INTRODUCTION

Voltage-gated H⁺ channels conduct H⁺ current with extremely high selectivity and exhibit voltage-dependent gating that is strongly modulated by both extracellular and intracellular pH (pH_o and pH_i, respectively).¹ Here we explore the effects of substituting heavy water (deuterium oxide, D₂O), for water (protium oxide, H₂O), on both the conductance and the pH dependence of channel gating. The isotope effect on conductance should provide insight into the mechanism by which permeation occurs. Isotope effects on the regulation by pH (or pD) of the voltage dependence and kinetics of gating provide clues to the possible protonation/deprotonation reactions that have been proposed to play a role in channel gating (Byerly et al., 1984; Cherny et al., 1995).

Several chemical properties of D₂O and H₂O are compared in Table I. From the perspective of this study, the main differences between D₂O and H₂O are: (a) the viscosity of D₂O is 25% greater than H₂O, (b) the conductivity of H⁺ in H₂O is 1.4-1.5 times that of D⁺ in D₂O, (c) H⁺ has a much greater tendency than D⁺ to tunnel, (d) D⁺ weighs twice as much as H⁺, and (e) D⁺ is bound more tightly in D₃O⁺ and in many other compounds than is H⁺. Three main types of deuterium isotope effects are recognized: general solvent effects and primary and secondary kinetic effects. General solvent isotope effects reflect the different properties of D₂O and H₂O as solvents, such as viscosity or dielectric constant. As seen in Table I, these differences are rather moderate, and their effects are accordingly usually moderate as well. Kinetic isotope effects reflect involvement of protons or deuterons in chemical reactions. Primary kinetic isotope effects occur when H⁺ directly participates in a rate-determining step in the reaction, for example a protonation/deprotonation or H⁺ transfer reaction. For example, ionization of a number of bases is typically three to seven times slower in D₂O (Bell, 1973). Secondary isotope effects reflect D⁺ for H⁺ substitution at some site distinct from the primary reaction center. Secondary kinetic isotope effects tend to be small, 1.02-1.40 (Kirsch, 1977).

Table I. Properties of H2O and D2O at 20°C [View Table]

The conductivity of H⁺ is about five times higher than that of other cations with ionic radii like that of H₃O⁺; the limiting equivalent conductivity (⁰) at 25°C in water is 350 S cm²/equiv. for H⁺ but 73.5 S cm²/equiv. for NH₄⁺ (Robinson and Stokes, 1965). This anomalously high conductivity for H⁺ has been ascribed to conduction by a mechanism in which H⁺ jumps from H₃O⁺ to a neighboring water molecule (Danneel, 1905; Hückel, 1928; Bernal and Fowler, 1933; Conway et al., 1956). H⁺ hopping can occur faster than ordinary hydrodynamic diffusion (i.e., bodily movement of an individual H₃O⁺ molecule analogous to the diffusion of ordinary ions). After one H⁺ conduction event, a structural reorientation of the hydrogen-bonded water lattice is necessary before another proton can be conducted (Danneel, 1905; Bernal and Fowler, 1933; Conway et al., 1956). Proton conduction through channels is believed to occur by an analogous two-step "hop-turn" process through a hydrogen-bonded chain or "proton wire" spanning the membrane (Nagle and Morowitz, 1978; Nagle and Tristram-Nagle, 1983).

The mobility (measured as conductivity) of H⁺ in H₂O is 1.41 times that of D⁺ in D₂O (Table I); nevertheless, the mobility of D⁺ is still 4 times that of K⁺ in D₂O (Lewis and Doody, 1933). Thus, D⁺ also exhibits abnormally large conductivity, even though tunnel transfer of D⁺ is 20 times less likely than for H⁺ and one might have expected simple hydrodynamic diffusion of D₃O⁺ to play a larger role for D⁺, which would accordingly have a conductivity similar to that of other cations (Bernal and Fowler, 1933). Evidently the reorientation of hydrogen-bonded water molecules (the turning step of a hop-turn mechanism) is rate limiting for both H⁺ and D⁺ conduction. The nature of this rate-determining step has been proposed to be the reorientation of hydrogen-bonded water molecules in the field of the H₃O⁺ ion (Conway et al., 1956), "structural diffusion" or formation and decomposition of hydrogen bonds at the edge of the H₉O₄⁺ complex (i.e., the hydronium ion with its first hydration shell) (Eigen and DeMaeyer, 1958), or more recently, the breaking of an ordinary second-shell hydrogen bond converting H₉O₄⁺ to H₅O₂⁺ (Agmon, 1995, 1996). Some such reorganization of hydrogen bonds may also be the rate limiting step in proton translocation across water-filled ion channels such as gramicidin (Pomès and Roux, 1996).

A characteristic feature of voltage-gated H⁺ currents is their sensitivity to both pH_o and pH_i. Increasing pH_o and decreasing pH_i shift the voltage-activation curve to more negative potentials in every cell in which these parameters have been studied (reviewed by DeCoursey and Cherny, 1994). This effect of pH is reminiscent of its effects on many other ion channels, which may reflect the neutralization of negative surface charges (see Hille, 1992). However, the magnitude of the pH-induced voltage shifts for H⁺ currents has led to the suggestion that protonation of specific sites on or near the channel allosterically modulate gating (Byerly et al., 1984). In alveolar epithelial cells (Cherny et al., 1995), as well as in other cells (DeCoursey and Cherny, 1996a; Cherny et al., 1997), the shift produced by internal and external protons (H⁺_i and H⁺_o) is quite similar, 40 mV/U change in pH, within a large pH range encompassing physiological values. Thus the position of the voltage activation curve can be predicted from the pH gradient, pH, rather than by pH_o and pH_i independently. This behavior was explained by a model (Cherny et al., 1995) in which there exist similar protonation sites accessible from either the internal or external solution, but not both simultaneously. Protonation from the outside stabilizes the closed channel, whereas protonation from the inside stabilizes the open channel. Here we show that H⁺ channels are regulated in a similar manner by D⁺, but that D⁺ binds more tightly to the modulatory sites on the channel molecule.

