A voltage-gated proton-selective channel lacking the pore domain

Abstract

Voltage changes across the cell membrane control the gating of many cation-selective ion channels. Conserved from bacteria to humans¹, the voltage-gated-ligand superfamily of ion channels are encoded as polypeptide chains of six transmembrane-spanning segments (S1–S6). S1–S4 functions as a self-contained voltage-sensing domain (VSD), in essence a positively charged lever that moves in response to voltage changes. The VSD ‘ligand’ transmits force via a linker to the S5–S6 pore domain ‘receptor’², thereby opening or closing the channel. The ascidian VSD protein Ci-VSP gates a phosphatase activity rather than a channel pore, indicating that VSDs function independently of ion channels³. Here we describe a mammalian VSD protein (H_V1) that lacks a discernible pore domain but is sufficient for expression of a voltage-sensitive proton-selective ion channel activity. H_v1 currents are activated at depolarizing voltages, sensitive to the transmembrane pH gradient, H⁺-selective, and Zn²⁺-sensitive. Mutagenesis of H_v1 identified three arginine residues in S4 that regulate channel gating and two histidine residues that are required for extracellular inhibition of H_v1 by Zn²⁺. H_v1 is expressed in immune tissues and manifests the characteristic properties of native proton conductances ($G_{{\text{vH}}^{+}}$). In phagocytic leukocytes⁴, $G_{{\text{vH}}^{+}}$ are required to support the oxidative burst that underlies microbial killing by the innate immune system^4,5. The data presented here identify H_v1 as a long-sought voltage-gated H⁺ channel and establish H_v1 as the founding member of a family of mammalian VSD proteins.

Main

H_v1 was identified as a novel conserved human gene using bioinformatic searches based on known cation channels. The H_v1 gene is located on chromosome 12 at q24.11 and is encoded by seven exons, the first of which is predicted to be non-coding. H_v1 messenger RNA is encoded by multiple independent human expressed sequence tags that contain in-frame stop codons preceding the initiator ATG and 3′ poly-A⁺ tails. The predicted H_v1 protein is 273 amino acids in length (31.7 kDa, pI = 6.62) and forms four transmembrane segments (S1–S4), according to hydropathy analyses (Supplementary Fig. S1a, b). The overall transmembrane (TM) structure and placement of charged residues in hydrophobic domains is conserved among the vertebrate H_v1 orthologues (Supplementary Fig. S1a) and similar to both Ci-VSP (ref. 3) and the VSD of K_v1.2 (ref. 2). The amino-terminal ∼100 amino acids in H_v1 are homologous to protein and lipid phosphatases, but unlike the catalytically active Ci-VSP, core active site residues required for phosphatase activity⁶ are not conserved in H_v1.

Both the distribution of expressed sequence tags (UniGene, see http://www.ncbi.nlm.nih.gov/UniGene/ESTProfileViewer.cgi?uglist=Hs.334637) and high-throughput expression studies (GNF SymAtlas, see http://symatlas.gnf.org/SymAtlas/; search on MGC15619 protein) indicate that H_v1 mRNA is enriched in immune tissues such as lymph node, B-lymphocytes, monocytes and spleen. We found marked lymph node expression of H_v1 mRNA on a human RNA dot blot (Supplementary Fig. S1d, spot F7). H_v1 protein expression was detected with an affinity-purified rabbit polyclonal anti-peptide antibody that recognizes H_v1, green flurorescent protein (GFP)-H_v1 and H_v1-haemagglutinin (HA) at their appropriate molecular weights in western blots of transfected HEK cells, but not mock-transfected controls (Supplementary Fig. S2b). The antibody recognized H_v1 in human immune tissues including lymph node (Supplementary Fig. S2a), and the human leukocyte cell lines HL-60 and Jurkat (Supplementary Fig. S2b, c).

