by a different approach to be similar to bulk values (725). As discussed in section IVO, even in concentrated HCl solutions, the mobility of protons in gramicidin is still within an order of magnitude of its value in bulk solution (207) (Fig. 13). As mentioned in section IIIE, the calculated diffusion coefficient of the proton itself inside gramicidin was 40 times that in bulk water (901), but for continuous proton flux, the slower turning step must also occur. In summary, at pH >2, proton flux through gramicidin is constrained very little by the channel.

Proton transfer in gramicidin was found to be semicollective, i.e., neither concerted nor incoherent (816, 870). The proton did not cross the entire channel in a concerted manner, rather it appeared to fluctuate among a subset of two to five waters within the channel (816). FT-IR spectroscopy indicates large proton polarizability in gramicidin channels, suggestive of collective proton fluctuations (72).

The water inside the gramicidin channel is less mobile than in bulk solution (173, 271, 509, 870, but cf. Ref. 856), which is generally true for water in other confined spaces or channels (see sect. IIIE). Although its mobility may be reduced, water is still highly mobile in the pore; the very idea that gramicidin is a water-filled pore was established by elegant measurements of water flux through the channel (611).

Water molecules inside the gramicidin channel are probably oriented to some extent. Many molecular dynamics simulations indicate that the waters in a pore containing no ions are essentially entirely oriented (174, 316, 487, 640, 816, 856, 857), although some defects or breaks in the HBC may occur (173, 819, 856, 870). In contrast, Jordan (509) found that water alignment did not persist the entire length of the channel, but extended only over approximately two water molecules (509). Pomès and Roux (819) found that most channels contained a single bonding defect that was located preferentially near one end of the channel and that the remaining waters were polarized.

A proton (or other cation) will tend to orient the waters in the channel on either side of the permeating ion, although the extent to which this occurs varies among studies (316, 487, 640, 816, 819, 856, 856, 870). Calculations that include polarizability indicate that the ordering may extend only over a few water molecules (271, 272, 509, 870). If the flexibility of the channel is increased, the ordering of waters is reduced (271). In this regard, if the waters do not interact with the channel that confines them (a greasy channel), the water wire is uninterrupted (818), whereas when the waters can interact with the walls of the pore, interruptions (i.e., hydrogen bonding defects) often occur (816).

Because proton hopping through gramicidin is extremely rapid, the reorientation of water molecules (the turning step of the hop-turn mechanism) was almost universally believed to be slower, and at low pH, the rate-determining process (810, 814, 816, 818) (see sect. IIID). At pH >2, proton entry is rate determining, because it occurs infrequently. A recent study of gramicidins A, B, and M (374) and modeling of proton conduction in gramicidin (901) resulted in the surprising proposal that the rate-determining step for gramicidin A at pH <1.5 is not water reorientation, but instead is proton exit from the channel. The modeling also suggested that for gramicidin M, proton entry is rate determining at most pH values, with proton exit and water reorientation becoming rate determining only at pH <0. It will be interesting to see if more direct experiments will support these deductions.

Pomès (814) has emphasized the importance of the water coordination number for proton conduction. In bulk water, the waters ideally are four-coordinated (84, 814). Proton conduction in water is thought to reflect fluctuation between Eigen and Zundel cationic forms (see sect. IIB), in both of which water is three-coordinated. Gramicidin provides a better medium for rapid proton conduction than a hydrophobic cylinder, because its waters are three-coordinated, with hydrogen bonds with both neighbors and a third with the channel wall (814).

Cation-selective ion channels comprise narrow water-filled pores that exclude anions and accomplish cation selectivity by various techniques, including steric or electrostatic constraints, and ion-dependent compensation for removal of the waters of hydration around ions (89, 444, 761). The presence of a row of water molecules in a cation channel provides presumptive evidence that proton conduction by a water-wire mechanism is possible. Admittedly, by use of clever design, aquaporin channels manage to exclude protons, but at the same time, they also exclude other cations (see sect. IVD). To detect proton current through normal cation channels, it is usually necessary to remove other permeant ions and lower the pH to maximize the H⁺ current. If H⁺ and Na⁺, for example, were equally likely to enter and permeate Na⁺ channels, then at physiological concentrations, 3.5 × 10⁶ Na⁺ would permeate for every H⁺. In other words, 1 pA of Na⁺ current would be "contaminated" by only 1.8 protons. Even if normal ion channels, by virtue of their containing a water wire apparently begging for protons to hop through, conducted protons 100 times better than other cations, the total H⁺ flux in physiological solutions would be negligible.

Voltage-gated Na⁺ channels conduct protons at low pH in the absence of Na⁺ (60, 79, 219, 714, 715). The relative permeability (P_H/P_Na) calculated with the GHK voltage equation (368, 444, 456) is 252-274 (714, 715), reminiscent of the high P_H/P_Na of gramicidin channels (see sect. IVA). Amiloride-sensitive Na⁺ channels conduct protons, and this proton current is inhibited by amiloride (363, 364, 637, 638). Evidence has been presented that protons can permeate the Na⁺/H⁺ antiporter in gastrointestinal apical membrane vesicles from rabbits (1078). It should be emphasized that because of the low concentration of H⁺ even at pH 4 (1,000 times smaller than [Na⁺]), there still was no detectable shift of V_rev of Na⁺ or K⁺ channels in nerve studied at pH 4 (440, 1082). Thus proton permeability of normal ion channels is unlikely to play an important role under physiological conditions.

No reports of H⁺ current through K⁺ channels exist. Although it may be that no serious attempt at such a measurement has been made, K⁺ channels may just conduct H⁺ poorly. K⁺ channels appear normally to be occupied by several permeant ions (268, 445, 457, 496, 752), and complete removal of K⁺ seems to abolish channel function (19). In this context, H⁺ might be able to permeate only by sneaking through the channel in between K⁺.

Colicin and other bacterial channels have wide pores that conduct ions as large as tetraethylammonium⁺, yet surprisingly, they have anomalously high proton permeability (537). Based on reversal potential measurements, P_H/P_K > 1,000 under some conditions. This high P_H was not influenced by DEPC diethylpyrocarbonate (DEPC, a histidine modifying reagent), amantadine, or changes in buffer concentration. The mechanism responsible for this phenomenon remains obscure.

Artificial ion channels have been synthesized with simple, defined amino acid sequences. A 21-amino acid peptide containing no formal charges, (LSLLLSL)₃, forms proton-selective channels (599). A similar peptide, (LSSLLSL)₃, forms cation-selective channels that conduct protons approximately four times faster than any other cation (599). Energy minimization models of these channels suggest that (LSLLLSL)₃ assembles into trimers or tetramers that contain only a few waters, whereas (LSSLLSL)₃ assembles into hexamers or larger aggregates that contain a continuous row of waters (13c, 599). Molecular dynamics simulations indicate that the waters in (LSLLLSL)₃ are immobile, whereas those in (LSSLLSL)₃ are mobile, although less so than in bulk solution (834). Another synthetic proton-selective channel, (LSLBLSL)₃, was constructed including Aib, a conformationally constrained amino acid, to confirm that these channels are helical in their functional conformation. The tetrameric nature was established by attaching four helices together with a tetraphenylporphyrin template (13b). These synthetic channels reinforce ideas derived from gramicidin studies, that protons permeate a HBC or water wire without requiring translational movement of the water molecules, and that any cation-selective water-filled ion channel is likely to conduct H⁺ more rapidly than any other cation.

Long-chain polyamino acids (polyleucine, polyalanine) incorporated into lipid bilayers increased the proton permeability severalfold but did not form discrete channel-like openings (777). This behavior may reflect induction of a transient leak (between peptide and lipid) rather than the formation of channels.

A novel type of proton channel is formed by mutagenesis of Shaker voltage-gated potassium channels. Insertion of a His residue at position 371 (R371H) results in proton channel behavior (960). This channel is gated because His³⁷¹ is part of the channel's voltage sensor, and it happens that when the channel opens, His³⁷¹ becomes simultaneously accessible to solutions on both sides of the membrane. Other mutations in which a His is able to access only one side of the membrane at a time result in proton carrier function (see sect. IIIC).

Although water permeates cell membranes fairly readily, some cells have specialized channels called aquaporins that conduct water rapidly, 3.9 × 10⁹ H₂O/s (13). Permeation of water through aquaporin requires breaking hydrogen bonds, and the activation energy for water permeation, 3.1 kcal/mol (1096), is near that of hydrogen bond strength in water, 2.6 kcal/mol (1056). Because of the facility with which protons are conducted both in bulk water and in water-filled ion channels like gramicidin, one might expect that aquaporins would conduct protons as well (367, 413). However, no proton permeability is observed (1096). How does this water-filled pore exclude protons? The atomic structure of aquaporin-1 has been determined and the relevant structures identified (718). The channel lining is generally hydrophobic, but at a central constriction the amido groups of two Asn residues project toward the center of the pore. These positively charged groups form hydrogen bonds with one or two water molecules in the pore. During permeation, each water molecule is briefly hydrogen-bonded to both amido groups. The water is then conformationally constrained and prevented from forming the hydrogen bonds with adjacent waters that are necessary for Grotthuss-type proton conduction (718). In addition, the net positive charge in this region would tend to exclude all cations, including protons (12). The nonpolar walls of the pore also disfavor permeation of any ions including H⁺ (839). More recent molecular dynamics simulations of the Escherichia coli water channel, GlpF, reveal the same crucial two Asn residues at the central constriction but provide a different mechanism (989). The central H₂O molecule forms hydrogen bonds simultaneously with waters on either side, acting exclusively as a hydrogen bond donor. The waters in both half-channels are thus oriented oppositely, with their hydrogens pointing toward the channel mouths.

A membrane protein of the influenza A virus, M₂ plays a critical role in the infection process. The M₂ channel is believed to allow proton influx into the virion, which enables viral uncoating, i.e., the unpacking of the viral ribonucleoproteins that is necessary for them to enter the host cell's nucleus (661). A second proposed function is to dissipate the low pH of the trans-Golgi network, thereby preserving the hemagglutinin molecules from acid inactivation and preventing them from entering the noninfective acidic conformation prematurely (422). Inhibition of M₂ channels with amantadine or rimantadine inhibits viral replication (220, 416). Mutations of the M₂ channel that prevent its proton channel function resulted in an influenza A virus that could replicate through multiple cycles in cell culture, but not in mice (1065). However, viral replication in the M₂ mutant that lacked channel activity was considerably slower (1065), comparable to that of wild-type M₂ after inhibition by amantadine, and the mutant viruses could not compete with wild-type viruses in cell culture (992). Thus, although M₂ channel function is not absolutely essential for viral survival in cell culture, it promotes efficient viral replication in an infected organism.

The M₂ protein expressed in Xenopus oocytes (812), mammalian CV-1 cells (1058), or lipid bilayers (1016) functions as an ion channel when it is activated by acidic pH. Because the M₂ channel is not voltage gated, it can be studied only by inhibiting the currents with amantadine and subtracting to obtain the amantadine-sensitive current. This technical complication made it difficult to establish precisely the selectivity of the M₂ channel. Early studies reported distinct permeability to Na⁺ and a variety of cations (812, 923, 1016, 1058). Proton permeability of a 25-amino acid peptide comprising the membrane-spanning region of M₂ was demonstrated at pH 2.3 after its incorporation into bilayers (273). Indirect evidence for proton permeability of intact M₂ channels was provided by measurements of pH changes in lipid vesicles (898) and in Xenopus oocytes (923). Finally, direct measurements of H⁺ currents through M₂ channels expressed in murine erythroleukemia cells were reported, in which the reversal potentials for the rimantadine-sensitive current indicated nearly perfect proton selectivity (175). It has been demonstrated that much of the deviation of V_rev from E_H in Xenopus oocytes was the result of local pH changes resulting from proton flux through M₂ channels and that under physiological conditions M₂ is indeed highly H⁺ selective (711). Measurement of ion fluxes into vesicles supports the idea that M₂ channels are essentially perfectly proton selective (617).

The M₂ channel is a homotetramer formed by 96-amino acid monomers that span the membrane once. The M₂ mRNA codes for a 97-amino acid polypeptide, but the initiation Met is cleaved in the mature protein (1013). In its membrane-spanning region there are mainly hydrophobic residues, with the exception of Ser³¹ and His³⁷. Amantadine sterically occludes the channel near the center of the transmembrane domain, between Val²⁷ and Ser³¹ (274), or perhaps by binding to His³⁷ (873). Substantial evidence supports an important role for His³⁷ in the activation of M₂ channels by low pH. Deletion of His³⁷ or mutation to Ala, Gly, or Glu abolishes activation by low pH (812, 1059). Two types of roles for His³⁷ have been proposed: 1) as a proton shuttle and selectivity filter or 2) as a gating regulator, the protonation of which results in a conformational change (i.e., channel opening). Pinto et al. (811) proposed a structure for M₂ in which there is a channel large enough to contain a continuous water wire with a single occlusion at the point where the four His³⁷ protrude into the lumen of the pore (Fig. 6, yellow). This proposal was supported by cysteine scanning mutagenesis that showed extracellular accessibility of Ala³⁰ and Gly³⁴, and intracellular accessibility of Trp⁴¹ (925). The occlusion formed by His³⁷ could prevent cation conduction and would produce H⁺ selectivity by allowing H⁺ conduction by successive protonation of a His imidazole nitrogen, deprotonation of another nitrogen, and then a ring flip (tautomerization) to complete the cycle (811). A virtually identical His tautomerization mechanism had been proposed to occur at the His⁶⁴ at the entrance to the proton channel in carbonic anhydrase II (735, 932). In this model, His³⁷ actively and obligatorily participates in H⁺ conduction. This shuttle model was supported by molecular dynamics simulations in which only singly or doubly protonated channels were stable (906).