MATERIALS AND METHODS

Alveolar Epithelial Cells

Type II alveolar epithelial cells were isolated from adult male Sprague-Dawley rats under sodium pentobarbital anesthesia using enzyme digestion, lectin agglutination, and differential adherence, as described in detail elsewhere (DeCoursey et al., 1988; DeCoursey, 1990). Briefly, the lungs were lavaged to remove macrophages, elastase and trypsin were instilled, and then the tissue was minced and forced through fine mesh. Lectin agglutination and differential adherence further removed contaminating cell types. The preparation at first includes mainly type II alveolar epithelial cells, but after several days in culture, the properties of the cells become more like type I cells. No obvious changes in the properties of H⁺ currents have been observed. H⁺ currents were studied in approximately spherical cells up to several weeks after isolation.

Solutions

Most solutions (both external and internal) contained 1 mM EGTA, 2 mM MgCl₂, 100 mM buffer, and TMAMeSO₃ added to bring the osmolarity to ~300 mosM, and titrated to the desired pH with tetramethylammonium hydroxide or methanesulfonic acid (solutions using BisTris as a buffer). The pH 8, 9, and 10 solutions contained 3 mM CaCl₂ instead of MgCl₂. A stock solution of TMAMeSO₃ was made by neutralizing tetramethylammonium hydroxide with methanesulfonic acid. Buffers (Sigma Chemical Co., St. Louis, MO), which were used near their pK in the following solutions, were: pH 5.5, pD 6.0 Mes; pH 6.5, pD 7.0 Bis-Tris (bis[2-hydroxyethyl]imino-tris[hydroxymethyl]methane); pH 7.0, pD 7.0 BES (N,N-bis[2-hydroxyethyl]-2-aminoethanesulfonic acid); pH 7.5, pD 8.0 HEPES; pH 8.0 Tricine (N-tris[hydroxymethyl] methylglycine); pH 9.0, pD 9.0 CHES (2-[N-cyclohexylamino] ethanesulfonic acid); pH 10, pD 10 CAPS (3-[cyclohexylamino]- 1-propanesulfonic acid). The pH (or pD) of all solutions was checked frequently.

A series of solutions containing NH₄⁺ was made to impose a defined pH gradient across the cell membrane, as described by Grinstein et al. (1994). The principle is that if neutral NH₃ molecules permeate the membrane rapidly enough to approach identical concentrations on both sides of the membrane, then:

(1)

because the bath solution is heavily buffered (100 mM buffer) and diffuses freely but the pipette solution (for these measurements) is weakly buffered and diffusion is slowed by the pipette tip. The shift of pH_i occurs because [H⁺]_i = pK_a  log [NH₄⁺]_i/ [NH₃]_i. The extracellular solutions were made with 100 mM HEPES, 2 mM MgCl₂, 1 mM EGTA, and various concentrations of (NH₄)₂SO₄, at pH 7.5. TMAMeSO₃ was added to bring the osmolarity to ~300 mosM. The pipette solution, which was also used externally, included 25 mM (NH₄)₂SO₄, 5 mM BES, 2 mM MgCl₂, 1 mM EGTA, and TMAMeSO₃, brought to pH 7.0 with tetramethylammonium hydroxide.

We assume that when NH₄⁺ diffuses from the pipette into the cell, if D₂O is present in the bath (and hence inside the cell) there will be rapid exchange of D⁺ for H⁺ in NH₄⁺, and that therefore efflux of ND₃ will occur, leaving D⁺ rather than H⁺ behind inside the cell. Deuterons in deutero-ammonia, ND₃, exchange rapidly with protons (Cross and Leighton, 1938).

The osmolarity of solutions was measured with a Wescor 5500 Vapor Pressure Osmometer (Wescor, Logan, UT). Deuterium oxide (99.8% or 99.9%) was purchased from Sigma Chemical Co. A liquid junction potential of ~2 mV was measured between solutions identical except that D₂O replaced H₂O. If water did not permeate the cell membrane, correction for this junction potential would make the transmembrane potential 2 mV more negative. However, as described in Fig. 1, we feel that water permeates the cell membrane, and thus there would be offsetting junction potentials at the pipette tip and bath electrode even in whole cell configuration. Therefore no junction potential correction has been applied.

pD Measurement

The reading taken from a glass pH electrode, pH_nom, deviates from the true pD of D₂O solutions by 0.40 U, such that pD = pH_nom + 0.40 (Glasoe and Long, 1960). Another estimate of this difference is 0.45 ± 0.03 (Dean, 1985), and even more disparate values can be found in early studies. Given the uncertainty about the precise value, we tested our pH meter (Radiometer Ion83 Ion meter; Radiometer, Copenhagen, Denmark) following the approach taken by Glasoe and Long (1960). Our pH meter read 0.402 ± 0.006 (mean ± SD, n = 3) higher when 0.01 M HCl was added to H₂O than when added to D₂O. We therefore corrected the pD in D₂O solutions by adding 0.40 to the nominal reading of our pH meter.

