Research Article

Structure of the voltage-gated K+ channel Eag1 reveals an alternative voltage sensing mechanism

See allHide authors and affiliations

Science  12 Aug 2016:
Vol. 353, Issue 6300, pp. 664-669
DOI: 10.1126/science.aaf8070

A different gate design

The voltage-gated potassium channel Eag1 is overexpressed in tumor cells from a range of cancers, and inhibiting Eag1 reduces tumor growth. Whicher and Mackinnon determined the structure of a mammalian Eag1 bound to the inhibitor calmodulin at 3.78 Å resolution (see the Perspective by Toombes and Swartz). The organization of the voltage-sensing and pore domains differs from that of other potassium channels, indicating that the gating mechanism is distinct. The structure also shows how the channel can be closed by a ligand, independently of the position of the voltage sensor.

Science, this issue p. 664; see also p. 646

Abstract

Voltage-gated potassium (Kv) channels are gated by the movement of the transmembrane voltage sensor, which is coupled, through the helical S4-S5 linker, to the potassium pore. We determined the single-particle cryo–electron microscopy structure of mammalian Kv10.1, or Eag1, bound to the channel inhibitor calmodulin, at 3.78 angstrom resolution. Unlike previous Kv structures, the S4-S5 linker of Eag1 is a five-residue loop and the transmembrane segments are not domain swapped, which suggest an alternative mechanism of voltage-dependent gating. Additionally, the structure and position of the S4-S5 linker allow calmodulin to bind to the intracellular domains and to close the potassium pore, independent of voltage-sensor position. The structure reveals an alternative gating mechanism for Kv channels and provides a template to further understand the gating properties of Eag1 and related channels.

Excitable cells, such as neurons and muscle cells, transmit electrical impulses known as action potentials. The action potentials are propagated by voltage-gated ion channels, which function analogously to electrical transistors by gating in response to membrane voltage. Voltage-gated potassium (Kv) channels promote K+ efflux upon cell membrane depolarization and thereby return the voltage to the K+ Nernst or resting membrane potential. Kv channels are tetramers, and each subunit comprises six transmembrane segments (S1 to S6). The S1 to S4 segments form the voltage-sensor (VS) domain, which contains multiple positively charged residues that move in response to the value of the membrane voltage. The movement of the VS is coupled to the pore (S5-S6) by alinker connecting S4 to S5 (S4-S5 linker). Inknown Kv structures, as well as voltage-dependent Na+ and Ca2+ channel structures, the S4-S5 linker is an α helix that runs parallel to the membrane and is positioned above S6 (14). In this position, the S4-S5 linker was proposed to act as a lever that performs mechanical work on S6 to open and close the channel when the charged S4 helix is driven through the transmembrane electric field (5).

Kv channels can be divided into 12 subfamilies (Kvs 1 to 12). The Eag family, which includes Kv channels 10 to 12, displays sequence homology to Kv channels 1 to 9, the cyclic nucleotide–gated (CNG) channels, and the hyperpolarization-activated (HCN) channels (6). Members of the Eag family have the S1 to S6 transmembrane segments, as well as three intracellular domains, an N-terminal Per-ARNT-Sim (PAS) domain, a C-terminal C-linker domain, and a C-terminal cyclic nucleotide–binding homology domain (CNBHD). The different gating properties of the Eag family members can be partially explained by the sequence variations among their PAS domains, which demonstrate its importance in gating (711). The C-linker domain and CNBHD are homologous to the intracellular domains of HCN and CNG channels. In HCN and CNG, binding of cyclic nucleotides to the cyclic nucleotide–binding domain (CNBD) is coupled to the pore via the C-linker domain (12). In the Eag family, a portion of the CNBHD sequence occupies the putative cyclic nucleotide–binding site, which prevents binding of and renders the Eag channels insensitive to cyclic nucleotides (1316). Structures of the PAS domain and CNBHD both separately and in complex provided insights into their role in channel gating (9, 1416). However, the structures did not provide an understanding of how the domains assemble into tetramers or how these domains interact with transmembrane regions. As a result, there remain many questions as to how the intracellular domains gate the channel and how this gating mechanism compares with that of the other Kv channels and the CNG channels.