View larger version (82K): [in this window] [in a new window]   Fig. 6. Side view of the proposed transmembrane section of the M2 viral proton channel. Every fourth amino acid residue (in colors) is expected to face the pore. Starting at the extracellular end (top), the residues identified by cysteine-scanning mutagenesis to be accessible to the external solution are Ala30 and Gly34. The crucial His37 (yellow) is believed to occlude the pore to cations other than protons. The internally accessible Trp41 (red) may play a role in gating the channel (774). [From Shuck et al. (925), copyright 2000 American Society for Microbiology.]

The gating model retains the concept of His³⁷ as a regulator of channel activity (878). Molecular dynamics simulations (878) as well as site-directed dichroism studies (573) indicate that when all four His³⁷ in the M₂ channel are deprotonated they protrude into the pore, occluding it. When these residues are fully protonated, they retract due at least in part to electrostatic repulsion, which permits a continuous water wire to span the entire pore (878). This mechanism might require only two of the four His to be protonated at low pH to activate the channel (312). In support of the shuttle mechanism, the conductance of M₂ was reduced by 40-50% in deuterium, suggesting that H⁺ does not permeate as H₃O⁺ (712), because the isotope effect for proton/deuteron conduction in bulk water is weaker than this (see sect. VI) (65, 325, 612, 624, 844). However, this conclusion is compromised by the possibility that water inside the M₂ channel may be to some extent effectively "frozen" (i.e., less mobile) as has been suggested (317) (see also sect. IIIE). In this event, a larger isotope effect might be predicted by analogy with the larger deuterium isotope effect reported for conduction in ice (289, 575). In support of the "gating" model, Raman spectroscopy indicates that His³⁷ is indeed protonated when the channel is activated at low pH, and furthermore that it appears to interact with the indole ring of Trp⁴¹ (774). On this basis, a model similar to that of Sansom et al. (878) was proposed, in which low pH activation occurs when protonation of the four His³⁷ residues results in their retraction (774). A novel feature is that the protonated His³⁷ imidazolium in its retracted position is closer to Trp⁴¹, and this conformation is stabilized by interaction of the cationic imidazolium with  electrons of the indole ring of Trp⁴¹ (774). The importance of Trp⁴¹ in M₂ gating was demonstrated by alteration of the pH dependence of gating in W41A, W41C, W41F, and W41Y mutations (995).

The world record for the smallest unitary conductance of any ion channel was claimed recently for the viral M₂ proton channel (617). At room temperature and pH 7.4, the single-channel current was estimated to be 1.2 aA (2.7-4.1 aA at pH 5.7; , Fig. 13), corresponding with a conductance of 8-44 aS (aS = attoSiemens = 10¹⁸ Siemens). By calculating the net H⁺ flux, we find that one M₂ channel transports 12 yM (1 yoctomole = 10²⁴ mol; Ref. 1001) of protons per second at pH 7.4, or 7.5 H⁺/s. It should be noted that this conductance estimate was indirect and assumes that every M₂ molecule is functional, and therefore represents a lower limit. Whether this miniscule unitary current is truly a record may depend on comparison with a mammalian serotonin receptor, which can act as a proton channel at low pH (145, 146). The flux through a single transporter is 300 H⁺/s at pH 3.5, but only 20 H⁺/s at pH 5.5 (146) (, Fig. 13), thus when extrapolated to physiological pH the proton flux is even smaller than through M₂. Estimates of the M₂ unitary current based on other considerations are much larger, 0.5 fA at pH 6.2 to an upper limit of 10 fA (712). These latter values fall precisely in the range of estimates for voltage-gated proton channels (85, 136, 168, 236, 244).

An intriguing structural parallel with voltage-gated proton channels is provided by evidence that M₂ channels bind a heavy metal, Cu²⁺, and that this binding occurs at His³⁷ (340). Because His³⁷ is believed to regulate M₂ channel function depending on its degree of protonation, this is very similar to the suggestion that Zn²⁺ binds to voltage-gated proton channels at an externally accessible site formed by His residues (163), the protonation of which regulates the gating of these channels (166).

ATP is generated in plant and animal cells by H⁺-ATPases that utilize energy stored in the form of a proton gradient (694, 698). Some of these enzymes can reverse direction and function as proton pumps. H⁺-ATPases are divided into F (chloroplast, mitochondria, bacteria), V (vacuolar, in organelles), and P (plants, fungi, bacteria) types, with F and V sharing similar sequences and mechanisms (596). The intricacy and complexity of the operation of this enzyme is astonishing. The F_oF₁ ATP synthase consists of two main parts: a subunit that is embedded in the membrane and contains proton channels, F_o, and the ATP binding subunit, F₁, which is attached via a narrow stalk (Fig. 7).³ Recent studies in which large fluorescent molecules have been attached to the -subunit at the interface between the F_o and F₁ components have demonstrated in spectacular fashion that the oligomeric ring of F_o proton channel molecules spins around during ATP hydrolysis (760, 1089), confirming earlier proposals (4, 113, 275, 384, 697, 866). The whole complex is anchored in the membrane by subunit b, which acts as a stator. Rotation of subunit c relative to subunit b has been confirmed biochemically (503) and is required for proton translocation (978). This rotary engine motif is shared by the Na⁺-driven ATP synthase of Propionigenium modestum (265, 513) and the flagellar motor (86, 513, 681). The key amino acid in the F_o proton channel is Asp⁶¹ (or Glu⁶⁵ in Na⁺-driven F-ATPases, Ref. 264), one of which is present in each of the 9-14 c subunits (20, 495, 958). When the carboxyl group of Asp⁶¹ is protonated it rotates, and rotation in the correct direction is enforced by electrostatic interactions with Arg²¹⁰ on the stator, which lower the pK_a of Asp⁶¹ if it rotates incorrectly, with the result that the group becomes deprotonated. The proton gradient across the membrane thus ensures correct rotation by a probabilistic process (293). ATP synthesis is accomplished when protons at the side with high local [H⁺] bind to a low pK_a site (Asp⁶¹), are translocated, and then released to the side with low [H⁺]. Release occurs simultaneously with a conformational change that promotes ATP synthesis and also increases the proton affinity of the site (307). When the H⁺-ATPase reverses direction to act as a pump, Asp⁶¹ changes from high to low pK_a conformation as protons are pumped uphill (307). It appears that V-ATPases function generally like F-ATPases, but with certain differences (375). The V_o rotor has 6 protonation sites compared with 9-14 for the F_o rotor, and V-ATPases can pump against a larger pH but at a lower maximum rate than F-ATPases (375). Only F-ATPases are reversible under in vivo conditions; V-ATPases function exclusively as H⁺ pumps (596).

View larger version (83K): [in this window] [in a new window]   Fig. 7. The components of ATP synthase, a rotary motor. Protons enter the 10 proton channels in the membrane-bound Fo component, where they bind to Asp61, and are subsequently released to the opposite side of the membrane via an outlet channel after Fo rotation (arrow). The -subunits are attached to the Fo ring and also rotate relative to the F1 component, which catalyzes ATP synthesis (or breakdown if the reaction is driven backwards), which is thus driven by the proton-motive force. [From Jiang et al. (495), copyright 2001 National Academy of Sciences, USA.]

Recent cysteine scanning studies have shown that N-ethylmaleimide can access several residues well inside the F_o proton channel and that even more of them are accessible to Ag⁺, which is close to H₃O⁺ in size (306). These data suggest that the proton pathway through the access channel to Asp⁶¹ is a water wire. They further indicate that this part of the proton channel is not strictly H⁺ selective (Table 1).

Dicyclohexylcarbodiimide (DCCD), a classical inhibitor of H⁺-ATPase, inhibits by binding to Asp⁶¹ in the F_o component at alkaline pH, blocking proton conduction through F_o (148, 308, 392, 449, 464, 512, 1092). At low pH, DCCD also binds covalently to Glu¹⁹⁹ in the F₁ sector (299, 359, 1091). Despite this evidence of promiscuity, DCCD does not inhibit voltage-gated proton channels (Table 6). V-type H⁺-ATPases are inhibited by DCCD, but in addition are inhibited potently by bafilomycin A₁ (110, 269), which binds to the c subunit and blocks proton flux (193, 667).

Torque generation by the Na⁺-driven ATP synthase of Propionigenium modestum requires an electrical potential of ±90 mV, which cannot be substituted by a Na⁺ gradient (515). This voltage-dependent switch from idling mode to torque generation may conceptually resemble the voltage-gating mechanism of voltage-gated proton channels. However, a required voltage-sensitive step in the reaction cycle could also account for these data.

Many attempts have been made to demonstrate proton conduction through F_o after removing the F₁ component. Although some data to the contrary have been presented (129, 702, 1098), most studies conclude that F_o in isolation from F₁ (or V_o in the case of vacuolar proton pumps, or CF_o in the case of chloroplast proton pumps) can act as a passive proton conductor (21, 143, 150, 192, 193, 323, 615, 749, 891, 892, 928, 1097). One group found that native V_o did not conduct protons (1098), but that reconstituted V_o did (1097). Another found that F_o by itself conducted protons at a rate too low for kinetic competence during ATP synthesis, but that coexpression with F₁ increased H⁺ flux (795). Yoshida et al. (1092) found that F_o proton conduction was prevented by binding of the -subunit and suggested that this might comprise a gate (1092). Schindler and Nelson (887) reported that proteolipid derived from mitochondrial H⁺-ATPase incorporated into lipid bilayers formed H⁺-selective channels detectable at pH 2.2 (887). More recent attempts to record H⁺ currents electrically through CF_o (1053) have been complicated by evidence that either contaminating protein or the F_o component itself is capable of functioning as a nonselective cation-permeable channel (16, 143) that in some cases is inhibited by the standard F_o inhibitors venturicidin and DCCD (669, 893). Indirect measurements of the unitary conductance of CF_o at first provided estimates of 10 fS (892). Shortly thereafter, it was concluded that because only a small fraction of the CF_o was active, the actual conductance was much higher, 169 fS (616) or even 1 pS (21, 615). Surprisingly, the conductance is independent of pH between pH 5.6 and 8.0 (21, 512), suggesting that proton conduction through the channel, rather than the proton supply, is rate determining. The pH independence of the conductance at high pH appears to contradict the proportionality between [H⁺] and conductance at lower pH reported by Schindler and Nelson (887). The very large estimated unitary H⁺ conductance is indeed enigmatic (21, 512) and far exceeds the conductance of any other proton-conducting channel in this pH range with the exception of carbonic anhydrase II (see sect. IVK) (Fig. 13). An array of mechanisms that might increase the supply of protons to a channel with such prodigious H⁺ conductance in the face of such tiny H⁺ concentrations is discussed in section IVO. A recent estimate of the conductance of F_o reconstituted into liposomes is much lower, 0.1-0.2 fS, which amounts to a unitary flux of ~70 H⁺/s (143). In several respects, CF_o behavior strongly resembles that of voltage-gated proton channels (Table 1): the channel is extremely selective for H⁺ (P_H/P_K > 10⁷), the H⁺ current is 1.7-fold greater than D⁺ current, and the Q₁₀ for the conductance is 1.9 (21, 512).

Bacteria swim by means of rapidly rotating flagella. Flagellar rotation is driven by a motor that is powered by the proton gradient, or protonmotive force (or in some cases by a Na⁺ gradient). The membrane-bound part of the flagellar motor (the stator, which anchors the whole complex to the cell wall) consists of two molecular components, MotA and MotB. Proton conduction does not occur when MotA is expressed alone, in the absence of the rest of the flagellar motor components (970). H⁺ flux through a proton channel formed by MotA and MotB drives the rotation of the flagellum (97, 98, 344, 970). Because too few of the amino acids that form the proton channel have protonatable side chains to form a HBC that would span the membrane, it has been proposed that water molecules comprise most of the proton pathway (98, 920). Correct operation of this system requires an aspartyl residue in MotB (Asp³²) that is believed to comprise a proton binding site; surprisingly, no other individual conserved acidic residue was required for torque generation (1100). When Asp³² was replaced with 15 other amino acids, only D32E mutants (conservative replacement of Asp by Glu) exhibited any function (1100).

There are striking similarities between the flagellar motor proton channel (in situ in low-torque mode) and voltage-gated proton channels. Both have strong temperature dependence, large deuterium isotope effects, and weak dependence on pH (159, 160, 682). The flagellar motor speed is directly proportional to applied voltage (proton-motive force), demonstrating tight coupling between proton flux and motor rotation (328).

