Skip to main content
PMC home page
eLife logoLink to eLife
. 2019 Sep 6;8:e49188. doi: 10.7554/eLife.49188

Regulation of Eag1 gating by its intracellular domains

Jonathan R Whicher 1, Roderick MacKinnon 1,✉
Editors: Gary Yellen2, Olga Boudker3
PMCID: PMC6731095  PMID: 31490124

Abstract

Voltage-gated potassium channels (Kvs) are gated by transmembrane voltage sensors (VS) that move in response to changes in membrane voltage. Kv10.1 or Eag1 also has three intracellular domains: PAS, C-linker, and CNBHD. We demonstrate that the Eag1 intracellular domains are not required for voltage-dependent gating but likely interact with the VS to modulate gating. We identified specific interactions between the PAS, CNBHD, and VS that modulate voltage-dependent gating and provide evidence that VS movement destabilizes these interactions to promote channel opening. Additionally, mutation of these interactions renders Eag1 insensitive to calmodulin inhibition. The structure of the calmodulin insensitive mutant in a pre-open conformation suggests that channel opening may occur through a rotation of the intracellular domains and calmodulin may prevent this rotation by stabilizing interactions between the VS and intracellular domains. Intracellular domains likely play a similar modulatory role in voltage-dependent gating of the related Kv11-12 channels.

Research organism: Virus, Xenopus, Other

Introduction

Voltage-gated potassium channels (Kvs) conduct potassium ions in response to changes in membrane voltage. All Kvs are tetramers and consist of 6 transmembrane segments (S1-S6) (Jiang et al., 2003; Long et al., 2005a; Long et al., 2007; Sun and MacKinnon, 2017; Wang and MacKinnon, 2017; Whicher and MacKinnon, 2016). S1-S4 form the voltage sensor (VS) and S5-S6 form the potassium pore. The S4 helix of the VS is positively charged and was proposed to move within the membrane in response to changes in membrane voltage (Aggarwal and MacKinnon, 1996; Seoh et al., 1996). Upon membrane hyperpolarization the S4 was predicted to move ‘down’ towards in the intracellular side of the membrane to close the potassium pore and upon membrane depolarization the S4 was predicted to move ‘up’ to the extracellular side of the membrane to open the potassium pore. In Kvs 1–9 movement of S4 is coupled to the potassium pore by a ~ 15 residue, helical S4-S5 linker, which forms a domain-swapped linkage between S4 and S5 that positions it directly above the pore lining S6 helix (Long et al., 2005a; Long et al., 2007; Sun and MacKinnon, 2017). In this position, the S4-S5 linker was proposed to function as a mechanical lever to couple movement of the S4 to the S6 helices to open and close the pore (Long et al., 2005b). However, in Kvs 10–12 the S4-S5 linker is only six residues, which is not long enough to form the domain-swapped linkage observed in Kvs 1–9. Indeed, recent structures of Eag1 (Kv10.1) and hErg (Kv11.1) revealed that the S4-S5 linker forms a non-domain swapped linkage between S4 and S5 and is unlikely to be domain swapped in any conformation (Figure 1a,b,c) (Wang and MacKinnon, 2017; Whicher and MacKinnon, 2016). Due to the non-domain swapped transmembrane architecture, the S4-S5 linker is not in the same position and does not appear to form a similar mechanical lever, suggesting an alternative mechanism of voltage-dependent gating in Kvs 10–12 (Lörinczi et al., 2015; Tomczak et al., 2017).

Figure 1. Comparison of the PAS loop in Eag1 and hErg.

Figure 1.

Open in a new tab

(A) Structure of Eag1 (PDB: 5K7L) in the closed conformation (PAS is orange, VS is yellow, S4-S5 linker is blue, pore is green, C-linker is red, CNBHD is cyan, CaM is purple, and membrane is indicated by gray bars) with a black box indicating the view for panel (B). (B) View of interaction between PAS loop, VS and CNBHD in the closed conformation of Eag1 with the same coloring as in (A). The N-terminus of the protein is shown as a sphere. The PAS loop N-terminus is not observed in this structure. (C) View of the interaction between PAS loop, VS and CNBHD in the open conformation of hErg (PDB: 5VA2) with the same coloring and orientation as in (B). The N-terminus is shown as a sphere and the PAS loop N-terminus is shown in gray.

Kvs 10–12 have three intracellular domains: an N-terminal Per-ARNT-Sim domain (PAS), a C-terminal C-linker domain, and a C-terminal cyclic nucleotide binding homology domain (CNBHD). Interactions between the VS and intracellular domain observed in the structure of Eag1 suggests that the intracellular domains may function in voltage-dependent gating (Whicher and MacKinnon, 2016) (Figure 1a). For example, the PAS domain, which is positioned directly below the VS, has a 15-residue N-terminal loop (PAS loop) that is directed, through interactions with the VS and CNBHD, towards the S4-S5 linker (Figure 1b). The PAS loop has been implicated in voltage-dependent gating and the large Cole-Moore effect observed in Eag1, in which more hyperpolarized (negative) resting membrane potentials result in slower rates of activation (Cole and Moore, 1960; Ju and Wray, 2006; Ludwig et al., 1994; Terlau et al., 1997). In addition, the C-linker forms an intracellular ring directly below the S6 helices, which are positioned to couple movements of the intracellular domains to the pore (Figure 1a). The C-linker ring is near the S4-S5 linker and the S4, which adopts a depolarized or up conformation in the Eag1 structure (Whicher and MacKinnon, 2016). Since the S4-S5 linker of Eag1 appeared unlikely to function as a mechanical lever, we proposed a voltage-dependent gating mechanism in which the S4 helix moves towards the intracellular side of the membrane during hyperpolarization to interact with and rotate the C-linker and S6 helices to close the potassium pore (Whicher and MacKinnon, 2016). This proposal was based on structural data alone and awaits further examination by functional analysis.

In addition to voltage-dependent gating, Eag1 is also gated by the calcium sensor calmodulin (CaM) (Schönherr et al., 2000; Ziechner et al., 2006). CaM binds to Eag1 only in the presence of calcium and holds the pore closed even during membrane depolarization. Each PAS and CNBHD domain has a CaM binding site and thus there are eight binding sites per tetramer. In the Eag1 structure, each CaM molecule occupies two binding sites, one on the PAS and one on the CNBHD, clamping the two domains together (Figure 1a) (Whicher and MacKinnon, 2016). This binding orientation was shown to be essential for channel inhibition but it is unclear how CaM binding prevents opening of the pore. Here we investigate the role of the intracellular domains in voltage-dependent gating and CaM inhibition. We identify interactions between the VS, PAS loop, and CNBHD that modulate voltage-dependent gating and are essential for CaM inhibition. We provide evidence that VS movement during depolarization may destabilize this interface between the PAS loop and the CNBHD to promote channel opening and find that CaM seems to function by stabilizing this interface to inhibit the channel. Finally, we determine a new structure of an Eag1 channel mutant that is insensitive to CaM inhibition. The structure revealed a pre-open conformation of Eag1 (i.e. a conformation that plausibly lies on the conformational pathway leading to opening) and suggests that channel opening may occur through a rotation of the intracellular domains.

Results

Role of the Eag1 intracellular domains in voltage-dependent gating

Based on the previous structure of Eag1, we proposed that membrane hyperpolarization causes S4 to interact with and rotate the C-linker to close the pore. To test this hypothesis, we characterized an Eag1 channel in which the PAS, C-linker, and CNBHD were deleted (Eag1TM). The C-terminal assembly domain (887-962) was included in Eag1TM as this domain is needed for tetramer assembly (Ludwig et al., 1997). Eag1TM forms functional channels that are voltage dependent (Figure 2a,b), indicating that the intracellular domains are not essential for voltage-dependent gating and that an interaction between the S4 and C-linker is not required to close the potassium pore. However, the intracellular domains do modify the gating kinetics, which was demonstrated by both the Eag1TM construct and the Eag1/hErg chimera where the PAS, C-linker and CNBHD from hErg were inserted onto Eag1TM (Figure 2a,b). Eag1TM had a right shifted V0.5 of 46 mV (zero-slope on the activation curve was not reached up to 100 mV and therefore an accurate value for V0.5 could not be measured) compared to the WT Eag1 (19 mV). In addition, the Eag1/hErg chimera has a V0.5 (2.5 mV) that is in between that of WT Eag1 and hErg (−22 mV) and exhibits slow rates of channel closure (deactivation), a characteristic of hErg channels. We also recorded the Cole-Moore effect for each construct by holding the cell at increasing holding potentials, from −190 mV to the voltage of channel activation, and stepping to the same depolarized voltage (Figure 2c). To compare the Cole-Moore effect from different mutants we plotted holding potential as a function of current at 10 ms following the depolarization step. Then we fit the plot with a Boltzmann function (defined in the Materials and methods) to estimate the holding potential that produces half maximal rates of activation (V0.5CM) (Figure 2d). Neither mutant channel exhibits a Cole-Moore effect (Figure 2c), which, along with the slow deactivation of the Eag1/hErg chimera and the shifted V0.5 of both channels, demonstrates that the intracellular domains influence voltage-dependent gating kinetics. In addition, this result suggests that the Cole-Moore effect may arise from an interaction between the transmembrane and intracellular domains.

Figure 2. Role of intracellular domains in voltage-dependent gating.

Figure 2.

Open in a new tab

(A) Voltage family current trace of WT Eag1, Eag1TM, and the Eag1/hErg chimera with the voltage-pulse protocol shown above. (B) Normalized tail current vs. depolarization voltage plot for WT Eag1 (black square, n = 6), Eag1TM (green circle, n = 4), and the Eag1/hErg chimera (red triangle, n = 4) with V0.5 values (mean ± sd). Eag1TM did not reach saturation up to 100 mV. (C) Cole-Moore effect of WT Eag1, Eag1TM, and the Eag1/hErg chimera with the voltage-pulse protocol shown above. (D) Plot of normalized current at 10 ms following the depolarization step vs holding potential for WT Eag1 (Cole-Moore I-V plot). The Cole-Moore I-V plot was fit with a Boltzmann function to estimate the holding potential that produces half maximal rates of activation (V0.5CM = −126 ± 0.9 mV, mean ± sd, n = 6).

Interactions between the voltage sensor and intracellular domains

To search for contacts between the intracellular domains and the transmembrane domains that influence gating properties we searched for mutations that modified or resulted in the loss of the Cole-Moore effect in Eag1. Using the structures of Eag1 and hErg as a guide, we first modified by alanine scanning mutagenesis the C-terminus of S4 and the S4-S5 linker (residues 343–348), which are near the PAS loop in both structures (Figure 1b, Figure 3, and Figure 3—figure supplement 1a,b) (Gianulis et al., 2013; Wang et al., 1998). These results suggest that the S4-S5 linker plays a role in voltage-dependent gating as we observed both negative (H343A, Y344A) and positive (D342A, I345A, E346A) shifted V0.5 values. Furthermore, H343A, Y344A, I345A, E346A, and Y347A all exhibit a Cole-Moore effect with both negative (H343A, I345A) and positive (E346A, Y347A) shifts in the V0.5CM (Figure 3—figure supplement 1c,d). We note that only mutation of Asp 342 to Ala (D342A) results in complete loss of the Cole-Moore effect, suggesting that Asp 342 may interact with the intracellular domains (Figure 3).

