Voltage and pH difference across the membrane control the S4 voltage-sensor motion of the Hv1 proton channel

The voltage-gated proton channel Hv1 is expressed in a variety of cells, including macrophages, sperm, and lung epithelial cells. Hv1 is gated by both the membrane potential and the difference between the intra- and extracellular pH (ΔpH). The coupling of voltage- and ∆pH-sensing is such that Hv1 opens only when the electrochemical proton gradient is outwardly directed. However, the molecular mechanism of this coupling is not known. Here, we investigate the coupling between voltage- and ΔpH-sensing of Ciona intestinalis proton channel (ciHv1) using patch-clamp fluorometry (PCF) and proton uncaging. We show that changes in ΔpH can induce conformational changes of the S4 voltage sensor. Our results are consistent with the idea that S4 can detect both voltage and ΔpH.

Electrophysiological recordings. Prior to recording, oocytes were mechanically devitellinated under a stereoscope and placed in a recording chamber under an inverted IX71 microscope (Olympus, Tokyo, Japan) equipped with a 10 × or 20 × objective. Patch electrodes were pulled from 1.5 mm thick borosilicate glass capillaries (Hilgenberg, Malsfeld, Germany) on a DMZ puller (Zeitz Instruments GmbH, Martinsried, Germany) and subsequently fire polished with a Narishige MF-830 microforge (Narishige, Tokyo, Japan). The resulting initial electrode resistance was 0.6-1.5 MΩ (about 8-30 μm inner tip diameter) in the used recording solutions. Excised macro patches were obtained within seconds to minutes. Holding potentials were − 60 mV or − 80 mV. Recordings were performed at room temperature (RT, 22-25 °C) using an Axopatch 200B amplifier (Molecular Devices, Union City, CA, USA), connected via a Digidata 1440A acquisition board (Molecular Devices) to a PC running the ClampEx software (Molecular Devices). Data were filtered at 2 or 5 kHz and the sampling rate was 10 kHz. Pipette and bath solutions contained (in mM): 100 HEPES, 30 methanesulfonic acid, 5 tetraethylammonium chloride (TEA-Cl), and 5 EGTA, adjusted to pH 7.0 or pH 7.5 with TEA hydroxide (> 25 mM). HEPES was replaced by 2-(N-morpholino)ethanesulfonic acid (MES) in recording solutions adjusted to pH 6.5. Large proton currents generated by the activation of proton channels can lead to accumulation or depletion of protons at either side of the membrane, thereby changing the pH in the vicinity of the proton channel 29 . As a consequence, the electrochemical driving force for protons can change (reduction of outward currents during channel activation), precluding stable recording conditions. In order to keep the pH during Hv1 activation as stable as possible, we used solutions with high pH buffer concentrations (100 mM), used patch pipettes with large tip diameters to obtain a large pH-buffer volume-to-membrane surface ratio, and used oocytes with low Hv1 expression levels to record rather small proton currents. Chemicals were purchased from Sigma-Aldrich ( PCF recordings. On the day of recording, oocytes were labeled at 4 °C for 45-60 min in a solution containing (in mM): 92 KCl, 0.75 CaCl 2 , 1 MgCl 2 , 10 HEPES, and 0.05 (2-((5(6)-Tetramethyl-rhodamine)carboxylamino)ethyl)methanethiosulfonate (MTS-TAMRA), adjusted to pH 7.5 with KOH. Subsequently, oocytes were washed three times in ND96 and stored at 12 °C until recording. Excised patches were obtained under visual control using a 10× objective. The objective was changed to a 60× oil-immersion objective (Olympus Apo N 60XOTIRF NA1.49 or PLAPON 60XOTIRFM NA 1.45). TAMRA was excited with a Spectra X (Lumencor, Beaverton, OR) at 550/15 nm or a Polychrome 5 light source (Till Photonics, Martinsried, Germany), and the fluorescence emission was monitored through a TRITC filter cube (Semrock, Rochester, NY, USA; excitation FF01-543/22-25, dichroic FF562-Di03, and emission FF01-593/40, or excitation FF01-542/20, dichroic FF570-Di01, and emission FF01-620/52) and detected with an iXon Ultra DU-897U or a Luca S emCCD camera (iXon Ultra, Andor Technologies, Belfast, UK). The frame rate was 200 Hz, 8 × 8 pixels were binned and registered in the frame-transfer mode using the camera's conventional output amplifier. Acquisition was triggered externally via the ClampEx software. The light intensities, measured with a PS19Q sensor connected to a FieldMax-TOP power meter for visible light (Coherent, Dieburg, Germany) at the level of the recording stage, were ~ 0.16 mW/ mm 2 .
