PIP2 activation of the cardiac IKs potassium channel Authors: Lu Zhao, Xianjin Xu, Chenxi Cui, Rui Duan, Ali A. Kermani, Jingyi Shi, Lu Han, Ji Sun, Xiaoqin Zou, Jianmin Cui Published: 2026 ABSTRACT The IKs channel, composed of voltage-gated potassium channel KCNQ1 and regulatory subunit KCNE1, controls cardiac action potential durations. KCNQ1 and IKs activation requires PIP2, and its depletion abolishes channel opening. KCNQ1 adopts both bent and straight conformations and can bind two PIP2 molecules: one adjacent to VSD (V-PIP2), and the other at the VSD-pore interface (C-PIP2). Here we show that the two PIP2 perform essential yet distinct roles: V-PIP2 enables the bent-to-straight transition, whereas C-PIP2 mediates VSD-pore coupling. VSD activation elevates the V-PIP2 site, permitting the shift from the bent, intermediate open (IO) state associated with KCNQ1 to the straight, IKs-exclusive activated open (AO) state, which is further stabilized by C-PIP2. Leveraging this mechanism, we develop a compound CA1, which selectively targets the V-PIP2 site and modulates IKs channel activity without affecting KCNQ1, offering a promising conceptional path for specific and safe antiarrhythmic therapeutics. Introduction Phosphatidylinositol 4,5-bisphosphate (PIP2) is an important signaling molecule that regulates the function of a variety of membrane proteins1–3, including many of ion channels4,5. Among PIP2-regulated ion channels, the family of KCNQ1-5 potassium (K+) channels plays a crucial role in regulating membrane excitability in the brain, heart, and epithelium6–9. PIP2 is required for KCNQ channel function, and channel activity diminishes when membrane PIP2 levels are reduced7,10,11. Thus, investigating PIP2-dependent activation of KCNQ channels could yield valuable insights into the physiological effects and molecular mechanisms of PIP2 regulation in ion channels. We recently solved the structure of KCNQ1 in association with its regulatory subunit, KCNE1, and PIP2 molecules (Fig. S1)11, providing a structural foundation for understanding mechanisms of the PIP2-dependent activation of KCNQ1 channels. KCNQ1 channels are voltage-gated K+ (Kv) channels12–14. KCNQ1, together with the regulatory subunit KCNE1, forms the slow-delayed rectifier potassium (IKs) channels in the heart and inner ear, which are crucial for regulating heart rhythm and supporting K+ recycling in the endolymph15–17. When associated with the regulatory subunit KCNE3, KCNQ1 is constitutively active at physiological voltages, functioning as a K+ transporter to maintain ionic homeostasis in epithelial and endothelial cells18,19. Malfunctions of KCNQ1 caused by drugs, mutations, and single nucleotide polymorphisms are linked to arrhythmia, deafness, atrial fibrillation, and type-2 diabetes mellitus12,20–22. The voltage-dependent gating of Kv channels involves three fundamental processes, voltage sensor domain (VSD) activation, VSD-pore coupling, and pore opening. The VSD of Kv channels, such as Shaker and KCNQ1, can exist in three resolvable states: resting (R), intermediate (I), and activated (A). Upon membrane depolarization the VSD of these channels transition from the R state to the A state via the I state23–25. However, in Shaker channels, the pore conducts only at the activated open (AO) state when the VSD is in the A state23,26. In contrast, in KCNQ1, the pore opens when the VSD is in either the I or the A state, resulting in the IO (intermediate open) and AO states14,24,27,28 (Fig. S1a). The IO and AO states exhibit drastically different properties in their voltage dependence, activation kinetics, current amplitude, ion permeability, and pharmacology13,14,24. The IO and AO states of KCNQ1 are differentially regulated by various KCNE subunits, which exhibit distinct tissue distributions15,16,18,19. This differential regulation enables KCNQ1 + KCNE channels to function across diverse tissues, supporting a wide range of physiological functions. For instance, both KCNE1 and KCNE3 selectively shift the VSD activation to the I state at more negative voltages18. However, KCNE1 suppresses IO but enhances AO of KCNQ1, making the channel open exclusively in the AO state, with a more positively shifted voltage dependence, slower activation kinetics, and larger current amplitudes24,28. These features are essential for terminating cardiac action potential. In contrast, KCNE3 does not suppress IO, allowing KCNQ1 + KCNE3 channels open at more negative voltages and remain constitutively open at physiological voltages to maintain K+ homeostasis in the epithelium18,19. KCNQ1 channels adopt a homo-tetrameric assembly, where each subunit contains six transmembrane α-helices S1-S6, with S1-S4 forming the VSD and S5 and S6 forming the central pore29,30. Each KCNQ1 subunit interacts with a calmodulin (CaM) molecule mainly through its cytosolic-helix A (HA) and helix B (HB) (Fig. S1b). Our previous structural studies revealed two distinct conformations of the cytosolic domain in KCNQ1 channels; straight and bent29,30. In the straight conformation, HA forms a continuous helix with the transmembrane helix S6, positioning CaM deep in the cytosol, away from the membrane-spanning domain of KCNQ1. The bent conformation is facilitated by a structural kink at the “RQKH” motif, located between helix S6 and HA. This kink allows HA to adopt an upward-bent orientation, enabling CaM to undergo rotational displacement to interact with the cytosolic S2-S3 linker (S2S3L). Our previous functional studies demonstrated that mutations disrupting the S2S3L-CaM interface shift KCNQ1 channel properties from predominantly IO to characteristics resembling the AO state. Based on these findings, we proposed that the IO state likely adopts the bent conformation, while the AO state corresponds to the straight conformation, and that disrupting S2S3L-CaM interactions promotes the transition from the IO to the AO state31. Our prior studies demonstrated that PIP2 is required for VSD-pore coupling in KCNQ1 in both the IO and AO states7,10. Without PIP2, the VSD activates but fails to induce pore opening. Functional and mutagenesis studies indicated a PIP2 binding site at the interface between the VSD and the pore, essential for KCNQ1 channel opening7. However, our previous structural analyses of KCNQ1 and KCNQ1 + KCNE3 did not clearly show PIP2 binding at the VSD-pore interface but instead revealed a PIP2-binding site adjacent to the VSD30. We recently solved the structure of the KCNQ1-KCNE1 complex, which revealed not only the PIP2-binding site near the VSD, but also an additional PIP2 site at the VSD-pore interface, yielding a total of eight PIP2 molecules in the channel complex11 (Fig. S1b). In this study, we solve the structure of KCNQ1 in an I state. We examine the structural and functional properties of the two PIP2-binding sites in KCNQ1. These results, together with the identification of a compound, CA1, that selectively binds to one of PIP2 sites and specifically enhances the activation of KCNQ1 only in the presence of KCNE1, support that the IO and AO states adopt the bent and straight conformations, respectively. Through electrophysiological experiments and structural characterization of KCNQ1 channel at the I state, we find that the first PIP2 molecule located near the VSD and termed V-PIP2 for its role in VSD function (Fig. S1b) drives the conformational shift from IO to AO during VSD activation. VSD activation to the A state alters the conformation of V-PIP2 and its binding site, detaching CaM from the VSD and promoting the transition from bent to straight. The second PIP2 molecule, designated C-PIP2, was revealed in our recently determined KCNQ1-KCNE1 structure at the VSD-pore interface, promoting VSD-pore coupling to enable channel opening in both IO and AO states while stabilizing the straight conformation (Fig. S1b)11. VSD-pore coupling determines the channel’s ability to open, whereas the IO-to-AO transition is critical for the function of IKs channels in the heart. Therefore, both PIP2 molecules are vital for the physiological functions of KCNQ1 channels. Results Conformations of KCNQ1 in various functional states Voltage dependent gating of KCNQ1 involves two measurable steps of VSD activation to the I and A states, both of which trigger pore opening to the IO and AO (Fig. S1a). Previous structural studies indicate that KCNQ1 may adopt a bent conformation in the resting-closed (RC) state32 and a straight conformation in the AO state11,30. Then, what is the conformation, bent or straight, in the IO state? To address this question, we trapped the VSD of KCNQ1 in I state by introducing E160R/R231E (E1R/R2E) mutations24,33 (Fig. 1a). The structure of KCNQ1-E1R/R2E in the I state is determined with overall resolution of 3.47 Å (Table. S1) and shares high similarity with KCNQ1-PIP2 structure (PDB: 9VEN), and both adopt a bent conformation: the S6 and