Figure 3. Role of Arg 7, Arg 8, and Asp 342 in voltage dependent gating.

(A) Voltage family current trace for D342A, R7A/R8A, and Δ3–9 with the voltage-pulse protocol shown above. (B) Normalized tail current vs. depolarization voltage plot of WT Eag1 (black square, n = 6), D342A (green triangle, n = 5), R7A/R8A (orange circle, n = 5), and Δ3–9 (red diamond, n = 5) with V0.5 values (mean ± sd). D342A, R7A/R8A, and Δ3–9 did not reach saturation up to 100 mV. (C) Cole-Moore effect of D342A, R7A/R8A, and Δ3–9 with the voltage-pulse protocol shown above. (D) Cole-Moore I-V plot for WT Eag1 (black square, n = 6), R7A/R8A (orange cirlce, n = 5), and Δ3–9 (red diamond, n = 5) with V0.5CM values (mean ± sd).

Figure 3.

Figure 3—figure supplement 1. S4-S5 linker mutations.

Figure 3—figure supplement 1.

(A) Voltage family current trace for H343A, Y344A, I345A, E346A, and Y347A with the voltage-pulse protocol shown above. (B) Normalized tail current vs. depolarization voltage plot of WT Eag1 (black square, n = 6), H343A (cyan triangle, n = 4), Y344A (green triangle, n = 5), I345A (red diamond, n = 5), E346A (orange circle, n = 6), and Y347A (blue unfilled square, n = 5) with V0.5 values (mean ± sd). (C) Cole-Moore effect of H343A, Y344A, I345A, E346A, and Y347A with the voltage-pulse protocol shown above. (D) Cole-Moore I-V plot for WT Eag1 (black square, n = 6), H343A (cyan triangle n = 3), Y344A (green triangle, n = 5), I345A (red diamond, n = 4), E346A (orange circle, n = 4), and Y347A (blue unfilled square, n = 5) with V0.5CM values (mean ± sd).
Open in a new tab

Asp 342 is located at the C-terminus of S4 and is highly conserved in Kvs 10–12. In the closed conformation of Eag1, Asp 342 does not interact with the intracellular domains (Figure 1b) (Whicher and MacKinnon, 2016). However, in the open conformation of hErg the homologous Asp is near (~6 Å) two Arg residues in the PAS loop (Figure 1c) (Wang and MacKinnon, 2017). In all Kvs 10–12 the PAS loop has at least one positively charged residue. In Eag1, the corresponding Arg residues are Arg 7 and 8. We mutated the Arg residues to Ala (R7A/R8A) and deleted residues 3–9 (Δ3–9) (Figure 3). The V0.5 of R7A/R8A and Δ3–9 are right shifted to a similar extent as Eag1TM and the D342A mutant. Therefore, Arg 7 and 8 promote channel opening, like Asp 342, and mutation of these residues has a similar effect on V0.5 as loss of the intracellular domains. In addition, the R7A/R8A and Δ3–9 mutations result in a right-shifted V0.5CM, indicating that more depolarized holding potentials are required for fast activation of these mutants than WT Eag1. The right shifted V0.5 and the modified Cole-Moore effect of R7A/R8A and Δ3–9 suggest that Arg 7 and 8 might form a functional interaction with Asp 342. Furthermore, since mutation of Arg 7 and 8 did not result in complete loss of the Cole-Moore effect, Asp 342 likely interacts with additional residues on the intracellular domains. Taken together, the functional data along with the proximity of the PAS loop and Asp 342 in the open conformation of hErg suggest that an interaction between Asp 342 and the PAS loop may occur in the open conformation to promote channel opening.

Implications for voltage-dependent gating

How might an interaction between the PAS loop and Asp 342 promote channel opening? One hypothesis is that the interaction between the PAS loop and Asp 342 stabilizes the depolarized state of the voltage sensor. This hypothesis would explain why deletion of the intracellular domains (Eag1TM) and mutation of Asp 342 and Arg 7 and 8 disfavors channel opening as indicated by a right shift in the voltage-dependence of activation. However, further analysis of PAS loop residues that are adjacent to Arg 7 and 8, Leu 10, Val 11, Ala 12, and Pro 13, revealed an additional function of the PAS loop in voltage-dependent gating. Leu 10, Val 11, Ala 12, and Pro 13 interact with Tyr 639, a conserved residue in Kvs 10–12 that is located on the CNBHD, and Tyr 213, a Phe, Tyr or Cys in Kvs 10–12 that is located on the loop before S1 of the VS (Figure 1b) (Whicher and MacKinnon, 2016). In this position, residues 10–13 link the VS and the intracellular CNBHD. To study the functional consequences of altering this region of contact, we generated Tyr 213 to Ala (Y213A) and Tyr 639 to Arg (Y639R) mutant channels and PAS loop mutant channels with successive deletions of residues 10–13 (Δ3–10, Δ3–11, Δ3–12, and Δ3–13). These mutants show inactivation and hooked tail currents at more depolarized potentials (40–100 mV), as was previously shown (Terlau et al., 1997) (Figure 4a,b and Figure 4—figure supplement 1a,b). These mutants also produce channels that open at more negative (hyperpolarized) potentials than WT Eag1 (−80 mV) and exhibit slow deactivation, demonstrating that the interaction between residues 10–13, Tyr 213, and Tyr 639 promotes the closed state of Eag1.

Figure 4. Interaction between residues 10–13, Tyr 213, and Tyr 639.

(A) Voltage family current trace for the Δ3–13, Y213A, and Y639R with the voltage-pulse protocol shown above. (B) Normalized tail current vs. depolarization voltage plot of WT Eag1 (black square, n = 6), Δ3–13 (orange triangle, n = 5), Y213A (green circle, n = 5), and Y639R (cyan diamond, n = 5) (mean ± sd). (C) Cole-Moore effect of Δ3–13, Y213A, and Y639R with the voltage-pulse protocol shown above.

Figure 4.

Figure 4—figure supplement 1. PAS loop deletions.

Figure 4—figure supplement 1.

(A) Voltage family current trace for Δ3–10, Δ3–11, and Δ3–12 with the voltage-pulse protocol shown above. (B) Normalized tail current vs. depolarization voltage plot of WT Eag1 (black square, n = 6), Δ3–10 (orange circle, n = 5), Δ3–11 (red diamond, n = 3), and Δ3–12 (blue triangle, n = 3). (C) Cole-Moore effect of Δ3–10, Δ3–11, and Δ3–12 with the voltage-pulse protocol shown above (mean ± sd).
Open in a new tab

Based on these data, the PAS loop can be divided into two functionally distinct segments: the N-terminus (residues 1–9; not observed in the Eag1 structure) and the C-terminus (residues 10–13) (Figure 1c). The N-terminus seems to promote channel opening and may interact with Asp 342 of the S4-S5 linker. The C-terminus seems to promote channel closure and interacts with the CNBHD. Furthermore, in the open state structure of hErg the PAS loop C-terminus does not interact with either Tyr 403 (equivalent to Tyr 213 in Eag1) or Tyr 827 (equivalent to Tyr 639 in Eag1) (Figure 1c) suggesting that destabilization of this interface might be necessary for channel opening (Wang and MacKinnon, 2017). Therefore, we propose that the following structural interactions take place in association with voltage dependent gating. Upon depolarization, Asp 342 interacts with the PAS loop N-terminus to stabilize the open state of the VS and destabilize the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639 to promote channel opening (Figure 1c). When the VS is hyperpolarized, movement of the S4 disrupts the interaction between Asp 342 and the PAS loop N-terminus, allowing the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639 to form and promote channel closing (Figure 1b).

Two experiments support this hypothesis for the role of the intracellular domains in voltage-dependent gating. First, co-expression of two halves of a split Eag1 construct (L341 split), in which the N-terminal half of the channel includes Met 1-Leu 341 and the C-terminal half includes Asp 342-Ser 962, produces a partially constitutively open channel (Figure 5a) (Tomczak et al., 2017). This functional behavior can be explained by the above hypothesis because in the split Eag1 construct Asp 342 is no longer connected to S4 of the VS and thus is no longer forced to move when the VS moves. Consequently, Asp 342 can maintain its interaction with the PAS loop to promote channel opening. Therefore, if Arg 7 and 8 and Asp 342 functionally interact to promote channel opening then mutation of this interface should produce a split channel that is no longer constitutively open. In agreement with this conclusion, introduction of the D342A and Δ3–9 mutations into the Eag1 split construct produce channels that close at hyperpolarized potentials (Figure 5b,c) (Tomczak et al., 2017). Second, this mechanism provides an explanation for the Cole-Moore effect. The Cole-Moore effect was proposed to be due to the existence of multiple closed states that the VS must transition through in order to reach an active or depolarized conformation (Cole and Moore, 1960). At more negative potentials the VS must transition through more closed states to reach an active conformation, which results in slower activation times. In Eag, at more negative holding potentials the VS might have to transition through more closed states in order for the S4 and Asp 342 to interact with the PAS loop N-terminus, which will result in slower activation times. Therefore, if we remove the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639, which we propose is destabilized when Asp 342 interacts with the PAS loop N-terminus, then the Cole-Moore effect should be lost. In agreement with this line of reasoning, Δ3–12 and Δ3–13 do not exhibit a Cole-Moore effect while Y213A, Y639R, Δ3–10, and Δ3–11 show a reduced Cole-Moore effect compared to the WT channel (Figure 4c and Figure 4—figure supplement 1c).

Figure 5. L341 split channels.

Figure 5.

Open in a new tab

Voltage family current trace for the L341 split (A), D342A L341 split (B), and Δ3–9 L341 split (C) with the voltage-pulse protocol shown above. (D) Normalized current vs depolarization voltage for L341 split (black square, n = 11), D342A L341 split (red diamond, n = 7), and Δ3–9 L341 split (orange circle, n = 7) (mean ± sd).

Structure of constitutively open Eag1

To better understand how the channel opens and how CaM inhibits opening we sought to determine the structure of an open Eag1 channel. In the presence of Ca2+/CaM, Eag1 Δ3–13 remains open at hyperpolarized voltages (Figure 6a). Therefore, we determined the Cryo-EM structure of Eag1 Δ3–13 in the presence of calcium and bound to CaM (Eag1 Δ3–13/CaM) (Figure 6b,c and Figure 6—figure supplements 1, 2, 3 and 4). Two different conformations were identified for Eag1 Δ3–13/CaM: conformation 1 at 3.7 Å and conformation 2 at 4.0 Å resolution. In both conformations, the S4 helices adopt a depolarized conformation and intracellular domains are rotated in a counterclockwise direction when viewed from the extracellular side of the membrane, but in conformation two the extent of the rotation is larger (2.4° degrees for conformation 1 and 8.6° degrees for conformation 2) (Figure 6b,c and Figure 6—figure supplement 4). The rotation observed in these conformations is in a similar direction as the intracellular domains in the open conformation structure of hErg (Wang and MacKinnon, 2017) (Figure 6d). However, the extent of the rotation of the Eag1 intracellular domains is not as large as the 20° rotation observed in hErg and the S6 helices remain closed (Figure 6d,e). As a result, we believe that conformation 1 and conformation 2 represent pre-open conformations of Eag1 on the pathway from closed to fully open.

