Research Article

Crystal Structure of the Human Two–Pore Domain Potassium Channel K2P1

See allHide authors and affiliations

Science  27 Jan 2012:
Vol. 335, Issue 6067, pp. 432-436
DOI: 10.1126/science.1213274


Two–pore domain potassium (K+) channels (K2P channels) control the negative resting potential of eukaryotic cells and regulate cell excitability by conducting K+ ions across the plasma membrane. Here, we present the 3.4 angstrom resolution crystal structure of a human K2P channel, K2P1 (TWIK-1). Unlike other K+ channel structures, K2P1 is dimeric. An extracellular cap domain located above the selectivity filter forms an ion pathway in which K+ ions flow through side portals. Openings within the transmembrane region expose the pore to the lipid bilayer and are filled with electron density attributable to alkyl chains. An interfacial helix appears structurally poised to affect gating. The structure lays a foundation to further investigate how K2P channels are regulated by diverse stimuli.

The high resting permeability of the plasma membrane to potassium ions (K+) was appreciated for nearly 50 years (13) before these “background” K+ currents were partially attributed to a family of two–pore domain K+ (K2P) channels (4). A total of 15 K2P channels in humans can be divided into 6 K2P subfamilies based on sequence identity and functional characteristics: TWIK, TREK, TASK, TALK, THIK, and TRESK (4). Under normal physiological conditions, the K+ gradient across the cellular membrane (~150 mM inside and ~5 mM outside) causes the net efflux of K+ ions through K2P channels, which stabilizes the negative electrical potential (resting potential) of the cell (4). Expression of the K2P channel K2P1 (TWIK-1) has been detected in several tissues and plays important roles in excitable cells in the heart and brain (57). K2P1 has the unusual property that it becomes permeable to sodium ions (Na+) under physiological conditions when the extracellular K+ concentration is below normal levels (hypokalemia) (7). Hypokalemia occurs in up to 20% of hospitalized patients and increases the risk of sudden cardiac death (8). Because the extracellular concentration of Na+ is higher than the intracellular concentration, this altered permeability of K2P1 allows the influx of Na+ that has been shown to depolarize cardiomyocytes and may contribute to cardiac arrhythmia or arrest (7).

Although K2P channels were first considered simply “leak” channels, it has become clear that they are regulated by numerous factors (4). For example, TREK-1 is modulated by changes in temperature or pH, by membrane stretch, and by anesthetics and antidepressants (913). These regulatory properties may be involved in pain perception and neuroprotection (4). Regulation of TREK-1 activity by diverse stimuli has been partly ascribed to a C-terminal region of amino acids following the fourth transmembrane (M4) helix (912, 14, 15). Evidence suggests that the activity of K2P1 can be silenced by modification of a lysine residue in the corresponding C-terminal region (Lys274) with a small ubiquitin-like modifier (SUMO), consistent with a role for this region in channel regulation (16, 17) [although other studies have suggested that this modification does not occur (18, 19)].

A K+ channel pore is composed of four pore-forming (P) domains. Each P domain contains an outer transmembrane helix, a pore helix, a selectivity filter signature sequence, and an inner transmembrane helix (20). The crystal structures of K+ channels determined to date are of tetrameric channels, with four identical subunits that each contain one P domain. The genes of K2P channels contain two P domain sequences (P domain 1 and P domain 2) arranged in tandem, and these channels assemble as dimers (4, 5, 21) (Fig. 1C). So, each K2P protomer contains four transmembrane helices (M1 to M4). M1 and M3 are the outer helices of P domains 1 and 2, respectively. M2 and M4 are the inner helices of P domains 1 and 2, respectively. Although tetrameric K+ channels often have regulatory domains such as voltage sensors within the span of the membrane or ligand-binding domains located on the cytosolic side, they lack extracellular domains. K2P channels have an extracellular region of ~55 amino acids within P domain 1 (between M1 and pore helix 1) that is not present in P domain 2 or found in tetrameric K+ channels (fig. S1A). The function of this region and whether or not it forms a structured domain are largely unknown. Although existing structures of K+ channels give insights into principles of ion permeation, ion selectivity, and gating (22), structures of K2P channels are needed to understand the features specific to this class. Here, we present the crystal structure of human K2P1 (TWIK-1).

