Structure of the C-terminal region of an ERG channel and functional implications

Abstract

The human ether-à-go-go–related gene (hERG) encodes a K⁺ channel crucial for repolarization of the cardiac action potential. EAG-related gene (ERG) channels contain a C-terminal cyclic nucleotide-binding homology domain coupled to the pore of the channel by a C-linker. Here, we report the structure of the C-linker/cyclic nucleotide-binding homology domain of a mosquito ERG channel at 2.5-Å resolution. The structure reveals that the region expected to form the cyclic nucleotide-binding pocket is negatively charged and is occupied by a short β-strand, referred to as the intrinsic ligand, explaining the lack of direct regulation of ERG channels by cyclic nucleotides. In hERG channels, the intrinsic ligand harbors hereditary mutations associated with long-QT syndrome (LQTS), a potentially lethal cardiac arrhythmia. Mutations in the intrinsic ligand affected hERG channel gating and LQTS mutations abolished hERG currents and altered trafficking of hERG channels, which explains the LQT phenotype. The structure also reveals a dramatically different conformation of the C-linker compared with the structures of the related ether-à-go-go–like K⁺ and hyperpolarization-activated cyclic nucleotide-modulated channels, suggesting that the C-linker region may be highly dynamic in the KCNH, hyperpolarization-activated cyclic nucleotide-modulated, and cyclic nucleotide-gated channels.

The human ether-à-go-go-related gene (hERG) ion channel underlies the fast delayed rectifier current (I_Kr) in the heart and is the major contributor to the repolarization phase of the ventricular action potential (1–4). Disruption of the function of this channel, either genetically or as an unintended consequence of prescription medication, causes lengthening of the ventricular action potential, which can lead to a condition known as long-QT syndrome (LQTS) (5–8). LQTS is characterized by a prolonged QT interval on an electrocardiogram, cardiac arrhythmias, and predisposition to sudden cardiac death. hERG channel dysfunction is responsible for ∼45% of LQTS-related mutations identified so far and the vast majority of drug-induced LQTS (9–11).

The EAG-related gene (ERG) channel subfamily belongs to the KCNH family of voltage-gated K⁺ channels, which also includes the ether-à-go-go (EAG) and EAG-like K⁺ (ELK) channel subfamilies (Fig. 1A) (12). EAG and ELK channels are key regulators of tumor progression (13, 14) and neuronal excitability (15, 16). Similar to other K⁺ selective channels, KCNH channels are assembled from four subunits around a central ion conducting pore. Each of the four subunits contains six membrane spanning segments (S1–S6) and an intervening pore-forming loop (Fig. 1B). The S1–S4 segments comprise a voltage sensing domain, whereas the S5–S6 segments together with the intervening loop form a pore domain (17).

Fig. 1.

Topology and similarity of KCNH channels and the structure of the agERG C-linker/CNBHD. (A) Phylogenetic tree of the KCNH family of ion channels computed with Cobalt. (B) Cartoon of two opposing subunits of a tetrameric KCNH channel. The pore-forming loop and S5–S6 transmembrane domains are gray. The N-terminal α-helix and PAS domain are magenta. The elbow and shoulder regions of the C-linker are represented by the red and pink cylinders, respectively. The β9-strand is represented by a green rectangle. The rest of the CNBHD is blue. The dashed lines connecting the PAS domain and S1 transmembrane domain indicate intersubunit nature of the interactions between the PAS domain of one subunit and the CNBHD of the adjacent subunit (28). (C) Ribbon representation of the structure of the C-linker/CNBHD of agERG channels. (D) Electrostatic potential surface of the C-linker/CNBHD of agERG channels viewed in the same orientation as in C. The electrostatic potential surface was calculated by using the APBS plugin for PyMol with the PARSE force field and colored from red (−3 kT/e) to blue (+3 kT/e).

A characteristic feature of the channels in the KCNH family is the presence of a Per-Arnt-Sim (PAS) domain in their cytoplasmic amino-terminal region, and a cyclic nucleotide-binding homology domain (CNBHD) in their cytoplasmic carboxyl-terminal region coupled to the pore of the channel by a C-linker (12, 18–22) (Fig. 1B). The CNBHD of KCNH channels shares general architecture with the cyclic nucleotide-binding domains (CNBD) of related cyclic nucleotide-gated (CNG) and hyperpolarization-activated cyclic nucleotide-modulated (HCN) channels (12, 19). However, unlike the CNBD of CNG and HCN channels, the CNBHD of KCNH channels does not appreciably bind cyclic nucleotides (18, 23). Therefore, paradoxically, despite the presence of a CNBHD, the KCNH channels are not regulated by direct binding of cyclic nucleotides (23–26). Based on Förster resonance energy transfer (FRET) experiments, the CNBHDs and PAS domains are located in close proximity (27). Moreover, it has been proposed that the C-linker/CNBHD directly interacts with the PAS domain (28–30), forming an intersubunit interaction (28). In hERG channels, both the C-linker/CNBHD and PAS domain harbor genetically occurring mutations that are linked to LQTS (11, 31–34) (reviewed in ref. 32). These LQTS mutations have been found to speed up deactivation of hERG channels or lead to defects in protein trafficking and folding, precluding formation of functional channels (35–43).

