Zinc inhibition of monomeric and dimeric proton channels suggests cooperative gating

Abstract

Voltage-gated proton channels are strongly inhibited by Zn²⁺, which binds to His residues. However, in a molecular model, the two externally accessible His are too far apart to coordinate Zn²⁺. We hypothesize that high-affinity Zn²⁺ binding occurs at the dimer interface between pairs of His residues from both monomers. Consistent with this idea, Zn²⁺ effects were weaker in monomeric channels. Mutation of His¹⁹³ and His¹⁴⁰ in various combinations and in tandem dimers revealed that channel opening was slowed by Zn²⁺ only when at least one His was present in each monomer, suggesting that in wild-type (WT) H_V1, Zn²⁺ binding between His of both monomers inhibits channel opening. In addition, monomeric channels opened exponentially, and dimeric channels opened sigmoidally. Monomeric channel gating had weaker temperature dependence than dimeric channels. Finally, monomeric channels opened 6.6 times faster than dimeric channels. Together, these observations suggest that in the proton channel dimer, the two monomers are closely apposed and interact during a cooperative gating process. Zn²⁺ appears to slow opening by preventing movement of the monomers relative to each other that is prerequisite to opening. These data also suggest that the association of the monomers is tenuous and allows substantial freedom of movement. The data support the idea that native proton channels are dimeric. Finally, the idea that monomer–dimer interconversion occurs during activation of phagocytes appears to be ruled out.

Introduction

Human (H_V1) and mouse (mVSOP) voltage-gated proton channels in heterologous expression systems appear to function as dimers, with a conduction pathway in each monomer (Koch et al. 2008; Lee et al. 2008; Tombola et al. 2008). Very recent studies suggest that the monomers interact during gating, although the details of this interaction remain obscure (Gonzalez et al. 2010; Tombola et al. 2010). It is not certain that the native channel is a dimer, because dimerization has been demonstrated only in heterologous expression systems, and the gating of expressed and native proton channels differs significantly (Musset et al. 2008). Here we compare certain properties of monomeric and dimeric proton channels in an attempt to gain insight into possible interactions. We also focus on Zn²⁺ binding sites in the channel, which provide clues about the possible dimer interface.

The most potent inhibitor of voltage-gated proton channels is Zn²⁺ (Mahaut-Smith, 1989). The effects of Zn²⁺ are profoundly attenuated at low pH_o (Cherny & DeCoursey, 1999). This strong competition between H⁺ and Zn²⁺ for an external metal binding site in native proton channels could not be explained by simple 1:1 binding. Instead, the data were well described by models that assumed that the Zn²⁺ atom was coordinated between two to three titratable groups with pK_a 6–7, suggestive of His residues (Cherny & DeCoursey, 1999). When the human proton channel gene was identified, the H_V1 protein displayed two His residues accessible to the extracellular solution (Ramsey et al. 2006). Mutating either of these His to Ala (H140A, His¹⁴⁰ replaced by Ala, or H193A, His¹⁹³ replaced by Ala) reduced Zn²⁺ inhibition, and the double mutant had little Zn²⁺ sensitivity (Ramsey et al. 2006). One might conclude that Zn²⁺ inhibits proton current by binding simultaneously to His¹⁴⁰ and His¹⁹³. However, examination of a homology model of the proton channel structure (Fig. 1A) reveals a striking problem, namely that the two His residues are 14 Å apart – too far to coordinate a Zn²⁺ atom by overlap of their individual electron orbitals. Analysis of Zn²⁺ binding sites in 111 proteins revealed distances of 1.9–2.4 Å between Zn²⁺ and its coordinating atom (Alberts et al. 1998). Of course, Zn²⁺ binding to the proton channel is rapidly reversible, and a formal binding site is not expected. No plausible alternative sequence alignment, coordinate reassignment or torsioning of His sidechains resulted in a significantly shorter distance between His residues within the monomer.

Figure 1. Homology models of the H_V1 proton channel as a monomer (A) or a dimer (B), showing the locations of the two His residues exposed to the extracellular solution.

In A, the extracellular surface of the membrane is indicated by the upper line, the intracellular surface by the lower line. His¹⁴⁰ and His¹⁹³ are shown in aqua. In the monomer, His¹⁴⁰ and His¹⁹³ are 14 Å apart, too far to plausibly coordinate a Zn²⁺ atom. In the dimer (B), the His¹⁴⁰ from each monomer are closer together, as are the His¹⁹³; either or both could comprise high-affinity Zn²⁺ binding sites. Thus, a high-affinity Zn²⁺ binding site exists only in the dimer. The recent identification of proton channel genes allows homology-based structural predictions, because the proton channel molecule bears striking homology to the voltage-sensing domain of K⁺ and other voltage-gated ion channels (Ramsey et al. 2006; Sasaki et al. 2006). H_V1 contains four transmembrane domains resembling S1–S4 of other channels, but lacks the S5–S6 regions that comprise the ion conduction pathway in other channels.

This dilemma can be resolved by postulating that Zn²⁺ binds at the interface between the two monomers (Fig. 1B) to complementary His resides from each monomer. Binding between residues in different subunits can occur because the corresponding His residues approach more closely than is possible in a monomer. Thus, in the dimer model shown in Fig. 1B, the His¹⁹³ of the two monomers are 5.7 Å apart and the His¹⁴⁰ are 9.4 Å apart. In the His–Zn–His complex, the Zn–His distance would be somewhat larger than half of these distances, depending on the coordination angle (Alberts et al. 1998). Thus, both pairs of His residues form potential Zn²⁺ binding sites, particularly if Zn²⁺ binding pulls the two His towards each other. Empirically, both His residues contribute to Zn²⁺ effects, because both single mutants exhibited reduced, but still significant Zn²⁺ sensitivity (Ramsey et al. 2006). An alternate arrangement of the monomers in the dimer results in two identical Zn²⁺ binding sites, each comprising His¹⁴⁰ and His¹⁹³, one from each monomer (not shown). To obtain information about the Zn²⁺ binding site(s) on the human H_V1 proton channel, we generated mutants in which His¹⁴⁰ and His¹⁹³ were modified singly or together, as well as tandem dimers with various combinations of mutations.

