External Cd²⁺ and protons activate the hyperpolarization-gated K⁺ channel KAT1 at the voltage sensor

Eukaryotic voltage-gated cation channels share a core subunit structure consisting of a four-transmembrane helix (S1–S4) voltage sensor domain (VSD) linked to a two-transmembrane helix (S5–S6) pore domain. These channels function as tetramers (MacKinnon, 1991), with four independent voltage sensors surrounding a single tetrameric ion pore (Lee et al., 2005; Long et al., 2005a; Whicher and MacKinnon, 2016). While eukaryotic voltage-gated channels ultimately have a common evolutionary origin, they are structurally and functionally diverse and comprise multiple lineages traceable to distinct prokaryotic channel ancestors (Brams et al., 2014; Jegla et al., 2018; Jiang et al., 2001; Liebeskind et al., 2013; Loukin et al., 2005). Plant voltage-gated K⁺ channels, including the KAT1 channel from the model organism Arabidopsisthaliana, belong to a lineage referred to here as the cyclic nucleotide binding domain or CNBD superfamily that has eubacterial origins and arose from the fusion of the voltage-gated channel core to a cyclic nucleotide-binding domain in the cytoplasmic C-terminus (Brams et al., 2014; Jegla et al., 2018). This cytoplasmic domain is present in and diagnostic for all eukaryotic CNBD superfamily channels, but in some gene families, such as metazoan Ether-a-go-go (EAG) K⁺ channels, it regulates gating but no longer binds cyclic nucleotides (Brelidze et al., 2009). It is therefore sometimes referred to as a cyclic nucleotide-binding homology domain or CNBHD (Brelidze et al., 2012). Eukaryotic CNBD superfamily channels are also functionally diverse in terms of ion selectivity and voltage gating. For example, metazoan HCN and CNG channels are nonselective cation channels (Ludwig et al., 1998; Santoro et al., 1998; Zagotta and Siegelbaum, 1996) rather than K⁺ channels. Some CNBD superfamily channels are essentially voltage insensitive (metazoan CNG channels; Zagotta and Siegelbaum, 1996), while others are activated by depolarization (EAG family K⁺ channels and the plant voltage-gated K⁺ channels GORK and SKOR; Ache et al., 2000; Gaymard et al., 1998; Robertson et al., 1996) or hyperpolarization (HCN channels and plant voltage-gated K⁺ channels including KAT1; Ludwig et al., 1998; Santoro et al., 1998; Schachtman et al., 1992; Sentenac et al., 1992).

Voltage-gating mechanisms have been extensively studied in CNBD superfamily channels, but the structural basis for the opposite gating phenotypes in depolarization-gated and hyperpolarization-gated channels is not yet well understood. The S4 transmembrane helix, which contains basic (positive) gating charges, moves outward during depolarization for both depolarization-gated (EAG family; Bannister et al., 2005; Silverman et al., 2003; Whicher and MacKinnon, 2016) and hyperpolarization-gated (HCN1 and KAT1; Latorre et al., 2003; Lee and MacKinnon, 2017; Männikkö et al., 2002) channels, meaning that depolarization-gated channels activate when the S4 is “up” and hyperpolarization-gated channels, including HCNs and KAT1, activate when the S4 is “down.” A recent cryo-EM analysis revealed that when S4 is held down in HCN1 with an artificially introduced Cd²⁺ bridge between S4 and the neighboring transmembrane domain S2, the S4 helix bends on the cytoplasmic side into two helices, displacing the inner end of the pore’s S5 helix, and thus providing a possible mechanism to allow pore opening (Lee and MacKinnon, 2019). A similar S4 bend and S5 movement has been observed in molecular dynamics simulations of HCN1 under a hyperpolarizing electric field (Kasimova et al., 2019). However, the HCN1 pore was closed in both studies, so the mechanism for hyperpolarization gating in HCN channels remains unclear. Moreover, it remains to be determined if KAT1 and HCN channels share common gating mechanisms. The hyperpolarization gating in plant and animal CNBD family channels may have evolved independently, given the distinct lineage-specific phylogenetic origins of each channel type (Baker et al., 2015; Jegla et al., 2018). A recent KAT1 cryo-EM structure in a closed S4-up state shows remarkable homology to up-state structures of animal CNBD family channel structures (Clark et al., 2020), but it lacks a serine residue in S4 that is critical for the S4 bend that might mediate activation in HCN1. This suggests that S4-down states of KAT1 may be more similar to those of EAG family channels despite opposite gating polarity.

Much of what we know about VSD arrangement in closed S4-down states in CNBD superfamily channels comes from studies of external divalent cation and proton block of EAG family channels, comprising the Eag (Kv10), Erg (Kv11), and Elk (Kv12) gene subfamilies. Divalent ion inhibition was first studied in Drosophila melanogaster Eag, where external divalent cations such as Mg²⁺ and Ni²⁺ block outward movement of S4 by binding to a cluster of acidic residues in the S2 and S3 transmembrane helices located in a proposed externally accessible aqueous cleft in the core of the VSD (Silverman et al., 2000, 2004; Tang et al., 2000). This cleft has since been observed in all crystal and cryo-EM structures of voltage-gated channels (Lee and MacKinnon, 2017; Lee et al., 2005; Long et al., 2005a; Whicher and MacKinnon, 2016) and, in combination with a mirrored cleft on the intracellular side, serves to focus the transmembrane electric field to a narrow band in the middle of the voltage sensor (Long et al., 2005b; Whicher and MacKinnon, 2016). Basic S4 gating charges traverse through the clefts during gating and interact with the acidic residues in the extracellular cleft in the S4-up state (Long et al., 2005a, 2005b; Silverman et al., 2003; Whicher and MacKinnon, 2016). These interactions observed in the S4-up state could at least in part explain why divalent block is state dependent, with a preference for an S4-down conformation (Silverman et al., 2004). The external cleft’s cluster of acidic charges is conserved across the EAG family, and divalent block has since been observed in Erg and Elk channels, though each subfamily has a distinct divalent cation susceptibility profile (Fernandez et al., 2005; Jo et al., 1999; Zhang et al., 2009). In all three subfamilies, protons also inhibit voltage activation across a physiological pH range for metazoans (pH 6–8) via the external acidic charge cluster (Kazmierczak et al., 2013) and compete with divalent cations for the site (Jo et al., 1999; Kazmierczak et al., 2013; Terlau et al., 1996). Neutralization of any of the three acidic residues in the external cleft by mutation also inhibits activation of EAG family channels (Fernandez et al., 2005; Kazmierczak et al., 2013; Silverman et al., 2000; Zhang et al., 2009). The exact mechanism for divalent block of outward S4 movement remains unsolved, and it is possible that multiple proposed mechanisms, such as electrostatic charge repulsion (Silverman et al., 2000, 2003), narrowing of the cleft (Schönherr et al., 2002), and direct interactions with S4 (Tang et al., 2000; Zhang et al., 2009), come into play for various divalent ion/channel combinations. Independently from divalent cation block, the sensitivity of EAG family channel gating to protonation or neutralization of residues comprising the external acidic charge cluster underscores its importance in determining the energy barrier for S4 movement, perhaps by altering the local electric field or providing key salt bridges for S4 gating charges during or after transition of S4 to the up state observed in structures (Lee and MacKinnon, 2017; Wang and MacKinnon, 2017; Whicher and MacKinnon, 2016).