Figure 6. Structure of Eag1 Δ3–13/CaM.

(A) Left, Voltage family current trace for Eag1 Δ3–13 in the presence of 1 mM CaCl2 with the voltage-pulse protocol shown above. Right, normalized current vs depolarization voltage for Eag1 Δ3–13 in the presence of 1 mM CaCl2 (black square, n = 3) with reversal potential (Erev) (mean ± sd). (B) Structural superposition of Eag1 Δ3–13/CaM conformation 1 (cyan) and Eag1/CaM (PDB-5K7L, (gray) using the selectivity filter. Only the intracellular domains are shown from an extracellular view and the location of the S6 helices are indicated with spheres. Degree of rotation is indicated by the arrow. (C) Structural superposition of Eag1 Δ3–13/CaM conformation 2 (green) and Eag1/CaM (gray) using the selectivity filter with the same view as (B). (D) Structural superposition of Eag1 Δ3–13/CaM conformation 2 (green C, red O, blue N), Eag1/CaM (gray C, red O, blue N), and hErg (PDB-5VA2, yellow C, red O, blue N) using the selectivity filter. Location of the intracellular gate Gln (Q476 for Eag1 and Q664 for hErg) are shown as ball and stick. (E) Plot of pore diameter for Eag1 Δ3–13/CaM conformation 2 (green), Eag1 (gray), and hErg (yellow). The location of the selectivity filter and intracellular gate are indicated and the dashed gray line at 6 Å indicates the diameter of hydrated potassium.

Figure 6.

Figure 6—figure supplement 1. Single-particle cryo-EM structure determination of Eag1 Δ3–13/CaM.

Figure 6—figure supplement 1.

(A) SDS-PAGE gel of fractions from the final gel filtration column of Eag1 Δ3–13/CaM (molecular weight standards are in kDa). (B) Representative micrograph of Eag1 Δ3–13/CaM in vitreous ice (scale bar = 500 Å). (C) Representative class averages from 2D classification (scale bar = 100 Å). (D) Gold standard FSC of conformation 1 (green) and conformation 2 (black) of Eag1 Δ3–13 bound to CaM. The dotted line indicates the 0.143 FSC cutoff which corresponds to the indicated resolutions. Orientation distribution of particles in the final reconstruction of conformation 1 (E) and conformation 2 (F) of Eag1 Δ3–13 bound to CaM. Each column indicates one view and the number of particles in each view is indicated by the size and color of the column (larger columns colored red contain a higher number of particles).
Figure 6—figure supplement 2. Cryo-EM density map of Eag1 Δ3–13/CaM.

Figure 6—figure supplement 2.

Cryo-EM maps of conformation 1 (A) and conformation 2 (B) colored based on local resolution estimated by ResMap. Slices of the transmembrane domain (top panel) and intracellular domains (bottom panel) colored by local resolution are shown on the right.
Figure 6—figure supplement 3. Structure validation of the atomic model of Eag1 Δ3–13/CaM.

Figure 6—figure supplement 3.

( A) FSC cross validation between the Eag1 Δ3–13/CaM model and the full map used for refinement for conformation 1 (green) and conformation 2 (black). (B). Collection parameters and refinement statistics for Eag1 Δ3–13/CaM conformation 1 and 2.
Figure 6—figure supplement 4. Structural superposition of the intracellular domains.

Figure 6—figure supplement 4.

(A) Side view of conformation 1 of Eag1 Δ3–13 bound to CaM. Position of the membrane is indicated by gray lines and domain coloring is as follows: PAS-orange, VS-yellow, S5-S6-green, C-linker-red, CNBHD-cyan, and CaM-purple. Black trace indicates the subunit shown in (B) and (C). (B) Structural superposition of the intracellular domains of Eag1 Δ3–13/CaM conformation 1 (colored) and Eag1/CaM (PDB-5K7L, gray) using the selectivity filter for alignment. (C) Structural superposition of the intracellular domains of Eag1 Δ3–13/CaM conformation 1 (colored) and Eag1/CaM (gray) using the intracellular domains for alignment. (D) Side view of conformation 2 of Eag1 Δ3–13 bound to CaM. Position of the membrane is indicated by gray lines and domain coloring is the same as in (A). Black trace indicates the subunit shown in (E) and (F). (E) Structural superposition of the intracellular domains of Eag1 Δ3–13/CaM conformation 2 (colored) and Eag1/CaM (gray) using the selectivity filter for alignment. (F) Structural superposition of the intracellular domains of Eag1 Δ3–13/CaM conformation 2 (colored) and Eag1/CaM (gray) using the intracellular domains for alignment.
Open in a new tab

A hypothesis for why the pore remains closed in the structure of Eag1 Δ3–13/CaM is that, compared to hErg, Eag1 might be more stable in a closed conformation. This hypothesis is consistent with a number of observations on the function. First, insertion of the intracellular domains of hErg onto Eag1 (Eag1/hErg chimera) causes a 20 mV left shift in the V0.5 (Figure 2a,b). Second, the V0.5 of Eag1 is right-shifted by 40 mV compared with hErg and Eag1TM is right shifted by 80 mV when compared to a hErg channel lacking the intracellular domains (Figure 2a,b) (Hausammann and Grütter, 2013; Wang and MacKinnon, 2017). What might cause Eag1 to be more stable in a closed conformation? In Eag1, Phe 475 and Gln 477 are located at the interface of the S6 helices on either side of Gln 476, the intracellular gate. In Kv11 and Kv12, which have a left shifted V0.5 compared to Eag1, these residues are Ile and Arg respectively (Bauer and Schwarz, 2018). The Eag1 double mutant F475I/Q477R causes a 50 mV left shift in the V0.5 to −20 mV when introduced into the full-length channel and a 16 mV left shift in the V0.5 to 30 mV when introduced into Eag1TM (Figure 7a,b,c). In addition, when the F475I/Q477R mutation is introduced into the Eag1/hErg chimera the channel remains open at hyperpolarized potentials (Figure 7d). Taken together, these data suggest that the intracellular domains and Phe 475 and Gln 477 cause Eag1 to be more stable in a closed conformation. Therefore, we propose that the Eag1 intracellular domains, when viewed from the extracellular side, rotate in a counterclockwise direction to promote the opening of the pore. However, due to the stability of Eag1 in a closed conformation and the conditions under which the Cryo-EM structure was determined we suspect that pore opening is transient and thus not observed in the Eag1 Δ3–13/CaM structure.

Figure 7. Eag1 pore mutants.

Figure 7.

Open in a new tab

(A) Phe 475 and Gln 477 (shown as green sticks, with red O, and blue N) are at the interface of the S6 helices (green). C-linker is shown in red. (B) Voltage family current trace for F475I/Q477R and Eag1TM F475I/Q477R with the voltage-pulse protocol shown above. (C) Normalized tail current vs. depolarization voltage plot of WT Eag1 (black square, n = 6), F475I/Q477R (green triangle, n = 7), and Eag1TM F475I/Q477R (red diamond, n = 6) with V0.5 values (mean ± sd). (D) Top, Voltage family current trace for Eag1/hErg chimera F475I/Q477R with the voltage-pulse protocol shown above. Bottom, normalized current vs depolarization voltage for Eag1/hErg chimera F475I/Q477R (black square, n = 5) with reversal potential (Erev) (mean ± sd).

CaM inhibition

Rotation of the intracellular domains observed in the different conformations of Eag1 Δ3–13/CaM occurs with CaM bound to the channel in the same orientation as observed in the WT structure (Figure 6—figure supplement 4) (Whicher and MacKinnon, 2016). This finding suggests that binding of CaM does not clamp the PAS and CNBHD domains together to prevent rotation of the intracellular domains and channel opening as we previously proposed. Instead, we think that more likely CaM binding to the channel helps to stabilize the hydrophobic interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639. As discussed above, this interaction seems to be important for channel closure. Thus, by stabilizing this interaction CaM can inhibit the channel. In support of this hypothesis, deletion of this interaction by removing residues 3–13 results in a channel that is no longer inhibited by CaM (Figure 6a). In addition, a similar effect was observed in Eag1 mutants that lack the PAS cap (residues 1–26) or the entire PAS domain (27-135) (Lörinczi et al., 2016).

Discussion

In summary, the data presented here provide many insights into the mechanism of voltage-dependent gating and calmodulin inhibition in Eag1. First, the intracellular domains are not required for voltage-dependent gating but do modulate voltage-dependent gating kinetics and cause the Cole-Moore effect. Second, an interaction between Asp 342 at the intracellular side of the S4 helix and the intracellular domains is essential for the Cole-Moore effect. Through mutagenesis in full length and split channels, we identified Arg 7 and 8 in the PAS loop N-terminus as potential interaction partners for Asp 342. However, mutation of Arg 7 and 8 results in modification but not complete loss of the Cole-Moore effect, suggesting that Asp 342 may interact with other residues on the intracellular domains. Third, interaction between the PAS loop C-terminus, Tyr 213 of the VS, and Tyr 639 of the CNBHD plays an important role in gating of Eag1. Deletion of residues 3–13 from the PAS loop produces a channel that opens at more hyperpolarized potentials than WT Eag1 and has slow deactivation kinetics, suggesting that this interaction promotes channel closure. In addition, this interface is important for the mechanism of CaM inhibition as Eag1 Δ3–13 is constitutively open in the presence of Ca2+/CaM. Finally, the structure of Eag1 Δ3–13 bound to CaM in a pre-open conformation demonstrates that channel opening may occur through a rotation of the intracellular domains in a counterclockwise direction when viewed from the extracellular side of the membrane.

Based on the data, we propose the following general mechanism of modulation of voltage-dependent gating by the intracellular domains. In the depolarized conformation of the VS, Asp 342 interacts with the PAS loop N-terminus to stabilize the depolarized state of the VS as well as destabilize the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639 to promote channel opening (Figure 1c). In the hyperpolarized conformation of the VS, movement of the S4 disrupts the interaction between Asp 342 and the PAS loop N-terminus to allow for the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639 to form and promote channel closing (Figure 1b). The structure of Eag1 Δ3–13/CaM in a pre-open conformation suggests that channel opening may occur through a counterclockwise rotation of the intracellular domains and channel closing may occur through a clockwise rotation of the intracellular domains. This mechanism is consistent with previous functional data examining split Eag1 channels that do not have a covalent linkage between the VS and pore and provides an explanation for the Cole-Moore effect (Lörinczi et al., 2015; Tomczak et al., 2017). At more negative holding potentials the S4 and Asp 342 may have to transition through more closed states in order to interact with the PAS loop N-terminus, which will result in slower activation times. Furthermore, CaM seems to function through this proposed mechanism by stabilizing the interaction between residues 10–13, Tyr 213, and Tyr 639 to prevent pore opening. Perhaps by binding to the PAS domain, CaM may be able to stabilize the PAS loop in a closed conformation.