Fig. 1

Overall structure. (A) Tertiary structure of K2P1, showing a ribbon representation from the side. One subunit is colored blue-to-red from the N to the C terminus, and the other subunit is gray. K+ ions are shown as green spheres. Approximate boundaries of the lipid membrane are shown as horizontal lines. The intersubunit disulfide bond (involving Cys69) at the apex of the extracellular cap is colored green. Loop regions not included in the final model are indicated as dashed lines. The central cavity is located just below the K+ ions along the symmetry axis (behind Gly141). (B) An orthogonal view of the channel from the side. (C) Secondary structure of K2P1 colored according to (A). Dashed lines indicate disordered regions. Residues discussed in the text are labeled.

Structure determination. For crystallization, human K2P1 was truncated to amino acids 19 to 288 [removing 18 residues from the N terminus and 48 residues from the C terminus that are not well conserved among K2P channel families (fig. S1A)]. Cys22 and Asn95 were mutated to prevent nonspecific disulfide formation and glycosylation, respectively. The protein was expressed in Pichia pastoris, and a flux assay confirmed that the purified protein forms functional K+ channels when reconstituted into lipid vesicles (fig. S2).

K2P1 crystallized in the P21 space group with four copies of the polypeptide in the asymmetric unit assembled into two channels. The crystals were grown in the presence of 150 mM K+ and diffracted x-rays to 3.4 Å resolution. Experimental phases were derived from diffraction data collected from crystals containing mercury, gold, and thallium derivatives. The heavy atoms bound to cysteine residues within the M2, M3, and M4 transmembrane helices and served as guides in building the atomic model. Difference Fourier maps indicated that the thallium ions (Tl+), which are usually permeant in K+ channels (23), were bound within the selectivity filter. The phases, which were improved by noncrystallographic symmetry averaging, yielded an electron density map that allowed placement of the majority of the amino acid side chains (fig. S3). The coordinates were refined to yield a model with good stereochemistry and an Rfree value of 27.9% (table S1).

Molecular architecture. Although dimeric, K2P1 recapitulates the fundamental architecture of tetrameric K+ channels, and it also reveals differences that give insight into gating and ion permeation (Fig. 1). P domains 1 and 2 are arranged in a clockwise fashion when viewed from the extracellular side of the membrane. The ion conduction pathway below the selectivity filter is lined by inner helices that constitute the inner gate (also called the activation gate) in other K+ channels (Figs. 1 and 2) (24). The inner helices of K2P1 are separated by >11 Å, suggesting that the channel is crystallized in an open conformation (Fig. 2B). Near the center of the membrane, the M2 inner helix is kinked by approximately 20° at a proline residue (Pro143) that is conserved in K2P channels (except the THIK subfamily) (Fig. 1A and fig. S1A). As in tetrameric K+ channels, a central cavity is located just below the selectivity filter, near the center of the membrane (Figs. 1 and 2B). Pore helices, which surround the selectivity filter, are arranged with the negative ends of their helix-dipole moments oriented toward the central cavity and could serve to stabilize a cation within it (20, 25, 26).

Fig. 2

Interfacial C helix. Ribbon representation of the transmembrane and intracellular regions of K2P1 viewed from the membrane (A) and intracellular side (B). The inner helices are colored blue, the C helices are red, and the remaining portions are gray. K+ ions are shown as green spheres. In (A), residues on the C helix in the foreground are drawn as sticks (yellow, carbon; blue, nitrogen; red, oxygen; green, sulfur).

In addition to an extracellular domain, the salient differences between the structure of K2P1 and those of other K+ channels are an amphipathic helix located near the membrane/cytosol interface, a disparate extracellular ion pathway, and a subunit interface that exposes the central cavity to the lipid membrane.

Extracellular cap domain. In other K+ channel families, the connection between the outer helix and the pore helix (called the turret) is typically 5 to 20 amino acids (fig. S1A) (20). In K2P1, the 56 amino acids between M1 and pore helix 1 form a structured domain that extends approximately 35 Å above the lipid membrane (Fig. 1, A and B). The domain, which we refer to as the extracellular cap, contains two helices (E1 and E2) from each subunit that form a structure resembling an A-frame that is positioned directly above the extracellular side of the selectivity filter. The proximity of the extracellular region to the selectivity filter was predicted from studies of the TASK-3 K2P channel (27), suggesting that this is a common element of K2P channels. A proline residue that is conserved among K2P channels (Pro47) is located within a bend of ~30° between the M1 helix and the E1 helix (Fig. 1). Cys69, which is located in the turn between the E1 and E2 helices at the apex of the extracellular cap and conserved in most K2P channels, forms a disulfide bond that covalently links the two subunits together. This disulfide cross-link was predicted from biochemical experiments (21). The E2 helices form a hydrophobic core within the extracellular cap through the interaction of hydrophobic amino acids along their length with both the E1 helix of the same subunit and the E2 helix of the adjacent subunit (Fig. 1 and fig. S1A). The C-terminal ends of the E2 helices, one from each subunit, are positioned with the negative ends of their helix dipole moments above and to either side of the selectivity filter. Residues following the E2 helix (90 to 93) extend away from the selectivity filter following a sharp bend that occurs at a conserved glycine (Gly89) and pack against the E2 helix. This positions the remaining residues preceding pore helix 1 away from the selectivity filter. Several of these (94 to 99) are disordered. The extracellular cap creates an extracellular ion pathway that is markedly different from that observed in other K+ channel structures (discussed in greater detail below).