The question we address here using the plant voltage-gated K⁺ channel KAT1 is whether hyperpolarization-gated cation CNBD superfamily channels traverse to or through a VSD conformation homologous to the divalent ion/proton-induced S4-down state of depolarization-gated EAG family channels during voltage activation. If so, then protons and/or divalent cations should enhance voltage activation of KAT1 through interaction with a binding site in the external pocket of the VSD domain (Fig. 1 A). S4-up-state structures of hyperpolarization-gated channels (KAT1 and HCN1) confirm that these channels share the external aqueous cleft found in EAG family channels (Clark et al., 2020; Lee and MacKinnon, 2017). We chose KAT1 for this study because it lacks an extended string of basic S4 charges that are uniquely present in HCN channels (Fig. 1 C, dotted underline). The extra HCN S4 charges are predicted to interact with the external acidic cluster in the S4-down state (Lee and MacKinnon, 2019) and could potentially interfere with divalent ion binding. While external Cd²⁺ can access an introduced binding site at the base of the external VSD cleft in an S4-down state of HCN1 (Lee and MacKinnon, 2019), WT HCN1 is indeed minimally sensitive to Cd²⁺ (Bell et al., 2009). Whether or not a native divalent ion binding site exists in the external VSD of KAT1 is also an open question since KAT1 has only a single acidic residue (D95 in S2) at the S2/S3 charge cluster sites conserved in divalent ion/proton-sensitive EAG channels (Fig. 1 B). Neutralization of this residue (D95N) enhances KAT1 activation by depolarizing the voltage-activation (G-V) curve (Lefoulon et al., 2014), suggesting that D95 is important for transition to or stabilization of the closed S4-up states. In KAT1, neutral residues occupy the positions of the two other external cluster acidic residues conserved across the EAG family (N99 in S2 and Q149 in S3; Fig. 1 B). Nevertheless, in the S4-up-state structure of KAT1, these three residues reside in close proximity within the external cleft and are positioned to interact with S4 charges during gating (Clark et al., 2020).

Despite the absence of two of the three acidic residues found in EAG family channels, external acidification does activate KAT1 and closely related channels from other plant species by depolarizing the G-V curve (Hedrich et al., 1995; Hoth et al., 1997; Hoth and Hedrich, 1999; Wang et al., 2016). Pore residues implicated in pH regulation of KAT1-related channels do not account for the KAT1 G-V shift at low pH (Hoth et al., 1997; Hoth and Hedrich, 1999). A recent study identified a voltage sensor region overlapping D95 as responsible for differences in pH regulation between KAT1 and ZmK2.1, a KAT1-like channel from maize (Wang et al., 2016), but it did not examine the role of charge at the D95 position and did not interpret the results in the context of current accepted models of VSD structure. Zn²⁺ activates a chimera between KAT1 and the related channel KDC1, causing a small (∼10 mV) depolarization of the G-V curve, but the effect appears to be absent in KAT1 itself, and pore/voltage sensor residues outside the VSD pocket and specific to KDC1 have been implicated (Picco et al., 2004, 2008). Divalent cations can access the external aqueous cleft of KAT1 because a high-affinity Cd²⁺ binding site that promotes KAT1 activation can be engineered with an introduced cysteine on S4 (Clark et al., 2020), but the effects of Cd²⁺ on the native channel at concentrations more similar to those that modulate EAG family channels have not been quantified. We show here that Cd²⁺ is a potent activator of KAT1, depolarizing the G-V curve by 150 mV, and that D95 is required for modulation of KAT1 voltage gating by both Cd²⁺ and external protons. Our results suggest that KAT1 does indeed have a divalent ion/proton-sensitive S4-down state in its activation pathway.

An Arabidopsisthaliana KAT1 clone was amplified by PCR to add a Kozak consensus sequence (CCACC) immediately upstream of the initiator methionine and transferred into the Xenopus laevis oocyte expression plasmid pOX (Jegla and Salkoff, 1997) by restriction digestion and ligation using HindIII and XbaI sites introduced by PCR upstream of the Kozak sequence and downstream of the stop codon, respectively. A clone confirmed to be identical to the KAT1 reference sequence (GenBank accession no. NM_123993) was used in all subsequent experiments. KAT1 mutations were introduced using the Quickchange site-directed mutagenesis method (Stratagene), and sequences were confirmed. The following primers were used for mutagenesis: D95N, sense 5′-GCG​ATT​TTC​ATC​ATC​AAC​AAC​ATT​GTT​AAT​GG-3′, antisense 5′-CCA​TTA​ACA​ATG​TTG​TTG​ATG​ATG​AAA​ATC​GC-3′; N99D, sense 5′-CAT​CGA​CAA​CAT​TGT​TGA​TGG​CTT​CTT​CGC​CAT-3′, antisense 5′-ATG​GCG​AAG​AAG​CCA​TCA​ACA​ATG​TTG​TCG​ATG-3′; and Q149E, sense 5′-CCA​CAG​CAC​CAT​TTG​ACC​CAC​TAA​GCC​TCT​TG-3′, antisense 5′-CAA​GAG​GCT​TAG​TGG​GTC​AAA​TGG​TGC​TGT​GG-3′.

