We routinely observe robust inward I_STEP at negative potentials in cells expressing Hv1 R1H (Figure 1C,E). Whereas the I_STEP-V relation for WT Hv1 is outwardly rectifying, the I_STEP-V relation in R1H exhibits double rectification with an apparent plateau at intermediate voltages near −30 mV (Figure 1E). The inwardly-rectifying shape of the steady-state I_STEP-V relation in Hv1 R1H at negative potentials is similar to other R1H VS domain mutants (Starace and Bezanilla, 2004; Capes et al., 2012; Struyk and Cannon, 2007; Villalba-Galea et al., 2013) but distinct from the bell-shaped I_STEP-V relations in Shaker R2H and R3H that utilize a carrier-type (G_CA) mechanism for H⁺ transfer (Starace and Bezanilla, 2001; Starace et al., 1997). To discriminate resting- and activated-state conductances in Hv1, we use G_SH terminology in reference to the channel-like H⁺ conductances observed in R1H VS domain mutants.

The I_STEP-V curve in Hv1 R1H exhibits prominent inward rectification at negative potentials (Figure 1E), similar to Shaker and Ci VSP R1H mutants (Starace and Bezanilla, 2004; Villalba-Galea et al., 2013), whereas at potentials > −30 mV, I_STEP-V curve in Hv1 R1H exhibits outward rectification like WT Hv1 (Figure 1E). The apparent ‘plateau’ in the I_STEP-V relation near −30 mV appears to result when both G_SH and G_AQ are close to their respective minima (Figure 1E). Consistent with this interpretation, the doubly-rectifying I_STEP-V relation gives rise to a ‘U-shaped’ G_STEP-V relation in Hv1 R1H (Figure 1F). The net G_STEP-V may be interpreted to represent the amalgam of distinct conductances (G_SH and G_AQ) that have distinct voltage dependencies, opposite gating polarity and unequal maximal amplitudes. Classical ion channel gating theory predicts that G = N·γ·P_OPEN (where γ is unitary conductance, N is channel number and P_OPEN is open probability), and we therefore infer that G_SH = N_SH·γ_SH·P_OPEN-SH and G_AQ = N_AQ·γ_AQ·P_OPEN-AQ. If each Hv1 R1H mutant VS domain mediates both G_SH and G_AQ (albeit at different potentials), i.e., N_SH = N_AQ and γ_AQ≠ γ_SH (Figure 1F).

I_STEP-V and I_TAIL-V relations in WT Hv1 are apparently linear at negative voltages (Figure 1D,E), consistent with the expectation that P_OPEN-AQ will asymptotically approach its minimum as the membrane potential becomes more negative (Gonzalez et al., 2013; Cherny et al., 1995). In R1H however, the inward I_STEP clearly becomes larger as membrane potential becomes more negative (Figure 1E). G_STEP also rises with additional hyperpolarization, suggesting that the voltage-dependent increase in inward current results from an increase in P_OPEN-SH (Figure 1F). The notion that G_SH gating reflects a change in VS conformation is consistent with results from a study conducted showing that Ci Hv1 exhibits kinetically distinct fluorescence changes with distinct voltage dependencies (Qiu et al., 2013). However, G_AQ and G_SH gating measured here are more widely separated than the fluorescence changes (Qiu et al., 2013), suggesting that G_SH may report an earlier transition in the Hv1 activation pathway.

The G_STEP-V relation in Hv1 R1H exhibits a local minimum near −50 to −70 mV; at these intermediate potentials, G_STEP could reflect contributions from G_AQ and G_SH, in addition to voltage-independent membrane leakage (G_LEAK). Inspection of the G_AQ-mediated I_TAIL-V relation in R1H indicates that P_OPEN-AQ is negligibly small at voltages negative to −40 mV (Figure 1G), but the unambiguous dissection of G_SH gating is compromised by the contributions of G_SH and G_LEAK to the aggregate G_STEP. To measure G_SH gating in isolation, we sought to test the hypothesis that mutagenesis could be used to experimentally block G_AQ. We therefore combined R1H with D112V, which is reported to abrogate G_AQ (Musset et al., 2011), but so far we have been unable to measure either G_SH or G_AQ in cells expressing channels D112X-R1H double-mutant (where X is Val, Asn or Ala; not shown).

Previous studies show that N4R and N4K mutations attenuate outward I_STEP mediated by G_AQ, but have comparatively little effect on inward I_TAIL, indicating that basic side chains at this position block the H⁺ permeation pathway in a voltage-dependent fashion (Ramsey et al., 2010; Sakata et al., 2010). We therefore incorporated N4R into the background of Hv1 R1H (R1H-N4R) and measured expressed currents as described for the R1H single mutant. Hv1 R1H-N4R mediates robust inward currents carried by G_SH like R1H, but outward G_AQ-mediated I_STEP amplitude is substantially reduced, and the remaining current exhibit a linear dependence on membrane potential and is thus attributable to G_LEAK (Figure 2A,B). The time course of I_TAIL decay in N4R (Ramsey et al., 2010) and R1H-N4R (Figure 2A) is evidently monophasic and notably lacks the ‘hook’ seen in the presence of the gating modifier 2GBI (Hong et al., 2013), indicating that R1H-N4R channels are open, but blocked, at positive voltages, and relief of a block occurs instantaneously upon hyperpolarization (Figure 2A).

Figure 2 with 1 supplement see all

Download asset Open asset

Unlike R1H, the steady-state R1H-N4R I_STEP-V relation is inwardly rectifying (Figure 2B). A comparison of R1H and R1H-N4R G_STEP-V relations shows that the G_AQ component is absent in R1H-N4R, and G_SH approaches saturable minimum at voltages positive to ~−30 mV (Figure 2C). Linear subtraction of the leakage (G_LEAK = 1.5 ± 0.2 nS at +100 mV and G_LEAK = 1.6 ± 0.2 nS at 0 mV; n = 4 cells) yields a G_STEP-V relation that is readily fit to a single Boltzmann function (Figure 2F), although ambiguity about the maximal amplitude of G_SH at large negative potentials does not permit unambiguous determination of the V_0.5 or slope factors determined from curve fitting (Figure 2F). Although G_AQ is blocked at positive potentials in R1H-N4R, the inward I_TAIL carried by G_AQ remains measurable, and V_THR for activation of G_AQ is similar in R1H and R1H-N4R (Figure 2A,D; Table 2), Boltzmann fits of the respective I_TAIL-V relations illustrate that the fitted slope value is steeper and midpoint (V_0.5) is ~20 mV more negative in R1H-N4R compared to R1H (Figure 2E). Although N4R dramatically decreases outward current carried by G_AQ, the second-site mutation appears to have only a modest effect on G_AQ gating. Wide separation in the positions of the normalized G-V relations (Figure 2D) indicates that G_SH and G_AQ gating report thermodynamically distinct steps in the Hv1 activation pathway.

Our data suggest that voltage-dependent closure of G_SH reports initial VS activation while G_AQ gating reflects a late gating transition. Our results are similar, but not identical, to Shaker R1H (Starace and Bezanilla, 2004). For example, the time courses of G_SH opening and closing in Hv1 R1H-N4R (Figure 2A) are evidently faster than Shaker and Ci VSP R1H mutants (Starace and Bezanilla, 2004; Villalba-Galea et al., 2013), possibly indicating that G_SH gating in Hv1 does not require substantial conformational rearrangement of the protein backbone. In contrast, the time course of G_AQ-mediated I_STEP and I_TAIL (Figures 1C, 2A) are comparatively slow, suggesting that activation and deactivation gating requires more extensive protein conformational rearrangements. G_SH gating and G_AQ block by N4R are likely to involve rapid local changes in the orientation of side chains that lie in or near the focused electrical field, and G_SH gating phenomenologically resembles pore block by a permeant ion. G_Ω gating in R1A/C/S mutant Shaker channels is similarly attributed to block by the protein-associated side chains of S4 Arg residues, which permeate the ‘gating pore’ during VS activation. Consistent with the small apparent gating valence (0.5–0.7 e₀; Figure 2F) estimated from Boltzmann fits, local reorientation of the imidazole side chain in the introduced His at R1 could account for the voltage dependence of G_SH gating.