Bacteriorhodopsin is a light-driven proton pump in the purple membrane of Halobacterium halobium that creates a proton gradient used for the manufacture of ATP (101, 596). Like F_oF₁-ATPases (695, 698), bacteriorhodopsin contains a "proton channel" that is essential to its function (556, 969). This "channel" probably at no time forms a continuous pathway across the membrane (1009); rather, the proton moves through two partial channels separated by a Schiff base (Fig. 8). The proton pathway in bacteriorhodopsin is one of the first explicitly proposed to comprise HBC elements (969). The entire photocycle can be dissected into at least eight intermediate states. Energy is introduced by absorption of a photon, and as the cycle proceeds, the result is the translocation of one proton from the cytoplasm to the extracellular solution. Proton movement occurs in three main steps (588) that are "backwards" in a sense. First, light-induced isomerization of retinal in the Schiff base alters the local geometry, which results in transfer of a proton from the Schiff base to Asp⁸⁵, causing the exit of a proton through the extracellular-facing "release" channel (587, 589). Next, the Schiff base is reprotonated from the intracellular-facing (cytoplasmic) channel, receiving the proton from Asp⁹⁶. A proton then enters the cytoplasmic "uptake" channel to reprotonate Asp⁹⁶. Finally, the released proton is replaced as Asp⁸⁵ deprotonates. A minimal model requires five separate proton transfer events (436). The "reprotonation switch" serves to ensure directionality to proton transport through bacteriorhodopsin by preventing reprotonation from the exit side (586).

View larger version (64K): [in this window] [in a new window]   Fig. 8. Proposed proton pathways through bacteriorhodopsin. The proton moves from the cytoplasmic side at the top to the extracellular side at the bottom. The numbered steps indicate the chronology: 1) a proton is transferred from the Schiff base to Asp85, 2) the proton is released to the extracellular solution, 3) a proton is transferred from Asp96 to the Schiff base, 4) Asp96 is reprotonated from the cytoplasmic solution, and 5) transfer of a proton from Asp85 to a release site, perhaps comprising waters held in position by the two Glu residues near the extracellular surface. [From Luecke et al. (632), copyright 1999 American Association for the Advancement of Science.]

The proton pathway (Table 2) includes several water molecules in addition to titratable amino acid residues. The X-ray structure of bacteriorhodopsin crystals indicates eight or nine water molecules in the putative proton pathway (633, 801). The cytoplasmic pathway is hydrophobic but must acquire enough waters during the photocycle (2-4) to bridge the 10- to 12-Å gap between Asp⁹⁶ and the Schiff base (436, 633, 858, 1074). The isotope effect on proton transfer through this channel is weak, 1.3-1.7 (601), similar to that for ordinary hydrogen bond cleavage in water, 1.4 (1056), and for H⁺ conductivity in water, 1.4-1.5 (65, 325, 612, 624, 844). This suggests that proton movement through this channel is similar to that in bulk water (601, 1074), in which the rate-determining step is hydrogen bond cleavage (8). The extracellular pathway contains seven bound waters in an extensive three-dimensional hydrogen-bonded network with amino acid residues and the retinal Schiff base (633). The very large isotope effect on proton release (Table 1) may reflect a conformational change in the protein (601). An alternative interpretation that is supported by a highly curved proton inventory plot (127) is that it reflects multiple proton transfers or extensive hydrogen bond rearrangement within the exit channel. The proton release pathway may include several titratable amino acid residues (Table 2).

Any step in the photocycle that involves net charge moving across the membrane potential field should be influenced by membrane potential. The rate of proton pumping by bacteriorhodopsin has been shown to depend linearly on membrane potential (347, 650, 717, 727, 728). However, this does not mean that bacteriorhodopsin is obligatorily voltage gated. The influence of membrane potential on charge movement across the membrane is an unavoidable electrostatic property that is distinct from voltage-dependent gating. Nevertheless, to achieve proton pumping, proton translocation must be vectorial (unidirectional). A conformational change resulting from deprotonation of the Schiff base toward the extracellular side of the membrane, which may switch the accessibility of the active site of bacteriorhodopsin to the cytoplasmic side of the membrane, has been proposed on the basis of electron crystallography (971). The configuration of the retinal, from X-ray crystallography of an early M state (590), indicated a more local change for the switch, which reorients the direction of the N-H bond from the extracellular to the cytoplasmic direction after deprotonation of the Schiff base. The reprotonation switch mechanism closely resembles the proposed mechanism of gating of voltage-gated proton channels (Fig. 20), except that the result of the latter is believed to be an uninterrupted H⁺ conduction path across the membrane.

In photosynthetic bacteria (e.g., Rhodobacter sphaeroides), light induces a series of electron transfer events in the photosynthetic reaction center protein that are coupled to proton transport across the membrane. The catalysis involves the light-driven reduction of quinone to quinol, which involves the uptake of two protons by the reaction center protein (Q + 2e + 2hv + 2H⁺  QH₂). Normally electron transfer is rate limiting; proton transfer can be made to be rate limiting by Zn²⁺ or Cd²⁺ binding (789, 790) or by mutations that greatly slow proton transfer (788). One of several proposed proton pathways is illustrated in Figure 9 (43, 775, 788, 790). Two parallel pathways are illustrated which have the following elements



where BH is protonated buffer and the amino acid residues are labeled as described in Figure 9. The inner end of both pathways may be accessible to the external solution, because mutations of Glu^L212 and Asp^L213 inhibit proton transfer, and function is restored by "chemical rescue" with several small weak acids (990). For this reason, the bacterial reaction center (BRC) proton channel is listed as having low proton selectivity in Table 1. Part of the evidence that protons enter via this channel is based on the inhibition of proton uptake by divalent metal ions. Zn²⁺ and Cd²⁺ inhibition occurs when the metal ion is coordinated at the surface of the BRC by two His and one Asp (Asp¹²⁴) (43). The similarity of this proposed binding site to the external metal receptor (three His residues) proposed for the voltage-gated proton channel (163) is noteworthy. There is strong competition between metal cations and protons in both molecules. As a result of this competition, the apparent metal affinity of the metal binding sites on both BRC and voltage-gated proton channels decreases upon protonation (163, 355, 869). The pH effect observed for the BRC is evidently weaker: lowering pH from 8 to 5 reduced the affinity of Ni²⁺ for BRC by a factor 10², and for Cd²⁺ by a smaller amount (355), whereas over the same pH range the apparent affinity of Zn²⁺ for the voltage-gated proton channel decreased by 10³-10⁵ (163). For voltage-gated proton channels, the strong competition between H⁺ and Zn²⁺ required assuming that the Zn²⁺ receptor comprises two or three protonation sites (163). The BRC exhibits simple 1:1 competition (869) in spite of the presence of several protonatable groups near the putative metal receptor.

View larger version (35K): [in this window] [in a new window]   Fig. 9. Proposed proton pathways in the bacterial reaction center. The reaction center consists of three polypeptides (L, M, and H subunits) and nine cofactors. Catalysis involves the light-driven reduction of quinone to quinol, which involves the uptake of two protons by the reaction center (Q + 2e + 2hv + 2H+ QH2). Proton entry occurs from a buffer molecule in solution (BH) to a His, either His-H126 and/or His-H128; the subunit and position of the residue is indicated (e.g., His-H128 means His at position 128 of the H subunit). The evidence for the involvement of these His residues is relatively recent (47). Proton transfer proceeds via bridging water molecules (red spheres) to an Asp at either L210 or M17, shown in bold (789). These two Asp provide parallel branches, because the presence of either one is sufficient for function. They also act cooperatively in that the rate of proton transfer is much greater (>10-fold) with both present than with just one present. Proton transfer then proceeds via another water molecule to Asp-L213, which provides a branch point. Here the pathways for the two protons involved in the formation of QH2 diverge, with one proton proceeding to a carbonyl oxygen of the quinone via Ser-L223 [H+(1)] and the other proceeding to Glu-L212 via at least two water molecules, and then onto the other carbonyl oxygen of the quinone [H+(2)]. Direct transfer from Glu-L212 to QB is not likely, because they are separated by ~5 Å. Thus the transfer from Glu-L212 to the quinone may involve QB movement or a bridging water molecule that is not seen in the structure. [From Paddock et al. (788), copyright 2001 American Chemical Society.]

As for voltage-gated proton channels, proton uptake by the BRC appears not to be limited by diffusion of protons or protonated buffer, because it is weakly dependent on pH, has high temperature sensitivity, weak dependence on viscosity, and a large deuterium isotope effect (657). Maróti and Wraight (657) proposed that a conformational change or hydrogen bond rearrangement was required to enable proton transfer into the reaction center.

The proposed role of the two superficial His (His¹²⁶ and His¹²⁸) as proton donors was confirmed by mutating them to Ala. Surprisingly, replacing either one alone had no effect, but in the double mutant, the two proton-limited rate constants were reduced eight- and fourfold (10). The actual slowing of proton transfer is greater than this but cannot be determined directly because the rate constants include both proton transfer and the coupled electron transfer. In the native BRC electron transfer is rate limiting. This also means that the single mutants might have reduced proton transfer but that any such reduction is not evident because electron transfer is still rate determining. Function was completely restored by chemical rescue with 50 mM imidazole, which has no effect on the native BRC (10). The fact that single mutants retained full function indicates that the two His are redundant; only one is necessary for normal enzyme function. Proton uptake can be restored in the double mutant by chemical rescue with acids with a wide range of pK_a, to generate a Brönsted plot, which revealed that the rate constant for proton transfer across this 20-Å pathway is 10⁵ s¹ (787). The presence of protonatable groups at the entrance to the channel speeds proton transport by at least two orders of magnitude (787).

Cytochrome c oxidase is the final enzyme in the respiratory chain in the inner mitochondrial membrane, responsible for pumping protons across the membrane from the matrix into the intermembrane space, thus creating the proton gradient used to generate ATP (51, 52, 350, 352, 1072, 1075). Widely studied homologous heme-copper oxidases are found in the cell membranes of aerobic bacteria, such as Escherichia coli, Paracoccus denitrificans, and Rhodobacter sphaeroides. In each catalytic cycle, four protons are pumped across the membrane, and four additional protons combine with four electrons and O₂ to form water as a by-product. In some models, the translocated ("pumped" or "vectorial") protons and the substrate ("consumed," "scalar," or "chemical") protons were proposed to travel through different proton channels (483), D and K, respectively (see below). Proton and electron movements are tightly coupled, although the details remain controversial (517, 685, 686, 708, 842, 1041, 1075, 1076). Based on the crystal structure, three proton pathways were proposed, each comprising a number of amino acid side groups and several water molecules in an extensively hydrogen-bonded arrangement (483, 1018, 1093). Of three proposed proton channels, H, D, and K, the latter two have been shown by mutagenesis to be functionally important (305, 341, 352, 602, 689, 1004, 1040, 1074). The D and K channels are named for crucial Asp and Lys residues (D¹³² and K³⁶² in Figs. 10 and 11, respectively) that upon mutation prevent proton conduction (558). These two channels appear to function during distinct parts of the catalytic cycle (352, 558, 689, 862, 1051, 1077). In addition, there is another, more recently investigated proton channel, the "exit" channel, that allows pumped protons to reach the external medium (689, 828). Intriguingly, after block of the D channel, but not the K channel, the oxidase still can transfer electrons slowly, but cannot pump protons (305, 691, 1004). To support this activity, protons may be taken up through the exit channel, resulting in a futile cycle (689).

Figure 10 shows one of the proton uptake pathways, the D channel. As was the case in the bacterial photosynthetic reaction center (see sect. IVI), the proton pathway comprises a combination of amino acid side groups and water molecules (red dots). Calculations based on the known structure of cytochrome c oxidase predict that ~130 water molecules reside within subunits I and II of the enzyme (458). The proton pathway may consist mainly of water molecules that are stabilized within the channel by polar residues that line the channel, such as Ser and Asn (458, 843). Mutation of some of these polar residues reduces the proton-pumping ability of the enzyme only slightly (693), whereas mutation of others prevents proton translocation (341, 1004). The mutations D132N and E286Q abolish function (689, 1040, 1051). At the entrance to the D channel there is a cluster of negatively charged amino acids and six His, all of which combine to form an effective proton-collecting antenna (523, 653). The proton transport process in the D channel is reminiscent of that in bacteriorhodopsin (1074). First, a proton is transferred from Glu²⁸⁶ to the binuclear center or to the outside, and then Glu²⁸⁶ is reprotonated from bulk solution (949). The proton transfer via Glu²⁸⁶ is rate determining for one of the electron transport steps and displays a kinetic isotope effect of 7 for the R. sphaeroides enzyme (525). Thus not only the electron transport, but the proton displacement across 30 Å is rate limited by this single proton transfer step.

View larger version (20K): [in this window] [in a new window]   Fig. 10. One of the proton pathways inside cytochrome c oxidase, the D channel of Rhodobacter sphaeroides. This pathway is used for the transfer of both pumped and substrate protons during the reaction of the reduced enzyme with O2. Asp132 is at the surface and functions as the entrance to the channel from the inside (N side). The red dots illustrate waters that are present in the crystal structure and probably participate in proton conduction. The protons bind to E286 and then are somehow transferred to the binuclear center (top right, heme a3/CuB). Subsequently, the pumped protons exit through a different channel to the external solution (P side) (690). [From Svensson-Ek et al. (979).]

