Conformational rearrangements in the second voltage sensor domain switch PIP₂- and voltage-gating modes in two-pore channels

Significance

Membrane voltage and PIP₂ are different types of signals on endosomal and lysosomal membranes. The two signals are integrated into two-pore channels (TPCs), whose two repeating domains, DI and DII, play roles in PIP₂ binding and voltage sensing, respectively. We showed that the conformation of the S4 helix in DII determines the voltage-dependent or PIP₂-dependent gating mode, which explains the different preferences of the two signals between the TPC subtypes. The preference for these two gating modes can be changed by the flavonoid, naringenin. Our findings on the molecular mechanisms of the two gating modes in TPCs provide a clue to the understanding and pharmacological manipulation of the signaling by PIP₂ and voltage in intracellular organelles.

Abstract

Two-pore channels (TPCs) are activated by phosphatidylinositol bisphosphate (PIP₂) binding to domain I and/or by voltage sensing in domain II (DII). Little is known about how these two stimuli are integrated, and how each TPC subtype achieves its unique preference. Here, we show that distinct conformations of DII-S4 in the voltage-sensor domain determine the two gating modes. DII-S4 adopts an intermediate conformation, and forced stabilization in this conformation was found to result in a high PIP₂-dependence in primarily voltage-dependent TPC3. In TPC2, which is PIP₂-gated and nonvoltage-dependent, a stabilized intermediate conformation does not affect the PIP₂-gated currents. These results indicate that the intermediate state represents the PIP₂-gating mode, which is distinct from the voltage-gating mode in TPCs. We also found in TPC2 that the tricyclic antidepressant desipramine induces DII-S4-based voltage dependence and that naringenin, a flavonoid, biases the mode preference from PIP₂-gating to desipramine-induced voltage gating. Taken together, our study on TPCs revealed an unprecedented mode-switching mechanism involving conformational changes in DII-S4, and its active role in integrating voltage and PIP₂ stimuli.

Two-pore channels (TPCs) are members of the voltage-gated cation channel superfamily (1–3). TPCs play a major role in nicotinic acid adenine dinucleotide phosphate-dependent Ca²⁺ release from intracellular organelles such as lysosomes, endosomes, and cortical granules, which serve as Ca²⁺ stores that are different from the endoplasmic reticulum (4, 5). Consequently, TPCs are related to diverse physiological functions, including angiogenesis (6) and hair color determination (7), as well as pathophysiologies, such as fatty liver disease (8) and Parkinson's disease (9). Owing to the critical roles of TPCs in intracellular organelles, their pharmacological inhibition and gene knockout have been known to prevent intracellular entry of the Ebola virus (10). Similarly, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which has emerged as the cause of the COVID-19 pandemic, invades cells through angiotensin-converting enzyme 2-mediated endocytosis, involving endosomes and lysosomes (11–13). Consistent with this notion, some TPC2 inhibitors such as tetrandrine and naringenin suppressed SARS-CoV-2 infection in in vitro assays (14–17).

TPCs are composed of two homologous domains (DI and DII), each of which contains six transmembrane helices that are functional units of voltage-gated cation channels (1, 18, 19) and function as dimers (20, 21). The former four helices form a voltage sensor domain (VSD), and the latter two helices comprise a pore domain (PD). In VSDs, positively charged residues in helix S4 are important for voltage sensing (22–24). In the case of TPCs, while both DI-S4 and DII-S4 have positively charged residues, voltage sensitivity is primarily governed by DII-S4 (21, 25, 26). Electrophysiological recordings have shown that TPCs are activated by phosphatidylinositol bisphosphate (PIP₂), PI(3,5)P₂ and/or PI(3,4)P₂, and selectively allow Na⁺ permeation (27–31). Several cryo-electron microscopy (cryo-EM) structures of TPCs have shown that PIP₂ binds to a region composed of a DI-S4/S5 linker and VSD1 (25, 26, 32). Therefore, TPCs are a unique type of voltage-gated ion channel in which each domain is responsible for ligand binding and voltage sensing.

It is still unclear how PIP₂ binding and voltage sensing are integrated in TPCs, mainly owing to the lack of functional studies employing electrophysiological analysis, in contrast to many existing structural reports. Notably, structures with apo- and PIP₂-bound states have been previously reported (26, 32), where PIP₂-induced structural rearrangements around the activation gate were mainly discussed in terms of the gating mechanism. However, our previous functional study on TPC3 from Xenopus tropicalis (XtTPC3) showed that PIP₂ binding in DI affects the movement of distantly located DII-S4 (30), suggesting that the integration mechanisms of PIP₂ and voltage in TPC3 are global and complex. In addition, TPC2, which is believed to be nonvoltage-dependent, was recently reported to elicit a voltage-dependent current in the presence of a series of chemical compounds named lysosomal Na⁺ channel voltage-dependent activators (LyNaVAs) (33). Furthermore, another type of TPC2 agonist has been shown to alter cation selectivity (34). TPC2, and perhaps other TPC subtypes, have more complex integration mechanisms for PIP₂ and voltage than expected. Moreover, it remains unclear how different preferences and sensitivities to respective stimuli are achieved for each TPC subtype.

To explore the details of the activation mechanism of PIP₂ and voltage in TPCs, we first focused on XtTPC3. Voltage-clamp fluorometry (VCF) analysis of XtTPC3 revealed that DII-S4 has an intermediate state that opens in a strongly PIP₂-dependent manner. A mutation that stabilizes this intermediate state transforms XtTPC3, which is predominantly voltage gated, into a strong PIP₂-gated channel-type, such as TPC2. Next, we focused on human TPC2 (HsTPC2) and revealed similarities in the PIP₂ and voltage-gating modes in the TPC subtypes. Furthermore, considering the presence of the two modes, we found that naringenin is not a simple TPC2 inhibitor; rather, it acts to bias the equilibrium between the two modes. These results reveal a unique regulatory mechanism that is likely to be common among TPC subtypes, where DII-S4 conformations define PIP₂-gating and voltage-gating.