Kv10.1, or Eag1, is member of the Eag family that produces a potassium-selective, noninactivating current (17). Activation time of Eag1 is sensitive to hyperpolarization, a property that may be regulated by the interaction between the N terminus of the PAS domain and the S4-S5 linker (7, 8). In addition, Eag1 is inhibited by the binding of calmodulin (CaM) only in the presence of calcium (18, 19). Eag1 is found primarily in the brain, where it localizes to presynaptic terminals and regulates neurotransmitter release (20). Furthermore, Eag1 functions in the differentiation of myoblasts (21) and in the regulation of cell cycle progression (22). Overexpression of Eag1 in mammalian cells induces an oncogenic phenotype, and Eag1 is found overexpressed in tumor tissues from a wide range of cancers (23, 24). Inhibition or knockdown of Eag1 reduces tumor growth (2529). Therefore, a structure of Eag1 will provide an advance toward an understanding of the gating mechanism of the Eag family, as well as a template for the design of inhibitors that could be used as anticancer therapeutics.

Structure determination of rEag1

We determined the structure of the rat ortholog of Eag1 (rEag1) at 3.78 Å resolution using single-particle cryo–electron microscopy (cryo-EM) (fig. S1). In order to produce stable homogeneous rEag1, it was necessary to truncate the C terminus and purify rEag1 bound to the channel inhibitor CaM (fig. S1). The truncation removed 114 residues (773 to 886) from the unstructured C terminus but retained residues 887 to 962, which are necessary for tetramer formation (30). Truncated rEag1 (rEag1Δ) displayed all the properties of the wild-type (WT) rEag1, including (i) a non-inactivating current with V0.5 of 4.9 ± 0.6 mV (SEM) compared with 10 mV for WT Eag1 (31), (ii) activation kinetics that are sensitive to hyperpolarization, and (iii) inhibition by CaM (Fig. 1, A to D).

Fig. 1 Functional properties of rEag1Δ.

(A) Representative voltage family current trace of rEag1Δ with the voltage-pulse protocol shown above. (B) Tail current (I/Imax, means ± SD) versus voltage (I-V plot) from (A) was plotted and fit with a Boltzmann function to give a V0.5 of 4.9 ± 0.6 mV (SEM) and valence z = 1.4 ± 0.05 (SEM) (N = 4). (C) Representative current trace demonstrating that the activation time of rEag1Δ increases after more hyperpolarized (negative) holding potentials. Voltage-pulse protocol is shown above recording. (D) CaM inhibition of rEag1Δ. Inside-out patches from CHO cells expressing rEag1Δ excised in the presence of 10 μM free Ca2+ (left) and after local perfusion of 5 mM EDTA (right). Voltage-pulse protocol is shown above recording.

The overall resolution of the rEag1 cryo-EM density map was 3.78 Å. However, local resolution calculations suggest that the resolution for S1, S2, S4, S5, S6, C-linker, and CNBHD was between 3.3 and 4 Å (fig. S2). In agreement with this calculation, most side chains were observed in these regions of the map, which allowed for de novo model building (fig. S3). The local resolution for PAS, S3, CaM, and regions of the CNBHD (residues 683 to 722) near the CaM binding sites, was between 4 and 6 Å (figs. S2 and S3). As a result, fewer side chains were observed in these regions, but strong main-chain density allowed for accurate model building and placement of crystal structures. The rEag1 model was refined in both real and reciprocal space against one of the independently calculated half maps (working set) (fig. S4). Validation of the model against the other independently calculated half map (free set) and the full map demonstrates good agreement between the model and the maps and the absence of over fitting (fig. S4).