Xenopus ovaries shipped in saline were sourced from Xenopus 1 or Nasco. Ovaries were digested with 1–2 mg/ml type II collagenase (Sigma-Aldrich) in calcium-free saline (98 mM NaCl, 2 mM KCl, 1 mM MgCl₂, and 5 mM HEPES, pH 7.2) to release oocytes and remove the follicle cell layer. After digestion, oocytes were thoroughly rinsed to remove the collagenase and cultured for 1–3 d at 18°C in the above solution supplemented with 1.8 mM CaCl₂, 2.5 mM Na-pyruvate, 100 U/ml penicillin, and 100 µg/ml streptomycin (Sigma-Aldrich). For some batches of oocytes, 50 µg/ml tetracycline (Sigma-Aldrich) was added to eliminate infection from bacteria resistant to penicillin and streptomycin. Mature oocytes (stage V or VI) were injected with ∼15–75 ng in vitro transcribed cRNA in ∼50 nl with a Nanoject II injector (Drummond Scientific). Optimal cRNA concentrations and postinjection incubation times (1–3 d) for electrophysiological experiments were empirically determined. We made capped cRNA transcripts for injection from NotI linearized expression plasmid templates using the T3 mMessage mMachine system (Life Technologies). Transcripts were purified using lithium chloride precipitation, resuspended in nuclease-free water supplemented with a 1:20 dilution of SUPERase-In (Life Technologies), analyzed by gel electrophoresis to confirm synthesis, quantified using a Nanodrop spectrophotometer (Thermo Fisher Scientific), and stored at −80°C before injection.

Channel function was assessed using standard two-electrode voltage clamp methods using a CA-1B amplifier (Dagan) and the pClamp 10/Digidata 1440A suite (Molecular Devices) for data collection and initial analysis. Most electrophysiological data were sampled at 10 kHz and filtered at 5 kHz using the amplifier’s four-pole Bessel filter. For the time course of Cd²⁺ modification used to determine state dependence, data were sampled at 1 kHz. Borosilicate glass electrodes were prepared using a P-1000 Micropipette Puller (Sutter Instruments), filled with 3 M KCl, and had tip resistances of 0.3–1.5 MΩ. Bath electrodes were placed in 1 M NaCl and separated from the perfusion chamber using a 1-M NaCl agarose bridge. Recordings were made at room temperature (21–24°C). Base recording solution consisted of 78 mM NaOH, 2 mM NaCl, 18 mM KOH, 2 mM KCl, 1 mM CaCl₂, and 5 mM HEPES, adjusted to the indicated pH with methane sulfonic acid, making methanesulfonate the major anion. For solutions of pH <6, 5 mM Na-PIPES was used in place of 5 mM HEPES, and NaOH was adjusted to 73 mM. Solution pH was calibrated before recording sessions. CdCl₂, MgCl₂, MnCl₂, NiCl₂, and ZnCl₂ were added to the indicated concentrations from 1-M water-based stocks. Oocytes were placed in a low-volume OPC-1 chamber (AutoMate Scientific) for recording, and were constantly perfused via a ValveLink DTM 8 gravity perfusion system (AutoMate Scientific) with a flow rate of ∼1 ml/min (>15 bath volume exchanges/min). Chemicals for solutions were sourced from Sigma-Aldrich.

The KAT1 Cd²⁺ affinity was estimated using an adjusted Hill equation fit of the ΔV₅₀ (V₅₀[Cd²⁺] − V₅₀[control]) calculated for each tested Cd²⁺ concentration: ΔV₅₀ = start + (end − start)([Cd²⁺]^H/(K_d^H + [Cd²⁺]^H), where H is the Hill slope and K_d is the midpoint [Cd²⁺]. The baselines of the Hill fits were adjusted to account for small negative shifts in V₅₀ that occurred at low Ca²⁺ concentrations.

Reported activation time constants were derived from dual exponential fits of the activation time course, I(t) = A₁[1 − e^(−t/τ1)] + A₂[1 − e^(−t/τ2)] + b, where I(t) is the current at time t, A₁ and τ₁ are the amplitude and time constant of the fast component, A₂ and τ₂ are the amplitude and time constant of the slow component, and b is the baseline. To obtain the most accurate fit of the major components of activation, the fit window excluded the small sigmoidal delay that precedes activation and the small, slow increases in current size that occur after 600 ms. Deactivation time constants at pH 7.0 and pH 4.0 were determined from single exponential fits of tail current time course, I(t) = A[1 − e^(−t/τ) + b], while deactivation at pH 7.0 in the presence of 300 µM Cd²⁺ was fitted with two exponentials using the same equation described above for activation. To facilitate comparisons of the total deactivation time course between groups fitted with different numbers of exponentials, we used T_0.5, the time at which currents reached 50% of the amplitude calculated from exponential fits. The apparent gating charges associated with activation/deactivation components were estimated from single exponential fits of the voltage dependence of the time constant, τ(V) = τ(0)e^zVF/RT, where τ(V) is the time constant at voltage V, τ(0) is the time constant at 0 mV, z is the equivalent charge movement, T is the temperature, F is the Faraday constant, and R is the ideal gas constant.

Reported error values are SEM unless otherwise specified. Statistical significance and P values for pairwise comparisons were assessed using two-sample unpaired t tests.

Fig. S1 presents the effect of external Ca²⁺ concentration on the KAT1 V₅₀ in control conditions and in 300 µM Cd²⁺. Fig. S2 plots the contribution of native oocyte currents to the total outward currents recorded in 300 µM Cd²⁺ at extreme depolarized voltages.

Because the specificity of divalent cation block varies among Eag, Elk, and Erg subfamily channels, we first conducted a survey of extracellular divalent cations for the ability to activate KAT1 at pH 7.0 in 20 mM extracellular K⁺. Voltage activation of KAT1 did not appear to be sensitive to our standard bath concentration of Ca²⁺ (1 mM), because reducing bath Ca²⁺ to 200 µM did not noticeably alter KAT1 currents elicited by hyperpolarizing voltage ramps (Fig. 2 A) or KAT1 V₅₀ determined from voltage steps (Fig. S1). We therefore used 1 mM Ca²⁺ in our bath solutions for subsequent experiments to optimize recording stability at hyperpolarized voltages. We did not examine KAT1 sensitivity to lower Ca²⁺ concentrations, because we often observed significant leak current at hyperpolarized voltages with <200 µM Ca²⁺. Fig. 2, B–F, shows example KAT1 currents induced by hyperpolarizing voltage ramps in the presence of 300 µM Cd²⁺, Zn²⁺, Mn²⁺, Ni²⁺, and Mg²⁺. Cd²⁺ caused a large increase in KAT1 current across the entire voltage ramp and also at the −20 mV holding potential, indicating a large depolarizing shift in voltage activation (Fig. 2 B). This activating effect was reversible and could be washed off by reapplying control solution (Fig. 2 B). The other divalent cations we tested did not activate KAT1, showing little effect or, in the case of Ni²⁺ and Zn²⁺, causing a noticeable reduction in current (Fig. 2, C and E). We cannot rule out the possibility that some of these other divalent cations might be able to activate KAT1 in the absence of Ca²⁺, but Cd²⁺ was notable as the only divalent ion we tested that could robustly activate KAT1 at tractable Ca²⁺ concentrations. We therefore focused on Cd²⁺ for detailed analyses of divalent ion effects on KAT1 gating.