Results

Two Steps of Movements of DII-S4 in XtTPC3 Revealed by VCF Analysis.

VCF is a method used to track local structural changes using a fluorophore introduced into a specific amino acid. When a fluorophore is attached to the top of the S4 helix of typical voltage-gated ion channels, such as a Shaker channel, the fluorescence changes reflect, in most cases, the movement corresponding to the transfer of gating charges that precede the opening of the activation gate. Therefore, voltage-dependent changes in fluorescence intensity are generally detected at more hyperpolarized voltages than currents, and match the voltage dependence of the gating charge transfer (Fig. 1A) (35). To analyze the movement of DII-S4, we performed VCF analysis using XtTPC3 with a fluorophore attached to a cysteine introduced at Gln507, located at the extracellular end of DII-S4 (Fig. 1B and SI Appendix, Fig. S1 A and B). Current and fluorescence were simultaneously measured in the presence of PI(3,4)P₂, which potentiated the voltage dependence of XtTPC3 (31). PI(3,4)P₂ levels were increased by the coexpression of human INPP5D (HsIP5D), a PI(3,4) P₂-producing enzyme, and insulin treatment (36). We observed changes in the fluorescence intensity that were made of two distinct components (Fig. 1 C–E). One component with a gradual voltage-dependent decrease (F1) was observed at more hyperpolarized potentials than at those where channel opening occurred, whereas the other component (F2) showed a relatively steep increase at more depolarized potentials. As two components were observed in the distinct range of the membrane voltages, at a constant +90 mV step pulse, a rapid fluorescence decrease corresponding to F1 and a subsequent gradual increase derived from F2 could be resolved (Fig. 1D). These two types of fluorescence intensity changes with different properties indicate that DII-S4 has an intermediate conformation during the transition from the resting state to the fully activated state. Our previous recordings showed that the major gating charge movement in XtTPC3 occurred within the range of membrane potentials in which F2 was observed (Fig. 1E) (30). Taken together, the DII-S4 movement in XtTPC3 is thought to be quite different from the typical S4 movements (Fig. 1 A and E).

Fig. 1.

Two components of fluorescence changes in the VCF measurement of XtTPC3. (A) A schematic drawing for the typical voltage-gated ion channel with the S4 helix labeled by a fluorophore (Top) and an example of its VCF measurement (Bottom). Two VSDs and their linkers to the PDs were omitted for clarity. G/G_max, Q/Q_max and ΔF/F_max indicate the voltage dependence of normalized conductance, gating charge, and changes in fluorescence intensity, respectively. (B) A representation of the molecular structure of XtTPC3. VSD1 and VSD2 from one subunit and their linkers to the PDs were omitted for clarity. DII-S1 and DII-S2 (gray), and DII-S4 (aqua) are shown on a highlighted VSD2. Asn507 (green light spot) and three positively charged residues (red) are also depicted. (C) Representative current recordings (Upper) and fluorescence recordings (Lower) for Q507C attached with Alexa Fluor 488. The expanded view of the current trace elicited by a −40 mV pulse from a holding potential at −60 mV is also shown (Inset). The step pulses to elicit the current and fluorescence are shown on the right side of the fluorescence in matched colors of the corresponding traces. (D) The expanded view of the fluorescence traces in (C), in which the region onset of the current deviation by the step pulses to −40 mV and +90 mV is shown. The color codes are the same as in (C). (E) Normalized G–V and ΔF–V relationships of Q507C coexpressed with HsIP5D (G–V: black open circle, ΔF–V: filled green) and with HsIP4B (G–V: black filled square, ΔF–V: magenta filled square).

Our previous studies on XtTPC3 have shown that the potentiation of voltage dependence by PI(3,4)P₂ binding in DI is caused by the enhancement of voltage-dependent movement of DII-S4 via the interdomain interaction between DI-S6 and DII-S6 (30, 31). To investigate the effects of PI(3,4)P₂ binding on DII-S4 movement in detail, VCF measurements were performed in XtTPC3 coexpressed with human INPP4B (HsIP4B), which specifically degrades PI(3,4)P₂. In XtTPC3 coexpressed with HsIP4B, the conductance-voltage (G–V) relationship significantly shifted toward the depolarized direction, as previously shown (Fig. 1E). VCF measurements showed that an F1 component, similar to that in the presence of PI(3,4)P₂, was observed, even in the absence of PI(3,4)P₂ (Fig. 1E and SI Appendix, Fig. S1C). The F2 component shifted slightly in the depolarized direction compared with that in the presence of PI(3,4)P₂, as previously shown (30).

Introduction of an Electrostatic Interaction between Glu447 and Phe514 Provides an Additional Activation Component That Is Highly Sensitive to PI(3,4)P₂.

In contrast to the two clear components measured in the VCF analysis, only a single component in the G–V relationship was observed at depolarized membrane potentials, suggesting that the intermediate state does not lead to a sufficient level of channel opening in wild type (WT) XtTPC3 (Fig. 1E). A mutation in DII-S4 that uncovered this hidden state was searched to investigate the intermediate state in detail. Substitution of Phe514 with arginine resulted in channel opening, even at −60 mV (Fig. 2 A, Inset), and a clear two-step voltage dependence in the current in the presence of PI(3,4)P₂ (Fig. 2B). Of these two steps, the one on the high-voltage (HV) side occurred in a range similar to the voltage dependence in WT XtTPC3, while the other occurred on the very low-voltage (LV) side with a less steep slope. Notably, the LV component in F514R appeared in a similar membrane potential range, where the F1 component was observed in the VCF of Q507C (Fig. 2B).

Fig. 2.