HA helices form a helix-loop-helix structure. The R237 (R4) residue of the S4 helix points to the gate charge transfer center formed by F167, E170, and D202, while R231 (R2) and H240 (H5) of S4 helix sit within the gate charge transfer center in the resting state (PDB: 8SIN) and activated state (PDB: 9VEI), respectively. This supports that the structure is captured in the I state, with the S4 ~ 5 Å lower than in the A state34. The ion conductance pathway plotted using HOLE program shows that the pore radius (~1 Å at the narrowest point) is too small for hydrated potassium to pass, suggesting that the structure may represent an intermediate-closed (IC) state. In the structure, one PIP2 molecule is observed at the V-PIP2 site, which is mainly formed by the S2S3L, S3, and S4-S5 linker (S4S5L) (Fig. 1a–c).Fig. 1Conformations of KCNQ1 in various functional states.a Cryo-EM density map of KCNQ1-E160R/R231E IC state in the presence of PIP2. KCNQ1 and CaM are colored in blue and orange, respectively. The head and tail groups of PIP2 are colored red and black, respectively, to show its location. b Comparison of voltage sensor domains between KCNQ1-RC state (PDB: 8SIN), KCNQ1-I state (PDB: 8SIM), KCNQ1-AO state (PDB: 9VEO), and KCNQ1-E160R/R231E intermediate state in this study. Only the helical regions of S2-S4 and S4S5L are shown for clarity. The C-alpha atoms of the positive charged (or polar) resides on S4 are shown in yellow spheres with stick side chains, and the gating charge transfer center residues F167 (green), E170, and D202 are shown as sticks. Mutated residues are labeled with red color. c The ion conductance path and core radius for KCNQ1 bent conformation (PDB: 6UZZ), KCNQ1 with PIP2 IC bent conformation, and KCNQ1-KCNE1 with PIP2 AO straight conformation (PDB: 9VEI), respectively. Figures are plotted using the HOLE program. d Representative current traces of one interaction pairs between KCNQ1-S2S3L and CaM: KCNQ1-Y184W and CaM-D96R, displayed without (upper) and with (lower) KCNE1 co-expression. e Current amplitude comparison (+40 mV without KCNE1, +80 mV with KCNE1) of all four identified interaction pairs, shown without (upper) and with (lower) KCNE1. The color of bars was the same as the current traces. The holding potential is −80 mV for all electrophysiology recordings. For this and subsequent figures, statistical analysis was conducted using an unpaired two-tailed t-test with the control. The data points represent the mean ± SEM. Statistical significance was designated as “*” for p ≦ 0.05, “**” for p ≦ 0.005, and “***” for p ≦0.0005; “NS.” represents no statistical difference. For this figure, all mean ± SEM values, the actual number of samples (n), and the p values for each dataset are included in the legends of Figs. S3, 7, 9, 11 within the Supplementary information. Conformations of KCNQ1 in various functional states. a Cryo-EM density map of KCNQ1-E160R/R231E IC state in the presence of PIP2. KCNQ1 and CaM are colored in blue and orange, respectively. The head and tail groups of PIP2 are colored red and black, respectively, to show its location. b Comparison of voltage sensor domains between KCNQ1-RC state (PDB: 8SIN), KCNQ1-I state (PDB: 8SIM), KCNQ1-AO state (PDB: 9VEO), and KCNQ1-E160R/R231E intermediate state in this study. Only the helical regions of S2-S4 and S4S5L are shown for clarity. The C-alpha atoms of the positive charged (or polar) resides on S4 are shown in yellow spheres with stick side chains, and the gating charge transfer center residues F167 (green), E170, and D202 are shown as sticks. Mutated residues are labeled with red color. c The ion conductance path and core radius for KCNQ1 bent conformation (PDB: 6UZZ), KCNQ1 with PIP2 IC bent conformation, and KCNQ1-KCNE1 with PIP2 AO straight conformation (PDB: 9VEI), respectively. Figures are plotted using the HOLE program. d Representative current traces of one interaction pairs between KCNQ1-S2S3L and CaM: KCNQ1-Y184W and CaM-D96R, displayed without (upper) and with (lower) KCNE1 co-expression. e Current amplitude comparison (+40 mV without KCNE1, +80 mV with KCNE1) of all four identified interaction pairs, shown without (upper) and with (lower) KCNE1. The color of bars was the same as the current traces. The holding potential is −80 mV for all electrophysiology recordings. For this and subsequent figures, statistical analysis was conducted using an unpaired two-tailed t-test with the control. The data points represent the mean ± SEM. Statistical significance was designated as “*” for p ≦ 0.05, “**” for p ≦ 0.005, and “***” for p ≦0.0005; “NS.” represents no statistical difference. For this figure, all mean ± SEM values, the actual number of samples (n), and the p values for each dataset are included in the legends of Figs. S3, 7, 9, 11 within the Supplementary information. In functional studies, while the KCNQ1-E1R/R2E channels showed constitutive macroscopic currents independent of voltage24,33, single-channel recordings exhibited flickering openings with low open probability28. These results are consistent with the KCNQ1-E1R/R2E structure being in the IC conformation. Then, does the IO state adopt the bent conformation? In KCNQ1 channel structures, each subunit binds a CaM molecule at the cytosolic domain11,29,30,32,34. In the bent conformation, CaM interacts with the S2S3L on the cytosolic side of the VSD (Fig. 1a). During VSD activation (R → I → A), if the channel remains bent, with S2S3L and CaM likely interacting differently before reaching the straight A state, these S2S3L-CaM interactions would influence activation of the IO state more than to the AO state. To test this hypothesis, we performed mutational scanning in both CaM (D94 and D96) and S2S3L (C180, R181, S182, K183, Y184, L191, and R192) as well as in S3 (R195 and K196) to evaluate the function of individual mutations and their pairs in the activation of KCNQ1 and IKs, respectively (Figs. S2–12). Four pairs of these mutations are noteworthy: KCNQ1-C180D in S2S3L paired with CaM-D94K and R195Q in S3 paired with CaM-D94K (Figs. 1e and S3a, b, S11a, b), and KCNQ1-Y184W and L191D in S2S3L, each paired with CaM-D96R (Figs. 1d, e and S7c, d, S9c, d). Among these four pairs of mutations, three double mutations (KCNQ1-C180D and CaM-D94K, KCNQ1-R195Q and CaM-D94K, KCNQ1-Y184W and CaM-D96R) caused near complete loss of current compared to single mutations in KCNQ1 or CaM alone, whereas the KCNQ1-L191D and CaM-D96R double mutation pair significantly recued the current of KCNQ1-L191D mutation (Figs. 1e and S9c, d). Such drastic changes in current amplitude by the double mutations suggest that these paired residues may interact during channel opening. The four mutation pairs specifically affect KCNQ1, as their double mutations in IKs produced current amplitudes similar to those of single KCNQ1 or CaM mutations (Figs. 1d, e and S3a, b, S7c, d, S9c, d, S11a, b). It is unlikely that these double mutations reduce surface expression, as evidenced by the robust currents exhibited by the mutant IKs complexes, although it is known that KCNQ1 and KCNE1 traffic to the plasma membrane via independent pathways in both cardiac myocytes and Xenopus oocytes35. Since KCNQ1 opens predominantly to the IO state and IKs opens exclusively to the AO state, these results support that the channel adopts a bent conformation in the IO state and a straight conformation in the AO state. Unlike the mutations of KCNQ1-S182 and CaM-Y100 in our previous studies, which changed the kinetics and voltage dependence of activation of KCNQ1 to be more like those in IKs31, these mutations affected current amplitude of the channel (Fig. 1d, e). These results suggested that, instead of affecting the IO-AO transition31, the mutations reported here (Fig. 1d, e) may alter the interaction between KCNQ1-VSD and CaM during activation to change the conformation of the IO state. Further support for the correlation between the conformation and functional state is shown below. C-PIP2 binding is essential for channel activation In the KCNQ1-E1R/R2E structure, no obvious C-PIP2 was observed. Is C-PIP2 important for both the IO and AO states, and what role does C-PIP2 play? The KCNQ1-KCNE1 structure reveals that C-PIP2 interacts with residues in the S4S5 linker (V255 and F256), S5 (R259, Q260, L262 and L263) of one KCNQ1 subunit, and S6-HA (Q359, K362, and R366) of a neighboring KCNQ1 subunit, and the KCNE1 (F57, I61, L63, S64, R67, S68, K69, and K70) subunit (Fig. 2a). First, we mutated each C-PIP2-interacting residue in both KCNQ1 and KCNE1 to alanine (Ala) or neutralized charged amino acids (glutamine Q or asparagine N) and assessed the function of the mutant channels. All mutations, except KCNE1-F57A, significantly reduced IKs current amplitudes (Figs. 2b, c and S13). The similar effect of KCNE1-K70N was reported previously10. The C-PIP2 site mutations also reduced KCNQ1 current amplitude in