Differences in the voltage dependence of G_SH and G_AQ gating suggest that mutation of residues which selectively stabilize the activated-state Hv1 VS conformation may preferentially perturb G_AQ gating. An acidic residue in S3, D185/D^3.61, is conserved only in Hv1 (Figure 1—figure supplement 1). D185 mutations produce dramatic shifts in V_THR toward positive potentials (Ramsey et al., 2010) without altering H⁺ selectivity (Musset et al., 2011), consistent with the hypothesis that this residue participates in an interaction that stabilizes the G_AQ-open, activated-state conformation. We therefore introduced D185A and D185H mutations into the background of R1H and measured their effects on G_SH and G_AQ gating. As in Hv1 R1H and R1H-N4R, cells expressing D185-R1H double mutants manifest robust steady-state inward currents at negative membrane potentials (Figure 3A,B; Figure 3—figure supplement 1). As expected, the G_AQ gating is shifted positively in D185H-R1H and D185A-R1H (Figure 3C,D; Figure 3—figure supplement 1A,B). Compared to R1H, V_THR is shifted +65 mV in D185A-R1H and +105 mV in D185H-R1H (Table 2; Figure 3G,H); the effects of D185 mutations in the background of R1H are similar to the effects of single D185A or D185H mutations (Ramsey et al., 2010).

Figure 3 with 2 supplements see all

Download asset Open asset

Like R1H alone, D185H-R1H and D185A-R1H exhibit U-shaped G_STEP-V relations (Figure 3D; Figure 3—figure supplement 1A) that are similar to R1H (Figure 2), indicating that G_SH is not abrogated by D185 mutation. In contrast to R1H, the I_TAIL-V and G_STEP-V relations in D185A-R1H and D185H-R1H exhibit a wider plateau at intermediate potentials (Figure 3D; Figure 3—figure supplement 1A), which facilitates G_LEAK subtraction (Figures 3D, Figure 3—figure supplement 1) and analysis of G_SH gating. At negative voltages where G_SH is open, the D185H-R1H and D185A-R1H I_STEP-V and G_STEP-V relations are similar to R1H and R1H-N4R (Figure 3D; Figure 3—figure supplement 1A,B), indicating that G_SH gating is unaffected by D185 mutation. D185 mutations therefore appear to selectively destabilize the G_AQ-open conformation of the Hv1 VS domain, but do not alter interactions that are important for resting-state stabilization.

Next we sought to test the hypothesis that changes in G_SH gating can also be experimentally measured. However, the lack of G_SH saturation at negative potentials limits our ability to accurately determine G_SH gating parameters for from fits of G_STEP-V data to a Boltzmann function, even when the contributions to the net G_STEP from G_SH and G_LEAK are defined (Figure 2F). To circumvent this limitation, we reasoned that an analysis of the first derivatives of G_STEP-V relations (dG_STEP/dV) could be a useful alternative approach. First, we simulated ideal G_STEP-V relations using a Boltzmann function (Figure 3—figure supplement 2A,C). As expected, a plot of dG_STEP/dV vs. the applied potential (dG_STEP/dV-V) yields a bell-shaped distribution (Figure 3—figure supplement 2B,D). Fitting the data to a Gaussian function allows us to estimate the voltage at which the curve peaks (V_PEAK). Fitted V_PEAK values are correlated to V_0.5 in simulated G_SH and G_AQ Boltzmann distributions (Figure 3—figure supplement 2E). Altering the amplitude and position of simulated G_STEP-V relations to reflect the known effect of changing pH_O on G_AQ gating produces a commensurate shift (40 mV/pH unit) in V_PEAK (Figure 3—figure supplement 2A–E). We confirm that experimentally-measured V_0.5 (estimated by Boltzmann fitting of I_TAIL-V relations) and V_PEAK (from Gaussian fits to dI_TAIL/dV-V) values are similarly pH_O-dependent using experimental I_TAIL data in R1H-N4R (Figure 3—figure supplement 2F,G). The slopes of the V_0.5-pH_O and V_PEAK-pH_O relations in R1H-N4R are each close to 40 mV/pH unit (Figure 3—figure supplement 2G).

Our analyses of simulated and experimental data indicate that V_PEAK can be used to estimate the positions of G-V relations when experimental conditions preclude direct measurement of either G_min or G_max. We therefore compared estimated G_AQ gating parameters in R1H, D185A-R1H and D185H-R1H determined from analyses of V_PEAK and V_THR. In D185H-R1H, I_TAIL does not clearly reach saturation at voltages ≤+200 mV, but the dG_STEP/dV-V relation rises to a peak near +150 mV and falls again at more positive potentials (Figure 3E). Although we did not measure R1H-D185A over as wide a range of positive potentials, we observe a peak in the dG_STEP/dV-V data near +100 mV (Figure 3E), suggesting that the midpoint of the G_AQ-V relation was reached. Gaussian fits of data from R1H, D185A-R1H and D185H-R1H yield V_PEAK values of +23.3 mV, +98.9 mV and +144.3 mV, respectively (Figure 3E), and V_PEAK is well-correlated to V_THR (Figure 3G).

In contrast to G_AQ gating, dG_STEP/dV-V relations at negative voltages are similar in R1H, R1H-N4R, D185A-R1H and D185H-R1H (Figure 3F), and the fitted V_PEAK values indicate that G_SH gating is poorly correlated with the V_THR for G_AQ gating (Figure 3H). First derivative analyses of G-V relations therefore appear to quantitatively agree with results obtained using the established V_THR method (Musset et al., 2008). We noted earlier that the apparent maximal amplitudes of G_AQ and G_SH (G_AQmax and G_SHmax, respectively) are distinct (Figures 1F, 2C), but our estimate of G_SHmax remains tentative (Figure 2F). Using V_PEAK determined from first derivative analysis (V_PEAK = −189 mV = V_0.5) to constrain Boltzmann fits to the R1H-N4R data yields a revised estimate of G_SHmax and the slope factor for G_SH gating (G_SHmax = 4.6 nS, dx = 42.1; Figure 2F, dashed line).

By subtracting the voltage-independent leak (G_LEAK = 1.5 nS in R1H-N4R; Figure 2C), we calculated the net G_STEP-V for R1H (Figure 1F) and estimated the voltage dependence of G_AQ gating (V_0.5 = 29.4 mV; Figure 2—figure supplement 1), which compares favorably with the value determined from direct measurement of the R1H-N4R I_TAIL-V relation (V_0.5 = 26.3 mV; Figure 2E). The foregoing analysis allows us to directly compare G_AQmax (22.2 nS; Figure 2—figure supplement 1) and G_{SH max} (4.6 nS; Figure 2F); after leak subtraction, G_AQmax/G_SHmax = 4.8. Assuming that the maximum open probabilities for G_AQ and G_SH (P_OPENmax-AQ and P_OPENmax-SH) are equal, the data suggest that the respective unitary conductances (γ_AQmax and γ_SHmax, respectively) also differ by a factor of ~5. Stated differently, the data indicate that the capacity for H⁺ transfer via the His-dependent G_SH pathway is about 5 times smaller than that of the intrinsic G_AQ.