Interfacial C helix and inner helix gating. Comparing crystal structures of K+ channels with open and closed inner gates has allowed conformational changes underlying inner helix gating to be deduced (24). A glycine residue located on the inner helices near the midpoint of the membrane serves as a gating hinge that permits the inner helices to bend as they constrict and dilate the pore near the intracellular side (fig. S4, E and F) (24). In K2P channels, corresponding glycine residues (Gly141 and Gly256 in K2P1) are strictly conserved in both inner helices (Fig. 1 and fig. S1A), which suggests that the inner helices may move in analogous fashion. In the structure, the inner helices create a pore that has similar dimensions to other K+ channels with open inner gates (Fig. 2B and fig. S4E). Some studies support the hypothesis that the inner gates of K2P channels function similarly to other K+ channels (28, 29). In a study of TREK-1, however, quaternary ammonium compounds such as tetrahexylammonium (THA) freely accessed the central cavity before channel activation, which was interpreted to mean that the inner gate is constitutively open (30). The structure of K2P1 raises the alternative possibility that THA, which is lipophilic, might also be able to access its binding site via openings in the transmembrane region that expose the central cavity to the membrane.

In the structure of the voltage-dependent K+ channel Kv1.2, a helix at the membrane/cytosol interface (the S4-S5 helix) couples the voltage sensor to the inner gate (31, 32). In K2P channels, amino acids following the M4 helix have been experimentally shown to play a role in gating (912, 14, 15, 33). In K2P1, we find that these residues (His271-Tyr 281) form an amphipathic helix (the C helix) at the membrane/cytosol interface that is reminiscent of the S4-S5 helix of Kv1.2 (Fig. 2). The C helix lies roughly parallel to the membrane, following a bend of ~80° (at Leu270) from M4, with its hydrophobic and hydrophilic residues oriented toward and away from the membrane core, respectively (Fig. 2A). Lys274, the potential site of SUMO modification (16, 17), is located on it (Fig. 2A). The C-terminal end of the C helix reaches across the dimer interface and is positioned between the M1 and M2 helices of the adjacent subunit (Figs. 1A and 2). The C helix is therefore coupled to two inner helices: the M4 helix to which it is covalently connected and the M2 helix of the adjacent subunit (Fig. 2). The C helix appears structurally poised to affect inner helix gating.

The regions in other K2P channels that correspond to the C helix of K2P1 serve as part of the gating apparatus that responds to various stimuli, including certain phospholipids, intracellular acidification, and membrane stretch (912, 14, 15). This region of TREK-1, which is predicted to be helical from sequence analysis, associates with the plasma membrane in a dynamic manner that is correlated with channel gating, consistent with the idea that it may be a mobile gating element (14, 34). In the conformation of K2P1 observed in the crystal structure, the C helix may stabilize an open gate by separating the inner helices. Further experiments are necessary to understand the coupling of this region to channel gating and determine whether these mechanisms vary among K2P channels.