One way to test the novel idea that Zn²⁺ binds simultaneously to His residues in the two individual monomers is to express the channel in monomeric form. The Zn²⁺ sensitivity of the monomeric channel should be identical to that of the dimer if Zn²⁺ were coordinated between the His¹⁴⁰ and His¹⁹³ within each monomer. However, if Zn²⁺ is preferentially coordinated between two His residues, one on each monomer, then the monomeric channel should have reduced Zn²⁺ sensitivity. To study monomeric human proton channels, we used a construct of H_V1 that terminated at Lys²²¹ (H_V1ΔC), and thus lacked the intracellular C-terminus. Coiled-coil interactions in the C-terminus stabilize the dimeric form of the channel (Lee et al. 2008; Tombola et al. 2008). To generate monomeric mouse proton channels, we truncated both C- and N-termini of mVSOP (Koch et al. 2008), to produce mVSOPΔCΔN. These truncations result in monomeric, but fully functional, voltage-gated proton channels, confirming that each monomer contains its own conduction pathway (Koch et al. 2008; Tombola et al. 2008). Since Zn²⁺ effects in monomeric channels were weaker, we pursued the idea that Zn²⁺ may prevent opening by constraining movement of the monomers relative to each other. A comparison of the kinetics and temperature dependence of gating of monomers and dimers, suggests that channel opening is more complex in the dimer.

Methods

Electrophysiology

In some studies, green fluorescent protein (GFP)-tagged proton channels were studied. More often, GFP was co-transfected with the proton channel construct. Fluorescent cells were identified in the field using Nikon inverted microscopes with fluorescence capability. Micropipettes were pulled using a Flaming Brown automatic pipette puller (Sutter Instruments, San Rafael, CA, USA) from 7052 glass (Garner Glass Co., Claremont, CA, USA), coated with Sylgard 184 (Dow Corning Corp., Midland, MI, USA), and heat polished to a tip resistance ranging typically from 5 to 15 MΩ with the pipette solutions used. Electrical contact with the pipette solution was achieved by a thin sintered Ag–AgCl pellet (In Vivo Metric Systems, Healdsburg, CA, USA) attached to a Teflon-encased silver wire, or simply a chlorided silver wire. A reference electrode made from a Ag–AgCl pellet was connected to the bath through an agar bridge made with Ringer solution. The current signal from the patch clamp (EPC-9 from HEKA Elektronik, Lambrecht/Pfalz, Germany, or Axopatch 200B from Axon Instruments, Foster City, CA, USA) was recorded and analysed using Lab View, SCB-68 (National Instruments, Austin, TX, USA), Pulse and PulseFit (HEKA), or pCLAMP (Molecular Devices, Sunnyvale, CA, USA) software supplemented by Microsoft Excel, Origin 7, and Sigmaplot (SPSS Inc., Chicago, IL, USA). Seals were formed with Ringer solution (in mm: 160 NaCl, 4.5 KCl, 2 CaCl₂, 1 MgCl₂, 5 Hepes, pH 7.4) in the bath, and the potential zeroed after the pipette was in contact with the cell. No liquid junction potential correction was applied.

Inside-out patches were made by forming a seal and then lifting the pipette into the air briefly. For whole-cell recording, bath and pipette solutions contained 100–200 mm buffer, 1–2 mm CaCl₂ or MgCl₂ (pipette solutions were Ca²⁺ free), 1–2 mm EGTA, and tetramethylammonium methanesulfonate or N-methyl-d-glucamine to adjust the osmolality to roughly 300 mOsm, titrated with tetramethylammonium hydroxide or methanesulfonic acid. Buffers used at various pH values were Mes at pH 5.5–6.0, BisTris at pH 6.5, Pipes at pH 7.0 and Hepes at pH 7.5–8.0. We omitted EGTA from solutions for Zn²⁺ measurements. No leak correction has been applied to any current records. Except where noted, measurements were done at 21°C or at room temperature (20–25°C). The bath temperature was controlled by a system from Brooks Industries (Lake Villa, IL, USA), or by an in-house built system using Peltier devices in a feedback arrangement, monitored by a resistance temperature detector element (Omega Scientific, Stamford, CT, USA) immersed in the bath. Temperature changes were transmitted to the glass recording chamber through a supporting copper plate. The temperature probe was positioned as near the pipette tip as possible.

Currents were fitted to a rising exponential to obtain the activation time constant (τ_act) and the proton conductance (g_H), which was calculated from the steady-state current (the fitted current extrapolated to infinite time) using reversal potentials (V_rev) measured in each solution in each cell. In these fits, we ignored the initial delay in WT activation; the remaining current usually fitted a single exponential well (Fig. 3). The reversal potential (V_rev) was measured by two methods. When V_rev was negative to the threshold voltage at which the proton conductance (g_H) was first activated, V_threshold, V_rev was determined by the tail current method (Hodgkin & Huxley, 1952). If V_rev was within the range of active proton conductance, it was determined by interpolation between time-dependent inward or outward currents during test pulses (after scaling according to the tail current amplitude). The magnitude of the shift of the g_H–V relationship was determined by plotting g_H values semilogarithmically against voltage (e.g. Fig. 8), and then shifting the g_H–V relationships along the voltage axis until they superimposed upon the control curve.

Figure 3. WT dimeric proton channels open slowly with a sigmoidal time course (A), but monomeric H_V1ΔC channels open rapidly and exponentially (B).

Individual current records are shown at +50 mV from a holding potential of −40 mV in inside-out patches at pH_o 7.5, pH_i 7.5 at 23°C. Single exponential fits are superimposed and extrapolated on the data points shown as circles. For the records shown here, τ_act was 3.0 s and 0.59 s for WT and H_V1ΔC, respectively.

Figure 8. Greatly reduced Zn²⁺ sensitivity of the WT–H140A/H193A tandem dimer, compared with the WT–WT tandem dimer.

All current families in the WT–WT tandem dimer (A) or the WT–H140A/H193A tandem dimer (B) were recorded during pulses to the voltages shown in the inset at pH_o 7.0, pH_i 6.5. C and D, the g_H–V relationships were determined from the cells in A and B using reversal potentials measured in each cell.