A hallmark feature of G_AQ gating in native and expressed Hv1 channels is the sensitivity of G_AQ gating to changes in the pH gradient (Ramsey et al., 2006; Sasaki et al., 2006; Cherny et al., 1995). Mutations of candidate ionizable residues surprisingly failed to alter the sensitivity to changes in pH_O (Ramsey et al., 2010), and the molecular mechanism for △pH sensing remains unknown. A kinetic model of Hv1 gating predicts that a voltage-independent transition governs G_AQ opening, and this gating step could also be required for the channel’s strong sensitivity to changes in △pH (Villalba-Galea, 2014). In order to determine whether earlysteps in the Hv1 activation pathway are sensitive to changes in △pH, we measured G_SH gating in cells expressing Hv1 R1H-N4R at pH_O 5.5, 6.5 and 7.5 (Figure 4A–C). Consistent with the effect of extracellular acidification to increase the driving force for inward H⁺ current, I_STEP increases as pH_O is lowered (Figure 4D). The I_STEP-V relations remain inwardly rectifying for each pH tested, suggesting that G_AQ block by N4R is not perturbed by changing pH_O (Figure 4D).

Figure 4

Download asset Open asset

Congruent with the effect of changing pH_O on I_STEP, G_STEP amplitude also varies with pH_O in R1H-N4R (Figure 4E). To determine whether changing pH_O shifts the apparent position of the G_SH-V relation, we compared the dG_STEP/dV-V relations at pH_O 5.5, 6.5 and 7.5 (Figure 4F). Gaussian fits to the data reveal that V_PEAK (pH_O7.5, V_PEAK = −227 ± 9 mV, n = 11; pH_O 6.5, V_PEAK = −180 ± 7 mV, n = 14; pH_O 5.5, V_PEAK = −156 ± 10 mV, n = 8) is sensitive to changes in pH_O (Figure 4F). To directly compare the pH_O dependence of G_AQ and G_SH gating in R1H-N4R, we also measured I_TAIL at pH_O 5.5, 6.5 and 7.5 and fit the normalized data to a Boltzmann function (Figure 4G). Similar to WT Hv1, V_0.5 (G_AQ gating) shifts −41.0 mV/pH unit in R1H-N4R (Figure 4H). Interestingly, V_PEAK (G_SH gating) shifts −44.2 mV/pH unit (Figure 4H), indicating that G_SH and G_AQ gating are similarly sensitive to changes in pH_O. Together with the effect of D185 mutations on G_AQ gating, our findings imply that △pH-dependent gating occurs early in the Hv1 activation pathway, and later steps (like G_AQ opening) inherit their △pH sensitivity from a previous gating transition. The sensitivity of G_SH-V relations to changes in pH_O, but not to D185 mutation, further reinforces our conclusion that G_AQ and G_SH report thermodynamically distinct gating transitions.

R1H mutations are sufficient to confer phenomenologically similar G_SH in VS domains from Hv1, Shaker and Ci VSP, suggesting the mechanism of H⁺ transfer and resting-state VS structure are similar. A likely mechanism is H⁺ shuttling mechanism via ionizable nitrogen atom(s) in the imidazole ring of the introduced His (Starace and Bezanilla, 2004). H⁺ delivery to and removal from the introduced His presumably requires that hydrogen bonds are formed between nitrogen atoms and intra- and extra-cellular waters, and protons short-circuit the sharply-focused electrical field as they are shuttled by the introduced His (Starace and Bezanilla, 2004). The introduced His imidazole ring side chain is therefore likely to be in or near the hydrophobic barrier, and thus close to F150/F^2.50, in the G_SH-open, resting-state VS domain conformation. With the exception of At TPC1 DII VS domain X-ray structures (Guo et al., 2016; Kintzer and Stroud, 2016), R1/R^4.47 is not close to F^2.50 in putative resting-state VS domain X-ray structures, and the structural basis for H⁺ transfer via G_SH in R1H mutants remains unclear (Takeshita et al., 2014; Li et al., 2014; Vargas et al., 2012; Jensen et al., 2012; Delemotte et al., 2011; Chamberlin et al., 2014; Li et al., 2015).

We therefore generated a new resting-state Hv1 VS domain model (Hv1 D) in which R1 is located adjacent to F150/F^2.50 and subjected the Hv1 D model to all-atom molecular dynamics (MD) simulations as described previously (Ramsey et al., 2010). We also subsequently produced an R1H mutant resting-state Hv1 model structure (Hv1 E) and subjected the mutant model to MD simulation. The backbone structures of Hv1 D and the recently-solved X-ray structures of the domain II VS from At TPC1 (At TPC1 DII VS; pdb: 5W1J and 5DQQ), which adopts a resting-state conformation (Guo et al., 2016; Kintzer and Stroud, 2016), are remarkably similar (Figure 5F). The main difference between Hv1 D and At TPC1 DII VS domains is the tilt of S4 relative to membrane normal. S4 is more vertically oriented in Hv1 D than At TPC1 DII (Figure 5F), but given that S4 is likely to be highly mobile, the subtle difference in S4 tilt is perhaps not surprising. As suggested by protein sequence alignment (Figure 1—figure supplement 1) R537, rather than R531 (Guo et al., 2016; Kintzer and Stroud, 2016), in At TPC1 DII VS occupies a similar position as R1/R205/R^4.47 in Hv1 (Figure 5G), and we therefore define R537 as R1/R^4.47. R1/R^4.47 C_α positions in Hv1D and AtTPC1 DII are separated by 2.9 Å, and the side chains of these residues are similarly directed to the intracellular side of F^2.50 (Figure 5G). Small differences in C_α distances are also measured between D/N^1.51 (1.0 Å), F^2.50 (2.0 Å) and D/E^3.61 (3.5 Å) in Hv1 D/At TPC1 DII VS, and these side chains are also oriented similarly in both structures (Figure 5G). In summary, the structural similarity between the At TPC1 DII X-ray structure (Guo et al., 2016; Kintzer and Stroud, 2016) and Hv1 D model VS domains strongly argues that our Table 1new Hv1 model represents a thermodynamically stable protein conformation.

Figure 5 with 4 supplements see all

Download asset Open asset

As in At TPC1 DII VS X-ray structures, we find that the R1/R^4.47 terminal amine is oriented toward the intracellular vestibule in Hv1 D, where it is predicted to participate in a Coulombic interaction with a conserved acidic residue, D174/D^3.50 (Figure 5G; Figure 5—figure supplement 4F), that is part of the intracellular electrostatic network (Ramsey et al., 2010; Long et al., 2005). In Hv1 E, D174/D^3.50 interacts primarily with R2/R^4.50, rather than R1H (Figure 5—figure supplement 1G), which is consistent with our observation that the G_AQ-V relation is slightly shifted toward negative potentials. The Coulombic interaction between R1/R^4.47 and D174/D^3.50appears to help to stabilize a G_AQ-closed, VS resting-state conformation. R1/R^4.47 also forms a salt bridge with D112/D^1.51 in Hv1 D (Figure 5—figure supplement 1F), but experimental data show that the main effect of D112 mutations is to shift the G_AQ-V relation toward positive potentials, suggesting that D112 plays a more important role in activated-state stabilization than resting-state stabilization (Ramsey et al., 2010). Consistent with this interpretation, we find that although D112 makes a stable electrostatic interaction with a protonated nitrogen atom of the R1H imidazole ring in Hv1 E, R1H only moderately shifts the G_AQ-V relation (Table 2; Figure 1G). Later we explore possible activated-state interactions between D112 and R3/R^4.53 (Figure 6A). Although Coulombic interactions involving D112 are reorganized in Hv1 E compared to Hv1 D (Figure 5—figure supplement 1F,G), distances between C_α atoms of selected atoms (D112/D^1.51, F150/F^2.50, D185/D^3.61 and R1/R205/R^4.47) are nonetheless similar (Figure 5—figure supplement 3A,B), illustrating that the VS domain architecture is similarly stable in Hv1 D and Hv1 E models.