Two steps in voltage-dependent activation of F514R. (A) Step pulses to elicit each current are colored to indicate the corresponding current traces (Bottom Left), and the representative current recordings of WT (Left) and F514R (Right). In F514R, the expanded view, to highlight the currents elicited by hyperpolarized step pulses, is also shown (Inset). (B) Normalized tail current amplitude of WT (black, n = 5) and F514R (magenta, n = 4). The same plot for ΔF–V of Q507C as in Fig. 1E is shown (dotted green line) for comparison. (C) Amino acid sequence alignments of DII-S4 of XtTPC3 and HsTPC2 (Top), a representation of the molecular structure of VSD2 in XtTPC3 (Bottom Left), and the aligned structures of XtTPC3 and HsTPC2 (Bottom Right). Each residue is depicted in matching colors in all three panels. DII-S3 was omitted for clarity. The spheres indicate the C_α positions of each colored residue. The yellow dotted line indicates the hydrogen bond between Asn473 and Asn548 in HsTPC2. (D) Normalized tail current amplitudes in a series of F514R mutants at each membrane potential (n = 4 and 5). (E) Normalized tail current amplitude in F514R or WT, coexpressed with HsIP4B or HsIP5D and treated by insulin, respectively (n = 4).

We hypothesized that the LV component is caused by an electrostatic interaction between the introduced F514R and a negatively charged residue endogenous to VSD2, such as Asp437, Glu438, and Glu447 (Fig. 2C). While additional D437N or E438Q mutations to F514R still showed an LV component with an apparent shift in the HV component, the LV component was lost in E447Q/F514R in the presence of PI(3,4)P₂ (Fig. 2D and SI Appendix, Fig. S2 A–C). The XtTPC3 structural model, based on the cryo-EM structure of zebrafish TPC3, which is proposed to be in a deactivated state (25), shows that Phe514 is located near Glu447 in this conformation and is likely to approach Asp437 and Glu438 when DII-S4 is fully activated (Fig. 2C). Taken together, these results show that the conductance of the LV component appears when F514R is locked close to Glu447 in a deactivated-like VSD2 conformation, which is a distinct mechanism from the typical voltage-dependent opening accompanied by large S4 movement. Next, we investigated the effects of PI(3,4)P₂ on the two-step activation of F514R. In WT, the G–V in the presence of PI(3,4)P₂, which was achieved by HsIP5D coexpression and insulin treatment, appeared more negative than that in the absence of PI(3,4)P₂, which was achieved by the coexpression of HsIP4B (Fig. 2E and SI Appendix, Fig. S2D). In F514R, HsIP4B coexpression resulted in a shift of the HV component toward the depolarized direction, as in WT XtTPC3; however, it also showed a clear loss of the LV component, instead of a shift (Fig. 2E and SI Appendix, Fig. S2C), suggesting that the LV component depends not only on the VSD conformation in DII, but also on PI(3,4)P₂ binding to DI.

Disulfide Bond Formation between Glu447 and Phe514 Opens Channels in a PI(3,4)P₂-Dependent Manner.

To further demonstrate channel opening in the LV state, an oxidant that forms disulfide bonds, copper-phenanthroline (Cu-Phen), was applied to E447C/F514C and E438C/F514C. In the presence of PI(3,4)P₂, Cu-Phen treatment of both E447C/F514C and E438C/F514C resulted in an increase in the currents that depended on the formation of disulfide bonds (Fig. 3 A and B and SI Appendix, Figs. S3 and S4 A–C). The current increase was observed not only in the peak amplitude, but also in the basal current at the −60 mV holding potential, which is Na⁺ permeable (SI Appendix, Fig. S4A). The Na⁺-permeable conductance at hyperpolarized potentials, which are far more negative than the V_1/2 values (over 200 mV in E447C/F514C), indicates that the voltage-independent opening is added via disulfide bond formation between F514C and E438C or E447C (Fig. 3 C and D). In addition to this voltage-independent opening, voltage-dependent activation and deactivation by strong depolarization were still observed, although their time constants were modified by the Cu-Phen treatment (SI Appendix, Fig. S4D). These results suggest that the cross-linking between the two cysteines renders the channels constitutively active at hyperpolarized potentials, but DII-S4 still moves in a voltage-dependent manner to further increase the open probability (Fig. 3 C and D). We found that the constitutively active components at hyperpolarized potentials in each mutant showed the PIP₂ dependence. In the presence of PI(3,4)P₂, channel opening at hyperpolarized potentials was achieved in both conformations of DII-S4, where Phe514 was located in the proximity of Glu438 or Glu447. In the absence of PI(3,4)P₂, the currents after Cu-Phen treatment in E447C/F514C on the hyperpolarized side was smaller than in the presence of PI(3,4)P₂ with a statistical significance (P < 0.01, Fig. 3 C–E). On the other hand, those in E438C/F514C did not show a statistically significant difference. These results, consistent with those of F514R, show that the conductance that occurs when Phe514 is close to Glu447 depends strongly on PI(3,4)P₂.

Fig. 3.

Disulfide bond formation in E447C/F514C and E438C/F514C, coexpressed with HsIP5D or HsIP4B. (A and B) Representative current recordings of E447C/F514C (A) and E438C/F514C (B), coexpressed with HsIP5D and treated with insulin (Top), or coexpressed with HsIP4B (Bottom), respectively (n = 9 to 11). The recordings before (Left) and after (Right) Cu-Phen treatment are shown. The recordings were performed in the 20 mM Na⁺ solution to avoid larger inward currents through the constitutively open channels. (C and D) The normalized tail current amplitudes at each membrane potential in E447C/F514C (C) and E438C/F514C (D), before and after Cu-Phen treatment. In all the combinations, the tail current amplitude was normalized to the maximum current amplitude before Cu-Phen treatment. The color codes for each combination are the same as used in (A) and (B). The open and filled symbols indicate the amplitude before and after Cu-Phen treatment, respectively. (E) Normalized tail current amplitudes at −100 mV after Cu-Phen treatment. Statistical significance was examined by an unpaired t test. **P < 0.01, *P < 0.05, and n.s.: P > 0.05.