These interactions also help to understand the gating of related channels. For hErg, which does not exhibit a Cole-Moore effect, the PAS loop Arg residues (Arg 4 and 5) and the S4 Asp (Asp 540) function in slow deactivation (Morais Cabral et al., 1998; Muskett et al., 2011; Ng et al., 2011; Sanguinetti and Xu, 1999). How do equivalent residues in Eag1 and hErg function in different voltage-dependent gating outcomes? As discussed above, in Eag1 deletion of the intracellular domains or mutation of Asp 342 causes a positive shift in the V0.5, suggesting that the interaction between the PAS loop N-terminus and Asp 342 is important for channel activation and thus functions in the Cole-Moore effect. However, in hErg deletion of the intracellular domains and mutation of Asp 540 does not change the V0.5, suggesting that the interaction between the PAS N-terminus and Asp 540 is not important for channel activation (Hausammann and Grütter, 2013; Morais Cabral et al., 1998; Sanguinetti and Xu, 1999). Instead, we propose that in hErg, which we have shown apparently has a more stable open conformation than Eag1, the interaction between Asp 540 and the PAS loop forms after channel activation and prevents pore closing, which results in slow deactivation. Like Eag1, the interaction between Asp 540 and the PAS loop N-terminus in hErg may prevent pore closing by destabilizing the interaction between the PAS loop C-terminus, Tyr 403, and Tyr 827. In support of this idea, splitting the hErg channel, analogous to the Eag1 L341 split, so that Asp 540 is not covalently linked to the voltage sensor, produces a channel that displays slowed deactivation compared to WT hErg (de la Peña et al., 2018).

Materials and methods

Key resources table.

Reagent type
(species) or resource		 Designation Source or reference Identifiers Additional
information
Gene (Rattus norvegicus) Kv10.1/Eag1/Kcnh1 Synthetic Uniprot: Q63472
Gene (Homo sapiens) Calmodulin Synthetic Uniprot: P0DP24
Cell line (Homo sapiens) HEK293S GnTI- ATCC ATCC: CRL-3022
RRID:CVCL_A785
Cell line
(Spodopterafrugiperda)		 Sf9 ATCC ATCC: CRL-1711
RRID:CVCL_0549
Cell line (Cricetulus griseus) Chinese Hamster Ovary cells Sigma RRID: CVCL_0213
Recombinant DNA reagent pEG Bacmam doi: https://doi.org/10.1038/nprot.2014.173
Recombinant DNA reagent pGEM-T vector Promega Catalog number: A1360
Software, algorithm pClampfit 10.5 Molecular Devices RRID: SCR_011323
Software, algorithm MotionCor2 doi: 10.1038/nmeth.4193 RRID: SCR_016499 http://msg.ucsf.edu/em/software/motioncor2.html
Software, algorithm CTFFIND4 doi: 10.1016/j.jsb.2015.08.008 RRID: SCR_016732 http://grigoriefflab.janelia.org/ctffind4
Software, algorithm RELION-3 doi: 10.1016/j.jsb.2012.09.006 RRID: SCR_016274 https://www2.mrc-lmb.cam.ac.uk/relion/index.php?title=Main_Page
Ssoftware, algorithm ResMap doi:
10.1038/nmeth.2727 		http://resmap.sourceforge.net
Software, algorithm Coot doi:
10.1107/S0907444910007493 		RRID: SCR_014222 https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Software, algorithm Phenix doi:
10.1107/S0907444909052925 		RRID: SCR_014224 http://phenix-online.org/
Software, algorithm Pymol PyMOL Molecular Graphics
System, Schrödinger, LLC		 RRID: SCR_000305 http://www.pymol.org/
Software, algorithm UCSF Chimera UCSF Resource for
Biocomputing, Visualization,and Bioinformatics		 RRID: SCR_004097 http://plato.cgl.ucsf.edu/chimera/
Software, algorithm HOLE doi:10.1016/S0263-7855(97)00009-X http://www.holeprogram.org
Open in a new tab

Cloning of Eag1 constructs

Cloning of rat Eag1 into the BacMam (Goehring et al., 2014) expression vector with a C-terminal green fluorescent protein (GFP)-His6 tag was described previously (Whicher and MacKinnon, 2016). All constructs presented here are in the BacMam vector except for the L341 split constructs. For the Eag1TM construct, residues 197–481 (S1-S6) were fused to residues 887–962 (C-terminal tetramer assembly domain). For the Eag1/hErg1 chimera, residues 1–389 (PAS) and 670–1159 (C-linker, CNBHD, and C-terminal assembly domain) of hErg were fused to the N- and C-termini of Eag1 residues 197–481 (S1-S6), respectively. For the L341 split, the N-terminal half (1-341) and C-terminal half (342-963) were each cloned into a pGEM vector for oocyte expression. Mutagenesis and deletions were performed with standard protocols and constructs were confirmed by sequencing. Calmodulin (CaM) was cloned into a BacMam vector as described previously (Whicher and MacKinnon, 2016).

Electrophysiological recordings of Eag1 constructs in CHO cells

All recordings of Eag1 constructs in BacMam vectors were from Chinese hamster ovary (CHO) cells. CHO cells cultured in DMEM-F12 (Gibco) with 10% FBS were transfected with the Eag1 construct using the FuGENE HD transfection reagent (Promega). 48 hr following transfection, the media was replaced with bath solution (10 mM HEPES pH 7.4, 60 mM KCl, 95 mM NaCl, 1 mM CaCl2) and experiments were performed at room temperature using the whole cell patch clamp technique. Polished borosilicate glass pipettes with resistance between 2–4 MΩ were filled with 10 mM HEPES pH 7.4, 165 mM KCl, 5 mM EDTA. To record Eag1 Δ3–13 in the presence of calcium, the 5 mM EDTA was replaced with 1 mM CaCl2 in the pipette solution. Voltage-family recordings were measured by holding the cells at −100 mV, stepping to depolarized voltages up to 100 mV in 20 mV steps, and then stepping back to −80 mV. To determine the V0.5 value, normalized tail current vs. voltage was plotted and fit with a Boltzmann function. Cole-Moore effect recordings were measured by holding cells for 500 ms at increasing holding potentials from −190 mV to the voltage of channel activation (either −110 mV, −50 mV or −30 mV depending on the construct) in 20 mV steps followed by a step to 40 mV. To estimate the holding potential that produces half maximal rates of activation (V0.5CM), we plotted holding potential vs normalized current at 10 ms following the depolarization step and fit the plot with a Boltzmann function:

(I−Imin)/(Imax−Imin)=1/1+exp(−ZF/RT(V−V0.5cm))

where (I-Imin )/(Imax -Imin ) is the normalized current at 10 ms following the depolarization step, V is the hyperpolarization voltage preceding the depolarization step, V0.5cm is the hyperpolarization voltage that produces half maximal rates of activation, F is the Faraday’s constant, R is the gas constant, T is the absolute temperature, and Z is the apparent valence of voltage dependence.

To determine reversal potential, we plotted normalized outward current vs depolarization voltage and determined the X intercept. All recordings were measured with pClamp10.5 software (Molecular Devices), an Axopatch 200B amplifier (Molecular Devices), and an Axon digidata 1550 digitizer (Molecular Devices). Recordings were filtered at 1 kHz and sampled at 10 kHz. No leak current was subtracted from the current traces.

Electrophysiological recordings of Eag1 L341 split constructs in oocytes

The mMessage mMachine T7 transcription kit (Ambion) was used to produce cRNA of the Eag1 split constructs linearized with NdeI. The MEGAclear kit was used to purify cRNAs, which were injected into oocytes. A total of 10 ng of cRNA was injected per oocyte at a ratio of 1:1 N-terminal half:C-terminal half. Oocytes were stored at 18°C for 24–48 hr after injection in ND96 (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, 1.0 mM MgCl2, 5 mM HEPES pH 7.6 with NaOH, 50 μg/ml gentamycin) and used for recordings. The bath solution was 55 mM NaCl, 60 mM KCl, 1.8 mM CaCl2, and 10 mM HEPES pH 7.2 with NaOH and the pipette solution was 3M KCl. The voltage family protocol was as follows: hold at −20 mV, step to depolarized voltages from −120 to 100 mV in 20 mV steps, and then step back to −80 mV. All recordings were measured at room temperature with pClamp10.5 software (Molecular Devices), Gene Clamp 500 amplifier (Molecular Devices), and an Axon digidata 1440A digitizer (Molecular Devices) in two electrode voltage-clamp configuration. The recorded signal was filtered at 1 kHz and sampled at 10 kHz. No leak or capacitive currents were subtracted from the current traces.

Expression and purification of Eag1 Δ3–13/CaM

The C-terminal unstructured region (773-886) of the Eag1 Δ3–13 was removed as described previously to improve expression and stability. This mutation does not affect the functional properties of the channel (Whicher and MacKinnon, 2016). Baculovirus for Eag1 Δ3–13 and CaM, were produced by transfecting bacmids into SF9 cells in Grace’s media supplemented with 10% FBS with the cellfectin II reagent (Invitrogen). Then the baculovirus was amplified in 1L suspension cultures of SF9 cells at 27°C. 1L cultures of HEK293S GnTI- at 3 × 106 cells/mL in Freestyle 293 media (Gibco) supplemented with 2% FBS were infected with both Eag1 Δ3–13 and CaM baculovirus at a 4:1 Eag1 Δ3–13:CaM ratio. Following infection, the cells were incubated at 37°C for 18 hr, induced by adding 10 μM sodium butyrate, incubated at 30°C for 48 hr, and harvested.

4L of cell pellet was resuspended in lysis buffer (20 mM Tris pH 8, 1 mM CaCl2, 1 μg/ml leupepetin, 1 μg/ml pepstatin, 1 mM benzamidine, 1 μg/ml aprotonin, 0.01 mg/ml DNase, 1 mM PMSF), incubated at RT with stirring for 20 min, and centrifuged for 40 min at 35,000xg. Pellets were resuspended in extraction buffer (50 mM Tris pH 8, 300 mM KCl, 1 mM CaCl2, 8 mM Lauryl Maltose Neopentyl Glycol (LMNG), 2 mM Cholesteryl hemisuccinate (CHS), 1 μg/ml leupepetin, 1 μg/ml pepstatin, 1 mM benzamidine, 1 μg/ml aprotonin, 0.01 mg/ml DNase, 1 mM PMSF), incubated at 4°C for 2 hr with stirring, and centrifuged for 90 min at 35,000xg. The supernatant was incubated for 2 hr at 4°C with CNBR-activated sepharose beads (GE healthcare) coupled to a nanobody with high affinity for GFP (GFP-NB) (Kirchhofer et al., 2010). The beads were washed with superose 6 buffer (20 mM Tris pH 8, 300 mM KCl, 1 mM CaCl2, 0.05% Digitonin) first with and then without 10 mM MgCl2 and 5 mM adenosine triphosphate (ATP) to remove bound heat shock proteins. The washed beads were incubated overnight at 4°C with PreScission protease (10:1 w/w ratio) to remove the GFP tag from Eag1 Δ3–13. The protein was eluted with wash buffer, concentrated, and purified on a superose 6 column (GE healthcare) equilibrated with superose 6 buffer. Peak fractions of Eag1 Δ3–13 bound to CaM (Eag1 Δ3–13/CaM) (Figure 6—figure supplement 1a) were pooled and concentrated to 5 mg/ml for single particle Cryo-EM structure determination.