Estimation of the pK_a of the Buffers in H₂O and in D₂O

Most simple carboxylic and ammonium acids with pK_a between 4 and 10 have a pK_a 0.5-0.6 U higher in D₂O than in H₂O (Schowen, 1977). We titrated the buffers used in this study at room temperature (20-23°C). 10 mmol of buffer was added to 20 ml of H₂O or D₂O and titrated with 10 N NaOH, or 10 N HCl in the case of Bis-Tris. The resulting contamination of D₂O by the H⁺ from the base or acid titrating solutions is <3%. We corrected for this error in two ways. First, we increased the apparent change in pK_a, assuming a linear mole-fraction dependence (cf. Glasoe and Long, 1960), which increased the pK_a in D₂O by 0.02 U. We also carried out some titrations using deuterated acids and bases (DCl and NaOD, both from Aldrich Chemical Co, Milwaukee, WI). The results by these two methods were similar. The averages of two to three separate determinations for each buffer are given in Table II.

Table II. pKa of Buffers in H2O and D2O [View Table]

Electrophysiology

Conventional whole-cell, cell-attached patch, or excised inside-out patch configurations were used. Experiments were done at 20°C, with the bath temperature controlled by Peltier devices and monitored continuously by a thinfilm platinum RTD (resistance temperature detector) element (Omega Engineering, Stamford, CT) immersed in the bath. Micropipettes were pulled in several stages using a Flaming Brown automatic pipette puller (Sutter Instruments, San Rafael, CA) from EG-6 glass (Garner Glass Co., Claremont, CA), coated with Sylgard 184 (Dow Corning Corp., Midland, MI), and heat polished to a tip resistance ranging typically 3-10 M. Electrical contact with the pipette solution was achieved by a thin sintered Ag-AgCl pellet (In Vivo Metric Systems, Healdsburg, CA) attached to a silver wire covered by a Teflon tube. A reference electrode made from a Ag-AgCl pellet was connected to the bath through an agar bridge made with Ringer's solution. The current signal from the patch clamp (List Electronic, Darmstadt, Germany) was recorded and analyzed using an Indec Laboratory Data Acquisition and Display System (Indec Corporation, Sunnyvale, CA). Data acquisition and analysis programs were written in BASIC-23 or FORTRAN. Seals were formed with Ringer's solution (in mM: 160 NaCl, 4.5 KCl, 2 CaCl₂, 1 MgCl₂, 5 HEPES, pH 7.4) in the bath, and the zero current potential established after the pipette was in contact with the cell. Inside-out patches were formed by lifting the pipette into the air briefly.

For "typical" families of H⁺ currents, pulses were applied in 20-mV increments with an interval of 30-40 s or more, depending on test pulse duration and the behavior of each particular cell. Although 30 s is not long enough for complete recovery from the depletion of intracellular protonated buffer, it represents a compromise aimed at allowing multiple measurements to be made in each cell reasonably close together in time. For some measurements in which only small currents were elicited, such as pulses in 5-mV increments near V_threshold, a smaller interval between pulses was used, because negligible depletion was expected. We tried to bracket measurements in different solutions whenever possible.

Data Analysis

The time constant of H⁺ current activation, _act, was obtained by fitting the current record by eye with a single exponential after a brief delay (as described in DeCoursey and Cherny, 1995):

(2)

where I₀ is the initial amplitude of the current after the voltage step, I is the steady-state current amplitude, t is the time after the voltage step, and t_delay is the delay. The H⁺ current amplitude is (I₀  I). No other time-dependent conductances were observed consistently under the ionic conditions employed. Tail current time constants, _tail, were fitted either to a single decaying exponential:

(3)

where I₀ is the amplitude of the decaying part of the tail current, or to the sum of two exponentials:

(4)

where A_n are amplitudes and _n are time constants.

Conventions

We refer to the pL in the format pL_o//pL_i. In the inside-out patch configuration the solution in the pipette sets pL_o, which is defined as the pL of the solution bathing the original extracellular surface of the membrane, and the bath solution is considered pL_i. Currents and voltages are presented in the normal sense, that is, upward currents represent current flowing outward through the membrane from the original intracellular surface, and potentials are expressed by defining as 0 mV the original bath solution. Current records are presented without correction for leak current or liquid junction potentials.

As discussed in detail in Strategic Considerations and in Fig. 1, when the bath solvent differs from that in the pipette, the effective pH_i (or pD_i) will differ from the nominal value of the pipette solution by ~0.5 U. Therefore, when bath and pipette solvents differ, we provide values for the presumed effective internal H⁺ or D⁺ concentration, e.g., pH_i,eff 6.5 indicates a pD 7.0 pipette solution with any H₂O solution in the bath. The majority of experiments were done with D₂O rather than with H₂O pipette solutions because we wanted the measurements in D₂O to be contaminated as little as possible by H₂O.

RESULTS

Strategic Considerations

The nature of the problem under investigation introduces several complications, which require explanation, as well as a perhaps less-than-obvious approach. Ideally we would like to compare the behavior of the proton conductance in the same cell under identical conditions while varying only the solvent (D₂O or H₂O) on one side of the membrane and keeping pL_o and pL_i constant (pL_x refers to either pH_x or pD_x). However, the high membrane permeability of water means that only symmetrical solvent studies can be contemplated. Less obviously, due to the increased pK_a of buffer in D₂O (Table II), it is impossible to compare directly in the same cell identical pH_o and pD_o by simply changing the external solvent, without at the same time changing pL_i. However, it is desirable to make comparisons in the same cell, because H⁺ currents vary substantially from cell to cell. We therefore adopted two strategies. First, we compare currents measured with the same pH or pD gradient (e.g., pH_o 6.5//pH_i 6.5 and pD_o 7.0// pD_i 7.0), because the gradient, pH, appears to be a fundamental determinant of H⁺ channel gating (Cherny et al., 1995). This approach has the drawback of comparing the effects of different absolute concentrations of protons and deuterons, and there is some indication that H⁺ channel gating kinetics depend on the absolute pH_i, rather than pH alone (DeCoursey and Cherny, 1995). The second approach (see MATERIALS AND METHODS) overcomes this shortcoming by controlling pH_i by applying a known NH₄⁺ gradient (Roos and Boron, 1981), as illustrated by Grinstein et al. (1994). Varying the NH₄⁺ gradient allows resetting pH_i (or pD_i) in a cell under whole-cell voltage-clamp, and ideally, comparison of currents at the same pH and pD.

