Human EAG channels are directly modulated by PIP2 as revealed by electrophysiological and optical interference investigations

Voltage-gated ether à go-go (EAG) K+ channels are expressed in various types of cancer cells and also in the central nervous system. Aberrant overactivation of human EAG1 (hEAG1) channels is associated with cancer and neuronal disorders such as Zimmermann-Laband and Temple-Baraitser syndromes. Although hEAG1 channels are recognized as potential therapeutic targets, regulation of their functional properties is only poorly understood. Here, we show that the membrane lipid phosphatidylinositol 4,5-bisphosphate (PIP2) is a potent inhibitory gating modifier of hEAG1 channels. PIP2 inhibits the channel activity by directly binding to a short N-terminal segment of the channel important for Ca2+/calmodulin (CaM) binding as evidenced by bio-layer interferometry measurements. Conversely, depletion of endogenous PIP2 either by serotonin-induced phospholipase C (PLC) activation or by a rapamycin-induced translocation system enhances the channel activity at physiological membrane potentials, suggesting that PIP2 exerts a tonic inhibitory influence. Our study, combining electrophysiological and direct binding assays, demonstrates that hEAG1 channels are subject to potent inhibitory modulation by multiple phospholipids and suggests that manipulations of the PIP2 signaling pathway may represent a strategy to treat hEAG1 channel-associated diseases.

Scientific RepoRts | 6:23417 | DOI: 10.1038/srep23417 as a multifunctional molecule, PIP 2 plays pivotal roles in normal and pathological cellular functions [18][19][20] . For instance, PIP 2 is implicated in cell proliferation and neurological diseases [21][22][23] . Consequently, the lipid kinases including phosphatidylinositol 4-kinases (PI4Ks) and phosphatidylinositol-4-phosphate 5-kinases (PIP5KIs) responsible for PIP 2 synthesis are considered to be potential therapeutic targets for various disorders involving altered neuronal excitability such as chronic pain 19,22,[24][25][26] . In addition, the degradation of PIP 2 is subject to dynamic regulation by many important neuronal transmitters via G protein-coupled receptor (GPCR)-induced phospholipase C (PLC) activation, which in turn affects many important downstream targets including a wide array of ion channels 19,27,28 . The mechanisms of ion channel regulation by PIP 2 appear diverse. A crystal structure of an inward-rectifier K + channel with PIP 2 bound shows a direct interaction between the negatively charged head group of the lipid and a cluster of positively charged residues of the channel 29 . Besides the direct interaction mechanism, PIP 2 also exerts its modulatory influences on ion channels indirectly by binding to numerous channel-associated inositol-binding proteins [30][31][32] . Moreover, modulation of ion channel function by PIP 2 may involve alteration of the physical and chemical properties of the plasma membrane, which could then change the channel structure and function 33 . For instance, membrane lipid rafts enriched in PIP 2 34 may influence ion channels contained therein 35,36 . Despite the existence of these direct and indirect mechanisms, many studies investigating modulation of ion channels by PIP 2 relied solely on electrophysiological measurements and mutagenesis, and failed to clarify the exact mechanism of the PIP 2 action.
The potential importance of both hEAG1 channels and PIP 2 in neurological and developmental disorders 7-9 prompted us to investigate whether hEAG1 channels are regulated by PIP 2 . In this study, we show that gating of the hEAG1 channel is potently inhibited by exogenous PIP 2 and that physiological depletion of endogenous PIP 2 in the plasma membrane enhances the channel activity. Moreover, the bio-layer interferometry (BLI) assay demonstrates that PIP 2 directly binds to the hEAG1 channel complex. Removal of a short domain located in the N terminus of the channel abolished its interaction with PIP 2 as well as the lipid-induced ionic current inhibition. Our findings show that the hEAG1 channel is directly regulated by PIP 2 and that this regulation may contribute to normal human physiology and pathology.