We used 2-s voltage steps from a holding potential of 0 mV to generate G-V curves for KAT1 at pH 7.0 in the absence or presence of 300 µM Cd²⁺. Fig. 3, A and B, shows example currents for both conditions from the same oocyte. In the control condition, KAT1 channels are closed at voltages above the reversal potential (approximately −40 mV in our 20 mM extracellular K⁺ recording solution), so no outward currents are apparent in the voltage steps, and inward currents activated by hyperpolarization increase from a near-0 current baseline. In contrast, in the presence of 300 µM Cd²⁺, outward KAT1 currents are readily apparent at 0 mV; most of the inward current elicited during hyperpolarizing steps below reversal potential is instantaneous; and voltage steps above the reversal potential elicit large, instantaneous outward currents that decay slowly with depolarization >40 mV. These observations indicate that in the presence of 300 µM Cd²⁺, most KAT1 channels are already open at 0 mV, but that channels can be closed by strong depolarization. This is consistent with a large depolarizing shift in voltage activation in the presence of 300 µM Cd²⁺. Large native currents elicited >120 mV (Fig. S2) limited the voltage range we were able to test in G-V analyses, making it difficult to determine whether all KAT1 channels closed in response to depolarization. Nevertheless, we were able to determine a G-V relationship for the majority of channels that did close.

KAT1 G-V curves recorded at pH 7.0 from isochronal tail currents at −100 mV after 2-s voltage steps with and without 300 µM Cd²⁺ are shown with single Boltzmann fits in Fig. 3 C and do indeed reveal a large depolarizing shift in the G-V in the presence of Cd²⁺. KAT1 traverses multiple voltage-dependent closed states during activation (Latorre et al., 2003; Zei and Aldrich, 1998), so the G-V shape is not expected to precisely fit a single Boltzmann distribution. However, we found that single Boltzmann fits were sufficient to allow quantitative comparison of G-V curves obtained in different conditions. 300 µM Cd²⁺ caused a 149.8 ± 2.3-mV shift in the midpoint of the V₅₀ calculated from the single Boltzmann fit of voltage activation (see Table 1 for Boltzmann fit parameters). V₅₀ calculated from Boltzmann fits was dependent on the Cd²⁺ concentration (Fig. 3, C and D). A Hill equation fit of ΔV₅₀ versus Cd²⁺ concentration (Fig. 3 D) yielded a Cd²⁺K_d of 114.3 µM with a Hill slope of 2.2, suggesting multiple binding sites with some degree of cooperativity.

In Eag channels, the rate of divalent block is dependent on the availability of the sensitive S4-down state (Silverman et al., 2003; Terlau et al., 1996). We therefore compared the rate of Cd²⁺ activation of KAT1 at 0 and −80 mV to examine state dependence. The control KAT1 G-V recorded in the absence of Cd²⁺ (Fig. 3 C) predicts a large increase in channel open probability at −80 mV relative to 0 mV, suggesting that the Cd²⁺-sensitive S4-down state should be significantly more available at −80 mV. For KAT1, we found an approximately threefold increase in the rate of Cd²⁺ activation at −80 mV compared with 0 mV (Fig. 4, A and B). At −80 mV, the KAT1 current increased from −1.90 ± 0.15 µM to −6.41 ± 0.27 µM (n = 12) after addition of 300 µM Cd²⁺, which qualitatively reflects the change of open probability from ∼0.3 to ∼1.0 after application of 300 µM Cd²⁺ at −80 mV observed during G-V analyses (Fig. 3 C). In contrast, we observed much less current increase than predicted by change in open probability between control and 300 µM Cd²⁺ at 0 mV. The open probability reached ∼0.8 after addition of 300 µM Cd²⁺ in G-V analyses (Fig. 3 C). Given a similar driving force at 0 and −80 mV in 20 mM bath K⁺, one might expect ∼5-μA current after addition of 300 µM Cd²⁺ in the state-dependent experiment shown in Fig. 4 A. However, we observed only a current increase from 0.02 ± 0.004 to 0.54 ± 0.03 µA (n = 12) in the experiment. These data are nevertheless consistent with a Cd²⁺ binding site present in an S4-down conformation of KAT1 that occurs more frequently at −80 mV than at 0 mV. We hypothesize that at 0 mV, voltage sensors do not enter the Cd²⁺-sensitive S4-down state frequently enough to allow opening of the majority of KAT1 channels.

We next examined the effect of Cd²⁺ on the activation rate of KAT1. We reasoned that if the Cd²⁺ binding occurs in an S4-down state of the VSD, then the activation rate of KAT1 in 300 µM Cd²⁺ should show reduced voltage dependence because some VSDs might already be trapped in this S4-down state, and thus channel activation would require less charge movement. Because open probability is high at 0 mV in 300 µM Cd²⁺, a 2-s voltage pulse to 80 mV was necessary to close a sufficient number of channels for analysis of activation in subsequent hyperpolarizing steps. Because KAT1 has a significant open probability at 80 mV in in 300 µM Cd²⁺ (Fig. 3 C), Cd²⁺ is able to bind to KAT1 at this voltage, and it is therefore very likely that some VSDs in channels closed by the 80-mV prepulse are still Cd²⁺ modified, allowing us to measure the effect of Cd²⁺ on the voltage sensitivity of activation. A comparison of normalized example traces for control and 300 µM Cd²⁺ is shown in Fig. 5 A. We found that two exponentials accurately described the bulk of the activation time course in both the absence and presence of Cd²⁺ (Fig. 5 A, inset). We did not attempt to fit the previously noted sigmoidal delay in KAT1 activation (Latorre et al., 2003; Zei and Aldrich, 1998), which becomes too short to precisely quantify in two-electrode voltage clamp with large hyperpolarizing steps, and we did not fit small, slow increases in current occurring after 600 ms that were variable from oocyte to oocyte. 300 µM Cd²⁺ strongly promotes KAT1 activation, but the overall activation time course in Cd²⁺ is nevertheless slower at voltages more negative than −130 mV. Plots of the fast (τ_F) and slow (τ_S) time constants of activation versus voltage are shown for KAT1 with and without 300 µM Cd²⁺ in Fig. 5, B and C. τ_F and τ_S have less voltage dependence in the presence of Cd²⁺, and single exponential fits of the data (smooth curves) show a reduction in the equivalent charge movement (z) associated with activation. In control conditions, the total equivalent charge movement associated with these two activation components is 2.63e (0.86e + 1.77e for the fast and slow componentsb, respectively; Fig. 5, B and C), which is qualitatively similar to estimates of charge movement associated with KAT1 activation as derived from limiting slope measurements of open probability (Latorre et al., 2003) and detailed kinetic models (Zei and Aldrich, 1998). This suggests that conformational transitions responsible for τ_F and τ_S account for a significant fraction of the voltage sensor charge that moves during KAT1 activation. The more than threefold reduction in equivalent charge associated with activation that we observed in 300 µM Cd²⁺ (2.63e to 0.81e; Fig. 5, B and C) is consistent with a Cd²⁺ binding site in an S4-down conformation and fewer VSDs moving from an S4-up to an S4-down conformation during activation. Alternatively, Cd²⁺ might bind to transitional S4 conformations in which some but not all the gating charge has crossed the membrane electric field. An effect of this reduction in voltage sensitivity of activation is a slowing of the activation time course, because the time constants of activation in 300 µM Cd²⁺ become progressively larger than in control conditions as hyperpolarization is increased. The activation time course in Cd²⁺ is further slowed because the fractional amplitude of τ_F is reduced at all voltages (Fig. 5 D).

We found that Cd²⁺ also substantially slowed the rate of KAT1 deactivation (Fig. 6, A and B). Before deactivation, KAT1 was activated by a 1.35-s step at −140 mV with and without 300 µM Cd²⁺. In control conditions, the KAT1 deactivation time course was well fitted by a single exponential (Fig. 6 A, inset). This is consistent with the KAT1 gating model presented by Zei and Aldrich (1998) in which a single transition dominates the deactivation time course. The time constant of deactivation (τ_Deact) is plotted versus voltage in Fig. 6 A and fitted with a single exponential that yielded an equivalent charge movement of 0.59e. The deactivation time course in 300 µM Cd²⁺ required two exponentials to accurately fit (Fig. 6 B), and the smaller-amplitude fast component did not show a simple exponential relationship to voltage. We hypothesize that this component could derive from Cd²⁺ dissociation rather than a true gating phenomenon. A single exponential fit of the slow component (τ_Deact-S) yielded an equivalent charge movement of 0.42e, a modest reduction from control (0.59e) compared with what we observed for activation. If deactivation is dominated by the first closing transition as predicted by Zei and Aldrich (1998), then it is likely to occur upstream of the major charge-carrying outward S4 movements that occur during depolarization, and it would therefore make sense for Cd²⁺ to have less impact on the voltage dependence of deactivation. Although the voltage dependence of deactivation was not substantially altered by Cd²⁺, the deactivation time course was nevertheless much slower than what we observed for controls, implying that Cd²⁺ binds to an S4 conformation that occurs during the open state. Because of the large G-V shift in 300 µM Cd²⁺, we obtained measurements for both conditions only at a single voltage (+70 mV) where control τ_Deact was 7.5 ± 0.3 ms and τ_Deact-S in 300 µM Cd²⁺ was 1,157 ± 81 ms, a >100-fold difference. This comparison overestimates the overall difference in deactivation time course because it ignores the fast component observed only in Cd²⁺. We therefore calculated the time point of 50% deactivation (T_0.5). At 70 mV, the only voltage for which we could measure deactivation in both control conditions and 300 µM Cd²⁺, T_0.5 was 5.7 ± 0.2 and 298 ± 13 ms for control and Cd²⁺ conditions, respectively. This is still a much greater difference than we observed for the activation time course (Fig. 5). Fig. 6 C shows a plot of T_0.5 for deactivation versus voltage in which the values for 300 µM Cd²⁺ have been repositioned by −150 mV to account for the ∼150-mV G-V shift we observe in 300 µM Cd²⁺. Deactivation in Cd²⁺ remains approximately two- to threefold slower in Cd²⁺ even with this −150-mV adjustment.

We tested Cd²⁺ sensitivity of KAT1 D95N to assess whether the Cd²⁺ binding site resides in the external cleft of the VSD. The D95N KAT1 mutant has previously been reported to activate at more depolarized potentials than WT KAT1 (Lefoulon et al., 2014), consistent with a role for D95 in stabilizing a closed S4-up state of the VSD. We did not observe major changes in KAT1 D95N currents recorded in control conditions (pH 7) compared with 300 µM Cd²⁺ (Fig. 7, A and B), and the large depolarizing G-V shift we observed for WT KAT1 in 300 µM Cd²⁺ (Fig. 3 D) was absent in the D95N mutant (Fig. 7 C and Table 1). Because KAT1 D95N is easier to open than KAT1 WT, the lack of Cd²⁺ activation in this mutant is unlikely to be caused by the state dependence of Cd²⁺ binding we observed for KAT1 (Fig. 4). Note that we did observe a small inhibition of KAT1 D95N current and a small left shift of the G-V in the presence of Cd²⁺ (−3.8 ± 3.2 mV). We observed a similar inhibition and shift in WT KAT1 at Cd²⁺ concentrations below those required to activate KAT1 (Fig. 3 D). We reasoned that this minor effect of Cd²⁺ must involve a different site, and we did not study it further. The complete loss of Cd²⁺ potentiation indicates that D95 is required for the Cd²⁺ binding conformation.

We next examined the pH sensitivity of KAT1 gating and Cd²⁺ modulation. Example currents from the same oocyte are shown for KAT1 at pH 7.0 and pH 4.0 in Fig. 8, A and B. As expected from previous studies (Hedrich et al., 1995; Hoth et al., 2001; Hoth and Hedrich, 1999), KAT1 currents were strongly potentiated at pH 4. G-V curves for KAT1 are shown for pH 8–4 in Fig. 8 C, and Boltzmann fit parameters are given in Table 2. ΔV₅₀ values (relative to pH 8) are shown for pH 7–4 in Fig. 8 D. V₅₀ was significantly right shifted at pH values <6, with most of the shift occurring at pH 4 (Fig. 8 D), indicating a low proton affinity compared with EAG family channels, which have large G-V shifts in the pH 6–8 range (Kazmierczak et al., 2013; Li et al., 2015). External acidification from pH 7.0 to 4.0 induces a 49.1 ± 0.5-mV depolarizing shift of KAT1 V₅₀ and a 30.8 ± 4.9% increase of current amplitude at −160 mV. The current size increase at −160 mV in pH 4.0 is larger than expected from the change in open probability of ∼0.9 at pH 7.0 to ∼1 at pH 4.0, suggesting additional alterations of KAT1 gating beyond the G-V shift. We did not examine the state dependence of proton modulation, because KAT1 responses to pH were rapid compared with changes induced by Cd²⁺, making it possible for bath exchange rate to contribute to the observed modification rate. Note that the pH effect is unlikely to be saturated at pH 4, but we were not able to test lower pH values because of oocyte instability.

We compared activation and deactivation kinetics of KAT1 currents recorded at pH 4.0 and pH 7.0 (Fig. 9, A–F). At pH 7.0, the majority of the activation time course at pH 4.0 was well described by a dual exponential fit (Fig. 9 A), and protons accelerated both the fast (τ_F; Fig. 9 B) and slow (τ_S; Fig. 9 C) components. Protons had less effect than Cd²⁺ on the apparent gating charge (z) calculated from exponential fits of τ_F and τ_S versus voltage (Fig. 9, B and C) or on the fractional amplitude of the components (Fig. 9 D). This may simply reflect the smaller G-V shift at pH 4.0 compared with 300 µM Cd²⁺. Our results suggest that even at pH 4.0, most KAT1 VSDs are likely to sit in a closed S4-up conformation at the 0-mV resting potential used here, consistent with the fact there is no instantaneous current at 0 mV holding voltage (Figs. 8 B and 9 A, inset). The deactivation time course of KAT1 in pH 4.0 fit well with a single exponential (Fig. 9, E and F), and exponential fits of τ_Deact versus voltage (Fig. 9 F) showed similar equivalent gating charges associated with deactivation at pH 7.0 and pH 4.0. In contrast to what we observed for Cd²⁺ in Fig. 6 C, the slowing of deactivation at pH 4.0 could be almost entirely accounted for by the ∼50-mV G-V shift between pH 7.0 and pH 4.0 (Fig. 9 F). For example, τ_Deact at 0 mV and pH 4.0 (101.5 ± 7.2 ms) is similar to τ_Deact at −50 mV at pH 7.0 (93.0 ± 5.1 ms). These results indicate that protons may simply restrict S4 movement less than Cd²⁺ in KAT1. Indeed, in the human Erg1 channel, protons appear to access the external acidic cluster in both S4-up and S4-down conformations (Kazmierczak et al., 2013; Li et al., 2015; Shi et al., 2019).

We assessed the pH sensitivity of KAT1 D95N and the pH sensitivity of Cd²⁺ binding to examine whether the proton binding sites responsible for the shift in KAT1 G-V overlap with the Cd²⁺ binding site in the external aqueous pocket of the VSD. Fig. 10, A and B, shows example traces for KAT1 D95N at pH 7.0 and pH 4.0, and Fig. 10 C compares G-V curves for D95N for pH 8.0–4.0. Parameters for the G-V curves in Fig. 10 C are given in Table 2, and KAT1 D95N ΔV₅₀ values (relative to pH 8) are compared with WT ΔV₅₀ values for pH 7–4 in Fig. 10 D. For KAT1 D95N, only pH 4.0 causes a statistically significant shift in V₅₀ compared with pH 8. However, the ΔV₅₀ from pH 8.0 to pH 4.0 was significantly reduced in the D95N mutation, to 12.9 ± 1.2 mV from 48.8 ± 0.7 mV in WT. D95N also eliminated the current increase we observed in WT KAT1 at −160 mV and pH 4; in D95N, we instead observed a small 12.1 ± 2.6% (n = 6) decrease in current at −110 mV (equivalent G-V position to −160 mV in WT) upon switching from pH 7.0 to pH 4.0. D95 therefore plays a central role in both Cd²⁺ and proton potentiation of KAT1 activation. This D95 dependence suggests that the binding site for both cations lies in the external cleft of the VSD. Cd²⁺ binding was highly pH sensitive (Fig. 10, E and F), as expected from the observed overlap in divalent ion and proton modulation sites observed for EAG channels (Terlau et al., 1996; Kazmierczak et al., 2013). However, pH 5.0 was sufficient to eliminate potentiation of KAT1 by 300 µM Cd²⁺ (Fig. 10 E). We observed only a small left shift in G-V (−10.9 ± 0.7 mV) in 300 µM Cd²⁺ at pH 5.0 (Fig. 10 E), similar to what we observed at low Cd²⁺ concentrations and pH 7.0 for WT (Fig. 3 D) and D95N (Fig. 8 E). We performed dose–response experiments for Cd²⁺ at pH 9–5 and used Hill equation fits to estimate Cd²⁺ affinity (Fig. 10 F and Table 3). Cd²⁺ affinity ranged from 17.3 µM at pH 9.0 to 6.4 mM at pH 6, but the maximal effect of Cd²⁺ remained qualitatively similar, suggesting direct rather than allosteric competition between protons and Cd²⁺ within this pH range. Cd²⁺ potentiation at pH 5.0 was apparent only in 30 mM Cd²⁺, and we could therefore not obtain a full dose–response curve for estimation of affinity. Our results show that the Cd²⁺ binding site overlaps a high-affinity proton binding site and that Cd²⁺ binding can be eliminated by protonation of this site. This site appears distinct from the low-affinity proton binding site responsible for KAT1 activation, because the pH range for proton/Cd²⁺ competition and proton-dependent activation of KAT1 show little overlap. Our results therefore indicate the presence of two independent proton binding sites in the external VSD of KAT1 with distinct affinities.

To gain further evidence that the proton binding site that mediates KAT1 activation is in the outer voltage sensor, we examined whether introduction of acidic residues into the KAT1 VSD at positions N99 in S2 and Q149 in S3 alters proton sensitivity of KAT1 gating, since acidic residues at these positions play a key role in divalent ion and proton sensitivity of EAG family channels. Example traces for KAT1 N99D and Q149E channels recorded at pH 7.0 are shown in Fig. 11, A and B, and G-V values measured at pH 7.0 are compared with those of KAT1 WT in Fig. 11 C. Boltzmann fit parameters for N99D and Q149E KAT1 channel mutants are reported in Table 1. N99D channels had a small but significant hyperpolarizing shift in V₅₀ compared with WT, while Q149E had an ∼25-mV depolarizing shift in V₅₀. Both mutants showed a significant increase in pH sensitivity compared with WT, as shown in plots of ΔV₅₀ for pH 7–4 relative to pH 8.0 for N99D, Q149E, and WT channels in Fig. 11 D. While the first significant shift in G-V relative to pH 8.0 occurs for WT at pH 5.0 (Fig. 8 D), N99D and Q149E channels already had significant shifts relative to pH 8.0 at pH 6. ΔV₅₀ relative to pH 8.0 was significantly increased for Q149E mutants compared with WT at all pH values, while N99D mutants showed significant increases compared with WT at pH 7, 6, and 5. While the inability to test pH values <4 prevented us from accurately measuring the acid dissociation constant (pK_a), the increased sensitivity of these mutants to external pH changes in the pH 8–5 range suggests higher proton affinity for KAT1 activation. These results for N99D and Q149E do not differentiate an increase in the affinity of the native low-affinity proton binding site from a possible introduction of a new higher-affinity proton binding site that can independently mediate activation, but they do provide additional support for the external VSD cleft as the relevant proton modulation site for KAT1 voltage gating. N99D and Q149E channels also exhibited significant reductions relative to WT in the ΔV₅₀ for 300 µM Cd²⁺ at pH 7, to 31.4 ± 4.4 and 89.8 ± 2.7 mV, respectively (Fig. 11, E and F; and Table 1). A detailed examination of the mechanism for this reduction in Cd²⁺ sensitivity is beyond the scope of this study but suggests the potential for interaction between the VSD’s high- and low-affinity proton binding sites, or alternatively, structural rearrangements within the VSD of these mutants that could indirectly reduce Cd²⁺ affinity. Regardless, the result provides further support for the external VSD as the site of Cd²⁺ modulation.

Our results provide evidence that the plant hyperpolarization-gated CNBD superfamily K⁺ channel KAT1 traverses a divalent ion/proton-sensitive VSD conformation in its gating pathway. Several lines of evidence suggest that this conformation is homologous to the divalent ion/proton-sensitive conformation of metazoan depolarization-gated EAG family K⁺ channels. First, the voltage sensors of KAT1 and EAG family channels share a common voltage sensor response to depolarization: outward movement of S4 (Latorre et al., 2003; Schönherr et al., 2002). Because S4 movement is oppositely coupled to pore status in hyperpolarization-gated and depolarization-gated channels (Latorre et al., 2003; Männikkö et al., 2002), outward movement of S4 leads to opening of EAG channels and closing of KAT1. If divalent cations lock the VSD in a similar S4-down conformation, then this opposite coupling can explain why divalent cations and protons inhibit EAG family channels but activate KAT1. For both channel types, there is evidence from gating currents, S4 accessibility studies, gating models derived from structure function analysis, and direct fluorescence measurements of S4 movement that the outward S4 movement involves at least two sequential steps (Bannister et al., 2005; Latorre et al., 2003; Silverman et al., 2003; Zei and Aldrich, 1998). Mg²⁺ slows the most hyperpolarized detectable component of S4 movement in Drosophila Eag, which occurs before the outward S4 motion that corresponds to the charge movement detected in gating currents (Bannister et al., 2005). Our results suggest that Cd²⁺ binds to a similar deep S4-down state in KAT1, because Cd²⁺ reduces the voltage dependence of KAT1 activation (Fig. 5), an indication that the binding site resides in a conformation downstream of a significant fraction of the inward gating charge movement. This is consistent with the observation that charge movements captured in KAT1 gating currents correspond to early closed-closed transitions rather than channel opening (Latorre et al., 2003). While examining gating currents was beyond the scope of this study, our proposed model suggests that Cd²⁺ should reduce KAT1 On gating currents. KAT1 VSDs are very likely to sit in a Cd²⁺-sensitive conformation in the open state of the channel, since Cd²⁺ significantly slows deactivation (Fig. 6). The closing transition in KAT1 does not appear to correlate with Off gating charge movement (Latorre et al., 2003). However, the slower overall rate of activation at hyperpolarized potentials in Cd²⁺ (Fig. 5) indicates that the Cd²⁺ binding site might also be available in a closed state proximal to the open state. The results also do not rule out the presence of Cd²⁺ in the external VSD pocket in closed S4-up conformations but do suggest that Cd²⁺ binding must be preferential or must preferentially restrict gating movements in deep S4-down conformations.

Mutagenesis studies have firmly established the acidic charge cluster in the external aqueous cleft of the VSD as the site for divalent ion/proton inhibition of voltage gating in diverse EAG family channels, although the number of proton binding sites in the cleft has not been characterized. Here we find evidence for two proton binding sites in the external aqueous cleft of KAT1, a high-affinity binding site that lies within the Cd²⁺ binding site and a low-affinity site that mediates proton potentiation of KAT1 gating. In principle, both proton binding sites could lie within the Cd²⁺ binding pocket, since Cd²⁺ prefers at least four coordination points (Rulísek and Vondrásek, 1998; Seebeck et al., 2008), but we were unable to examine the possible contribution of the low-affinity proton binding site to Cd²⁺ coordination since loss of the high-affinity site is sufficient to eliminate Cd²⁺ modulation. A role for the single acidic residue conserved in the external aqueous cleft of KAT1, D95, in proton binding is suggested by the fact that it is required for both Cd²⁺ and proton potentiation (Figs. 8 and 10). While it is tempting to speculate that D95 would also contribute to the lower-affinity site since it has an acidic side chain, pK_a values of residues within proteins are context dependent and notoriously hard to predict, and our data do not resolve the issue. D95 might be required for both proton binding sites because it holds the VSD in a divalent ion/proton-sensitive conformation or contributes to a second divalent ion binding site by altering the pK_a of neighboring pocket residues in S2–S4. Additionally, Cd²⁺ and proton coordination points could be contributed by neighboring hydrogen bonding amino acid side chains on S2–S4 or even water molecules in the cleft itself. The recent cryo-EM structure of KAT1 (Clark et al., 2020) does not resolve which other polar residues might contribute to the proton binding sites we identified, because it was captured in an S4-up state. However, N99 and Q149, which we found influence both proton and Cd²⁺ modulation (Fig. 11), are two of the polar residues that lie within the pocket in proximity to D95 in this structure. Note that there is some residual G-V shift at pH 4.0 in the D95N mutant. Because we did not explore this residual pH sensitivity further, we cannot differentiate whether the native low-affinity site retains some pH sensitivity in the N95 background or whether there is an additional proton binding site with a minor effect on pH sensitivity of the KAT1 G-V. While our results stop short of conclusively identifying the proton and Cd²⁺ binding residues, they nevertheless strongly suggest that the KAT1 Cd²⁺/proton modulation sites are homologous to the divalent ion binding site of Eag channels in that they lie within the external aqueous cleft of the VSD and favor the S4-down conformation.

The KAT1 proton modulation site in the VSD has a distinctly lower affinity than the proton modulation site in the VSD of EAG family channels, where proton inhibition is significant between pH 8.0 and 6, with a pK_a more than pH 6.5 calculated for Elk channels (Kazmierczak et al., 2013). A possible explanation for the difference in proton affinity between the low-affinity proton binding site mediating activation in KAT1 and the binding sites that mediate inhibition of EAG channels is the different number of acidic residues in the external VSD pocket in EAG channels compared with KAT1. While EAG family channels contain a cluster of three acidic residues in S2 and S3 that project into the external VSD aqueous pocket (Whicher and MacKinnon, 2016), KAT1 contains a single predicted pocket-facing acidic residue in S2 (Fig. 1). Clusters of acidic side chains that allow carboxy-carboxylate pairing often have higher pK_a values compared with isolated acidic residues (Harris and Turner, 2002), and this has been proposed as a possible explanation for the high pK_a of the EAG family proton modulation site (Kazmierczak et al., 2013). The >6 pK_a for proton activation in a metazoan ASIC family acid-sensing channel appears to depend on carboxy-carboxylate pairing (Jasti et al., 2007). Here we found that separate insertion of additional acidic residues into KAT1 at N99 and Q149 (the S2 and S3 positions where acidic residues are found in EAG family channels) enhanced proton sensitivity of KAT1 activation at pH 5–7, implying increased proton affinity (Fig. 11). In EAG channels, the acidic residues at the equivalent positions of D95, N99, and Q149 in KAT1 are close enough to each other to contribute to coordination of a divalent cation (Fernandez et al., 2005; Silverman et al., 2000; Zhang et al., 2009), so it seems reasonable to postulate that they could be close enough to each other to alter site pK_a by exchanging or sharing protons. The S4-up state structure of KAT1 shows that D95, N99, and D149 do indeed lie in close proximity within the external aqueous cleft (Clark et al., 2020), suggesting that they have the potential to sit within hydrogen bonding distance in S4-down states. However, we cannot rule out that D99 and E149 alter the pH sensitivity of KAT1 independently from interactions with D95 or that EAG family channels simply lack the second low-affinity proton binding site we identified in KAT1. In the latter case, proton effects on EAG channel gating might be mediated by a site homologous to the high-affinity proton binding site we identified in KAT1 that alters Cd²⁺ modulation but does not promote activation.

The low affinity of the proton binding site that promotes KAT1 activation could represent an adaptation for its role in K⁺ uptake into guard cells that helps drive osmotic uptake of water to promote stomatal opening. Microscopic stomatal pores are the conduit through which plants both take up CO₂ for photosynthesis and lose water vapor to the atmosphere. External (apoplastic) pH surrounding plant cells is typically much lower than that in metazoans (Barbez et al., 2017; Felle and Hanstein, 2002; Sharp and Davies, 2009), and KAT1 appears tuned for proton modulation within this pH range. Stomatal opening is initiated by proton efflux (Assmann et al., 1985; Shimazaki et al., 1986), which has a hyperpolarizing effect on membrane voltage and thus promotes KAT1 activation to increase K⁺ uptake. In addition, this proton efflux may decrease apoplastic pH (Felle and Hanstein, 2002) and thus further enhance KAT1 activation (González et al., 2012; Hoth and Hedrich, 1999). While an in vivo role for pH sensitivity of KAT1 has not yet been definitively established, proton activation of native inward K⁺ currents has been recorded from guard cells of several plant species (Blatt, 1992; Ilan et al., 1996; Müller-Röber et al., 1995), including observation of an analogous pH-dependent shift in G-V typified by external proton acceleration of current activation and slowing of current deactivation (Blatt, 1992). Drought, which opposes stomatal opening and thus reduces transpirational water loss from the plant during water stress, often elicits increases in external pH (Hartung et al., 1988; Sharp and Davies, 2009). It therefore seems likely that pH-dependent tuning of KAT1 constitutes a mechanism for active pH responsiveness with functional advantages.

Structural and functional analyses show a remarkable degree of conservation in voltage sensor conformation and gating movements among CNBD superfamily channels, despite an impressive diversity of voltage-gating phenotypes. Indeed, the S4-up state conformation of KAT1 determined by cryo-EM is essentially the same as the S4-up conformations observed for EAG channels despite over a billion years of evolutionary divergence and opposite voltage-gating phenotypes (Clark et al., 2020; Wang and MacKinnon, 2017; Whicher and MacKinnon, 2016). Here we show that the VSD domain of the plant hyperpolarization-gated channel KAT1 adopts a divalent ion/proton-sensitive conformation in and/or near the open state, analogous to the divalent ion/proton-sensitive deep closed conformation of the depolarization-gated EAG family K⁺ channels of metazoans. Whether this conformation is relevant to gating in all voltage-sensitive CNBD superfamily channels remains unknown. HCN channels, while divalent insensitive, are strongly activated by low pH (Stevens et al., 2001), although the mechanism has not been established. The role of the VSD in gating of both prokaryotic CNBD channels and the large family of plant cyclic nucleotide-gated channels has not yet been addressed. Furthermore, much of the sequence diversity of CNBD superfamily channels lies in single-celled eukaryotes such as ciliate protozoans and algae (Aury et al., 2006; Haynes et al., 2003; Jegla et al., 2018; Jegla and Salkoff, 1994, 1995; Merchant et al., 2007), and the structure–function relationships of these channels have yet to be examined. Divalent ion specificity and sensitivity are modulated by the composition of residues at the external acidic cluster positions as well as the outer edge of S4, and these regions are not highly conserved across the CNBD superfamily, or even within the plant voltage-gated K⁺ channel family (Jegla et al., 2018). Modulation of the VSD by divalent ions, while a useful tool for biophysical analysis of gating and pH sensitivity of EAG family channels and KAT1, is unlikely to play a conserved physiological role in the regulation of most CNBD superfamily channels. We suggest that physiologically relevant proton sensitivity is likely to be more widespread because proton binding may have looser structural requirements and may occur in multiple states. The external aqueous cleft of the VSD of any channel that gates via voltage-dependent displacement of S4 gating charges is likely to have some form of a proton-sensitive hydrogen bonding/charge network for up-state stabilization. Whether this network mediates physiologically significant pH effects could simply depend on the pK_a of the constituent proton binding sites.

Christopher J. Lingle served as editor.

We thank Benjamin Simonson and Aditya Pisupati their assistance.

Work was supported by NSF Division of Integrative Organismal Systems (16–21027 to T. Jegla and S.M. Assmann).

The authors declare no competing financial interests.

Author contributions: Y. Zhou designed experiments, collected, analyzed, and interpreted data, and wrote the manuscript. T. Jegla conceived the study, analyzed and interpreted data, and wrote the manuscript. S.M. Assmann interpreted data and wrote the manuscript.