Comparisons of available resting-state model and X-ray VS domain structures suggest an emerging pattern: the vertical position of S4 relative to S1-S3 is characteristically different in Hv1 D and At TPC1 DII compared to other VS domain X-ray structures. For example, the register of S4 Arg residues is shifted by one helical turn in the Kv1.2 resting-state model, where R2 (R^4.50) occupies the same position as R1/R^4.47 in Hv1 D (Figure 5E; Video 1). Similar differences in the register of S4 Arg residues are noted when Hv1 D is compared to putative resting-state conformations in Kv1.2–2.1 chimera (Kv chimera) VS domain models (Jensen et al., 2012; Delemotte et al., 2011) or the Ci VSD ‘down’ (Ci VSD_D, pdb: 4G80) (Li et al., 2014) X-ray structure (Figure 5—figure supplement 1). In Hv1 FL, the R1/R^4.47 side chain extends into the extracellular vestibule and the R2 side chain is close to F^2.50, similar to Kv 1.2, Kv chimera and Ci VSD_D (Figure 5—figure supplement 1E; Video 1). Despite divergent approaches used to elucidate possible resting-state structures, the backbone structure and positions of other key residues in Ci VSD_D, including D^1.51/D129 and F161/F^2.50 are nearly superimposable with their positions in Hv1 D (Figure 5—figure supplement 1D,E). As expected, the backbone structures of S1-S4 helices in Hv1 D (WT) and Hv1 E (R1H) are also quite similar to the Kv1.2 resting-state (Pathak et al., 2007) template structure (Figure 5A–E; Video 1).

Video 1

Download asset

Consistent with the experimental observation that R1H confers G_SH, and R1H is therefore readily accessible to intra- and extra-cellular solvent (Starace and Bezanilla, 2004), we observe that R1H is accessible to water molecules from both sides of the membrane, and the central crevice is similarly hydrated during Hv1 D and Hv1 E MD simulations (Figure 5—figure supplement 3C–I). The imidazole group of R1H is adjacent to F150/F^2.50, midway between D112/D^1.51 and D174/D^3.50, and appears to be appropriately positioned to shuttle protons between waters in the intracellular and extracellular vestibules (Figure 5E; Figure 5—figure supplement 3C–I; Videos 2, 3). In contrast, the central crevices in a Ci Hv1 VS domain resting-state model structure (Chamberlin et al., 2014) and the mHv1cc Hv1/VSP/GCN4 chimeric protein (mHv1cc; pdb: 3WKV) are occupied by hydrophobic side chains (Figure 5—figure supplement 4A–C; Videos 2, 3). In mHv1cc, a cluster of aliphatic side chains caps the central crevice on the extracellular side of R1 (Figure 5—figure supplement 4A; Videos 2, 3), preventing formation of a continuous hydrated pathway for H⁺ transfer (Takeshita et al., 2014), and extracellular water access to R1/R^4.47 is also evidently precluded in Ci Hv1 (Figure 5—figure supplement 4B; Videos 2, 3). Although hydrophobic side chains form a ring that surrounds the central hydrophilic crevice in Hv1 D and Hv1 E, R1 and R1H side chains are clearly visible within the gating pores when the model structures are viewed from the extracellular space (Figure 5—figure supplement 4C; Video 3).

Video 2

Download asset

Video 3

Download asset

In both Hv1 D and Hv1 E, we find that D112/D^1.51 is on the extracellular side of F^2.50 and is readily accessible to solvent (Figure 5E; Figure 5—figure supplement 3C–I). The position of D112/D^1.51 is consistent with experimental data showing that D112 is required for exquisite H⁺ selectivity via G_AQ and mutant channels (other than D112V or D112E, which are either non-functional or similar to WT Hv1, respectively) are permeable to anions (Musset et al., 2011), strongly arguing that the environment around D^1.51 is solvent-exposed. Consistent with experimental data showing that V116 (V^1.55) is functionally redundant with D^1.51 in supporting G_AQ (Morgan et al., 2013), we find that V^1.55 is physically close to D^1.51 in Hv1 D and Hv1 E models (Figure 5—figure supplement 4C). However, we have so far been unable to measure currents associated with G_SH (or G_AQ) in HEK-293 cells expressing R1H-D112A, R1H-D112N or R1H-D112V double-mutant channels (not shown), and it remains unclear whether R1H-D112X mutations disrupt the structure of the permeation pathway, displace necessary water molecules, or attenuate plasma membrane targeting. In contrast to Hv1 D and Hv1 E, the D^1.51, R1 and R2 side chains are closely packed into a hydrophobic crevice (Figure 5—figure supplements 4A, Videos 2, 3) and evidently shielded from waters in mHv1cc (Takeshita et al., 2014).

D185/D^3.61 mutations dramatically shift the G_AQ-V relation toward positive potentials, indicating that this acidic side chain is likely to participate in an interaction that stabilizes the activated-state VS conformation. Interestingly, the Hv1 D185-equivalent (E511/E^3.61) is also present in the At TPC1 DII VS domain (Figure 1—figure supplement 1), but in the X-ray structures E511/E^3.61 interacts with R531/R^4.41 (Guo et al., 2016; Kintzer and Stroud, 2016). R531/R^4.41 is not conserved in Hv1 (Figure 1—figure supplement 1), suggesting that D^3.61 has a specific function in Hv1 that is not shared among other VS domains. Identifying the interacting partner(s) of D185/D^3.61 in Hv1 model structures is therefore of interest. In Hv1 FL, D185/D^3.61 is close enough to engage in a Coulombic interaction with R1 and appears to be solvent-accessible (Figure 5—figure supplement 4F; Video 4), but in mHv1cc, hydrophobic side chains fill the extracellular vestibule and the D^3.61/D181 carboxylate is tightly packed between non-polar side chains, including F178/F^3.58, L184/L^3.64, L197/L^4.43 and L200/L^4.46 (Figure 5—figure supplement 2D,F; 4A; Figure 6—figure supplement 4F), and evidently inaccessible to solvent (Takeshita et al., 2014). Consistent with experimental data showing that D185 mutations do not alter G_SH gating, we observe that D185 is located at the extracellular end of S3 in a solvent-accessible location that is distant from R1 in Hv1 D (Figure 5—figure supplements 3H, 4F).

Video 4

Download asset

In contrast to the Hv1 D resting-state model, D185/D^3.61 is located close to R3/R^4.53 in the activated-state Hv1 B model (Ramsey et al., 2010). We show here that introduction of the N214R (N4R) mutation into Hv1 B does not alter the overall structure or stability of the VS domain, and D185/D^3.61 remains close enough to R3 to participate in a stable Coulombic interaction with R3/R^4.53 (Figure 6; Figure 6—figure supplement 1). As observed previously in Hv1 B (Ramsey et al., 2010), R3/R^4.53 makes a bidentate interaction with D112/D^1.51and D185/D^3.61 in Hv1 B N4R (Figure 6; Figure 6—figure supplement 1). However, N4R addition allows D112/D^1.51 to form a new salt bridge with N4R (Figure 6—figure supplement 1) that may help stabilize the G_AQ-open conformation; this arrangement is in good agreement with experimental effects of N4R and N4K mutations, which markedly slow the timecourse of I_TAIL decay (Ramsey et al., 2010). Although we have not explicitly tested this hypothesis using computational approaches, the N4R side chain appears to be appropriately positioned to sense changes in the electrical field that is thought to be focused near F150/F^2.50 (Ahern and Horn, 2005; Starace and Bezanilla, 2004). Rapid movement of a cationic N4R terminal amine within the electrical field is consistent with the experimental observation that outward currents carried by G_AQ exhibit rapid voltage-dependent block/unblock in Hv1 R1H-N4R (Figure 2).