Results
Inhibition of hEAG1 channels by PIP 2 . The effect of PIP 2 on heterologously expressed hEAG1 channels was studied by direct application of exogenous PIP 2 to the intracellular side of excised patches from Chinese hamster ovary (CHO) cells transiently expressing hEAG1 channels. Acute application of 3 μM brain-derived PIP 2 with 18 to 20 carbon tail groups 31 to the intracellular side caused a rapid and potent current inhibition (Fig. 1A,B). Because currents through hEAG1 channels frequently underwent progressive rundown after patch excision, our analysis was typically limited to the results obtained within 5-8 min of patch excision. We also tested the effect of PIP 2 on the whole-cell current by perfusing PIP 2 into the HEK293T cells stably expressing hEAG1 channels using a fast intracellular perfusion system and observed a similar inhibitory effect (Fig. 1C,D). The development of the inhibitory effect was slower with lower concentrations of PIP 2 (Fig. 1E) and the results obtained with 200-s PIP 2 application epochs showed that the fractional decrease in current size was also concentration dependent with an apparent half-maximal inhibitory concentration (IC 50 ) of 0.35 ± 0.01 μM and the Hill coefficient of 0.92 ± 0.32 (Fig. 1F). By contrast, dioctanoyl-PIP 2 (diC8-PIP 2 ) and its analogs with different numbers of negative charges (3 μM), which have 8-carbon tails and higher water solubility 27 than brain-derived PIP 2 , showed less overall potency in inhibiting hEAG1 channels ( Fig. 1F-H). The IC 50 value for diC8-PIP 2 was 18.6 ± 0.6 μM, suggesting that the inhibitory potency of PIP 2 depends on the length of the fatty acid chains.
In addition to PIP 2 , other types of phosphatidylinositol (PIs) are also involved in cellular signaling and they may show specificities in ion channel regulation 37,38 . We therefore tested whether these PI lipids inhibited the hEAG1 channel. When applied at 3 μM, PI, PI(4)P, PI(4,5)P 2 , and PI(3,5)P 2 differed noticeably in their abilities to alter the hEAG1 current, from no effect by PI to the strongest inhibition by PIP 2 and PI(3,5)P 2 (Fig. 1I), showing that the number of negative charges of the head group is also critical for the channel inhibition. PIP 2 (3 μM) virtually annihilated currents through hEAG1 at 40 mV at which the channel was maximally activated (Fig. 1A). However, even with PIP 2 present (2 μM), depolarization to an extremely positive voltage (e.g., 200 mV) did elicit a robust outward current and the peak tail current size following the depolarization was similar to that after depolarization to 40 mV without PIP 2 ( Fig. 2A,B). This observation suggests that hEAG1 remained functional even in the presence of 3 μM PIP 2 ( Fig. 1A) but the voltage dependence of activation was shifted markedly to the positive direction. Consistent with this possibility, the conductance-voltage (GV) curve obtained with a lower concentration of PIP 2 (0.1 μM) showed a clear positive shift; the half-activation voltage (V 0.5 ) changed from −16.1 ± 4.0 to 40.8 ± 6.1 mV (Fig. 2E,F). Additionally, the GV curve became less steep; the slope factor (k) increased from 14.7 ± 1.2 to 44.2 ± 9.3 mV (Fig. 2C-F). hEAG1 channels interact directly with PIP 2 . To test whether PIP 2 directly interacts with the hEAG1 channel, we used bio-layer interferometry (BLI) to measure the kinetics of binding of PIP 2 to the purified hEAG1 channel complex. BLI is a novel methodology that has emerged recently for detecting protein-protein interactions as well as those between proteins and small molecules. Binding events between a protein and its ligands can be measured by changes in optical interference (Fig. 3A). The quality and the specificity of purified biotinylated hEAG1 channel protein was confirmed by Western blot (Fig. 3B). Addition of PIP 2 to the protein anchored to the BLI sensor tip rapidly increased the optical interference signal (Fig. 3C). The kinetics of the BLI signal resembled those observed in the patch-clamp measurements in the excised patches (Fig. 1B). The kinetics of the increase in the BLI signal by PIP 2 was concentration dependent, becoming faster with greater concentrations, and the BLI signal increase was partially reversible after 3 min of wash (Fig. 3C,D). Using a 200-s application epoch at each concentration as utilized in the electrophysiological experiments, the cumulative concentration dependence Scientific RepoRts | 6:23417 | DOI: 10.1038/srep23417 suggested a dissociation constant (KD) value of 0.35 ± 0.04 μM (Fig. 3E), similar to the IC 50 value obtained from the electrophysiological measurements (Fig. 1F).