F1 Component Reflects the Movement of DII-S4 to a Conformation in Which the Channel Opens with a Strong PI(3,4)P₂-Dependence.

To confirm the correlation between the F1 component observed in the VCF of Q507C and the LV component of the F514R current, we performed VCF analysis of the mutant that stabilized the LV conformation. E447R/F514E was designed with reference to the strategy used by Taylor et al. for the analysis of the KCNQ1 channel (37). The E447R/F514E mutation is likely to strongly stabilize the LV conformation by inducing two electrostatic repulsions and one interaction (Fig. 4A). While further movement of DII-S4 from the LV conformation toward the extracellular side is expected to be prohibited, movement to the inside would still be allowed. Therefore, it was expected that the F2 fluorescence change was suppressed, but F1 was observed. E447R/F514E showed currents with a very gentle voltage dependence over the entire range of measured membrane potentials (Fig. 4 B and C). Notably, when HsIP4B was coexpressed to suppress PI(3,4)P₂ levels, E447R/F514E produced almost no current over the entire range of membrane potentials. This is in sharp contrast to the WT current and HV component of F514R, where channel currents remained even in the absence of PI(3,4)P₂ (Figs. 2E and 4C), indicating that the E447R/F514E current consisted only of the LV component. VCF analysis of E447R/Q507C/F514E labeled with the fluorophore showed that F1 was observed, as in the case of Q507C, whereas F2 was almost completely lost (Fig. 4 D and E). These results demonstrate that the conductance of the LV component, which has weak voltage dependence and strong PI(3,4)P₂-dependence, is caused by the DII-S4 movement that corresponds to F1. Taken together with the previous results, it was shown that the LV component reflects channel opening in a conformation with a VSD2/activation gate in the intermediate/open (I/O) state, and that the HV component, corresponding to the F2 components, reflects opening in a fully activated/open (A/O) state.

Fig. 4.

Current and fluorescence measurements of the E447R/F514E-type mutants. (A) A representation of the VSD2 domain to outline the strategy to stabilize the LV component. The color codes are the same as in Fig. 2C. Lines with arrowheads indicate electrostatic repulsion. (B) Representative current recordings of E447R/F514E (Left) and WT (Right), coexpressed with HsIP5D and treated with insulin (Top), or HsIP4B (Bottom), respectively. Currents are elicited by the step pulses from −150 mV to +200 mV from a holding potential of −60 mV, followed by a 0 mV pulse for 50 ms to elicit the tail currents. (C) Normalized tail current amplitudes of E447R/F514E (magenta) and WT (gray), coexpressed with HsIP5D and treated with insulin (filled), or HsIP4B (open), respectively (n = 5). In both WT and E447R/F514E, the tail current amplitude was normalized to the average amplitude of the tail currents in the HsIP5D coexpression and insulin treatment conditions. (D) Representative current recordings (Upper) and fluorescence recordings (Lower) for E447R/Q507C/F514E coexpressed with HsIP5D and treated by insulin, and labeled by Alexa Fluor 488. Currents and fluorescence were elicited by the step pulses for 300 ms, ranging from -150 mV to +200 mV, from a holding potential of −60 mV, followed by a 0 mV pulse for 100 ms to elicit the tail currents. The recordings were performed in the ND96-based solution containing 20 mM NaCl to decrease the currents at the holding potential. (E) Normalized ΔF–V (green; n = 4) and normalized tail current amplitude (black, n = 4) at each membrane potential of E447R/Q507C/F514E, coexpressed with HsIP5D and treated by insulin.

In E447R/Q507C/F514E, VCF analysis revealed the critical importance of PI(3,4)P₂ in LV and I/O currents. Therefore, we next investigated the existence and contribution of PI(3,4)P₂-dependent I/O currents, even in the WT. The small conductance observed at the hyperpolarized potential in the presence of PI(3,4)P₂ in the WT could be attributed to the I/O state (Fig. 1 C, Inset). In the case of Q507C, a comparison of ΔF–V and G–V showed that the voltage dependence of F2 almost overlapped with G–V in the absence of PI(3,4)P₂, while F2 followed G–V in the presence of PI(3,4)P₂ (Fig. 1E). We also compared the kinetics of the current and fluorescence elicited by step pulses at which the F2 component is dominant. Consistent with the analysis of ΔF–V and G–V, the current concurrently occurred with F2 in the absence of PI(3,4)P₂, while the development of the current was faster than that of F2 in the presence of PI(3,4)P₂ (Fig. 5 A and B and SI Appendix, Fig. S5 A and B). These results suggested two important mechanisms for the PI(3,4)P₂ dependence of XtTPC3. First, in the absence of PI(3,4)P₂, the DII-S4 movement that reflects F2 triggers concurrent channel opening. Second, in the presence of PI(3,4)P₂, before the F2 movement, the channel is already opened. Considering the existence of PI(3,4)P₂-dependent I/O currents in the E447R/F514E mutant, the opening prior to the F2 component in the presence of PI(3,4)P₂ is likely to be caused by the I/O state in the WT.

Fig. 5.

