Crystal Structure of the Mammalian GIRK2 K⁺ Channel and Gating Regulation by G-Proteins, PIP₂ and Sodium

Summary

G-protein-gated K⁺ channels (Kir3.1–Kir3.4) control electrical excitability in many different cells. Among their functions relevant to human physiology and disease, they regulate the heart rate and govern a wide range of neuronal activities. Here we present the first crystal structures of a G-protein-gated K⁺ channel. By comparing the wild-type structure to that of a constitutively active mutant, we identify a global conformational change through which G-proteins could open a G-loop gate in the cytoplasmic domain. The structures of both channels in the absence and presence of PIP₂ show that G-proteins open only the G-loop gate in the absence of PIP₂, but in the presence of PIP₂ the G-loop gate and a second inner helix gate become coupled, so that both gates open. We also identify a strategically located Na⁺ ion-binding site, which would allow intracellular Na⁺ to modulate GIRK channel activity. These data provide a mechanistic description of multi-ligand regulation of GIRK channel gating.

Introduction

G-protein-gated K⁺ (GIRK) channels are members of the inward rectifier (Kir) channel family, so named because the outward flow of K⁺ ions is inhibited by intracellular polyamines and Mg²⁺, which block the pore in a voltage-dependent manner. Kir channels play an essential role in many physiological processes including neuronal signaling, kidney function, insulin secretion and heart rate control. Mutations of Kir channels underlie numerous diseases including primary aldosteronism, Andersen syndrome, Bartter syndrome, and congenital hyperinsulinism (Choi et al., 2011; Hibino et al., 2010).

All Kir channels share the same basic topology: four subunits combine to form a canonical K⁺ pore-forming transmembrane domain (TMD) and a large cytoplasmic domain (CTD) (Fig. 1A). It is thought that ion conduction may be regulated by two gates in series: one is formed by the inner helices of the TMD (Doyle et al., 1998; Jiang et al., 2002), and the other by the G loop at the apex of the CTD (Nishida et al., 2007; Pegan et al., 2005). Various regulatory molecules are thought to control these gates, but the control mechanisms are still unknown.

Figure 1. Structure and function of the GIRK2 channel.

(A) Cartoon diagram of the GIRK2 structure. Each subunit of the tetramer is a different color. Unmodeled segments of the turret and N-terminal linker are drawn with dashed lines. The approximate boundary of the phospholipid bilayer is indicated by the thick black lines.

(B) A cartoon diagram of key residues that mediate the contacts at the interface between the cytoplasmic and transmembrane domains. The same coloring scheme as in panel A is used.

(C) Representative example of a two-electrode voltage-clamp recording from Xenopus laevis oocytes expressing the truncated GIRK2 construct used for crystallography, held at −80 mV. The white bar indicates a physiological extracellular solution containing 96 mM NaCl and 2 mM KCl, whereas the gray bars represent a solution of 98 mM KCl only. The application of 10 μM acetylcholine (ACh) or 1 μM tertiapin-Q (TPN-Q) is also indicated. The dashed line represents zero current – all traces under this line represent negative, inward currents.

(D) Voltage ramps of oocytes expressing the same truncated GIRK2 construct. Currents were measured using two-electrode voltage clamp with an extracellular solution of 98 mM KCl, in the presence of 1 μM ACh (total current, red) or 1 μM ACh + 100 nM TPN-Q (background current, green). GIRK2-specific current (total minus background) is shown as the blue trace. The inset graph is the same data, but with a different scale for the axes to illustrate the degree of rectification.

See also Figure S1 and Table S1.

The anionic lipid phosphatidylinositol 4,5-bisphosphate (PIP₂) is essential for the activation of all Kir channels (Hibino et al., 2010). GIRK channels (Kir3.x) are unique in that they also require G proteins, in combination with PIP₂, for activation (Huang et al., 1998; Logothetis et al., 1987; Reuveny et al., 1994; Sui et al., 1998; Wickman et al., 1994). Certain GIRK channel subtypes are also modulated by intracellular Na⁺ ions (Ho and Murrell-Lagnado, 1999a, b; Lesage et al., 1995; Sui et al., 1996; Sui et al., 1998). GIRK channel activation elicits the flow of K⁺ ions across the cell membrane and thus drives the membrane voltage towards the Nernst potential for K⁺. Near the K⁺ Nernst potential, voltage-dependent Na⁺ and Ca²⁺ channels tend to be silenced and therefore electrical excitation is diminished. This is an important signaling mechanism by which hormone and neurotransmitter stimulation of G protein coupled receptors (GPCRs) regulates many essential physiological processes (Luscher and Slesinger, 2010). For example, acetylcholine secreted by the vagus nerve controls heart rate through stimulation of muscarinic GPCRs in cardiac pacemaker cells (Logothetis et al., 1987; Pfaffinger et al., 1985).

Electrophysiological studies have sought to understand how the various ligands – G proteins, PIP₂, and Na⁺ – interact simultaneously with GIRK channels to regulate their gating. These studies suggest that G proteins and Na⁺ function in a co-dependent manner with PIP₂ to open GIRK channels (Huang et al., 1998; Sui et al., 1998; Zhang et al., 1999). Here we present crystal structures of a quiescent, closed GIRK2 channel and of a point mutant that is constitutively active, independent of G protein stimulation. Further, both structures are determined in the absence and presence of PIP₂. These structures render a molecular mechanistic description of multi-ligand regulation of GIRK channels.

Results

Closed structure of GIRK2

Four GIRK channel isoforms (Kir3.1 – Kir3.4) associate into various homo/hetero-tetrameric complexes. GIRK2 (Kir3.2) forms functional homotetramers and is thus a good candidate for crystallographic structure determination (Kofuji et al., 1995). To obtain suitably diffracting crystals, we modified the mouse GIRK2 cDNA to remove unstructured regions of the N- and C-termini (Nishida et al., 2007; Nishida and MacKinnon, 2002; Pegan et al., 2005; Tao et al., 2009). The resulting channel differs from the corresponding human ortholog by only one amino acid near the structured end of the C-terminus (Asn377 is Ser in human GIRK2) (Figure S2). This channel with unstructured regions of the N- and C-termini removed, which we refer to as wild type in this study, exhibits the fundamental characteristics of the full-length GIRK2: G protein activation, inhibition by tertiapin-Q, and a strongly rectifying current-voltage curve (Figure 1C,D).

Wild type GIRK2 crystals diffracted X-rays to 3.6 Å resolution. Initial phases were determined by molecular replacement using a GIRK2 CTD structure and a model was built and refined to working and free residuals (R_w/R_f) of 26.0/27.3% (Figure 1A,B and Table S1) (Inanobe et al., 2007). The overall architecture of GIRK2 is similar to the G protein-independent ‘classical inward rectifier’ channel Kir2.2 (Tao et al., 2009), but has two significant differences. First, the turrets surrounding the extracellular entryway to the pore form a wider, more open vestibule in GIRK2 (Figure S1B,C). This structural difference may provide a simple explanation for pharmacological differences between classical inward rectifiers and GIRK channels. Many GIRK channels, including GIRK2, are inhibited by certain pore-blocking toxins such as tertiapin, as shown in Figure 1C, whereas classical inward rectifier channels are not (Jin and Lu, 1999). The more open turrets in GIRK2 would allow tertiapin to fit into the vestibule, whereas the more restrictive turrets in classical inward rectifiers appear to prevent toxin binding. The second structural difference occurs at the interface between the TMD and CTD. In Kir2.2 the CTD is extended away from the TMD, whereas in GIRK2 the two components are tightly juxtaposed (Figure 1A,B and Figure S1A).

The TMD-CTD interface in GIRK2 is mediated by both hydrophilic and hydrophobic interactions between the interfacial helices of the TMD, the TM-CTD linker, and the βC-βD loop of the CTD (Figure 1B). These interactions were absent in the more extended Kir2.2 structure (Tao et al., 2009). It seems likely that they play an important role in the control of GIRK2 channel activity because they are in close proximity to the two constrictions along the ion pathway that have been hypothesized to function as gates. One gate - the inner helix gate - is formed by the inner helices of the TMD, just inside the membrane, above the level of the interfacial helix (Figure 1A and Figure S1A). Another gate - the G-loop gate - is formed by the G-loop at the apex of the CTD, just outside the membrane, below the level of the interfacial helix (Figure 1A and Figure S1A). In this structure of GIRK2, both gates are tightly closed.

Binding Site for Regulation by Intracellular Sodium Ions

Near the TMD-CTD interface, immediately beneath the βC-βD loop, there is electron density that cannot be attributed to protein atoms (Figure 2A). Given the surrounding protein chemical groups, this density most likely represents either a metal ion, for example Na⁺, or a water molecule. To distinguish between these possibilities we crystallized the channel in the presence of Tl⁺, a monovalent metal ion identifiable by its X-ray anomalous signal, which has been used previously to analyze the Na⁺ binding site in the transport protein LeuT (Boudker et al., 2007). An anomalous peak, the third strongest in the anomalous difference electron density map (4.6 σ, after sites in the selectivity filter and at the intracellular pore entryway), identifies this extra density as a metal cation (Figure 2B). In the native structure, the density is undoubtedly due to a Na⁺ ion coming from the crystallization solution of 1M NaCl.

Figure 2. The Na⁺ binding site in GIRK2 is located between the βC-βD and βE-βG loops and is dependent on the negatively charged residue Asp228.

(A) The key side chain and main chain atoms involved in coordinating a Na⁺ ion (purple sphere) are shown. A weighted 2F_o-F_c Na⁺-omit electron density map is shown as a blue mesh, contoured at 1.2 σ.

(B) A Tl⁺ anomalous difference electron density map is shown at the same site, contoured at 4 σ (magenta mesh). The data are derived from WT GIRK2 crystals grown in KNO₃ and then soaked in TlNO₃, all in the absence of any Na⁺.

(C) The same site is shown for the structure of a D228N mutant. Electron density is also shown at the same contour level as for wild-type, showing the lack of density for Na⁺.

(D) A model of the whole channel is shown. Each subunit is a different color with the front subunit removed for clarity. The Na⁺ ions are again depicted as purple spheres to show their location relative to the whole channel. The region of the channel shown in panels A-C is highlighted by the black box.

See also Figure S2.

The discovery of a Na⁺ binding site is interesting because GIRK channels that contain either Kir3.2 or Kir3.4 subunits are known to be activated by elevated levels of intracellular Na⁺ with an EC₅₀ of 30–40 mM (Ho and Murrell-Lagnado, 1999a, b; Lesage et al., 1995; Sui et al., 1996; Sui et al., 1998; Zhang et al., 1999). Sodium activation of GIRK channels is thought to serve an important physiological function by producing negative feedback on excessive electrical excitability: during excitation, Na⁺ entry can elevate intracellular Na⁺ concentrations above the normal range of 5–15 mM, enough to activate GIRK channels and drive the membrane potential negative again. Mutagenesis studies have pinpointed an aspartic acid in the βC-βD loop as a critical determinant of Na⁺ activation (Asp228 in Kir3.2), and the presence of aspartic acid rather than asparagine accounts for the Na⁺ sensitivity of GIRK channels with Kir3.2 and Kir3.4 subunits (Ho and Murrell-Lagnado, 1999a, b; Zhang et al., 1999) (Figure S2). Asp228 is one of the coordinating residues of the Na⁺ site in the crystal structure and when Asp228 is mutated to asparagine, the Na⁺ density is no longer observed (Figure 2C). This direct correlation between structure and function observed when mutating Asp228 suggests that we have identified the regulatory Na⁺ binding site. In the refined model, Na⁺ is coordinated not only by the side chain carboxylate of Asp 228, but also by main chain carbonyl oxygen atoms from Arg230, Asn231 and Ser232 (Figure 2A). Main chain carbonyls from Leu275 and Val276 in the βE-βG loop may also participate, as well as the flanking histidines (His69 and His233), which may help to coordinate a water molecule near this site. As we present additional data below it will become clear why this Na⁺ ion is strategically located in the channel to modulate gating.

The R201A Mutant Appears to Stabilize the G Protein Activated State

We suspected that the TMD-CTD interface likely plays an important role in controlling the channel’s gates, and in particular in transmitting conformational changes that allow G proteins to regulate the gates. Using the closed structure of GIRK2 to guide our experiments, we introduced point mutations into the TMD-CTD interface and assessed their effects using a functional assay developed by Jan and coworkers (Kubo et al., 1993). In this assay, the GIRK2 channel was co-expressed in Xenopus oocytes with the M2 muscarinic G protein coupled receptor (Kubo et al., 1993). Stimulation of the M2 receptor by acetylcholine (ACh) causes GIRK2 channel opening, mediated by endogenous G proteins in the oocyte. When the wild type GIRK2 channel is expressed, acetylcholine application stimulates large K⁺ currents (Figure 1C). Subsequent application of tertiapin-Q (TPN-Q) distinguishes GIRK2 currents from endogenous oocyte channel currents. For the wild type channel, the magnitude of current stimulated by acetylcholine nearly equals that inhibited by TPN-Q, which means there is little or no G protein-independent GIRK2 current (Figure 1C).

Mutations in the TMD-CTD interface were introduced at many of the most conserved positions and in a number of cases the mutations we chose corresponded to naturally occurring mutants that underlie channelopathies (Decher et al., 2007; Donaldson et al., 2003; Lin et al., 2006; Plaster et al., 2001; Zhang et al., 1999) (Figure S2). Not surprisingly, the majority of these mutations yielded nonfunctional GIRK2 channels (listed in Figure 3 legend). The mutations that did express are shown in Figure 3A. Grey bars show that functional expression levels in each case are comparable to wild type, while black bars show the fraction of GIRK2 current that is stimulated by the addition of acetylcholine. This fraction is near unity for wild type and four of the mutants (Figure 3A). In contrast, the R201A mutation stands out because its acetylcholine-stimulated fraction is less than 0.2, which means that its activity is largely independent of acetylcholine (Figure 3B).

Figure 3. The R201A mutant is constitutively active, which is partially the result of a rearrangement of the βC-βD loop and a widening of the G loop gate.

(A) A summary of oocyte currents from two-electrode voltage clamp experiments for various GIRK2 mutants. The experiments were performed the same as in Figure 1C. Mutations were selected based on the interactions shown in Figure 1B. Specific GIRK2 current (gray bars) is evaluated as the total ACh-stimulated current minus the remaining current after TPN-Q blockage. Fraction ACh stimulation (black bars) is evaluated as the ACh-stimulated current (the difference in the current before and after the addition of ACh) divided by the total GIRK2-specific current. (*) All currents were measured 16–24 hrs after RNA injection, except for Y78L and R201A, which did not show appreciable current until after 3 days of expression. All RNAs were injected undiluted, except for the D228N mutant, which was diluted 10-fold. Error bars represent the standard deviation of the mean from 3 different oocytes for each mutant. Data are not shown for the following mutants that did not show any detectable currents: D81Y/N/R/A/E, Y78D/A, R230D/K/A, R201K/D/Q, D228R/E/A, R201D/D228D, R230D/D81R, D81Y/Y78D.

(B) Representative example of a two-electrode voltage-clamp recording of Xenopus laevis oocytes expressing the R201A GIRK mutant. The experiment was performed exactly as in Figure 1C.

(C) A comparison of the wild-type (red with orange side chains) and R201A mutant structures (green).

(D and E) A comparison of the conformation of the G loop gate between the wild-type (red, D) and R201A (green, E) structures. The top panels are a cross-section of the channel, showing the G loops of just two opposing subunits. Key residues that form the main constriction in the ion conduction pathway are highlighted (there was no electron density for M319 in (E), so it was modeled as alanine). The bottom panels show a top-down view of the cytoplasmic domain tetramer with the side chains that were highlighted in the top panels shown as space-filling models.

To analyze the structural alterations underlying constitutive activation in the R201A mutant, we determined its crystal structure at 3.1 Å resolution. Compared to wild type, this channel shows a large rearrangement of the βC-βD loop, movement of the His233 side chain to fill the void left by deletion of the Arg201 side chain, and a shifting of the strand comprising residues 235–237 to interact more closely with residues 272–275 of the βE-βG loop (Figure 3C). These conformational changes are associated with a displacement and rotation of the G loop and a change in the positions of three amino acids that form the gate’s constriction: Met313, Gly318 and Met319 (Figure 3D, E Supplementary Movie 1). The net effect of these concerted changes is a widening of the G loop gate from about 6.0 Å to nearly 12.0 Å, and the appearance of oxygen atoms from the Gly318 main-chain carbonyl and the Thr320 side chain to face the pore (Figure 3D,E). These conformational changes effectively create a 9 Å diameter hydrophilic pore (as delimited by the van der Waals surfaces of the pore-lining oxygen atoms), which should be sufficient to allow passage of an 8 Å diameter hydrated K⁺ ion.

The R201A mutant also shows a propagated conformational change that extends to the perimeter of the CTD, a distance of approximately 30 Å from the site of the R201A mutation, and encompasses the Na⁺ binding site (Figure 4, Supplementary Movie 1). This propagated conformational change is mediated by a domino-like displacement of β strands, which in turn leads to a reorganization of several hydrophobic amino acids (Val67, Leu257, and Val276) and a rotamer switch in the conformation of the Tyr58 side chain near the CTD perimeter. Several independent lines of evidence suggest that this propagated conformational change is similar to the conformational change that occurs when G proteins bind to the channel.

Figure 4. The R201A mutant shows a propagated conformational change through the cytoplasmic domain that mimics the same changes induced by G proteins.

A comparison of the conformational changes over the whole channel between the wild-type (red) and the R201A mutant (green) structures. The front subunit has been removed for clarity. The intensity of the color is related to the absolute conformational change between identical residues (a combination of the distance between both the alpha and gamma carbons of the two structures, to account for both main-chain and rotamer conformational changes). The large deviations in the transmembrane domain arise from a slight rigid-body twist of the cytoplasmic domain relative to the transmembrane domain. The close-up view on the right highlights key side chains involved in propagating the conformational change through the channel. Important ligand-binding sites are identified by blue circles.

See also Figure S3 and Movie S1.

First, site-directed mutagenesis studies of GIRK/Kir2.1 chimeric channels have identified several buried and surface resides as being essential in mediating G protein activation of GIRK channels. The critical surface residues are located around the βL-βM loop of the CTD, especially residue Leu344, thus implicating this region as a binding site for G proteins (Finley et al., 2004; He et al., 1999). Therefore it is reasonable to think that G protein binding to this region could regulate this conformational change, leading to G loop gate opening. It was also shown that mutation of the buried residue Leu273 to isoleucine eliminated G protein activation (He et al., 2002). Until now it was unclear how this would affect channel function, but here we can see that Leu273 is located on the βE-βG loop next to Val276 and that it undergoes a significant conformational change in the R201A mutant. It is conceivable that a mutation at this position could alter the propagation of a conformational change from the surface of the protein to the G loop.

Second, NMR studies of the interaction between G protein subunits and the GIRK1 CTD using transferred cross saturation and chemical shift perturbation experiments have identified regions on the CTD that either directly contact a G protein or undergo a conformational change in response to G protein binding (Yokogawa et al., 2011). These studies concluded that G proteins bind to a surface mainly comprised of the βL-βM and βD-βE loops (Figure S3B). They also concluded that G protein binding elicits conformational changes near this surface as well as at locations within the N-terminus, especially Tyr58 (Figure S3C,D). The NMR data match the conformational changes that we observe in the βL-βM loop, N-terminus, and in particular Tyr58, which is the residue that undergoes a rotamer switch in the R201A mutant (Figure S3A and Figure 4). We will show below that the R201A mutant in the presence of PIP₂ reveals an additional conformational change that is consistent with the NMR data on the βD-βE loop (Figure S3E,F).

Taken together, the mutagenesis and NMR data support the hypothesis that constitutive activation in the R201A mutant channel results because this mutant favors a conformation similar to that induced by G protein binding. The G loop gate is open in the R201A mutant, but the inner helix gate remains closed. We note, however, that G protein stimulation alone is insufficient to achieve ion conduction in GIRK channels. In electrophysiological assays, the signaling lipid PIP₂ is also required, in addition to G proteins, for channel activation.

The Role of PIP₂ in GIRK2 Channel Activation

We next determined the crystal structure of wild type GIRK2 in the presence of C8-PIP₂ (8 carbon acyl chains) at 3.0 Å resolution. The structure shows one ordered PIP₂ lipid molecule per subunit bound near the TMD-CTD interface. The negatively charged phosphates of the PIP₂ molecule are coordinated by several positively charged residues: Lys64, Lys194, Lys199, and Lys200, along with backbone amides at the junction of the interfacial and outer helices (Figure 5, Figure S4A,B). The side chains for residues Lys90 and Arg92 did not show any appreciable electron density, but may still contribute to an overall positive electrostatic potential of the binding site (Figure 5, Figure S4A,B). The presence of PIP₂ induced a modest displacement of the protein main chain near the PIP₂ binding site, where the interfacial helix turns into the outer helix (Figure S4C). PIP₂ also produced a slight rotation of the inner helices, accompanied by a weakening of electron density for the Phe192 side chain, which forms the most constricted region of the inner helix gate (Figure S4D,E,F). Clearly, however, both the G loop gate and inner helix gate remain closed in the presence of PIP₂, consistent with the observation that PIP₂ alone is insufficient to open GIRK channels in electrophysiological experiments.

Figure 5. PIP₂ binds at the interface between the transmembrane and cytoplasmic domains, and is coordinated by several positively charged residues.

PIP₂ molecules are colored yellow, orange, and red (for carbon, phosphorus, and oxygen atoms) and are shown in the context of the whole channel (gray) on the left, and a closeup view on the right. The thick black lines indicate the approximate boundary of the plasma membrane and the black box highlights the region of the close-up view on the right. On the right, the main coordinating residues are shown as sticks. Residues Lys90 and Arg92 were modeled as alanines due to a lack of electron density, but probably still contribute to the positive electrostatics of the binding site. The important gating residue Phe192 is also shown for reference.

See also Figure S4.

In contrast to the small effect of PIP₂ on the conformation of the wild type channel, PIP₂ had a profound effect on the R201A mutant. A crystal structure in the presence of C8-PIP₂, determined at 3.5 Å resolution, shows a large change in both the G loop and the inner helix gates (Figure 6A,B, Supplementary Movie 2). Because of the way in which channel molecules are packed against each other in this orthorhombic crystal, only two of the four subunits are free to bind PIP₂ and undergo the conformational change (Figure S5). In the PIP₂ bound subunits, we observe conformational changes similar to those observed in the R201A mutant. We also observe a rigid body rotation of the CTDs, which necessitates a movement of the βD-βE loop (Figure 6C,D, Figure S3E,F). The rotation of the CTDs is propagated across the TMD-CTD interface and is associated with a rotation and splaying apart of the inner helices. The net effect of PIP₂ in combination with the R201A mutation is an opening of the G loop gate to 15 Å and the inner helix gate to 11 Å (Figure 6A,B, Supplementary Movie 2).

Figure 6. PIP₂ binding to the R201A mutant causes the inner helix and G loop gates to open.

(A) The inner helix gate is shown as a Cα-trace for the apo structure of R201A (thin, green ribbons) and the structure of the R201A mutant with PIP₂ (thick, blue and purple ribbons). The side chain of Phe192 is also shown. For the R201A + PIP₂ structure, the subunits that bound PIP₂ are shown in blue, whereas the PIP₂-free subunits are shown in purple.

(B) The same coloring scheme is used as in panel A, except now the G loop gate is highlighted.

(C) One subunit of R201A + PIP₂ is shown (blue) compared to the apo structure of R201A (green). The residues for the βL-βM loop are hidden to more clearly show the rigid-body twisting of the cytoplasmic domain. The PIP₂ molecule is shown in ball and stick representation.

(D) A top down, orthogonal view of panel C. Only the cytoplasmic domain is shown to more clearly show the rigid-body twisting.

See also Figures S3 and S5, and Movie S2

Figure 7A shows a surface rendering of the entire pore lining for the wild type channel and for the R201A mutant in the presence of PIP₂. Although we do not know for sure whether in a cell two or four PIP₂ molecules bind to the channel in its G protein-activated state, this rendering shows the case in which we allow four subunits to bind PIP₂ and undergo the conformational change, which we anticipate would likely occur in the unconstrained environment of the membrane. It is evident that in the setting of dual activation by G protein subunits (mimicked by the R201A mutant) and PIP₂, the highly constricted pore of the quiescent wild type channel opens wide enough to allow a hydrated K⁺ ion passage from the cytoplasm to the selectivity filter.

Figure 7. The R201A mutant in the presence of PIP₂ creates a continuous cavity from the cytoplasm to the selectivity filter large enough for a hydrated K⁺ ion.

(A) Surface representations for the interior surfaces of the wild-type apo (left) vs. a four-fold symmetrical model of R201A + PIP₂ (right). The K⁺-accessible space was determined with the HOLLOW script (Ho and Gruswitz, 2008). The R201A + PIP₂ model was generated by rotating the PIP₂-bound subunits 90 degrees about the central tetrameric axis. The surface of the cavities are colored according to the hydrophobicity of the surrounding side chains. For reference, a cartoon representation of the channel is shown on the far left. The key residues of the inner helix (Phe192) and G loop gates (Met313 and Met319) are shown in cyan.

(B) A schematic model depicting the effects that PIP₂ and the R201A mutation have on GIRK2 channels. The three main constriction points are labeled in the top-left panel: a. selectivity filter, b. inner helix gate, c. G loop gate. PIP₂ binding to the wild type channel opens the inner helix gate slightly, but the G loop gate is still closed (top right). The R201A mutant induces a series of conformational changes that mimic the effect of G protein binding – this results in the opening of the G loop gate, but the inner helix gate is still closed (bottom left). The combination of the R201A mutant and PIP₂ causes a rotation of the CTD, which in turn splays apart the inner helix gate and further opens the G loop gate (bottom right). The net effect of these conformational changes is an open channel.

See also Figure S6.

Discussion

This study presents the first molecular structures of a G protein gated K⁺ channel, GIRK2. Crystallization of the wild type channel and of a constitutively active mutant, both in the absence and presence of the signaling lipid PIP₂, reveal four distinct structures that we believe represent physiologically relevant conformations that underlie GIRK channel gating. In contrast to voltage-dependent and other ligand-gated K⁺ channels, GIRK channels clearly contain two functional gates – an inner helix gate and a G loop gate – that are regulated by a combination of cytoplasmic and membrane stimuli.

Until this study, all previous structures of Kir family channels including several bacterial family members, a chimera with a bacterial TM and eukaryotic CTD, and a eukaryotic ‘classical inward rectifier’, exhibited a tightly closed inner helix gate (Clarke et al., 2010; Kuo et al., 2003; Nishida et al., 2007; Tao et al., 2009). The conformation of the inner helix gate in three structures presented here – wild type without PIP₂, wild type with PIP₂, and the R201A mutant without PIP₂ – have a closed inner helix gate conformation similar to previous structures. In the R201A mutant in the presence of PIP₂ we observe an open inner helix gate conformation. As discussed above, we think it is likely that crystal lattice contacts prevented PIP₂ binding and opening in two of the subunits, but that in a membrane all four subunits would bind PIP₂ and undergo the conformational change. If four subunits undergo the change, then the inner helix gate will open wide enough to permit passage of a hydrated K⁺ ion, but not as wide as in voltage-gated K⁺ and MthK Ca²⁺ gated K⁺ channels (Jiang et al., 2002; Jiang et al., 2003; Long et al., 2005; Long et al., 2007). A narrower passageway from the cytoplasm to the selectivity filter in inward rectifier K⁺ channels might be functionally significant because strong voltage-dependent block by intracellular polyamines and Mg²⁺ give rise to their namesake conduction property – inward rectification (Lopatin et al., 1994; Matsuda et al., 1987). The strong voltage dependence of these blocking ions is in part due to the coupled movement of blocking and permeant ions in the pore (Spassova and Lu, 1998). A very wide ion pathway would allow blocking and permeant ions to interchange their positions along the pathway (i.e. slip by each other), but a narrower pathway will force the ions to move in a queue, and thus the movement of K⁺ will impart excess voltage dependence onto the blocking ion and give rise to stronger rectification (Spassova and Lu, 1998).

We discovered the constitutively active R201A mutant by evaluating the initial wild type closed structure and then perturbing the TM-CTD interface because it appeared as if it ought to be involved in regulating the gates. Not surprisingly, most mutations in this region abolished channel function. Fortuitously, in the R201A mutant we observe constitutive activation associated with an open G loop gate and a propagated conformational change across the CTD. The extent and distribution of this conformational change matches very well previous mutational studies implicating the βLβM loop as a site for binding G protein subunits, and NMR studies on the effect of G protein binding to isolated CTD structures (Finley et al., 2004; He et al., 2002; He et al., 1999; Yokogawa et al., 2011). We thus hypothesize that the R201A mutant stabilizes a conformation that is similar to the G protein activated state.

Electrophysiological experiments have shown that elevated levels of Na⁺, along with the presence of PIP₂, can activate certain GIRK channels that have an aspartic acid residue at position 228 (Ho and Murrell-Lagnado, 1999a, b; Lesage et al., 1995; Sui et al., 1996; Sui et al., 1998; Zhang et al., 1999). Our data may start to address the mechanism of this mode of activation. We show that a Na⁺ ion binds in a pocket between the βC-βD and βE-βG loops. This lies directly on the pathway of the propagated conformational changes that we observed in the R201A structure, which we believe to mimic the effects of a bound G protein. Thus, Na⁺ may work to influence the conformation of this pathway. One possible mechanism by which the Na⁺ ion could do this is by binding to Asp228 and thereby weakening the ionized hydrogen bond that it forms with Arg201. Through such an interaction we imagine that Na⁺ could in part mimic the effect of the R201A mutation and, by extension, the effect of G proteins as well.

There is a very good correlation between the four crystal structures and past electrophysiological studies showing a dual requirement for both G proteins and PIP₂ to open GIRK channels (Huang et al., 1998; Sui et al., 1998; Zhang et al., 1999). We can now view this dual requirement through a thermodynamic cycle of the four structures: PIP₂ alone binds but does not open either gate, G proteins (R201A) alone open the G loop gate but not the inner helix gate, but G proteins in addition to PIP₂ open both gates (Figure 7B, Figure S6). In words, the relations suggest that PIP₂ functions to couple tightly the two gates so that upon G protein stimulation both gates open to activate the channel.

One well established role of the CTD is to form an extended pore that is important for producing rectification (Nishida and MacKinnon, 2002; Tao et al., 2009; Yang et al., 1995). In this study we observe how the CTD also has the ability to undergo conformational changes that allow cytoplasmic ligands to allosterically control gates in the pore. In realizing that G proteins apparently regulate the gates through a pathway of conformational change that includes a modulatory Na⁺ binding site, we have begun to see the structural underpinnings of a very complex form of gating regulation.

Experimental Procedures

Molecular Biology

A truncated GIRK2 cDNA (consisting of residues 52–380 – hereafter referred to as wild type) was subcloned by PCR into pPICZ and pGEM vectors for expression in Pichia pastoris, or Xenopus laevis oocytes, respectively. The ORF in the pPICZ vector contained a C-terminal PreScission protease site, followed by green fluorescent protein (GFP), and a His10 tag. The pGEM vector contained a C-terminal FCYENE tag (Ma et al., 2001)) and then a c-Myc tag.

Protein Expression and Purification

Protein expression was induced in stable P. pastoris cell-lines by the addition of methanol for 20–24 hrs at 24 °C. Cells were harvested by centrifugation, frozen in liquid N₂, and stored at −80 °C until needed. Frozen cells were lysed in a mixer mill, then solubilized for 1 hour at room temperature (RT) by resuspending in 50 mM Hepes pH 7.35, 150 mM KCl, 4 % (w/v) n-decyl-β-D-maltopyranoside (DM), and protease inhibitor cocktail. Clarified supernatant was then incubated with Talon metal affinity resin for 1 hr at RT with gentle mixing. The resin was washed in batch with 5 column volumes (cv) of Buffer A (50 mM Hepes pH 7.0, 150 mM KCl, 0.4 % (w/v) DM, then loaded onto a column and further washed with 5 cv Buffer A + 40 mM imidazole, then 2 cv Buffer A + 80 mM. The column was then eluted with Buffer A + 300 mM imidazole. Peak fractions were pooled and 20 mM DTT, 3 mM TCEP, 1 mM EDTA were added, and then cut with PreScission protease for either 2 hrs at RT or overnight at 4C. The cleaved protein was then concentrated to run on a Superdex-200 gel filtration column in 20 mM TRIS-HCl pH 7.5, 150 mM KCl, 0.2 % (w/v) DM (anagrade), 20 mM DTT, and 1 mM EDTA.

Crystallization & Structure Determination

Purified protein representing the peak tetramer fractions was pooled and concentrated in a 50 K MWCO concentrator to 6–7 mg/mL, mixed 1:1 with crystallization solution, then crystallized using the hanging drop vapor diffusion method. For the PIP₂ complexes, 1,2-dioctanoyl-sn-glycero-3-phospho-(1′-myo-inositol-4′,5′-bisphosphate) (C8-PIP₂, Avanti Polar Lipids) was added to the concentrated protein at a final concentration of 2 mM right before setting up the drops. For R201A + PIP₂, 10 mM spermine was also added, although it was not seen in the electron density maps. Crystallization drops were incubated at 20 °C and crystals (D-shaped plates) usually appeared after 1–3 days. The wild type and R201A mutant crystals grew in 50 mM Na citrate pH 6.0, 1 M NaCl, 30–35 % PEG 400. The D228N mutant crystals grew in 50 mM Na citrate pH 6.0, 1 M NaNO₃, 24–26 % PEG 400. The wild type + PIP₂ crystals grew in 50 mM Na citrate pH 6.0, 1 M NaCl, 20 % PEG 400. The R201A mutant + PIP₂ crystals grew in 50 mM Na Hepes pH 7.25, 0.5 M NaCl, 25 % PEG 400. For collecting Tl⁺ anomalous diffraction data, wild type protein was crystallized in the presence of KNO₃ instead of NaCl. These crystals were then transferred to solutions containing TlNO₃. For data collection, the crystals were cryoprotected and flash-frozen in liquid N₂. The structures were solved by molecular replacement with the MOLREP (Vagin and Teplyakov, 2000) program, using the GIRK2 cytoplasmic domain structure as a search model (Inanobe et al., 2007). The models were built using COOT (Emsley and Cowtan, 2004) and refined with REFMAC (Murshudov et al., 1997). Data collection and refinement statistics are shown in Table S1.

Electrophysiology

X. laevis oocytes were injected with 50 nL cRNA and incubated for 1–3 days before recording currents. All experiments were performed at room temperature using two-electrode voltage-clamp. For gap-free recordings, oocytes were held at −80 mV in ND96 (96 mM NaCl, 2 mM KCl, 0.3 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, pH 7.6 with KOH), then perfused with a highK solution (98 mM KCl, 0.3 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, pH 7.6 with KOH) that also contained either 10 μM acetylcholine (ACh) or 1 μM tertiapin-Q (TPN-Q). GIRK2-specific current was calculated by subtracting the current remaining after TPN-Q blockage (background oocyte currents) from the maximum current elicited with highK + ACh. Basal GIRK2 current was calculated by subtracting the current remaining after TPN-Q blockage from the maximum current elicited with highK only. For measuring rectification, the membrane potential was ramped from −80 mV to +80 mV over 100 ms in the presence of either highK buffer + 1 μM ACh to measure total currents, or highK + 1 μM ACh + 100 nM TPN-Q to measure background oocyte currents. GIRK2-specific currents were calculated by subtracting the background currents from the total currents.