the absence of KCNE1 (Figs. 2b, c and S13). These results are consistent with observations made when membrane PIP2 levels were reduced10,11, indicating that these mutations diminish PIP2 binding to the KCNQ1 channel and that C-PIP2 is critical for the conductance of KCNQ1 and IKs. Importantly, all these mutations caused only minor shifts of the voltage-dependent activation, with the conductance-voltage (G–V) relationship deviating from the wild-type (WT) less than 25 mV (Fig. 2d and S13). This is consistent with the mechanism whereby C-PIP2 binding is crucial for coupling of VSD movements and pore opening but not for VSD activation7.Fig. 2C-PIP2 binding is essential for channel activation.a C-PIP2 binding site in the KCNQ1-KCNE1-PIP2 complex (PDB: 9VEI). KCNQ1, KCNE1 and PIP2 are colored in blue, magenta and black, respectively. C-PIP2 is shown with ball-and-stick model. Orange color represents the Phosphorus (P) atoms, black represents Carbon (C) atoms, and red represents Oxygen (O) atoms. Residues including interactions with C-PIP2 are displayed. Key interactions are labeled with dashed lines. b Representative current recordings of KCNQ1-WT (left) and mutant KCNQ1-R259Q (middle) channels, displayed without (upper) and with (lower) KCNE1 co-expression. Right: G–V relationship comparison with WT channels. c Current amplitude comparison (+40 mV without KCNE1, +80 mV with KCNE1) of mutagenesis screens of all labeled C-PIP2 binding sites in (a). The black bars show without KCNE1, and the red bars show with KCNE1. d V1/2 of mutagenesis screens (black for without KCNE1, red for with KCNE1) of all labeled C-PIP2 binding sites in (a). e Structure depiction of the straight conformation of one monomer of tetrameric KCNQ1-KCNE1-CaM complex (PDB: 9VEI) in the presence of C-PIP2. KCNQ1, KCNE1, PIP2, and CaM are colored in blue, magenta, black, and orange, respectively. “RQKH” motif is colored in green. f Molecular docking of C-PIP2 into the bent conformation of KCNQ1-E1R/R2E. Secondary structure of the neighboring KCNQ1 subunit is labelled with S6’ and HA’. g Structure depiction of C-PIP2 binding residues specific to KCNQ1 from molecular docking. h Representative current recordings of KCNQ1-WT (left) and mutant KCNQ1-T264A (middle) channels, displayed without (upper, black) and with (lower, red) KCNE1 co-expression. Right: G–V relationship comparison with WT channels. i Current amplitude comparison of C-PIP2 binding residue mutations identified by docking in (g). Black: without KCNE1, red: with KCNE1. All mean ± SEM values, the actual number of samples (n), and the p values for each data are included in the Supplementary Table3 within the Supplementary information. C-PIP2 binding is essential for channel activation. a C-PIP2 binding site in the KCNQ1-KCNE1-PIP2 complex (PDB: 9VEI). KCNQ1, KCNE1 and PIP2 are colored in blue, magenta and black, respectively. C-PIP2 is shown with ball-and-stick model. Orange color represents the Phosphorus (P) atoms, black represents Carbon (C) atoms, and red represents Oxygen (O) atoms. Residues including interactions with C-PIP2 are displayed. Key interactions are labeled with dashed lines. b Representative current recordings of KCNQ1-WT (left) and mutant KCNQ1-R259Q (middle) channels, displayed without (upper) and with (lower) KCNE1 co-expression. Right: G–V relationship comparison with WT channels. c Current amplitude comparison (+40 mV without KCNE1, +80 mV with KCNE1) of mutagenesis screens of all labeled C-PIP2 binding sites in (a). The black bars show without KCNE1, and the red bars show with KCNE1. d V1/2 of mutagenesis screens (black for without KCNE1, red for with KCNE1) of all labeled C-PIP2 binding sites in (a). e Structure depiction of the straight conformation of one monomer of tetrameric KCNQ1-KCNE1-CaM complex (PDB: 9VEI) in the presence of C-PIP2. KCNQ1, KCNE1, PIP2, and CaM are colored in blue, magenta, black, and orange, respectively. “RQKH” motif is colored in green. f Molecular docking of C-PIP2 into the bent conformation of KCNQ1-E1R/R2E. Secondary structure of the neighboring KCNQ1 subunit is labelled with S6’ and HA’. g Structure depiction of C-PIP2 binding residues specific to KCNQ1 from molecular docking. h Representative current recordings of KCNQ1-WT (left) and mutant KCNQ1-T264A (middle) channels, displayed without (upper, black) and with (lower, red) KCNE1 co-expression. Right: G–V relationship comparison with WT channels. i Current amplitude comparison of C-PIP2 binding residue mutations identified by docking in (g). Black: without KCNE1, red: with KCNE1. All mean ± SEM values, the actual number of samples (n), and the p values for each data are included in the Supplementary Table3 within the Supplementary information. Among the exanimated mutations, KCNQ1-I263A, Q359A, K362A, and R366A produced less pronounced reductions in KCNQ1 current amplitude than the reduction observed in the IKs current (Fig. 2c). These results indicate that the C-PIP2 binding sites in KCNQ1 may alter in the absence of KCNE132. Notably, residues Q359, K362 and R366 within or around the “RQKH” motif, which is the hinge for the bent and straight conformation switch (Fig. 2e), interact with C-PIP2 primarily in the presence of KCNE1 (Fig. 2c). Thus, the binding of C-PIP2 to these residues may help KCNE1 stabilize the straight conformation7,10, while the disruption of the “RQKH” motif from it α-helical structure into a loop may alter C-PIP2 binding. To verify this notion, we evaluated the structure of the C-PIP2 binding sites in the bent conformation. Since C-PIP2 binding was not observed in any available structures of KCNQ1 in the bent conformation, we performed molecular docking of C-PIP2 to the bent conformation of KCNQ1 (Figs. 2f, g and S14a–c). The results revealed that the residues I263, Q359, K362, and R366 no longer interact with C-PIP2 (Fig. 2f). These findings align with the functional data showing that mutations of these residues reduce KCNQ1 current amplitudes to a lesser extent than IKs (Figs. 2c and S14d–g). Molecular docking further revealed that, in the bent conformation, residues T264, K358, and Q361 interacted with the C-PIP2 (Fig. 2g), which are not part of the C-PIP2 site in the IKs structure. As shown in Fig. 2g, the negatively charged phosphomonoester groups of C-PIP2 form salt bridge with the positively charged residues K354 and K358 and may also form a hydrogen bond with the side chain of Q361. The side chain of T264 may form a hydrogen bond with an oxygen atom on the tail of C-PIP2. The surface representation indicates that the binding region near the C-PIP2 tail is relatively hydrophobic, consistent with the hydrophobic nature of the PIP2 tail (Figs. S14a, b). Mutation of these residues reduced KCNQ1 current amplitudes to a greater extent than IKs (Figs. 2g-i and S14h, i). These results indicate that the opening of KCNQ1 channel is associated with the bent conformation, whereas the opening of IKs channel is associated with the straight conformation, and the C-PIP2 site alters between the two conformations. Our previous findings indicated that both KCNQ1 and IKs require PIP2 for channel function, but the EC50 of PIP2 dose-response for KCNQ1 was over 100-fold greater than that for IKs, suggesting a higher PIP2 affinity for IKs10. The differences in C-PIP2 binding sites within KCNQ1 across different states, combined with the role of KCNE1 in facilitating C-PIP2 binding, explain the ﹥100-fold increase in PIP2 sensitivity observed in IKs channels compared to KCNQ1 alone. This mechanism also accounts for the observation that no PIP2 is resolved in the C-site in the bent conformation due to its low affinity. V-PIP2 binding is essential for channel activation to AO but not to IO In the structure of KCNQ1-E1R/R2E, V-PIP2 interacts with residues in the S2S3L (R181, K183, and Y184), the S3 (K196 and I198) and S4S5L (Q244, W248, and R249) (Figs. 1a and 4a). In the structure of KCNQ1-KCNE1 (PDB: 9VEI), V-PIP2 additionally interacts with the N-terminus S0 (Y111 and R116) of KCNQ1, as well as with KCNE1 (E72 and H73) (Fig. 3a). We mutated these V-PIP2 interacting residues to A or N/Q using mutational scanning. Most mutations in KCNQ1 (R181, K183, Y184, K196, and R249) did not reduce but rather increased KCNQ1 current amplitudes (Figs. 3b, c and S15, S16), with the exceptions of Q244N and W248A (Figs. 3c and S16a, b), which are critical for VSD-pore coupling25. Y111A and R116A reduced KCNQ1 current amplitudes (Figs. 3c and S15a, b), but these effects may not be due to a direct disruption of V-PIP2 binding because in the KCNQ1-E1R/R2E structure, which may correspond to the bent conformation and the IO state, Y111 and R116 do not interact with V-PIP2. On the other hand, most of the mutations significantly reduced the current amplitudes of IKs except for the KCNQ1-R181Q + KCNE1 and KCNQ1-K183N + KCNE1 (Figs. 3c and S15c, d). Although these two mutations