With overexpression of channels in small cells, large proton currents remove enough protons from the cell to increase pH_i substantially (DeCoursey, 1991; Kapus et al. 1993; Demaurex et al. 1993; DeCoursey & Cherny, 1994, 1995, 1998; Musset et al. 2008; Kuno et al. 2009). As proton channel gating kinetics depends strongly on pH, proton depletion is a significant source of error. To minimize this problem, we applied pulses of different length to encompass different voltage ranges (Fig. 2). Longer pulses gave information about small depolarizations, and shorter pulses were used at large depolarizations, where activation was faster. Values for τ_act obtained in one cell for pulses of different lengths are plotted in the inset of Fig. 2. It is evident that despite a long interval between pulses, depletion progresses during families of pulses. Since cells survive a finite time and it is desirable to acquire data at multiple Zn²⁺ concentrations, some compromise is required between full pH_i recovery between pulses and ideal data. We applied infrequently repeated small test pulses after each family to determine when pH_i recovery was complete. Experiments were conducted with the intent of minimizing depletion, but also with the expectation that any systematic errors would apply equally to all constructs studied.

Figure 2. Use of different pulse lengths to evaluate τ_act at different voltage ranges, to minimize distortion of kinetics by proton depletion.

Superimposed are four pulse families of different durations in the same cell, transfected with the H193A–H140A tandem dimer at pH_o 7.0, pH_i 6.5 with 10 μm Zn²⁺. Pulses were applied in increasingly positive steps in 10 mV increments from a holding potential of −60 mV up to +20 mV (8 s pulses, ○), +40 mV (4 s, ▪), +70 mV (2 s, ▵) or +90 mV . Pulses were applied at 22 s intervals except for a 38 s interval for the 8 s family. The order of families was 4 s, 8 s, 2 s and 1 s. The inset shows τ_act estimated from these records.

The slowing of τ_act by Zn²⁺ was evaluated over a wide voltage range, and each value was an average of the ratio at several voltages. The slowing by Zn²⁺ is roughly independent of voltage (Cherny & DeCoursey, 1999). We gave preference to τ_act from longer pulses, when data with different length pulses were available. As the τ_actvs. voltage relationship is often non-monotonic, with smaller values near V_threshold that first increase and then decrease with depolarization (Musset et al. 2008), as evident in Fig. 2, we compared values positive to the maximum τ_act value.

Except where noted, statistical comparisons used Student's unpaired t test.

Culture, mutation and transfection

HEK-293 cells were maintained in 5% CO₂ and 95% air in a humidified incubator at 37°C in Dulbecco's modified Eagle's medium containing 10% heat-inactivated fetal bovine serum, 2 mm l-glutamine, 100 U ml⁻¹ penicillin, 250 ng ml⁻¹ Fungizone and 100 μg ml⁻¹ streptomycin. The coding sequence of human H_V1 (Hvcn1) was cloned into either pcDNA3.1(–) or pQBI25-fC3 (to make GFP–H_V1) vectors as described previously (Ramsey et al. 2006). The mouse orthologue (mVSOP) was derived from RIKEN cDNA 0610039P13 as described previously (Sasaki et al. 2006). HEK-293 cells were grown to ∼80% confluency in 35 mm cultures dishes, usually by seeding cells 1 day ahead of transfection. Cells were transfected in media without fetal bovine serum with ∼0.7 μg of the appropriate cDNA using Lipofectamine 2000 (Invitrogen). After 6 h incubation at 37°C in 5% CO₂, the medium was replaced with medium containing fetal bovine serum. The next day, the cells were trypsinized and re-plated onto glass coverslips at low density for patch clamp recording. We selected green cells under fluorescence for recording.

The C-terminus of mouse mVSOP was truncated by introducing a stop codon at K217 (Quikchange site directed mutagenesis kit, Stratagene, San Diego, CA, USA). For N- and C-terminal truncations, mVSOP was amplified with primers containing a start codon at T77 and a stop codon at K217, and the PCR product was cloned in pCDNA 3.1(+). Single mutations, H140A and H193A, were generated by site-directed mutagenesis. The WT–WT, H193A–H140A and WT–H140A/H193A tandem constructs were made by removing the stop codon in the respective constructs in pCDNA and introducing a Nhe1 site and three alanines forming a six amino acid linker, ASGAAA, between the two proteins.

Homology model

A homology model of the voltage-sensing domain (VSD) of human H_V1 (NCBI NP_115745.2) was created using Insight II (Accelrys, San Diego, CA, USA). The basis set consisted of the isolated voltage sensor of KvAP (1ORS) and the voltage sensor of Kv1.2 (2R9R) (Jiang et al. 2003a; Long et al. 2007). Basis set crystal structures were structurally aligned, and the sequence of the human H_V1 VSD was aligned to a structure-based sequence alignment in BioEdit (Ibis Biosciences). Direct coordinate assignment was performed on structurally conserved regions. Loop assignments were performed via loop searches using a local subset of the Protein Data Bank, and using the Archpred server (Fernandez-Fuentes et al. 2006). Energy minimization routines were run to convergence on the splice points, the loops and the side-chains of the structurally conserved regions. In the final model, Procheck (Laskowski et al. 1993) revealed very few unrealistic bond lengths and bond angles, all of which resulted from crystal-based assignments. The general features of this model, including the distance between His¹⁴⁰ and His¹⁹³, were supported by an independent model generated using the automated I-TASSER method (Zhang, 2008).

Monomer models were visualized in Insight and in MolSoft ICM Pro (MolSoft), and the two monomers were moved manually (‘visual docking’), to verify the plausibility of Zn²⁺ binding between monomers. Further support for plausible Zn²⁺-binding dimer configurations was sought by submitting the model to the GrammX protein–protein docking server (Tovchigrechko & Vakser, 2006), using the homodimer option and with varying numbers of residues suggested as participating in the interface. Results of docking runs were initially visualized and inspected in VMD (visual molecular dynamics) (Humphrey et al. 1996). Dimer models retained from initial inspections were superimposed on a model of the 2R9R crystal structure in the plasma membrane obtained from the Orientations of Proteins in Membranes database (Lomize et al. 2006) to assess their ability to span the membrane appropriately. Approximately 1% of the dimers returned from the docking runs were in realistic orientations. Figure 1B is a representative example from the realistic dimers that were identified, all of which had His–His distances of 4–10 Å.

Results

Activation kinetics in monomeric and dimeric proton channels

The activation kinetics of monomeric and dimeric proton channel currents differs in two respects. First, WT dimeric proton channels activate with a sigmoidal time course (Fig. 3A), like all native proton currents (DeCoursey, 2003). In contrast, monomeric H_V1ΔC proton currents activate with an exponential time course (Fig. 3B). We fitted all currents to a single rising exponential to obtain the time constant of activation, τ_act. For dimeric channel currents, we ignored the initial delay. Exponential opening is consistent with a simple first order opening transition in the monomer. Sigmoidal kinetics can be explained by a cooperative gating process in which multiple subunits must undergo an opening transition before conduction can occur (Hodgkin & Huxley, 1952). As the proton channel is a dimer in which each monomer contains its own conduction pathway (Koch et al. 2008; Tombola et al. 2008; Lee et al. 2008), such cooperativity is surprising.