In preliminary experiments, we observed large (> 1 nA at +100 mV) voltage-dependent, outwardly rectifying whole-cell currents in transfected HEK cells bathed in standard Ringer's, low-Cl^- (where all but 8 mM Cl^- was substituted with , SCN^- or Br^-), or NMDG⁺(N-methyl-d-glucamine)·Cl (Na⁺, K⁺, Ca²⁺ and Mg²⁺-free) solutions, suggesting that H_v1 currents were primarily permeable to H⁺. To study H⁺ currents under better pH control, we used solutions with high H⁺ buffering capacity (100 mM buffer used near its pK_a: MES, pH 5.5; Bis-Tris, pH 6.5; HEPES, pH 7.5) similar to those used previously to study native $G_{{\text{vH}}^{+}}$ (refs 7, 8). In response to depolarizing voltage steps, cells transfected with H_v1 generated time-, voltage-, and pH-dependent currents with relatively slow activation kinetics and a rapidly decaying inward tail current (Fig. 1a–c). The kinetics of H_v1 activation (τ_act = 715 ± 124 ms) and deactivation (τ_deact = 65.0 ± 12.1 ms) were estimated from single exponential fits to the currents at +80 mV or -80 mV, respectively. Both τ_act and τ_deact are strongly voltage-dependent (Fig. 1e) and τ_act was noticeably variable between experiments, perhaps owing to small fluctuations in room temperature and the highly temperature-sensitive gating of H_v1 (Supplementary Fig. S3c, d).

Figure 1: Biophysical properties of expressed H _v 1 currents.

Depolarizing voltage steps (5 mV increments) were applied to an H_v1-transfected HM1 cell (a–c) to elicit outward H⁺ currents and deactivating inward tail currents (- 80 mV). TMA5.5_o, TMA6.5_o or TMA7.5_o (TMA6.5_i for all) were used to impose pH gradients indicated in the diagrams. a, = 0.1, V_h = -40 mV (V_step = +30 mV to +80 mV, scale bar 0.2 nA, 1 s; b, = 1, V_h = -40 mV (V_step = -30 mV to +20 mV), scale bar 0.1 nA, 1 s; c, = 10, V_h = -70 mV (V_step = -60 mV to -10 mV), scale bar 0.2 nA, 1 s. d, H_v1 currents at the end of the depolarizing step (I_step, open symbols) and the absolute value of I_tail (- 80 mV, filled symbols) are plotted as a function of the step voltage. Data shown are from the cell shown in a–c. Circles, = 0.1 (V_thr = 70 ± 5.8 mV, n = 3); triangles, = 1.0 (V_thr = 11.7 ± 2.1 mV, n = 6); squares, = 10 (V_thr = -31.7 ± 3.3 mV, n = 3). e, Representative currents for E_rev measurement (V_h = -40 mV, V_step = +40 mV, V_tail = -100 mV to +40 mV). Small I_step (< 200 pA) was chosen to minimize H⁺_i depletion (note constant outward current level). f, Monoexponential fits of I_tail were extrapolated to t = 0 and E_rev was estimated from the zero-current intercept of linear fits to the data. Open circles, TMA6.5_i, TMA5.5_o ( = 0.1); filled triangles, TMA6.5_i, TMA6.5_o ( = 1); open squares, TMA6.5_i, TMA7.5_o ( = 10). Average E_rev values: E_rev = 53.3 ± 1.4 mV, = 0.1, n = 10; E_rev = 0.9 ± 0.7 mV, = 1.0, n = 14; E_rev = -56.5 ± 2.6 mV, = 10, n = 5. Data represent mean ± s.e.m. from n experiments. g, V_thr is plotted against the Nernst potential for H⁺ at 24 °C. The data are fitted to V_thr = 0.82E_rev + 13.8 mV (solid line). Data represent mean ± s.e.m. from n = 3–7 experiments. The V_thr versus E_rev relationship for native $G_{{\text{vH}}^{+}}$ (V_thr = 0.79 E_rev + 23 mV⁴, dotted line) is shown for comparison.

In the absence of a H⁺ gradient (the ratio of internal to external H⁺ concentration = 1.0), the voltage threshold (V_thr) for activation of outward currents (I_step) was 11 mV (Fig. 1d). An outwardly directed H⁺ gradient, equivalent to net intracellular acidification ( = 10), caused V_thr to shift 40 mV negative (Fig. 1d). The reversal potential (E_rev) of open H_v1 channels determined from I_tail (Fig. 1e) was ∼0 mV under symmetrical recording conditions ( = 1) and shifted >50 mV negative or positive in inward ( = 0.1) or outward ( = 10) H⁺ gradients, respectively (Fig. 1f). A plot of E_rev versus [H⁺] was well fitted to a line with slope = 54.7 mV/log[H⁺] (data not shown), indicating that the H_v1 current was highly selective for H⁺. A plot of V_thr versus the calculated Nernst potential for H⁺ was well fitted by the line V_thr = 0.82E_rev + 13.8 mV (Fig. 1g).