Selectivity filter. In the selectivity filters of tetrameric K+ channels, dehydrated K+ ions are coordinated by backbone carbonyl oxygens and a threonine side chain that are arranged in a four-fold symmetric configuration that mimics the hydration of K+ ions (35). In K2P1, the amino acid sequences comprising the selectivity filter differ between P domains 1 and 2 (TTGYGH and TIGLGD) and deviate from the “canonical” K+ channel filter sequence TXGYGD, where X represents an aliphatic amino acid (Fig. 3 and fig. S1A). In the assembled channel, the packing of the side chains surrounding the selectivity filter is two-fold symmetric (Fig. 3D). However, at the limit of accuracy with which atomic positions can be determined in the K2P1 structure, the coordination of ions within the filter is four-fold symmetric and the amino acids comprising the selectivity filter adopt a conductive configuration (Fig. 3) (35). Thr118, the second T of the TTGYGH filter sequence, is associated with the ability of K2P1 to conduct Na+ when the extracellular K+ concentration is below ~3 mM, which occurs during hypokalemia (7). In the crystal, which was grown in 150 mM K+, the side chain of Thr118 occupies approximately the same three-dimensional space as aliphatic residues normally found in this position, and the backbone is indistinguishable from a canonical conductive conformation (Fig. 3, A to C). Whereas the conformation of the selectivity filter of the prokaryotic K+ channel KcsA, which has a canonical sequence, has been shown to change when the K+ concentration is low (35), the filter of another K+ channel (MthK) maintains its conformation in low K+ (36). What conformation the K2P1 filter adopts in low K+, what effect Thr118 has on this, and how Na+ permeates through it are undetermined. The selectivity filter observed in the dimeric K2P1 channel underscores the idea of a universal mode of K+ selectivity that incorporates four-fold symmetry of K+ ion coordination.

Fig. 3

Selectivity filter. (A to C) Conformation of amino acids comprising the selectivity filter from P domain 1 (A) and P domain 2 (B) compared with the selectivity filter of Kv1.2 (C) (PDB ID 2R9R). Two subunits are shown from the side, with the extracellular solution located above. The filter sequence is labeled in single-letter amino acid code (D, Asp; G, Gly; H, His; I, Ile; L, Leu; T, Thr; V, Val; Y, Tyr). Amino acids are drawn as colored sticks (yellow, carbon; blue, nitrogen; red, oxygen; and green, sulfur), and K+ ions are depicted as green spheres. (D) Depiction of a ~10 Å–thick cross section of K2P1, viewed from the extracellular side, showing the packing of residues (orange-colored carbons) surrounding the selectivity filter (yellow-colored carbons). Circles highlight differences in packing surrounding the selectivity filter residues of P domains 1 and 2. K+ ions are drawn as green spheres.

Extracellular ion pathway. K2P1 has a distinct extracellular ion pathway compared with the structures of other K+ channels (Fig. 4). The extracellular cap restricts access to the selectivity filter, allowing the passage of K+ ions only through two side portals (Fig. 4A, arrows). The portals are funnel-shaped, and their narrow ends meet above the selectivity filter. The van der Waals surface of the opening at the narrow end is approximately 6 Å across (~9 Å measured between atom positions) (Fig. 4B). The size of the portals would allow K+ ions to move through while remaining hydrated (Fig. 4B). The walls lining the side portals are predominantly negatively charged (contributed in part by Glu84, Ser86, Asn87, Ser91, and Glu235) (fig. S5C). Notably, the C-terminal ends of the E2 helices (one from each subunit) also line the portals above the selectivity filter (Fig. 1B and Fig. 4C). The negative electrostatic potential at the C-terminal end of these helices, which arises from the helix dipole effect (3740), would raise the local concentration of cations near the mouth of the selectivity filter while lowering the concentration of anions. The electron density maps are consistent with a K+ ion that is coordinated by two water molecules bound in an external site outside the filter (Fig. 4C and fig. S6). This assignment is confirmed by the 6-σ peak in a difference Fourier electron density map calculated using diffraction data from a crystal soaked in Tl+, an electron-rich surrogate for K+ (23) (Fig. 4C).

Fig. 4

Extracellular ion pathway. (A) Cutaway view of the molecular surface of K2P1 (gray) from the side with the approximate boundaries of the lipid membrane shown as horizontal lines. K+ ions (four sites in the selectivity filter and one external site) are shown as green spheres. Side portals (arrows) located beneath the extracellular cap connect the selectivity filter with the extracellular solution. Orange mesh depicts electron density assigned to alkyl chains. (B) Diameter of the pore (measured as the separation between van der Waals surfaces) as a function of distance along the ion conduction pathway. A dashed line indicates the diameter of a K+-H2O complex. The 0 coordinate on the y axis corresponds to the location of the external K+ ion. (C) Ion binding in the extracellular pathway. Amino acids of the selectivity filter (yellow-colored carbons, P domain 2) and E2 helices (cyan-colored carbons) are shown as sticks. The orientation is from the side, with a 90° rotation about the vertical axis with respect to (A). Electron density corresponding to potassium ions is shown from a simulated-annealing omit Fo-Fc map (blue mesh, calculated from 20 to 3.4 Å resolution and contoured at 4 σ). A Tl+ isomorphous difference Fourier map is also shown (red mesh, calculated from 20 to 6.2 Å resolution and contoured at 4 σ).