In these experiments we varied the solvent in the pipette and bath solutions. Because water has a high membrane permeability, it seemed likely that the solvent in the bath solution would enter the cell much faster than solvent would diffuse from the pipette, and thus the solvent in the bath would also be present in the cell, regardless of the pipette solution. This expectation was tested theoretically and experimentally.

How fast does water enter the cell?

A critical question in the interpretation of the data is whether solvent in the bath diffuses across the cell membrane fast enough to dominate the intracellular solution in spite of the presence of the pipette tip which is a continuous source of solvent from the pipette solution. The water permeability, P_osm, of planar lipid bilayers or liposomes ranges from 10⁴ cm/s to 10² cm/s; P_osm in various epithelial cell membranes similarly ranges from 10⁴ cm/s to >10² cm/s (Tripathi and Boulpaep, 1989). Because both HgCl₂-sensitive and HgCl₂-insensitive water channels occur in lung tissue (Folkesson et al., 1994; Hasegawa et al., 1994), it is likely that P_osm is relatively high in alveolar epithelial cells, at least in situ. Osmotic water permeability (P_f) is 1.7 ± 10² cm/s and diffusional water permeability, P_d, is 1.3 ± 10⁵ cm/s across the alveoli of intact mouse lung (Carter et al., 1996). However, P_d was probably grossly underestimated because of unstirred layer effects (Finkelstein, 1984; Carter et al., 1996). We calculated the steady-state distribution of normal or heavy water when one species was in the pipette solution and the other in the bath solution. The compartmental diffusion model used has been described in detail previously (DeCoursey, 1995), and simplifies the calculation by placing the pipette tip at the center of a spherical cell. The diffusion coefficient of H₂O was taken as 2.1 × 10⁵ cm²/s (Robinson and Stokes, 1965), the pipette tip was assumed to have a diameter of 1.0 µm, the cell diameter was 20 µm, and we assume that D₂O and H₂O have similar membrane permeabilities (Perkins and Cafiso, 1986; Deamer, 1987; Gutknecht, 1987). A range of P_osm was assumed. For P_osm > 10³ cm/s the membrane presented essentially no barrier to diffusion, and the solvent in the bath was the main solvent inside the cell. Nevertheless, because the pipette is a constant source, there is always a finite concentration of the pipette solvent. For the pipette tip at the center of a 20 µm diameter cell, the limiting submembrane concentration at infinite P_osm is ~2% due to that in the pipette. Lowering P_osm to 10⁴ cm/s caused the membrane to become a significant diffusion barrier, with the steady-state concentration of solvent near the inside of the membrane 24% due to the pipette and 76% due to the bath. The fraction of solvent near the membrane originating in the pipette would be larger in a smaller cell but would be smaller if the pipette tip diameter were smaller. In conclusion, the pipette solvent is present in the cell at significant levels only for a quite conservative estimate of P_osm, and in all likelihood the solvent in the bath permeates the membrane rapidly enough that most of the solvent near the membrane originated in the bath. We therefore assume that the membrane is exposed to nearly symmetrical solvent, with a finite but small contribution from the pipette.

What is the pL (pH or pD) inside the cell?

The actual pL_i can be deduced from knowledge of pL_o and the reversal potential, V_rev. In the experiment illustrated in Fig. 1, the pipette contained pD 7.0 solution, and the tail current reversal potential, V_rev, was measured in several different bath solutions. V_rev was near 0 mV when the bath contained pD 7.0 (Fig. 1 A) or pH 6.5 (Fig. 1 C), and was 27 mV at pH_o 7.0 (Fig. 1 B). In eight cells, V_rev was 29.9 ± 4.5 mV (mean ± SD) more negative at pH_o 7.0 than at pD_o 7.0, both with pD_i 7.0. Reversal near 0 mV is expected for symmetrical pD 7.0//7.0. Why was V_rev near 0 mV at pH_o 6.5 but not at pH_o 7.0, under nominally symmetrical bi-ionic conditions? The explanation arises from the fact that many molecules bind D⁺ more tightly than H⁺. Most simple carboxylic and ammonium acids with pK_a between 4 and 10, including buffers, have a pK_a 0.5-0.6 U higher in D₂O than in H₂O (Schowen, 1977). We confirmed this generalization by titrating the buffers used in this study in both H₂O and D₂O and found pK_a shifts ranging 0.60- 0.69 U (Table II). Fig. 1 D illustrates diagrammatically the effect of this pK_a difference on a cell studied in the whole-cell configuration. The cell nominally contains the pipette solution with its buffer titrated to some pH or pD, in this example pD 7.0. If the solvent in the bath differs from that in the pipette, the bath solvent will replace the pipette solvent inside the cell, as discussed above. Because H⁺ has a lower affinity for buffer than does D⁺, fewer H⁺ will be bound to buffer than were D⁺, and hence the actual pH_i will be lower by ~0.5 U than was the pD of the pipette solution. This is true regardless of the actual value of pH_o, because it results from the solvent dependence of the pK_a of the buffer. The chart in Fig. 1 summarizes the experiment illustrated. Given the bath and pipette solutions, the observed V_rev agrees well with E_H calculated with the assumptions that (a) the solvent in the bath completely replaces that in the cell, and (b) the effective pH_i will be ~0.5 U lower than pD in the pipette when H₂O replaces D₂O in the bath. By similar logic, when H₂O is in the pipette solution and D₂O is in the bath, the actual pD_i will be ~0.5 U higher than pH_i with H₂O in the bath.