Although the outwardly rectifying I_step–V relation was linear at voltages ∼10–70 mV positive to E_rev (Fig. 1d, Fig. 2a, c), we were surprised to find that the H_v1 I_step–V deviated from linearity and appeared to saturate at voltages ∼100 mV positive to E_rev (Fig. 2a, c). We infer that the apparent nonlinearity in I_step results from depletion of H⁺_i caused by rapid channel-mediated H⁺ efflux that is not adequately compensated by delivery of H⁺_i from the buffer during the voltage step. Slow H_v1 kinetics near V_thr further hinder our ability to measure accurately the channels' voltage dependence because extraordinarily long voltage pulses are required to reach true steady state. Erosion of the H⁺ gradient during I_step and underestimation of I_step are both expected to flatten the I_tail–V relation. We were nonetheless able to fit H_v1 I_tail–V relations ( = 10) to a Boltzmann function in order to approximate the voltage-dependence of the H_v1 open channel probability (P_o, Fig. 2c). In an outward H⁺ gradient, H_v1 channels appear to be half open at ∼60 mV and to move the equivalent of about one electronic charge across the membrane during channel opening (zδ = 0.9). The midpoint of the I_step–V relation shifted >40 mV when the H⁺ gradient was altered, consistent with the observed shift in V_thr (Fig. 1).

Figure 2: H _v 1 voltage-dependent gating.

a, H_v1 currents (- 60 mV to +120 mV, V_h = -40 mV, = 10, Na6.5_i, Na7.5_o) in a representative cell. Scale bar 2 nA, 400 ms. Under symmetrical conditions (TMA6.5, = 1), τ_act = 715 ± 124 ms (+ 80 mV), n = 9 and τ_deact = 65.0 ± 12.1 ms (- 80 mV), n = 7. b, R205A (- 60 mV to +180 mV, V_h = -60 mV, = 10, Na6.5_i, Na7.5_o). Scale bar 2 nA, 10 ms. Note the 40-fold difference in timescale compared to a. c, The I_tail–V relation (- 80 mV, = 10, Na6.5_i, Na7.5_o) was normalized to the maximum current obtained from a Boltzmann fit to the data for each cell expressing H_v1 (filled circles) or R205A (open circles) to estimate P_o - V. Data were fitted to a Boltzmann function (solid lines) and normalized to the extrapolated maximum current. Points represent mean ± s.e.m. of normalized data. A comparison of curve fits from individual experiments (H_v1, V_0.5 = 58.0 ± 5.6 mV, zδ = 0.90 ± 0.04, n = 6; R205A, V_0.5 = 99.5 ± 18.3 mV, zδ = 0.57* ± 0.03, n = 3; *P = 0.03 by Student's non-paired t-test) indicates that significantly less effective charge is moved in R205A than in H_v1 during channel gating. Data represent mean ± s.e.m. from n experiments.

In an initial attempt to determine whether voltage-dependent gating in H_v1 involved basic residues in S4, as it does in other voltage-gated channels, we individually replaced each of three arginine residues with alanine. Representative traces from cells expressing H_v1 or the point mutant R205A are shown in Fig. 2a, b. Relative to H_v1, two arginine mutants (R205A, R208A) exhibited dramatically faster activation and deactivation kinetics, whereas R211A exhibited only faster deactivation kinetics (Supplementary Fig. S3e–g). R205A exhibited a positive shift in the half-maximal voltage for channel activation (V_0.5) and a statistically significant decrease in the slope factor (zδ) derived from Boltzmann fits to I_tail–V (Fig. 2c).

Like $G_{{\text{vH}}^{+}}$, H_V1 current was rapidly and reversibly blocked by bath superfusion of ZnCl₂ in the absence of EGTA (Fig. 3a, e). We therefore mutated putatively extracellular residues that might coordinate metal ions to inhibit H_v1 currents. Zn²⁺ potently inhibited I_Hv1 (half-maximal inhibitory concentration, IC₅₀ ≈ 2 µM) and slowed τ_act (Fig. 3a–c, data not shown). Apparent Zn²⁺ potency decreased 10- or 30-fold when we mutated H193 or H140 to alanine, respectively. Mutation of both residues (H140A/H193A) practically abolished Zn²⁺ sensitivity (Fig. 3d). All three histidine mutants exhibited the expected H⁺-selectivity and -dependent shift in V_thr that are characteristic of H_v1 (data not shown).

Figure 3: Mutations in H _v 1 reveal residues required for Zn ²⁺ inhibition.