Sequence alignments indicate that several of the residues that line the side portals vary among K2P channels (fig. S1A). This may account for some of the electrophysiological differences between K2P channels (4) and suggests that it may be possible to block the extracellular ion pathway with channel-specific drugs. The presence of the extracellular cap explains why protein toxins that block the pores of many tetrameric K+ channels from the extracellular side are ineffective against K2P channels (5, 41, 42). Members of the Kir family of tetrameric K+ channels are also relatively insensitive to these toxins, and this is attributed to the particularly large turrets (containing ~20 amino acids) of these channels on the extracellular side (43). In addition to conferring resistance to toxins, it is possible that the extracellular cap serves as a binding site for endogenous extracellular factors that modulate K2P channel activity.

Subunit interface. K2P1 has an unusual transmembrane molecular surface that may have important functional consequences. The two P domains within each K2P1 subunit form a continuous molecular surface throughout the transmembrane region (Fig. 5A). However, the surface formed at the interface between the two subunits is markedly different from the intramolecular surface and from the transmembrane surfaces of tetrameric K+ channel structures determined to date (Fig. 5B). In K2P1, openings between the subunits expose the central cavity to the hydrophobic core of the lipid membrane. The two openings, one on each side of the dimer, are sealed off at the top at the level of the selectivity filter and at the bottom by the C helix and consequently span much of what would correspond to the inner leaflet of the membrane. They are located between the M2 helix of one subunit and the M4 helix of the other (Figs. 1A and 5B). The separation between these helices arises from M4 being more perpendicular to the membrane than M2 and from a bend in M2 that occurs at Pro143. Pro143 is conserved in most K2P channels, suggesting that analogous openings may be present in other K2P channels.

Fig. 5

Molecular surface and subunit interface. (A) View of the molecular surface, which has been made partially transparent to reveal a ribbon representation of the channel, from the side with the extracellular solution above. One subunit of the dimer is colored blue and the other red. (B) Orthogonal view from the side. Electron density (orange mesh, Fo-Fc simulated annealing omit map calculated from 20 to 3.4 Å resolution and contoured at 3 σ) is present within an opening between the subunits. Stereo representations of (A) and (B) are shown as supplementary information (fig. S5, A and B). (C) Cutaway view of the molecular surface viewed from the intracellular side, with one subunit colored blue and the other red. The intracellular-most K+ ion in the selectivity filter is depicted as a green sphere (center). Electron density (orange mesh), calculated as in (B), is shown for what is interpreted as alkyl chains (yellow sticks).

Electron density maps revealed tubes of density within the openings just below the selectivity filter (Fig. 5, B and C). The density is consistent with two alkyl chains of about 11 carbons each. The alkyl chains could arise from detergent used in purification or from bound lipids that copurified with K2P1. Two lines of evidence support the binding of alkyl chains within the openings. First, the recently determined crystal structure of a prokaryotic voltage-dependent Na+ channel contains analogous openings (termed fenestrations) (44). The fenestrations were filled with electron density that could be attributed to lipid alkyl chains, suggesting that this may be a property common to other channels. Second, lipids are known to affect TREK-1 channel function, and residues that play a role in lipid sensing correspond to positions along the C helix of K2P1, which is located directly below the openings (912, 45). The openings may provide a way for hydrophobic K+ channel inhibitors, such as THA, to reach the central cavity via the membrane and suggest that screening of lipophilic compounds may be an effective strategy for ion channel drug discovery. The openings create a physical connection between the membrane and the ion pore and thus may be a means of coupling properties of the membrane such as lipid composition or mechanical tension with K2P channel function.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1

References (4662)

References and Notes

  1. Acknowledgments: We thank the staff of beamlines X25 and X29 (National Synchrotron Light Source, Brookhaven National Laboratory) for assistance at the synchrotron; C. D. Lima, N. P. Pavletich, Z. Lu, and members of the Long laboratory for helpful discussions; and N. P. Pavletich, M. B. Long, and J. P. Greenfield for comments on the manuscript. This work was supported in part by a Burroughs Wellcome Career Award in the Biomedical Sciences and by a Gerstner Young Investigator Award to S.B.L. Coordinates and structure factors have been deposited in the Protein Data Bank (PDB ID 3UKM).
View Abstract

Navigate This Article