Consistent with previous reports (Koch et al. 2008; Tombola et al. 2008), channel opening in H_V1ΔC or mVSOPΔNΔC monomers was much faster than in WT dimers (Fig. 3, and not shown). Inside-out patches were used for this measurement, because proton depletion notoriously can distort gating kinetics in whole-cell studies (DeCoursey, 1991; Kapus et al. 1993; Demaurex et al. 1993; DeCoursey & Cherny, 1994, 1995, 1998; Musset et al. 2008; Kuno et al. 2009). The average ratio of τ_act in H_V1/H_V1ΔC measured (at each voltage) in inside-out patches at pH_o 7.5, pH_i 7.5 at +30 to +80 mV in 10 mV increments was 6.58 ± 0.44 (mean ±s.e.m. for 6 voltages, each with 4–7 cells for H_V1 and 5–7 cells for H_V1ΔC). A 5-fold slower time-to-half-peak was reported for mVSOP than mVSOPΔNΔC (Koch et al. 2008). Faster activation of monomeric channels suggests that interaction between monomers in the dimer slows opening or that the presence of cytoplasmic domains slows gating (Koch et al. 2008). Proton channels in monomeric form obligatorily gate independently; apparently a qualitatively different gating process occurs in the dimer.

Temperature dependence is weaker in the monomer

A complex gating mechanism for WT dimeric proton channels is consistent with the extreme temperature dependence of gating of native proton channels (Q₁₀= 6–9, or 30–38 kcal mol⁻¹ for both opening and closing kinetics) (DeCoursey & Cherny, 1998). If part of the energetic barrier arises from inter-subunit interaction, the temperature dependence of gating in the monomeric construct might be weaker. Figure 4 illustrates that this was the case for the human proton channel: gating kinetics in the H_V1ΔC monomer had temperature dependence only about half that of wild type (WT) H_V1. Families of WT (A and B) and H_V1ΔC (C and D) proton currents are shown at roughly 10°C different temperatures. The time scales differ by a factor of 8 for WT and 3 for H_V1ΔC, illustrating the much stronger temperature dependence of channel opening kinetics in the dimer. We used excised, inside-out patches of membrane for these measurements to avoid complications due to proton depletion during large currents (DeCoursey & Cherny, 1998; Kuno et al. 2009). Estimates were based on several currents from families of pulses recorded at two or more temperatures, with the assumptions that within the error of the estimates, Q₁₀ for gating kinetics is independent of voltage and pH (DeCoursey & Cherny, 1998; Kuno et al. 2009). On average, the Q₁₀ of channel opening (τ_act) was 7.16 for WT and 3.57 for H_V1ΔC (Fig. 4E). The Q₁₀ for channel closing (τ_tail, tail current time constant) was 7.46 for WT and 2.33 for H_V1ΔC. Measurements were made over a pH range of 5.5–7.5 and within the temperature range 9–32°C. Thus, channel opening and closing are both substantially more energetically demanding in WT dimeric channels, suggesting a different gating mechanism.

Figure 4. Temperature dependence of WT (A and B) and H_V1ΔC proton currents (C and D) recorded in inside-out patches of membrane.

For each patch, families of pulses to the same voltages were applied, with the pulse duration adjusted as shown. A, WT proton currents during pulses from −60 mV to +20 mV in 10 mV increments at 20.8°C (A) and 31.2°C (B), at pH_o 7.5, pH_i 6.5. C, H_V1ΔC proton currents in a patch elicited by families of pulses from −40 mV to +80 mV in 10 mV increments, at 11.9°C. D, pulses to the same voltages in the same patch at 23.3°C. Both at pH_o 7.5, pH_i 7.5. E, summary of the temperature dependence of WT and H_V1ΔC proton channel gating kinetics and conductance. Measurements were made in excised inside-out patches to avoid depletion effects. Means ±s.e.m. are shown for 6–8 WT (H_V1) patches and 6–9 H_V1ΔC patches. *P < 0.01 by Student's t test.

Zn²⁺ sensitivity of the monomer is weaker than the dimer

Figure 5 shows that the Zn²⁺ sensitivity of the monomeric H_V1ΔC channel (Fig. 5B) was distinctly weaker than that of the dimeric H_V1 (Fig. 5A). The main effects of Zn²⁺ on proton currents are to slow channel opening (larger activation time constant, τ_act) and to shift the proton conductance–voltage relationship, g_H–V, positively. The slowing of τ_act was significantly greater in the WT channels than in the H_V1ΔC (P < 0.01 at 1, 10 and 100 μm Zn²⁺, data not shown).

Figure 5. The Zn²⁺ sensitivity of the human monomeric proton channel is weaker than that of the dimer.

A, WT H_V1 proton current families in the same cell at pH_o 7.0, pH_i 6.5 during pulses in 10 mV increments to +30 mV (control) or to +60 mV (Zn²⁺). B, weaker Zn²⁺ sensitivity of H_V1ΔC monomeric channels at pH_o 7.0, pH_i 6.5. Pulses were applied in 10 mV increments up to +100 mV.

Measurements were also performed on mouse proton channels and the monomeric mVSOPΔNΔC construct, in order to determine whether the phenomena described here apply to proton currents in other species. WT mouse proton currents (mVSOP) activated sigmoidally, and the monomeric mVSOPΔNΔC currents activated exponentially (not shown). Slowing of τ_act by Zn²⁺ was evaluated as described for H_V1. The Zn²⁺ sensitivity of the monomeric construct was significantly weaker than WT at 1, 10 and 100 μm Zn²⁺ (P < 0.05 for each, data not shown).

Some slowing of activation is expected to result from the shift by Zn²⁺ of the gating of most voltage-dependent channels, although depending on the mechanism, different gating parameters may be shifted by different amounts (Frankenhaeuser & Hodgkin, 1957; Hille, 2001; Elinder & Åhrem, 2004). To evaluate this electrostatic effect, we determined the shift of the g_H–V relationship produced by each [Zn²⁺] in each cell and shifted the τ_act–V relationship negatively by that amount. Figure 6 shows the residual slowing of τ_act after this correction. WT murine and human proton channels are slowed by Zn²⁺ beyond what is expected from a simple voltage shift, but for the mVSOPΔNΔC and H_V1ΔC constructs, most of the observed slowing was attributable to the voltage shift.