The Glu residue (Glu²⁸⁶ in E. coli and R. sphaeroides; Glu²⁷⁸ in P. denitrificans) deep in the D channel near the heme-copper (Cu_B) bimetallic center is highly conserved and is crucial to proton translocation and O₂ reduction (59, 1040). The Glu side chain may have to move to mediate proton translocation (458, 815, 843). Mutation of Glu to anything but Asp abolishes proton translocation and grossly impairs enzyme function, but if a Tyr is inserted nearby, function and proton translocation are restored (59).

Like voltage-gated proton channels (and many other proton pathways), the D channel is inhibited by Zn²⁺, and less potently by Cd²⁺ or Cu²⁺ (1). Although the Zn²⁺ affinity is high, K_i ~2.6 µM, inhibition of decay of the peroxy intermediate, which involves proton uptake, saturates at a 56% lower rate of O₂ catalysis by the enzyme (1). However, proton uptake per se into the D channel is slowed >20-fold (2). The D132N mutation abolishes Zn²⁺ effects in the purified enzyme, suggesting that Zn²⁺ binds and exerts its effects near this Asp residue (2).

Figure 11 shows a second important pathway in cytochrome c oxidase, the K pathway. This pathway comprises mainly protonatable amino acids with only a few waters interspersed. Although it was considered possible that the K pathway subserves OH release rather than H⁺ uptake (855), the latter role is supported by mutagenesis studies (862, 929). An alternative proposal is that the K channel is not a proton channel at all, but instead functions as a "dielectric well" in which electron movement within the binuclear center is compensated by movement of charges within the K channel (511).

View larger version (30K): [in this window] [in a new window]   Fig. 11. The K channel is another proton pathway in cytochrome c oxidase of Rhodobacter sphaeroides. The entrance was first thought to be near Ser299 (115), but recent evidence supports Glu101 as a more probable entrance (116). Red dots indicate waters. [From Brändén et al. (116), copyright 2002 American Chemical Society.]

Based on the principle that an ion that interrupts a water wire should inhibit proton conductivity (421), Kornblatt (564) proposed that the inhibition of cytochrome c oxidase by formamide and formaldehyde reflects this mechanism.

The exit pathway in cytochrome c oxidase exhibits several intriguing parallels with voltage-gated proton channels. Both are potently inhibited by Zn²⁺ (163, 642, 692, 1008). The only other polyvalent cation with comparable effects was Cd²⁺, another classical voltage-gated proton channel inhibitor (134, 1008). Another similarity is that Zn²⁺ is strongly competitive with protons in inhibition of the exit channel, with the Zn²⁺ efficacy 10-fold weaker at pH 6 than pH 7; technical limitations prevented extending the measurements to lower pH (692) at which in voltage-gated proton channels protons compete with Zn²⁺ at >1:1 stoichiometry (163). The apparent pK_a for the cytochrome c oxidase exit channel was 6.8 (692), identical to the pK_a of 6.2-7.0 for the Zn²⁺ receptor of voltage-gated proton channels (163). Based on this apparent pK_a, the Zn²⁺ receptor in voltage-gated proton channels was speculated to comprise three His (163). Surprisingly, mutation of individual titratable amino acids near the cytochrome c oxidase exit channel did not prevent inhibition by Zn²⁺ (692), perhaps indicating redundancy or that multiple groups contribute to the Zn²⁺ receptor. Nevertheless, potent, selective inhibition by Zn²⁺ strongly supports the idea that electron-coupled proton backflow occurs through a specific channel rather than through the phospholipid bilayer. Zn²⁺ inhibition was lost in the presence of valinomycin or uncouplers, suggesting that proton flux through the exit channel is required to sustain electron transfer (692). The loss of Zn²⁺ binding when the membrane potential was removed could indicate that the exit channel is gated by membrane potential, allowing Zn²⁺ entry only when the channel is opened by hyperpolarization (692). In that case, the cytochrome c oxidase exit channel might belong to the rarified class of proton channels gated by voltage (Table 1). Alternatively, the loss of Zn²⁺ binding from the external solution when the normally large negative membrane potential is uncoupled could indicate that the Zn²⁺ binding site is within the membrane electrical field and that block is voltage dependent.

Carbonic anhydrase (CA), which comes in seven isoforms in mammals, catalyzes the following hydration/dehydration reaction: CO₂ + H₂O  HCO + H⁺ (620). The maximal turnover rate for human CA II is ~10⁶/s in either direction at 25°C, making it one of the fastest enzymes known (533, 621). The rapidity of this reaction can be appreciated when one considers that the conversion of CO₂ to HCO almost reaches equilibrium during the passage of blood through systemic capillaries, and the reverse reaction (including diffusion and other in vivo complications) in alveolar capillaries equilibrates within ~300 ms (1052). The catalytic center (active site) of the molecule (Fig. 12) is formed by a zinc atom coordinated to three His and to one solvent ligand and is located in a deep funnel-shaped pocket within the molecule (298). During hydration of CO₂, a zinc-bound hydroxide reacts with CO₂ and releases HCO, which is replaced at the zinc by a water molecule. In a separate stage of catalysis, H⁺ is transferred from zinc-bound water out to His⁶⁴ along a chain of two or more water molecules, and then H⁺ is transferred from His⁶⁴ to buffer in solution (1023). The position of His⁶⁴ differs substantially in crystals formed at low or high pH (735), suggesting that perhaps the imidazole ring of His⁶⁴ shuttles the proton by flipping its conformation into or out of the active site cavity, although this has not been shown definitively. In support of this idea, catalysis by the slower isozyme, CA V, is not enhanced by introducing His at position 64 (Y64H) unless the F65A mutation is simultaneously introduced; the bulky side chain of Phe⁶⁵ evidently prevents the requisite motion of His⁶⁴ (419) or disrupts water structure in the active site cavity (485).

View larger version (65K): [in this window] [in a new window]   Fig. 12. The proton pathway inside human carbonic anhydrase II. The active site is formed by a zinc atom coordinated between three His residues (blue). A chain of three (or more) waters (red spheres) relays a proton from the zinc-bound hydroxyl to His64. At high pH, the His64 is in the "in" conformation, but upon protonation (or at low pH), the imidazole ring flips to the "out" conformation to allow the proton to be transferred to a buffer for transport to the outside world. [From Christianson and Fierke (176), copyright 1996 American Chemical Society.]

The catalytic rate of CA II is limited by an intramolecular proton transfer at high buffer concentrations, but is buffer-limited at low buffer concentrations (620). When the buffer concentration is decreased below 10 mM, the catalytic rate of hydration at steady state decreases greatly because the supply of external protons is compromised (507, 621, 860, 933, 934, 1022). This result indicates that the proton is transferred from His⁶⁴ to a buffer molecule in the bulk solution. Rate constants are consistent with a diffusion-limited bimolecular proton transfer between enzyme and buffer (cf. Fig. 1) that depends only on the pK_a difference between buffer and the binding site, now known to be His⁶⁴ (860). When His⁶⁴ is replaced by Ala in human CA II, the maximum rate of CO₂ hydration is reduced 20-fold in the absence of buffers, and this can be overcome by chemical rescue by certain exogenous buffers (1023).

Intriguingly, the presence of His⁶⁴ appears to facilitate proton transfer, rather than slowing it. Two isoforms that lack any equivalent of His⁶⁴, CA III and CA V, are slower enzymes (104, 1020). Human CA III does not have His⁶⁴ or any comparable proton shuttle, and its catalytic rate is only 10³ s¹ (491). The absence of His⁶⁴ does not account for all of this effect, however. In the mutant K64H HCA III in which a His is introduced at position 64, the activity is enhanced only ~10-fold (491), leaving K64H HCA III still ~100-fold slower than HCA II. Evidently, there are other differences in the active-site cavity of HCA III that slow catalysis. Replacing Lys⁶⁴ in CA III with Asp or Glu (K64E and K64D mutations) increases the proton transfer rate 20-fold (829). It appears that the presence of a titratable residue (His, Asp, or Glu) at the entrance to the proton channel in carbonic anhydrases enhances proton conduction by enabling direct Eigen-type proton transfer between buffer and the enzyme.

Like voltage-gated proton channels, the proton pathway in CA II is inhibited by divalent cations (Table 1).

The fundamental mechanism of energy transduction in animal cells is the electron transfer chain in mitochondrial membranes, which generates a proton gradient that is used to synthesize ATP. There is a measurable proton conductance in some mitochondrial inner membranes, which results in "slippage" or dissipation of the proton gradient (125, 342, 572, 753, 759). Brown fat cells also have an inducible proton leak mediated by a 32-kDa uncoupling protein (UCP), formerly called thermogenin, in the inner mitochondrial membrane (754). The UCP-mediated proton leak is activated by fatty acids and suppressed by purine nucleotides (ATP, GDP) or serum albumin, at least in isolated mitochondria, and functions to dissipate the mitochondrial proton gradient, which forces the mitochondria to work harder and generate heat metabolically. Substantial interest in UCPs is based on the hope that these molecules can be exploited to combat obesity by dissipating energy. There are now three isoforms of UCP, with the one in brown fat being UCP1. The sequences of UCP2 and UCP3 are 55-59% identical with UCP1, but 71-73% identical with each other (109, 314, 1042), and their contribution to physiological uncoupling is less well established. UCP3 is highly expressed in skeletal muscle (109, 369, 1043). However, UCP3 knock-out mice are not obese (369, 1043). Although some evidence suggests an uncoupling function of UCP3 (369, 484, 1043), Cadenas and co-workers (138, 277) found no change in mitochondrial proton conductance in UCP3 knock-outs and concluded that UCP3 does not form a basal proton conductance pathway. The thermogenic response to norepinephrine or fatty acids was abolished in UCP1 knock-out mice in spite of high expression of UCP2 and UCP3, indicating that UCP1 alone is responsible for thermogenesis in brown fat cells (665, 666).

It is not clear that UCP1 is a proton channel. A recent novel proposal is that the thermogenic function of UCP1 is secondary and that its main role is to protect mitochondria from reactive oxygen species when stimulated by superoxide anion (277). In light of other examples of H⁺ channels with low conductance (Fig. 13), the low turnover rate of UCP1 (50-600 H⁺/min at 11°C) (549) does not preclude its being a channel. The proton conductance increases with the proton-motive force (549) and is greatly and nonlinearly enhanced at large values (e.g., >100 mV) of proton-motive force (373). Similar observations were made in skeletal muscle mitochondria, although it was not clear whether UCP3 was involved (137). Two His residues in UCP1 appear to facilitate proton transport; H⁺ flux was reduced 90% by mutation of either and abolished by the double mutation (94). Neutralization by mutation of either of two Asp residues reduced proton flux by >75% (278). Replacement of Asp with Glu preserved almost normal proton transport (278). These results are suggestive of a HBC conduction mechanism, as explored in detail by Klingenberg and Echtay (548). A novel feature is the inclusion of carboxyl groups from fatty acids in the HBC (Table 2).

View larger version (21K): [in this window] [in a new window]   Fig. 13. Compilation of data in the literature for single-channel H+ currents through various proton-conducting channels, mostly obtained at symmetrical pH. Lines connect data from each source. All values have been scaled linearly with voltage to 100 mV. The pH is simply taken as the bulk value; the local pH near a membrane made of charged phospholipids can differ substantially from bulk pH (215, 554), and the H+ conductance of gramicidin channels has been shown to reflect the local pH (253a). The dashed line shows the maximum H+ current if diffusion of H+ to the channel were rate limiting (Eq. 2), assuming a capture radius ro of 0.87 Å as found for gramicidin (230). Note that the H+ conductance is proportional to [H+] over a wide range for several channels that are believed to be water-filled pores. The product of the conductivity of concentrated HCl at 25°C (786) and its concentration is plotted for comparison (+), after arbitrary scaling. The parallel behavior of this parameter and the conductance of ion channels studied at high [HCl] suggests that the apparent saturation of single-channel H+ current may simply reflect bulk properties of HCl (244). If so, then the apparent saturation of H+ conductance of the gramicidin water wire reflects diffusion limitation external to the channel, and the maximum H+ flux that a water wire can sustain has not been attained experimentally. Possible explanations of the "shoulder" in the gramicidin data between pH 2 and pH 1 are discussed in Reference 244. Sources of data for the channels are as follows: () gramicidin A (14, 208, 230, 453, 748); () covalently linked gramicidin dimer in GMO membranes (206, 208); (small dot within a circle) gramicidin A (292); (large dot within a circle) gramicidin A (421); (blue ) M2 proton channel of the influenza A virus (low pH, Ref. 273; high pH, Ref. 617); (gold ) "5-HT" serotonin receptor H+ selective "leakage" current (146); (brown ) "MotA" flagellar motor torque generator (682); (aqua ) CFo the proton channel component of H+-ATPase (low pH, Ref. 887; high pH, Ref. 21); (orange hexagons) electron-coupled proton transfer rates for both proton transfer steps in the Rhodobacter sphaeroides bacterial reaction center (788); (blue ) H+ flux through the Fo proton channel of H+-ATPase in vesicles (143); (green ), H+ flux through the R371H K+ channel voltage sensor (960); (green ) cytochrome c oxidase proton uptake rate (526); and (green ) (LSLLLSL)3 and (green ) (LSLBLSL)3 are 21-amino acid synthetic channels (254). For carbonic anhydrase (1020), the turnover rate is plotted in purple letters without symbols for carbonic anhydrase II (CA2), carbonic anhydrase III (CA3), which lacks His64, and K64H, a CA III mutant into which the His64 that guards the entrance to the channel in CA II has been inserted. The voltage-gated H+ channel estimates (red ) are based on noise measurements in inside-out patches of human eosinophils (168, 720). The value given for MotA is based on H+ flux/torque generator (1,200 H+/revolution, 8 torque generators/motor) during flagellar rotation at 100 Hz (682), with the proton-motive force scaled linearly with voltage (328) to 100 mV; the passive H+ flux through MotA incorporated into bilayers is substantially less (97, 970). The 5-HT receptor current appears to be an acid-induced "leak" mode of the serotonin receptor, induced by 5-HT addition, but not part of the normal transport cycle (146). [Modified from DeCoursey and Cherny (245).]