VCF analysis of the PI(3,4)P₂-dependence and the simulation analysis with a state transition model. (A) Superimposed recordings of current and fluorescence from Q507C coexpressed with HsIP5D and treated with insulin (Top) and HsIP4B (Bottom), elicited by a +110 mV step pulse. In both cases, the fluorescence recordings (blue and magenta, respectively) and current recordings (aqua and pink, respectively) were fitted by a single exponential function. The yellow arrowheads indicate the F1 components, while the blue and magenta arrowheads indicate the F2 components. (B) Time constants for activation were obtained from (A) at each membrane potential (n = 6 to 7). The color codes match those used in (A). (C) A scheme of the kinetic model for state transitions of XtTPC3 in the presence of PI(3,4)P₂ (Top) and the simulated G–V and ΔF–V relationships based on the model (Bottom). R, I, Ix, and A in the spheres indicate the VSD2 conformations in resting, intermediate, intermediate-x, and activated states, respectively. C and O indicate the conformations of the activation gate in the closed and open states, respectively. The regions colored and denoted as F1 or F2 indicate the transitions corresponding to each fluorescent intensity change. The dotted compartments are the grouping based on the fluorescence changes. The faint spheres and lines with arrowheads indicate the lack of the state occupancy and zero rates of state transitions, respectively. Dotted sphere and line with arrowheads in I/O state indicate its low occupancy and small rates. (D) A kinetic model and the simulated G–V and ΔF–V relationships in the absence of PI(3,4)P₂ as shown in (C). A paired t test was used to examine statistical significance. **P < 0.01, *P < 0.05, n.s.: P > 0.05.

Kinetic Modeling of PI(3,4)P₂- and Voltage-Dependent Gating in XtTPC3.

To solidify the existence of intermediate states and their PI(3,4)P₂ dependence, we attempted to reproduce the unique features of XtTPC3 that were revealed in this study using a simulation based on kinetic models. However, a quantitatively accurate reconstruction of the experimental data with the Markov model, which requires the parameters for rate constants between several states, was difficult owing to the lack of experimental values. Therefore, this simulation was performed, as supporting evidence, to examine qualitatively whether any kinetic models that reproduce atypical features of XtTPC3 could exist.

A model with one intermediate state (I) of VSD2 failed to reproduce various features of XtTPC3 in the simulation (SI Appendix, Fig. S5C). On the other hand, a model with two intermediate states (I and Ix, where “x” was used due to its uncertainties) is more likely as it recapitulated various experimental observations (Fig. 5 C and D). This model assumes that VSD2 has four distinct states connected by voltage-dependent rate constants. The conformational rearrangements that correspond to F1 and F2 represent the transitions from resting (R) to I and from Ix to A, respectively. Each VSD2 state has closed and open pore states, with fixed rate constants between them. Ix is a state that has not been characterized experimentally but was found to be essential to reproduce the features of XtTPC3 by simulation.

In this model, PI(3,4)P₂ is responsible for determining the occupancies of both the I/O and Ix/O. In the absence of PI(3,4)P₂, channel opening is directly dependent on the transition from Ix/C to A/C, explaining why F2 concurrently occurs with currents (Fig. 5D). In the presence of PI(3,4)P₂, channel opening was mainly caused by Ix/O via Ix/C; hence, currents occurred at voltages lower than the F2 fluorescence change and currents in the absence of PI(3,4)P₂ (Figs. 1E and 5C). Consistent with these relationships, similar kinetics of the increase in the currents and fluorescence in the absence of PI(3,4)P₂ and the delayed appearance of fluorescence in the presence of PI(3,4)P₂ were also reproduced (SI Appendix, Fig. S5D). The LV component in F514R was observed when the entry into the I/O state, whose occupancy was quite low in WT, increased (SI Appendix, Fig. S5 E and F). The LV component was also lost in the absence of PI(3,4)P₂, because the occupancy of I/O was assumed to be dependent on PI(3,4)P₂. Overall, the simulation based on the 4-VSD2-states model qualitatively reproduced the unique features of XtTPC3, showing that it is a plausible model for explaining the characteristic gating mechanisms of XtTPC3.

Correspondence of the Two Gating Modes in XtTPC3 to LyNaVA- and PI(3,5)P₂-Dependent Gating Modes in HsTPC2.

The results thus far show that XtTPC3 is intrinsically designed to open in two distinct DII-S4 conformations: one is the A/O state with DII-S4 fully activated, and the other is the I/O state with less voltage dependence and strong PI(3,4)P₂-dependence. To investigate whether these two gating mechanisms are shared in the TPC family, we focused on TPC2, which has distinct features from those of TPC3.

It is known that TPC2 has LyNaVA-mediated voltage dependence. Because the molecular mechanisms of LyNaVA-induced currents remain elusive (33), we examined whether this unusual voltage dependence was regulated by DI- or DII-S4 (Fig. 6A). A ΔN37 mutant of HsTPC2, in which the N-terminal region critical for lysosomal retention was deleted (38, 39), was used to translocate HsTPC2 to the plasma membrane. HsTPC2 ΔN37, hereafter referred to as WT, showed a current in response to intracellular injection of PI(3,5)P₂ (Fig. 6B and SI Appendix, Fig. S6 A and B). Extracellular perfusion with desipramine, a LyNaVA, also induced currents in the absence of PI(3,5)P₂ (Fig. 6B). While PI(3,5)P₂-evoked currents showed no voltage-dependence, desipramine-induced currents showed a clear voltage-dependence and inward rectification (V_1/2 = −44.9 ± 4.2, n = 6) (Fig. 6 B and C and SI Appendix, Fig. S6C), demonstrating that desipramine-induced HsTPC2 currents were reproduced in the oocyte expression system, as previously reported (33). Neutralizing mutations of arginine residues in DI-S4, such as R185Q and R188Q, did not significantly change the V_1/2 values of the desipramine-induced currents. However, significant shifts toward the depolarized direction were observed in the R554Q and R557Q mutants of DII-S4 (Fig. 6 B and C and SI Appendix, Fig. S6 D and E and Table S2). The I551R mutant, which adds a positive charge at the upper position in DII-S4 (Fig. 6A), did not show a current with voltage dependence under basal conditions. In the presence of desipramine, the current with the G–V remarkably shifted toward a positive direction compared with the WT, possibly due to the perturbation of the optimal voltage-dependent movement of DII-S4 by introducing an extra positive charge. These results show that LyNaVA-induced voltage-dependent currents in HsTPC2 depend on DII-S4.

Fig. 6.

Comparative analysis of desipramine-induced and PI(3,5)P₂-evoked currents in HsTPC2. (A) A representation of the molecular structure of HsTPC2 as shown in Fig. 1B. The positively charged residues are shown as (+) in DI-S4 (blue) and DII-S4 (red). (B) Step pulses to elicit the currents that are colored to indicate the corresponding current recordings (Bottom Left) and the representative current recordings of HsTPC2 WT and mutants. Desipramine-induced currents and PI(3,5)P₂-injection-evoked currents (gray box) in WT and the two mutants, with ΔN37 as common background, are shown. (C) G–V relationships of WT and the neutralizing or charge-introducing mutants obtained from the tail current at −100 mV shown in (B) (n = 4 to 6). (D) Representative recordings of desipramine-induced currents in each HsTPC2 construct. Currents were evoked by ramp pulses ranging from −30 mV to +50 mV. TPC2 currents were induced by the coapplication of 1 mM desipramine and 100 µM riluzole, followed by the addition of 100 µM CdCl₂. (E) Time courses of the change in current amplitude of HsTPC2 in response to the application of desipramine/riluzole and then CdCl₂ (n = 4). Desipramine/riluzole-induced currents were normalized to the maximum induced currents. (F) The plots of the current amplitude evoked by PI(3,5)P₂ or water injection into the oocytes expressing TPC2 WT and mutants or no exogenous proteins (n = 5 to 8). Each current amplitude was measured at −30 mV. Statistical significance was examined by an unpaired t test or one-way ANOVA followed by Dunnett’s test in (F). **P < 0.01, n.s.: P > 0.05.

It is noteworthy that the cryo-EM structures of HsTPC2 in both its apo- and PI(3,5)P₂ bound-states showed that Asn473 and Asn548, which correspond to Glu447 and Phe514 of XtTPC3, are located close enough to form a hydrogen bond (Figs. 2C and 6A). Assuming that the mechanisms for the PIP₂- and voltage-gating modes in TPC2 are similar to those in TPC3, further locking of this conformation would perturb the desipramine-induced voltage-dependent opening mechanism owing to the restricted movement of DII-S4, but still allow the PI(3,5)P₂-dependent mechanism. N473C/N548C, aimed at a Cd²⁺ coordination-induced bridge between the two residues, produced a very small desipramine-induced current. To measure the time course of the effect of Cd²⁺ with a larger current amplitude, we coperfused riluzole, another type of agonist different from LyNaVAs (33), and desipramine, to N473C/N548C (Fig. 6D and SI Appendix, Fig. S7A). Further perfusion with Cd²⁺ strongly inhibited the desipramine/riluzole-induced current in N473C/N548C (Fig. 6 D and E). DTT pretreatment of N473C/N548C did not cause a statistically significant increase in the current, confirming that the two cysteines were free without the oxidant (SI Appendix, Fig. S7 B and C). Under the same conditions, the N473C current was undetectable and no Cd²⁺ effects were observed in N548C. The application of desipramine alone to N473C/N548C also evoked small currents, which were similarly attenuated by Cd²⁺ perfusion (SI Appendix, Fig. S7D). In addition, a Cd²⁺-induced shift in the G–V relationship of the desipramine-induced current toward the depolarized direction was also confirmed in N473C/N548C (SI Appendix, Fig. S7 E and F and Table S2). In contrast, the presence or absence of Cd²⁺ caused no significant differences in the PI(3,5)P₂-evoked currents in terms of both the amplitude and shift in the reversal potential, which signifies the existence of Na⁺ selective currents (Fig. 6F and SI Appendix, Fig. S8 A and B). These observations were in good agreement with the results for Glu447 and Phe514 in XtTPC3. We also observed that PI(3,5)P₂ injection in the presence of desipramine shifted the voltage dependence of HsTPC2 WT toward a negative direction, indicating the existence of PIP₂-dependent potentiation of voltage dependence, as shown in XtTPC3 (SI Appendix, Fig. S8 C–E). Taken together, these results highlight the similarities between the two distinct gating mechanisms defined by the DII-S4 conformation in TPC subtypes.

Naringenin Has Opposite Effects on Desipramine/Voltage- and PI(3,5)P₂-Dependent Modes in HsTPC2.

Naringenin is a plant flavonoid that exhibits various biological activities and has been reported to inhibit the HsTPC2 current induced by PI(3,5)P₂ (40). In our experiments using the Xenopus oocyte expression system, although the results were not statistically significant, naringenin tended to suppress the currents induced by PI(3,5)P₂ intracellular injection, which is consistent with a previous report (Fig. 7 A and B). In contrast, naringenin unexpectedly showed potentiation of desipramine-induced currents and a shift of the G–V relationship toward a negative direction (Fig. 7 C–F and SI Appendix, Fig. S8F and Table S2). Naringenin alone had no effect on HsTPC2 basal currents (Fig. 7 D and E). Taken together, it is likely that naringenin is not a simple inhibitor of HsTPC2 but rather a so-called “biased modulator” that changes the equilibrium between the two modes in favor of the LyNaVA mode (Fig. 8). These complex pharmacological effects of naringenin demonstrate the existence of a multimodal gating mechanism in TPC2 and provide some insights into the function and pharmacology of TPCs.

Fig. 7.

Different effects of naringenin on PI(3,5)P₂-evoked and desipramine-induced currents in HsTPC2. (A) Representative recordings of PI(3,5)P₂-evoked currents of WT (ΔN37) HsTPC2 in the absence (Top) and the presence (Bottom) of 0.5 mM naringenin in the bath solution. Currents were evoked by ramp pulses ranging from −30 mV to +50 mV. (B) Plot for the current amplitude at −30 mV evoked by PI(3,5)P₂ obtained in (A) (n = 7 to 8). (C) Representative current recordings induced by desipramine and by the following coapplication of desipramine and naringenin. Currents were elicited by step pulses, ranging from −120 mV to +60 mV, from a holding potential at 0 mV, followed by a −100 mV pulse to record the tail current. (D) Representative current recordings before (Left) and after (Right) the application of 0.5 mM naringenin. Currents were elicited by the same pulse protocol used in (C). (E) Normalized peak current amplitude at each membrane potential, obtained from the currents shown in (C and D)  (n = 4 to 5). The peak currents were normalized to those in the absence of reagents in each recording. (F) G–V relationships for the currents induced by desipramine alone or by the coapplication of desipramine and naringenin (n = 5). Statistical significance was examined by an unpaired t test or one-way ANOVA followed by Dunnett’s test in (B) and (E and F), respectively. **P < 0.01, *P < 0.05, n.s.: P > 0.05.

Fig. 8.

Schematic drawings for the state transitions in TPC3 and TPC2. (A) State transitions of TPC3 in the absence (Left) and presence (Middle) of PI(3,4)P₂. Voltage-dependent transition is shown in a vertical direction from resting (Top), intermediate, intermediate-x, and activated states (Bottom). Thick painted lines with an arrowhead between the states indicate the F1 (yellow) and F2 (magenta and blue) components in the VCF measurement, respectively. In the intermediate state (I) of WT (Middle), the activation gate could open with a low open probability in the presence of PI(3,4)P₂. In the F514R mutant (Right), the electrostatic interaction between Glu447 and F514R (red dotted oval) stabilizes the I conformation, resulting in a high open probability. Ix states were also depicted as they were shown to be necessary by the simulation analysis. Ix states are expected to have open or closed states in the presence or absence of PI(3,4)P₂, respectively. In the A states, the open probability in the presence of PI(3,4)P₂ is much higher than in its absence, as PI(3,4)P₂ binding facilitates the DII-S4 movement. (B) State transitions in TPC2. In TPC2, it is unclear whether the states corresponding to the resting states of TPC3 exist or not. A hydrogen bond between Asn473 and Asn548 (dotted oval) stabilizes the PI(3,5)P₂-preferable intermediate state (Top). Voltage-dependent state transitions from the intermediate states to the activated states do not occur as currents in the absence of LyNaVA. LyNaVA binding allows the voltage-dependent transition. Further increase in the open probability occurs when PI(3,5)P₂ and voltage are provided simultaneously. Naringenin is thought to have different effects on the two gating modes, suppression of the PI(3,5)P₂-gating and potentiation of the voltage-dependent gating.

Discussion

In this study, we revealed that the VSD2 conformations define the complex switching and integrating mechanisms for PI(3,4)P₂ and voltage gating in XtTPC3. We also confirmed the similarities between these mechanisms in XtTPC3 and LyNaVA-induced voltage gating and PI(3,5)P₂-gating in HsTPC2. The notion of complex gating modes in XtTPC3 is supported by the qualitative reproduction in the simulation based on a kinetic model, as well as the finding in HsTPC2 that naringenin is a unique effector that changes the equilibrium between the two modes.

Our VCF analysis of XtTPC3 revealed characteristic relationships between the transfer of gating charges and multistep state transitions. F2 movement induces a transition to the fully activated state (A/O). F2 movement occurs along with most of the gating charge transfer (30), which is consistent with its large movement, where Phe514 on DII-S4 becomes closer to Glu438 (Fig. 2C). In the absence of PI(3,4)P₂, the F2 component and the current increase occurred almost concurrently (Fig. 5 A and B). This coincidence between the current and fluorescence, which tracks the movement of the gating charge, is occasionally observed in some voltage-dependent ion channels, including KCNQ1 (41). This means that, owing to the poor cooperativity between the four VSDs and the pore gate, a single VSD movement can lead to gate opening (42). Therefore, F2 movement is still typical compared with S4 movement in other voltage-gated cation channels. However, the F1 component is quite atypical in that it can induce channel opening even before a large transfer of gating charges. A very small transfer of gating charges in the F1 component indicates limited and local movement of DII-S4 (30). Consistent with this, VCF measurements with the fluorophore attached to the cytosolic end of DII-S4 in XtTPC3 showed no F1-corresponding movement (30). In addition, the gentle voltage dependence of the F1 component suggests that the intermediate conformation of VSD2 is similar to, and easily returns to, the resting conformation (Fig. 1E). This instability could explain why the I/O state was minor in the WT XtTPC3 (Fig. 8A). An artificial interaction between Glu447 and Phe514, which stabilizes the intermediate conformation (I), increases the chance of channel opening by reaching the I/O state. Interestingly, a recent report on the cryo-EM structures of plant TPC1, which does not have any PIP₂ sensitivity, showed the conformational rearrangement of DII-S4 that is limited around the extracellular end by comparing the two “S4-down” conformations (43). This rearrangement induced a slight expansion of the entire VSD2 structure as if it released DII-S4 for further movement. The F1 movement may correspond to this local structural rearrangement to prepare for the subsequent voltage-dependent F2 movement as well as for PIP₂-gating (Fig. 8). These observations indicate that the multiple-state transition mechanism is achieved by atypical F1 and typical F2 movements of DII-S4 in TPCs.