Figure 6. Monomeric proton channels (mVSOPΔNΔC or H_V1ΔC) are less sensitive to the slowing effect of Zn²⁺ than WT dimeric channels (mVSOP or H_V1).

Plotted is the mean ±s.e.m. residual slowing after correction for the observed shift of the g_H–V relationship in each cell at each [Zn²⁺]. Slowing is defined as the ratio τ_act(Zn²⁺)/τ_act, which was determined at moderate depolarizations in the voltage range where τ_act is approximately exponentially dependent on voltage. For mVSOPΔNΔC and mVSOP, n= 3–4; for H_V1 and H_V1ΔC, n= 8–11. *P < 0.05 for truncated vs. WT channels.

Zn²⁺ sensitivity of histidine mutants

In order to further test the hypothesis that externally applied Zn²⁺ exerts its characteristic effects at one or more high affinity binding sites at which Zn²⁺ is coordinated between two His residues on the proton channel, we studied a series of His mutants and constructs, illustrated in the diagram in Fig. 7. The various constructs in Fig. 7 were expressed and studied at pH_o 7.0 in the presence of Zn²⁺. Of the main effects of Zn²⁺ (Cherny & DeCoursey, 1999), we focused on the shift of the g_H–V relationship (Figs 8 and 9A) and the slowing of τ_act (Fig. 9B).

Figure 7. Diagram of the histidine mutants and constructs used and their nomenclature.

His¹⁴⁰ is on S2 near the external surface of the membrane; His¹⁹³ is on the S3–S4 linker. Filled symbols indicate His, open symbols indicate Ala substitution. The C-terminal truncation H_V1ΔC expresses as a monomer (Tombola et al. 2008; Koch et al. 2008). The dimer assembles mainly by C-terminal interactions (Tombola et al. 2008; Koch et al. 2008; Lee et al. 2008). Tandem dimers were linked by insertion of –ASGAAA– between C- and N-termini of two monomers.

Figure 9. Effects of Zn²⁺ on the voltage dependence and gating kinetics of several proton channel constructs with His mutations.

A, shifts of the g_H–V relationship produced by Zn²⁺ in various human H_V1 constructs. Inset diagrams illustrate the positions of His¹⁴⁰ and His¹⁹³ residues (•, ▪) or their substitution by Ala (○, □), respectively, as also shown in Fig. 7. Tandem dimer data are connected by dashed lines. Mean ±s.e.m. are shown for numbers of cells studied at up to 100 μm Zn²⁺: 9–12 WT; 5–7 H140A; 5–6 H193A; 5–6 H140A/H193A; 3–12 WT–WT tandem; 5–7 H193A–H140A tandem; 5–6 WT–H140A/H193A tandem. by Student's t test compared with WT. B, slowing of activation kinetics produced by Zn²⁺ in various H_V1 constructs. Mean ±s.e.m. slowing of τ_act by Zn²⁺ (defined as the ratio τ_act(Zn²⁺)/τ_act) in the indicated proton channel constructs, for numbers of cells studied at up to 100 μm Zn²⁺: 8–11 WT; 5–7 H140A; 5–8 H193A; 4 H140A/H193A; 4–11 WT–WT tandem; 5–7 H193A–H140A tandem; 5–6 WT–H140A/H193A tandem. by Student's t test compared with WT.

With regard to the shift of the g_H–V relationship by Zn²⁺, H140A was similar to WT, but H193A was distinctly less sensitive (Fig. 9A; P < 0.001 at all [Zn²⁺]). This result suggests that Zn²⁺ bound to His¹⁹³ exerts a stronger effect on the voltage sensor of the proton channel. Surprisingly, both single His mutants H140A and H193A exhibited more similar slowing effects (Fig. 9B), although H193A was again more sensitive (P < 0.05 vs. H140A at 1 and 10 μm Zn²⁺). The slowing of τ_act by Zn²⁺ was significantly weaker than WT in either single mutant, indicating that both His residues contribute to the slowing of channel opening by Zn²⁺. The double His mutant H140A/H193A exhibited essentially no slowing by Zn²⁺, but a weak shift of the g_H–V relationship remained. The shift may reflect non-specific divalent cation effects or Zn²⁺ binding to sites on the channel other than His. Nevertheless, that no slowing was seen in H140A/H193A suggests that virtually the entire slowing effect can be attributed to Zn²⁺ binding to His¹⁴⁰ and His¹⁹³.

Zn²⁺ sensitivity of histidine mutants and constructs: tandem dimers

A control WT–WT tandem dimer was generated by linking two WT H_V1 channels. The gating and Zn²⁺ sensitivity of this WT–WT tandem dimer (Fig. 8A) were indistinguishable from WT channels. This result supports the validity of the use of tandem dimers to evaluate the characteristics of Zn²⁺ binding to H_V1. In addition, this result further supports the conclusion that the Hvcn1 gene product assembles as a dimer in several expression systems (Koch et al. 2008; Lee et al. 2008; Tombola et al. 2008).

If a high-affinity binding site for Zn²⁺ can exist within a single monomer, then a high-affinity site should exist on the WT–H140A/H193A tandem dimer, which retains both His residues in one monomer but neither in the other. However, Zn²⁺ effects were drastically attenuated in this construct (Fig. 8B). For example, in the WT–WT tandem dimer, 1 μm Zn²⁺ already shifts the g_H–V relationship positively by ∼10 mV (Fig. 8C). In contrast, in the WT–H140A/H193A tandem dimer, little shift occurs until 100 μm Zn²⁺ is introduced (Fig. 8D).

The properties of the WT–H140A/H193A tandem dimer were unique in one respect. Most of the channel constructs tested activated exponentially or exponentially after a delay, with or without Zn²⁺. An exception is the H193A–H140A tandem dimer, whose currents were sometimes better fitted with two components in the absence of Zn²⁺. Proton currents in the WT–H140A/H193A tandem dimer activated with two components, but only in the presence of ≥100 μm Zn²⁺ (Fig. 8B). When fitted with two exponentials, the faster component had time constants <1 s, whereas the time constant of the slower component was close to τ_act measured in the absence of Zn²⁺. Thus, Zn²⁺ not only failed to slow activation, it appeared to introduce a novel rapid component of activation. Assuming that cooperative gating normally occurs, it might be speculated that in this construct, Zn²⁺ may uncouple the two monomers. In any case, the absence of slowing by Zn²⁺ of any component of opening of the WT–H140A/H193A tandem dimer suggests that slowing occurs when Zn²⁺ binds between His residues in the two monomers at the dimer interface.