Inhibition of I_step (+ 25 mV to +50 mV) in cells superfused with EGTA-free TMA6.5 or Na6.5 ( = 1) was typically faster than the sampling interval (10 s) and washout was always complete. Average IC₅₀ = 2.2 ± 0.6 µM, n_H = 1.0 ± 0.2, n = 4; = 1, + 40 mV, TMA6.5. Data represent mean ± s.e.m. from n experiments. The applied voltages and [Zn²⁺] for the records shown (Na6.5, = 1) were: a, H_v1, V_step = +40/V_tail = -80 mV, [Zn²⁺] = 0, 0.1, 1, 10, 100 µM; b, H140A, +40/ - 60 mV, [Zn²⁺] = 0, 1, 10, 100 µM; c, H193A, +30/ - 80 mV, [Zn²⁺] = 0, 10, 100 µM; d, H140A/H193A, +50/ - 20 mV, [Zn²⁺] = 0, 1 mM. Scale bars: 100 pA, 1 s. e, Representative Zn²⁺ concentration-response curves for H_v1 (filled squares, IC₅₀ = 1.9 µM, n_H = 0.9), H193A (open triangles, IC₅₀ = 17.9 µM, n_H = 1.2), H140A (filled circles, IC₅₀ = 74.3 µM, n_H = 1.0), and H140A/H193A (open squares, maximum inhibition = 11.1 ± 3.4%, n = 3).

The surprising finding of this study is that expression of a voltage-sensor domain, without its normally associated pore domain, is sufficient to reconstitute voltage-dependent proton channel activity. $G_{{\text{vH}}^{+}}$ have been recorded in a wide variety of cells from molluscs⁹ to mammals (reviewed in ref. 4). In mammalian phagocytes, $G_{{\text{vH}}^{+}}$ provides necessary charge compensation for the gp91^phox electron current that underlies Zn²⁺-sensitive superoxide anion production^4,5,10,11. Loss of functional gp91^phox in humans causes X-linked chronic granulomatous disease, which is characterized by decreases in microbial killing by phagocytes^12,13.

The properties of expressed H_v1 are consistent with those of $G_{{\text{vH}}^{+}}$ (refs 4, 14, 15). H_v1 is H⁺-selective, exhibits a -40 mV shift in V_thr when pH_i is decreased one unit, is blocked by micromolar Zn²⁺, and is highly temperature sensitive. H_v1 kinetics fall within the range reported for $G_{{\text{vH}}^{+}}$ in a variety of mammalian cells⁴. Kinetic heterogeneity among native $G_{{\text{vH}}^{+}}$ could reflect the existence of unidentified H_v1 homologues, temperature differences during recording, post-translational modification of H_v1, or second-messenger modulation of H⁺ channels. The small endogenous $G_{{\text{vH}}^{+}}$ reported in HEK¹⁶ and CHO¹⁷ (but not COS-7¹⁸) cells appeared neither to facilitate nor interfere with expression of H_v1 current in this study. Slow activation kinetics and large H⁺ currents in H_v1-transfected cells will tend to exacerbate difficulties in measuring steady-state channel behaviour and in maintaining the imposed H⁺ gradient during strongly depolarizing voltage pulses, and could explain why the voltage dependence of expressed H_v1 channels appears shallow relative to native currents⁴.

An accepted model for voltage-gated channels is one in which positively charged residues in S4 function as the primary voltage sensors^19,20. Charge-neutralizing mutagenesis of S4 arginine residues (R205A, R208A, R211A) demonstrates that these basic residues are intimately involved in H_v1 gating and may directly sense transmembrane voltage. Voltage-dependent gating in H_v1 is distinguished from other cation channels by its sensitivity to a H⁺ gradient, indicating that further studies are required to clarify whether the mechanism of voltage sensing in H_v1 is similar to that in other voltage-gated channels. Starace and Bezanilla²¹ replaced an S4 arginine in Shaker with histidine (R362H) to create a H⁺-permeable channel at negative membrane potentials. In H_v1, neutralizing S4 arginines or histidines in the extracellular linkers did not prevent the H⁺ conductance, indicating that H⁺ permeation does not require proton acceptors at these positions. Mutagenesis of putative H⁺ acceptors may be of limited use in defining the conduction pathway given that H⁺ conductance could be indirectly attenuated owing to allosteric effects. The mechanism of proton permeation may involve a voltage-dependent movement of S4 to align residues that create a ‘proton wire’, as in gramicidin²², or exposure of an H⁺ acceptor that bridges two water-filled crevices²¹. The latter model unites both pore and wire features in that the proton acceptor can also be thought of as the binding site for H⁺ within a pore that confers H⁺ selectivity.