UCP1 appears also to mediate Cl conductance (754), leading to the suggestion that the anion channel also binds fatty acids that activate proton transport (492). Skulachev (946) proposed that although the fatty acids that activate UCP1 do not mediate proton transport by a simple weak acid mechanism (see sect. IIIA3), UCP1 completes a weak acid circuit by facilitating the flip-flop of the anionic form RCOO. In this view, UCP1 does not mediate proton permeation per se; instead, protons are transported in the form of protonated fatty acids, RCOOH. More recent studies support this conclusion (343, 484, 493). In conclusion, although the primary sequences of the UCPs are known, it is unclear whether they directly translocate protons, much less whether they might do so as channels or as carriers.

There are several reports of proton conductance associated with expression of various membrane proteins that are not yet sufficiently characterized to speculate on the mechanism involved. In Xenopus oocytes expressing the excitatory amino acid transporter EAAT4, arachidonic acid substantially enhances the substrate-gated (not voltage-gated) current by eliciting a proton-selective conductance (301). The rat serotonin receptor expressed in Xenopus oocytes mediates a proton-selective conductance at low pH, with exceedingly small unitary currents (~20 protons · s¹ · transporter¹ at pH 5.5) (145, 146). A non-voltage-gated conductance in Rana catesbeiana taste receptor cells is reportedly permeable to K⁺ and H⁺ (555). One conformation of the Na⁺-K⁺-ATPase appears to be capable of proton conduction (1060). However, the reversal potential varied 82.7 mV/unit pH, suggesting that the Na⁺-K⁺-ATPase can transport protons in a manner like the H⁺-K⁺-ATPase, rather than as a passive channel (841). Envelope proteins of Semliki Forest virus can function as proton channels, possibly as a triple helical structure (888). Sarcoplasmic reticulum membranes appear to contain pathways permeable to protons that may serve to compensate for Ca²⁺ flux (679, 680). Renal cortex membranes may contain proton pathways that dissipate pH gradients (836). A light-induced H⁺ current is mediated by Channelrhodopsin-1 in the alga Chlamydomonas reinhardtii (285, 729). A H⁺-coupled oligopeptide transporter from Caenorhabditis elegans may function mainly as a proton channel (302). Uncoupled H⁺ leak occurs through a H⁺-coupled amino acid transporter in Arabidopsis thaliana (103). Diphtheria toxin induced nonselective cation channels are permeable to protons (876).

In Table 1, selected properties of a number of proton-conducting molecules are compared. Most channels listed are highly proton selective, a few are nonselective but can conduct protons, and aquaporins are impermeable to protons and other cations. Selectivity is determined by the nature of the proton-conducting pathway. The nonselective proton-permeable channels, gramicidin and voltage-gated sodium channels, are known to be water-filled pores. Gramicidin conducts many cations, whereas Na⁺ channels preferentially conduct Na⁺ and Li⁺, but both conduct protons efficiently across the water wire that they contain. The proton impermeability of water channels (aquaporins) shows that not every water-filled pore conducts protons, although aquaporins exclude all cations by a similar mechanism. It is likely that any water-filled cation-selective channel will also conduct protons, although the H⁺ flux may be quite small (see sect. IVB). In all of the highly H⁺-selective channels, the pathway is formed at least in part by protonatable sites on amino acids or groups capable of forming hydrogen bonds. These are termed HBC, implying that at least part of the pathway is formed by the protein itself, although strictly speaking, a pure water wire is a subset of HBC. In most of the HBCs, the pathway includes both protonatable amino acid residues and water molecules. Three channels, M₂, carbonic anhydrase II, and the voltage sensor of a voltage-gated potassium channel, comprise a water-filled pore interrupted by a single His residue (one His per channel subunit, which is present in multiple copies in the tetrameric channels). One conclusion that can be drawn immediately from current information is that extremely high selectivity (e.g., P_H > 10⁶ P_cation) occurs when the pathway is a HBC, not a simple water wire. The "selectivity filter" (442) of proton-selective channels is thus a protonatable amino acid residue that forms part of the conduction pathway. Selectivity is accomplished because such a group can bind protons but not other cations. Implicit in this mechanism of selectivity is the requirement that protons permeate as H⁺ rather than as H₃O⁺. The high selectivity of voltage-gated proton channels suggests a HBC conduction mechanism (238).

The maximum H⁺ flux varies over seven orders of magnitude. Gramicidin channels conduct protons at a higher rate than any other narrow-pore ion channel conducts any ion, reflecting the efficiency of the water-wire conduction pathway. The rates of H⁺ conduction through HBCs that include amino acids vary widely. The high turnover rate of carbonic anhydrase II reflects efficient proton transfer directly between buffer and a His at the mouth of the pore. Lower rates can occur for various reasons, one of which is limited proton supply to the channel, which will depend on pH. This problem is discussed in section IVO.

Solvent isotope effects are often substantially larger in HBCs in Table 1 than in bulk solution (65, 325, 612, 624, 844) or in water wires such as gramicidin (14, 171). An isotope effect of 7 was found for a single rate-limiting proton transfer at a Glu residue in cytochrome c oxidase (525); thus HBCs can have quite large isotope effects. Similarly, the temperature dependence of proton conduction through water-filled pores is not much different from that of H⁺ conduction in bulk solution, whereas HBCs often exhibit strong temperature dependence. The strong deuterium isotope effect and temperature dependence of H⁺ permeation through voltage-gated proton channels are consistent with the conduction pathway being a HBC and not a water-filled pore (242, 243).

The frequency with which divalent cations appear in Table 1 as inhibitors of various H⁺ channels is striking. The titratable groups on amino acids, in particular the sulfhydryl group on Cys and the imidazole nitrogen on His, as well as carboxylate oxygens of Glu and Asp, are notorious metal binders (119, 1031, 1032). Competition between divalent metal ions and protons for binding sites on proteins is common. Zinc is a potent inhibitor of the mitochondrial bc₁ complex, binding with a K_i of 10⁷ M to a protonation site at which H⁺ and Zn²⁺ exhibit negative cooperativity (622). Two His residues in the glutamate transporter EAAT4 bind Zn²⁺, which then inhibits an anion conductance (699). Zinc binds to titratable His residues in serum albumin and insulin (119). The binding of Cu²⁺ to CA (possibly to a His residue) is inhibited at low pH (1021). A Cys residue, in cooperation with a His, binds Zn²⁺ competitively with protons and mediates proton inhibition of an inward rectifier K⁺ channel (187). Zinc inhibits N-methyl-D-aspartate receptors by interacting with His residues, apparently by "flexible coordination"; mutation of each of the 3 most critical His lowered Zn²⁺ affinity >100-fold, mutation of any of 4 other His lowered Zn²⁺ affinity 4- to 8-fold (626). It is not too surprising that metals would tend to bind at the same protonatable sites on a protein that comprise the HBC conduction pathway, although this requires the sites to be accessible to the solution. A skeptic could argue that any protein is apt to have many sites that could bind divalent metals, and certainly the effects of heavy metals on normal ion channels are qualitatively similar to their effects on voltage-gated proton channels, although H⁺ channels are demonstrably more sensitive (163, 642).

Table 2 shows the proposed composition of proton pathways in a number of molecules whose function involves proton translocation across membranes. Several points emerge. The specific amino acid composition of proton pathways varies dramatically. It would require both intuition and exceedingly good luck to identify the voltage-gated proton channel by deducing possible proton pathways and searching the GenBank for candidates. The pathways comprise a variety of amino acid residues intercalated by water molecules as required to fill in gaps. Most of the amino acids are titratable and presumably are protonated transiently during proton permeation, although the specific function of each residue is controversial in many instances. Some of the amino acids in the pathways in Table 2 may act to stabilize within the channel the water molecules that actually translocate protons. Some are critical because they interact with other parts of the molecule in essential ways. Interpretation of site-directed mutagenesis studies is rarely unambiguous, demonstrating that an amino acid mutation that prevents proton conduction does not discriminate which role this residue plays. As discussed in section IIID, the archetypal proton channel may comprise a water wire interrupted by only one or two titratable amino acids residues, as originally proposed for the H⁺-ATPase (111). Ten amino acids that are proposed to form the F_o proton channel are shown in Table 2, but these may mainly provide a matrix for the water molecules that conduct protons. The actual proton pathway may be a water wire leading to the Asp⁶¹, as emphasized by recent studies (306).

Three titratable amino acids with pK_a lower than physiological pH, Glu, Asp, and His, are prominent in Table 2, presumably reflecting their suitability for H⁺ conduction. Groups with higher pK_a are reluctant to release the proton and thus as HBC elements would slow conduction. However, the pK_a of an amino acid inside a protein can differ substantially from that in free solution, depending on the extent to which the group is buried in the protein and also on the hydrophobicity or hydrophilicity of the local microenvironment (677, 899). In several cases, conformational changes in the protein dramatically change the pK_a of crucial amino acids during proton transport (e.g., bacteriorhodopsin and H⁺-ATPase). His seems to appear with regularity in proton pathways. This should perhaps not be too surprising, since His has a nominal pK_a of 6.0, which means that it can be protonated at physiological pH, but does not hold the proton too tightly. Three minimal HBC pathways comprise aqueous pores occluded by His at one point: the M₂ viral proton channel, CA II, and a channel constructed artificially by mutation of the voltage sensor of voltage-gated K⁺ channels, resulting in aqueous pathways that are occluded at one point by a His residue. The His occlusion imparts strong proton selectivity and may allow protons to pass by means of a hypothetical ring flip (811), as was first proposed for the CA proton channel (735, 932).

Titratable amino acids can serve two important functions in proton channels. In addition to forming part of the conduction pathway, they can mediate molecular functions, in particular, pH sensitivity. It is vital to the function of most proton channels that they respond appropriately to pH. The exquisite regulation by pH of the voltage dependence of voltage-gated proton channels has been proposed to be mediated by the degree of protonation of titratable sites that are accessible to external and internal solutions (166), with the external sites comprising three His residues (163). There are numerous other examples of titratable residues serving as regulatory protonation sites. Proton-pumping nicotinamide nucleotide transhydrogenases use a proton gradient to synthesize NADPH from NADH. The activity of this enzyme is regulated by the degree of protonation of a His residue (95). Gating of the ClC-1 Cl channel is regulated by pH_o (475, 476, 791, 864, 1063). The bulk of site-directed mutagenesis evidence points to Cys residue involvement in ClC-0 and ClC-1 channel gating as well as Zn²⁺ and Cd²⁺ binding and block by anthracene-9-carboxylic acid (7, 577, 578, 618, 863). The inhibition of ROMK1 channels by low pH_i is mediated by four His residues (151), but also by a Lys (899).

A general rule seems to be that titratable groups at the entrance to a proton channel can increase the proton flux by an order of magnitude or more. Mutational excision of the two His at the mouth of the BRC proton channel slows proton uptake by at least two orders of magnitude (787). Introducing His near the entrance to the CA III proton channel enhances activity ~10-fold (491), and Asp or Glu (K64E or K64D mutations) increases the proton transfer rate 20-fold (829). Proton currents through gramicidin channels are enhanced 12.5-fold by formic acid, which evidently supplies protons by binding transiently to the mouth of the channel (230). Apparent rate constants for protonation of membrane-associated protonophores that greatly exceed the diffusion limit can be accounted for by a direct proton transfer reaction of the Eigen variety (83).

The "chemical rescue" technique (1011, 1015) has become a fashionable way to support the idea that a particular titratable group plays a role in H⁺ conduction. The amino acid in question is eliminated by mutation, resulting in loss of function, and then function is restored by addition of titratable compounds such as imidazole or pyridine derivatives (25, 10, 787, 1017, 1023), or azide, formate, or cyanate for amino acids with lower pK_a (126, 990, 1011). The titratable group must be accessible to the solution, such as His⁶⁴ in CA II (25). A novel variant of the chemical rescue approach was the restoration of proton uptake via the D channel in D132N or D132A mutants of cytochrome c oxidase by exogenous long-chain unsaturated fatty acids, such as arachidonic acid (304).