The close correspondence between the result from the BLI and electrophysiological measurements was also observed in the ligand selectivity. PI(3,5)P 2 (3 μM), which effectively inhibited the hEAG1 channel electrophysiologically (Fig. 1I), elicited robust BLI signals (Fig. 3F,G). In contrast, diC8-PIP 2 (3 μM), which failed to inhibit the hEAG1 channel in our electrophysiological experiments (Fig. 1G,H), also failed to elicit any noticeable BLI signal (Fig. 3F,G). Representative whole-cell hEAG1 currents (C) and normalized hEAG1 channel peak current (D) elicited by pulses to 40 mV from − 80 mV using an automated patch-clamp instrument (Patchliner, Nanion) before and after intracellular perfusion of PIP 2 (3 μM). **P < 0.01 compared to control levels before lipid applications. Depletion of endogenous PIP 2 increases hEAG1-mediated K + currents. Some inward-rectifier K + channels require PIP 2 for function, indicating that these channels are subject to basal tonic modulation by PIP 2 20 .
We investigated whether a similar tonic modulation of hEAG1 channels exists by selectively depleting PIP 2 in HEK293T cells stably expressing hEAG1 channels. We used the plasma membrane-targeted, rapamycin-induced translocation system based on inducible dimerization of PM-FRB-CFP and 5-phosphatase-FKBP-mRFP to acutely deplete PIP 2 by a specific dephosphorylation of PIP 2 at the 5′ position 39,40 . PIP 2 depletion in the plasma membrane was monitored by PLCδ 1PH-GFP, which binds to PIP 2 at the plasma membrane 40 . When hEAG1-expressing cells were transfected with PLCδ 1PH-GFP alone, the GFP fluorescence was found primarily near the plasma membrane (  The results obtained in the presence of PIP 2 (2 μM) were normalized to the maximal conductance inferred from the tail currents before PIP 2 application in each patch. The normalized conductance values were 0.94 ± 0.02, 0.51 ± 0.04 and 0.81 ± 0.04, respectively (n = 6). (C) Illustrative current traces elicited by pulses from − 80 to 50 mV in 10 mV increments before and after application of 0.1 μM PIP 2 . (D) Peak current-voltage curves from the results shown in (C). (E) Normalized conductance-voltage (GV) curves constructed from hEAG1 channel tail currents before and 3-5 min after application of 0.1 μM PIP 2 . The curves were fitted with a Boltzmann function from the results obtained using the same protocol as in (C). The fit parameters, half-activation voltage (V 0.5 ) and slope factor (k), of voltagedependent activation of hEAG1 channels before and after PIP 2 application (0.1 μM) are shown in (F). **P < 0.01 compared to the control before PIP 2 application and † † P < 0.01 compared to after PIP 2 application at 40 mV. decrease in the level of plasma membrane PIP 2 . Rapamycin also increased the whole-cell hEAG1 current (Fig. 4B  bottom). The fractional increase in current was especially noticeable at more negative voltages (Fig. 4C), consistent with the idea that PIP 2 depletion shifted the current-voltage (IV) relationship to the negative direction.
One of the physiological ways by which PIP 2 is hydrolyzed is by the action of phospholipase C (PLC) following activation of GPCRs such as serotonin receptors 41 , which are in turn implicated in tumorigenesis 42 and also in neuronal signaling. We thus examined if serotonin increases the hEAG1 channel activity by depleting PIP 2 . We transfected the hEAG1-expressing HEK293T cells with human 5-hydroxytryptamine (serotonin) receptor 2A (HTR 2A ) 43 and PLCδ 1PH-GFP. Hydrolysis of PIP 2 by PLC leads to IP 3 release and increases the intracellular Ca 2+ concentration ([Ca 2+ ] i ), which may confound any effect of PIP 2 depletion 16 . To avoid this potential complication, the "fast" Ca 2+ chelator 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA, 10 mM) was used in the whole-cell recording solution to rapidly control [Ca 2+ ] i . Serotonin (100 μM) had no detectable effect on the GFP signal or the current size in the control hEAG1-expressing cells (Fig. 4D,F). However, when the channels  were co-expressed with the serotonin receptor HTR 2A , serotonin dimmed the GFP signal reporting the PIP 2 level (Fig. 4E top and middle) and noticeably increased the whole-cell current (Fig. 4E bottom, F). The fractional increase in current was greater at more negative voltages (Fig. 4F), suggesting that the IV curve shifted to the negative direction. The results collectively suggest that endogenous PIP 2 exerts a detectable tonic inhibitory influence on hEAG1 channels in intact cells.

