Phosphatidylinositol 4,5-bisphosphate (PIP2) regulates Shaker K+ channels and voltage-gated Ca2+ channels in a bimodal fashion by inhibiting voltage activation while stabilizing open channels. Bimodal regulation is conserved in hyperpolarization-activated cyclic nucleotide–gated (HCN) channels, but voltage activation is enhanced while the open channel state is destabilized. The proposed sites of PIP2 regulation in these channels include the voltage-sensor domain (VSD) and conserved regions of the proximal cytoplasmic C terminus. Relatively little is known about PIP2 regulation of Ether-á-go-go (EAG) channels, a metazoan-specific family of K+ channels that includes three gene subfamilies, Eag (Kv10), Erg (Kv11), and Elk (Kv12). We examined PIP2 regulation of the Elk subfamily potassium channel human Elk1 to determine whether bimodal regulation is conserved within the EAG K+ channel family. Open-state stabilization by PIP2 has been observed in human Erg1, but the proposed site of regulation in the distal C terminus is not conserved among EAG family channels. We show that PIP2 strongly inhibits voltage activation of Elk1 but also stabilizes the open state. This stabilization produces slow deactivation and a mode shift in voltage gating after activation. However, removal of PIP2 has the net effect of enhancing Elk1 activation. R347 in the linker between the VSD and pore (S4–S5 linker) and R479 near the S6 activation gate are required for PIP2 to inhibit voltage activation. The ability of PIP2 to stabilize the open state also requires these residues, suggesting an overlap in sites central to the opposing effects of PIP2 on channel gating. Open-state stabilization in Elk1 requires the N-terminal eag domain (PAS domain + Cap), and PIP2-dependent stabilization is enhanced by a conserved basic residue (K5) in the Cap. Our data shows that PIP2 can bimodally regulate voltage gating in EAG family channels, as has been proposed for Shaker and HCN channels. PIP2 regulation appears fundamentally different for Elk and KCNQ channels, suggesting that, although both channel types can regulate action potential threshold in neurons, they are not functionally redundant.
Phosphatidylinositol 4,5-bisphosphate (PIP2) is a regulator of a wide variety of ion channels, including evolutionarily diverse members of the voltage-gated cation channel superfamily (Hilgemann et al., 2001; Suh and Hille, 2008; Rodríguez-Menchaca et al., 2012b; Zhou and Logothetis, 2013). PIP2 is found in the inner membrane leaflet where its negatively charged headgroup is ideally positioned to interact with the gating machinery of voltage-gated ion channels. This machinery includes the activation gate of the channel pore and the S4–S5 linker, which couples the voltage-sensor domain (VSD) to that gate (Long et al., 2005a). In KCNQ K+ channels, PIP2 is required to couple voltage-sensor movement to pore opening and interacts with positively charged residues in the S4–S5 linker and immediately downstream of the S6 activation gate (Zaydman et al., 2013). In contrast, the Shaker channel Kv1.2 is dually modulated: PIP2 inhibits voltage activation but also increases maximal currents, suggesting open-state stabilization (Rodriguez-Menchaca et al., 2012a). Interactions between the S4–S5 linker of Kv1.2 and PIP2 inhibit voltage activation by restricting outward movement of the voltage sensor (Rodriguez-Menchaca et al., 2012a), but sites involved in PIP2-dependent enhancement have not been identified. However, some voltage-gated Ca2+ channels show the same dual modulation by PIP2 (Wu et al., 2002; Suh et al., 2010), and a residue at the intracellular face of the domain III S6 has been shown to regulate PIP2-dependent current enhancement (Zhen et al., 2006).
Bimodal modulation of gating by PIP2 has also been observed for the sea urchin hyperpolarization-activated CNG (HCN) channel SpIH (Flynn and Zagotta, 2011), which belongs to a separate superfamily of voltage-gated cation channels that share a cytoplasmic cyclic nucleotide–binding domain (CNBD). This CNBD superfamily includes HCN channels, CNG channels, and Ether-á-go-go (EAG) family voltage-gated K+ channels in animals (Yu and Catterall, 2004; Jegla et al., 2009), as well as K+ channels in plants (Schachtman et al., 1992; Sentenac et al., 1992), ciliate protozoans (Jegla and Salkoff, 1994, 1995), and prokaryotes (Brams et al., 2014). In SpIH, PIP2 depolarizes the voltage activation range as in Shaker (Flynn and Zagotta, 2011), but this enhances rather than inhibits activation in the physiological range because SpIH is opened by hyperpolarization. The site of action was localized to the transmembrane channel core, but specific residues were not identified. PIP2 also reduces maximal SpIH current, and this opposing inhibitory effect depends on basic residues in the C-linker, which connects the cytoplasmic CNBD to the activation gate (Flynn and Zagotta, 2011). The C-linker has been shown to play a critical role in the gating of HCN and CNG channels (Decher et al., 2004; Craven and Zagotta, 2006). CNG channels are also inhibited by phosphatidylinositol 3,4,5-bisphosphate (PIP3; Womack et al., 2000; Zhainazarov et al., 2004; Bright et al., 2007), although identified sites of action appear to be in the proximal N terminus and distal C terminus outside the conserved channel core domains (Brady et al., 2006; Dai et al., 2013).