Figure 6 with 4 supplements see all

Download asset Open asset

The availability of experimentally-refined resting- and activated-state Hv1 model structures suggests that the models could provide insights into the conformational changes associated with VS activation. Consistent with a generally accepted model of VS activation (Vargas et al., 2012), we find that the main difference between our experimentally-constrained activated- and resting-state Hv1 VS domain model structures is the position of S4 relative to the S1-S3 bundle, which appears to form a relatively immobile scaffold (Figure 6—figure supplements 2A–E and 3A,B; Video 5). To estimate the amplitude of S4 displacement in resting- vs. activated-state VS domain X-ray and model structures, we measured distances between equivalent atoms after performing structure-based alignments. Comparing C_α-C_α distances between R1 side chains in Hv1 D or Hv1 E and Hv1 B indicates that the S4 backbone is displaced ~15 Å; somewhat smaller distances (11 Å–13 Å) are measured when Hv1 D is compared to other activated-state structures (Figure 6—figure supplements 3, 4). Most of the calculated difference in R1-R1 C_α distance is observed in the vertical (z, i.e., membrane normal) axis, but differences in helical tilt and twist are also observed (Figure 6—figure supplements 2–4). In summary, our comparisons of VS domain structures suggest that the S4 helix is likely to undergo an ~11–15 Å vertical translation during activation of the Hv1 VS domain.

Video 5

Download asset

Among various G_GP, G_SH measurement has unique properties that offer deep insight into VS activation mechanism and structure: (1) The sufficiency of R1H to confer G_SH implies that the introduced His imidazole side chain ‘short-circuits’ a highly focused electrical field in the VS domain resting conformation (Starace and Bezanilla, 2004; Villalba-Galea et al., 2013; Starace et al., 1997). (2) R1 appears to contribute ~1 e₀ to the gating valence in both Shaker and Hv1 (Gonzalez et al., 2013; Seoh et al., 1996; Aggarwal and MacKinnon, 1996), and G_SH gating exhibits a similarly small (~0.7 e₀) apparent valence, constraining possible side chain positions within the electric field (Gonzalez et al., 2013; Ahern and Horn, 2005; Tao et al., 2010). (3) Voltage-dependent block of G_AQ by N4R places terminal nitrogen atoms at the intracellular entrance of the H⁺ permeation pathway and therefore close to the hydrophobic barrier formed by conserved hydrophobic side chains, including F^2.50. Biophysical properties of R1H and the effects of second-site mutations can thus be used to experimentally constrain the relative positions of specific side chains in VS domain model structures (Figure 6—figure supplements 2–4).

Our comparison of new and existing resting-state VS domain model and X-ray structures highlights structural features that are required for G_SH. The VS domain contains an hourglass-shaped aqueous central crevice with a central hydrophobic barrier (Ramsey et al., 2010; Takeshita et al., 2014; Wood et al., 2012; Chamberlin et al., 2014; Kulleperuma et al., 2013). The electrical field is highly focused across the hydrophobic barrier, and side chain chemistry at this location is therefore exquisitely sensitive to changes in membrane potential (Lacroix et al., 2014; Vargas et al., 2012; Tao et al., 2010). Although VS domains share a common protein fold, subtle differences in local structure and chemistry have the potential to imbue different voltage sensors with divergent functional properties (i.e., H⁺ permeation or pH-dependent gating). A detailed understanding of the similarities and differences in VS domain structure is therefore essential for dissecting VS mechanism.

Grotthuss-type H⁺ shuttling by the R1H imidazole side chain demonstrates that in the resting-state conformation, R1 is located at the hydrophobic constriction and the central crevice is hydrated and accessible to both intra- and extra-cellular water molecules. Resting-state VS domain structures (Figures 5, 6) in which the R1 side chain extends away from F^2.50 and into the extracellular vestibule (Takeshita et al., 2014; Li et al., 2014; Jensen et al., 2012; Delemotte et al., 2011; Li et al., 2015) may therefore represent intermediate-state conformations rather than the full resting-state conformation. A distinguishing feature of the Hv1 D resting state model and At TPC1 DII VS domain X-ray structures is the orientation of the R1 side chain, which extends into the intracellular vestibule (Figure 5). The position of the R1 side chain is consistent with the hypothesis that a local, voltage-dependent conformational rearrangement of R1 (or R1H) constitutes an initial step in the VS activation pathway, and that G_SH gating directly reports this transition.

Consistent with our data, D233/D^3.61 and R1/R255/R^4.47 are distant in the Ci Hv1 resting-state model (Chamberlin et al., 2014); however, the R255 side chain is intracellular to F^2.50 in the Ci Hv1 model and R1 does not appear to be appropriately positioned to mediate G_SH if it were mutated to His (the ability of R1H mutation to confer G_SH in Ci Hv1 remains to be tested experimentally). In mHv1cc, the D181/D^1.61 faces away from both R1/R201/R^4.47 and R2/R204/R^4.50, and these ionizable side chains are uncharacteristically packed into hydrophobic crevices (Figures 6A; Figure 5—figure supplement 4A; Video 2) (Takeshita et al., 2014). Finding these ionizable side chains in hydrophobic environments is unexpected because R1-R3 are expected to contribute cationic gating charge (Gonzalez et al., 2013; Seoh et al., 1996; Aggarwal and MacKinnon, 1996) and D^1.51 and D^3.61 appear to engage in Coulombic interactions that stabilize the activated-state conformation of the Hv1 VS domain (see below). In contrast to mHv1cc, D/N^1.51, D/E^3.61, R^4.47, R^4.50 and R^4.43 are readily solvent-accessible in Hv1 D and At TPC1 DII VS domains.

Outward current carried by G_AQ is selectively blocked in R1H-N4R, and rapid (< 1 ms) relief of block upon subsequent hyperpolarization (Figure 2A) strongly implicates that the N4R side chain functions as a tethered blocker operating from the intracellular side of the H⁺ permeation pathway. In agreement with a widely-accepted prevailing model of structural rearrangement during VS activation, we find that the position of F^2.50 is similar in G_AQ-open (Hv1 B) and G_SH-open (Hv1 D) model structures, and the central hydrophobic barrier is evidently maintained throughout the Hv1 gating cycle. Nonetheless, the gating pore remains well-hydrated in both resting and activated-state conformations (Figure 5—figure supplement 3). By analogy, we hypothesize that R3 needs to move outward, past F^2.50, to unblock the central crevice and open G_AQ. The dramatic positive shifts in G_AQ-V relations imparted by D185/D^3.61 (Figure 3) and R3 (Ramsey et al., 2010) mutations argue that interactions between these side chain are required for activated-state stabilization in WT Hv1 channels.

Consistent with the observation that D112/D^1.51 mutations also cause large positive shifts in the G_AQ-V relation (Ramsey et al., 2010), we find that R3 also interacts with D112 in the Hv1 B model structure. D185 mutations do not alter G_SH-V gating, indicating that this residue does not meaningfully contribute to stabilization of the Hv1 VS resting-state conformation, and D185 is appropriately distant from R1 in the Hv1 D resting-state model. In contrast to our experimental observations, a D185-R1 interaction is predicted to stabilize the Hv1 FL model activated-state conformation (Li et al., 2015). We conclude that G_AQ opening is directly controlled by a late step in the Hv1 activation pathway that requires interactions between D185 and one or more S4 Arg residues, most likely R3. We hypothesize that D185/^D3.61 functions to pull S4 upward, and thus helps to stabilize the G_AQ-open conformation. D112/D^1.51 also appears to play an important role in stabilizing S4 in the G_AQ-permissive conformation of S4, and may indirectly interact with D185 through R3, as seen in Hv1 B N4R (Figure 6—figure supplement 1). D^1.51 and D^3.61 are selectively conserved in Hv1 channel VS domains (Figure 1—figure supplement 1), and their contributions to activated-state stabilization are predicted to be necessary for H⁺ channel activity.