enhanced IKs current amplitudes, the enhancement was significantly smaller compared to the enhanced KCNQ1 current amplitudes of the same mutant. All mutations induce small GV shifts in both KCNQ1 (less than 15 mV) and IKs (less than 25 mV) (Figs. 3d and S15, S16), indicating that the mutations did not modify current amplitudes by altering voltage dependence. Additionally, we observed that mutations at the V-PIP2 sites did not significantly affect the current amplitudes of KCNQ1 + KCNE3 (Fig. 3e, f), whereas mutations at the C-PIP2 sites drastically reduced the current amplitudes of KCNQ1 + KCNE3 (Fig. S17). Thus, these results suggest that V-PIP2 is essential for the AO state but not the IO state, as both KCNQ1 and KCNQ1 + KCNE3 primarily open to the IO state, whereas IKs exclusively open to the AO state24,25,28,34.Fig. 3V-PIP2 binding is essential for channel activation to AO but not to IO.a V-PIP2 binding site in the KCNQ1-KCNE1-PIP2 complex (PDB: 9VEI). KCNQ1, KCNE1, and V-PIP2 are colored in blue, magenta, and black, respectively. Orange color represents the Phosphorus (P) atoms, black represents Carbon (C) atoms, and red represents Oxygen (O) atoms. Residues interaction with V-PIP2 are displayed. Key interactions are labeled with dashed lines. b Representative current recordings of KCNQ1-WT (left) and mutant KCNQ1-Y184W (middle) channels, displayed without (upper, black) and with (lower, red) KCNE1 co-expression. Right: G-V relationship comparison with WT channels. c Current amplitude comparison of mutagenesis screens (+40 mV without KCNE1, +80 mV with KCNE1) of all labeled V-PIP2 binding sites on (a). The black bars show without KCNE1, and the red bars show with KCNE1. d V1/2 of mutagenesis screens (black bars: without KCNE1, red bars: with KCNE1) of all labeled V-PIP2 binding sites in (a). e Representative current recordings of KCNQ1-WT + KCNE3-WT (black), mutant KCNQ1-Y184W (red) + KCNE3-WT and KCNQ1-K196Q (green) + KCNE3-WT channels. f The current amplitudes (at +40 mV) for the respective channels. The color of bars was the same as the current traces. g Representative current recordings of KCNQ1-S338F (upper, IO only) and KCNQ1-F351A (lower, AO only), and mutant KCNQ1-Y184W on KCNQ1-S338F and KCNQ1-F351A, respectively. Right: average G–V relationship comparison for the respective channels with KCNQ1-S338F (upper) and KCNQ1-F351A (lower), respectively. h Current amplitude comparison of selected V-PIP2 mutants (+40 mV with KCNQ1-S338F, brown; +80 mV with KCNQ1-F351A, orange). The color of bars was the same as the current traces. All mean ± SEM values, the actual number of samples (n), and the p values for each data are included in the Supplementary Tables 4–7 within the Supplementary information. V-PIP2 binding is essential for channel activation to AO but not to IO. a V-PIP2 binding site in the KCNQ1-KCNE1-PIP2 complex (PDB: 9VEI). KCNQ1, KCNE1, and V-PIP2 are colored in blue, magenta, and black, respectively. Orange color represents the Phosphorus (P) atoms, black represents Carbon (C) atoms, and red represents Oxygen (O) atoms. Residues interaction with V-PIP2 are displayed. Key interactions are labeled with dashed lines. b Representative current recordings of KCNQ1-WT (left) and mutant KCNQ1-Y184W (middle) channels, displayed without (upper, black) and with (lower, red) KCNE1 co-expression. Right: G-V relationship comparison with WT channels. c Current amplitude comparison of mutagenesis screens (+40 mV without KCNE1, +80 mV with KCNE1) of all labeled V-PIP2 binding sites on (a). The black bars show without KCNE1, and the red bars show with KCNE1. d V1/2 of mutagenesis screens (black bars: without KCNE1, red bars: with KCNE1) of all labeled V-PIP2 binding sites in (a). e Representative current recordings of KCNQ1-WT + KCNE3-WT (black), mutant KCNQ1-Y184W (red) + KCNE3-WT and KCNQ1-K196Q (green) + KCNE3-WT channels. f The current amplitudes (at +40 mV) for the respective channels. The color of bars was the same as the current traces. g Representative current recordings of KCNQ1-S338F (upper, IO only) and KCNQ1-F351A (lower, AO only), and mutant KCNQ1-Y184W on KCNQ1-S338F and KCNQ1-F351A, respectively. Right: average G–V relationship comparison for the respective channels with KCNQ1-S338F (upper) and KCNQ1-F351A (lower), respectively. h Current amplitude comparison of selected V-PIP2 mutants (+40 mV with KCNQ1-S338F, brown; +80 mV with KCNQ1-F351A, orange). The color of bars was the same as the current traces. All mean ± SEM values, the actual number of samples (n), and the p values for each data are included in the Supplementary Tables 4–7 within the Supplementary information. To validate this mechanism, we introduced some of the V-PIP2 site mutations into the mutant KCNQ1-S338F and KCNQ1-F351A, which were previously established as opening only in the IO and AO state, respectively24,28. Notably, these mutations either had no impact (Y184, K196, I198, Q244, and R249) or enhanced KCNQ1-S338F current amplitudes (K183), while significantly reducing KCNQ1-F351A current amplitudes (Figs. 3g, h and S18), supporting the notion that V-PIP2 is specifically required for the AO state, but not the IO state. By contrast, KCNQ1-R259Q, which affects C-PIP2 binding, reduced the current amplitudes of both KCNQ1-S338F and KCNQ1-F351A (Fig. S19). V-PIP2 modulates IO to AO transition Comparing V-PIP2 and its binding site in the structures of KCNQ1-E1R/R2E and KCNQ1-KCNE1 (Fig. 4a), notable differences were observed in the interaction network of PIP2. In the I state, the head group of PIP2 contacts S4, S4S5L, S2S3L, and CaM. Upon further activation of the VSD, S4 moves upward, followed by S4S5L and S2S3L, a motion that appears to elevate the center of the PIP2-binding pocket. This motion weakens interactions between CaM and the rest of the V-PIP2 site, facilitating the alteration of the V-PIP2 site and repositioning of V-PIP2 to form closer interactions with S2S3L (Figs. 4a and S20). The upward movement of the PIP2-interaction network, the detachment of CaM from the V-PIP2 site, and the alteration of V-PIP2 binding are associated with the “bent-to-straight” conformational transition of the KCNQ1 channel.Fig. 4V-PIP2 modulates IO to AO transition.a V-PIP2 binding pocket with key residues displayed for KCNQ1-E1R/R2E bent conformation (left, this study) and KCNQ1-KCNE1 straight conformation (right, PDB: 9VEI), respectively. KCNQ1, CaM, KCNE1, and V-PIP2 are colored in blue, orange, magenta, and black, respectively. Orange color represents the Phosphorus (P) atoms, black represents Carbon (C) atoms, and red represents Oxygen (O) atoms. Residues including interactions with V-PIP2 are displayed. Key interactions are labeled with dashed lines. b In silico screening of CA1 to the straight conformation of KCNQ1-KCNE1 (PDB: 9VEI) and the bent conformation of KCNQ1 (PDB: 6UZZ). KCNQ1, KCNE1, and CA1 are colored in tan, sky blue and red, respectively. The CA1 binding pocket is represented in orange. c CA1 effects on IKs channels. Current traces were recorded before (black) and after (red) adding 10 μM CA1. Average G-V relationship (lower) of IKs channels before (black) and after (red) adding 10 μM CA1. The EC50 of CA1 is 7.22 ± 0.57 μM (Fig. S13b). d Current amplitude comparative effects of CA1 on KCNQ1-WT (yellow), KCNQ1-WT + KCNE3-WT (green), IKs, and putative binding residues (rose pink) from molecular docking. Changes in current amplitudes were evaluated at +40 mV without KCNE1 and at +80 mV with KCNE1. e Representative current recordings of CA1 effects on KCNQ1-WT (upper) and KCNQ1-WT + KCNE3-WT (lower). Current traces were recorded before (black) and after (red) adding 10 μM CA1. All mean ± SEM values, the actual number of samples (n), and the p values for each data are included in the Supplementary Table 8 within the Supplementary information. f Representative structures from the Metadynamics MD simulation illustrating the conformational transition from the bent conformation to the straight conformation (left to right). CaM is shown in cyan, V-PIP2 in green, S2S3L in pink, S4 in orange, and S6 in yellow. V-PIP2 modulates IO to AO transition. a V-PIP2 binding pocket with key residues displayed for KCNQ1-E1R/R2E bent conformation (left, this study) and KCNQ1-KCNE1 straight conformation (right, PDB: 9VEI), respectively. KCNQ1, CaM, KCNE1, and V-PIP2 are colored in blue, orange, magenta, and black, respectively. Orange color represents the Phosphorus (P) atoms, black represents Carbon (C) atoms, and red represents Oxygen (O) atoms. Residues including interactions with V-PIP2 are displayed. Key interactions are labeled with dashed lines. b In silico screening of CA1 to the straight conformation of KCNQ1-KCNE1 (PDB: 9VEI) and the bent conformation of KCNQ1 (PDB: 6UZZ). KCNQ1, KCNE1, and CA1 are colored in tan, sky blue and red, respectively. The CA1 binding pocket is represented in orange. c CA1 effects on IKs channels. Current traces were recorded before (black) and after (red) adding 10 μM CA1. Average G-V relationship (lower) of IKs channels before (black) and after (red) adding 10 μM CA1. The EC50 of CA1 is 7.22 ± 0.57 μM (Fig. S13b). d Current amplitude comparative effects of CA1 on KCNQ1-WT (yellow), KCNQ1-WT + KCNE3-WT (green), IKs, and putative binding residues (rose pink) from molecular docking. Changes in current amplitudes were evaluated at +40 mV without KCNE1 and at +80 mV with KCNE1. e Representative current recordings of CA1 effects on KCNQ1-WT (upper) and KCNQ1-WT + KCNE3-WT (lower). Current traces were recorded before (black) and after (red) adding 10 μM CA1. All mean ± SEM values, the actual number of samples (n), and the p values for each data are included in the Supplementary Table 8 within the Supplementary information. f Representative structures from the Metadynamics MD simulation illustrating the conformational transition from the bent conformation to the straight conformation (left to right). CaM is shown in cyan, V-PIP2 in green, S2S3L in pink, S4 in orange, and S6 in yellow. To validate the observed structural difference in V-PIP2 binding between the I and A states, we employed small-molecule screening approaches36. We reasoned that a compound selectively binding to the V-PIP2 site in the straight conformation would modulate the IKs channel without affecting KCNQ1 alone in the I state. We screened the Available Chemical Database (ACD, Molecular Design Ltd.) by docking its compound to the V-PIP2 binding sites of KCNQ1-KCNE1 in the straight conformation. Docking results revealed that the compound CA1 (Fig. S21a) selectively binds to the straight conformation of the channel, with the binding site nearly overlapping with the V-PIP2 site. While the CA1 binding pocket is observed in the straight conformation of the KCNQ1-KCNE1 complex (Fig. 4b), this pocket is not present in the bent conformation of KCNQ1 (Fig. 4b). The binding pocket undergoes structural rearrangement in the bent conformation, preventing CA1 from binding (Fig. 4b). Our functional data showed that CA1 significantly enhanced the current amplitudes of IKs and shifted its GV to more negative voltages (Fig. 4c). Mutations in the putative CA1-binding residues (Fig. S21) attenuated the effects of CA1 (Figs. 4d and S21), supporting our docking results for the CA1 binding site in the straight conformation of the IKs channel, which opens exclusively to the AO state. Conversely, CA1 exhibited no enhancing effect on channels predominantly open to the IO state in the bent conformation, including KCNQ1 alone and the KCNQ1-KCNE3 complex (Fig. 4d, e). Notably, KCNQ1-KCNE3 could adopt a straight conformation and open in the AO state, albeit with a low probability30. Docking of CA1 onto the straight conformation of KCNQ1-KCNE3 showed that CA1 could bind to the same pocket (Fig. S21). Thus, the absence of CA1-induced current augmentation in KCNQ1 + KCNE3 channels is consistent with the complex primarily conducting in the IO state via the bent conformation, which precludes CA1 binding. To support the above mechanism that the VSD movement to the A state and the change in V-PIP2 trigger the bent to straight conformational switch, we performed molecular dynamics (MD) simulations by first constructing a hybrid starting model. In this model, the VSD and pore domains (residues 1–236) preceding the bending point (RQKH, residues 251–254), including the bound PIP2, were taken from the straight conformation, whereas the remaining regions were retained from the bent conformation. We then performed metadynamics simulations to accelerate sampling along the “bent-to-straight” transition. Representative snapshots and the time evolution of the distance between S2S3L and CaM are shown in Figs. 4f and S22a. Along the biased trajectory, CaM progressively dissociated from the VSD and rotated while KCNQ1 spontaneously adopted a straight configuration. These conformational changes align with our structures of the I and A states and support the mechanism in which upward S4 movement facilitates VSD activation and conformational changes in the V-PIP2 binding site, displacing CaM and stabilizing a straight cytosolic conformation, thus promoting the AO state. To visualize the conformational transition from the bent conformation to the straight conformation, we also performed targeted MD simulations. Representative snapshots illustrate this transition, which is further quantified by the time evolution of the RMSD and its correlation with the S2S3L-CaM distance (Figs. S22b, S23 and Movie S1). During the transition, the S4 helix exhibits an upward movement compared to its position in the bent conformation (Fig. S23a). The phosphatidylinositol head group of V-PIP2 undergoes a reorientation toward S2S3L, establishing contacts with this linker (Fig. S23b). Concurrently, CaM gradually dissociates from the VSD. A model of the activation of KCNQ1 and IKs by VSD, PIP2 and CaM In this study, we show that the correlation between the conformation (bent and straight) and the functional state (IO and AO) with the following evidence. (1) The structure of KCNQ1-E1R/R2E, which was functionally in a constitutive IO state24, adopted a bent conformation (Fig. 1a). (2) Paired mutations in the KCNQ1 VSD and CaM exhibited interactions specifically in KCNQ1 without KCNE1 (primarily opens to IO) but not in the presence of KCNE1 (exclusively opens to AO) (Fig. 1d, e), corroborating with the structure of KCNQ1-E1R/R2E (Fig. 1a). (3) C-PIP2 binding site differed in KCNQ1 and KCNQ1 + KCNE1, which were consistent with the bent conformation in KCNQ1 and the straight conformation in KCNQ1 + KCNE1 (Fig. 2). (4) The V-PIP2 site differed in the bent and straight conformation as shown by the structure (Fig. 4a) and CA1 docking (Fig. 4b), while functionally CA1 specifically modulated KCNQ1 + KCNE1 but not KCNQ1, consistent with the bent-IO and straight-AO correlations. Based on our findings and prior studies, we propose a model for KCNQ1 and IKs activation (Fig. 5). In this model, V-PIP2 interacts with the channel via a voltage-dependent mechanism to trigger the bent to straight and IO to AO transition, while C-PIP2 is associated with the channel to mediate the VSD-pore coupling and stabilize the straight conformation. The recently published structure of KCNQ1 with the VSD in the RC state indicated that the channel adopts a bent conformation, which obstructs the V-PIP2 site and prevents its binding32. Therefore, V-PIP2 binds to the channel upon VSD activation to the I state, in both IC and IO, which adopt the bent conformation (Fig. 5). V-PIP2 is subsequently reoriented upon VSD further activation to the A state, triggering the switch from bent to straight, which culminates in the activating-closed (AC) and AO states (Figs. 3 and 4f). C-PIP2 binds to the channel in both the I and A states to facilitate the VSD-pore coupling during the IC-IO and AC-AO transitions, but whether C-PIP2 binds to the RC state is not clear. In the I state, C-PIP2 binds to the channel with a lower affinity than in the A state owing to the alteration of the C-PIP2 binding sites during the “bent-to-straight” transition (Fig. 2). The KCNE1 association inhibits the IO state and enhances the AO state (Fig. S1a)24,25,28.Fig. 5A model depicting the activation of KCNQ1 by VSD, C-PIP2, V-PIP2, and CaM.In this model, KCNQ1-RC, IC, and IO states adopt a bent conformation, while the AC and AO states adopt a straight conformation. V-PIP2 associates with the channel during activation of the VSD to the I state, including IC and IO, through a voltage-dependent mechanism that triggers the IO-AO transition. CaM and V-PIP2 compete for binding to the S2S3L; when CaM binds, the channel adopts a bent conformation, whereas binding of V-PIP2 results in a straight conformation. C-PIP2 interacts with the channel in both the I and A states, facilitating the coupling of VSD and pore during the IC-IO and AC-AO transitions. A model depicting the activation of KCNQ1 by VSD, C-PIP2, V-PIP2, and CaM. In this model, KCNQ1-RC, IC, and IO states adopt a bent conformation, while the AC and AO states adopt a straight conformation. V-PIP2 associates with the channel during activation of the VSD to the I state, including IC and IO, through a voltage-dependent mechanism that triggers the IO-AO transition. CaM and V-PIP2 compete for binding to the S2S3L; when CaM binds, the channel adopts a bent conformation, whereas binding of V-PIP2 results in a straight conformation. C-PIP2 interacts with the channel in both the I and A states, facilitating the coupling of VSD and pore during the IC-IO and AC-AO transitions. Although V-PIP2 and C-PIP2 play distinctive roles in the activation of KCNQ1 channels, the two PIP2 molecules may have allosteric connections in the function of IKs channels since both PIP2 molecules bind to KCNE1 in the straight conformation (Figs. 2 and 3). This result would suggest that both V-PIP2 and C-PIP2 binding in the straight conformation and AO state may be positively cooperative and stabilize the straight conformation synergistically due to this allosteric connection via KCNE1. We propose that, even in association with KCNE1, the KCNQ1 subunit undergoes the same bent and straight transition as the VSD moves from the I state to the A state, based on the following observations. First, V-PIP2 is essential for the activation of the KCNQ1 + KCNE1 channel (Fig. 3), potentially by stabilizing the straight conformation. Second, C-PIP2 in IKs exhibits a state-dependent change in binding affinity; in the open state (AO, straight), the affinity is enhanced, and C-PIP2 is resistant to digestion by the lipid phosphatase CiVSP. By contrast, the C-PIP2 affinity decreases following channel deactivation at hyperpolarized voltages (IC and RC, bent), allowing for digestion by CiVSP37,38. The correspondence between RC, IC, and IO (all adopting bent conformation) and AO (adopting the straight conformation) is supported by structural and functional evidence from previous studies24,25,28–31,34 and from this study. However, whether the AC state corresponds to the straight conformation is not supported by available data and remains an assumption in the model. This model accounts for many aspects of KCNQ1/IKs channel function and underscores the pivotal roles of PIP2 in their gating mechanism. To date, three intermediate-state structures of human KCNQ1 are available; one in liposomes without PIP2 (PDB: 8SIM)32, the KCNQ1-E1R/R2E mutant bound to PIP2 from this study, and KCNQ1 bound to both PIP2 and KCNE1 (PDB: 9U7F)39. In all three, R4 from S4 occupies the gating charge transfer center, with the KCNQ1-KCNE1 complex with PIP2 showing the most pronounced downward displacement of the S4 helix and S4-S5 linker (Figs. S24a, b). The VSD exhibits flexibility across these structures, accompanied by rotational movement, and the cytosolic CaM and pore helices also displaced relative to one another (Figs. S24b–d), suggesting the dynamic nature of the intermediate states. While our manuscript was in preparation, the intermediate-state structure of Xenopus KCNQ1 (PDB: 9MY4)40 was published using the same E1R/R2E construct. Detailed comparison of human KCNQ1 intermediate structure with the Xenopus counterpart reveals shared features, including straightened S1 and S2 helices displaced away from the pore domain (Fig. S25) and R4 of S4 positioned in the charge-transfer center (Fig. S25b). Compared with activated state, human KCNQ1 intermediate structure shows a counterclockwise rotation and displacement of the VSD when pore domains are aligned (Figs. S17f, h). The S1/S2 straightening likely stems from the introduced mutations, as it is absent in non-mutated intermediate structures (PDB: 8SIM and PDB: 9U7F) (Fig. S24a). Notable differences between the two include a counterclockwise rotation (Fig. S25c) and upward displacement of the human VSD, when pore domains were aligned (Figs. S25d, e). Most importantly, different from the Xenopus structure, our human KCNQ1-E1R/R2E structure was captured with PIP2 bound at the V site (Figs. 1a and 3a). Compounds, such as CA1, which target the V-PIP2 binding site, modulate the activity of IKs channel (Fig. 4b–e). Since CA1 and V-PIP2 bind to almost identical sites, their binding is likely to have competing effects. CA1 may specifically modify cardiac physiology without affecting other tissues, such as epithelium, due to the tissue-specific distribution of KCNE1. This feature potentially aids in the development of safe and effective antiarrhythmic therapies. Methods Ethical statement Stage Ⅴ or Ⅵ oocytes were procured from Xenopus laevis via laparotomy, in accordance with the protocol approved by the Washington University Animal Studies Committee (protocol #24-0405). Construct design DNA sequences encoding human KCNQ1 (Homo sapiens: NP_000209.2), KCNE1 (Homo sapiens: NP_000210.2), were synthesized by Genewiz. N- and C-terminus loop of KCNQ1 are truncated for stability, resulting in a construct including residues 76–620. I145C and K41C mutations were introduced to KCNQ1 and KCNE1, respectively, for KCNQ1-KCNE1 stabilization and purification. To trap KCNQ1-VSD into I state, E160R (E1R) and R231E (R2E) were introduced to KCNQ1. Constructs were cloned into a BacMam expression vector using EcoR I and Xho I restriction sites, featuring a C-terminal green fluorescent protein (GFP)-6xHis tag linked by a PreScission protease site to the C-terminal of KCNQ1. The full length KCNE1 was cloned into a BacMam expression vector with an IL2 signal peptide followed by an mCherry fluorescence protein tag at N-terminus. Recombinant baculoviruses for KCNQ1-I145C-E1R-R2E and KCNE1-K41C were generated separately using the Bac-to-Bac system according to the manufacturer’s instructions (Invitrogen). Subsequently, P2 virus mixture of KCNQ1-I145C-E1R-R2E and KCNE1-K41C (volume ratio 1:1) was used to transduce HEK293S GnTI- cells (ATCC, Cat #CRL-3022) for protein expression. For 1-liter cultures of HEK293S GnTI- cells (2.0 × 106 cells/mL) in Freestyle 293 media (GIBCO) supplemented with 2% FBS (GIBCO), approximately 100 mL of P2 virus mixture was used. Infected cells were incubated overnight at 37 °C, and protein expression was induced by adding 10 mM sodium butyrate. Cells were then cultured at 30 °C for an additional 48 h before harvest. Protein expression and purification Cell pellet of 2.4 liters culture was resuspended in 50 mL lysis buffer (20 mM HEPES pH 7.4, 150 mM KCl, 4 mM EGTA, 2 μM leupeptin, 1 μM pepstatin, 100 μM PMSF, and 1 μg/mL aprotinin) and subsequently solubilized by adding 50 mL detergent mixture of LMNG/CHS (8:1 mass ratio, 2% stock solubilized in 20 mM HEPES pH 7.4, 150 mM KCl) to a final concentration of 1%, with stirring at 4 °C for 2 h. The solubilized membrane protein was separated from the insoluble fraction by high-speed centrifugation (38,000 × g for 1 h). 3 mL of CNBR-activated Sepharose beads (GE Healthcare) coupled with 3 mg of high-affinity GFP nanobodies (GFP-NB) were equilibrated with 20 mM HEPES pH 7.4, 150 mM KCl, 4 mM EGTA, and 0.05% Digitonin. Protein was incubated with GFP-NB beads for 1 h while rotating in the cold room, followed by column washing with 30 mL buffer containing 0.05% digitonin. The GFP tag was cleaved on the column by using PreScission protease for 2 h or overnight at 4 °C. Solubilized protein was eluted with 3 mL buffer containing 0.05% digitonin. Protein was kept on ice or at 4 °C throughout the purification process. During purification, the pink colored mCherry-tag KCNE1 was dissociated from the KCNQ1 after detergent application. Resolved structure lacks KCNE1, but Calmodulin (CaM, homo sapiens: NP_001734.1) binds to KCNQ1 at a 1:1 stoichiometry, as an essential structural component. Nanodisc reconstitution To prepare the lipid stock, a mixture of soybean polar lipid extract (Soy-PL, Avanti) and brain PIP2 (Avanti) in a 1:1 mass ratio was dissolved in chloroform and dried as a thin film in a glass tube using mild argon stream. The lipid film was resuspended in pentane and dried by argon twice. Then lipid film was solubilized in 20 mM HEPES buffer at pH 7.4 containing a 0.8% DDM/CHS (5:1 mass ratio) mixture, resulting in a final lipid concentration of 10 mM, calculated based on the average molecular weight of the lipid species. KCNQ1-I145C-E1R-R2E was reconstituted into nanodiscs containing a mixture of Soy-PL and brain PIP2. The protein was reconstituted with MSP2N2 and mixture of Soy-PL and brain PIP2 in a final ratio of 1:6:360, prepared as follows. The KCNQ1-I145C-E1R-R2E, eluted from the GFP-NB column, was mixed with the DDM-CHS solubilized lipid and incubated for 30 min at 4 °C to form mixed lipid-detergent micelles. Subsequently, MSP2N2 was added to the protein-lipid mixture for an additional 30-minute incubation before introducing an equal volume of freshly activated Bio-Beads SM2 to remove the detergent from the system. The mixture was incubated overnight at 4 °C with constant rotation. Following the removal of Bio-Beads, the reconstituted mixture was further purified using size-exclusion chromatography with a Superose 6 10/300 GL column (GE Healthcare) equilibrated with 20 mM HEPES pH 7.4, 150 mM KCl, and 4 mM EGTA. The peak fractions were pooled, concentrated to 3-4 mg/mL using a 100-kDa MWCO centrifugal device (Amicon filters), and immediately used for cryo-EM grid preparation. Cryo-EM sample preparation and data collection Quantifoil R1.2/1.3 (400 mesh) holey carbon gold grids were glow-discharged for 30-s. The concentrated protein sample was mixed with 120 mM Fos-Choline-8 at a volume ratio of 1:50 immediately before applying to the grid. 3.5 μL of ~2.5 mg/mL protein sample was applied to each grid, after 20-s waiting, which were double-blotted for 5-s under blot force -3 at 100% humidity and 16 °C, then vitrified by plunging into liquid ethane cooled by liquid nitrogen using Vitrobot Mark IV (FEI). Dataset was acquired on 300 keV Titan Krios microscope (FEI) equipped with a K3 direct electron detector (Gatan) using EPU software with magnification of 130,000. Data collection was conducted in super-resolution mode with pixel size of 0.6485 Å. Images were recorded with a defocus range of −1.0 to −2.0 μm at a dose rate of 14.7 e/frame/s with images captured over 50 frames. Cryo-EM image processing and 3D reconstruction Image stacks were gain-normalized and corrected for beam-induced motion using MotionCor241. The contrast transfer function parameters were estimated from motion-corrected summed images without dose-weighting using CTFFIND442. All subsequent processing steps were performed on motion-corrected, dose-weighted summed images. Data processing was performed in CryoSPARC43. 2D classification was conducted to remove junk particles. Good particles were subjected to Ab-Initio Reconstruction and Heterogeneous Refinement with C1 symmetry. Correct conformational particles were sorted out and input to Non-Uniform Refinement to generate a final reconstruction map with C4 symmetry. Cryo-EM structural refinement and model building Initially, KCNQ1 bent (PDB 6UZZ) was docked into the cryo-EM map KCNQ1-I145C-E160R-R231E. Model was manually built in Coot44. PIP2 molecule was generated as CIF file by the phenix.eLBOW45 and imported as PT5. The structural model was iteratively refined using phenix.real_space_refine46 with secondary structure restraints and checked in Coot. The quality of the structures was assessed using the MolProbity server47 (Fig. S26). The pore radii were calculated using HOLE48. Figures were created using PyMOL (The PyMOL Molecular Graphics System, Version 2.6.0 and 3.1.1, Schrödinger, LLC) and UCSF Chimera49. Molecular docking The interaction between C-PIP2 and KCNQ1 in the bent conformation (Fig. 2f, g) was modeled using AutoDock Vina50. The docking box was set to include residues surrounding the “RQKH” motif, and the side chains of residues Q360, K354, K358, and Q361 at the C-PIP2 binding site were treated as flexible during docking. It is noteworthy that PIP2 has two long and highly flexible fatty acid chains, which pose challenges for molecular docking. To address this, we used a PIP2 structure with shortened tails (each containing four carbon atoms) for docking. The PIP2 structure was treated as flexible. The exhaustiveness parameter was increased from the default value of 8 to 64. A predicted model of the C-PIP2-KCNQ1 complex in the bent conformation is consistent with mutagenesis results (Figs. 2g–i and S14). In silico compound screening The in silico screening strategy employed our recently developed template-guided docking method36 as the search engine to screen a subset of the Available Chemical Database (ACD, Molecular Design Ltd.), consisting of approximately 10,000 compounds. Each compound in this subset carries two formal charges and was screened against the V-PIP2 binding site in the KCNQ1-KCNE1 complex structure. For each compound in the chemical library, OMEGA2 (Version 3.0.1.2, OpenEye, Cadence Molecular Sciences, Santa Fe, NM, USA, http://www.eyesopen.com/)51,52 was used to generate up to 200 conformers. These 3D conformers were then superimposed onto the co-bound V-PIP2 (focused on the head group) using the 3D similarity calculation program SHAFTS53. Subsequently, the molecular docking program AutoDock Vina50 was employed to refine the complex structures generated during the superposition step. Compounds were ranked using a hybrid scoring function implemented in our method. Briefly, the ranking score combines the 3D similarity between each compound and the template PIP2 with the binding score of the compound at the V-PIP2 site. Additional details of our template-guided docking method are provided in ref. 36. The predicted binding modes of the top-ranked compounds were further evaluated by visual inspection. Given that the V-PIP2 binding site contains several positively charged residues (e.g., R181, K183, and K196), we prioritized compounds with negatively charged groups capable of forming salt bridges with these residues. As a result, compound CA1 was identified as an active binder of KCNQ1 + KCNE1 at the V-PIP2 site in experimental assays (Fig. 4c). Metadynamics simulations We examined the conformational transition of the KCNQ1 channel from the bent to straight conformation without applying external steering forces, as is done in targeted MD (see the next subsection). Well-tempered metadynamics simulations were performed to accelerate this transition using Amber2254 patched with PLUMED version 2.9.355. Simulations were carried out using a single subunit of KCNQ1 in complex with CaM. The initial hybrid structure was constructed by combining the VSD and pore domains (residues 1–236) preceding the bending point (“RQKH”, residues 251–254), including the bound PIP₂, from the straight conformation, with the remaining regions taken from the bent conformation. Protonation states of residues were assigned using PDB2PQR v1.9.056. The system was embedded in a POPC bilayer and solvated with TIP3P water and 0.15 M NaCl using the CHARMM-GUI server57. PIP2 structures were optimized at the HF/6-31 G** level using the Gaussian 16 package58, followed by single-point energy calculations at the B3LYP/cc-pVTZ level to obtain the electrostatic potential (ESP). Restrained ESP (RESP) charges were derived from these calculations for use in force field parameterization. Protein parameters were assigned using the AMBER ff14SB force field59, and ligand parameters were generated using the general AMBER force field 2 (GAFF2)60. The detailed setup of the metadynamics simulations is summarized in Table S2. The system was processed using a two-stage energy minimization. In the first stage, the system was minimized for 4000 steps, beginning with 1000 steps of steepest-descent minimization followed by conjugate-gradient minimization, with a 12 Å nonbonded cutoff and positional restrains of 10.0 kcal mol−1 Å−2 applied to all non-hydrogen solute atoms. This step allowed relaxation of lipids and solvents while maintaining the initial structure of the complex. In the second stage, minimization was performed using steepest-descent minimization for the first 10,000 steps and conjugate-gradient minimization until the energy gradient reached 0.01 kcal mol−1 Å−1, with positional restrains of 10.0 kcal mol−1 Å−2 applied to all heavy atoms in residues 1–236 and the PIP2 ligand. After minimization, the system was heated from 0 to 298 K over 100 ps using a 1 fs timestep in the NVT ensemble with Langevin thermostat for temperature control and a collision frequency of 4 ps−1. SHAKE61 was used to constrain covalent bonds involving hydrogen atoms. A 12 Å cutoff and positional restrains of 5.0 kcal mol−1 Å−2 were applied to all atoms in residues 1–236 and the PIP2 ligand. Random seeds were generated automatically (ig = −1) to ensure distinct initial velocities for independent replicates. After heating, the system was equilibrated for another 1 ns using a 2 fs timestep under isothermal-isobaric (NPT) ensemble with restrains of 5.0 kcal mol−1 Å−2 to maintain the structure taken from the straight conformation. Temperature was maintained at 298 K. Pressure was controlled at 1 atm using an isotropic Berendsen barostat62. Well-tempered metadynamics63 simulations were carried out for a total simulation time of 100 ns. The collective variable (CV) was defined as the root-mean-square deviation (RMSD) of KCNQ1 with respect to a reference straight conformation, using optimal rotational and translational alignment. Bias potentials were deposited every 100 steps with an initial Gaussian height of 1 kJ/mol and a sigma of 0.4 nm. A bias factor of 10 was employed at a temperature of 298 K. All other parameters for the simulations are the same as the previous equilibration part. Three independent metadynamics simulations were performed with distinct initial velocity seeds. Targeted molecular dynamics (TMD) simulations We performed targeted molecular dynamics (TMD) simulations to model the conformational transition of the KCNQ1 channel from the IO state to the AO state. The IO-state system was constructed using the tetrameric KCNQ1-CaM structure (4:4 stoichiometry), embedded into a POPC bilayer using the CHARMM-GUI web interface, similar with the metadynamics simulations setup. Protonation states of residues were assigned using PDB2PQR v1.9.0 prior to submission to CHARMM-GUI. The membrane-protein assembly was solvated with TIP3P water and neutralized with 0.15 M KCl. The PIP2 parameters were derived by optimizing the geometry at the HF/6-31 G** level and computing the electrostatic potential at the B3LYP/cc-pVTZ level, followed by restrained ESP (RESP) fitting. Protein