Overall structure of rEag1

Like previously known Kv structures, the transmembrane region of Eag1 has six helical segments (S1 to S6) (1, 2, 32). S1 to S4 form the VS, and S5, the pore helix, and S6 form the potassium pore (Fig. 2). A 40–amino acid turret (residues 378 to 417) between S5 and the pore helix extends ~25 Å out of the membrane and surrounds the extracellular opening of the pore (Fig. 2 and fig. S5). An extended extracellular turret was not observed in previous Kv structures (1, 2, 32), and sequence analysis suggests that, of the eukaryotic Kv channels, the Eag family are the only channels that have such an extended extracellular turret (fig. S6). Part of the turret forms an α helix (391 to 399 turret helix), which runs parallel to the membrane and interacts with the pore helix and the loop between the selectivity filter and S6. There are two glycosylation sites in the turret (N388 and N406) (33), and density was observed for the first N-acetyl glucosamine on N388 (figs. S3 and S5). In this position, the sugar chain may surround the extracellular opening to the pore and prevent binding of inhibitory toxins, which explains why none have been identified for rEag1.

Fig. 2 Model of rEag1Δ bound to CaM.

Each domain is colored as indicated, and the cell membrane is shown as gray lines. (Left) One subunit is represented as a cartoon, and the remaining subunits are shown in surface representation. (Right) Transmembrane domains are represented as a cartoon, and the intracellular domains are represented as a surface. The length and width of transmembrane and intracellular domains are indicated. The N (orange) and C terminus (cyan) are labeled and shown as spheres. The glycosylation site observed in the turret is shown as sticks with green C, red O, and blue N.

The intracellular domains extend ~65 Å into the cytoplasm. The PAS domains are located at the periphery of the intracellular region, where they interact primarily with the CNBHD from a neighboring subunit and are positioned directly underneath the VS (Fig. 2 and fig. S7). In the context of the full channel, we observed structural elements of the PAS domain not previously seen, including the N-terminal loop (residues 11 to 16), the helix that binds CaM (residues 147 to 157), and a C-terminal loop that connects to S1 of the VS (residues 198 to 213) (figs. S3 and S7). The N terminus of the PAS domain, which influences the rate of voltage-dependent channel opening (activation) and closing (deactivation), is directed toward the VS (Fig. 2 and figs. S3 and S7) (8).

The S6 helix extends into the intracellular region and connects to the C-linker (Fig. 2 and fig. S7). The C-linker has four helices and, as in the HCN channel, the first two helices fold into a helix-turn-helix motif that interacts with the second two helices from the neighboring subunit (12). With this structure, the C-linker forms an intracellular ring directly above the CNBHD ring that can couple the movements of the S6 and CNBHD (Fig. 2 and fig. S7). The CNBHD tetramer displays a quaternary structure similar to that of HCN (12). However, the intracellular vestibule formed by the CNBHD does not contain many negative charges, as is found in many cation channels, which may explain the low conductance of these channels (8 pS) (fig. S7) (34). Like previous CNBHD structures, a portion of the Eag1 sequence occupies the cyclic nucleotide–binding site, which prevents cyclic nucleotide binding (figs. S3 and S7) (1416).

VS structure

The overall architecture of the Eag1 VS is similar to that of previous Kv structures, but in Eag1, the region of S3 and S4 near the outer membrane leaflet—known as the VS paddle—is shifted toward S2 (fig. S8) (1, 2, 32). The S4 of Eag1 is a long 310 helix that contains six positively charged residues (K327, R330, R333, R336, R339, and K340). In addition, F261 and D264 from S2 and D299 from S3 form a charge-transfer center, which facilitates the movement of the positively charged residues on S4 in response to changes in membrane voltage (35).

Closer inspection of the Eag1 VS reveals three important differences when compared with the Kv1.2-2.1 VS (2). The first difference is a 13-residue linker between S2 and S3 (S2-S3 linker) in Eag1 that extends into the cytoplasm (Fig. 3 and figs. S3 and S8). This linker is a conserved feature of the Eag family, and the only other Kv subfamily that contains a similar linker is Kv7 (fig. S6).