Molecular determinants of hEAG1 channel inhibition by PIP 2 . PIP 2 -binding sites typically consist
of positively charged amino acid residues 44 . To identify the molecular locus required for the PIP 2 -dependent regulation of hEAG1, we first investigated whether the positively charged residues in S4 of hEAG1 are involved in PIP 2 -induced channel inhibition in part because PIP 2 alters the GV characteristics of the hEAG1 channel (Fig. 2E). The hEAG1 S4 segment, like those in other voltage-gated K + channels 45,46 , contains multiple positively charged residues that move within the electric field (Fig. 5A). If these mobile charges interact with PIP 2 applied to the intracellular side, some voltage/S4 state dependence of the effect of PIP 2 on hEAG1 may be expected as observed in KCNQ channels 47 . PIP 2 quickly inhibited the hEAG1 channel when the membrane potential was held at 40 mV (Fig. 5B,D). At this voltage, assuming that hEAG1 and Kv1.2/2.1 46 activate in a similar fashion, the voltage sensors are expected to be fully activated, placing hEAG1 K327 through R336 on the extracellular side and R339 interacting with negatively charged residues in S2 and S3. A similar inhibition was observed when the membrane potential was held at − 80 mV at which R330 through K340 electrically should face the intracellular side (Fig. 5C,D). Furthermore, neutralization of the positively charged residues at R333, R336, R339, and/or K340 had no appreciable effect on the PIP 2 -induced channel inhibition (Fig. 5E).
Amino-acid sequences of hEAG1, hELK1, and hERG1 show considerable similarity (~40%). A recent study has shown that PIP 2 inhibits hELK1 by interacting with some of the positively charged residues in the S4-S5 linker, S6, and EAG domain 48 . However, these residues are not well conserved in hEAG1 channels. By contrast, PIP 2 increases currents through hERG1 and a short region rich in positively charged residues ( 883 RQRKRKLSFRRR 894 ) in the intracellular C terminus of hERG1 has been implicated 49 . The distal C terminus of hEAG1 also contains a similar segment with multiple positively charged residues (C3 806 KRKSWARFK 814 , Fig. 6A). Deletion of this segment did not alter the inhibitory effect of exogenous PIP 2 (∆ C3, Fig. 6B). In many ion channels such as KCNQ 50 , TRPC 51 , SK2 52 , and CNG channels 53 , the putative PIP 2 interaction sites overlap with CaM binding sites. Therefore, we hypothesized that the CaM interaction domains in the hEAG1 channel 15-17 may be involved in the inhibitory effect of PIP 2 . Up to three potential CaM binding segments in hEAG1 have been reported: CaM-N in the N terminus and CaM-C1 and CaM-C2 in the C terminus (Fig. 6A) [15][16][17] . We deleted these CaM binding sites individually (∆ CaM-N, ∆ CaM-C1 and ∆ CaM-C2). Currents through hEAG1 ∆ CaM-C1 and hEAG1 ∆ CaM-C2 maintained wild-type-like sensitivity to PIP 2 ; however, those from hEAG1 ∆ CaM-N were virtually unaltered by PIP 2 , up to 3 μM (Fig. 6B,C). Moreover, no BLI signals were observed from the purified hEAG1 ∆ CaM-N in response to PIP 2 (3 μM; Fig. 6D,E). In addition to hEAG1, we also found that PIP 2 significantly inhibited hEAG2 (KCNH5, Kv10.2) channels, which have virtually the same sequence near the CaM-N area as hEAG1 (Supplementary Fig. S1).