atoms were parameterized using the AMBER ff14SB force field, and ligand parameters were generated using GAFF2. Details of the TMD simulation system setup are given in Table S2. All simulations were performed with AMBER22 patched with PLUMED 2.9.3. Here, TMD was implemented by applying harmonic restraints that gradually reduced the RMSD between the evolving structure and a set of target AO-state reference structures. For this purpose, each of the four protein subunits and each of the four corresponding ligand molecules was assigned to its own RMSD collective variable, defined relative to the pre-aligned AO-state reference for that specific subunit. The simulations used the PLUMED MOVINGRESTRAINT feature, in which the RMSD target was decreased linearly from its initial value to zero over the entire 50 ns trajectory. The force constant was ramped simultaneously from 100 to 300 kcal mol⁻¹ Å⁻², providing a smoothly increasing bias toward the AO conformation. To prevent artificial global rotation or translation from influencing the RMSD-based restraints, a correction was applied every ten simulation steps, aligning each frame to the AO-state reference. As in the metadynamics simulations, the system was prepared using a two-stage minimization before TMD. In the first stage, the system was minimized for 4000 steps, beginning with 1000 steps of steepest descent followed by conjugate-gradient minimization, with a 12 Å nonbonded cutoff and positional restraints of 5.0 kcal mol−1 Å−2 applied to all non-hydrogen solute atoms. This allowed relaxation of lipids and solvents while maintaining the overall geometry of the protein-ligand complex. In the second stage, an unrestrained minimization was performed using the steepest descent for the first 10,000 steps and conjugate-gradient minimization until the energy gradient reached 0.01 kcal mol−1 Å−1. After minimization, the system was heated from 0 to 298 K over 100 ps using a 1 fs timestep in the NVT ensemble, with Langevin thermostat for temperature control and a collision frequency of 4 ps−1. SHAKE was used to constrain covalent bonds involving hydrogens, and a 12 Å nonbonded cutoff was applied. To preserve the structural integrity of the protein and PIP₂ during heating, positional restraints of 2.0 kcal mol−1 Å−2 were applied to all non-hydrogen atoms. Random seeds were generated automatically (ig = −1) to ensure distinct initial velocities for independent replicates. After heating, the system was equilibrated without restraints for another 1 ns using a 2 fs timestep under isothermal-isobaric (NPT) conditions. Temperature was maintained at 298 K. Pressure was controlled at 1 atm using isotropic Berendsen barostat. TMD simulations were performed starting from the equilibrated IO-state structure. Prior to initiation of biasing, the initial and target structures were aligned to remove global rotational and translational differences. Each TMD trajectory was run for 50 ns with the RMSD targets for the protein subunits decreasing from 25 Å to 0 Å and those for the PIP2 decreasing from 3.5 Å to 0 Å. The force constants increased linearly from 100 to 300 kcal mol⁻¹ Å⁻² over the same time interval. Three independent TMD simulations were performed with distinct initial velocity seeds. Mutagenesis for functional studies Human wildtype (WT) KCNQ1, KCNE1, KCNE3, and CaM cDNA were each cloned into the pcDNATM 3.1 (+) vector (Cat. No: V79020, Thermo Fisher Scientific). Site-directed mutations in these cDNA constructs were generated using overlap extension and high-fidelity PCR. Mutant sequences were verified through Sanger DNA sequencing in both directions. cRNA was made by in vitro transcription from the linearized cDNA using the mMESSAGE mMACHINETM T7 Transcription Kit (Cat. No: AM1344, Thermo Fisher Scientific). The quality and quantity of the cRNA was assessed by agarose gel electrophoresis and Nanodrop. The stock and diluted cRNA were stored at −80 °C until further use. To study the interactions between KCNQ1 and CaM residues (Figs. 1d, e and S3–12), we mutated two negatively charged residues (D) in CaM with positively charged residues (K and R). We then made mutations on KCNQ1 residues primarily to charged amino acids (D and E) to augment interaction strength. At the same time, we also made some KCNQ1 mutations to other residues (Q, N, W, etc.) to explore possible CaM-VSD interactions (hydrophobic, steric, etc.) as well as electrostatic interactions. In this study, we found that many mutations in KCNQ1 or CaM altered KCNQ1 current amplitude when expressed in oocytes. To identify the pairs of KCNQ1 and CaM residues that interact to maintain KCNQ1 currents, we use a “black or white” principle and suggest that the interaction occurs only when the current of the double mutation changes from robust to near-zero or vice versa comparing to single mutations. This stringent criterion is necessary since we cannot quantify the strength of interaction of the residues by a gradual change of current amplitude when the residues were mutated. We show our results on mutations of all residues that were screened (Figs. S3–12), but only the mutation pairs that match our criterion are identified as the interaction pairs (Fig. 1d, e). Ion channel expression in Xenopus oocytes Stage Ⅴ or Ⅵ oocytes were procured from Xenopus laevis via laparotomy, in accordance with the protocol approved by the Washington University Animal Studies Committee (protocol #24-0405). Oocytes were subjected to digestion with collagenase (0.5 mg/ml, Sigma-Aldrich, St. Louis) and subsequently injected with channel cRNAs using a Nanoject (Drummond, Broomall). Each oocyte was injected an identical amount of cRNAs corresponding to either WT or mutant KCNQ1. In tests using KCNE1 or CaM co-expression, KCNE1 and CaM cRNAs were co-injected at mass ratios of 4:1 (KCNQ1: KCNE1) and 1:1 (KCNQ1: CaM), respectively. Injected oocytes were incubated in ND96 solution [96 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2, 5 mM Hepes, 2.5 mM CH3COCO2Na, and 1: 100 penicilin-streptomycin (pH 7.6)] at 18 °C for 3–5 days before recording. Two-electrode voltage clamp (TEVC) Microelectrodes were fabricated using thin wall borosilicate glass (B150-117-10) and a micropipette puller (P-1000, Sutter Instrument, Novato, CA). The pipette resistance ranged from 0.5 to 3 MΩ when filled with a 3 M KCl solution and immerged in ND96 solution, which comprises 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2, 5 mM Hepes, and 2.5 mM CH3COCO2Na at pH 7.6. The experiments were recorded in ND96 solutions at room temperature. Whole-cell currents were recorded using a GeneClamp 500B amplifier (Axon Instruments, CA) driven by Patchmaster (HEKA, Holliston, MA). To prevent aliasing, the device applied a low-pass filter to the measured currents at 2 kHz. The testing voltages were from −100 mV to +60 mV (without KCNE1) or +80 mV (with KCNE1 or KCNQ1-F351A) with 10 mV increments and then returned to −40 mV for measuring the tail currents. The holding potential is −80;mV for all electrophysiology recordings. In compound CA1 application experiments, CA1 was initially dissolved in Dimethyl Sulfoxide (DMSO) to make a 10 mM stock solution, and 1 μL of this stock was added to 1000 μL bath solution (ND96) to achieve a final concentration of 10 μM using a manual pipette. The recording chamber was thoroughly rinsed with 70% ethanol and the deionized water following each experiment involving CA1 administrations. All the electrophysiological recordings were repeated with at least two different batches of oocyte. Electrophysiological data analysis Data were analyzed using MATLAB (MathWorks, MA) and Sigmaplot (SPSS) software. The G-V relationship calculation involved normalizing the instantaneous tail currents subsequent to test pulses to the maximum current. The G-V relationship was fitted using a single Boltzmann equation in the form of1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G\left(V\right)={\left(1+\exp \left(\right.-{V}_{s}\left(V-{V}_{1/2}\right)\right)}^{-1}$$\end{document}GV=1+exp−VsV−V1/2−1where V represents the test pulse voltage, V1/2 represents the half-activation voltage, and Vs regulates the steepness of the Boltzmann equation. Vs is related to RT/zF, where R denotes the gas constant, T represents the temperature, z signfies the equivalent valence, and F indicates the Faraday constant. The current amplitude comparison was determined using the steady-state current amplitude at the end of the 4 s test pulse. Statistical analysis Data were presented as mean ± standard error of mean (SEM), with n specifying the number of independent oocytes. Statistical analysis was performed in SigmaPlot and GraphPad. Statistical differences were considered significant whenever p < 0.05. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Reporting Summary Transparent Peer Review file Source data Source Data Source Data Source Data Source Data