V_rev measurements are consistent with high water permeability and the 0.5 U pK_a correction for intracellular buffer in D₂O

We proposed above that the bath solvent will "fill" the cell regardless of the pipette solvent and that when the bath solvent differs from that in the pipette, pL_i will change by ~0.5 U from its nominal value. To a first approximation these assumptions seem reasonable, but two possible sources of error should be considered. First, some finite fraction of solvent in the cell is derived from the pipette. We could not determine from our data the extent of this "contamination." Second, we assume that the buffer pK_a increases exactly 0.5 U when D₂O replaces H₂O, although the true change may be slightly higher and may differ for different buffers. Our titration of several buffers used (Table II) revealed an average pK_a shift of 0.67 U in D₂O. To test the adequacy of our approximation of a 0.5 U shift, we compared the value for V_rev measured in the same cell in D₂O and in H₂O at 0.5 U lower pL_o. The difference in V_rev averaged 2.9 ± 0.7 mV (mean ± SEM, n = 21) for pD_o 6.0-pH_o 5.5, pD_o 7.0-pH_o 6.5, and pD_o 8.0-pH_o 7.5. We could not detect any significant difference between buffers in this respect. By this measure the actual pH_i may be ~0.05 U more acidic than our assumed value, i.e., pH_i may be 0.55 U lower than pD_i. However, considering that the slope of the V_rev vs. pH relationship in water was 52.4 mV (Cherny et al., 1995) compared with 58.2 mV for E_H, possibly indicating a ~10% attenuation of the pH applied across the membrane, one might suggest that the change in buffer pK_a should also be attenuated by 10% for internal consistency.

A complementary comparison can be made between V_rev measured in the same bath solution, but with H₂O or D₂O in the pipette solution. At pD_o 7, V_rev averaged +4.5 ± 1.2 mV (mean ± SEM, n = 4) with pH_i 6.5 and +4.3 ± 0.8 mV (n = 12) with pD_i 7. At pH_o 6.5, V_rev averaged +2.0 ± 1.6 mV (n = 4) with pH_i 6.5, and +0.5 ± 1.1 mV (n = 10) with pD_i 7. Thus, no systematic difference was observed in V_rev with D₂O or H₂O in the pipette. Together these data support the validity of the assumptions used to interpret these experiments.

Reversal Potential of D⁺ Currents

Values of V_rev obtained from tail current measurements, such as those illustrated in Fig. 1, A-C, in bilateral D₂O are plotted as a function of the pD gradient in Fig. 2. In most experiments, V_rev was slightly positive to the calculated Nernst potential for D⁺, E_D (dark line), reminiscent of the small positive deviations of V_rev from E_H reported in most studies of H⁺ currents. Most of the data points for each pD_i parallel E_D, clearly establishing the selectivity of this conductance for D⁺. The largest deviation occurred at pD_o 10//pD_i 8. Parallel experiments in H₂O solutions (not shown) produced a similar but more exaggerated resultV_rev followed E_H closely up to pH_o 8, with a smaller shift at pH_o 9, and no further shift at pH_o 10. The simplest interpretation of this result is that at high pH_o there is a loss of control over pH_i.

A more traditional but less attractive interpretation of the deviations of V_rev from E_D is that the selectivity of the conductance for D⁺ is not absolute, and that at high pL the permeability to some other ion (e.g., TMA⁺) is increased. However, the observed deviations are not consistent with a constant permeability of TMA⁺ relative to D⁺, because they were roughly the same at a given pD gradient, pD, at various absolute pD. Thus, the ratio P_TMA/P_D calculated using the GHK voltage equation was 2 × 10⁷, 2 × 10⁸, and 5 × 10⁹ at pD · 6, pD · 7, or pD · 8, respectively, all at pD = 2.0. Barring a bizarrely concentration-dependent permeability ratio, it appears that the conductance is extremely selective for D⁺ (or H⁺), with a relative permeability >10⁸ greater for D⁺ than for TMA⁺.

Behavior of the Proton Conductance in D₂O

After complete replacement of water with heavy water, D⁺ currents behaved qualitatively like H⁺ currents in normal water. Typical families of currents are illustrated in Fig. 3, with pD_i 6 and pD_o 8, 7, or 6. At relatively negative potentials only a small time-independent leak current was observed. During depolarizing pulses a slowly activating outward current appeared. The current has a sigmoid time course, and activation was faster at more positive potentials. Decreasing pD_o produced two distinct effects on the currents. The voltage at which the conductance was first activated, V_threshold, became more positive by about 40 mV/U decrease in pD_o, and the rate of current activation became slower. This shift in the position of the voltage-activation curve is more apparent in Fig. 4. The currents measured at the end of 8-s pulses are plotted (solid symbols), as well as the amplitude extrapolated from a single-exponential fit to the rising phase (open symbols). This latter value corrects for the fact that the currents did not always reach steady state by the end of the pulses, as well as correcting for any time-independent leak current. In this example, and in other experiments, the shift in the current-voltage relationship was very nearly 40 mV/U decrease in pD_o. These effects are quite similar to those of changes in pH_o in water (Cherny et al., 1995).