The expectation of the GHK equation (368, 444, 456) is that the single-channel H⁺ current should be proportional to [H⁺], and thus should increase 10-fold/unit decrease in pH. This prediction is borne out for several channels. Figure 13 illustrates the pH dependence of single-channel proton current for every ion channel reported to conduct protons. For gramicidin and certain synthetic channels, the unitary H⁺ current is nearly proportional to [H⁺] between pH ~0 and pH 3.75, except for a "shoulder" region between pH 1 and pH 2 (see sect. IVA3). This proportionality may be extended to pH 4.6 by including noise measurements of Neher, Sandblom, and Eisenman (292) (two left-most points in Fig. 13), or based on more indirect measurements, to pH 7.5 (571) or even pH 8.5 (396). Simple extrapolation of the H⁺ current measured in various channels at low pH into the physiological range, on the GHK assumption of direct proportionality, provides an estimate of <0.1 fA at pH 7. Voltage-gated proton channels and several other channels deviate from this pattern in two important respects. First, their H⁺ current has a weaker dependence on pH. Second, at pH >6, the estimated single-channel H⁺ current is higher than the nominal diffusion limit, indicated on the graph by a dashed red line.

Some lack of proportionality between H⁺ concentration and H⁺ current is evident for voltage-gated proton channels, CF_o, M₂, the 5-HT receptor, and the K⁺ channel voltage sensor mutant R371H, all studied reasonably near the physiological pH range. The macroscopic g_H,max of voltage-gated proton channels increased only approximately twofold/unit decrease in pH_i (234) (see sect. VF). Surprisingly, single H⁺ channel currents appear to depend more strongly on pH_i (168); nevertheless, the dependence is less than directly proportional. The proportionality between gramicidin H⁺ current and [H⁺] generally has been taken to indicate that the current is limited by the diffusional supply of protons to the mouth of the channel, rather than by permeation through the channel (230). Protons diffuse through gramicidin channels essentially unhindered, at a rate comparable to that in bulk solution (206). The anomalous behavior of voltage-gated proton channels has been interpreted to indicate that the rate-determining step in conduction occurs within the channel itself and not in the diffusional approach (166, 238, 239, 244, 245). However, it should be emphasized that strict proportionality has not been demonstrated for any proton channel in the physiological pH range.

The second complication in Figure 13 is that for several molecules (CF_o, MotA, the BRC, cytochrome c oxidase, and voltage-gated proton channels), the estimated elementary proton flux exceeds the diffusion limit, indicated on the graph by a dashed red line, by 1-2 orders of magnitude. The problem of proton supply is most severe for CF_o proton channels, whose estimated unitary conductance is nearly 1 pS and appears to be pH independent between 5.5 and 8.0 (21, 512). The maximum current permitted by diffusion (I_max; dashed line in Fig. 13), is given by (27, 71, 441, 593)

(2)

for a hemispherical approach, where F is Faraday's constant, r_o is the capture radius, D_H is the H⁺ diffusion constant, 8.65 × 10⁵ cm²/s at 20°C (845), and c_H is the H⁺ concentration. Applying Equation 2 to H⁺ channels is complicated by uncertainty as to the appropriate value of r_o, the effects of buffer, and the actual c_H near the membrane. On a macroscopic scale, r_o is defined as the difference between the radius of the (spherical) permeating ion and the radius of the (cylindrical) pore. The probability of permeation is assumed to be unity when the entire molecule enters the pore without hitting the edges but is presumed to be zero when any part of the molecule collides with the pore mouth (303). On the molecular scale, r_o becomes an operationally defined parameter (26, 452), which may be effectively increased by various mechanisms. Due to the special mechanism of H⁺ movement through water, its effective reaction distance is large and thus r_o might be larger than for ordinary ions (287). For proton currents, r_o has been estimated to be 1 Å for a synthetic proton channel (599) and 0.87 Å for gramicidin (230), much larger than 0.12-0.33 Å for "ordinary" monovalent cations in gramicidin (26, 27). Calculations by Sacks et al. (867) suggest that even larger distances (i.e., perhaps tens of Å) might apply for H⁺ movement in the plane of the membrane (see sect. IID).

Because protonated buffer is present at 10⁶ higher concentration than H⁺, buffer ought to compensate to some extent for H⁺ depleted by current flow. However, the calculated effects of buffer on H⁺ diffusion toward a channel, based on the model of Nunogaki and Kasai (768), indicate that the effective diffusion limit is increased less than twofold by 100 mM buffer (240). Similarly weak enhancement by weak acid buffers was calculated by Decker and Levitt (230). However, they found that formic acid increased the H⁺ current through gramicidin ~12-fold, apparently by transiently binding to the mouth of the channel and then dissociating (230). The formic acid effect appears to be analogous to the enhanced H⁺ conductance in CA, where large buffer effects occur (see sect. IVK) when His or another protonatable amino acid is present at the mouth of the proton channel (507, 860, 933, 934, 1022). Evidently, the proton supply can be enhanced by buffer to a much greater extent if there is direct proton transfer between buffer and a proton acceptor on the target molecule. In fact, if protonated buffer were the permeating species, the unitary current limit according to Equation 2 increases to 1.9 pA (238). However, the insensitivity of voltage-gated proton currents to buffer concentration between 1-100 mM rules out direct proton transfer from buffer to the channel as a rate-determining step for this channel (240).

If we assume that r_o is 0.87 Å as in gramicidin (230), then the diffusion-limited single-channel H⁺ current given by Equation 2 is 1.4 fA at pH 5.5 and only 14 aA at pH 7.5. Incorporating the effects of 100 mM buffer according to Nunogaki and Kasai (768) increases these estimates only by a factor of 1.26 or 1.42, respectively for MES or HEPES buffers (240). Up to a point, arbitrary scaling of r_o can be justified by postulating that the entire membrane acts as a "proton-collecting antenna" (see sect. IID) and then simply calculating how much surface area each channel could use for this purpose (244). With the use of values derived below (see sect. VIH6), each H⁺ channel in an eosinophil, with the highest H⁺ channel expression of any native cell, has a surface area of 5-8 × 10³ nm² to draw from. With the use of r_o determined for H⁺ current in gramicidin, 0.87 Å (230), to estimate the effective "capture distance," i.e., the distance from the membrane surface from which a proton could leap to the membrane in a single bound, converting the resulting volume enclosed by the disk to its hemispherical equivalent, and determining its radius, for the voltage-gated H⁺ channel at pH 6.5, the effective capture radius is ~60-70 Å.⁴ This value is 70- to 80-fold greater than r_o assumed in Figure 13. As evident from Equation 2, the "diffusion limit" scales up in direct proportion to r_o. The estimated unitary H⁺ currents for voltage-gated proton channels fall within this new upper limit, indicating that sufficient membrane area is available to provide an adequate supply of protons to each H⁺ channel. Among several crude assumptions in this calculation is the nontrivial assumption that surface H⁺ conduction is infinitely rapid.

The unitary H⁺ currents for MotA-MotB, the bacterial reaction center, cytochrome c oxidase, as well as voltage-gated proton channels, are all reasonably close to these rough limits, although the CF_o proton channel at high pH remains well above. Several mechanisms that could increase the supply of protons to a channel and thus effectively increase r_o in the above equation (593), include 1) a negatively charged vestibule that funnels protons toward the mouth of the channel (26, 155, 214, 508); 2) hydrolysis (527); 3) rapid conduction of protons at the surface of the membrane (404, 418, 706, 724, 820, 821); 4) buffers (399, 768); 5) electrodiffusion, which may increase r_o by one Debye length (806); 6) increased local proton concentration due to negative surface charges on the membrane (32, 215, 378, 401, 554, 988) or on the channel itself (31, 362, 382), a single negative charge can double the conductance at low salt concentration (140, 383); 7) titratable membrane groups acting as proton antennae (253a, 400); and 8) buffering by phospholipid head groups (390, 536). Some of these mechanisms overlap conceptually. A fundamental problem is that although there are ample protons in H₂O and bound to buffers, any proton concentrating mechanism that involves charge (mechanisms 1, 3, 5, 6, 7, 8) will only work with free H₃O⁺. Furthermore, negative charges concentrate protons at the surface most effectively when the ionic strength is low (214, 988) or if H⁺ is the only cation present, although the smaller size of H₃O⁺ than a hydrated cation may be advantageous from this perspective (342). No mechanism that invokes the consumption of a transient store of protons is viable, because voltage-gated proton channels do not inactivate (238) and carry sustained current for minutes (85, 236). The effect of surface charges (on the membrane or channel) on local pH can be large. In positively charged ethylated diphytanoyl phosphatidylcholine membranes, the surface pH can be 2 pH units higher than in bulk solution, although this result applies to pure HCl solutions (253a). Conversely, at high pH, the pH near negatively charged biological membranes can be 2 pH units lower than in bulk solution (215, 554), with larger effects seen at low ionic strength (987, 988). If the surface pH were two units lower than bulk, the problematic supply of protons would immediately be resolved for all proton channels except CF_o.

Kasianowicz, Benz, and McLaughlin (527) considered three mechanisms by which H⁺ might reach the membrane from bulk solution. H⁺ can arrive as a free proton (protolysis mechanism), by hydrolysis of water, or by direct proton transfer from protonated buffer. The rate at which H⁺ channels are protonated, assuming the protolysis mechanism, was 4.9 × 10⁹ M¹ · s¹ at pH_i 5.5 and 2.0 × 10¹¹ M¹ · s¹ at pH_i 7.5 (240). Kasianowicz et al. (527) obtained higher apparent rates than these and thus concluded that hydrolysis was the only possible source of protons. The apparent protonation rate of H⁺ channels is faster than the fastest reaction occurring in free solution, the recombination of H⁺ with OH at 1.3-1.4 × 10¹¹ M¹ · s¹ (80, 287), but within those reported for proton transfer reactions between groups at the surface of the membrane (see sect. IID). Thus protolysis remains a viable mechanism, but hydrolysis is not ruled out as a possible source of protons.

Because five different molecules (the BRC, MotA-MotB, cytochrome c oxidase, CA III, and voltage-gated proton channels) studied in very different ways all exhibit comparably large unitary H⁺ flux in the physiological pH range, it seems evident that nature does not view proton supply as a problem, whether or not we fully understand the mechanisms involved. In at least two cases, CA II and the BRC, proton flux is enhanced by the presence of one or more His residues at the channel entrance (10, 419, 491, 787). The higher proton flux through CA II is facilitated by direct proton transfer from protonated buffer to the channel (see sect. IVK).

	V. VOLTAGE-GATED PROTON CHANNELS: GENERAL PROPERTIES

Top Previous Next References

Voltage-gated proton channels were first identified explicitly as a distinct entity by Thomas and Meech (1008) in their pioneering study of snail neurons (Fig. 14). Proton currents likely had been observed in earlier studies in Helix, Limnea, and Planorbis neurons but were misidentified as Cd²⁺-sensitive nonspecific currents (133, 569). Earlier, Thomas (1005) had observed Na⁺-independent pH_i recovery after HCl injection in snail neurons depolarized with high [K⁺] that probably was mediated by voltage-gated proton channels, but suggested K⁺/H⁺ exchange as a possible mechanism. Byerly, Meech, and Moody (134) thoroughly characterized the electrophysiological properties of voltage-gated proton currents in snail neurons under voltage clamp. Shortly thereafter, Barish and Baud (70) described similar voltage-gated proton currents in Ambystoma (axolotl) oocytes. For almost a decade, snail neurons and newt oocytes were the only cells known to express these channels. In 1991, voltage-gated proton channels were first described in mammalian cells, rat alveolar epithelial cells (232). H⁺ currents in human cells were reported in 1993 (236, 258). To date, nearly 100 voltage-clamp studies and reviews of voltage-gated proton channels have been published.

View larger version (23K): [in this window] [in a new window]   Fig. 14. The first description of voltage-gated proton channels was based on elegant simultaneous measurements (from top to bottom) of membrane potential (Em), voltage-clamp current, HCl injection current, and pHi in snail neurons (Helix aspersa) by Thomas and Meech (1008). Several HCl injections lowered pHi to <6.5, then the membrane was depolarized stepwise. At 10 mV and at more positive voltages, a decaying outward current was observed. The recovery of pHi (bottom trace) from the acid load clearly is faster at more positive voltages, consistent with the outward current being carried by H+. The outward H+ current decays because the driving force (V-EH) decreases as pHi increases (due to the continuous H+ efflux). Large H+ currents in small cells decay two orders of magnitude faster because of the much smaller cell volume than in snail neurons. The neuron was pretreated with CsCl to inhibit K+ currents and SITS to inhibit other endogenous pH regulatory mechanisms. [From Thomas and Meech (1008), as modified by Eder and DeCoursey (280).]