EM sample preparation and imaging of Eag1 Δ3–13/CaM

In a Vitrobot Mark IV (FEI), 3.5 μl of 5 mg/ml Eag1 Δ3–13/CaM was pipetted onto Quantifoil R1.2/1.3 gold holey carbon grids (Quantifoil) with 400 mesh that were glow-discharged for 10 s. The grids were blotted for 4 s at 100% humidity and frozen in liquid nitrogen cooled liquid ethane. Images were collected in a 300keV Titan Krios (FEI) with a Gatan K2 Summit direct electron detector (Gatan) with Serial EM (Mastronarde, 2005) in super-resolution counting mode, with a super resolution pixel size of 0.5 Å, and a defocus range of 1.2 to 2.4 μm. Data were collected with a dose of 8 electrons per physical pixel per second (pixel size of 1.0 Å at the specimen) and images were recorded with a 10 s exposure and 200 ms subframes (50 total frames) to give a total dose of 80 electrons per Å2 (1.6 electrons per Å2 per subframe).

Image processing and map generation

Dose fractionated subframes were binned by 2 (giving a pixel size of 1.0 Å), aligned, and summed using MotionCor2 (Zheng et al., 2017) with 5 × 5 patches (Figure 6—figure supplement 1b). The contrast transfer function was estimated for each summed image using CTFFIND4 (Rohou and Grigorieff, 2015). Three projection averages from the previous structure of Eag1 bound to CaM (Whicher and MacKinnon, 2016) were used as templates for automated picking in RELION (Scheres, 2012). The automatically selected particles were manually inspected to remove false positives and subjected to 2D classification in RELION specifying 200 classes (Figure 6—figure supplement 1c). The lowest populated classes were removed resulting in a data set of 378,000 particles. 3D classification of this data set, with Eag1/CaM as a reference (Whicher and MacKinnon, 2016), resulted in five classes with similar numbers of particles and resolution. Therefore, all 378,000 particles were combined for 3D refinement, with C4 symmetry imposed, producing a map at 4.5 Å resolution estimated by gold standard FSC at the 0.143 cutoff criteria (Scheres and Chen, 2012). The refined particles were subjected to further rounds of 3D classification without image alignment, which produced 2 subsets of particles: conformation 1 and conformation 2. Conformation 1 of Eag1 Δ3–13/CaM has 43,137 particles and a similar overall structure to Eag1/CaM (Whicher and MacKinnon, 2016). Conformation 2 of Eag1 Δ3–13/CaM has 54,530 particles and the intracellular domains are rotated with respect to the transmembrane domains. Bayesian particle polishing and 3D refinement, with C4 symmetry imposed, of the particle subsets in RELION resulted in 3.67 Å for conformation 1 and 4 Å for conformation 2. Gold standard FSC curves were calculated with a mask that excludes the detergent micelle and resolution values were estimated with the FSC = 0.143 cutoff criteria (Figure 6—figure supplement 1d–f) (Scheres and Chen, 2012). Local resolutions were estimated by ResMap (Figure 6—figure supplement 2) (Kucukelbir et al., 2014).

Model building

The models of conformation 1 and conformation 2 were built in Coot (Emsley et al., 2010). For both conformations, first the S1-S6 and then the intracellular domains (PAS, C-linker, CNBHD, and CaM) from the structure of Eag1/CaM (Whicher and MacKinnon, 2016) (pdb-5K7L) were placed into the density as a rigid body. Following rigid body fitting, the model was manually inspected to fix regions that did not agree with the map or delete regions where there was no density. In conformation 1, we did not observe density for residues 244–246 (S1-S2 linker), 305–322 (S3-S4 linker), 407–411, 697–703, and 721-C-terminus. In conformation 2, we did not observe density for residues 202–213, 243–246 (S1-S2 linker), 274–283 (S2-S3 linker), 305–323 (S3-S4 linker), 407–411, 697–705, and 721-C-terminus. The side chains were modeled as alanine in lower resolution regions. Phenix real space refinement was used to refine the tetramer model of conformation 1 and conformation 2. Final models were validated using MolProbity and by comparing FSCs between the refined model and the EM map (Figure 6—figure supplement 3). Figures were generated with Chimera (Pettersen et al., 2004), Pymol (The PyMOL Molecular Graphics System, Version 1.8 Schrödinger, LLC.), HOLE (Smart et al., 1996), and structure calculations were performed with the SBgrid suite of programs (Morin et al., 2013).

Acknowledgements

We thank Mark Ebrahim and Johanna Sotiris at the Rockefeller University Cryo-EM resource center for help with data collection and Jue Chen and members of the MacKinnon laboratory for helpful discussions. This work was supported in part by National Institutes of Health grant GM43949. JRW is a Damon Runyon Fellow supported by the Damon Runyon Cancer Research Foundation (DRG-2212–15) and RM is an investigator in the Howard Hughes Medical Institute.

The low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 Δ3–13/CaM conformation 1 (accession code: EMD-20295) and conformation 2 (accession code: EMD-20294) have been deposited in the electron microscopy data bank. Atomic coordinates for Eag1 Δ3–13/CaM conformation 1 (accession code: 6PBY) and conformation 2 (accession code: 6PBX) have been deposited in the protein data bank.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Roderick MacKinnon, Email: mackinn@mail.rockefeller.edu.

Gary Yellen, Harvard Medical School, United States.

Olga Boudker, Weill Cornell Medicine, United States.

Funding Information

This paper was supported by the following grants:

  • Damon Runyon Cancer Research Foundation DRG-2212-15 to Jonathan R Whicher.

  • National Institutes of Health GM43949 to Roderick MacKinnon.

  • Howard Hughes Medical Institute to Roderick MacKinnon.

Additional information

Competing interests

No competing interests declared.

Author contributions

Conceptualization, Formal analysis, Investigation, Methodology, Writing—original draft.

Conceptualization, Supervision, Funding acquisition, Writing—review and editing.

Additional files

Transparent reporting form
elife-49188-transrepform.docx (244.8KB, docx)
DOI: 10.7554/eLife.49188.015

Data availability

The low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 1 (accession code: EMD-20295) and conformation 2 (accession code: EMD-20294) have been deposited in the electron microscopy data bank. Atomic coordinates for Eag1 3-13/CaM conformation 1 (accession code: 6PBY) and conformation 2 (accession code: 6PBX) have been deposited in the protein data bank.

The following datasets were generated:

Whicher JR, MacKinnon R. 2019. Low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 1. Electron Microscopy Data Bank. EMD-20295

Whicher JR, MacKinnon R. 2019. Low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 2. Electron Microscopy Data Bank. EMD-20294

Whicher JR, MacKinnon R. 2019. Atomic coordinates for Eag1 3-13/CaM conformation 1. Protein Data Bank. 6PBY

Whicher JR, MacKinnon R. 2019. Atomic coordinates for Eag1 3-13/CaM conformation 2. Protein Data Bank. 6PBX