Polyvalent cations, particularly Zn²⁺, are the only known blockers of native $G_{{\text{vH}}^{+}}$ (refs 4, 15). Zn²⁺ block of native proton current was predicted to involve two extracellularly accessible amino acids; apparent Zn²⁺ potency is modulated by both [H⁺]_o and membrane voltage¹⁵. Strikingly, we identified two histidine residues that appear to function coordinately to confer Zn²⁺ inhibition to H_v1. In addition, Zn²⁺ slows τ_act, mirroring its effect on $G_{{\text{vH}}^{+}}$ (ref. 15) and further supporting the conclusion that H_v1 activity underlies native $G_{{\text{vH}}^{+}}$.

Expression of H_v1 alone is sufficient to reconstitute most properties of $G_{{\text{vH}}^{+}}$, making it unlikely that $G_{{\text{vH}}^{+}}$ is mediated by gp91^phox (refs 23, 24). Future studies are required to ascertain whether H_v1 physically associates with or is regulated by components of the NADPH oxidase complex, protein kinase C, or arachidonic acid^25,26,27, whether H_v1 channel activity is sufficient to control O₂^- secretion and reactive oxygen species production^28,29, and what factors account for differences between ‘phagocyte’ and ‘epithelial’ type $G_{{\text{vH}}^{+}}$ (ref. 4). Finally, future structural and biophysical studies should help to elucidate the proton permeation pathway within this voltage sensor domain protein.

Note added in proof: Okamura and colleagues also report that the mouse Hv1 gene encodes a proton channel, which they call VSOP (voltage-sensor only protein)³⁰.

Methods

H_v1 complementary DNA was amplified by polymerase chain reaction (PCR) from an expressed sequence tag clone (IMAGE 6424182) and subcloned into pQBI25-fC3 (QBiogene) to create GFP–H_v1, pcDNA3.1(- ) for expression of non-tagged H_v1, or pBSIISK - (Stratagene) for in vitro mRNA transcription. Site-directed mutagenesis was used to create point mutations in GFP–H_v1: H140A, H193A, R205A, R208A, R211A, and the H140A/H193A double mutant; H_v1-HA was used for biochemical studies. A ∼300 nucleotide antisense RNA probe labelled with α-³³P-UTP was synthesized by in vitro transcription (Ambion) and hybridized (68 °C) to a human multiple tissue mRNA panel (MTE3, Clontech). Incorporated radioactivity was detected by a phosphorimager (Molecular Dynamics).

Currents reported here were recorded 18–36 h after transfection of GFP–H_v1 cDNA in the HEK-293 cell clone HM1. Recording solutions were similar to those described previously¹⁸ and contained 100 mM pH buffer (MES, Bis-Tris or HEPES) near its pK_a (5.5, 6.5 or 7.5, respectively) in tetramethylammonium methanesulphonate or NaCl adjusted to ∼300 mOsm. H⁺ gradients ( = [H⁺]_{intracellular}/[H⁺]_{extracellular}) were imposed by gravity-fed bath superfusion of differentially buffered solutions. Currents were recorded with an Axopatch 200A (Axon Instruments; 1 kHz filter) and digitized at 2 kHz (10 kHz for Arg mutants) using Clampex9 (Axon Instruments). Data were analysed using Clampfit9 (Axon Instruments) and Origin 6 (Microcal). Recordings were performed at 22–24 °C unless otherwise stated. Data shown and stated in text and legends represent mean ± s.e.m. for the indicated number of experiments.

Affinity-purified 4234 polyclonal rabbit antibodies (Chemicon) were raised against the keyhole limpet haemocyanin (KLH)-conjugated H_v1 peptide CSEKEQEIERLNKL. Transfected cells were lysed in 1% Triton X-100 and crude membranes were subjected to SDS–polyacrylamide gel electrophoresis (PAGE) (NuPage, Invitrogen). Following transfer to nitrocellulose, blots were probed with the indicated concentration of 4234 followed by an anti-rabbit HRP-conjugated secondary antibody (Zymed). Immunoreactivity was detected by enhanced chemiluminescence (Pierce). A human immune tissue western blot (1525, ProSci) was probed with 4234 as described.