Fig. 3 Structure of VS domains (S1 to S4).

Stereoview of the rEag1 (A) and the Kv1.2-2.1 (PDB ID 2R9R) (B) VSs (S1 to S4). The five equivalent positions on S4 (blue spheres and sticks with yellow C and blue N; K327, R330, R333, R336, and R339 in rEag1) and the Phe (green spheres and sticks with green C; F261 in rEag1), as well as the negative charges (red spheres and sticks with yellow C and red O; D264 and D299 in rEag1) in the charge transfer. The S2-S3 linker and S4-S5 linker are labeled.

The second difference is the register of the positively charged residues on S4 relative to the gating-charge–transfer center (Fig. 3). The S4 of Eag1 and Kv1.2-2.1 both have five positively charged residues in equivalent positions (charges 1 to 5; the Gln in position 1 of Kv1.2-2.1 is an Arg in Kv1.2) (Fig. 3 and fig. S6); however, in the VS of Kv1.2-2.1, which is in the open or depolarized conformation, charges 1 to 4 are above (external to) the charge-transfer center, whereas in Eag1 only charges 1 to 3 (K327, R330, and R333) are above the charge-transfer center (2). In Eag1, charge 4 (R336) interacts with D264 and D299 of the charge-transfer center, and although side-chain density was not observed for the charge 5 Arg (R339), it is near D222, which is conserved in the Eag family. We believe this is the depolarized conformation of the Eag1 VS for the following reasons. First, another member of the Eag family, known as hERG or Kv11.1, was shown to have a smaller gating charge than Kv1.2 (6 to 8 compared with 12 to 14 elementary charge units), which suggests that fewer charges move through the charge-transfer center during hERG gating (3638). Second, charge-neutralizing mutations indicated that only charges 1 to 3 of hERG are involved in sensing membrane voltage (36). Finally, Eag1 is open at 0 mV, the voltage at which the structure was determined (Fig. 1, A and B). Therefore, in the depolarized conformation of Eag1, the S4 appears to be displaced one 310 helix turn toward the intracellular membrane when compared with Kv1.2-2.1. This allows the S4-S5 linker of Eag1 to interact with the intracellular region of S6 in the up or depolarized conformation (Fig. 4A).

Fig. 4 Architecture of transmembrane domains (S1 to S6).

The transmembrane segments (S1 to S6) of rEag (A) and Kv1.2-2.1 (B) (PDB ID 2R9R) are shown with each subunit represented as a different color and the S4-S5 linker shown in red. (C) Stereoview of S4, the S4-S5 linker, and S5 segments of rEag1 (shown as sticks with yellow C, red O, and blue N) with cryo-EM density.

The third difference is the structure of the S4-S5 linker. In Eag1, the S4-S5 linker is a short five-residue loop (Fig. 3A; Fig. 4, A and C; and fig. S6). In Kv1.2-2.1, the S4-S5 linker is a 15-residue helix that runs parallel to the membrane (Figs. 3B and 4B). This change in the length and structure of the S4-S5 linker has important implications in the architecture of S1 to S6 and channel gating.

Architecture of S1 to S6

The 15-residue helical S4-S5 linker in Kv1.2-2.1 structure results in domain-swapped transmembrane segments (VS interacts with the S5 from a neighboring subunit) (Fig. 4B) (1, 2). This domain swap positions the S4-S5 linker of Kv1.2-2.1 above S6 from within its own subunit and creates a ring of S4-S5 linker helices that wrap around the S6 helices (1, 2). The five-residue S4-S5 linker of Eag1 cannot span the ~20 Å distance between S4 and S5 of the domain-swapped Kv1.2-2.1. As a result, the transmembrane segments of Eag1 are not domain-swapped (VS interacts with the S5 from the same subunit) and likely remain in a non–domain swapped conformation domain in all channel states (Fig. 4, A and C). In the non–domain swapped architecture of Eag1, the S4-S5 linker is positioned above the S6 from the neighboring subunit, and the interaction of S1 and S5 makes up the largest interface between the VS and pore (Fig. 4A and fig. S8).