Functional independence between PIP 2 and Ca 2+ /CaM in the regulation of hEAG1 channels. The observation that the potential binding sites for PIP 2 and Ca 2+ /CaM in the N terminus overlap suggests that these two inhibitory factors may interfere with each other. Furthermore, both Ca 2+ /CaM and PIP 2 may be present together under physiological conditions. The exposure of inside-out membrane patches to 200 nM CaM in 1 μM [Ca 2+ ] i resulted in a complete inhibition of hEAG1 currents and this effect, unlike that by PIP 2 , was rapidly and fully reversible (Fig. 7A). The second exposure to Ca 2+ /CaM at the same concentrations together with PIP 2 (3 μM) inhibited the current as rapidly as the first exposure. However, no recovery was observed after wash out of Ca 2+ /CaM (Fig. 7A), an indication of the long-lasting action of PIP 2 (see Fig. 1). To assess whether the presence of PIP 2 alters its sensitivity to Ca 2+ /CaM, we utilized the following protocol. We first verified that Ca 2+ /CaM (30 nM CaM in 1 μM [Ca 2+ ] i ) inhibited the current in a rapidly reversible fashion (Fig. 7B, first "CaM"). Following wash out of Ca 2+ /CaM, application of PIP 2 (0.5 μM) decreased the current by 50% (Fig. 7B, "PIP 2 "). Subsequent application of Ca 2+ /CaM results in a near 100% inhibition (Fig. 7B, second "CaM"), suggesting that Ca 2+ /CaM and PIP 2 may have additive effects on hEAG1 channel inhibition.

Discussion
Increasing evidence suggests that hEAG1 K + channels are closely linked with tumorigenesis and developmental diseases [7][8][9] . However, the regulation of the hEAG1 channel function is only beginning to be elucidated and only a few physiologically relevant regulators have been identified 12,14,16,54 . Here we have demonstrated that PIP 2 profoundly shifts the voltage dependence of activation of the hEAG1 channel to the positive direction and acts as a potent inhibitory modulator of the channel. Further, we have identified that a region near the channel's N-terminal CaM interaction site is essential for both the electrophysiological effects and binding of PIP 2 .
Membrane lipids may be an important class of modulators of EAG channels. A recent study has shown that mouse neurons have two populations of EAG channels, one of which is sensitive to the alteration of lipid raft integrity such as cholesterol depletion 54 . In our study, temporally-controlled depletion of PIP 2 by rapamycin-induced translocation of 5-phosphatase 55 in whole cells noticeably increased hEAG1 currents especially at negative voltages. Thus, PIP 2 -mediated inhibition of hEAG1 channels is well suited to influence action potential firing in neurons. Physiological relevance of the tonic inhibitory influence of PIP 2 as an endogenous modulator of hEAG1 is further suggested by the finding that activation of serotonin HTR 2A receptors, known to activate PLC 43 to promote hydrolysis of PIP 2 , also increases hEAG1 currents at negative voltages. However, the magnitude of hEAG1 current increase in vivo mediated by GPCR activation is difficult to predict. Under such conditions, GPCR activation leads to generation of multiple signaling molecules capable of affecting hEAG1 channels in distinct ways. For example, PIP 2 depletion by PLC produces IP 3 , which in turn can promote Ca 2+ release from the endoplasmic reticulum. Ca 2+ /CaM could then inhibit hEAG1 channels, dampening the direct stimulatory effect of PIP 2 depletion on the channels described here. In addition, activation of many GPCRs including serotonin receptors can stimulate phospholipases A2 (PLA2) thereby inducing the release of arachidonic acid (AA), a lipid messenger which has been reported to activate hEAG channels in human melanoma cells 14 . Thus, depending on the comparative potencies and kinetics of these direct and indirect effects, complex regulation of hEAG1 channels by PIP 2 under physiological conditions is possible.
Numerous ion channels have been reported to be PIP 2 sensitive 20 . However, elucidation of the exact mode of action of PIP 2 , for example direct binding vs. indirect effects, has been hampered by the lack of appropriate PIP 2 binding assays with a good time resolution that detect physiologically relevant binding events. Here, we used bio-layer interferometry (BLI), a novel assay capable for measuring the kinetics of binding between proteins and their ligands 56 , to detect binding of PIP 2 to the isolated hEAG1 protein. The BLI measurements showed the existence of two kinetic components in association and dissociation (Fig. 3C). The fast BLI association and slow dissociation components especially well correspond to the electrophysiological results in their kinetics and concentration dependence. The slow BLI association component and the fast dissociation signals, which became more prominent at high concentrations of PIP 2 , may be electrophysiologically silent. Furthermore, the deletion of the N-terminal Ca 2+ /CaM binding area in hEAG1, which eliminates the electrophysiological effect of PIP 2 , also abolishes the BLI signal. A Hill coefficient (~1) of the concentration dependence curve of the current inhibition by PIP 2 hints that binding of one PIP 2 is capable of exerting a full inhibitory effect; this idea needs further investigation. The close correspondence between the electrophysiological and BLI results demonstrates that BLI, which requires no labeling and has a time resolution comparable to that of the patch-clamp method, could be a very promising method for studying the pharmacology of ion channels, thus complementing the established electrophysiological assays.
Our electrophysiological results show that PIP 2 functions as a potent inhibitory gating modifier of the hEAG1 channel by shifting the overall voltage dependence of activation to the positive direction. We did not measure the exact shift at high concentrations of PIP 2 ; however, the observation that a functionally saturating concentration of PIP 2 (3 μM) essentially eliminates the current at 40 mV where the normalized conductance without PIP 2 is near unity suggests that PIP 2 may induce a > 100 mV shift. Such a large shift in the voltage dependence of activation may partially explain the profound current inhibition. How binding of PIP 2 to the N terminus of hEAG1 leads to this marked shift in the overall voltage dependence can only be speculated. Studies show that the N-terminal region of the members of the KCNH channel family (K V 10, K V 11 and K V 12) including hEAG1 (K V 10.1) plays pivotal roles in channel gating 57 . An area termed the EAG domain in the N terminus directly interacts with an area in the C terminus called the cyclic nucleotide-binding homology domain (CNBHD), which in turn regulates behavior of the ion conduction gate in S6 via the C-linker domain 13 . Downstream of the EAG domain close to S1, the N terminus contains one of the CaM binding sites, CaM-N, which is involved in Ca 2+ /CaM-induced inhibition of hEAG1 17 . Our results show that the amino-acid residues involved in PIP 2 binding partially overlap with the CaM-N site, suggesting that PIP 2 and Ca 2+ /CaM may interact. Such an interaction is in fact seen in many other ion channels such as TRPC6 in which PIP 2 binding disrupts the association of CaM with the channel protein 51 . A structural study shows that PIP 2 activates type 2 small-conductance Ca 2+ -dependent K + channels (SK2) by binding to the positively charged residues located at the interface of the SK2-CaM complex and that phosphorylation of CaM reduces the affinity of PIP 2 for the channel complex 52 . However, in hEAG1, the long-lasting inhibitory effect of PIP 2 is unaltered by Ca 2+ /CaM while we cannot formally exclude the possibility that binding of PIP 2 somehow renders the interaction of the CaM-N domain and CaM essentially irreversible. Thus, PIP 2 and Ca 2+ /CaM with vastly different kinetics are capable of regulating the channel in each other's presence, expanding the modulatory versatility of the hEAG1 channel. The N terminus of the EAG channel has been suggested to interact directly with CNBHD 13 . Perhaps, this interaction is modified by PIP 2 to alter the voltage dependence of activation (Fig. 8).
In summary, we have demonstrated that PIP 2 is a potent endogenous gating modulator of the hEAG1 channel that exerts its tonic inhibitory influence by direct binding to the N terminus of the channel protein.
Gain-of-function mutations in the gene coding for hEAG1 have been associated with developmental and neurological disorders such as Zimmermann-Laband and Temple-Baraitser syndromes 7,8 . Thus, the tonic inhibition of the hEAG1 channel by PIP 2 may be beneficial to the aforementioned disorders and dysregulation of the PIP 2 signaling pathway may contribute to disease through hEAG1 channels.

Materials and Methods
Channel expression. hEAG1 (NM_002238.3) and hEAG2 (NM_139318.4) in the expression vector pcDNA3.1 were transiently expressed in CHO cells using X-tremeGENE 9 (Roche Diagnostics). The mutant channels were constructed by using the QuickChange Site-Directed Mutagenesis kit (Agilent) following the manufacturer's instruction. Purification and biotinylation of FLAG-tagged channel proteins. Wild-type and mutant channels with a FLAG tag at the distal C terminus were prepared in the pCDH lentiviral expression plasmid (System Bioscience). To generate cells stably expressing hEAG1 or the mutant channels, ~2.45 μg of the DNA construct was transfected into 5 × 10 5 HEK293T cells using the calcium phosphate transfection method in a 35 mm diameter tissue-culture dish. The cells were cultured in DMEM (Invitrogen) supplemented with 10% FBS (Invitrogen) and 5 μg/ml puromycin (Invitrogen). After 2 days of culture, cells were trypsinized and seeded onto 10 cm diameter tissue-culture plates. The cells were passaged every 2-3 days and supplied with fresh puromycin 5 μg/ml contained selection medium for 12-14 days. Puromycin-resistant cells were trypsinized and seeded onto 96-well tissue-culture dishes at a density of ~1 cell per 2 wells. Monoclonal puromycin-resistant cells were cultured and expanded to prepare the seed stocks for future use and concomitantly to electrophysiologically examine the channel expression. To purify hEAG1 wild-type and the mutant channel proteins, cells stably expressing the channel of interest were harvested from DMEM containing 2 μg/ml puromycin, and were solubilized in a lysis buffer (10 mM HEPES, 1.5 mM MgCl 2 , 10 mM NaCl, 1% NP-40, pH 7.0) for 30 min followed by sedimentation by centrifugation (15,000 × g, 30 min). The supernatant was collected and mixed with anti-FLAG beads (Sigma) overnight. The beads-captured protein was collected based on the manufacturer's instruction. All steps were performed at 4 °C or on ice and all buffers contained Complete Protease Inhibitor (Roche Diagnostics). The purified proteins were quantified using BCA kit (Biyuntian, China) and labeled with biotin (Pierce). The biotinylated membrane proteins were stored in the buffer (PBS with 0.02% Tween-20 and 0.1% BSA, pH 7.4) for the bio-layer interferometry (BLI) assay.
Western blot. The purified hEAG1 protein samples (2 μg) were separated by 8% SDS-polyacrylamide gel and then transferred to nitrocellulose membranes. After blocking with 5% nonfat milk in Tris-buffered saline containing 0.1% Tween-20 (TBST) for 1 h at room temperature, the transferred membranes were incubated overnight at 4 °C with M2 monoclonal anti-FLAG antibody (1: 1000; Sigma), followed by a goat anti-mouse HRP-conjugated secondary antibody (1: 5000; Santa Cruz Biotechnology). The ECL chemiluminescence was visualized and captured by GE Amersham Imager 600 Imaging System.

Bio-layer interferometry (BLI).
Binding of phospholipids to the channel protein was measured and analyzed on an Octet Red96 instrument (FortéBio) at room temperature. The buffer-equilibrated streptavidin biosensors were loaded with 100 μg/ml protein. A duplicate set of sensors was incubated in the buffer without protein for a background binding control. The assay was performed in black 96-well plates (Thermo Fisher Scientific) with the total working volume of 0.21 ml per well. The signal was analyzed using a double reference subtraction protocol to subtract the non-specific binding, background, and signal drift caused by sensor variability. The binding event between the protein and the lipids was quantified by the shift of interference pattern of the light 58 .
Electrophysiology and analysis. Macroscopic currents were recorded in the inside-out patch-clamp configuration from CHO cells transiently expressing hEAG1 channels by an EPC-10 amplifier (HEKA Electronics) controlled by PatchMaster software (HEKA Electronics). Patch pipettes pulled from borosilicate glass (Warner) had a typical initial resistance of 1 to 3 MΩ when filled with the pipette solution described below and ~60% of the initial input resistance was electronically compensated in the macroscopic current measurements. Leak and capacitative currents were subtracted with a P/6 protocol. The external solution contained (in mM): 140 KCl, 2 MgCl 2 , 10 HEPES, 15 glucose, pH 7.2 with N-methyl-D-glucamine (NMG). The internal solution without Ca 2+ contained (in mM): 140 KCl, 10 EGTA, 10 NaCl, 1 MgCl 2 and 10 HEPES, pH 7.2 with NMG. Internal solutions with higher concentrations of free Ca 2+ were prepared using various amounts of EGTA as described previously 59 . The current was elicited to different voltages from the holding potential of − 80 mV every 2 s. The lipids were applied to the intracellular side when the current reached a stable level after patch excision. Half-activation potential (V 0.5 ) was obtained by fitting the maximal peak tail current versus membrane voltage with a Boltzmann function, where V m is the membrane potential, and k is the slope factor. Whole-cell currents were recorded from HEK293T cells stably expressing wild-type or mutant hEAG1 channels using an EPC10 amplifier and in some experiments using an automated planar patch-clamp instrument (Nanion) as described previously 60 . The external solution contained (in mM): 134 NaCl, 6 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 glucose, 10 HEPES, pH 7.4 (with NMG). The intracellular solution contained (in mM): 110 K aspartate, 30 KCl, 10 NaCl, 2 MgCl 2 , 10 HEPES, pH 7.2 (with NMG).