Early reports of voltage-gated proton channels were met with normal scientific skepticism. Are H⁺ currents actually conducted through other ion channels? Are H⁺ currents an artifact of the unusual pH and ionic conditions used to study them? Do H⁺ currents serve any useful function in cells of interest to anthropocentrists? Because these issues arise logically, they will be discussed briefly. First, the idea that H⁺ current might simply reflect H⁺ permeation through other ion channels, specifically voltage-gated K⁺ channels, can be refuted unequivocally. First, it is possible to inhibit other ionic currents that are superimposed on H⁺ currents, including Ca²⁺, Ca²⁺-activated K⁺, voltage-gated K⁺, or Cl currents by various channel blockers listed in Table 6 or by ion substitution, without effect on H⁺ currents (70, 237, 499, 641, 710). In searching (in vain) for specific inhibitors, numerous investigators have tried batteries of blockers of other ion channels and ion transporters, without identifying any that block voltage-gated proton channels (Table 6). Conversely, H⁺ channels are more sensitive to inhibition by Zn²⁺ than are other ion channels in the same cells (85, 163, 474, 642). H⁺ current could not be mediated by delayed rectifier K⁺ channels, because these currents had different Q₁₀ values and different distributions in both whole cell and excised patch configurations (136). H⁺ current could not be mediated by Ca²⁺ channels, because Ca²⁺ currents "wash out" with time, but H⁺ currents do not (498). Finally, it is clear that the unitary conductance of voltage-gated proton channels is very small indeed (see sect. VG), with the best estimates being ~10³ smaller than that of voltage-gated K⁺, Na⁺, or Ca²⁺ channels in physiological solutions. The size of whole cell H⁺ currents at low pH_i in several cells is as large or larger than conventional ionic currents in the same cells (134, 232, 281, 372, 574, 641). Therefore, there must be ~10³ more voltage-gated proton channels than other ion channels. For example, if one assumed that the 130 voltage-gated K⁺ channels in the average rat alveolar epithelial cell (250) could conduct protons, these K⁺ channels would carry <1 pA of H⁺ current at physiological pH, assuming a gramicidin-like H⁺ conductance, which would be difficult to detect in the presence of macroscopic voltage-gated proton currents approaching 1 nA. Clearly, voltage-gated H⁺ current is mediated by a unique channel.

Voltage-gated proton channels are usually studied by removing permeant ions (K⁺, Ca²⁺, Na⁺, and Cl) from the solutions to preclude contamination by other types of ionic currents. As a result, the question arises whether H⁺ currents might somehow be caused by unusual ionic conditions and whether they are still detectable under physiological ionic conditions. This question has been resolved unambiguously. H⁺ currents can be seen at physiological pH, in quasi-physiological solutions such as Ringer solution (68, 70, 74, 232, 236, 372, 473, 498, 499, 574, 676, 1008), in large cells with intact cytoplasm (70, 74, 473, 641, 676, 1008), in cell-attached patches exposed to cytoplasm (239), in permeabilized patch studies (165, 246-248, 387), and at physiological temperatures (136, 243, 259, 311, 574) where H⁺ currents are much larger than at room temperature (136, 170, 243, 574).

Finally, strong evidence exists that H⁺ currents in intact cells serve a variety of important physiological functions. Voltage-gated proton currents are activated by various agonists of the respiratory burst in phagocytes that are studied using permeabilized patch technique, i.e., with intact cytoplasm and during physiological responses (165, 246, 248). H⁺ efflux attributed to H⁺ current is elicited in airway epithelial monolayers by histamine or ATP (311). Voltage-gated proton channels in intact cells are activated during recovery from acid loads (574, 641, 676, 762, 1006, 1008) and during various other physiological responses (see sect. VI).

The cells in which voltage-gated proton channels have been demonstrated by direct voltage-clamp measurement are listed in Table 3. Table 3 is dominated by leukocytes and related cell lines, although epithelia are also well represented. The proposal of a specific function for H⁺ channels in human neutrophils (429), even before their existence was confirmed by voltage clamp (236), provided strong motivation to study neutrophils and related cells. Essentially every leukocyte that has been examined expresses voltage-gated proton channels. It is tempting to remark on the prevalence in Table 3 of cells that are subject to extreme pH in their local environment, namely, epithelia, kidney cells, oocytes, and leukocytes. The pH_o in abscesses, tumors, or synovial fluid in septic arthritis is typically lower and more variable than in normal tissue (365, 423, 942, 996, 1061) and can drop to pH 5.8 (996). On the other hand, the difficulty in finding cells that do not express H⁺ channels, e.g., to serve as a heterologous expression system, suggests that other cells that have not yet been examined systematically may also express these channels. Each year H⁺ currents are reported in several new cell types.

The H⁺ current density in different cells varies over approximately three orders of magnitude. The highest expression is in eosinophils, whose respiratory burst (see sect. VIH) is more intense than in other phagocytes (246, 360). Very few cells have been identified as having no H⁺ channels. The common expression systems, CHO and HEK-293 cells, both express low levels of voltage-gated proton channels. A literature search for "lymphocyte" and "ion channel" reveals 735 publications, but until recently (886), the presence of voltage-gated proton channels had not been detected. One explanation is that H⁺ currents in T lymphocytes are quite small: the whole cell current at +60 mV is only 1.5 pA at pH_o 7.5 and pH_i 6.0 (886) and presumably is smaller at physiological pH_i. Thus far, only COS-7 cells have been reported to lack detectable H⁺ currents (668, 705).

Table 3 excludes reports that deduce the possible existence of voltage-gated proton channels from pH measurements, because of the difficulty in excluding other mechanisms, such as other proton transporters. ZnCl₂-inhibitable H⁺ efflux in chicken enterocytes induced by phorbol esters or by an acid load might reflect voltage-gated proton channels (141, 802). A Zn²⁺-sensitive proton influx reported in leech central neurons during recovery from an intracellular alkaline load (322) appears distinct from voltage-gated proton channels, which mainly conduct outward currents. The OK cell line has a passive proton pathway that conducts inward or outward currents and is insensitive to Zn²⁺ (376). The plasma membrane of Elodea densa leaves is extremely selective for H⁺ at high pH, and the existence of H⁺ channels has been proposed (687). A La³⁺-sensitive, arachidonic acid-activatable proton conductance has been described in platelets, based on pH measurements (149). Intriguingly, aldosterone activates a proton conductance in Madin-Darby canine kidney (cultured kidney) cells that is inhibited by ZnCl₂ and is voltage sensitive in the sense that hyperpolarization increases H⁺ influx (349). However, a large H⁺ influx was observed at 90 mV with a small pH gradient (pH_o 7.4, pH_i approaching 6.7), apparently incompatible with the threshold potential (V_threshold) of voltage-gated proton channels, even in phorbol 12-myristate 13-acetate (PMA)-activated phagocytes in which V_threshold is shifted by 40 mV (cf. Fig. 19).

Although voltage-gated proton channels share a number of distinctive properties, it is possible to distinguish four or five varieties based on several functional properties that are summarized in Table 4. The functional categories are certainly arbitrary but will have to suffice until proton channel molecules are definitively identified. Future structure-function studies will reveal whether the apparent differences reflect structural or regulatory differences. All voltage-gated proton channels are activated by membrane depolarization, and all are sensitive to pH. The most dramatic difference between types of H⁺ channels is in gating kinetics. The rates of channel opening and closing (indicated by the time constants _act and _tail, respectively, which are larger when gating is slower) vary over three orders of magnitude. The sigmoidicity of the activation time course appears to differ, but this is a subtle distinction. The H⁺ channel tail current decays with two clear exponential components in alveolar epithelial cells at potentials positive to V_rev (166). In other cells, deactivation is well described by a single exponential, except at voltages positive to V_threshold at which in the steady-state some channels remain open. The slower component of channel closing appears to be related kinetically to the process that governs channel opening, and both are exquisitely sensitive to pH_o, in contrast to the rapid tail current component that is weakly dependent on either pH_o or voltage (166).

The type x (oxidase-related) channel complicates the picture. This was proposed to be a distinct H⁺ channel that is active only when NADPH oxidase is functioning (67c). More recent work strongly suggests that type x behavior is instead a gating mode of the type p (phagocyte) channel (see sect. VIH2). Several types of evidence suggest that type p channels shift into the type x gating mode upon stimulation of phagocytes with agonists that activate NADPH oxidase (165, 246-248). After removal of the stimulus, type x gating behavior sometimes gradually reverts back to type p behavior (165, 246). For these reasons, the properties of type x behavior are somewhat elusive and have not been characterized thoroughly.

The sensitivity of voltage-gated proton channels to [Ca²⁺]_i is controversial and is discussed in more detail below (see sect. VIB2). Although modulatory effects have been reported (372, 895), it is clear that most voltage-gated proton channels are not Ca²⁺ activated (134, 739, 886). A Ca²⁺-activated H⁺ current in HEK-293 cells transfected with the gp91^phox homolog Nox5 has been reported but not characterized (67b).

Measurement of the V_rev of voltage-gated proton currents at various pH indicates extremely high H⁺ selectivity. The relative permeability of H⁺ to other cations can be demonstrated by comparing V_rev with the Nernst potential for H⁺, E_H (751), or by comparing the slope of V_rev values as a function of pH. In most studies there is not perfect agreement between V_rev and E_H and the slope of the V_rev versus E_H relationship is more often 40-50 mV/unit pH (85, 134, 232, 242, 258, 281, 519, 668, 710, 762, 886, 895), or even <40 mV/unit (259, 391, 574, 641), than near the ideal (at 20°C) of 58 mV/unit (70, 164, 166, 170, 241, 247, 372, 518). However, in part because [H⁺] is 4-7 orders of magnitude smaller than the concentration of the predominant cation, the relative permeability of H⁺ to the other cations when calculated by the GHK voltage equation (368, 444, 456) is typically >10⁶ (166, 170, 237, 238, 241, 242, 247, 258, 372, 519, 886), and is as high as 2 × 10⁸ in deuterium solutions (242). Even though these permeability ratios are impressive, they probably underestimate the true selectivity of H⁺ channels. In fact, they should not be taken to reflect finite permeability of the H⁺ channel to other cations. Most telling is that the measured V_rev does not change when the predominant cation or anion in the solution is changed, once one corrects for liquid junction potential differences (70, 85, 236, 237, 258, 281, 372, 473, 519, 574, 641, 830). Most likely, small deviations of V_rev from E_H reflect experimental error in measuring V_rev, a contribution from leak current, or imperfect control over pH, rather than genuine permeability of H⁺ channels to other cations.

In principle, it is difficult to distinguish between conductance of protons per se or proton equivalents, such as H₃O⁺ permeation or flux of OH or HCO in the opposite direction. The Nernst potential for OH is always equal to that for H⁺. However, several types of evidence indicate that the current carried through voltage-gated proton channels is in fact a proton current.

The maximum conductance (g_H,max) increases when pH_i is decreased, consistent with the increase in [H⁺] on the side of the membrane from which the current flows, whereas intracellular [OH] decreases and extracellular [OH] (the source of any OH current) is not changed. The increase is less than proportional to the change in [H⁺] (see sect. VF).

The proton conductance decreases almost 50% when deuterium replaces water (242). In terms of the classical square-root dependence of reaction rates on the mass of the reactants (366), one would predict isotope effects of 41% (for D⁺ vs. H⁺) and 3% (for OD vs. OH). The same argument speaks against bodily (hydrodynamic) permeation of H₃O⁺, which compared with D₃O⁺ has a predicted isotope effect of only 8% (242).

The extremely high selectivity can be explained readily if the conduction pathway is a HBC (see sect. IIID) but is difficult to explain if the H⁺ channel is a water-filled pore. There is strong evidence that the pathway is a HBC; this mechanism permits H⁺, but not H₃O⁺ permeation.

Indirect but compelling evidence thus supports the conclusion that the permeating ionic species is the proton (H⁺) and not the hydronium ion (H₃O⁺) or an anion like OH moving in the opposite direction. If the proton channel were a HBC and the transported species were OH, then the actual mechanism of transport would still comprise H⁺ hopping in the same direction across the membrane, but from H₂O to OH. This would result in net OH translocation in the opposite direction.

The expectation of the GHK equation is that the single-channel H⁺ current should be proportional to [H⁺] and should thus increase 10-fold/unit decrease in pH. Figure 13 shows that this prediction is borne out for several proton-conducting channels, but not for others. The g_H,max increased only ~2-fold/unit decrease in pH_i in essentially every whole cell study of voltage-gated proton currents in which the data permit such a comparison (summarized in Fig. 1 of Ref. 234), with the exceptions of a 2.5-fold/unit increase in human basophils (170) and a 3-fold/unit increase in human lymphocytes (886). This remarkable property was confirmed directly in inside-out patches from alveolar epithelial cells. Compared in the same patch of membrane, the H⁺ conductance increased only 1.7-fold/unit decrease in pH_i (239). This anomalously weak dependence of g_H on [H⁺] has been interpreted to indicate that the rate-determining step in conduction occurs within the H⁺ channel itself (234, 238, 239).

Surprisingly, single H⁺ channel currents appear to depend more strongly on [H⁺] than macroscopic currents (168). The unitary H⁺ conductance, based on H⁺ current noise (see sect. VG), increased approximately fourfold at pH_i 5.5 compared with pH_i 6.5 in excised patches from human eosinophils (168). Because the macroscopic g_H,max increases only ~2-fold/unit, the number of functional channels must decrease substantially at lower pH_i. Evidently intracellular protons inhibit H⁺ channel function, in a sort of self-inhibition. Although these single-channel data still indicate a less-than-proportional dependence of H⁺ current on [H⁺], the discrepancy has been attenuated.