In Kv1.2-2.1, the position and structure of the S4-S5 linker, as well as mutational studies on the closely-related Shaker Kv channel, suggest that the S4-S5 linker functions as a mechanical lever to couple movement of the VS to S6 (Fig. 4B) (5, 39). A conserved proline in S6 was proposed to act as a hinge (Pro-Val-Pro hinge), which allows S6 to move up and down in the membrane in response to movement of the S4-S5 linker to open and close the channel (5, 40). However, the Eag1 S4-S5 linker is a short loop that is not domain swapped and, thus, is not in a position to function as a mechanical lever. Indeed, cutting the S4-S5 linker in Eag1, that is, deleting the covalent link between VS and the pore, produces a functional voltage-gated channel (41). Furthermore, the S6 hinge proline is replaced by G469 in Eag1 (fig. S6), which mutational experiments suggest does not function as a hinge (42). In agreement, S6 of Eag1 is entirely helical with no noticeable kink before the S6 bundle crossing (Fig. 4A). These structural differences suggest that the mechanism of coupling VS movement to pore opening and closing is fundamentally different between Eag1 and Kv1.2-2.1. Sequence analysis indicates that all Eag family members have a short S4-S5 linker and no hinge proline in S6, whereas Kvs 1 to 9 have a longer S4-S5 linker and a hinge proline (Figs. 3 and 4 and fig. S6). Therefore, we predict that all Eag family members will have a non–domain swapped architecture and a gating mechanism similar to Eag1, and Kvs 1-9 will have a domain-swapped architecture and a gating mechanism similar to that of Kv1.2-2.1. Why is the mechanism of coupling VS movement to pore opening and closing different between the Eag family and Kvs 1 to 9 and how is movement of the VS coupled to the pore in Eag1? To answer these questions, it is necessary to first understand the mechanism of CaM inhibition and the interaction between the intracellular domains and the VS.

Mechanism of CaM inhibition

CaM binds Eag1 in the presence of Ca2+ and inhibits ion conduction (Fig. 1D) (18, 19). In agreement with this, when bound to CaM, the rEag1 pore is closed with a diameter (~1 Å between van der Waal surfaces) less than that of hydrated potassium (6 to 8 Å) (Fig. 5A). In this closed conformation, the S6 of rEag1 has an ~55° bend directly after the inner helix gate (Q476) in a direction that would tighten the S6 helical bundle (Fig. 5B).

Fig. 5 Intracellular domains and CaM interaction.

(A) Two subunits of the tetramer are shown, and sticks (green C, red O, and blue N) indicate the selectivity filter (436 to 441), constriction sites (F468 and T472), and the bundle crossing (Q476) (left). The plot of pore diameter (right) is aligned with the molecular model. At all points, the pore diameter is smaller than that of hydrated potassium (dashed gray line at 6 Å). (B) Organization of rEag1 intracellular domains (extracellular view). Directly C-terminal to the bundle crossing [spheres colored as in (A)], the S6 helix (green) is connected to the C-linker (red). The CaM (purple) is bound to both the CNBHD (cyan) and the PAS (orange). (C) Two helices (683 to 695 and 708 to 722) at the C terminus of the CNBHD (cyan, a dashed line shows the connection between helices) act as a claw to grab the CaM C-lobe (purple). The N-lobe of the same CaM molecule binds to a helix on the PAS domain (147 to 157, orange) that interacts with the neighboring CNBHD (gray). (D) The pore is closed with a 55° bend in S6 in a direction that tightens the helical bundle that forms the intracellular gate [Q476, spheres colored as in (A)]. The VS (yellow) is in an up or depolarized conformation with the S4 directed toward the C-linker at a distance of 8 Å. The N terminus of the PAS domain (orange sphere) and the S2-S3 linker are directed toward the interface between VS and C-linker.