The shifts of the g_H–V relationship and the slowing of τ_act produced by Zn²⁺ are summarized for this and other constructs in Fig. 9. Figure 9A reveals that the shift in the g_H–V relationship produced by Zn²⁺ was almost as weak in the WT–H140A/H193A construct which has one pair of dissimilar His residues as in the double mutant H140A/H193A that lacks His altogether.

Upon examination of the slowing of channel opening by Zn²⁺ (Fig. 9B), the constructs fall into three categories. Slowing is most profound for constructs in which both pairs of His¹⁹³ and His¹⁴⁰ exist (WT and WT–WT tandem). Weaker but distinct slowing also occurs when a single symmetrical pair of His¹⁴⁰ or His¹⁹³ are present (H193A or H140A) or the asymmetrical pair (H193A–H140A tandem dimer). Remarkably, there was practically no slowing of activation in the WT–H140A/H193A tandem dimer or the double mutant H140A/H193A. The simplest interpretation is that slowing occurs when Zn²⁺ can interact with both monomers simultaneously.

Discussion

Monomer interactions during channel opening

An apparent contradiction between structural predictions based on electrophysiological measurements of competition between Zn²⁺ and H⁺ (Cherny & DeCoursey, 1999), and a molecular homology model of the proton channel VSD (Fig. 1) led to the hypothesis that Zn²⁺ binds between monomers in the proton channel dimer. The obvious prediction that monomeric channels should have lower Zn²⁺ affinity was borne out. That Zn²⁺ could prevent channel opening at such a location further suggested that the two monomers are closely apposed and may interact during channel opening. Several types of evidence are consistent with this hypothesis. Monomeric channels open with an exponential time course, rather than the sigmoidal kinetics typical of all native proton channels (DeCoursey, 2003) and of cooperative gating mechanisms in general (Hodgkin & Huxley, 1952; Horrigan et al. 1999). Channel opening was 6.6 times faster in monomeric channels, consistent with a more facile opening process than in the dimer. The activation energy of WT channel gating was double that of monomeric channels, indicating a more complex opening process in the dimer. The native voltage-gated proton channel has much larger activation energy for opening than most other voltage-gated ion channels. That the Q₁₀ was found to be similar for the delay, τ_act and τ_tail, led to the suggestion that the channel comprises multiple subunits that each undergoes a single rate-determining conformational change (DeCoursey & Cherny, 1998). Finally, Zn²⁺ retards proton channel opening beyond the slowing expected from a simple voltage shift (Cherny & DeCoursey, 1999), as was observed for Cd²⁺ effects on snail neuron proton channels (Byerly et al. 1984). Together, these observations strongly suggest pronounced interaction between monomers during proton channel opening. Two recent studies support this conclusion by providing evidence for cooperativity (Gonzalez et al. 2010; Tombola et al. 2010). In one proposal, both monomers must activate before either can conduct current (Gonzalez et al. 2010). In the other proposal, the subunits can open individually, but positive cooperativity results in both usually being in the same open or closed state (Tombola et al. 2010).

The physical interpretation of the present results is not clear-cut. Ascribing the delay in H⁺ current activation to a gating step (between closed states) preceding opening in each (independent) monomer would not account for the absence of delay in monomeric channels. One possibility is that in the dimer, the gating of each monomer is slowed by the presence of the other monomer. Alternatively, perhaps both channel subunits must enter the ‘open’ configuration before either can conduct current. This interpretation is suggested by the observation that the monomeric channel activates with an exponential time course, whereas dimeric channel currents exhibit a sigmoidal time course (Fig. 3), reminiscent of classical Hodgkin–Huxley kinetics (Hodgkin & Huxley, 1952). Similarly, for the Ciona proton channel (CiVSOP), monomeric constructs activated exponentially, whereas the dimer opened with sigmoidal kinetics (Gonzalez et al. 2010). The relatively brief delay and slower subsequent activation suggest that a slow concerted gating process occurs after both monomers have activated (Horrigan et al. 1999). In either case, the slower activation of dimeric channels suggests a complex gating process involving interaction between the two monomers. Intriguingly, direct measurements of the single-channel current of voltage-gated proton channels provided estimates that were roughly twice the value determined from current fluctuation analysis (Cherny et al. 2003). This observation is compatible with a gating process in which both channels in the dimer open approximately synchronously often enough to be interpreted as a single opening event (with twice the unitary conductance), but in which most gating transitions (that contribute to the fluctuations) involve individual conduction pathways. This phenomenology appears to be more consistent with the proposal of Tombola et al. (2010) than that of Gonzalez et al. (2010).

The sigmoidal activation kinetics (DeCoursey, 2003) and strong temperature dependence of gating of native proton currents (DeCoursey & Cherny, 1998; Kuno et al. 2009) closely resemble the corresponding behaviour of expressed WT (dimeric) proton channels, but not monomeric constructs. These properties therefore support the idea that the native proton channel, like expressed proton channels, assembles as a dimer, despite the difference in voltage dependence of expressed and native proton channels (Musset et al. 2008).

Teleological advantages of cooperative gating

Why would cooperative gating at the expense of slower activation be advantageous for proton channels? Cooperativity increases the voltage sensitivity of channel opening (Sigworth, 1993). Proton channels exist in many non-excitable cells. In many situations that result in channel opening, speed typically is not an important consideration. For example, the phagocyte respiratory burst continues for roughly 1 min to 1 h depending on the stimulus (DeCoursey & Ligeti, 2005). It may be more important that proton channels open with steep voltage dependence than rapidly. Proton efflux limits membrane depolarization, which is desirable during the respiratory burst, because NADPH oxidase enzyme activity is voltage dependent, and is inhibited at large positive voltages (DeCoursey et al. 2003; Petheő & Demaurex, 2005).

Deducing channel architecture from Zn²⁺ effects

Simple models of competition between H⁺ and Zn²⁺ indicated that Zn²⁺ binds to multiple titratable groups with pK_a 6.2–7.0 (Cherny & DeCoursey, 1999). The human proton channel has two His residues predicted to be exposed to the extracellular solution, His¹⁴⁰ and His¹⁹³, and mutation of either one to Ala reduced Zn²⁺ sensitivity of the channel; the double mutant H140A/H193A was even less sensitive (Ramsey et al. 2006; see also Fig. 9). These observations point to Zn²⁺ coordination between two His residues. The two His in each monomer were too far apart to coordinate Zn²⁺ in the homology model (Fig. 1A), and the weaker Zn²⁺ effects in monomeric H_V1ΔC are consistent with the distance between His¹⁴⁰ and His¹⁹³ being too great for Zn²⁺ to bind to both His within a single monomer. However, potential high-affinity Zn²⁺ binding sites occur at the interface between monomers in the dimer model. Plausible dimer models were identified in which potential binding sites were comprised of His¹⁴⁰–His¹⁴⁰ or His¹⁹³–His¹⁹³ pairs (Fig. 1B). An alternative site formed by His¹⁴⁰–His¹⁹³ from complementary subunits can be obtained by small, opposite rotations of each monomer in the dimer shown in Fig. 1B, around an axis parallel to the membrane. The resulting His¹⁴⁰–His¹⁹³ distance is approximately 7 Å (not shown).

An alternate interpretation is that in the WT channel, strong Zn²⁺ effects are the result of Zn²⁺ binding independently to each channel subunit. Two arguments oppose this interpretation. First, the competition between H⁺ and Zn²⁺ required that at least two titratable groups coordinate Zn²⁺ at its primary site of action (Cherny & DeCoursey, 1999). Second, assuming that neither channel can conduct until both undergo an opening transition (Gonzalez et al. 2010), if Zn²⁺ could bind independently within each monomer, then the Zn²⁺ effects observed in the WT–H140A/H193A tandem dimer (Fig. 9) would be expected to be more profound. In contrast, this construct was nearly as insensitive to Zn²⁺ as the double mutant that lacks both His (H140A/H193A).

Since Zn²⁺ effects do not saturate at achievable concentrations, distinguishing differential binding affinity from differential efficacy at distinct sites is not straightforward. Nevertheless, some conclusions appear warranted. When Zn²⁺ binds either to His¹⁴⁰ (in H193A) or His¹⁹³ (in H140A) the g_H–V relationship shifts, although the shift is greater at any given [Zn²⁺] when Zn²⁺ binds at His¹⁹³ (in H140A or H193A–H140A tandem). This result suggests that Zn²⁺ exerts a stronger effect on the voltage sensor of the proton channel when bound to His¹⁹³ than to His¹⁴⁰. However, it is also possible that the affinity of Zn²⁺ is greater for His¹⁹³ than for His¹⁴⁰.

Interpreting Zn²⁺ effects on the shift of the g_H–V relationship and slowing of τ_act is further complicated by the interrelatedness of these parameters. If the channel opening rate is decreased, the g_H–V relationship will shift positively, because the open probability will decrease at all potentials. Electrostatic effects generally shift voltage-dependent parameters along the voltage axis (Frankenhaeuser & Hodgkin, 1957; Hille, 2001; Elinder & Åhrem, 2004); in the simplest form of this type of mechanism, all parameters shift equally. However, the slowing of τ_act in proton channels by Zn²⁺ not only exceeds that predicted from the shift of the g_H–V relationship, but results in τ_act values at high Zn²⁺ that are slower than at any voltage without Zn²⁺ (Cherny & DeCoursey, 1999). It appears that Zn²⁺ slows channel opening by a distinct mechanism, and the data presented here indicate that this slowing effect occurs when Zn²⁺ binds to His residues. That no slowing was seen in H140A/H193A suggests that virtually the entire slowing effect can be attributed to Zn²⁺ binding to His¹⁴⁰ and His¹⁹³. In addition, because the rank order of sensitivity to Zn²⁺ effects is different among the constructs tested for slowing and shift of the g_H–V relationship, these effects are, at least to some extent, separable.

The slowing of channel opening in the constructs tested (Fig. 9B) supports the hypothesis that the dimer interface allows His¹⁴⁰ and His¹⁹³ from each monomer to pair with the corresponding His from the other monomer to form a Zn²⁺ binding site. Profound slowing by Zn²⁺ occurs when all four His residues are present in the dimer (WT and WT–WT tandem). Moderate slowing also occurs when a single His residue is present in each monomer (H140A, H193A or H193A–H140A tandem dimer). The slowing of τ_act by Zn²⁺ was significantly weaker than WT in either single mutant, indicating that both His residues contribute to the slowing of channel opening. The weak but distinct slowing in the H193A–H140A tandem dimer might reflect coordination of Zn²⁺ between dissimilar His or between His¹⁹³ and Glu¹⁹² or Glu¹⁹⁶ in the opposite monomer. Glu¹⁹² and Glu¹⁹⁶ are 9 Å and 11 Å, respectively, from His¹⁹³ in the model, but they are in the flexible S3–S4 linker and thus might approach more closely. Remarkably, there was no slowing of τ_act in the WT–H140A/H193A tandem dimer. Taken together, these data suggest that when Zn²⁺ binds to His¹⁴⁰ or His¹⁹³ residues at the interface between monomers, neither monomeric channel can undergo the physical movement required for opening. Zn²⁺ may retard the movement of one monomer relative to the other during gating. Alternatively, Zn²⁺ may constrain the dimer in a conformation from which it cannot open.

The double His mutant H140A/H193A exhibited essentially no slowing by Zn²⁺ although a weak but distinct shift of the g_H–V relationship remained. The shift may reflect non-specific divalent cation effects or binding to residues on the channel other than His. In enzymes containing structural or catalytic Zn²⁺, three or four amino acids (His, Cys, Glu or Asp) coordinate the Zn²⁺ (Vallee & Auld, 1990; Alberts et al. 1998). Of course, Zn²⁺ binding to the proton channel is neither structural nor catalytic and is rapidly reversible and consequently, a formal binding site is not expected.

Conformational changes during gating are unlikely to bring His¹⁴⁰ and His¹⁹³ closer together in the monomer

The homology model in Fig. 1 was based on what is presumed to be an open conformation of K⁺ channels (Jiang et al. 2003a; Long et al. 2007). Closing of K⁺ channels is thought to involve inward motion of S4, changing its orientation relative to other parts of the voltage-sensing domain (VSD; the S1–S4 complex) (Bezanilla, 2000; Jiang et al. 2003b; Starace & Bezanilla, 2004; Tombola et al. 2006; Swartz, 2008). Roughly analogous motion of S4 in H_V1 is suggested by recent studies (Gonzalez et al. 2010; Tombola et al. 2010), although evidence that truncation of the C-terminus between the second and third Arg residues in S4 did not abolish proton selectivity or gating raises questions about the extent of similarity (Sakata et al. 2010). Positioned within a long helix, His¹⁴⁰ is probably stationary. The kink in the S3 helix might allow His¹⁹³ to swing toward the S2 helix and His¹⁴⁰, but the length of the S3b helix probably precludes an approach close enough for intramonomer Zn²⁺ binding. A model of the closed state of Kv1.2 (Pathak et al. 2007) does not show any significant decrease in the intramonomer distance between residues in positions homologous to those of His¹⁴⁰ and His¹⁹³. This buttresses the idea that the H_V1 monomer does not support tight Zn²⁺ binding. In the dimer, conformational rearrangements (caused by S4 motion, for example) could orient the His pairs more favourably for inter-monomer Zn²⁺ binding in the closed, rather than in the open conformation, which would be consistent with the preferential effect of Zn²⁺ on channel opening over closing (Cherny & DeCoursey, 1999; Elinder & Åhrem, 2004).

Comparison of the proposed dimer interface with previous models

The dimer interface in Fig. 1B, which is supported by the Zn²⁺ studies, differs from previous proposals. FRET studies in the Ciona proton channel indicated 42 Å between corresponding Ser²⁴² residues in the dimer (Koch et al. 2008), which like His¹⁹³ in the human channel, is located in the S3–S4 linker. A study of H_V1 based on a series of Cys mutants led to a topological model (Lee et al. 2008) with a dimer interface different from that in Fig. 1B, in which the two His¹⁹³ are far apart. However, in that study, distinct cross-links were also detected at position 194 in H_V1, at the tip of the voltage-sensor paddle (Lee et al. 2008), which indicates that His¹⁹³ residues from each monomer are sometimes in close proximity. All of these observations can be reconciled if the proton channel dimer is loosely organized and occasionally samples alternative associations. This idea is supported by the fact that the channel appears as a monomer in Western blots (Ramsey et al. 2006; Koch et al. 2008; Lee et al. 2008, 2009; Capasso et al. 2010). Configurations in which His residues from both monomers approach each other may not occur frequently, but when they do, the opportunity for Zn²⁺ to bind exists. Zn²⁺ binding may then constrain the dimer in a conformation that precludes channel opening. Solving the crystal structure (Lee et al. 2009; Li et al. 2009) might shed light on these possibilities, but it may simply illustrate one of many possible orientations, or perhaps H_V1 will crystallize as a monomer. An alternative possibility that cannot be ruled out is that when the H_V1 dimer adopts a conformation with close apposition of S1 regions as proposed by Lee et al. (2009), Zn²⁺ binds between His residues exposed at the edges of dimers, promoting tetramer formation. Evidence consistent with tetramer or higher order polymer formation at least under specific conditions has been observed (Lee et al. 2009).

The C-terminus of H_V1 was crystallized recently, and structural changes were observed at different pH values (Li et al. 2010). Given that both the WT–WT tandem dimer (in which coiled-coil interaction along the entire C-termini may be disrupted by linking N- and C-termini together) and constructs lacking the C-terminus altogether, such as H_V1ΔC, appeared to have normal pH dependence (Koch et al. 2008; Sakata et al. 2010), it seems unlikely that the pH dependence of gating resides in the C-terminus.

The enhanced gating mode does not reflect monomer–dimer interconversion

During the respiratory burst in neutrophils and eosinophils that accompanies phagocytosis and reflects NADPH oxidase activation, proton channels exhibit profoundly enhanced gating (Bánfi et al. 1999; DeCoursey et al. 2000). The enhanced gating strongly promotes proton channel opening and was predicted by modelling to improve the efficiency of NADPH oxidase by limiting the depolarization required to open enough proton channels to compensate for the electrogenic activity of the oxidase (Murphy & DeCoursey, 2006). Depolarization directly inhibits NADPH oxidase activity (DeCoursey et al. 2003; Petheő & Demaurex, 2005). Koch et al. (2008) speculated that the conversion of phagocyte proton channels to the ‘enhanced gating mode’ might reflect interconversion between monomeric and dimeric channel arrangements. As Zn²⁺ binds with high affinity between monomers in the dimer, but weakly to monomeric proton channels, one would predict different Zn²⁺ sensitivity of the two modes of gating. However, in eosinophils, the Zn²⁺ sensitivity of the proton channel in resting and enhanced gating modes is identical (DeCoursey et al. 2001). Since the monomeric proton channel (H_V1ΔC) has distinctly weaker Zn²⁺ sensitivity, it appears that the phagocyte proton channel remains a dimer during enhanced gating. Enhanced gating can be explained by phosphorylation of the dimeric channel (Morgan et al. 2007; Musset et al. 2010).

Summary

Taken together, the evidence suggests profound intermonomer interaction during opening in the dimeric voltage-gated proton channel, despite each monomer having a separate conduction pathway. Interaction between monomers during gating is suggested by the slower opening rate of the dimeric than the monomeric channel, the higher activation energy of channel gating of the H_V1 dimer, and sigmoidal activation kinetics of the dimer compared with exponential kinetics of the monomer. Perhaps a concerted opening process occurs in the dimer. Alternatively, both monomers in the dimer may undergo identical gating movements as a prerequisite to conduction. Zn²⁺ slows channel opening most effectively when symmetrical His residues are available in both monomeric channels. As Zn²⁺ effects are weaker in monomeric than dimeric proton channels, the phagocyte proton channel in both resting and enhanced gating modes is evidently a dimer.

Glossary

Abbreviations

g_H

proton conductance

g_H–V

proton conductance–voltage relationship

H_V1

human proton channel protein

H_V1ΔC

C-terminal truncation of H_V1

mVSOP

mouse proton channel

mVSOPΔNΔC

N- and C-terminal truncation of mVSOP

τ_act

activation time constant

τ_tail

deactivation (tail current) time constant

VSD

voltage-sensing domain

Author contributions

S.M.E.S., B.M., V.V.C., S.R. and T.E.D. contributed to conception and design; S.M.E.S., S.S. and S.R. contributed new tools; S.M.E.S. built the molecular model; B.M., V.V.C. and D.M. collected, analysed and interpreted data. T.E.D. wrote the paper; all authors critiqued the manuscript and approved the final version. Patch clamp studies were done at Rush University, constructs were generated at the University of Chicago and Emory University, molecular modelling was done at Emory University.