G_AQ closing requires only a small inward translation of R3 toward F^2.50. The R3-associated cation may disrupt the hydrogen bond network required for H⁺ transfer and/or electrostatically prevent inward H⁺ flux through the central crevice. Membrane hyperpolarization presumably drives the VS through several non-conducting intermediate states similar to those seen in Kv channel VS domain simulations (Jensen et al., 2012; Delemotte et al., 2011), and the full resting conformation is achieved when R1 reaches the position near F^2.50 seen in Hv1 D (Figure 5E). The mechanism outlined above is generally consistent with a widely accepted model of the VS activation process (Vargas et al., 2012), and G_SH data reported here extend this model to Hv1 channel gating.

An intriguing but as yet unresolved question is whether the amplitude of S4 movement is similar in Hv1, Ci VSP and voltage-gated channels like Shaker and Kv1.2. The gating valence in Shaker K⁺ channels is ~3 electronic charges (e₀) per VS domain, and likely reflects the movement of R1-R4 side chains in or through the electrical field (Bezanilla, 2008; Seoh et al., 1996; Aggarwal and MacKinnon, 1996). A limiting slope analysis of Hv1 gating suggests that the effective gating valence (~2.5 e₀/VS) is slightly smaller than Shaker (Gonzalez et al., 2013), consistent with the substitution of a neutral polar Asn (N^4.56; N214 or N4 in Hv1) at the R4 position (Figure 1A, Figure 1—figure supplement 1). The decreased gating charge in Hv1 suggests that VS activation (and thus G_AQ opening) might require a smaller displacement of S4 than is seen in prototypical VGCs like Shaker. However, except for state-dependent mapping of chemical sensitivity in Ci Hv1 (Gonzalez et al., 2013), experimental data that constrain S4 position in resting- and activated-state conformations of the Hv1 VS domain have not been reported.

Atomic distances measured in resting- and activated-state Hv1 models suggest that R1 (R^4.47) C_α atoms in S4 could move as much as 14–16 Å (Figure 6—figure supplements 3 and 4). Shorter distances (11–13 Å) are measured when Hv1 D is compared to other activated-state VS domain models (Figure 6—figure supplements 3 and 4). The apparent flexibility of Arg side chains in VS domains (Li et al., 2014) suggests that the magnitude of S4 translation may not be easily inferred from measurements of gating valence alone. Proton transfer via G_SH and voltage-dependent block of G_AQ appear to place stringent constraints on the relative positions of target side chain atoms, and may offer advantages over alternative approaches, such as chemical accessibility in Cys mutant proteins, for ascertaining structural changes that occur during VS activation. However, a systematic comparison of experimental and structural strategies in each model system is needed to identify specific advantages and liabilities of various approaches. The combination of electrophysiological and computational approaches used here allows researchers to iteratively refine model structures and experimentally interrogate new structure-based hypotheses of mechanism in the context of biophysically-determined kinetic and thermodynamic parameters of protein function, and is thus faster and more flexible than structural determination by X-ray crystallography alone. Although our experimental data probably do not offer sufficient spatial resolution to discriminate whether S4 moves ~12 Å vs. ~14 Å, it is difficult to reconcile our data with models in which S4 movement is closer to 5 Å, such as is seen when Ci VSD_U and Ci VSD_D X-ray structures are compared (Figure 6—figure supplements 3 and 4) (Li et al., 2014).

Direct comparisons of G_AQ and G_SH gating reveal additional insight into the VS activation mechanism. G_AQ-V and G_SH-V relations are oppositely sensitive to changes in membrane potential and gated over widely-separated ranges of voltage change, and thus report thermodynamically distinct gating transitions. We show for the first time that G_AQ and G_SH gating is similarly sensitive to changes in pH_O (Figure 4). In a previously proposed Hv1 gating scheme, the pH dependence of G_AQ gating attributed to closed-state transitions that occur early in the Hv1 activation pathway (Villalba-Galea, 2014), and the pH_O dependence of G_SH gating reported here is consistent with this model. Voltage clamp fluorimetry (VCF) in Ci Hv1 also supports the conclusion that VS conformational rearrangements are detectable prior to G_AQ opening (Qiu et al., 2013), but the pH dependence of fluorescence changes has not been investigated in Hv1. Intriguingly, a VCF study conducted in hERG (Shi et al., 2014) suggests that pH-dependent gating could be a more widespread property of VS activation mechanism than has previously been appreciated.

The mechanism of pH_O-dependent gating in Hv1 is enigmatic. pH_O sensitivity is surprisingly refractory to neutralizing mutagenesis of ionizable residues in Hv1 (Ramsey et al., 2010; Musset et al., 2011; Morgan et al., 2013). Recently, W207/W^4.49 mutations were shown to alter the pH sensitivity of G_AQ gating at alkaline pH_O, but pH-dependent gating at physiological pH_O is similar to WT Hv1 (Cherny et al., 2015). W207 is not predicted to face the hydrated central crevice in either resting- or activated-state Hv1 VS domain models, and the mechanism by which W207X mutations affect pH-dependent G_AQ gating remains mysterious (Cherny et al., 2015). Given that G_SH and G_AQ appear to share the requirement for a hydrated central crevice H⁺ permeation, a plausible hypothesis is that changes in pH_O or pH_I exert their effects mainly by affecting hydrogen bonding patterns in the central crevice. For example, pH-dependent changes in Coulombic interactions within the extracellular vestibule could be coupled with reciprocal conformational changes in the structure of the intracellular electrostatic network, thus altering the VS resting-activated equilibrium. However, the mechanism of pH-dependent conformational coupling remains to be elaborated.

The difference in the apparent maximal amplitudes of G_SH and G_AQ suggests that the mechanisms of H⁺ transfer could be distinct. We and others previously hypothesized that proton permeation via G_AQ occurs in a water wire (Ramsey et al., 2010; Wood et al., 2012; Freites et al., 2006); DeCoursey and colleagues subsequently argued side chain ionization of D112/D^1.51 is required for H⁺ transfer (Musset et al., 2011; Dudev et al., 2015). If proton transfer via G_AQ and G_SH operate by a ‘shuttle’ mechanism requiring explicit ionization of D112 or R1H, respectively, we might expect the unitary conductances (γ_AQ and γ_SH) to be similar. If G_AQ = N·γ_AQP_OPEN-AQ and G_SH = N·γ_SH·P_OPEN-SH, the observation that apparent G_AQmax is ~five-fold larger than G_SHmax (Figure 2) argues that γ_AQ ≈ 5·γ_SH. The smaller G_SH unitary H⁺ transfer capacity is consistent with the hypothesis that hydrogen bonds, which are necessary for H⁺shuttling, are constrained by H⁺ donor and acceptor atom geometry and distance (Cherny et al., 2015). Although D112/D^1.51 is necessary for maintaining the exquisitely high H⁺ selectivity measured in WT Hv1 (Dudev et al., 2015), necessity for D112/D^1.51 to directly catalyze G_AQ H⁺ transfer (Dudev et al., 2015) has not been experimentally determined, and a water-wire mechanism for G_AQ (Ramsey et al., 2010) is equally compatible with the available experimental data. We hypothesize that G_AQ utilizes ensemble of highly dynamic hydrogen bonds between and among waters and protein atoms diffusive in the central crevice for H⁺ transfer in a water wire. Functional redundancy imbued by a water wire is consistent with the resiliency of Hv1 to mutagenesis and potentially explains the more robust H⁺ transfer capacity of G_AQ.

Our experimental and computational results suggest a mechanism for H⁺ conduction and selectivity in Hv1 that is distinct from the interpretation of Dudev, et al. (Dudev et al., 2015). An acidic residue in S1 (D^1.51; E^1.51 or D^1.55 in mutant channels) located in the hydrated central crevice prevents permeation of solution anions (i.e., Cl^-, MeSO₃^- or OH⁻) while R3/R^4.53 limits cation (Li⁺ or Na⁺) permeability (Musset et al., 2011; Berger and Isacoff, 2011; Morgan et al., 2013). D112/D^1.51 and R3/R^4.53 mutations allow ions other than H⁺ to permeate, as reported (Musset et al., 2011; Berger and Isacoff, 2011), demonstrating that these side chains remain ionized in WT Hv1 when G_AQ is open. Because monovalent ions (other than H⁺) are unlikely to permeate as dehydrated ions in D112/D^1.51 and R3/R^4.53 mutant channels, the central crevice remains well-hydrated in these mutant channels (Ramsey et al., 2010). We hypothesize that the previously proposed water-wire mechanism for G_AQ remains operational in D112/D^1.51 and R3/R^4.53 mutants, but permeating ions like Na⁺ and Cl^- transiently disrupt the hydrogen bond structure that is necessary for Grotthuss-type H⁺ transfer via G_AQ. However, intervals between diffusive ion permeation events, rapid H⁺ transfer in the water wire continues unabated. The eroded selectivity reported for D112 and R3 mutants therefore reflects the time-averaged amalgam of two distinct conduction mechanisms: 1) monovalent ion diffusion through a water-filled gating pore, and 2) Grotthuss-type proton transfer in a water wire. In short, G_AQ in both WT and mutant Hv1 channels is mediated water-wire proton transfer, but mutant channels allow more diffusive anion/cation leakage through the hydrated central crevice.

Taken together, our experimental data and model structures indicate that G_AQ and G_SH share a common H⁺ permeation pathway within the hydrated central crevice, but the underlying mechanisms of H⁺ transfer are distinct. A water wire supports G_AQ, while H⁺ shuttling via G_SH requires explicit ionization of the introduced His side chain. The unitary conductance of G_SH, which is ~5 times smaller than G_AQ, reflects the additional complexity that is inherent to the H⁺ shuttle process. The His side chain must first accept a proton from water in the extracellular vestibule, likely undergo a rotation or tautomerization event that delivers the associated proton across the hydrophobic barrier, donate H⁺ to water in the intracellular vestibule, and finally return to the initial conformation to repeat the cycle. The H⁺ shuttle mechanism is channel-like in the sense that voltage-dependent conformational changes gate G_SH and the I_STEP-V relation appears linear (Ohmic) at large negative voltages but transporter-like with respect to the necessity for side chain ionization. G_AQ, on the other hand, requires only water molecules, and the myriad possible hydrogen bonding patterns within the hydrated crevice confers a functionally robust, rapid, and H⁺-selective proton transfer pathway. Systematic testing of the hypotheses elaborated here will require additional computational and experimental strategies, but the results of future studies are likely to produce fundamentally important insights into the mechanisms of VS activation gating by changes in voltage and pH gradients and strategies that underlie H⁺-selective transport in VS domains and other protein systems.

Human Hv1 cDNA (NM_032369) carrying an N-terminal Venus tag was subcloned from pBSTA (gift of Carlos A. Villalba-Galea) into pcDNA5/FRT/TO using standard methods and isogenic tetracycline-inducible FlpIn293-TREx stable cell lines were generated according to the manufacturer’s directions (ThermoFisher Scientific, Waltham, MA ). Parental FlpIn293-TREx cells were obtained directly from the manufacturer and cultured as instructed; cells were not independently authenticated or tested for mycoplasma. Hygromycin B (100 µg/ml) was used for selection and propagation of isogenic stable cell lines. Cells were plated onto glass coverslips and expression of mutant Hv1 proteins was induced by addition of tetracycline (0.5–1 µg/ml) to the culture medium 12–48 hr prior to electrophysiology. Close to 100% of tetracycline-induced cells typically express Venus fluorescence, and both the intensity and pattern of fluorescence was similar among cells expressing a given mutation. Absolute current amplitudes appeared to correlate positively with increasing [tetracycline] and induction time, although this pattern was not studied systematically.

Whole-cell currents were measured at 22–24°C using a List EPC-7 or A-M Systems model 2400 amplifier. Data were low-pass filtered at 2–5 kHz digitized at 10–20 kHz using a National Instruments USB-5221 or USB-5251 DAQ interfaced to a PC computer running a custom LabVIEW 7-based data acquisition and amplifier control program (C. A. Villalba-Galea; details and software distribution available on request). Data were analyzed using Clampfit9 (Molecular Devices) and Origin 6.0 (Microcal). The standard intracellular and extracellular solutions contained (in mM): 100 Bis (2-hydroxyethyl) amino-tris(hydroxymethyl) methane (Bis-Tris), 1 ethylene glycol tetraacetic acid (EGTA), 8 HCl and pH was adjusted to 6.5 and final osmolality of 310–320 mOsm by addition of tetramethylammonium hydroxide (TMA·OH) and methanesulfonic acid (HMeSO₃). Current reversal potentials and pH_O-dependent gating were measured in bath solutions containing either 100 mM 2-(N-morpholino) ethanesulfonic acid (MES, pH 5.5) or 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.5 in place of Bis-Tris, as previously described (Ramsey et al., 2010). Series resistance was not routinely compensated and liquid junction potential corrections are not applied.

Unless otherwise indicated, the data represent means ± SEM of values measured in n cells. I_STEP represents the peak current during steps to the indicated potentials. In most cases, I_STEP was stable, but in cells with large currents we sometimes observed a decay in the amplitude of I_STEP during the voltage step that we attribute to a change in the pH gradient, which may not be sufficiently controlled by 100 mM pH buffer in the recording solutions when G_SH is open. I_STEP is measured during voltage steps and I_TAIL represents peak current immediately after a subsequent hyperpolarizing step determined by fitting current time course to a single exponential function G_STEP was calculated from G_STEP = I_STEP/V-E_REV where E_REV is the zero-current potential determined from inspection of the I_STEP-V relation. I_TAIL amplitude is determined by fitting current decay to a mono-exponential function of the form I_TAIL = I₀ + Ae^-V/t (where I₀ is the minimum current after decay of I_TAIL, A is current amplitude, V is membrane potential and t is time) and extrapolating fits to the instant at which the voltage was changed. V_THR, the apparent threshold for activation of I_TAIL is estimated from visual inspection of I_TAIL as previously described (Musset et al., 2008). Steady-state conductance during voltage steps (G_STEP) is calculated from G_STEP = I_STEP/V-E_REV where E_REV is the zero-current potential determined from inspection of the I_STEP-V relation. In some experiments (see Figure 1—figure supplement 2), we changed V_TAIL (following a constant V_STEP) to determine E_REV of tail currents as previously described (Ramsey et al., 2006). Offline linear leak subtraction of I_STEP-V relations was performed only in cases where the I-V relations are clearly linear (i.e., I_STEP-V in R1H-N4R at V_m > 0 mV and I_TAIL-V in R1H or R1H-N4R at V_m < −50 mV). I_TAIL-V relations are fit to a Boltzmann function of the form: ${\displaystyle I_{TAIL}=\frac{\left(I_{TAIL\mathrm{m}\mathrm{a}\mathrm{x}}\right)-\left(I_{TAIL\mathrm{m}\mathrm{i}\mathrm{n}}\right)}{1+e^{V-V_{0.5}╱dx}}+I_{TAI{L}_{\mathrm{m}\mathrm{i}\mathrm{n}}}}$, where V_0.5 is the voltage at which 50% of the maximum current is reached, dx is a slope factor, and I_TAILmax and I_TAILmin represent the maximum and minimum tail current amplitudes, respectively. G_STEP-V relations are fit to a single Boltzmann of the form ${\displaystyle G_{STEP}=\frac{\left(G_{STEP\mathrm{m}\mathrm{a}\mathrm{x}}\right)-\left(G_{STEP\mathrm{m}\mathrm{i}\mathrm{n}}\right)}{1+e^{V-V_{0.5}╱dx}}+G_{STE{P}_{\mathrm{m}\mathrm{i}\mathrm{n}}}}$ where V_0.5, dx, G_STEPmax and G_STEPmin have the same meanings as defined for I_TAIL. In some cases, effective gating valence (z_G) was calculated from fitted dx values by z_G = RT/F·dx, where F, R and T have their usual meanings (e.g., RT/F = 25.3 mV at 20°C). dG_STEP/dV relations are fit to a Gaussian function of the form $d{G}_{STEP}/dV={(d{G}_{STEP}/dV)}_0+\left(\frac{A}{\varpi \sqrt{\pi /2}}\right)e^{-2}{}^{\frac{{(V-V_{PEAK})}^2}{{\varpi }^2}}$, where V_PEAK is the voltage at which the function reaches its maximum and ω is a width factor.