Uncaging of NPE-caged-proton. 1 Data analysis. Data were analyzed with Igor Pro (Wavemetrics, Portland, OR, USA). Conductance-voltage relationships (GV) were obtained from normalized tail currents (I tail ) measured 5-25 ms after the end of the depolarizing voltage step and fitted with the Boltzmann equation: G/G max = 1/(1 + exp(− (V − V 1/2 )/S)), where G max is the maximal conductance, S the slope, V the membrane voltage, and V 1/2 the voltage at which 50% of the maximal amplitude is reached. From the PCF image frames, the mean (i.e. spatial average) fluorescence intensity was calculated from the brightest pixels (around 5-30 binned pixels, marked with stars in Fig. 2F), and the dark count of the camera was subtracted to obtain the fluorescence F. F was then normalized to the initial baseline level at − 80 mV to obtain ∆F/F. For the average fluorescence traces in different pH conditions (Figs. 2G, 3B, 5C and 6B), only traces in steady-state conditions (i.e. when solution exchange was completed) were used. The voltage-induced change in fluorescence is denoted as F signal , and the amplitude of F signal is reported as the difference between the fluorescence at − 80 mV and the steady-state fluorescence at the end of the voltage step. Activation time constants of current (I) and fluorescence (F) were obtained from double-exponential fits, and the deactivation time constants of F were obtained from mono-exponential fits. The minimum of F was taken as starting point for the deactivation. The relationship between voltage dependence or kinetics and pH or ΔpH was tested with linear regression analysis and tested for significant deviations from a zero-slope line with a significance level of α = 0.05. For ciHv1-I175C-TAMRA, the ΔpH dependence of fast and slow activation time constants of F was compared to the respective slopes of fast and slow activation time constants of I by two-tailed paired t tests with a significance level of α = 0.05. r 2 denotes the coefficient of determination. The ∆pH-induced effects on the F signal amplitude and F(− 80 mV) were tested with one-way ANOVA, followed by Tukey's test post-hoc analysis. All values are reported as mean ± SD.

Results
Voltage dependence of Hv1 is sensitive to ΔpH, but not to pH itself. To study the coupling between voltage-and ΔpH-sensing, we heterologously expressed Hv1 from Ciona intestinalis (ciHv1) in Xenopus laevis oocytes. CiHv1 displays robust heterologous expression 18 and faster activation kinetics than human Hv1, which facilitates data acquisition. Hv1 currents were recorded from excised inside-out membrane patches under various pH conditions (Fig. 1). Voltage steps of increasing amplitudes (up to + 100 mV) gave rise to outward currents (Fig. 1A). From the tail currents (Fig. 1A, arrows), the normalized conductance-voltage relationship (GV) was calculated (Fig. 1B, see methods). We tested the action of ∆pH on the GV relation by changing pH i .  Table 1   www.nature.com/scientificreports/ When pH i was changed from 7.0 to 6.5 (ΔpH = 0.5) or from 7.0 to 7.5 (ΔpH = − 0.5), the voltage of half-maximal activation (V 1/2 ) of the GV relation shifted towards more negative or positive potentials, respectively (Fig. 1B). Similar to previous reports 1,2,24 , the V 1/2 shifted by − 48.6 ± 9.5 mV/ΔpH unit (Fig. 1D,E, Table 1). By contrast, at symmetrical pH (ΔpH = 0), V 1/2 shifted only by − 9.1 ± 4.6 mV/pH unit between pH 6.5 and 7.5 ( Fig. 1C,F,G), indicating that within this pH range, the voltage dependence of ciHv1, like the voltage dependence of its orthologues from human and mice, is relatively insensitive to pH itself (Fig. 1G). Similar to the ΔpH-dependent changes in kinetics seen in voltage-gated proton currents in rat alveolar epithelial cells 24 , we also find ΔpHdependent activation kinetics of ciHv1: the fast and slow activation time constants τ fast and τ slow become faster when ΔpH becomes more positive, i.e. when pH i < pH o (Fig. 1H, Table 2). The relationship between log(τ fast ) and ΔpH, and between log(τ slow ) and ΔpH was linear in the investigated range. As previously reported for human Hv1 30 , we also find that the activation time constants of ciHv1 are faster at acidic symmetric pH than at alkaline symmetric pH ( Fig. 1I and Table 2). However, the dependence of channel activation kinetics on ΔpH is steeper than the dependence of channel activation kinetics on symmetric changes in pH itself (compare Fig. 1H,I).
Taken together, the activation kinetics suggest that channel gating depends more on pH i than on pH o .
The motion of the S4 segment is similar for different symmetric pH conditions. To identify the molecular mechanism of coupling between voltage-and ΔpH-sensing, we studied the underlying conformational changes of Hv1. Previous studies showed that S4 is the main voltage sensor that moves outwardly upon depolarization 18,19,22 . Here, using the PCF technique 22,31,32 , we investigated whether altering ΔpH alone can change the S4 conformation. As readout for conformational changes, we used the thiol-reactive fluorophore MTS-TAMRA ( Fig. 2A). TAMRA is environmentally sensitive: the fluorescence intensity in a polar solvent, like H 2 O, is lower compared to that in a less polar solvent, like ethanol or methanol (Fig. 2B). This property has been used to track conformational changes of ion channels 19,33,34 . By contrast, TAMRA is insensitive to changes in pH: aqueous solutions of pH 6.5 to 7.5 do not change the fluorescence intensity (Fig. 2B). Thus, TAMRA is suitable to track protein conformational changes at different pH. For MTS-TAMRA labeling, a single cysteine was introduced at position L245C, which is located at the extracellular end of S4 35 (Fig. 2C); this site corresponds to F195C used for labeling of human Hv1 19 . The dependence of V 1/2 on ∆pH in this labelled mutant, called ciHv1-L245C-TAMRA, is preserved ( Fig. 2D and Table 1). Fluorescence in voltage-clamped, excised inside-out macro-patches was recorded (excitation light: 550 nm) with a camera (Fig. 2E). The membrane patch was visible as a curved fluorescent stripe (Fig. 2F). Pixels were binned to allow for fast frame rates (200 Hz), and voltage-evoked changes in the fluorescence of the membrane patch (ΔF) were analyzed. A previous study using voltage-clamp fluorometry identified complex voltage-evoked fluorescence signals that were interpreted as two consecutive S4 conformational changes that lead to Hv1 channel opening 36 . Here, we probed S4 conformational changes by stepping from − 80 to − 40 or + 40 mV such that the Hv1 channels preferentially populate the activated closed state or the activated open state, respectively. Upon a voltage step from a holding potential of − 80 to − 40 mV (activated closed state), the fluorescence of ciHv1-L245C-TAMRA decreased. The change was reversible: upon stepping back to − 80 mV, the fluorescence intensity returned to its original value (Fig. 2G, left). This voltage-evoked fluorescence signal (F signal ) is consistent with a previous study 36 Table 1. Fit parameters of the GV relationships of ciHv1 and ciHv1-L245C-TAMRA (mean ± SD). www.nature.com/scientificreports/ and was interpreted as the initial outward S4 motion during voltage sensing. F signal is similar in overall shape for various symmetric pH conditions (pH i = pH o = 6.5, 7.0, or 7.5, thus ΔpH = 0; Fig. 2G, left). The activation kinetics tend to be faster in acidic conditions; the differences were, however, not significant (Fig. 2H). Likewise, the deactivation kinetics do not depend on the pH itself (Fig. 2J). A voltage step from a holding potential of − 80 to + 40 mV (activated open state) led to a more complex F signal : a biphasic decrease of the fluorescence intensity. In addition, upon stepping back to − 80 mV, the fluorescence further decreased and then returned back to baseline, producing a characteristic "hook" in the fluorescence (Fig. 2G, right). Such an F signal was also observed in a previous study 36 and was interpreted as the result