Each Eag1 subunit has three CaM contact regions, which form two binding sites (19). In rEag1, all three regions are occupied by one CaM molecule and thus four CaM molecules are bound to the rEag1 tetramer (Fig. 5, B and C). The two contact regions at the C terminus of the CNBHD form two α helices that act as a claw to grab the CaM C-lobe. The first helix (residues 683 to 695), which has been shown to bind with lower affinity, interacts with the top of the CaM C-lobe near the Ca2+ binding sites and the second helix (residues 708 to 722), which has been shown to bind with higher affinity, is bound to the hydrophobic pocket of the CaM C-lobe (Fig. 5, B and C, and fig. S9) (19). The linker helix in CaM that connects the C- and N-lobes breaks and forms a loop from M76 to D80, a segment of high flexibility in CaM (fig. S9) (43). The break in the linker helix allows the hydrophobic core of the CaM N-lobe to bind a helix (residues 147 to 157) on the PAS domain (Fig. 5, B and C, and fig. S9). Note that this PAS domain interacts with the neighboring CNBHD. We propose the modeled orientation of the CaM lobes for two reasons. First, the linker constraints will only allow for this orientation (fig. S9). Second, further classification of the rEag1 particles generated a cryo-EM density map with improved CaM density that shows the position of the linker connecting the CaM lobes (fig. S9).

By binding to the CNBHD and PAS from neighboring subunits, we propose that the CaM acts as a molecular clamp to pull the two domains together and, thereby, alters the orientation of neighboring CNBHD domains (cyan and gray in Fig. 5C). Such a clamping mechanism would translate the CNBHD interacting with the CaM (cyan in Fig. 5C) toward the neighboring PAS domain (Fig. 6A). Because the CNBHD domain is connected to S6 via the C-linker, we hypothesize that the movement of the CNBHD toward the PAS domain would cause a rotation of both the C-linker and S6 to induce the 55° bend in a direction that tightens the helical bundle that forms the intracellular gate and to close the pore (Fig. 6A). Thus, CaM binding appears to twist the pore helices shut.

Fig. 6 Proposed mechanism of CaM inhibition and voltage-dependent gating.

(A) By binding to the PAS (orange) and CNBHD (cyan), the CaM (purple) molecule seems to pull the CNBHD, which is connected to the S6 (green) via the C-linker (red), toward the PAS domain. This would cause a clockwise rotation (when viewed from the extracellular side of the membrane) of the C-linker and S6 to close the potassium pore. (B) In the structure of rEag1 bound to CaM, the pore is closed with a 55° bend in S6 (green cylinder) in a direction that tightens the helical bundle that forms the intracellular gate, but the VS is in an up or depolarized conformation with the S4 (yellow cylinder) directed toward the C-linker (red cylinders) at a distance of 8 Å (top right, same conformation as Fig. 5D). In an up or depolarized state of the VS in the absence of CaM, the C-linker might rotate in a direction that loosens the helical bundle to relieve the curve in S6 and open the channel (left). In the closed or hyperpolarized state of the VS, the downward movement of the S4 might contact the C-linker to induce a bend in S6 similar to that imposed by CaM binding to close the channel (bottom right). Inner leaflet of the membrane is indicated by gray lines.

Insights into voltage-dependent gating

Functionally, CaM binding induces pore closure independent of voltage (Fig. 1D) (18, 19). The basis for this effect is observed in the structure of Eag1 bound to CaM, as the closed pore is decoupled from the up or depolarized VS (Figs. 3A and 5D). The non–domain swapped architecture of the Eag1, in which the S4-S5 linker is a short loop that forms a weak interaction with S6 from the neighboring subunit (Fig. 4A) rather than a helix that wraps around the S6 helices as in Kv1.2-2.1 (Fig. 4B), allows for pore closure by cytoplasmic domains independent of VS conformation. In this closed conformation, the C terminus of S4 is directed toward the C-linker at a distance of ~8 Å (Fig. 5D). We hypothesize that upon release of CaM, the C-linker and S6 would rotate toward the S4 in a direction that loosens the helical bundle. Such a rotation would relieve the high-energy 55° bend in S6 and open the channel (Fig. 6B).