Models for Hv1 in putative activated (Hv1 C) and resting (Hv1 D) states were developed from the Kv1.2 X-ray structure (pdb:3LUT) and resting state model structure of the Shaker voltage-gated K⁺ channel (Pathak et al., 2007) templates, respectively, using standard homology modelling procedures as described previously (Ramsey et al., 2010; Mokrab and Sansom, 2011). Hv1 B model construction was published previously (Ramsey et al., 2010). Briefly, homologous sequences were obtained for the target sequences and structures from UniRef100 (Bairoch et al., 2005) using non-iterative BLAST (e-value < 10). The two proteins were aligned using MAFFT (Katoh et al., 2002) based on the BLOSUM62 substitution matrix (Henikoff and Henikoff, 1992). Next, a structural profile (i.e. Position-Specific Substitution Matrices - PSSMs) was calculated for the structure and a sequence profile for the target sequence was created. The structural profile was then aligned against the sequence profile using FUGUE (Shi et al., 2001). The resulting structure-sequence alignment was manually adjusted to ensure conservation of key residues, then used as input for MODELLER (Sali and Blundell, 1993) to generate ten models per alignment. The best models were selected based on the energy and constraint violation values of MODELLER and the sequence-structure compatibility scores of pG (Sánchez and Sali, 1998), PROSA2003 (https://prosa.services.came.sbg.ac.at/prosa.php) (Sippl, 1993) and VERIFY3D (http://nihserver.mbi.ucla.edu/Verify_3D/) (Lüthy et al., 1992) as previously described (Ramsey et al., 2010; Mokrab and Sansom, 2011). Any unreliable regions in the model were improved by altering the alignments manually using ViTO (http://abcis.cbs.cnrs.fr/VITO/DOC/index.html) (Catherinot and Labesse, 2004).

All-atom molecular dynamics (MD) simulations were prepared as described (Sands and Sansom, 2007). Side-chain ionization states were determined based on pK_a calculations performed using PROPKA (http://propka.ki.ku.dk/). Ionizable residues were predicted to be in the default states at pH 7 based on standard pK_a values for each residue. We adopted lipid parameters as used previously (Berger et al., 1997). Prior to the production run, a 1 ns equilibration run was performed during which all of the heavy (i.e., not H⁺) protein atoms were harmonically restrained with an isotropic force constant of 1000 kJ mol⁻¹ nm⁻¹. Restrained MD runs were performed at 300K for each protein-bilayer system. Finally, all positional restraints were removed and 20 ns duration production run simulations were performed for each system. MD simulations were performed using GROMACS 3.3 (Van Der Spoel et al., 2005), implementing the GROMOS96 force field (http://www.gromos.net). Lipid parameters were based on GROMOS96, supplemented with additional bond, angle and dihedral terms (Berger et al., 1997). All energy minimization procedures used < 1000 steps of the steepest descent method in order to relax any steric conflicts generated during system setup. Long-range electrostatic interactions were calculated using the particle mesh Ewald (PME) method, with a 12 Å cutoff for the real space calculation (Sagui et al., 2004). A cut-off of 12 Å was used for the van der Waals interactions. The simulations were performed at constant temperature, pressure and number of particles. The temperature of the protein, lipid and solvent (waters and ions) were separately coupled using the Nosé-Hoover thermostat (Popov and Knyazev, 2014) at 310°K, with a coupling constant, τ_T = 0.1 ps. System pressures were semi-isotropically coupled using the Parrinello-Rahman barostat (Parrinello and Rahman, 1981) at 1 bar with a coupling constant, τ_P = 1 ps and compressibility = 4.5 × 10^–5 bar⁻¹. The LINCS algorithm (Hess, 2008) was used throughout to constrain bond lengths. The time step for integration in both simulations was 2 fs. All analyses used GROMACS tools and locally written code.

The final snapshots from Hv1 D GROMOS96 MD simulation was used as the template for the introduction of the R1H mutation using the Mutator plugin (VMD 1.9.2). The final snapshot of the GROMOS96 MD of the Hv1 B model structure (Ramsey et al., 2010) was used as the template for production of Hv1 B N4R using Modeller 9.16 (Sali and Blundell, 1993). All side chains were assumed to have the typical solution pK_a defined in PROPKA (Dolinsky et al., 2004; Olsson et al., 2011), and His residues were modeled with the delta nitrogen (HSD) protonated, which was the ionization state predicted by PROPKA. WT and mutant resting- and activated-state models were subsequently imbedded in a POPC membrane and solvated with a 150 mM KCl solution and energy-minimized in order to remove any unfavorable contacts. After energy minimization POPC lipid tails were allowed to equilibrate around the protein for 0.5 ns, after which the system was simulated according the NPT ensemble with harmonic constraints (5 kcal/mol·Å) applied to the alpha carbons for 1.5 ns. Once the system reached equilibrium, as judged by protein RMSD, stable system volume, and converged energy terms, Hv1-POPC systems were then simulated for 10 ns with the NPT ensemble at 300K. All energy minimizations were carried out in 5000 steps using conjugate gradient and line search algorithms. Simulations were carried out according to the CHARMM36 force field (Best et al., 2012) with the NPT ensemble at 300K and 1 bar using a CUDA build of NAMD 2.10 (Phillips et al., 2005) on a GPU server. Long range electrostatic interactions were calculated using a PME method with a 12 Å cutoff distance. Constant temperature is accomplished using Langevin dynamics and constant pressure control is accomplished using a Nose-Hoover Langevin piston. 2 fs time steps were used. All analysis was carried out in VMD1.9.2. Protein structures were aligned using MultiSeq STAMP (Roberts et al., 2006) implemented in VMD1.9.2; structures in Figure 5—figure supplement 2 were aligned using DeepAlign (http://raptorx.uchicago.edu/DeepAlign/submit/). Structural comparisons to Hv1 FL were conducted on chain A of the dimer, which is not identical to chain B (Li et al., 2015). Coordinates for Ci VSD_U (pdb: 4G7V), Ci VSD_D (pdb: 4G80), mHv1cc chimera (pdb: 3WKV), Kv1.2 (pdb: 3LUT) and At TPC1 DII (pdb: 5E1J and 5DQQ) VS domain X-ray structures are available at http://www.rcsb.org.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

The authors wish to thank Wendy Calchary, Audrey Le and Xuanyu Meng for technical assistance and Louis J De Felice and Carlos A Villalba-Galea for helpful discussions.

© 2016, Randolph et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.