of voltagedependent Hv1 channel activation from a resting closed state (high fluorescence intensity) via an activated closed state (low fluorescence intensity) to an activated open state (intermediate fluorescence intensity). Like F signal at − 40 mV, F signal at + 40 mV is similar in overall shape at various symmetric pH conditions (pH i = pH o = 6.5, 7.0, or 7.5; thus ΔpH = 0; Fig. 2G, right). However, some kinetics change for different symmetric pH conditions: at + 40 mV, but not at − 40 mV, the fast and slow activation kinetics of F signal are significantly faster at acidic than at alkaline pH (Fig. 2H,I,  Table 3). By stepping from − 80 mV to − 40 mV, small inward currents were elicited, which indicates channel opening. However, F signal lacked the characteristic "hook" (Fig. 2G, left and Fig. 3A, B), suggesting that F signal evoked by a voltage step to − 40 mV mainly reports on closed-state channel transitions. This is supported  Table 1 Table 3). The dashed lines are linear fits with the following slopes: slope(τ fast ) = 0.5 log(s)/pH unit, r 2 = 0.2, n.s., and slope(τ slow ) = 0.3 log(s)/pH unit, r 2 = 0.1, n.s., at − 40 mV; slope(τ fast ) = 0.5 log(s)/pH unit, r 2 = 0.6, p < 0.05 and slope(τ slow ) = 0.9 log(s)/pH unit, r 2 = 0.5, p < 0.05, at + 40 mV. (J-K) mean deactivation time constants τ deact of F Signal during repolarization from − 40 mV (panel J) or + 40 mV (panel K) to − 80 mV as function of pH (see also Table 3). The dashed lines are linear fits with the following slopes: slope(τ deact ) = 0.03 log(s)/pH unit, r 2 = 0.005, n.s., for − 40 mV; slope(τ deact ) = 0.2 log(s)/pH unit, r 2 = 0.3, n.s., for + 40 mV. Error bars indicate the SD.
Motion of the S4 segment is sensitive to changes in ∆pH. We next tested whether changes in ∆pH affect the voltage-dependent motion of the S4 voltage sensor. While recording from an excised inside-out patch containing ciHv1-L245C-TAMRA, we changed ∆pH by switching to solutions of different pH i , whereas pH o was kept constant (7.0), and repetitively stepped from -80 mV to − 40 mV and back (Fig. 3A). For clarity, averages of the amplitudes of F signal from Fig. 3A are summarized in Fig. 3B. When a more acidic solution was washed in (pH i = 6.5, ΔpH = 0.5), the fluorescence under symmetric pH conditions (pH i = pH o = 7, ΔpH = 0) was drastically altered in several aspects: first, the baseline fluorescence intensity was lowered at − 80 mV, indicating that changes in ΔpH affect the S4 conformation in a non-conducting closed state (Fig. 3A, B). Second, the voltage step to − 40 mV induced an increase rather than a decrease of the amplitude of F signal compared to the ∆pH = 0 condition; upon stepping back to − 80 mV, the fluorescence returned to baseline fluorescence intensity (Fig. 3A,  B). At more alkaline pH i (pH i = 7.5, ΔpH = − .5), the baseline fluorescence and F signal resembled those under ∆pH = 0 conditions (Fig. 3A, B). In Fig. 3C, the mean amplitudes of F ignal (left) and the normalized baseline fluorescence (F (− 80 mV), right) at different ∆pH are summarized. Taken together, this data indicates that changes Table 3. Activation (τ fast , τ slow ) and deactivation (τ deact ) kinetics of the F signal of ciHv1-L245C-TAMRA during a voltage step to − 40 or + 40 mV at different symmetric pH (∆pH = 0; mean ± SD).  www.nature.com/scientificreports/ in pH i that introduce a ΔpH > 0 can alter the S4 conformation in a non-conducting state, suggesting that the S4 conformation is not only sensitive to the membrane potential, but also sensitive to changes in ΔpH.
Proton uncaging in the patch pipette transiently alters pH. So far, we exchanged the pH in the recording chamber by exchanging solution using a gravity-driven perfusion system connected to the recording chamber (Figs. 1B, 2D, 3A). While the exchange of solutions in patch pipettes during a recording is also possible 38 , it is relatively slow and likely to be incomplete. To change the pH (i.e. lower the pH) in the patch pipette during a recording rapidly and without solution exchange (which can introduce artifacts), we used the photolytically cleavable cage 1-(2-nitrophenyl)ethyl sulfate (NPE-caged-proton). NPE-caged-proton releases a sulfate and a proton upon UV irradiation 39 (Fig. 4A). We tested UV-light induced acidification by uncaging NPE-caged-proton at the extracellular side in excised inside-out patches containing ciHv1. Prior to UV light, robust outward proton currents were recorded in response to voltage steps from − 60 to + 60 mV (Fig. 4A). During depolarization, a 1-s long UV-light stimulus first rapidly diminished the outward current, then led to a brief and small inward current, and finally abolished the current (Fig. 4A). The recording is consistent with a UVlight induced extracellular acidification (and hence a change in the electrochemical driving force for protons) and the coupled ΔpH-and voltage-sensing of Hv1, as depicted in Fig. 4B: depolarization to + 60 mV opens the proton channels and leads to a proton outward current (Transition 1). UV-light induced extracellular acidification changes the electrochemical driving force for protons to such an extent (∆pH < − 1) that the proton current reverses direction (Transition 2, indicated by arrow in Fig. 4A). The negative ΔpH also closes the proton channels (Transition 3), implying that the open state of ciHv1 is also sensitive to ΔpH. Thus, it is indeed possible to change the pH in the patch pipette during recording. Subsequent depolarizing voltage steps immediately after UV-light application did not elicit proton currents (Fig. 4C). After several minutes, however, the proton-current amplitude recovered (Fig. 4C), showing that the initial pH at the extracellular side of the membrane (facing the lumen of the patch pipette) is reestablished. This suggests that proton uncaging took predominantly place locally in the direct vicinity of the patch membrane (in the focal plane where UV-light intensity is maximal) and not in the bulk volume of the patch pipette, and that subsequent diffusion of buffer and protons reestablishes the initial pH. If this is the case, NPE-caged-protons should diffuse from the bulk volume back to the vicinity of the patch membrane and allow further local uncaging. Indeed, after several minutes, a second UV-light application could again diminish the proton current (Fig. 4D, E). This also shows that the UV-light stimulus does not degrade the functionality of the ciHv1 proton channel. Next, we tested UV-light induced acidification www.nature.com/scientificreports/ by uncaging NPE-caged-proton at the intracellular side in excised outside-out patches containing ciHv1. Prior to UV light, no or minimal outward proton currents were recorded in response to voltage steps from − 50 to + 5 mV (Fig. 4F). During depolarization, a 1-s-long UV-light stimulus increased the proton-current amplitude significantly (I pre = 0.96 ± 0.9 pA, vs. I post = 5.5 ± 4.1 pA; student´s paired t-test, p = 0.01; n = 6) (Fig. 4F), suggesting that uncaging of NPE-caged-proton acidified the intracellular side of the membrane and increased the chemical driving force for protons and increased the opening probability of Hv1. Taken together, UV-light induced uncaging of NPE-caged-proton can rapidly and transiently acidify the pipette solution at the excised membrane patch.
Motion of the S4 segment can be induced by proton uncaging. Next, we tested whether changing pH by proton uncaging can also induce conformational changes of the S4 voltage sensor. NPE-caged-proton was uncaged at the intracellular side of an outside-out patch containing ciHv1-L245C-TAMRA while at the same time changes of the fluorescence were recorded with PCF (Fig. 5A). The membrane patch was clamped at − 80 mV throughout the experiment, keeping ciHv1-L245C-TAMRA in a non-conducting, closed state (Fig. 5B). During the 1-s long UV-light stimulus, which bled through to the camera, the fluorescence recording could not be interpreted. After the UV-light stimulus, the fluorescence decreased monotonically to a lower steady-state fluorescence intensity (Fig. 5B,D). In contrast, a UV-light stimulus did not change the fluorescence when the patch pipette did not contain caged protons (Fig. 5E), suggesting that it is not the UV-light stimulus directly but indeed the proton uncaging that is responsible for the change in fluorescence. The mean steady-state fluorescence ratio between before and after the light stimulus at − 80 mV (F(− 80 mV, post UV) / F(-80 mV, pre UV)) decreased (Fig. 5D), which is in agreement with the change in the fluorescence ratio in response to intracellular acidification (ΔpH > 0) by gravity-driven perfusion (Fig. 3C, right). Therefore, acidification by proton uncaging corroborates our results from acidification by gravity-driven perfusion, suggesting that the decrease in fluorescence reports on movement of S4 in response to ΔpH. To test whether F signal is also changed by proton uncaging, proton currents were elicited by depolarizing voltage steps during PCF recordings. The mean F signal (same patch as in Fig. 5B) is shown in Fig. 5C. When pH i was lowered by proton uncaging, F signal was altered similarly as seen with gravity-driven perfusion of a solution with a lower pH i : The voltage step to -5 mV induced an increase rather than a decrease of the fluorescence as compared to the ΔpH = 0 condition; upon stepping back to − 80 mV, the fluorescence returned to baseline fluorescence intensity (compare Fig. 5 C with Fig. 3B).Taken together, both methods proton uncaging and gravity-driven perfusion, show that the S4 segment changes its conformation when a ΔpH is applied (by lowering pH i ). ing, Hv1 undergoes another conformational change that can be monitored fluorometrically at the extracellular end of S1 (position I175C in ciHv1) 22 . To test whether a change in ΔpH also directly affects S1 motion, we used mutant ciHv1-I175C to label S1 with TAMRA and performed PCF recordings (Fig. 6A). During recording from an excised inside-out patch containing ciHv1-I175C-TAMRA, we changed ΔpH by switching to solutions of different pH i , while pH o was kept constant (7.0) and repetitively stepped from − 80 mV to + 40 mV and back (Fig. 6A). To better compare outward currents and fluorescence signals, normalized averages of the data from Fig. 6A are shown in Fig. 6B. As reported previously 22 , in response to a voltage step from − 80 mV to + 40 mV, the S1 F signal and the outward current both increase with similar activation kinetics ( Fig. 6A-C, Table 4). In contrast to the S4 F signal , the S1 F signal does not change drastically when ∆pH is changed: the baseline fluorescence at − 80 mV did not change and the sign of the S1 F signal remained positive (Fig. 6D). To show that the direction of current does not determine the sign and kinetics of the fluorescence change, we recorded from the double mutant ciHv1-I153C-I175C-TAMRA (Fig. 6E): as shown previously 22 , the I153C mutation shifts the GV relationship to more negative potentials as compared to wild type, so that inward currents can be recorded at negative membrane potentials. The same is true for ciHv1-I153C-I175C-TAMRA, allowing us to record inward currents at − 10 mV and outward currents at + 10 mV. For both conditions, the S1 F signal recorded at position I175C is positive www.nature.com/scientificreports/ and does not change the sign, confirming that the direction of current does not change the sign of S1 F signal . The kinetics of the S1 F signal and the outward current co-varied for different ΔpH: intracellular acidification accelerated and alkalization decelerated kinetics of both the S1 F signal and the outward current (Fig. 6C). Thus, current and S1 F signal kinetics stay coupled even when ΔpH is changed, corroborating that ∆pH-and voltage-sensing are linked to each other via the S4 voltage sensor.

Discussion
Several ion channels are gated or allosterically modulated by pH 40 . For some channels (e.g. two P-domain K + channels), the amino-acid residues that convey pH sensitivity, presumably by titration of acidic or basic side chains, are well characterized 41,42 . However, Hv1 is modulated by ΔpH rather than pH i or pH o itself. Although some mutations have been reported to affect ΔpH sensing of Hv1 26,28,30 , a clear identification of amino-acid residues that constitute the ΔpH sensor(s) is lacking. Here, we provide direct evidence that the S4 segment alters its conformation in response to changes in ΔpH, suggesting that S4 serves as both voltage-and ΔpH-sensor. During voltage-dependent gating, Hv1 proceeds through multiple transitions 36,37,[43][44][45] . VCF recordings show that membrane depolarization moves the S4 voltage sensor to an activated state, followed by a second transition that leads to the open state 36 . Taking advantage of precise intra-and extracellular pH control in the PCF configuration, we show that the transitions monitored at S4 are, except for the activation kinetics, relatively similar for different symmetric pH conditions (Fig. 2G, ΔpH = 0), suggesting that the voltage dependence of voltage-sensor motion, like voltage-dependent gating, does not depend on pH itself. By contrast, asymmetric pH conditions can lead to a drastic change of the F ignal on S4 (Figs. 3 and 5), suggesting that voltage and ΔpH are the key stimuli to drive S4 conformational changes.
Recently, two studies investigated Hv1 mutants with diminished ionic currents and could identify gating currents, which reflect the movement of charged amino-acids side chains across the electric field of the membrane 44,45 . Interestingly, gating currents of human Hv1-W207A-N214R were shown to be sensitive to pH o : extracellular acidification by one pH unit shifts the gating charge-voltage relationship towards positive membrane potentials by 40 mV 45 . Because S4 contains the majority of gating charges 43 , the pH o sensitivity of gating currents suggests that the S4 conformation depends on pH o . Our results show that the S4 conformation is also sensitive to pH i . This is in agreement with a previous study that suggested that the gating transitions for Hv1 activation depend on pH i 37 . Taken together, these results are consistent with the idea that excessive protons at the intracellular side (low pH i ) or a positive membrane potential push S4 to the extracellular side; excessive protons at the extracellular side (low pH o ) or a negative membrane potential push S4 to the intracellular side (Fig. 7A). Therefore, we suggest that both ΔpH and voltage determine the position of S4 in the membrane (Fig. 7B). Hv1 might enter states by changes in ΔpH that are different from states by changes in membrane potential. Whether ΔpH sensing is restricted to only certain conformational states, remains unclear and thus a subject for future studies. Our PCF experiments provide strong evidence that S4 conformation can be changed by pH i in the resting-state (Figs. 3,5). Our proton uncaging experiments show that the open state is sensitive to changes in pH o (Fig. 4A). In a previous study of human Hv1, mutation of the first of the three arginines of the S4 segment to histidine ("R1H") creates an additional, hyperpolarization-activated proton conductance, termed shuttle conductance 46 . The shuttle conductance was found to be pH o -sensitive, which might reflect the ΔpH-sensitivity of the S4 segment for transitions between resting (non-open) states at negative membrane potentials.
Different molecular mechanism of ΔpH sensing are currently under debate 47 . Protonation or deprotonation of amino-acid residues can induce conformational changes in proteins. In Hv1, there might be distinct, protonable amino-acid residues serving as pH sensors at either side of the membrane for detecting pH o and pH i , respectively 24 . S4 contains only three arginines as potentially titratable amino-acid side chains, and those might be difficult to deprotonate under physiological conditions. If any of those arginines were deprotonated over the tested pH range, we would expect a significant change in voltage sensitivity, which should result in shallower slopes of the GVs. However, slopes for different pH conditions did not change significantly (Table 1). Alternatively, the negatively charged amino-acid residues in transmembrane segments S1-S3, which serve as countercharges of the arginine residues, might participate in pH sensing 48 ; titration of those residues might weaken their interaction with the arginine residues and, therefore, influence the S4 conformation. However, because ΔpH and not the pH itself determines gating, sensors for pH o and pH i need to be influenced by or functionally linked to each other. The strict dependence of V 1/2 on ΔpH, which is remarkable linear over a large pH range 26 , argues against titratable amino-acid residues as pH sensors: the pK a s of the intra-and extracellular sensor would need to be coupled in a symmetric fashion which appears implausible.
It has been suggested that water molecules could be an important component in mediating ∆pH sensitivity. Molecular dynamics simulations suggest that the core of the VSD of Hv1 contains more water molecules than Table 4. Activation kinetics of the current and F signal of ciHv1-I175C-TAMRA recorded at pH o = 7.0 (mean ± SD).  26 . Indeed, several solvent accessible amino-acid residues are located deep inside the crevice of the VSD 49 . The water molecules might form a robust water wire network with limited mobility within the VSD. Protonation of this network could lead to a rearrangement of the hydrogen bonding and exert electrostatic forces that change S4 conformation 26 . Such a mechanism is not localizable to the Hv1 channel itself; it would be a distributed effect acting on the channel. A previous study on Zn 2+ sensitivity in Hv1 provides evidence for allosteric coupling between intra-and extracellular residues induced through coulombic interactions 50 . This supports the idea that changes in electrostatic forces might be processed over a long distance across the membrane. We did not find evidence for direct ∆pH-induced conformational changes of the S1 segment, reinforcing the idea that the mechanisms of ΔpH-and voltage-sensing are intimately linked to each other via the S4 voltage sensor. However, we cannot rule out the possibility that the S4 segment is coupled to other -still unknown-∆pHsensing elements. Further studies are therefore needed to further characterize the ∆pH-induced conformational changes of the S4 segment as well as possible additional conformational changes involving other segments.
It is not known whether only the S4 segment of Hv1 is ΔpH sensitive, or whether the S4 segments of other voltage-gated ion channels are also ∆pH sensitive. The ΔpH-induced S4 movement during non-conducting states suggests that actual proton permeation is not a prerequisite for ∆pH sensing. Possibly, changes in ∆pH can induce motions in the VSDs of other voltage-gated ion channels that do not necessarily couple to channel opening and therefore might have escaped from being noticed in electrophysiological recordings. The higher water densities in Hv1 compared to other VSDs might play a key role for ∆pH sensing by enabling an interaction of protons with S4.
Our data suggest that the functional coupling between pH o and pH i sensor(s) is provided by the mobile S4 voltage sensor, serving as a whole as the "∆pH-sensing element". In this setting, the electrochemical potential for protons across the membrane, together with the membrane potential, determines the position of S4 in the membrane (Fig. 7B), and thereby sets the voltage of half-aximal channel activation. Of note, the ∆pH-induced shift of V 1/2 , reported to amount to − 0 mV/∆pH unit in the literature and to − 47 to − 57 mV/∆pH unit for the channels investigated in our study, is close to the electrochemical potential for protons (-59 mV/∆pH unit), suggesting an immediate link between gating and the permeant ion. Deviations might arise from differences between ion concentration and activity. Mutations that limit water exposure to S4 should decrease ∆pH sensing; those mutations, however, are likely to limit conduction as well. Clearly, the identification of the detailed molecular mechanism behind ∆pH sensing of Hv1 requires further studies using electrophysiological, proton uncaging, and fluorometric techniques on mutants of Hv1 as well as modeling approaches that take the potential profiles around Hv1 into consideration. . S1-S3 are omitted for clarity. " + " signs denote the charged arginines in S4. (B) cartoon depicting how S4 position is determined by both voltage and ∆pH across the membrane. Protons might exert electrostatic forces on Hv1, i.e. by protonation of a water wire in the VSD. The position of the mobile S4 segment depends on both, the electrochemical potential for protons and the membrane potential: excessive protons at the extracellular side (pH o < pH i ) and/or hyperpolarization push S4 to the intracellular side, stabilizing the closed state. Excessive protons at the intracellular side (pH o > pH i ) and/or depolarization push S4 to the extracellular side, stabilizing the activated state.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.