Much less is known about PIP2-dependent modulation of EAG family K+ channels. These channels can be identified by a unique subunit structure and separate into three gene subfamilies, Eag (Kv10), Erg (Kv11), and Elk (Kv12), based on sequence conservation (Warmke and Ganetzky, 1994; Ganetzky et al., 1999). The EAG family is ancient but metazoan specific. The EAG subunit structure appears to have arisen in basal metazoans, and the Eag, Erg, and Elk subfamilies first appeared and functionally differentiated in a common ancestor of cnidarians and bilaterians (Martinson et al., 2014; Li et al., 2015b). Virtually all data on PIP2 modulation of the EAG channel family comes from a single gene, the human Erg subfamily channel Erg1 (hErg1, Kv11.1), which underlies the IKr current critical to cardiac action potential repolarization (Sanguinetti et al., 1995; Trudeau et al., 1995; Keating and Sanguinetti, 2001). The application of exogenous PIP2 enhances hErg1 activation (Bian et al., 2001, 2004), and depletion of native PIP2 causes a small reduction in hErg1 currents (Kruse and Hille, 2013). The effect of PIP2 on Eag and Elk subfamily channel gating has not yet been described, and bimodal modulation by PIP2 has not been observed for the EAG family.
The EAG channel family subunit domain structure consists of a typical voltage-gated K+ channel core, a C-linker/cyclic nucleotide–binding homology domain (CNBHD) structurally equivalent to the C-linker/CNBD gating domain of HCN and CNG channels (Brelidze et al., 2012), and a unique N-terminal eag domain consisting of a Per-Arndt-Sim (PAS) domain (Morais Cabral et al., 1998) with an EAG channel–specific Cap (Gustina and Trudeau, 2012). The CNBHD receives its designation as a homology domain because although structurally similar to CNBD, it appears self-liganded in crystal structures (Brelidze et al., 2012, 2013), and EAG family channels do not appear to respond to the application of intracellular cyclic nucleotides (Brelidze et al., 2009, 2012). The eag domain docks on the CNBHD (Gianulis et al., 2013; Haitin et al., 2013), and in hErg1 it stabilizes open channels, producing the characteristic slow deactivation of hErg1 (Morais Cabral et al., 1998; Wang et al., 1998). This time-dependent open-state stabilization causes deactivation to occur in a more hyperpolarized voltage range than activation, a phenomenon that has been referred to as “mode shift” (Alonso-Ron et al., 2008; Villalba-Galea et al., 2008; Tan et al., 2012; Hull et al., 2014; Goodchild et al., 2015). Deletion of the eag domain or Cap in hErg1 reduces open-state stabilization and therefore speeds deactivation and diminishes mode shift (Morais Cabral et al., 1998; Wang et al., 1998; Alonso-Ron et al., 2008; Muskett et al., 2011; Tan et al., 2012; Hull et al., 2014; Goodchild et al., 2015). The N-terminal of the Cap of hErg1 is highly conserved and forms an amphipathic helix by NMR; neutralization of conserved basic residues within the helix speeds deactivation (Muskett et al., 2011). The location of the N-terminal of the Cap is not resolved in EAG family crystals (Morais Cabral et al., 1998; Haitin et al., 2013), but it has been proposed to interact directly with the channel core (Wang et al., 1998) or alternatively with the CNBHD (Haitin et al., 2013). Interestingly, the proposed site of PIP2 interaction with hErg1 is a cluster of positive charges in the C terminus downstream of the CNBHD (Bian et al., 2004) in an area that is poorly conserved even within the Erg channel subfamily, and thus it is not clear how PIP2 might regulate other EAG family channels.
We examined PIP2 regulation of the human Elk family channel Elk1 (HsElk1, Kv12.1) to gain a broader understanding of how PIP2 influences the activity of EAG family channels. We also examined the role of the eag domain in Elk channel gating to see if it plays a functionally equivalent role to what has been observed for hErg1, or inhibits activation and inactivation as has been proposed for the mammalian Eag subfamily channel Eag1 (Kv10.1; Terlau et al., 1997). Elk channels, including Elk1, characteristically have a hyperpolarized voltage-activation range with significant activation at neuronal resting potentials (Trudeau et al., 1999; Zou et al., 2003; Zhang et al., 2010; Li et al., 2015b). Elk2, like the KCNQ-encoded M current, contributes to subthreshold K+ current in hippocampal pyramidal neurons, and genetic or pharmacologic block of either channel causes hyperexcitability and epilepsy (Peters et al., 2005; Zhang et al., 2010). Comparing how Elk and KCNQ channels differ in terms of PIP2 regulation is therefore important for determining the degree to which these channels are functionally redundant or occupy separate physiological niches. We chose Elk1 for this study because it is the only mammalian Elk channel that expresses at high enough levels in heterologous systems to enable excised patch recording for direct PIP2 application.
The results we present here show that the eag domain of Elk1 is functionally equivalent to that of hErg1, and that PIP2 regulates Elk1 in a bimodal fashion, as has been observed for Shaker, HCN, and calcium channels. PIP2-dependent regulation of Elk1 requires basic residues in the S4–S5 linker (R347), proximal to the S6 activation gate (R479), and in the Cap of the eag domain (K5).
MATERIALS AND METHODS
Human Elk1 (Kv12.1, KCNH8) was amplified from whole brain RNA by standard RT-PCR techniques and cloned into the pOX plasmid (Jegla and Salkoff, 1997) for expression in Xenopus laevis oocytes. The WT clone used for these studies encodes a channel identical in amino acid sequence to the channel protein encoded by GenBank accession number NM_144633. Mutations and truncations were incorporated via standard oligo-mediated PCR mutagenesis techniques and sequence verified. Ciona intestinalis VSP (CiVSP) was provided by J. Cui (Washington University, St. Louis, MO). Expression plasmids were linearized and used as templates for capped cRNA synthesis with the T3 mMessage mMachine kit (Life Technologies). RNA transcripts were purified by LiCl precipitation and rinsed with ice-cold 70% ethanol. Air-dried pellets were dissolved in an appropriate volume of nuclease-free water supplemented with RNase inhibitor SUPERase-In (Life Technologies) and stored at −80°C until use.
Xenopus oocyte preparation, injection, and culture
Xenopus ovaries were obtained from either Xenopus 1 or Nasco. Oocytes were isolated from the ovaries and de-folliculated with 1 mg/ml Type II collagenase (Sigma-Aldrich) in Ca2+-free ND98 (98 mM NaCl, 2 mM KCl, 2 mM CaCl2, 1 mM MgCl2, and 5 mM HEPES, pH 7.2). Snipping the ovaries with scissors to increase fluid access to the interior and pre-rinsing in Ca2+-free ND98 increased the speed of ovary digestion and the quality of oocytes obtained. After digestion, mature, high quality oocytes were transferred into culture solution that included the addition of 1.8 mM Ca2+, 2.5 mM Na-pyruvate, 100 U/ml penicillin, and 100 µg/ml streptomycin to the ND98 base solution. Oocytes were cultured in this solution at 18°C before and after injection. Oocytes were injected with 50–70 nl of RNA solution using a Nanoject II injector (Drummond Scientific Company) and cultured 1–5 d before recording. The amount of RNA injected was adjusted to optimize current size based on empirical observations and varied between constructs. Chemical reagents for solutions were obtained from Sigma-Aldrich.
Oocytes were placed in a hypertonic stripping solution (culture solution supplemented with 200 mM sucrose) for 3–10 min to allow separation of the plasma membrane from the vitelline envelope. Forceps were used to remove the vitelline envelopes, and oocytes were transferred to the recording dish filled with internal solution (142 mM K+, 138 mM methanesulfonate, 4 mM Cl−, 10 mM HEPES, and 5 mM EGTA, pH 7.2). Patch pipettes were coated with Sticky Wax (Kerr Dental Laboratory Products) to minimize capacitance and had a resistance of 0.4–1 MΩ after fire polishing. The pipette solution contained 142 mM K+, 1 mM Mg+, 0.2 mM Ca2+, 140 mM methanesulfonate, 4.4 mM Cl−, and 10 mM HEPES, pH 7.2. The Ag-AgCl ground was placed in 1 M NaCl and connected to the bath with a 1-M NaCl agarose bridge. Junction potential was cancelled before patch formation, and pipette capacitance was compensated. Inside-out patches were immediately moved into a focal stream of internal solution after excision and maintained under constant perfusion. Data were collected using a Multiclamp 700A amplifier and the pClamp 9 acquisition suite (Molecular Devices). Data were sampled at 20 kHz and filtered at 1.4 kHz. In some cases, multiple sweeps were averaged to improve the signal-to-noise ratio. Phosphatidylinositol-4,5-bisphosphate C-8 (diC8-PIP2), phosphatidylinositol-4-phosphate C-8 (diC8-PI(4)P), and phosphatidylinositol-3,4,5-trisphosphate C-8 (diC8-PIP3) were purchased from Cayman Chemicals. 2.5-mM stock solutions were made with internal solution and stored at −80°C. The phosphoinositides were diluted to final concentrations with internal solution and applied directly to the intracellular face of excised patches through the perfusion stream.
G-V curves were determined from isochronal tail currents at −100 mV. Data were fitted with a single Boltzmann distribution in Origin 8.1 (OriginLab) using the equation f(V) = (A1 − A2)/(1 + e(V−V50)/s) + A2, where V50 represents the midpoint; s is the slope factor; and A1 and A2 represent the upper and lower asymptotes, respectively. Data from individual patches were normalized and averaged before plotting, and the Boltzmann fits shown in the figures are simulated using the arithmetic means of V50 and s values. For measurements of activation time course, the major phase of current activation (initial 80% of current rise after sigmoidal delay) was fitted with a single exponential in Clampfit (Molecular Devices) using the equation f(t) = y0 + Ae−t/τ, where y0 is the current offset, A is the amplitude, and τ is the time constant. Data for dose–response relationships were fit in Origin 8.1 with the modified Hill equation with an offset: y = START + (END − START)xn/(kn + xn), where START and END are the bottom and top of the curve, respectively; k is the PIP2 concentration producing half-maximal effect; and n is the Hill coefficient. Statistical significances were assessed using two-tailed Student’s t tests.
Patch excision alters voltage-dependent gating of Elk1
We first characterized gating changes that occur in Elk1 after excision of inside-out patches from Xenopus oocytes to assess the potential for the existence of PIP2-dependent regulation. Patch excision leads to gradual loss of PIP2 from the plasma membrane, and gating alterations that accompany patch excision therefore can reflect PIP2 dependence. For instance, rundown of KCNQ channels in inside-out patches can be reduced by manipulations that slow the loss of PIP2 and can be reversed by the application of a soluble PIP2 analogue diC8-PIP2 to the intracellular face of the patch (Li et al., 2005; Zaydman et al., 2013). Fig. 1 A shows Elk1 currents in symmetrical K+ evoked by voltage ramps before patch excision and several minutes after patch excision when currents had stabilized. Two obvious changes (arrows) include the enhanced activation of Elk1 at hyperpolarized potentials and a dramatic speed up of deactivation. Channel activity as measured by peak tail current at −100 mV after the voltage ramp increased for several minutes after excision and then remained stable through 15 min (Fig. 1 B). Example Elk1 current traces recorded in response to 1-s voltage steps from on-cell and excised patches are shown in Fig. 1 C, and voltage activation (G-V) curves measured from isochronal tail currents are shown in Fig. 1 D (fit parameters are provided in Table 1). Patch excision shifts the V50 of the voltage-activation curve −18.6 ± 1.3 mV. It leads to an increase in the activation rate of Elk1 (Fig. 1 E), as measured by a decrease in the time constant of activation (τact) (Fig. 1, E and F) that occurred over several minutes in parallel with the run-up of tail current size. Patch excision also increases the rate of deactivation, as measured by a significant decrease in the fraction of tail current remaining 60 ms after repolarization from 60 to −100 mV (Fig. 1 G). Measurements of channel activity in excised patches described later in this paper were taken only after currents elicited by voltage ramps had stabilized.
Elk1 voltage gating has a robust VSD mode shift that is lost upon patch excision
In Erg1, slow deactivation is caused by time-dependent mode shift of the activated voltage sensor (Piper et al., 2003; Tan et al., 2012; Goodchild et al., 2015); the mode-shifted VSD returns to the resting state in a more hyperpolarized voltage range than the original transition from resting to activated. Mode shift is believed to be caused by a relaxation of the VSD into a more stable activated conformation subsequent to its initial activation (Villalba-Galea et al., 2008). We examined mode shift in Elk1 to determine if it could explain the slow component of deactivation on cell, which is lost upon patch excision. Fig. 2 A shows families of on-cell Elk1 currents recorded in response to voltage steps from −120 to 60 mV from a holding potential of −100 mV (our standard G-V measurement) or with a 1-s prepulse to 60 mV to induce mode shift. The G-V V50 measured from isochronal tail currents recorded at −100 mV shifted from −30.8 ± 1.5 mV to −72.3 ± 1.0 mV when we introduced the prepulse to 60 mV (Fig. 2 B). This mode shift reduced from −41.6 ± 1.0 mV to −3.7 ± 0.7 mV after patch excision (Fig. 2 B). The magnitude of the slow component in Elk1 tail currents recorded at −100 mV on cell (but not excised) was highly sensitive to depolarizing step length (Fig. 2, C–E). We estimated the time course of mode shift at 60 mV by plotting the increase in tail current remaining 60 ms after repolarization to −100 mV (Fig. 2 E). An exponential fit of the data yielded a time constant of 505 ± 10 ms (n = 34).
PIP2 slows activation of Elk1 in excised patches and partially restores mode shift
We applied PIP2 to excised patches to determine if the gating changes we observed in Elk1 upon patch excision could be explained by loss of PIP2. We used the diC8 analogue of PIP2 for all experiments because of its enhanced solubility relative to native PIP2. Fig. 3 A shows ramp and step currents from an excised patch before and after the application of 10 µM diC8-PIP2 (referred to as “PIP2” from here on). 10 µM PIP2 depolarized the G-V curve measured from a −100-mV resting potential by 21.7 ± 1.1 mV; the resulting G-V curve was similar to what we observed on cell (Fig. 3 B). 10 µM PIP2 slowed the activation rate of Elk1 relative to excised control conditions (Fig. 3 C) as measured by the ratio of the time constant of activation in PIP2, τact(PIP2), to the time constant of activation in control solution, τact(Ctrl) (Fig. 3 D). The application of 10 µM PIP2 to excised patches also introduced a slow component into tail currents and significantly increased mode shift by 16.2 mV from −3.7 ± 0.7 mV to −19.9 ± 1.3 mV (Fig. 3, E–G).
We calculated the half-maximal PIP2 concentration (K1/2) for effects on τact at −20 mV, G-V V50, and mode shift using Hill-equation fits of data at various PIP2 concentrations (Fig. 3, H–J). The K1/2 values calculated by each method were similar: 12.7 ± 2.2 µM for τact, 9.4 ± 2.7 for G-V V50, and 11.3 ± 1.7 µM for mode shift. PIP2 could fully restore the slow activation of Elk1 observed on cell at a concentration just above these K1/2 values (between 10 and 30 µM; Fig. 3, D and H). Mode shift was not fully restored to excised patches at the same concentrations (Fig. 3 J), suggesting that PIP2 may not fully account for the large on-cell mode shift. These results show that PIP2 regulates the voltage gating of Elk1 in a bimodal fashion by inhibiting voltage activation while enhancing stabilization of open channels. Furthermore, the K1/2 values we determined for the application of diC8 PIP2 to excised patches suggest that Elk1 gating could be sensitive to changes in native PIP2 concentrations.
Depletion of native PIP2 with CiVSP modulates Elk1 voltage gating
Although a huge diversity of channels is regulated by PIP2 as measured by patch excision and direct PIP2 application to the intracellular face of patches, only a subset of channels appears to be regulated by changes in PIP2 concentration in intact cells (Hilgemann, 2012; Kruse et al., 2012; Kruse and Hille, 2013). In cases where channel activity requires PIP2, but is not modulated by physiological changes in PIP2 concentration, it has been suggested that constitutive PIP2 binding might be necessary for normal channel function (Hilgemann et al., 2001; Suh and Hille, 2008; Hilgemann, 2012; Kruse et al., 2012; Zhou and Logothetis, 2013). We therefore examined whether depletion of native PIP2 by the depolarization-activated phosphoinositide 5-phosphatase CiVSP (Murata et al., 2005; Halaszovich et al., 2009) could alter the gating of Elk1 in on-cell patches. Repeated depolarization of whole oocytes expressing CiVSP under two-electrode voltage clamp depletes PIP2 sufficiently to inhibit PIP2-dependent KCNQ currents (Li et al., 2015a). Oocytes for patches are often damaged during removal of the vitelline envelope, and we bathed the oocytes in high K+ intracellular recording solution, so it is reasonable to assume that they were sufficiently depolarized for some basal CiVSP activation and PIP2 depletion before patch formation. Indeed, activation of Elk1 currents was already fast in most CiVSP-expressing versus control patches during the first depolarizing voltage step applied to the patch. Nevertheless, repeated depolarizing steps to 60 mV further increased the rate of activation in some patches, a change that was never observed in control oocytes (Fig. 4 A). The mean activation time constants for on-cell patches recorded from CiVSP-expressing oocytes in response to increasing voltage steps were significantly smaller than for control oocytes (Fig. 4 B). However, CiVSP coexpression does not speed up activation as much as patch excision (Fig. 4 B), possibly because of incomplete depletion of native PIP2 during the steps or rapid PIP2 resynthesis between steps. Similarly, the deactivation time course was significantly faster in CiVSP-expressing oocytes (Fig. 4, C and D), and mode shift was significantly reduced from −41.6 ± 1.9 mV in control oocytes to −22.0 ± 1.4 mV in CiVSP-expressing oocytes (Fig. 4 E). These results suggest that native PIP2 in oocytes modulates Elk1 the same way as diC8-PIP2 in excised patches.
PIP2 and PIP3 regulate Elk1
We tested the ability of the diC8 analogues of two other plasma membrane phosphoinositides, phosphatidylinositol 4-phosphate (PI(4)P) and phosphatidylinositol (3,4,5)-trisphosphate (PIP3), to modulate the voltage gating of Elk1. PI(4)P is the precursor for PIP2 synthesis and is the major metabolite produced from PIP2 by CiVSP (Halaszovich et al., 2009). PIP3 is typically >100-fold less abundant than PIP2 in cells (Vanhaesebroeck et al., 2001) but strongly modulates CNG channel gating (Zhainazarov et al., 2004; Brady et al., 2006; Bright et al., 2007). Elk1 activation rates with the application of 10 µM of each diC8 phosphoinositide are compared in Fig. 5 (A and B). PIP3 slowed activation significantly more than PIP2, whereas PI(4)P had only a small effect on activation at 10 µM. The same rank order of effect (PIP3 > PIP2 > PI(4)P) was observed at the applied 10-µM concentration for mode shift (Fig. 5, C and D). The small effects observed for PI(4)P relative to PIP2 are consistent with the ability of the phosphoinositide 5-phosphatase CiVSP to enhance activation rate and reduce mode shift in Elk1. These data suggest that phosphoinositide regulation of Elk1 could occur through either PIP2 or PIP3, although we did not fully explore the affinity and efficacy of PIP3 for this study. We instead focused on PIP2 modulation for this study to facilitate comparison to the broad literature on PIP2 regulation of channels.
The role of the eag domain in mode shift
We next examined which domains of Elk1 might be responsible for the PIP2 dependence of voltage gating. Because the eag domain and conserved basic residues at the N terminus of the Cap support mode shift in hErg1 (Tan et al., 2012), we deleted the entire eag domain of Elk1 (Δ2–136), or alternatively neutralized a single conserved basic residue in the N-terminal Cap (K5Q), to test the role of the Elk1 eag domain in open-state stabilization. K5 aligns with a Cap residue (R5) that plays a role in mode shift in Erg1 (Fig. 6 A). Elk1 Δ2–136 and K5Q currents recorded on cell in response to voltage steps are shown in Fig. 6 (B and C), and the time course of on-cell tail currents recorded after a 1-s 60-mV pulse is compared with WT in Fig. 6 D. Elk1 Δ2–136 tail currents were extremely rapid on cell compared with WT, whereas K5Q tail currents were intermediate between the two. Mode shift on cell was almost completely eliminated in Elk1 Δ2–136 (−6.2 ± 3.1 mV; n = 9) and reduced significantly to −26.7 ± 1.7 mV (n = 11; P < 0.0001; t test) for Elk1 K5Q (Fig. 6, E and F, and Table 2). Neither mutant exhibited mode shift after excision; 10 µM PIP2 actually accelerated deactivation and introduced a small but significant increase in mode shift for K5Q but not Δ2–136 (Fig. 6, G–J, and Table 2). However, the increase in mode shift in excised patches caused by 10 µM PIP2 relative to control solution (ΔΔV50) was significantly reduced in both K5Q and Δ2–136 versus WT (Fig. 6 K). The accelerated deactivation we observed for both mutants in 10 µM PIP2 (Fig. 6 G) is likely a consequence of the large depolarized G-V shifts in the presence of PIP2 (Fig. 6, H and I), and we did not study it further. Finally, the on-cell mode shift of Elk1 K5Q was not reduced by coexpression with CiVSP (Fig. 6, L and M, and Table 3) and was identical in magnitude to the mode shift remaining for Elk1 WT after depletion of PIP2 by CiVSP (−21.3 ± 2.0 mV vs. −22.0 ± 1.4 mV; P = 0.76; n = 8 and 4, respectively). These results indicate that the eag domain is required for open-state stabilization and the accompanying mode shift in Elk1, and that K5 in the Cap increases the ability of PIP2 to promote stabilization and mode shift.
We reasoned that the ability of PIP2 to inhibit voltage activation in Elk1 K5Q and Δ2–136 mutants might be intact because of the large rightward shift of the G-V after 10 µM PIP2 application to excised patches (Fig. 6, H and I) and the leftward G-V shift observed for K5Q in the presence of CiVSP (Fig. 6 L). The activation rates of Elk1 Δ2–136 and K5Q were indeed slowed by PIP2 (Fig. 7, A and B), indicating that other regions of the Elk1 channel must mediate inhibitory effects of PIP2. The fold changes of τact caused by 10 µM PIP2 relative to excised controls were similar for Δ2–136, K5Q, and WT (Fig. 7, A and B). However, the rightward G-V shift caused by 10 µM PIP2 was significantly increased for both mutants (Fig. 7 C). PIP2-dependent G-V shifts reflect a composite of PIP2-dependent inhibition of activation counterbalanced by mode shift. We hypothesize that the depolarizing G-V shifts are bigger in K5Q and Δ2–136 because the counterbalancing effect of mode shift is significantly reduced.
Basic residues in the S4–S5 linker and post-S6 are required for PIP2-dependent inhibition of Elk1
Basic residues at the cytoplasmic interface of the VSD and S6 activation gate have been implicated in PIP2-dependent inhibition of Shaker channel gating (Rodriguez-Menchaca et al., 2012a) and PIP2-dependent coupling of VSD activation to pore opening in KCNQ channels (Zaydman et al., 2013). We therefore tested whether basic residues in the S4–S5 linker (R347) and post-S6 (R475 and R479) of Elk1 (Fig. 8, A and B) were required for PIP2-dependent modulation of Elk1 gating using charge-reversal and neutralization substitutions. R347E was functional (Fig. 9 A) but exhibited significantly smaller changes in activation rate relative to WT upon the application of 10 µM PIP2 (Fig. 9, A and B). Furthermore, deactivation was more rapid on cell in R347E, and PIP2 did not introduce a slow component in excised patches (Fig. 9 C). In contrast, PIP2 strongly inhibited activation of R347C and introduced a small slow component into the tail (Fig. 9, D–F). On-cell mode shift was significantly reduced in in R347E, and PIP2 did not significantly increase mode shift in excised patches (Fig. 9 G and Table 2). We hypothesize that R347E retains the ability to mode shift but lacks PIP2-dependent enhancement of mode shift. In contrast, R347C had normal mode shift on cell but significantly reduced PIP2-dependent enhancement of mode shift in excised patches (ΔΔV50, −7.3 ± 1.2 mV [n = 6] vs. −15.8 ± 1.1 mV [n = 18] for WT; P < 0.001). Patch excision had little effect on the G-V of R347E, but 10 µM PIP2 caused a 20.6 ± 3.6–mV shift in V50 (Fig. 9 H and Table 1). This is similar to what we observed for WT but less than what we observed for K5Q, which has normal PIP2-dependent inhibition of activation but (like R347E) lacks the counterbalance of PIP2-dependent enhancement of mode shift. This reduction in the PIP2-dependent depolarizing G-V shift relative to K5Q supports our observation that PIP2 has a reduced ability to inhibit activation in R347E. The PIP2-dependent depolarizing G-V shift was similar in R347C (22.9 ± 2.5 mV; Fig. 9 I) but was likely constrained by residual PIP2-dependent mode shift (Fig. 9 G).
Charge-reversal mutations at R475 (R457E) and R479 (R479E) had differential effects on PIP2-dependent inhibition of voltage activation (Fig. 10, A–C). PIP2 slowed voltage activation of R475E in a similar fashion to WT, whereas the effects of PIP2 on activation were significantly reduced in R479E, as quantified by the fold change in the activation rate induced by 10 µM PIP2. R475E also had a large mode shift on cell and normal PIP2-dependent enhancement of mode shift when excised (Fig. 10 D and Table 2). In contrast, on-cell mode shift was significantly reduced for R479E, and PIP2 did not recover mode shift in excised patches (Fig. 10 D and Table 2). The R475E G-V shifts upon excision and 10 µM PIP2 application were unexpectedly less than we observed for WT (Fig. 10 E and Table 1), but we did not explore this further because PIP2-dependent mode shift and inhibition of activation were normal when assayed individually. R479E had a depolarizing shift in G-V upon patch excision, and 10 µM PIP2 did not depolarize the G-V further (Fig. 10 F and Table 1). This phenotype is consistent with a large reduction in bimodal PIP2 regulation. Neutralization of R479 (R479Q) showed a parallel but less severe phenotype to R479E (Fig. 10, G–I, and Tables 1 and 2). Inhibition of activation in R479Q was significantly reduced relative to WT. R479Q on-cell mode shift was also significantly reduced, but a small amount of PIP2 enhancement of mode shift remained. These data argue that R479 (but not R475) has a strong influence on PIP2-dependent modulation of voltage gating in Elk1.
We further explored the roles of R347 and R479 in PIP2-dependent gating by examining the effect of native PIP2 depletion by CiVSP on these mutants. The activation rate of R347E was consistently faster at a given voltage in the presence of CiVSP, but the changes were small and only significant at −40 mV (Fig. 11, A and B). CiVSP did not alter the R347E G-V curves or significantly reduce on-cell mode shift (Fig. 11, C and D). R479E had essentially the same phenotype in the presence of CiVSP, except no significant changes in activation rate were observed and there was a small right shift in the G-V curves (Fig. 11, E–G). R479E mode shift on cell was slightly smaller than what we observed in previous experiments, but it was not further reduced by CiVSP (Fig. 11 H). The relative insensitivity of R347E and R479E to CiVSP further supports the conclusion that these mutants have reduced bimodal PIP2 modulation.
Elk1 is the first example of an EAG family channel that displays bimodal regulation of voltage gating by PIP2. This regulation is in part mediated by two basic residues in the S4–S5 linker (R347) and post-S6 region (R479). The residues are required for both PIP2-dependent inhibition of voltage gating and PIP2-dependent stabilization of open channels. One important open question regarding the bimodal regulation of voltage-gated channels by PIP2 has been whether the sites for inhibition and enhancement are spatially separate or overlapping (Zhou and Logothetis, 2013). Our functional data suggest that the sites could overlap in Elk1. Because the S4–S5 linker and cytoplasmic edge of S6 are the site of coupling between the VSD and activation gate (Chen et al., 2001; Tristani-Firouzi et al., 2002; Long et al., 2005a,b), these basic residues could lie in close proximity and conceivably have direct electrostatic interactions with PIP2, as has been suggested for similarly placed residues in KCNQ channels (Zaydman et al., 2013; Zaydman and Cui, 2014). Direct S4–S5 linker and post-S6 (TRP-box) interactions with PIP2 mediated by basic residues have also been proposed for Kv1.2 (Rodriguez-Menchaca et al., 2012a) and TRP channels (Rohács et al., 2005; Poblete et al., 2015), respectively. This mechanism is appealing because PIP2–protein interactions are typically mediated by direct binding to basic residues (Suh and Hille, 2008). However, our data do not differentiate between direct binding and allosteric interaction between the S4–S5 linker/post-S6 and PIP2 bound at an alternate site(s). Basic residues in the C-linker have been shown previously to be involved in PIP2-dependent inhibition of HCN channels (Flynn and Zagotta, 2011), which may be equivalent to the open-state stabilization we observed for Elk1 in our study. The Elk1 C-linker has similarly positioned basic residues, but we did not examine their role in PIP2-dependent modulation in this study.
Suh and Hille (2008) proposed two alternate models of how PIP2 can interact with ion channels: (1) a tight pocket with high affinity and specificity for PIP2, or (2) a more general electrostatic interaction with lower affinity and less selectivity between phosphoinositides. Our Elk1 data are supportive of a binding site with the latter characteristics. The affinity for PIP2 (K1/2 of roughly 10 µM) is relatively low, and PIP3 also had a strong bimodal effect on gating. The main requirement for strong modulation appears to be the presence of multiple phosphate groups on the inositol ring, as PI(4)P has much less effect on gating. A more general association between PIP2 and the cytoplasmic gating machinery of voltage-gated channels is also suggested by sequence conservation of the S4–S5 linker and post-S6. For instance, basic residues within these regions are required for PIP2-dependent modulation of gating in Elk1, KCNQ1, and Kv1.2, but the precise side chain, position, and pattern of basic residues are variable (Fig. 8, A and B). We suggest that PIP2 might be able to be accommodated at the activation gates of most voltage-gated channels, but its effects on gating will vary based on the precise contacts made. In this manner, PIP2 regulation could be conserved across the voltage-gated superfamily yet plastic through evolutionary time; various aspects of PIP2 modulation might come and go rapidly as the gating machinery sequence evolves. For instance, evidence so far suggests that KCNQ channels, which strictly require PIP2 for activation, probably evolved from the Shaker family sometime after the evolution of the first nervous systems (Li et al., 2015a). Shaker family channels vary with respect to PIP2 modulation (Hilgemann et al., 2001; Kruse et al., 2012; Rodriguez-Menchaca et al., 2012a; Kruse and Hille, 2013), but most do not require PIP2. Shakers from cnidarians are stable in excised patches (Jegla et al., 2012), so this lack of a requirement for PIP2 is ancient and was present in Shaker family channels at roughly the time the KCNQ family first appeared. Thus, it is possible that KCNQ channels gained strict PIP2 dependence de novo. It would be interesting to see if de novo insertion of single or a few basic residues within the gating machinery of diverse PIP2-agnostic voltage-gated channels could awaken a latent PIP2 sensitivity.
It remains to be seen if bimodal PIP2 modulation is a conserved property of Elk channels, but the basic residues we identified as being important for PIP2 modulation in Elk1 (K5, R347, and R479) are universally conserved as basic residues in mouse, human, fly, and sea anemone Elk channels, with one major exception (Fig. 8). The Nematostella vectensis (sea anemone) Elk orthologue lacks a basic residue at the 479 position, but it does have an adjacent R (at the 480 position equivalent) that is found in all Elk channels except Elk1. Erg1 is not modulated by PIP2 in a bimodal fashion (Bian et al., 2001; Rodriguez et al., 2010) and like all Erg and Eag subfamily channels intriguingly lacks a basic residue at the equivalent position of R479 (Fig. 8 B). PIP2 modulation has not been studied in the Eag subfamily to our knowledge, but it seems unlikely that they would have PIP2-dependent mode shift given their evolutionarily conserved rapid deactivation (Li et al., 2015b). Thus, it appears that PIP2 will have highly variable effects within the EAG family of channels. The auxiliary subunit KCNE1 enhances the PIP2 sensitivity of KCNQ1, suggesting that PIP2 modulation of EAG family channels could be further influenced by KCNEs, which are known to interact with some Erg and Elk channels (McDonald et al., 1997; Abbott et al., 1999; Clancy et al., 2009).
The position of the Cap domain in EAG family channels is somewhat controversial, with functional data from Eag1 (Terlau et al., 1997) and cross-linking experiments in Erg1 (de la Peña et al., 2011, 2015) suggesting close proximity to the S4–S5 linker, and structural and FRET data presenting an alternative that the Cap might instead interact through the CNBHD (Gianulis et al., 2013; Haitin et al., 2013). Structural evidence clearly shows the eag domain including part of the cap docks on the CNBHD (Haitin et al., 2013), but the complete position of the Cap has not been structurally resolved. However, it has been suggested that the Cap could regulate the S4–S5 linker while the eag domain is docked on the CNBHD (Gustina and Trudeau, 2012), and we find that the highly conserved basic residue near the tip of the Cap (K5 in Elk1) and R347 in the S4–S5 linker are both important for PIP2-dependent mode shift. However, K5 could allosterically influence voltage gating through the eag domain, which we find is required for mode shift (Fig. 6). Basic residues R4 and R5 in the Cap of hErg1 are required for slow deactivation (Muskett et al., 2011) and the hErg1 Cap is required for mode shift (Tan et al., 2012), but the dependence of this effect on PIP2 has not been explored. PIP2 does appear to enhance hErg1 through open-state stabilization (Rodriguez et al., 2010), but currently only the distal C terminus of hErg1 has been linked to PIP2-depenent modulation (Bian et al., 2004; Bian and McDonald, 2007).
The in vivo role of Elk channels has not been extensively studied, so the physiological importance of PIP2-dependent modulation of Elk voltage gating is not yet clear. However, two of our findings suggest a possible role for PIP2 in Elk1 modulation in vivo: (1) the Elk1 PIP2 affinity is relatively low (Fig. 3), indicating that channel activity could be affected by physiological changes in PIP2 (Suh and Hille, 2008); and (2) depletion of native PIP2 with CiVSP in oocytes was sufficient to alter Elk1 gating properties (Fig. 4). Gene knockout and pharmacological experiments in mouse show that Elk2 in pyramidal neurons contributes to subthreshold K+ conductance and alters firing threshold (Zhang et al., 2010). This is a qualitatively similar role to that played by KCNQ channels (M current), which strictly require PIP2 for opening (Li et al., 2005; Zaydman et al., 2013; Zaydman and Cui, 2014). Elk1 current has not been identified in vivo, but it also activates at subthreshold potentials, is highly expressed in sympathetic ganglia (Shi et al., 1998), and is present at modest levels in various brain regions (Engeland et al., 1998; Zou et al., 2003). Our results suggest that the physiological roles of Elk and KCNQ channels fundamentally diverge at the level of PIP2 regulation. Elk1 still activates after PIP2 depletion, and the net effect of bimodal PIP2 regulation is inhibitory. PIP2 serves to slow both the activation and deactivation rates of Elk1, which might effectively reduce the channel’s response to rapid voltage changes. PIP2 hydrolysis might then increase the response of the channel to brief depolarizations (accelerated activation) without a lasting increase in resting K+ conductance (accelerated deactivation). Characterization of Elk1 currents in neurons with genetic and pharmacologic tools will be necessary to test these hypotheses.
This work was supported by National Institutes of Health grant R01 NS069842 (to T. Jegla) and the Department of Biology at Pennsylvania State University.
The authors declare no competing financial interests.
Kenton J. Swartz served as editor.
Ciona intestinalis VSP
cyclic nucleotide–binding domain
cyclic nucleotide–binding homology domain