References

  1. Aggarwal SK, MacKinnon R. Contribution of the S4 segment to gating charge in the shaker K+ channel. Neuron. 1996;16:1169–1177. doi: 10.1016/S0896-6273(00)80143-9. [DOI] [PubMed] [Google Scholar]
  2. Bauer CK, Schwarz JR. Ether-à-go-go K+ channels: effective modulators of neuronal excitability. The Journal of Physiology. 2018;596:769–783. doi: 10.1113/JP275477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Cole KS, Moore JW. Potassium ion current in the squid giant axon: dynamic characteristic. Biophysical Journal. 1960;1:1–14. doi: 10.1016/s0006-3495(60)86871-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. de la Peña P, Domínguez P, Barros F. Gating mechanism of Kv11.1 (hERG) K+ channels without covalent connection between voltage sensor and pore domains. Pflügers Archiv - European Journal of Physiology. 2018;470:517–536. doi: 10.1007/s00424-017-2093-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and development of coot. Acta Crystallographica. Section D, Biological Crystallography. 2010;66:486–501. doi: 10.1107/S0907444910007493. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Gianulis EC, Liu Q, Trudeau MC. Direct interaction of eag domains and cyclic nucleotide-binding homology domains regulate deactivation gating in hERG channels. The Journal of General Physiology. 2013;142:351–366. doi: 10.1085/jgp.201310995. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Goehring A, Lee CH, Wang KH, Michel JC, Claxton DP, Baconguis I, Althoff T, Fischer S, Garcia KC, Gouaux E. Screening and large-scale expression of membrane proteins in mammalian cells for structural studies. Nature Protocols. 2014;9:2574–2585. doi: 10.1038/nprot.2014.173. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Hausammann GJ, Grütter MG. Chimeric hERG channels containing a tetramerization domain are functional and stable. Biochemistry. 2013;52:9237–9245. doi: 10.1021/bi401100a. [DOI] [PubMed] [Google Scholar]
  9. Jiang Y, Lee A, Chen J, Ruta V, Cadene M, Chait BT, MacKinnon R. X-ray structure of a voltage-dependent K+ channel. Nature. 2003;423:33–41. doi: 10.1038/nature01580. [DOI] [PubMed] [Google Scholar]
  10. Ju M, Wray D. Molecular regions responsible for differences in activation between heag channels. Biochemical and Biophysical Research Communications. 2006;342:1088–1097. doi: 10.1016/j.bbrc.2006.02.062. [DOI] [PubMed] [Google Scholar]
  11. Kirchhofer A, Helma J, Schmidthals K, Frauer C, Cui S, Karcher A, Pellis M, Muyldermans S, Casas-Delucchi CS, Cardoso MC, Leonhardt H, Hopfner KP, Rothbauer U. Modulation of protein properties in living cells using nanobodies. Nature Structural & Molecular Biology. 2010;17:133–138. doi: 10.1038/nsmb.1727. [DOI] [PubMed] [Google Scholar]
  12. Kucukelbir A, Sigworth FJ, Tagare HD. Quantifying the local resolution of cryo-EM density maps. Nature Methods. 2014;11:63–65. doi: 10.1038/nmeth.2727. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Long SB, Campbell EB, Mackinnon R. Crystal structure of a mammalian voltage-dependent shaker family K+ channel. Science. 2005a;309:897–903. doi: 10.1126/science.1116269. [DOI] [PubMed] [Google Scholar]
  14. Long SB, Campbell EB, Mackinnon R. Voltage sensor of Kv1.2: structural basis of electromechanical coupling. Science. 2005b;309:903–908. doi: 10.1126/science.1116270. [DOI] [PubMed] [Google Scholar]
  15. Long SB, Tao X, Campbell EB, MacKinnon R. Atomic structure of a voltage-dependent K+ channel in a lipid membrane-like environment. Nature. 2007;450:376–382. doi: 10.1038/nature06265. [DOI] [PubMed] [Google Scholar]
  16. Lörinczi É, Gómez-Posada JC, de la Peña P, Tomczak AP, Fernández-Trillo J, Leipscher U, Stühmer W, Barros F, Pardo LA. Voltage-dependent gating of KCNH potassium channels lacking a covalent link between voltage-sensing and pore domains. Nature Communications. 2015;6:6672. doi: 10.1038/ncomms7672. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Lörinczi E, Helliwell M, Finch A, Stansfeld PJ, Davies NW, Mahaut-Smith M, Muskett FW, Mitcheson JS. Calmodulin regulates human ether à Go-Go 1 (hEAG1) Potassium channels through interactions of the eag domain with the cyclic nucleotide binding homology domain. Journal of Biological Chemistry. 2016;291:17907–17918. doi: 10.1074/jbc.M116.733576. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Ludwig J, Terlau H, Wunder F, Brüggemann A, Pardo LA, Marquardt A, Stühmer W, Pongs O. Functional expression of a rat homologue of the voltage gated either á go-go potassium channel reveals differences in selectivity and activation kinetics between the Drosophila channel and its mammalian counterpart. The EMBO Journal. 1994;13:4451–4458. doi: 10.1002/j.1460-2075.1994.tb06767.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ludwig J, Owen D, Pongs O. Carboxy-terminal domain mediates assembly of the voltage-gated rat ether-à-go-go potassium channel. The EMBO Journal. 1997;16:6337–6345. doi: 10.1093/emboj/16.21.6337. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Mastronarde DN. Automated electron microscope tomography using robust prediction of specimen movements. Journal of Structural Biology. 2005;152:36–51. doi: 10.1016/j.jsb.2005.07.007. [DOI] [PubMed] [Google Scholar]
  21. Morais Cabral JH, Lee A, Cohen SL, Chait BT, Li M, Mackinnon R. Crystal structure and functional analysis of the HERG potassium channel N terminus: a eukaryotic PAS domain. Cell. 1998;95:649–655. doi: 10.1016/S0092-8674(00)81635-9. [DOI] [PubMed] [Google Scholar]
  22. Morin A, Eisenbraun B, Key J, Sanschagrin PC, Timony MA, Ottaviano M, Sliz P. Collaboration gets the most out of software. eLife. 2013;2:e01456. doi: 10.7554/eLife.01456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Muskett FW, Thouta S, Thomson SJ, Bowen A, Stansfeld PJ, Mitcheson JS. Mechanistic insight into human ether-à-go-go-related gene (hERG) K+ channel deactivation gating from the solution structure of the EAG domain. The Journal of Biological Chemistry. 2011;286:6184–6191. doi: 10.1074/jbc.M110.199364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Ng CA, Hunter MJ, Perry MD, Mobli M, Ke Y, Kuchel PW, King GF, Stock D, Vandenberg JI. The N-terminal tail of hERG contains an amphipathic α-helix that regulates channel deactivation. PLOS ONE. 2011;6:e16191. doi: 10.1371/journal.pone.0016191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE. UCSF chimera--a visualization system for exploratory research and analysis. Journal of Computational Chemistry. 2004;25:1605–1612. doi: 10.1002/jcc.20084. [DOI] [PubMed] [Google Scholar]
  26. Rohou A, Grigorieff N. CTFFIND4: fast and accurate defocus estimation from electron micrographs. Journal of Structural Biology. 2015;192:216–221. doi: 10.1016/j.jsb.2015.08.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Sanguinetti MC, Xu QP. Mutations of the S4-S5 Linker alter activation properties of HERG potassium channels expressed in xenopus oocytes. The Journal of Physiology. 1999;514 ( Pt 3:667–675. doi: 10.1111/j.1469-7793.1999.667ad.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Scheres SH. A bayesian view on cryo-EM structure determination. Journal of Molecular Biology. 2012;415:406–418. doi: 10.1016/j.jmb.2011.11.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Scheres SH, Chen S. Prevention of overfitting in cryo-EM structure determination. Nature Methods. 2012;9:853–854. doi: 10.1038/nmeth.2115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Schönherr R, Löber K, Heinemann SH. Inhibition of human ether à go-go potassium channels by Ca(2+)/calmodulin. The EMBO Journal. 2000;19:3263–3271. doi: 10.1093/emboj/19.13.3263. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Seoh SA, Sigg D, Papazian DM, Bezanilla F. Voltage-sensing residues in the S2 and S4 segments of the shaker K+ channel. Neuron. 1996;16:1159–1167. doi: 10.1016/S0896-6273(00)80142-7. [DOI] [PubMed] [Google Scholar]
  32. Smart OS, Neduvelil JG, Wang X, Wallace BA, Sansom MSP. HOLE: a program for the analysis of the pore dimensions of ion channel structural models. Journal of Molecular Graphics. 1996;14:354–360. doi: 10.1016/S0263-7855(97)00009-X. [DOI] [PubMed] [Google Scholar]
  33. Sun J, MacKinnon R. Cryo-EM structure of a KCNQ1/CaM complex reveals insights into congenital long QT syndrome. Cell. 2017;169:1042–1050. doi: 10.1016/j.cell.2017.05.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Terlau H, Heinemann SH, Stühmer W, Pongs O, Ludwig J. Amino terminal-dependent gating of the potassium channel rat eag is compensated by a mutation in the S4 segment. The Journal of Physiology. 1997;502 ( Pt 3:537–543. doi: 10.1111/j.1469-7793.1997.537bj.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Tomczak AP, Fernández-Trillo J, Bharill S, Papp F, Panyi G, Stühmer W, Isacoff EY, Pardo LA. A new mechanism of voltage-dependent gating exposed by KV10.1 channels interrupted between voltage sensor and pore. The Journal of General Physiology. 2017;149:577–593. doi: 10.1085/jgp.201611742. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Wang J, Trudeau MC, Zappia AM, Robertson GA. Regulation of deactivation by an amino terminal domain in human ether-à-go-go-related gene potassium channels. The Journal of General Physiology. 1998;112:637–647. doi: 10.1085/jgp.112.5.637. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Wang W, MacKinnon R. Cryo-EM structure of the open human Ether-à-go-go-Related K+ Channel hERG. Cell. 2017;169:422–430. doi: 10.1016/j.cell.2017.03.048. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Whicher JR, MacKinnon R. Structure of the voltage-gated K⁺ channel Eag1 reveals an alternative voltage sensing mechanism. Science. 2016;353:664–669. doi: 10.1126/science.aaf8070. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Zheng SQ, Palovcak E, Armache JP, Verba KA, Cheng Y, Agard DA. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nature Methods. 2017;14:331–332. doi: 10.1038/nmeth.4193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Ziechner U, Schönherr R, Born AK, Gavrilova-Ruch O, Glaser RW, Malesevic M, Küllertz G, Heinemann SH. Inhibition of human ether à go-go potassium channels by Ca2+/calmodulin binding to the cytosolic N- and C-termini. FEBS Journal. 2006;273:1074–1086. doi: 10.1111/j.1742-4658.2006.05134.x. [DOI] [PubMed] [Google Scholar]

Decision letter

Editor: Gary Yellen1
Reviewed by: Gary Yellen2, Kenton Jon Swartz3

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Regulation of Eag1 gating by its intracellular domains" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Gary Yellen as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Olga Boudker as the Senior Editor. The following individual involved in review of your submission has also agreed to reveal his identity: Kenton J Swartz (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This paper addresses the molecular mechanisms by which the voltage gating of the Eag1 (Kv10.1) channel and other Kv10-12 family members are regulated by their intracellular PAS, C-linker, and CNBHD domains. These channels have a different architecture from the Kv1-9 channels, which couple voltage sensing to gating using an apparently mechanical linkage (the S4-S5 linker). The authors show that the intracellular domains of Eag1 are not required for voltage gating (contrary to their earlier proposal) but that they both substantially modulate the open-closed bias of the channels and can produce the classic "Cole-Moore" effect (a pronounced delay in channel opening after strong hyperpolarization). This extends earlier functional exploration by Pardo and colleagues (2015) and Tomczak et al., 2017, and others, as well as previous structural work by the present authors. The work also addresses the interesting structural and functional analogy between the Cole-Moore effect (delayed opening after strong hyperpolarization) of the Kv10 channels and the slowed deactivation of hErg (Kv11.1) channels.

Essential revisions:

No new experiments are requested, but revisions to the text and figures are important for making the complex arguments here more transparent to the readers, for improving the precision of the explanations, and for providing access to relevant previous studies that bear on the conclusions here.

The paper should include at least one and probably two additional display elements to help explain the proposed model. The Introduction discusses many interesting structural elements in EAG and would greatly benefit from adding a new Figure 1 (or an additional panel) to help the reader see these features. The existing Figure 2 doesn't really suffice for this, and it also doesn't suffice to show all of the residues mutated or deleted. We would suggest a figure that shows everything in detail that is relevant to this study, including residues in the S4-S5 linker or loop. It would also be helpful to include a figure or table that explains the different general gating states essential to thinking about the authors' proposal (something similar to stably shut, pre-open, open, and stably open) and which interactions are thought to occur (for instance PAS-C/Y213/Y639, PAS-N/D342, clockwise or counterclockwise rotation), for both the Eag1 and hErg channels. This would help readers follow the proposal, and it might also help explain more precisely the analogy between the Cole-Moore shift in Eag1 and the slow closure of hErg.

The explanation (in the Discussion, third paragraph) for the slow deactivation of hErg is confusing. The idea is vaguely apparent, but the statement that the slow deactivation in hErg (which arises from open-state stabilization) "arises from the same mechanism as the Cole-Moore effect" (which is a pronounced closed-state stabilization) is not helpful. For the hErg channels, the N-PAS – D342 interaction stabilizes the open state (kinetically if not thermodynamically); for the Eag1 channels, it stabilizes the open state thermodynamically and speeds entry into it. In hErg channels, this produces slow deactivation; in Eag1 channels, the disappearance of this interaction at extreme negative voltages produces the Cole-Moore shift, a kinetic delay in subsequent opening that appears due to the slowness of the re-establishment of this interaction. However, as the authors point out, a simple stabilization in hErg channels does not quite fit, because there is no shift in activation midpoint. Instead, the interaction "prevent<s> pore closing" (but through a kinetic effect). The parallels between the Cole-Moore shift and the slow deactivation are interesting, but the authors should state their case more precisely here.

The authors propose a hypothesis for why the Eag1 delta3-13/CaM structures (the new structures here) remain closed – the Eag1 channels are more biased to closing than the hErg channels [p 10 "A hypothesis for why the pore remains closed in the Eag1 Δ3-13/CaM structure is that, compared to hErg, Eag1 may be more stable in a closed conformation."]. But the fact remains that this exact construct is constitutively open when studied functionally (Figure 6A). So the discrepancy cannot really be explained as a difference between family members – it is clearly a difference between the bilayer and the cryoEM conditions.