The unique PIP₂-dependence revealed in XtTPC3 could be a new type of PIP₂-dependent regulatory mechanism in voltage-gated ion channels. A PI(4,5)P₂ binding site in KCNQ1 is located at the interface between the S4-S5 linker and VSD (44, 45), similar to those of PI(3,4)P₂ and PI(3,5)P₂ in TPCs that are near their DI-S4/S5 linkers (25, 26, 31, 32). The binding of PI(4,5)P₂ to KCNQ1 plays a pivotal role in the coupling of VSDs and PDs and its dissociation results in a remarkable decrease in channel activity (45). As the loss of activity in the absence of PI(4,5)P₂ occurs even if KCNQ1 is coexpressed with its auxiliary subunit E1 or E3, which changes the ratio of AO to IO, PI(4,5)P₂ binding is considered to be required regardless of the state (37, 46, 47). In the case of XtTPC3, analysis of the LV components of F514R and E447R/F514E indicated that the binding of PI(3,4)P₂ was indispensable for the I/O state (Figs. 2E and 4C). The loss of PI(3,4)P₂ shifted the voltage dependence toward the depolarized direction in the WT current and in the HV component of F514R, which mainly reflects the A/O state; however, the channel currents were not lost. Qualitative simulation using a state transition model also supported the regulation mechanism by PI(3,4)P₂, focusing on the intermediate states (Fig. 5 C and D), which reproduced various characteristic features of XtTPC3. For modeling, the Ix states are required, although the detailed nature of the state transition remains elusive. VSD2 is known to adopt multiple conformations, some of which can represent the I and Ix states (43). The lack of fluorescence change corresponding to entry into the Ix state may be due to the modest differences in the structures between the I and Ix states, which would allow PI(3,4)P₂ to regulate both states similarly. These observations demonstrate a unique PIP₂-dependent regulatory mechanism that focuses more specifically on intermediate states than activated states.

In this study, DII-S4 in HsTPC2 was found to play a major role in the LyNaVA-induced voltage dependence. In the absence of LyNaVA, the voltage sensitivity of DII-S4 in TPC2 may not be impaired, but it may be decoupled from the activation gate. The binding of LyNaVAs is thought to restore this coupling via an unknown mechanism. The synergy between LyNaVAs and PI(3,5)P₂ in HsTPC2, which has been shown in a previous report (33), as well as the G–V shift caused by PI(3,5)P₂ injection in the presence of desipramine (SI Appendix, Fig. S8 C–E), appears to be homologous to the PI(3,4)P₂-dependent potentiation of voltage-dependence in XtTPC3. This homology strongly suggested shared and fundamental activation mechanisms in the TPC family (Fig. 8). A hydrogen bond between Asn473 and Asn548 observed in the cryo-EM structure of HsTPC2 suggested that it was originally tuned to be locked in the intermediate state to allow PIP₂-gating (Fig. 2C). Furthermore, a comparison between HsTPC2 and the XtTPC3 model, as well as mouse TPC1 (MmTPC1), showed that the relative positions of DII-S4 within VSD2 were almost identical and possibly captured in the intermediate state (SI Appendix, Fig. S9), indicating that this conformation is homologous and stable in all three subtypes. TPC subtypes are thought to share the fundamental framework enabling both PIP₂-gating and voltage-gating and achieving optimal gating by changing the mode and extent of suppression. TPC2 is biased to PIP₂-gating, while TPC3 is biased to voltage gating. Agents such as LyNaVAs, whose binding can induce conformational rearrangements that release these suppressions, may have the potential to act as TPC agonists (Fig. 8). In summary, TPC subtypes might share gating mechanisms and can be tuned to achieve the optimal extent of balanced gating by voltage and PIP₂.

The different effects of naringenin on PI(3,5)P₂-evoked currents (Fig. 7B) and desipramine-induced currents in the absence of PI(3,5)P₂ (Fig. 7E) suggest that these are distinct activation mechanisms, proving the existence of multimodal gating mechanisms in HsTPC2 (Fig. 8). Naringenin inhibits PI(3,5)P₂-induced TPC2 currents (40) and prevents the intracellular entry of several respiratory infectious viruses, including SARS-CoV-2, in in vitro experiments (17). The effects of naringenin revealed in this study provide interesting insights into the strategies for the development of TPC2-targeted drugs. Although there is currently no evidence that LyNaVA mode functions under physiological conditions, its distinct gating mechanism from the PIP₂-mode raises the possibility that it potentially reflects physiological activation under specific conditions. This notion seems to be analogous to a report stating that TPC2 currents with different ion selectivity, which were induced by two distinctive synthetic agonists, have different physiological consequences in lysosomes (34). Given these notions, the mode-targeted pharmacology of TPC2, with naringenin-like “biased modulators,” might be interesting and worth considering as a target to prevent the invasion of diverse viruses.

In summary, the present study showed that VSD2 in TPCs plays an active role in integrating or switching PIP₂-gating and voltage-gating modes, which is an unprecedented mechanism among voltage-gated ion channels. The intermediate conformation was defined as the PIP₂-dependent mode. The balance between the two modes is likely to be tuned in each TPC subtype and could be switched or biased by changing the VSD2 conformation, or especially in TPC2, by a unique type of “biased modulator.” Their multimodality and regulation may be a direction for drug development targeting TPCs.

Materials and Methods

Current recordings were performed with two-electrode voltage-clamp technique using Xenopus oocytes. Synthesized cRNAs were injected into oocytes to obtain the XtTPC3 and HsTPC2 currents. VCF measurements were performed under an upright fluorescent microscope, and the intensity of the emission light was quantified with the photomultipliers. In order to simulate the kinetic model for reproducing the measurements of XtTPC3 currents, the software NEURON was used. Detailed methods for cRNA preparation and injection into oocytes, two-electrode voltage clamp, VCF, simulation, data analysis, and structure modeling are described in the SI Appendix. Experiments using Xenopus oocytes were approved by Animal Care Committee of the National Institutes of Natural Sciences (umbrella institution of the National Institute for Physiological Sciences, Japan), and were performed in accordance with its guidelines.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.