Proton channels at physiological pH have a very small conductance. To some extent, the small unitary conductance may reflect the tiny concentration of the permeant ion H⁺. Extrapolating the H⁺ current of single gramicidin channels, which at low pH conduct larger H⁺ currents than any other channel (see sect. IVA2), to pH 7 (Fig. 13) indicates a predicted current of 44 aA (aA = attoampere = 10¹⁸ amperes). This value is far too small to be detectable directly by present technology. The estimated conductance of voltage-gated proton channels is in fact about an order of magnitude larger, and recently, apparent single-channel H⁺ currents in the range 5-15 fA were observed by direct electrical recording at low pH_i (5.0 or 5.5) (168).

The single-channel conductance can also be estimated from current variance analysis. Early attempts to resolve H⁺ channel gating-induced current fluctuations (85, 136, 236) met with limited success, due to poor signal-to-noise ratios (S/N). Byerly and Suen (136) established an upper bound at <50 fS at pH_i 5.9 (136). No excess fluctuations were seen (S/N = 0), but the data had to be filtered at 1 kHz because of the rapid gating kinetics in snail neurons. Bernheim et al. (85) filtered at 5 kHz, and from a 6% reduction of variance in the presence of Cd²⁺ (S/N = 0.06), estimated 90 fS at pH_i 5.5 (85). DeCoursey and Cherny improved the S/N ratio to 0.5 and estimated the unitary conductance to be ~10 fS at pH_i 6.0 in human neutrophils (236). Distinct excess fluctuations (presumably generated by H⁺ channel gating) were detected at 200 Hz but not 2-kHz low-pass filtering. All of these estimates are compromised by poor S/N ratios and should be considered very rough.

More recently, we have exploited the very slow gating in human eosinophils, combined with appropriate filtering, to improve the S/N ratio to >100 routinely, and sometimes >1,000 (168). Distinct excess fluctuations ascribable to H⁺ channel gating can be seen at voltages where H⁺ current is activated (Fig. 15). At low pH_i the variance was maximal near the midpoint of the g_H-V relationship, precisely as expected from the simple assumption that gating events (random opening and closing of channels) are most frequent when half the channels are open, decreasing with further depolarization because most channels stay open most of the time. This behavior strongly supports the idea that the noise is generated by H⁺ channel gating. These recent estimates place the unitary H⁺ channel conductance at 30-40 fS at pH_i 6.5 and perhaps four times higher at pH_i 5.5 (168).

View larger version (28K): [in this window] [in a new window]   Fig. 15. Membrane current fluctuations recorded in an excised outside-out patch of membrane from a rat alveolar epithelial cell. The current fluctuation (noise) is clearly greater at voltages where the gH is activated, at 0 mV and more positive. Currents at various voltages are shown on the same absolute scale without correction for leak; the distance between the records at 60 and 20 mV, both subthreshold, reflects a seal resistance of ~900 G. Filter, 20 Hz; pipette solution, pH 6.5; bath solution, pH 6.0 (V. V. Cherny and T. E. DeCoursey, unpublished data.)

Although the conductance of H⁺ channels is miniscule compared with other ion channels, in view of the low concentration of protons at physiological pH, the conductance seems implausibly large. Estimated unitary H⁺ currents are an order of magnitude greater than the diffusion limit (Fig. 13). Mechanisms that might reconcile this apparent paradox are discussed elsewhere (see sect. IVP). Judged solely on their conductance, voltage-gated proton channels cannot be distinguished from carriers. For example, the H⁺ efflux through Na⁺/H⁺ antiporters at pH_i 6.0 at their maximum turnover rate in human fibroblasts is equivalent to 0.5-1.7 fA (927). Nevertheless, the presence of H⁺ current fluctuations provides strong evidence of gating, a defining property of ion channels, and thus support for the designation of voltage-gated proton channels as ion channels.

Voltage-gated proton channels are extraordinarily sensitive to temperature (136, 170, 243, 574). Both the open-channel conductance and the kinetics of gating have higher temperature sensitivity than almost any other ion channel. In a survey of voltage-gated proton channels in six cell types, the time course of H⁺ current activation was fit by a single exponential after a delay to obtain _act and deactivation was fit with a single exponential to obtain the closing time constant _tail. Surprisingly, the delay, _act, and _tail all had Q₁₀ values of 6-9 (243). The Q₁₀ is the relative change in rate for a 10°C increase in temperature; these values correspond with activation energies of 30-38 kcal/mol. Only 1 of 24 studies of gating of other ion channels reported a higher value (825). The large Q₁₀ suggests that gating involves substantial conformational changes in the channel. The similarity of temperature sensitivity of the three gating parameters was surprising, because they had been envisioned as reflecting different, if overlapping, kinetic processes. The delay presumably reflects transitions between closed states; _act is slow and must reflect the entire opening process that by analogy with other voltage-gated channels might require conformational changes in multiple channel subunits, whereas _tail ought to reflect the first closing step. Nevertheless, the similar activation energies for these three processes, 30-38 kcal/mol, suggest that the same kinetic process underlies each. This might occur if H⁺ channel gating occurred generally as in the Hodgkin-Huxley model for K⁺ and Na⁺ channels (454). Each H⁺ channel might have several subunits, each of which must undergo a kinetically similar first-order conformational change during opening, whereas closing would occur as soon as one subunit underwent the reverse transition.

It is more difficult to determine the temperature dependence of the open-channel conductance, for several reasons. One must distinguish between gating and conductance. A Q₁₀ of 9.9 reported for the g_H in mast cells (574) is an overestimate that reflects a complex mixture of smaller conductance and less complete activation of H⁺ channels at lower temperatures. In this study, the H⁺ current was measured at the end of a voltage ramp of fixed duration applied at different temperatures; at low temperature a much smaller number of H⁺ channels would open by the end of the depolarizing voltage ramp. It is difficult to achieve steady-state H⁺ current at low temperature even with long pulses, because of the slow activation of H⁺ current (_act can be >100 s). The pH changes due to large H⁺ fluxes cause droop and distortion of the current, especially at high temperatures. Single-channel H⁺ currents have not been estimated at different temperatures and would be technically difficult to resolve. It is therefore necessary to make assumptions, the most significant of which is that increasing the temperature does not change the number of channels that open during a depolarizing pulse (P_open, if the total number of channels is constant). The g_H-V relationship is either unchanged or may shift negatively by ~10 mV at high temperature, and all kinetic parameters scale with temperature (243), suggesting that no qualitative changes in gating occur. More convincing is evidence from noise analysis that indicates that P_open reaches ~0.95 during large depolarizations at pH_i 5.5 at 20°C (168), which means that any increase in P_open at high temperature could not exceed 5%. Correction for the temperature dependence of the pK_a of buffer slightly decreased observed Q₁₀ values (243).

With all of these caveats in mind, the g_H is still strongly temperature sensitive. The Q₁₀ was 2.1 in snail neurons (136) and 2.1-3.1 in whole cell studies in six mammalian cell types (243). The temperature sensitivity was greater in excised patches, perhaps because there is less propensity toward H⁺ depletion during large currents. The Q₁₀ in patches was 2.8 at >20°C (18.3 kcal/mol) and 5.1 at <20°C (27 kcal/mol). The temperature dependence of several processes external to the channel (buffer diffusion, H⁺ diffusion) is expected to be much weaker than this, suggesting that permeation itself is rate determining (243). One possible exception is protolysis, donation of a proton from neutral water to the channel, which has a Q₁₀ of ~2 (371). The Q₁₀ for H⁺ conductance is larger than in any of 20 studies of other ion channels (references in Ref. 243), indicating that H⁺ permeation through voltage-gated proton channels is more demanding than ion permeation through most ion channels, presumably reflecting the HBC conduction mechanism. The Q₁₀ values for all proton channels in Table 1 that are not water-filled pores are substantially higher than typical values for ordinary ion channels, ~1.2-1.5.

The biological implication of the extraordinary temperature sensitivity of voltage-gated proton channels is that at body temperature H⁺ channels open much faster and conduct far more current than at room temperature. With the assumption of a Q₁₀ value of 2.8, the g_H,max at 37°C would be 5.8 times larger than that measured at room temperature. Although a _act of 1 s at room temperature seems very slow, if Q₁₀ is 8, _act is 30 ms at 37°C.

Most ion channels are only subtly affected by heavy water, typically exhibiting a slightly lower conductance and slightly slower gating. Voltage-gated proton currents are more sensitive to deuterium, a result that perhaps is not surprising in light of H⁺ or D⁺ being the permeating ion species, as well as the profound sensitivity of H⁺ channel gating to pH. The currents in deuterium are qualitatively similar to those in water, indicating that the channels are permeable to D⁺ and continue to exhibit voltage- and time-dependent gating. Quantitatively, the current is 1.9 times larger in H₂O than D₂O; the activation time constant _act is 3-fold slower in D₂O; the deactivation time constant _tail is hardly changed, being 1.0-1.5 times slower in D₂O; and the position of the g_H-V relationship appears similar in both solvents (242).

The isotope effect on permeation through most ion channels is similar to the isotope effect on the conductivity of the ion in bulk aqueous solution (references in Ref. 242). In contrast, the 1.9-fold larger H⁺ current than D⁺ current in voltage-gated proton channels is much larger than the ratio of mobilities or conductivities of H⁺/D⁺ in bulk solution, which are consistently 1.390-1.404 when measured electrically (65, 325, 612, 624), 1.32-1.35 in recent measurements (172), and slightly higher, 1.43-1.52 when measured polarographically (844). These values correspond with process 3 in Figure 16. The isotope effects for dielectric relaxation and for viscosity are also substantially weaker than this (180, 412). The low D⁺ conductance suggests that 1) the rate-determining step in conduction occurs within the channel, and 2) the conduction pathway is a HBC and not a water-filled pore (242) (see sect. VJ).

View larger version (18K): [in this window] [in a new window]   Fig. 16. Diagram illustrating the various processes that occur when H+ current flows through a proton channel. Many biophysical measurements on voltage-gated proton channels have been directed toward determining the rate-limiting process. Inset: predicted values for the deuterium isotope effect (predicted ratio in H2O to that in D2O) and the Q10 for H+ current if each step, in turn, were rate determining. The predicted value for the isotope effect on buffer diffusion (step 1) is based on the higher viscosity (412), and H+ diffusion (step 3) is based on the relative conductivity (65, 325, 612, 624, 844). The isotope effect for step 5 assumes a square-root dependence on the masses of OH vs. OD (366). Both the isotope effect and temperature dependence of step 4a are based on measurements of H+ current through gramicidin channels (14). The Q10 predictions are based on the temperature dependence of the viscosity of water (300, 813) (step 1), the temperature dependence of H+ mobility (606) and conductivity (786, 845) (step 3), and the conductivity of KOH solutions (628) (step 5). Observed values for voltage-gated proton channels are from Refs. 242, 243. See text for further details.

The slower activation of H⁺ current in D₂O is consistent with a rate-limiting step in channel opening being deprotonation of a site on the channel, as we proposed previously (166) (see sect. VM) (Fig. 20). The pK_a of most carboxylic and ammonium acids increases in D₂O by 0.5-0.6 units (894), because deuterons generally are held more tightly than protons. If the pK_a of the putative external regulatory protonation site were increased by 0.5 units, this would slow activation by threefold. The minimal slowing of deactivation in D₂O, in contrast, suggests that if deprotonation of a regulatory site occurs during channel closing, it is not rate determining. Alternatively, the chemical nature of the site might be different; sulfhydryl acids tend to exhibit smaller pK_a changes (894). In view of the different sensitivity of _act and _tail to D₂O, it was surprising that the overall g_H-V relationship did not shift. In the oversimplified case of a two-state channel, slowing the opening rate more than the closing rate ought to shift the g_H-V relationship toward more negative voltages. Not only was no such shift detected, there was remarkable similarity between the relationship between V_rev and V_threshold in both H₂O and D₂O (+ and , respectively, in Fig. 19). This paradox awaits explanation.

This question provides a useful framework for evaluating such properties as deuterium isotope effects and temperature dependence, as illustrated in Figure 16. Following Andersen's lucid analysis of the determinants of ionic currents through gramicidin channels (27, 30), we can identify various steps that must occur. The fact that almost all protons are not free ions, but are bound to buffer or are constituents of H₂O, increases the complexity of the problem compared with ordinary ion channels. It is necessary to consider several additional requisite processes, any of which could conceivably affect the rate of conduction. These include diffusion of protonated buffer, BH (step 1), and direct proton transfer from BH to a site at the mouth of the channel (step 2). Complementary processes that occur at the distal end of the channel could also be involved. Steps 4a and 4b indicate that conduction through the channel could occur by two qualitatively different mechanisms, diffusion through a water-filled pore like an ordinary ion channel (step 4a) and transmission through a HBC (step 4b). Figure 16, inset, gives estimates of the deuterium isotope effect and Q₁₀ expected if each process in turn were rate determining.