In general, please consider alternatives to the proposed model, and explain why you think that your model is better. For instance, the proposed mechanism has the interaction between D342 and the PAS loop as promoting opening by destabilizing closed state interactions between Y213/Y639 and the PAS loop. Instead, or in addition, could the D342-PAS interaction directly stabilize the open state? Or the activated state of the voltage sensor?

Please also address the following questions in the revision, to help clarify the model and the interpretation of the data:

1) If the primary effect of the intracellular domains is to inhibit opening, as suggested in the Abstract ("…provide evidence that VS movement destabilizes these interactions to promote channel opening") and other places in the paper, then why does deletion of intracellular domains, mutation of D342, mutations R7A/R8A, and Δ3-9, destabilize opening (shift the voltage-dependence to the right)?

2) Since there are no estimates of the energetics of the interactions in the structure or the energetic effects of the mutations, the paper should be more cautious about attributing the interactions proposed to the total effects of the intracellular domains.

3) Abstract: "The structure of the calmodulin insensitive mutant suggests that rotation of the intracellular domains promotes channel opening." This sentence is misleading. It should be indicated in the Abstract that the structure presented here is closed, like the original structure of Eag1. It is not clear how a closed structure tells us anything about promoting opening since Ca-CAM doesn't promote opening in Eag1Δ3-13 (does it?). Also, while it seems plausible that the counterclockwise rotation of the intracellular domains is helpful for opening, do we know that it is necessary (it is clearly not sufficient)?

4) Introduction section: "Due to the non-domain swapped transmembrane architecture, the S4-S5 linker is not in the correct position and does not have the required structure to function as a mechanical lever, suggesting an alternative mechanism of voltage-dependent gating in Kvs 10-12." This sentence should be supported by citing previous split experiments.

5) Subsection “Interactions between the voltage sensor and intracellular domains”: "…we first investigated the S4-S5 linker.…by alanine scanning mutagenesis " Wasn't alanine scanning and cysteine scanning mutagenesis previously done on the S4-S5 linker of hErg channels (Gianulis et al., 2013, and Wang et al., 1998 respectively)? This work should be cited.

6) Subsection “Insights into voltage-dependent gating”: "Furthermore, in the open state structure of hErg the PAS loop and the CNBHD Tyr do not interact (Figure 2B)…" What about interaction of the PAS loop and the 213 equivalent residue in the hErg structure?

7) Subsection “Insights into voltage-dependent gating” and elsewhere "…this mechanism provides an explanation for the Cole-Moore effect because at more negative holding potentials the S4 and Asp 342 will be driven further down and away from the PAS loop N-terminus, which will result in slower activation times." Further down from what? Are you suggesting that there is a down state of the VS that is further down than the normal resting state? Wouldn't any movement down prevent interactions of the N-terminal end of the PAS loop with D342?

8) Subsection “Structure of constitutively open Eag1”: "… this opening is transient and thus not observed in the Eag1 Δ 3-13/CaM structure." This is the first mention I found of the previous finding that PAS loop deletions in Eag cause inactivation. This should be mentioned earlier and properly cited.

9) Could the authors discuss why they think the new structures are not inactivated? Could they comment on the position of the S4 helix? Does it appear to be activated as suggested by the earlier CaM-inhibited structure?

10) The IV plots in Figure 5D, 6A and 7D look rather odd because they reverse near -20 mV (the holding voltage) rather than at a voltage near EK+. Oocytes normally have internal K+ set to 100K+, and according to the Materials and methods, these recording were obtained using an external solution with only 2.5 mM K+, which puts EK near to -100 mV. The simplest explanation I can think of is that these plots were obtained with much higher external K+, perhaps around 50-60 mM, but again the Materials and methods state it is 2.5 mM. Might the authors have plotted the difference between current measured at the holding voltage of -20 mV and each test potential, in some cases then normalized to that difference measured at +100 mV? But this is rather unconventional, and nothing is stated about how this was done. If -20 mV really is the Vrev for these constructs and external K- is 2.5 mM, then either these constructs are no longer K-selective, or the internal concentration of K is way off due to expressing these constructs. Please explain and fix as appropriate. In addition, are the oocytes currents being manipulated in any way? The capacitive currents seem to be largely compensated or subtracted, but nothing is stated in the Materials and methods. Are all the whole-cell CHO currents are not leak subtracted? It would be good to state this specifically. It would also be good to add dashed lines to all traces to indicate the zero-current level. Finally, in the traces shown in Figure 6A, it appears you have lost voltage-clamp in the last few traces. Perhaps replace with another recording or remove those traces?

eLife. 2019 Sep 6;8:e49188. doi: 10.7554/eLife.49188.026

Author response

Essential revisions:

No new experiments are requested, but revisions to the text and figures are important for making the complex arguments here more transparent to the readers, for improving the precision of the explanations, and for providing access to relevant previous studies that bear on the conclusions here.

The paper should include at least one and probably two additional display elements to help explain the proposed model. The Introduction discusses many interesting structural elements in EAG and would greatly benefit from adding a new Figure 1 (or an additional panel) to help the reader see these features. The existing Figure 2 doesn't really suffice for this, and it also doesn't suffice to show all of the residues mutated or deleted. We would suggest a figure that shows everything in detail that is relevant to this study, including residues in the S4-S5 linker or loop. It would also be helpful to include a figure or table that explains the different general gating states essential to thinking about the authors' proposal (something similar to stably shut, pre-open, open, and stably open) and which interactions are thought to occur (for instance PAS-C/Y213/Y639, PAS-N/D342, clockwise or counterclockwise rotation), for both the Eag1 and hErg channels. This would help readers follow the proposal, and it might also help explain more precisely the analogy between the Cole-Moore shift in Eag1 and the slow closure of hErg.

A new Figure 1 was added to the manuscript. This figure displays 2 views of the entire Eag1 channel (panel A) and views of the interactions between the PAS loop, VS, S4-S5 linker, C-linker, and CNBHD in the closed state of Eag1 (panel B) and the open state of hErg (panel C). All Eag1 residues that were mutated in this manuscript are shown in panel B.

Figure 1 panels B and C show the interactions that we propose occur in the open (panel C) and closed (panel B) states of Eag1. Instead of creating a new figure or table these figures were referenced when describing the mechanism.

The explanation (in the Discussion, third paragraph) for the slow deactivation of hErg is confusing. The idea is vaguely apparent, but the statement that the slow deactivation in hErg (which arises from open-state stabilization) "arises from the same mechanism as the Cole-Moore effect" (which is a pronounced closed-state stabilization) is not helpful. For the hErg channels, the N-PAS – D342 interaction stabilizes the open state (kinetically if not thermodynamically); for the Eag1 channels, it stabilizes the open state thermodynamically and speeds entry into it. In hErg channels, this produces slow deactivation; in Eag1 channels, the disappearance of this interaction at extreme negative voltages produces the Cole-Moore shift, a kinetic delay in subsequent opening that appears due to the slowness of the re-establishment of this interaction. However, as the authors point out, a simple stabilization in hErg channels does not quite fit, because there is no shift in activation midpoint. Instead, the interaction "prevent<s> pore closing" (but through a kinetic effect). The parallels between the Cole-Moore shift and the slow deactivation are interesting, but the authors should state their case more precisely here.

We agree that the statement that the slow deactivation in hErg "arises from the same mechanism as the Cole-Moore effect" is confusing and it was removed. The data suggest that equivalent residues (D342, the PAS loop N-terminus, and the PAS loop C-terminus) in Eag1 and hErg function in different voltage- dependent gating outcomes. In Eag1 these residues function in the Cole-Moore effect and in hErg these residues function in slow deactivation. In Eag1, mutation of D342 and the PAS loop N-terminus results in a right-shifted V0.5 suggesting that these residues are important for channel activation. In hErg, mutation of D540 and the PAS loop N-terminus does not affect V0.5 suggesting that these residues are not important for channel activation. Therefore, we propose that in Eag1, D342 and the PAS loop N-terminus interact during channel activation to promote pore opening and thus function in the Cole-Moore effect. Whereas in hErg, D540 and the PAS loop N-terminus interact after channel activation to prevent pore closing and thus function in slow deactivation. In both cases, we believe the interaction between the S4 Asp (342 in Eag1 and 540 in hErg) and the PAS loop N-terminus destabilizes the interaction between the PAS loop C- terminus, the S1 Tyr (213 in Eag1 and 403 in Eag1), and the CNBHD Tyr (639 in Eag1 and 827 in hErg) and thus the closed state of the channel. The paragraph was edited to convey these points.

The authors propose a hypothesis for why the Eag1 delta3-13/CaM structures (the new structures here) remain closed – the Eag1 channels are more biased to closing than the hErg channels [p 10 "A hypothesis for why the pore remains closed in the Eag1 Δ3-13/CaM structure is that, compared to hErg, Eag1 may be more stable in a closed conformation."]. But the fact remains that this exact construct is constitutively open when studied functionally (Figure 6A). So the discrepancy cannot really be explained as a difference between family members – it is clearly a difference between the bilayer and the cryoEM conditions.

The conditions of structure determination, such as detergent solubilization and cryogenic freezing temperatures, can affect a proteins structure. However, the open conformation structure of hErg was determined under similar Cryo-EM conditions with DDM instead of Digitonin as the detergent. In addition, the data presented in this manuscript and previous manuscripts suggest that Eag1 is more stable in a closed conformation than hErg. Therefore, we believe that the closed pore of Eag1 Δ3-13/CaM is likely the result of a combination of Eag1 being more stable in a closed conformation than hErg and the structure determination conditions. This is now reflected in the paper.

“Therefore, we propose that the Eag1 intracellular domains, when viewed from the extracellular side, rotate in a counterclockwise direction to promote the opening of the pore. However, due to the stability of Eag1 in a closed conformation and the conditions under which the Cryo-EM structure was determined we suspect that pore opening is transient and thus not observed in the Eag1 Δ3-13/CaM structure.”

In general, please consider alternatives to the proposed model, and explain why you think that your model is better. For instance, the proposed mechanism has the interaction between D342 and the PAS loop as promoting opening by destabilizing closed state interactions between Y213/Y639 and the PAS loop. Instead, or in addition, could the D342-PAS interaction directly stabilize the open state? Or the activated state of the voltage sensor?

The large right shift in voltage-dependence for the Eag1 TM mutant suggests that the intracellular domains help stabilize the open state of the channel. Therefore, it is possible that the interaction between D342 and the PAS loop N-terminus may promote channel opening by stabilizing the depolarized state of the VS as well as destabilizing the interaction between Y213, Y639, and the PAS loop C- terminus. A few sentences were added to the Results and Discussion sections to account for this possibility.

“How might an interaction between the PAS loop and Asp 342 promote channel opening? One hypothesis is that the interaction between the PAS loop and Asp 342 stabilizes the depolarized state of the voltage sensor. This hypothesis would explain why deletion of the intracellular domains (Eag1TM) and mutation of Asp 342 and Arg 7 and 8 disfavors channel opening as indicated by a right shift in the voltage-dependence of activation.”