Another effect of changes in pD_o evident in Fig. 3 is that the conductance was activated more slowly at lower pD_o. The time course of activation of H⁺ or D⁺ currents was fitted by a single exponential after a delay (Eq. 2). In some cases the fit was good, as in the example shown in the inset to Fig. 5, but sometimes the time course was more complex, with fast and slow components. Deviations from an exponential time course seemed most pronounced at large positive voltages and when there was a large pD gradient. Activation time constants, _act, in the same cell at pD_o 8, 7, and 6 are plotted in Fig. 5. At each pD_{o act} is clearly voltage dependent, decreasing with depolarization. Lowering pD_o appears to shift the _act-V relationship to more positive potentials and upwards, slowing activation in addition to shifting the voltage dependence. Similar results were obtained in other cells. Although the magnitude of _act varied from cell to cell, the effects of changes in pD_o in each cell were quite similar to those illustrated.

The effects of pD_i on D⁺ currents were studied both in whole-cell experiments and in excised patches. Studying patches allows a direct comparison in the same membrane. Fig. 6 illustrates D⁺ currents in an inside-out patch at pD_o 8.0 and pD_i 6.0 (A) or pD_i 7.0 (B). In this and in several other patches V_threshold was shifted by about 40 mV/U decrease in pD_i. Time-dependent outward current first appeared at 40 mV at pD_i 6.0 and at 0 mV at pD_i 7.0. The small amplitude of most patch currents in D₂O limits the quantitative accuracy of any conclusions. However, the conductance approximately doubled when pH_i was reduced 1 U, comparable with the 1.7-fold increase/U decrease in pH_i reported previously in inside-out patches (DeCoursey and Cherny, 1995). It is also obvious that activation was much faster at lower pD_i.

The effects of changes in pD_i in whole-cell experiments were explored in individual cells by varying the NH₄⁺ gradient across the cell membrane (MATERIALS AND METHODS). Fig. 7 illustrates families of D⁺ currents in a cell at two NH₄⁺ gradients. In each case pD_o was 7.5, but pD_i decreased as the NH₄⁺ in the bath was lowered. With a 1//50 NH₄⁺ gradient (A) V_rev was 66 mV, and with a 15//50 NH₄⁺ gradient (B) V_rev was 27 mV. On the basis of this change in V_rev, pD_i was ~0.7 U lower in A than in B. At lower pD_i the currents activated more rapidly and the conductance appeared to be increased. Qualitatively similar effects of changes in pH_i were seen in H₂O solutions at various NH₄⁺ gradients in alveolar epithelium (not shown) and in macrophages (Grinstein et al., 1994).

Deuterium Isotope Effects on H⁺ (D⁺) Currents

Families of currents in the same cell in H₂O and D₂O are illustrated in Fig. 8. To keep pL approximately constant, we compared pH_o 6.5//pH_i,eff 6.5 and pD_o 7// pD_i 7 (Fig. 8, A and B, respectively). In D₂O the currents are smaller and activate more slowly.

The average ratios of the current measured in individual cells both in effectively symmetrical H₂O and symmetrical D₂O are plotted in Fig. 9. The "steady-state" current amplitudes were obtained by extrapolation of single exponential fits (Eq. 2). At all potentials the currents were substantially larger in H₂O. The ratio decreased at more positive potentials, but two sources of error would tend to cause a voltage-independent effect to deviate in this direction. First, during large depolarizations there is depletion of protonated (or deuterated) buffer from the cell, which tends to reduce the currents in a current-dependent manner. Because the currents were larger in H₂O, there would be more attenuation than in D₂O. Second, to the extent that the position of the voltage-activation curve may be shifted slightly positive in D₂O relative to H₂O (e.g., see Figs. 10 and 11), a smaller fraction of the total conductance would be activated in D₂O, and this would mainly affect smaller depolarizations to the steep part of the g_H-V relationship. Thus, it is not clear whether this effect was voltage dependent. The average ratio at +80 and +100 mV was 1.92 at pD 8 compared with pH 7.5, 1.91 at pD 7 compared with pH 6.5, and 1.65 at pD 6 compared with pH 5.5. In summary, the current carried by H⁺ through proton channels is about twice as large as that carried by D⁺.

In symmetrical D₂O the conductance-voltage relationship shifted about 40 mV/U change in pD just as in H₂O. However, the absolute voltage dependence might be different in the two solvents. To address this possibility we compared similar pH and pD in the same cell, varying the NH₄⁺ gradient to regulate pL_i. Fig. 10 illustrates a typical experiment. Measurements were made in D₂O (filled symbols) and in water (open symbols) at 1// 50 NH₄⁺ (), 3//50 mM NH₄⁺ (), and 15//50 mM NH₄⁺ (). At each NH₄⁺ gradient, the g_D-V relation was shifted 10-15 mV positive to the corresponding g_H-V relation. Moreover, V_rev was consistently more positive in D₂O at any given NH₄⁺ gradient. Apparently NH₄⁺ gradients were less effective at clamping pL_i in D₂O, perhaps reflecting the higher viscosity of D₂O (Table I), or the higher pK_a of NH₄⁺ in D₂O (Lewis and Schutz, 1934)at any given pL there would be a smaller concentration of neutral ND₃ than NH₃ available to permeate the membrane. The cytoplasmic acidifying power of 3 mM NH₄⁺ in D₂O might be roughly equivalent to that of 1 mM NH₄⁺ in H₂O, as was observed in the experiment illustrated in Fig. 10, if the neutral form were present at equal concentration, because the NH₄⁺ gradient changes pL_i in a dynamic manner through a sustained flux of neutral NH₃. Indeed, Grinstein et al. (1994) found that methylamine⁺, with a pK_a 10.19 compared with 9.24 for NH₄⁺ (Dean, 1985), acidified the cytoplasm more slowly given the same gradient than did NH₄⁺. If one assumes that V_rev accurately reflects pL_i then correcting for the difference in V_rev reduces the average shift in D₂O (compared with H₂O) to only ~5 mV. Scaling the D₂O data up to correct for the smaller limiting conductance further reduces the size of the shift. A residual shift of a few mV cannot be ruled out, but any such shift is not large, and it is possible that there is no shift.

Fig. 10 also shows that the conductance near threshold potentials changed e-fold in 4-5 mV at each NH₄⁺ gradient. We could not detect any difference in this limiting slope in D₂O and H₂O. Measured at 10² to 10³ of its maximal value, the conductance changed e  - fold in 4.65 ± 0.16 mV (mean ± SEM, n = 22) in D₂O and H₂O combined; the lines drawn through the data in Fig. 10 illustrate this average slope. This slope corresponds with the translocation of 5.4 charges across the membrane during gating, which should be considered a lower bound for the actual gating charge movement.

Finally, examination of the limiting maximum conductance at large depolarizations (Fig. 10) reveals that over the range of pL_i studied, the conductance was about twice as large in H₂O as in D₂O. This result is an important corroboration of the conclusion drawn from Figs. 8 and 9, because those comparisons were at ~0.5 U different absolute pL_i. The higher conductance in H₂O than in D₂O in Fig. 10 cannot be ascribed to different pL_i and must reflect a fundamental difference in the rate at which D⁺ and H⁺ permeate the channel.

The potential at which the H⁺ conductance is first activated by depolarization, V_threshold, is plotted in Fig. 11 as a function of V_rev in H₂O (open symbols) and in D₂O (filled symbols). Data obtained at pH_o 6.5-10.0 and pD_o 7-10 are included, as well as from experiments in which pL_i was changed by varying the NH₄⁺ gradient across the membrane. The data describe a remarkably linear relationship, with no suggestion of saturation at either extreme. The data for effectively symmetrical H₂O and D₂O fitted independently by linear regression yielded identical slopes (0.76 for H₂O and 0.75 for D₂O). Thomas (1988) observed a similarly linear relationship between E_H and V_rev in snail neurons, over a range of pH_i ~7-8. This result shows clearly that the fundamental determinant of the position of the voltage-activation curve of the g_H is the pH gradient across the membrane, as was concluded previously (Cherny et al., 1995).

The regression line in Fig. 11 for D₂O is shifted 3.9 mV from that for H₂O, indicating a more positive V_threshold for a given V_rev. This small shift may be an artifact resulting from the greater difficulty in detecting small currents in D₂O because the conductance is smaller and activation is slower. In any case, there was little or no solvent dependence of the relationship between V_rev and V_threshold, suggesting the position of the voltage-activation curve of the proton conductance is fixed in a very similar manner by pD as by pH.

The time-course of H⁺ or D⁺ current activation during depolarizing pulses was fitted by a single exponential after a delay to obtain _act, as was shown in the inset in Fig. 5. Mean values for _act at various pD (solid symbols) and pH (open symbols) are plotted in Fig. 12, all for pL = 0. It is unclear from these data whether there might be some effect of the absolute value of pL on _act. However, all the mean _act values in D₂O are slower at each potential than any of the values in H₂O. The average of the ratios at all potentials 60 mV of the mean _act values in D₂O to H₂O at 0.5 U lower pL_i was 3.21 at pD 8, 3.19 at pD 7, and 2.96 at pD 6. In summary, D₂O slows _act by about threefold.

Because there was substantial variability of _act from one cell to another, comparisons were also made in individual cells at effectively symmetrical pH or pD. The average ratio of _act in D₂O to that in H₂O plotted in Fig. 13 reveals that _act was 2.0-3.6 times slower in D₂O. The slowing was not noticeably voltage dependent. There is a suggestion that the slowing effect was greater at higher pD (or pH). If the ratios at all voltages in each solution are averaged, the slowing effect was 2.17 at pD 6 compared with pH 5.5, 3.06 at pD 7 compared with pH 6.5, and 3.21 at pD 8 compared with pH 7.5. The solid symbols include only cells studied with D₂O pipette solutions, the open squares show data from cells with H₂O in the pipette. The slowing of _act by D₂O appears to be attenuated in these cells, possibly reflecting the small amount of H₂O inside the cell, although the difference is not significant. In summary, D₂O slows _act about threefold, and this effect appears to be voltage independent.

The channel closing rate was examined by fitting the time course of the decay of tail currents (MATERIALS AND METHODS), such as those illustrated in Fig. 1, A-C. The average values of _tail obtained in effectively symmetrical solutions are plotted in Fig. 14. There is a suggestion in the data that _tail was slightly slower at higher pL, and in D₂O compared with H₂O. The average ratios at all potentials of the mean _tail data for essentially symmetrical pL are 1.31 (pD 8/pH 7.5), 1.04 (pH 7.5/pD 7), 1.23 (pD 7/pH 6.5), 1.05 (pH 6.5/pD 6), and 1.51 (pD 6/ pH 5.5). The apparent slowing by D₂O was thus 23- 51%, and some part of this effect may be ascribable to increasing pL_i.

In some cells _tail is independent of pH_o (DeCoursey and Cherny, 1996a; Cherny et al., 1997), but the effects of pH_i have not been clearly determined. Therefore, we attempted to compare _tail in H₂O and D₂O at similar pL_i in the same cell by varying the NH₄⁺ gradient. Increasing pH_i in individual cells at constant pH_o consistently slowed _tail by a small amount (not shown). When D₂O was compared with H₂O at a constant NH₄⁺ gradient, i.e., at nearly constant pL_i (see above), there was also a consistent slowing of _tail in nearly every cell, by roughly 50%, consistent with the average values given above.

Fig. 15 illustrates putative H⁺ currents in a cell-attached patch. The cell was bathed with isotonic KMeSO₃ solution to depolarize the membrane to near 0 mV. During depolarizations positive to 0 mV, there are slowly activating outward currents that resemble H⁺ currents (cf. DeCoursey and Cherny, 1995), as well as brief discrete openings of some other channel(s). When H₂O in the bath was replaced with D₂O, the outward currents became much smaller and appeared to activate even more slowly. This isotope effect is comparable to the effects seen in whole-cell configuration, but larger than reported for other ion channels (Table III). Therefore, we conclude that the slowly activating outward currents were in fact H⁺ currents.

Table III. Deuterium Isotope Effects on Other Channels (temperature, °C) [View Table]

The "leak" current at subthreshold voltages usually decreased when D₂O replaced H₂O. However, it appears extremely unlikely that the leak is carried primarily by H⁺ or D⁺. Attempts to calculate the H⁺ permeability, P_H,, of the leak current using the Goldman-Hodgkin-Katz (GHK) current equation (Goldman, 1943; Hodgkin and Katz, 1949):

(5)

where I_H and P_H are expressed normalized to membrane area estimated assuming that the specific capacitance is 1 µF/cm², revealed numerous inconsistencies with this idea. The slope conductance of leak currents (defined as time-independent currents at subthreshold potentials) rarely changed by more than twofold/U change in pH or pD, and not always in the same direction. For a large pL gradient (e.g., pD 8//6), leak currents at negative potentials but positive to E_L were inward, giving a negative calculated P_L. Calculated values for P_L decreased substantially at low pL_o, even when the observed leak slope conductance was increased. Finally, the apparent reversal potential of the leak current, which was not well defined because the leak currents were often small, was usually closer to 0 mV than to E _L, and did not always change in the "right" direction when pL_o was varied. In summary, there is no evidence that H⁺ carries a significant fraction of the leak current. An upper limit on the passive membrane permeability to H⁺ or D⁺ can be given as <<10⁴ cm/s at pH_i 5.5 or pD_i 6. By comparison, when the g_H is fully activated, P_H exceeds 1 cm/s at pH 8.0//7.5 (calculated from data in Cherny et al., 1995).

DISCUSSION

The deuterium isotope effects observed provide information about H⁺ permeation as well as the regulation of gating by protons (or deuterons). The main results are: (a) D⁺ permeates proton channels. (b) The relative permeability of proton channels is >10⁸ greater for D⁺ than for TMA⁺. (c) The H⁺ conductance through proton channels is ~1.9 times that of D⁺. (d) D⁺ regulates the voltage dependence of H⁺ channel gating much like H⁺. (e) The threshold for activating the proton conductance is a linear function of V_rev and changes 40 mV/U change in pH or pD. (f) D⁺ currents activate with depolarization ~3 times slower than H⁺ currents, but deactivation is at most 1.5-fold slower in D₂O. (g) At least 5.4 equivalent gating charges move across the membrane field during proton channel opening in D₂O and in H₂O. (h) The upper limit of any proton leak conductance of the membrane of rat alveolar epithelial cells must be <<10⁴ cm/s. When the g_H is fully activated, P_H exceeds 1 cm/s.

Properties of Proton Channels

At high pD, the D⁺ permeability was >10⁸ greater than the TMA⁺ permeability. The calculated permeability ratio P_TMA/P_D decreased as pD increased, by about 10-fold/U change in pD_i. Although a concentration dependent permeability ratio cannot be ruled out, it seems more reasonable to suppose that deviations of V_rev from E_D are due to imperfect control of pD, rather than to finite permeability of the channel to other ions. Several other H⁺ channels have been reported to have comparably high selectivity for H⁺, including the F₀ component of H⁺-ATPase (Althoff et al., 1989; Junge, 1989) and the M2 viral envelope protein (Chizhmakov et al., 1996).

The substantially lower conductance of proton channels in D₂O than in H₂O suggests that the charge-carrying species is H⁺ (or D⁺) rather than OH (or OD). The isotope effect for D⁺ is large because its mass is twice that of H⁺, but OD is only 6% heavier than OH, and thus a much smaller isotope effect is to be expected: 41% for D⁺ vs. 3% for OD for a classical square-root dependence on the mass of reactants (Glasstone et al., 1941). A similar argument can be made against H₃O⁺ which would have a predicted isotope effect of just 8% over D₃O⁺. However, the extremely high selectivity of the g_H has been ascribed to a Grotthuss-type or proton-wire permeation mechanism, which could exist for L⁺ or OL, but not L₃O⁺ (Nagle and Morowitz, 1978; DeCoursey and Cherny, 1994). Additional evidence supporting H⁺ rather than OH as the charge carrying species is that the g_H increases ~1.7-fold/U decrease in pH_i over the range pH_i 7.5-4.0 (DeCoursey and Cherny, 1995, 1996a), i.e., as [H⁺]_i increases a