In the up or depolarized conformation of the VS, the S4-S5 linker interacts with the intracellular portion of S6 and thereby directs the C terminus of S4 toward the C-linker (Fig. 5D). In a down or hyperpolarized conformation of the VS, we hypothesize that the C terminus of S4 will move toward the cytoplasm to interact with the C-linker and will induce a bend in S6 similar to that imposed by CaM binding to close the channel (Fig. 6B). This mechanism, in which S4, rather than the S4-S5 linker, directly interacts with the C-linker to close the channel, would not require a covalent link between S4 and S5 to produce a voltage-dependent channel and, thus, is consistent with previous reports (41). The N terminus of the PAS domain extends from the helical PAS cap (residues 17 to 25) and is directed toward the interface between S4 and C-linker through interactions with the PAS C-terminal loop, the S2-S3 linker, and the CNBHD (Fig. 5D and fig. S7). In this position, the N terminus of the PAS domain may interact with the VS in a down or hyperpolarized conformation to stabilize the closed state and confer activation time sensitivity to hyperpolarization (8). The S2-S3 linker is also directed toward the S4/C-linker interface and may play a yet-unknown role in channel gating.

Summary

The structure of rEag1 reveals a non–domain swapped architecture of the S1 to S6 that is due to a short five-residue S4-S5 linker. This represents a divergence from the domain-swapped architecture of previous voltage-gated ion channel structures (Fig. 4) (14) and suggests a new paradigm for voltage-dependent gating for the Eag family of Kv channels. On the basis of the structure of Eag1, we propose a gating mechanism in which S4 enters the cytoplasm in a down or hyperpolarized state to interact with and induce a rotation of the C-linker and S6 in a direction that tightens the helical bundle to close the channel (Fig. 6B). In the up or depolarized state of the VS, S4 moves into the membrane, which allows the C-linker and S6 to rotate in a direction that loosens the helical bundle and thus relieves the high-energy bend in S6 to open the channel (Fig. 6B).

Two important consequences result from a gating mechanism in which the VS interacts with the cytoplasmic domains to gate the channel. First, this allows the cytoplasmic domains to close the channel independent of VS conformation. This is observed in the structure of Eag1, as binding of CaM to the cytoplasmic domains closes the pore, but the VS is in the up or depolarized conformation. Second, this provides an added level of regulation through the interaction of intracellular domains with the voltage-dependent gating machinery. In Eag1, the N terminus of the PAS domain, which confers sensitivity to hyperpolarization (8), is poised to interact with the S4 and S4-S5 linker in a closed conformation (Fig. 5D).

Supplementary Materials

www.sciencemag.org/content/353/6300/664/suppl/DC1

Materials and Methods

Figs. S1 to S9

References (4465)

References and notes

  1. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
Acknowledgments: We thank M. Ebrahim at the Rockefeller University cryo-EM resource center for help with data collection and J. Chen and members of the MacKinnon laboratory for helpful discussions. This work was supported in part by NIH grant GM43949. J.R.W. is a Damon Runyon Fellow supported by the Damon Runyon Cancer Research Foundation (DRG-2212-15) and R.M. is an investigator in the Howard Hughes Medical Institute. The low-pass filtered and amplitude-modified three-dimensional cryo-EM density maps for rEag1 have been deposited in the EM Data Bank with accession code EMD-8215. Atomic coordinates for rEag1 have been deposited in the Protein Data Bank under accession code 5K7L. J.R.W. performed functional experiments and expressed, purified, and determined single-particle cryo-EM structure of rEag1. J.R.W. and R.M. designed experiments, analyzed and interpreted results, and wrote the manuscript.
View Abstract

Navigate This Article