Please also address the following questions in the revision, to help clarify the model and the interpretation of the data:

1) If the primary effect of the intracellular domains is to inhibit opening, as suggested in the Abstract ("…provide evidence that VS movement destabilizes these interactions to promote channel opening") and other places in the paper, then why does deletion of intracellular domains, mutation of D342, mutations R7A/R8A, and Δ3-9, destabilize opening (shift the voltage-dependence to the right)?

We now explain that upon depolarization we propose that Asp 342 interacts with the PAS loop N-terminus to promote channel opening through the concomitant stabilization of the open state of the VS and destabilization of the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639 (discussed in the previous comment). Therefore, we would expect deletions of the intracellular domains and mutations to Asp 342 or the PAS loop N-terminus (residues 3-9) to destabilize channel opening.

2) Since there are no estimates of the energetics of the interactions in the structure or the energetic effects of the mutations, the paper should be more cautious about attributing the interactions proposed to the total effects of the intracellular domains.

To the point of the reviewers the description of the mechanism in the discussion was edited as shown:

“Based on the data, we proposed the following general mechanism of modulation of voltage-dependent gating by the intracellular domains. In the depolarized conformation of the VS, Asp 342 interacts with the PAS loop N-terminus to stabilize the depolarized state of the VS as well as destabilize the interaction between the PAS loop C-terminus, Tyr 213, and Tyr 639 to promote channel opening (Figure 1C). In the hyperpolarized conformation of the VS, movement of the S4 disrupts the interaction between Asp 342 and the PAS loop N-terminus to allow for the interaction between the PAS loop C- terminus, Tyr 213, and Tyr 639 to form and promote channel closing (Figure 1B). The structure of Eag1 Δ3-13/CaM in a pre-open conformation suggests that channel opening may occur through a counterclockwise rotation of the intracellular domains and channel closing may occur through a clockwise rotation of the intracellular domains”

3) Abstract: "The structure of the calmodulin insensitive mutant suggests that rotation of the intracellular domains promotes channel opening." This sentence is misleading. It should be indicated in the Abstract that the structure presented here is closed, like the original structure of Eag1. It is not clear how a closed structure tells us anything about promoting opening since Ca-CAM doesn't promote opening in Eag1Δ3-13 (does it?). Also, while it seems plausible that the counterclockwise rotation of the intracellular domains is helpful for opening, do we know that it is necessary (it is clearly not sufficient)?

In the presence of Ca2+/CaM Eag1 Δ3-13 is constitutively open. Therefore, the combination of the structure of Eag1 Δ3-13/CaM, which shows a rotation of the intracellular domains compared with the original Eag1 structure, and the open hErg structure strongly suggest that a rotation of the intracellular domains is important for channel opening. However, the sentence in the Abstract may mislead readers into thinking that the structure of Eag1 Δ3-13/CaM is in an open pore conformation. To avoid misleading readers, we have edited the paper to state that the structure of Eag1 Δ3-13/CaM represents a pre-open conformation of the channel. The Abstract was edited as shown:

“The structure of the calmodulin insensitive mutant in a pre-open conformation suggests that channel opening may occur through a rotation of the intracellular domains and calmodulin may prevent this rotation by stabilizing interactions between the VS and intracellular domains.”

4) Introduction section: "Due to the non-domain swapped transmembrane architecture, the S4-S5 linker is not in the correct position and does not have the required structure to function as a mechanical lever, suggesting an alternative mechanism of voltage-dependent gating in Kvs 10-12." This sentence should be supported by citing previous split experiments.

Both Eag1 split channel manuscripts (Lorinczi et al., 2016 and Tomczak et al., 2017) were cited.

5) Subsection “Interactions between the voltage sensor and intracellular domains”: "…we first investigated the S4-S5 linker.…by alanine scanning mutagenesis " Wasn't alanine scanning and cysteine scanning mutagenesis previously done on the S4-S5 linker of hErg channels (Gianulis et al., 2013, and Wang et al., 1998 respectively)? This work should be cited.

These citations were added to the manuscript.

6) Subsection “Insights into voltage-dependent gating”: "Furthermore, in the open state structure of hErg the PAS loop and the CNBHD Tyr do not interact (Figure 2B)…" What about interaction of the PAS loop and the 213 equivalent residue in the hErg structure?

In the open conformation of hErg, Tyr 403 (equivalent to Tyr 213 in Eag1) is near Arg 5 (equivalent to Arg 8 in Eag1) and no longer interacting with the PAS loop C-terminus (Figure 1C). This sentence was added to the manuscript:

“Furthermore, in the open state structure of hErg the PAS loop C-terminus does not interact with either Tyr 403 (equivalent to Tyr 213 in Eag1) or Tyr 827 (equivalent to Tyr 639 in Eag1) (Figure 1C) suggesting that destabilization of this interface might be necessary for channel opening”

7) Subsection “Insights into voltage-dependent gating” and elsewhere "…this mechanism provides an explanation for the Cole-Moore effect because at more negative holding potentials the S4 and Asp 342 will be driven further down and away from the PAS loop N-terminus, which will result in slower activation times." Further down from what? Are you suggesting that there is a down state of the VS that is further down than the normal resting state? Wouldn't any movement down prevent interactions of the N-terminal end of the PAS loop with D342?

The Cole-Moore effect was proposed to be due to the existence of multiple closed states that the VS transitions through in order to reach an active or depolarized conformation. At more negative potentials the VS transitions through more closed states to reach an active conformation, which results in slower activation times. So yes, we (and others before us) think that there are down states of the voltage sensor that are further down than the distribution of states associated with normal resting conditions. In Eag1 specifically, we propose that at more negative holding potentials the VS may have to transition through more closed states in order for the S4 and Asp 342 to interact with the PAS loop N-terminus, which will result in slower activation times. In the absence of a structure, we cannot accurately predict the position of the VS in a hyperpolarized conformation so the phrase “the S4 and Asp 342 will be driven further down and away from the PAS loop N-terminus” was removed from the manuscript. The explanation of the Cole-Moore effect in Eag1 was edited as follows:

“The Cole-Moore effect was proposed to be due to the existence of multiple closed states that the VS must transition through in order to reach an active or depolarized conformation (Cole and Moore, 1960). At more negative potentials the VS must transition through more closed states to reach an active conformation, which results in slower activation times. In Eag1, we propose that at more negative holding potentials the VS might have to transition through more closed states in order for the S4 and Asp 342 to interact with the PAS loop N-terminus, which will result in slower activation times.”

8) Subsection “Structure of constitutively open Eag1”: "… this opening is transient and thus not observed in the Eag1 Δ 3-13/CaM structure." This is the first mention I found of the previous finding that PAS loop deletions in Eag cause inactivation. This should be mentioned earlier and properly cited.

This is cited in subsection “Implications for voltage-dependent gating” when discussing PAS loop deletions.

9) Could the authors discuss why they think the new structures are not inactivated? Could they comment on the position of the S4 helix? Does it appear to be activated as suggested by the earlier CaM-inhibited structure?

The Eag1 Δ3-13 does not inactivate at 0 mV when bound to Ca2+/CaM (Figure 6A) so we would not expect the structure of Eag1 Δ3-13/CaM to be in an inactivated conformation.

The S4 helix is in depolarized conformation similar to the original structure of Eag1 bound to CaM. This was included in the manuscript:

“In both conformations, the S4 helices adopt a depolarized conformation and intracellular domains are rotated in a counterclockwise direction when viewed from the extracellular side of the membrane but in conformation 2 the extent of the rotation is larger (2.4° degrees for conformation 1 and 8.6° degrees for conformation 2) (Figure 6B,C).”

10) The IV plots in Figure 5D, 6A and 7D look rather odd because they reverse near -20 mV (the holding voltage) rather than at a voltage near EK+. Oocytes normally have internal K+ set to 100K+, and according to the Materials and methods, these recording were obtained using an external solution with only 2.5 mM K+, which puts EK near to -100 mV. The simplest explanation I can think of is that these plots were obtained with much higher external K+, perhaps around 50-60 mM, but again the Materials and methods state it is 2.5 mM. Might the authors have plotted the difference between current measured at the holding voltage of -20 mV and each test potential, in some cases then normalized to that difference measured at +100 mV? But this is rather unconventional, and nothing is stated about how this was done. If -20 mV really is the Vrev for these constructs and external K- is 2.5 mM, then either these constructs are no longer K-selective, or the internal concentration of K is way off due to expressing these constructs. Please explain and fix as appropriate.

A higher external K+ concentration of 60mM was used for these experiments. This is now updated in the Materials and methods section.

In addition, are the oocytes currents being manipulated in any way? The capacitive currents seem to be largely compensated or subtracted, but nothing is stated in the Materials and methods.

No leak or capacitive currents were subtracted from the oocyte current traces. This was added to the Materials and methods.

Are all the whole-cell CHO currents are not leak subtracted? It would be good to state this specifically.

The whole-cell CHO current traces are not leak subtracted. This was added to the Materials and methods.

It would also be good to add dashed lines to all traces to indicate the zero-current level.

Dashed lines to indicate zero current level were added to all traces.

Finally, in the traces shown in Figure 6A, it appears you have lost voltage-clamp in the last few traces. Perhaps replace with another recording or remove those traces?

The last few traces were removed from this figure.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Whicher JR, MacKinnon R. 2019. Low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 1. Electron Microscopy Data Bank. EMD-20295
    2. Whicher JR, MacKinnon R. 2019. Low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 2. Electron Microscopy Data Bank. EMD-20294
    3. Whicher JR, MacKinnon R. 2019. Atomic coordinates for Eag1 3-13/CaM conformation 1. Protein Data Bank. 6PBY
    4. Whicher JR, MacKinnon R. 2019. Atomic coordinates for Eag1 3-13/CaM conformation 2. Protein Data Bank. 6PBX

    Supplementary Materials

    Transparent reporting form
    elife-49188-transrepform.docx (244.8KB, docx)
    DOI: 10.7554/eLife.49188.015

    Data Availability Statement

    The low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 1 (accession code: EMD-20295) and conformation 2 (accession code: EMD-20294) have been deposited in the electron microscopy data bank. Atomic coordinates for Eag1 3-13/CaM conformation 1 (accession code: 6PBY) and conformation 2 (accession code: 6PBX) have been deposited in the protein data bank.

    The following datasets were generated:

    Whicher JR, MacKinnon R. 2019. Low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 1. Electron Microscopy Data Bank. EMD-20295

    Whicher JR, MacKinnon R. 2019. Low pass filtered and amplitude modified 3D cryo-EM density maps for Eag1 3-13/CaM conformation 2. Electron Microscopy Data Bank. EMD-20294

    Whicher JR, MacKinnon R. 2019. Atomic coordinates for Eag1 3-13/CaM conformation 1. Protein Data Bank. 6PBY

    Whicher JR, MacKinnon R. 2019. Atomic coordinates for Eag1 3-13/CaM conformation 2. Protein Data Bank. 6PBX

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES