Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels contribute to the rhythmic firing of pacemaker neurons and cardiomyocytes. Mutations in HCN channels are associated with cardiac arrhythmia and epilepsy. HCN channels belong to the superfamily of voltage-gated K+ channels, most of which are activated by depolarization. HCN channels, however, are activated by hyperpolarization. The mechanism behind this reversed gating polarity of HCN channels is not clear. We here show that sea urchin HCN (spHCN) channels with mutations in the C-terminal part of the voltage sensor use the same voltage-sensor movement to either close or open in response to hyperpolarizations depending on the absence or presence of cAMP. Our results support that non-covalent interactions at the C-terminal end of the voltage sensor are critical for HCN gating polarity. These interactions are also critical for the proper closing of the channels because these mutations exhibit large constitutive currents. Since a similar voltage-sensor movement can cause both depolarization- and hyperpolarization-activation in the same channel, this suggests that the coupling between the voltage sensor and the pore is changed to create channels opened by different polarities. We also show an identical voltage-sensor movement in activated and inactivated spHCN channels and suggest a model for spHCN activation and inactivation. Our results suggest the possibility that channels open by opposite voltage dependence, such as HCN and the related EAG channels, use the same voltage-sensor movement but different coupling mechanisms between the voltage sensor and the gate.
Introduction
Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels generate currents (called If, Ih, or Iq) that contribute to the initiation and modulation of the slow depolarization between action potentials in cardiac and neuronal pacemaker activity (Biel et al., 2009; Brown et al., 1979; Difrancesco, 1993; Gauss et al., 1998; Halliwell and Adams, 1982). Many mutations in the genes encoding HCN channels cause cardiac arrhythmia and neurological diseases, such as epilepsy (Biel et al., 2009; Rivolta et al., 2020). Activated by hyperpolarization and modulated by cyclic adenosine monophosphate (cAMP), HCN channels conduct Na+ and K+ ions, resulting in a reversal potential of around −20 mV under physiological conditions. HCN channels are members of the superfamily of voltage-gated K+ (Kv) channels. HCN channels have four subunits and each subunit consists of six transmembrane segments, S1–S6 (Lee and MacKinnon, 2017). The S1–S4 segments in each subunit form the peripheral voltage-sensing domain (VSD), whereas the S5 and S6 segments form the central pore domain (PD; Fig. 1; Lee and MacKinnon, 2017, 2019; Saponaro et al., 2021). The S4 segment harbors several positively charged residues and therefore senses voltage changes across the cell membrane. However, in contrast to most Kv channels that are activated by depolarizations (voltage changes to more positive voltages), HCN channels are activated by hyperpolarizations (voltage changes to more negative voltages). In addition, HCN channels have a cyclic-nucleotide binding domain (CNBD; Chen et al., 2001; Gauss et al., 1998; Ulens and Siegelbaum, 2003; Wang et al., 2001; Fig. 1). In mammalian HCN channels, cAMP binds to CNBD and increases the open probability and shifts the voltage dependence to more positive voltages by inducing movements in the C-linker that connects the CNBD and the pore (Fig. 1). In sea urchin HCN (spHCN) channels, cAMP removes a rapid inactivation that is only seen in spHCN channels (Gauss et al., 1998).
The reversed gating polarity of HCN channels remains unclear, although different mechanisms have been proposed (Cowgill et al., 2019; Flynn and Zagotta, 2018; Männikkö et al., 2002; Vemana et al., 2004). In contrast to most Kv channels, HCN channels have a non-swapped structure (Lee and MacKinnon, 2017), i.e., the VSD from one subunit is next to the PD from its own subunit (Fig. 1). However, that is not the reason for its reversed voltage dependence because Kv10.1 and Kv11.1 have non-swapped structures (Wang and MacKinnon, 2017; Whicher and MacKinnon, 2016) but are depolarization-activated channels. One proposed mechanism is that the reversed gating is due to a similar S4 movement in both Kv and HCN channels but a reversed VSD–PD coupling in HCN channels (Männikkö et al., 2002; Vemana et al., 2004): the depolarization-activated Kv channels open as S4 moves upward upon depolarizations (Long et al., 2005; Lu et al., 2002), whereas the hyperpolarization-activated HCN channels open as S4 moves downward upon hyperpolarizations (Dai et al., 2019; Lee and MacKinnon, 2019; Männikkö et al., 2002; Vemana et al., 2004; Wu et al., 2021). Zagotta and his colleagues (Flynn and Zagotta, 2018) proposed that the S4 C-terminal end acts as an inhibitory domain on the pore when S4 is in the upstate, and that upon hyperpolarization, the downward movement of S4 relieves this inhibitory effect and allows the pore to open. Our recent studies supported this model by showing that a hydrogen bond between a glutamate residue in the C-terminal end of S4 and an asparagine residue in S5 stabilizes S4 in its upstate and the gate in its closed state (Fig. 1; Ramentol et al., 2020). Another mechanism was proposed by Chanda et al. (Cowgill et al., 2019) suggesting that a bipolar S4 in HCN channels could move upward and downward in response to depolarization and hyperpolarization, respectively, thereby opening the pore from a common closed state with S4 in an intermediate position. In this model, mammalian HCN channels have two different opening states, a hyperpolarization-activated open state and a depolarization-activated open state—in which the depolarization-activated open channels quickly undergo inactivation thereby making the channels only hyperpolarization-activated. However, in some chimeras of HCN and the related depolarization-activated hERG channels, the depolarization-activated open states are very prominent. Note that HCN channel structures have been solved for three states (S4 up/closed [Lee and MacKinnon, 2017], S4 down/closed [Lee and MacKinnon, 2019], and S4 up/open [Saponaro et al., 2021]; Fig. 1) with the last two states being rare under physiological conditions. There is still no structure of the open state with S4 down, which is the most prominent open state under physiological conditions. Therefore, it is not clear how HCN channels open by downward S4 movement during hyperpolarizations.
The spHCN channel undergoes a cAMP-dependent inactivation (Gauss et al., 1998): in the absence of cAMP, spHCN channels undergo a fast inactivation and cAMP removes this inactivation. Different mechanisms underlying this inactivation have been proposed (Dai et al., 2021; Flynn and Zagotta, 2018; Shin et al., 2004). An early model (Shin et al., 2004) suggested a decoupling between the voltage sensor and the gate because the same gate seems to underlie closing by deactivation and closing by inactivation. Another mechanism (Flynn and Zagotta, 2018) suggested a separate voltage sensor movement for inactivation, separate from the voltage sensor movement that opens the channels because inactivation shows a steep voltage dependence. A recent study (Dai et al., 2021) suggested a third mechanism with an additional voltage sensor-to-gate coupling in the absence of cAMP involving a movement of the C-linker that recloses the gate (inactivating the channels) after downward S4 movement has opened the gate. Note that there is no structure of an inactivated spHCN channel. So, it is not clear how inactivation is triggered or what generates the voltage dependence of inactivation in spHCN channels.
We hypothesize that a similar S4 movement occurs in both depolarization-activated channels (such as EAG and mutant HCN channels) and hyperpolarization-activated channels (such as wt HCN channels) but that EAG channels open when S4 moves up and wt HCN channels open when S4 moves down. We also hypothesized that a similar S4 movement underlies the voltage dependence of the inactivation of spHCN channels.
We show here that spHCN channels with mutations at the C-terminal end of S4 are activated by depolarization rather than hyperpolarization in the absence of cAMP. cAMP removes this depolarization-dependent activation and even enables the mutants to be activated by hyperpolarization. This suggests that the depolarization-dependent activation in the mutants is due to recovery from inactivation. Furthermore, the S4 movement in response to a hyperpolarization is similar in hyperpolarization-activated wt spHCN channels and depolarization-activated spHCN mutants, as well as in inactivated spHCN channels in the absence of cAMP. This inward S4 movement leads to an opening in hyperpolarization-activated wt spHCN channels in the presence of cAMP, closing in depolarization-activated spHCN mutants, and inactivation in wt spHCN channels in the absence of cAMP. This suggests that the reversed gating polarity of spHCN channels and the inactivation of spHCN channels are not due to different S4 movements but due to differences in the VSD–PD coupling. Here, we propose a novel model for spHCN channel gating with similar S4 movement in closed, open, and inactivated channels.
Materials and methods
Molecular biology
The gene codifying for the sea urchin (Strongylocentrotus purpuratus) HCN (spHCN) channel was in the pGEM-HE expression plasmid. All mutations were introduced using QuikChange site-directed mutagenesis kit (Qiagen). In vitro spHCN cRNA was transcribed using mMessage mMachine T7 RNA Transcription Kit (Ambion). cRNA at 1–5 µg/μl was injected into defolliculated Xenopus laevis oocytes (Ecocyte). The oocytes were incubated in ND96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl2, 1.8 mM CaCl2, and 5 mM HEPES; pH = 7.5) for 2–3 d for membrane expression of the channels. 1 mM ZD7288 (Tocris Bioscience) was added to the bath solution. In voltage clamp fluorometry experiments, to test the voltage-sensor movement of spHCN channels with high cAMP, oocytes were injected with 50 nl of 5 mM cAMP and incubated at 10°C for 15 min prior to electrophysiological recordings.
Voltage clamp fluorometry (VCF)
After 2–3 d of incubation, oocytes were labeled with 100 µM Alexa-488 C5-maleimide (Molecular Probes) for 30 min at 4°C. Following labeling, they were kept on ice to prevent the internalization of labeled channels. Oocytes were recorded in ND96 solution. 100 μM LaCl3 was used to block endogenous currents induced by hyperpolarized voltages (Ramentol et al., 2020). Whole-cell ionic currents were measured with the two-electrode voltage clamp technique using an Axon Geneclamp 500B amplifier (Axon Instruments, Inc.). Data were filtered at 1 kHz, digitized at 5 kHz (Axon Digidata 1322 A), and monitored and collected using pClamp software (Axon Instruments, Inc.). Fluorescence signals were low-pass Bessell filtered (Frequency Devices) at 200 Hz and digitized at 1 kHz. From a holding potential of −10 mV, steps from +100 mV to −160 mV (in −20 mV steps) were applied to activate the S4 movement and current of the channel followed by a tail voltage of +40 mV to obtain the tail current. We measured the conductance–voltage (GV) relation of channels by measuring the tail currents at +40 mV. We measured the fluorescence–voltage (FV) relation of channels by measuring the steady-state fluorescence signal upon activation at different voltages.
Patch clamp
Currents were measured in the inside-out configuration using an Axon 200B amplifier and pClamp 9.0 (Axon Instruments, Inc.). Patches were recorded with 1–2 MΩ pipettes containing, in mM: 89 KCl, 15 HEPES, 0.4 CaCl2, and 0.8 MgCl2 (pH = 7.4 adjusted with KOH). The intracellular solution contained, in mM: 100 KCl, 10 HEPES, 1 EGTA, and 0.1 CaCl2, and, in µM: 0 or 100 cAMP (pH 7.1 adjusted with KOH). The internal solution was continuously perfused to the patches using a rapid perfusion system (Biologic Science Instruments RSC-160). From a holding potential of −80 mV, steps from +100 mV to −160 mV (in −20 mV steps) were applied to activate the current of the channel followed by a tail voltage of +40 mV to obtain the tail current. We measured the GV relation of channels by measuring the tail currents at +40 mV.
Molecular modeling
The fluorescence and currents for the 15-state model of the spHCN channel (Fig. 10) were simulated using the Berkeley Madonna program (University of California, Berkeley). The model is based on the 10-state (5 closed and 5 open states) allosteric Altomare model of mammalian HCN channels (Altomare et al., 2001) with an additional row of 5 inactivated states to reproduce the inactivation of spHCN channels. The four S4s are assumed to move independently of each other with activation rate α and deactivation rate β when the channels are closed. α = k exp[−z(V − V1/2)/kBT] and off-rates β = k exp[+z(V − V1/2)/kBT], k is the opening rate and closing rate at V = V1/2, z is the gating charge moved in each S4 transition, kB is the Boltzmann constant, and T is temperature. Each activated S4 increases the voltage-independent opening rate kopen and decreases the voltage-independent closing rate kclose with the allosteric factor L (Fig. 10). In the presence of cAMP, spHCN channels only stay in the 10 closed and open states because cAMP bound to the CNBD prevents inactivation (Fig. 10, red dashed box). In the absence of cAMP, spHCN channels can populate all 15 states (Fig. 10, blue dashed box), and each activated S4 increases the inactivating rate kin and decreases the recovery rate krec (both voltage-independent rates) with the allosteric factor M (Fig. 10). cAMP also increases the opening rate kopen with a factor of 2 independently of the effect on inactivation to recapitulate the effect of cAMP to increase the open probability in HCN channels (Proenza et al., 2002; Proenza and Yellen, 2006). The S4 activation and deactivation rates are modified in the open and inactivated state by the factors L and M to obey detailed balance.
Statistics and reproducibility
GV curves were obtained by plotting the normalized tail currents versus different test pulses to determine the steady-state voltage dependence of current activation. Tail currents were measured at +10 mV following test pulses. The GV curves were fit with a single Boltzmann equation: where Amax and Amin are the maximum and minimum, respectively, V1/2 is the voltage where 50% of the maximal conductance level is reached, and K is the slope factor. Data were normalized between the Amax and Amin values of the fit. Fluorescence signals were bleach-subtracted, and data points were averaged over tens of milliseconds at the end of the test pulse to reduce errors from signal noise. FV curves were obtained by plotting the normalized steady-state fluorescence signal versus different test pulses. The FV curves were fit with a single Boltzmann equation.
All experiments were repeated more than three times from at least two batches of oocytes. Statistical data analysis was performed using Student’s t test or ANOVA. Data are presented as mean ± SEM and n represents the number of experiments. All n numbers are biological replicates.
Results
QWE mutation at the C-terminal S4 activated by depolarization
HCN channels have several extra residues in the C-terminal end of the S4 helix compared with other voltage-gated channels (Fig. 2 a). These residues are highly conserved in the HCN channel family and are suggested to form non-covalent interactions with residues in the S5 helix to stabilize the closed state with S4 up (Cowgill et al., 2019; Flynn and Zagotta, 2018; Lee and MacKinnon, 2017; Fig. 2 b). Previously, we found that in spHCN channels, the deletion of the QWE motif (at the position of 354–356; Fig. 2 b) renders the channels voltage-independent and S4 immobilized (Ramentol et al., 2020). To understand the mechanism behind it, we made one mutant channel spHCN-QWE-3G (where QWE was replaced by GGG) to remove all the potential interactions formed by the side chains of these residues. We used VCF to measure S4 movement by the fluorescence emitted from fluorophores attached to S4 and gate opening by the current through the mutant channels. To conduct VCF, we introduced the mutation R332C in S4 and then labeled these mutant channels with the fluorophore Alexa-488 C5-maleimide (we indicate channels with Alexa-488 C5-maleimide attached to R332C with an *; Fig. 2 a). We have previously shown that the fluorescence from Alexa-488-labeled R332C (wt*) channels is a good reporter for S4 movement in spHCN channels (Ramentol et al., 2020; Wu et al., 2021).
Wt* channels are closed at the holding potential of −10 mV. They open and generate inward currents in response to voltage steps more negative than −40 mV (Fig. 2 c, left, black traces). Wt* channels close in response to the tail voltage at +40 mV, as can be seen by the decaying tail currents (Fig. 2 c, left, inset). We measured the GV relation of wt* channels by measuring the tail currents at +40 mV (Fig. 2 d, dashed black line, fits from previous data; Ramentol et al., 2020). The GV relation shows that wt* channels are activated by hyperpolarization. The fluorescence signals for wt* channels are complex and multicomponent (Fig. 2 c, left, red traces), and have been well described earlier (Wu et al., 2021). The fluorescence increases in response to voltage steps more negative than −20 mV, assumed to be reporting on S4 inward movement, and then decreases in response to the tail voltage at +40 mV, assumed to be reporting on outward S4 movement (Fig. 2 c, left, red traces). We measured the voltage dependence of the S4 movement by measuring the FV relation at the end of each voltage step. QWE-3G* channels shift the FV relation by more than +50 mV (Fig. 2 d, compare red dashed and solid lines). This shift is, most likely, mainly due to the mutation E356G in the QWE-3G* mutant because mutations of E356, such as E356A, have previously been shown to shift the FV relation by more than +50 mV (Wu et al., 2021). The fluorescence from the QWE-3G* mutant is already close to its maximum value at the holding potential of −10 mV as if S4 is already in the inward state at −10 mV. The fluorescence decreases in response to more positive voltages as if S4 moves outwards at more positive voltages (Fig. 2 c, right, red traces, and Fig. 2 d, red solid line).
At the holding voltage of −10 mV, QWE-3G* channels are partly open as seen by the instantaneous currents in response to both positive and negative voltage steps from −10 mV (Fig. 2 c, right, black traces). The QWE-3G mutation seems to destabilize the closed state of spHCN channels as they display large constitutively currents at positive voltages at which spHCN channels normally close, which is consistent with previous reports showing that the most C-terminal part of S4 is necessary for closing the pore at positive voltages (Flynn and Zagotta, 2018; Ramentol et al., 2020). The currents from QWE-3G* channels were blocked by the HCN-channel blocker ZD7288 (Fig. 3), suggesting that the mutant has not altered the open pore structure. Surprisingly, in contrast to wt* channels, QWE-3G* channels increase their conductance in response to depolarization and decrease their conductance in response to hyperpolarization (Fig. 2 d, black squares). The tail currents and GV relation from QWE-3G* channels show larger currents after steps to positive voltages (for example, after 100 mV, Fig. 2 c, right) and smaller currents after steps to negative voltages (for example, after 140 mV, Fig. 2 c, right), with reopening of channels with a sigmoidal time course during the tail voltage steps to +40 mV following steps from negative voltages (Fig. 2 c, right, inset). The relationship of the FV and the normalized GV of QWE-3G* channels (Fig. 2 e) is similar to that in wt spHCN channels (Fig. 2 d, dashed lines). The FV relations occur at more positive voltages than GV relations in both the wt* and QWE-3G* channels. In addition, the time course of the fluorescence slightly precedes that of the current in QWE-3G* channels (Fig. 2 f), as if the movement of S4 triggers the voltage-dependent changes in currents in QWE-3G* channels. A previous study has reported this time course difference in wt spHCN channels (Bruening-Wright and Larsson, 2007).
Taken together, our VCF data show that QWE-3G* channels show a similar S4 movement as wt* channels, albeit shifted by >50 mV, but, in contrast to wt* channels, the conductance of QWE-3G* channels is increased by the outward movement of S4 upon depolarization. This suggests that the QWE-3G* mutant changes the gating polarity of spHCN channels so that the mutant is now activated by depolarization rather than hyperpolarization. It further suggests that disrupting the interactions at the C-terminal end of S4 reverses the polarity of activation in spHCN channels.
cAMP reverses the voltage-dependent activation in QWE mutation
Zagotta and colleagues (Flynn and Zagotta, 2018) have shown that the deletion of QWE in spHCN channels exhibits depolarization-dependent activation in the absence of cAMP. This depolarization-dependent activation is eliminated by the application of cAMP. Since spHCN channels rapidly inactivate during hyperpolarization in the absence of cAMP, the authors concluded that the depolarization-dependent activation is because of the depolarization-dependent recovery from inactivation from a hyperpolarization-dependent inactivation. To determine whether the depolarization-dependent activation in QWE-3G is also due to the recovery from inactivation, we applied cAMP on QWE-3G using a patch clamp in the inside-out configuration.
In the absence of cAMP, QWE-3G channels show depolarization-activated currents. At the holding potential of −80 mV, QWE-3G channels are partly open as can be seen by the instantaneous currents in response to both positive and negative voltage steps (Fig. 4 a, left). The currents increase as the voltage steps become more positive. More importantly, the tail currents are larger following voltage steps from positive voltages and smaller following voltage steps from negative voltages with the reopening of channels with a sigmoidal time course during the tail voltage steps to +40 mV following steps from negative voltages (Fig. 4 a, left, inset, and Fig. 4 a). The GV relation and the steady-state I-V plot show that QWE-3G channels in 0 cAMP are activated by depolarization (Fig. 4, b and c). The QWE-3G currents in patch-clamp experiments (Fig. 4 a, left) are slightly different from the QWE-3G currents in whole oocytes in VCF experiments (Fig. 2 c, left, black traces). Note that whole oocytes contain 3–5 µM cAMP (Mulner et al., 1983), so a complete congruence would not be expected between data from excised patches with 0 cAMP and whole oocyte recordings. In addition, patch excision has been shown to shift the voltage dependence of wt spHCN channels to more negative voltages due to the loss of intracellular factors, such as phosphatidylinositol 4,5-bisphosphate (PIP2; Flynn and Zagotta, 2011).
The application of cAMP on the cytosolic side of the patch removes the depolarization-dependent activation of QWE-3G channels (Fig. 4 a, right). Instead, the QWE-3G channels seem now maximally open at the holding potential of −80 mV and the currents decrease during positive voltage steps, which leads to smaller tail currents at +40 mV following steps from positive voltages and larger tail currents following steps from negative voltages (Fig. 4 a, right, inset). The GV relation and the steady-state I-V plots show that the conductance of QWE-3G channels now increases with hyperpolarizing voltages in the presence of 100 µM cAMP (Fig. 4, b and c). This suggests that in the absence of cAMP, QWE-3G channels show inactivation at negative voltages and that the depolarization-dependent activation is due to the recovery of the inactivation at positive voltages. However, in the presence of cAMP, the inactivation is removed and the mutant shows hyperpolarization-dependent activation instead, as would be expected for wt spHCN channels (compare Fig. 4 b, blue, and Fig. 2 d, black dashed lines). In addition, cAMP also increases the current amplitudes at all voltages, which is consistent with previous studies that cAMP increases the open probability both at positive and negative voltages in HCN channels (Flynn and Zagotta, 2018; Proenza et al., 2002; Proenza and Yellen, 2006).
Increasing cAMP alters S4 movement only slightly
Since high concentrations of cAMP switched the polarity of gate opening in the QWE-3G* mutation, we used voltage-clamp fluorometry to test whether increasing cAMP would also alter the S4 movement in the QWE-3G* mutant. As the current from whole oocytes (Fig. 2 c) in control conditions (basal level of cAMP is around 3–5 μM; Mulner et al., 1983) was more similar to the currents found in excised patches with 0 cAMP than to currents from patches with 100 μM cAMP (Fig. 4 a), we assumed that QWE-3G* mutant channels in whole oocytes are mainly without bound cAMP. To increase cytosolic concentrations of cAMP in whole oocytes, we injected 5 mM of cAMP (final concentration around 500 μM cAMP) into oocytes. We found that fluorescence traces from the QWE-3G* mutant channel look similar in control conditions and after the injection of 5 mM cAMP (Fig. 5 a), although cAMP slightly shifts the FV relation by −20 mV (Fig. 5 b and Table 1), as if a similar voltage sensor movement occurs without significantly bound cAMP (control conditions) or in the presence of high concentrations of cAMP (5 mM injected cAMP). These results suggest that S4 moves downward in a similar manner in response to hyperpolarization in channels that open by depolarization, such as the QWE-3G* channels in low or no cAMP (Fig. 2 c, right, and Fig. 4 a, left), and in channels that open by hyperpolarization, such as the QWE-3G* channels in high cAMP (Fig. 4 a, right). However, this downward S4 movement leads to more closing of QWE-3G* channels in low or no cAMP but leads to more opening in QWE-3G* channels in high cAMP.
Voltage-sensor movement is similar in activated and inactivated spHCN channels
It has previously been proposed that the voltage dependence of the inactivation of spHCN channels is due to a different voltage sensor than during activation (Flynn and Zagotta, 2018). However, a recent study suggested that the same voltage sensor is responsible for both the activation and inactivation of spHCN channels (Dai et al., 2021). As our data here have suggested that cAMP switches the polarity of channel opening in our mutant spHCN channels by altering the occupancy of activated and inactivated states, we wanted to measure S4 movement in activated and inactivated spHCN channels. We have already shown the voltage-sensor movement in spHCN channels that are activated by hyperpolarization (wt channels) or depolarization (QWE-3G channels) in Fig. 2. Next, to investigate the voltage-sensor movement of spHCN channels upon inactivation, we introduced mutations at the position of R620 in the cAMP binding site in the CNBD to remove the cAMP binding and therefore to induce inactivation in spHCN channels. The homologous arginine R620 is important for cAMP binding to the CNBD in mammalian HCN and cyclic-nucleotide-gated channels, and mutating it to glutamate (R620E) disrupts the interaction between cAMP and CNBD (Chen et al., 2001; Ulens and Siegelbaum, 2003; Zagotta et al., 2003).
Fig. 6 a shows the current and fluorescence from several R620* mutations. The R620* mutations show similar GV (no GV for R620A due to no detectable currents) as wt* spHCN channels (Fig. 6 b). However, for the R620* mutants, the inward currents at negative voltages were relatively much smaller than the outward tail currents at the positive voltage compared with wt* channels (Fig. 6 a), as if the mutants undergo hyperpolarization-dependent inactivation and fast recovery from inactivation during the tail voltage. We compared the currents at −120 mV between wt* and R620* mutant channels after normalizing the currents to the maximum tail currents for each mutant (Fig. 7). Compared with wt* channels with a normalized current at −120 mV of −0.72 ± 0.11 (n = 8), R620E* and R620G* show significantly reduced currents of −0.29 ± 0.02 (n = 3) and −0.21 ± 0.06 (n = 4), respectively, while R620A* show no detectable currents (n = 4) as R620 mutations decrease the spHCN affinity to cAMP such that the mutants are now partly or completely inactivated. To confirm that R620 mutants cause inactivation, we introduced a single point mutation F459L, which has been shown to remove the inactivation in spHCN channels (Idikuda et al., 2019; Shin et al., 2004). We found that the F459L mutation increases the normalized current in all R620* mutants (Fig. 7), suggesting that the decreased currents in R620* mutants are mainly due to inactivation.
Next, we tested the voltage-sensor movement in inactivated R620* mutants (Fig. 6 a). Fig. 6 c shows that wt* and R620* mutant channels display similar FV relations. V1/2 for wt*, R620E*, R620G*, and R620A* channels is −54.70 ± 1.07 mV (n = 3), −52.76 ± 2.41 mV (n = 3), −54.75 ± 1.46 mV (n = 4), and −46.83 ± 2.44 mV (n = 4), respectively (Table 1). These data suggest that S4 moves in a similar way in the activated (wt*) or inactivated (R620* mutants) states of spHCN channels.
Model for activation and inactivation in spHCN channels
We here propose a model for the activation and inactivation of spHCN channels, which assumes that the same S4 movement occurs in activated or inactivated spHCN channels (Fig. 8 a, Fig. 9 a, and Table 2). In this model, S4 is in the up/resting state upon depolarization and the down/activated state upon hyperpolarization. When S4 is up, the closed gate is stabilized by interactions between the C-terminal ends of S4 and S5, as shown here (Fig. 2 b) and previously (Flynn and Zagotta, 2018; Ramentol et al., 2020). This is the resting-closed (RC) state. The downward movement of S4 upon hyperpolarization promotes gate opening by removing the interactions between S4 and S5, inducing the activated-open (AO) state. However, in the absence of cAMP, the downward movement of S4 also promotes inactivation, thereby inducing the activated-inactivated (AI) state. Inactivation is here proposed as an additional coupling of the voltage sensor to the gate due to the interactions between the C-linker of CNBD and S4 (Dai et al., 2021), thereby imposing an additional voltage sensor-to-gate coupling that closes the gate by inactivation. The initial downward motion of S4 and gate opening is relatively fast followed by a slower rearrangement of the CNBD, which is due to the interactions of the C-linker with S4 that more slowly close the gate by inactivation. Therefore, it is a similar downward S4 movement that induces gating opening and inactivation in the presence and absence of cAMP, respectively. Upon depolarization, the upward movement of S4 in the inactivated state promotes recovery from inactivation, which is here shown as a decoupling of the C-linker from S4 due to the upward movement of S4 away from the C-linker. This is the resting-inactivated (RI) state. In the presence of cAMP, inactivation is prevented due to the cAMP-induced conformational changes in the CNBD that decouples the C-linker from the VSD–PD interface by moving the C-linker away from S4 as proposed before (Dai et al., 2021). We also have a gating model for QWE-3G channels (Fig. 9 b). Using the same S4 movement, QWE-3G channels inactivate in the absence of cAMP and activate in the presence of cAMP upon hyperpolarization, thereby creating hyperpolarization-closed channels and hyperpolarization-opened channels, respectively.
Our simple kinetic model can reproduce the GV relations from wt spHCN and QWE-3G channels in this study (Fig. 8 b) using the full tetrameric channel gating scheme (Fig. 10). In wt channels, the open probability (Popen) increases in 100 cAMP but stays around 0% (as inactivation) in 0 cAMP when voltages become more negative. However, QWE-3G channels are partly open at positive voltages with a Popen of 50–60%. The Popen increases in 100 cAMP and decreases (as inactivation) in 0 cAMP when voltages become more negative, which explains the switch in gating polarity of QWE-3G channels that is dependent on cAMP (Fig. 4).
Discussion
Despite many previous studies (Cowgill et al., 2019; Dai et al., 2021; Flynn and Zagotta, 2018; Ramentol et al., 2020), the gating mechanism in hyperpolarization-activated HCN channels is still not completely understood. Unlike most depolarization-activated Kv channels, spHCN channels are activated by hyperpolarization and modulated by cAMP. The spHCN channel is unique in the HCN family in that they display a rapid inactivation upon hyperpolarization in the absence of cAMP (Gauss et al., 1998). The mechanism underlying the inactivation is not completely clear. We (Vemana et al., 2004) and others (Flynn and Zagotta, 2018) have shown the importance of the CNBD in the inactivation of spHCN channels since deleting the CNBD removes the inactivation. Recent cryo-EM structures of human HCN1 channels and FRET studies of spHCN channels show a large S4 movement between resting and activated states (Dai et al., 2021; Lee and MacKinnon, 2019). However, even in the activated state of S4, the gate is closed in the cryo-EM structure, making it unclear how HCN channels are opened by hyperpolarization. Our data here show that the mutation QWE-3G at the C-terminal end of S4 induces opposite voltage dependence in the absence or presence of cAMP (Fig. 4). In the presence of cAMP, similar to wt spHCN channels, the QWE-3G mutant increases its conductance at negative voltages. In contrast, in the absence of cAMP, the mutant instead decreases its conductance at negative voltages. This is most likely due to inactivation at negative voltages and recovery from the inactivation at positive voltages. So the same mutant can display both hyperpolarization-dependent and depolarization-dependent activation depending on the presence or absence of cAMP. In addition, the QWE-3G mutation seems to destabilize the closed state of spHCN channels as they display large constitutively open currents at positive voltages at which spHCN channels normally close, supporting that non-covalent interactions at the C terminal end of S4 contribute to the VSD–PD coupling of spHCN channels to close the channels at depolarized voltages (Flynn and Zagotta, 2018; Ramentol et al., 2020).
Using VCF, we found that the QWE-3G mutant undergoes a similar downward S4 movement in response to hyperpolarization in the absence or presence of cAMP, although cAMP slightly shifts the voltage dependence of S4 movement to more negative voltages (Fig. 5). However, this downward S4 movement leads to more closing of the QWE-3G mutant channel in the absence of cAMP (i.e., the channels being depolarization-activated) but more opening in the presence of cAMP (i.e. the channels being hyperpolarization-activated). Therefore, cAMP switches the gating polarity of these mutant spHCN channels by altering the voltage sensor-to-gate coupling but not the voltage sensor movement. These results agree with previous studies that the voltage sensor movement is conserved in ion channels gated by depolarization or hyperpolarization (Männikkö et al., 2002). This also supports the previously proposed model (Männikkö et al., 2002; Vemana et al., 2004) that it is a different voltage sensor-to-gate coupling that contributes to the reversed gating polarity of HCN channels compared with depolarization-activated Kv10 and Kv11 channels. One alternative explanation of our data is that the QWE-3G mutation renders the channels insensitive to S4 movement, which would explain the large constitutive conductance in the QWE-3G channels. The voltage dependence of the remaining voltage-dependent currents could be due to charge movements in some other parts of the channels (e.g., the selectivity filter or the activation gate itself). However, because the relationships of the voltage dependence and kinetics of the fluorescence and currents are similar in both the wt and QWE-3G channels, we think this alternative explanation is less likely than our suggestion that the voltage dependence of the currents in QWE-3G channels is due to S4 movement.
Zagotta’s group (Dai et al., 2021) recently showed that the extent of S4 movement (i.e., the end states of S4 conformational changes) is similar in the presence and absence of cAMP in spHCN channels using patch-clamp fluorometry. This is based on the fact that the tmFRET efficiency between a fluorophore in S4 and a transition metal attached to the HCN domain in the N terminus increased to a similar extent in the presence or absence of cAMP upon hyperpolarization, suggesting a similar downward movement of S4 in both activated and inactivated spHCN channels. Using R620 mutations to prevent cAMP binding to the spHCN channels or by injecting high concentrations of cAMP, we showed here that the voltage dependence of S4 movement in spHCN channels is similar in the absence and presence of bound cAMP (Fig. 5 b and Fig. 6 c). Our results together with the previous tmFRET study (Dai et al., 2021), therefore, suggest that the same S4 movement occurs during activation and inactivation in spHCN channels.
Taken together, our results here suggest that spHCN channels can open with different polarities (hyperpolarization-activated versus depolarization-activated spHCN channels) with a similar S4 movement. This supports a previous model that a different VSD–PD coupling contributes to the reversed gating polarity of HCN channels compared with most Kv channels. Our results open up the possibility that mammalian Kv10/Kv11 channels and HCN channels (similar non-swapped channel structures but open by opposite polarity) use similar S4 movements but different coupling mechanisms between the voltage sensor and the pore. For example, wt spHCN channels and QWE-3G mutant without cAMP would use voltage sensor-to-gate coupling formed by interactions between S4 and S5 for activation and the additional voltage sensor-to-gate coupling via the C-linker for inactivation (Fig. 10, blue box). In contrast, mammalian hyperpolarization-activated HCN channels and spHCN in the presence of cAMP would use exclusively the voltage sensor-to-gate coupling formed by interactions between S4 and S5 (Fig. 10, red box), whereas the depolarization-activated Kv10/Kv11 channels would use exclusively the voltage sensor-to-gate coupling via the C-linker (Fig. 10, purple box).
Data availability
All data are available from the corresponding author upon reasonable request.
Acknowledgments
Crina M. Nimigean served as editor.
We thank Dr. Fredrik Elinder, Linköping University, and Rene Barro-Soria, University of Miami, for their helpful comments on the manuscript.
This project was funded by National Institutes of Health grant R01 GM109762 (to H.P. Larsson).
The authors declare no competing financial interests.
Author contributions: X. Wu, R. Ramentol, and H.P. Larsson designed the research. X. Wu, K.P. Cunningham, R. Ramentol, M.E. Perez, and H. P. Larsson performed experiments. X. Wu and H.P. Larsson analyzed the data and wrote the manuscript.