Here, we present the crystal structure of the C-linker/CNBHD of an Anopheles gambiae ERG channel (agERG) at 2.5-Å resolution. The agERG and hERG channels share 78% amino acid identity and 90% similarity in the C-linker/CNBHD (Fig. 1A and Fig. S1). The structure of the C-terminal region of ERG channels reveals the structural basis of ERG channel regulation by the CNBHD. The C-linker in the ERG structure is in a dramatically different conformation in comparison with the C-linkers in related zELK (18) and mHCN2 (21) channels. The diversity of the C-linker conformations observed in the ERG, ELK, and HCN structures suggests that the C-linker region may be highly dynamic in the KCNH, HCN, and CNG channels.

Results

Structure of the C-Linker/CNBHD of an ERG Channel.

To gain molecular insight into the gating and regulation of ERG channels, we crystallized the C-linker/CNBHD of Anopheles gambiae (mosquito) ERG (agERG) channels. The C-linker/CNBHD of agERG was identified as a suitable candidate for crystallization by using a screen based on the fluorescence-detection size-exclusion chromatography (FSEC) (44). The C-linker/CNBHD of agERG channels crystallized in the space group P3₂21 with a single molecule in the asymmetric unit and diffracted X-rays to 2.5-Å resolution (Table S1). The structure was solved with molecular replacement, using the structure of the CNBHD of zebrafish ELK (zELK) channels (18) as a search model, combined with energy- and electron density-guided structure optimization with Rosetta (45).

The C-linker region in the crystal structure consists of four α-helices (αA′-αD′) with αA′-αB′ and αC′-αD′ helices forming two antiparallel helix-turn-helix motifs (Fig. 1C). This arrangement is markedly different from the topology of the C-linkers of zELK (18) and mHCN2 (21) channels that consist of six α-helices. The CNBHD of agERG is formed by eight β-strands forming an antiparallel β-roll, three α-helices (αA-αC), and a short β-strand (β9) following the αC-helix (Fig. 1C). The structure of the CNBHD is similar to the crystallized CNBHDs from the two other subfamilies of KCNH channels, mouse EAG (mEAG) (rmsd of 1.9 Å) (22) and zELK (rmsd of 2.4 Å) (18) (Fig. 2 A and B). With the exception of the β9-strand, the general fold of the CNBHD is similar to the fold of canonical cyclic nucleotide-binding proteins, including channels, kinases, transcription factors, and guanine nucleotide-exchange factors (46).

Fig. 2.

Structural comparison of the C-linker/CNBHDs of KCNH channels and the close-up view of the agERG intrinsic ligand. (A and B) Structural alignment of the CNBHD of agERG (the β9-strand is green and the rest is blue) and mEAG1 (gray) channels (22) (A) and agERG and zELK (yellow) channels (18) (B). The rmsd for the α carbons of the agERG and mEAG1 structures is 1.9 Å, and the rmsd for the α carbons of the agERG and zELK structures is 2.4 Å. (C) A 2F_o−F_c omit electron-density map of residues Y727, N728, and M729 of the intrinsic ligand bound inside the β-roll cavity of the CNBHD of agERG channels. The map is contoured at 1.0 σ. (D) Residues in the β-roll cavity interacting with residues Y727 and M729 of the intrinsic ligand. Dashed lines show both polar and nonpolar interactions. cAMP from the HCN2 structure (21) structurally aligned with the agERG is shown in yellow.

Unlike canonical cyclic nucleotide-binding proteins, however, the electrostatic profile of the C-linker/CNBHD of agERG channels displays a negatively charged β-roll cavity in the CNBHD (Fig. 1D). The β-roll cavity is the site of cyclic nucleotide binding in canonical cyclic nucleotide-binding proteins (46). In HCN2 channels, which are directly regulated by cyclic nucleotides, the β-roll cavity is positively charged, favoring binding of negatively charged cyclic nucleotides (21). In contrast, for agERG the β-roll cavity is negatively charged (Fig. 1D). The negatively charged β-roll cavity would be expected to impede binding of negatively charged cyclic nucleotides, explaining, in part, the absence of direct regulation of KCNH channels by cyclic nucleotides (23–26).

The β9-strand is also unique to the KCNH family and is not present in other CNBD-containing proteins. It is part of a short sequence of amino acids following the αC-helix that occupies the β-roll cavity where cyclic nucleotides normally bind in canonical cyclic nucleotide-binding proteins (Fig. 2C). We refer to this region as the “intrinsic ligand,” a portion of the protein that occupies the ligand binding site whose displacement regulates the channel. The sequence and structure of the intrinsic ligand is well conserved in the KCNH family of channels (18, 22) (Fig. 2 A and B and Fig. S1). In agERG, residues Y727 and M729 in the intrinsic ligand form a network of direct interactions with the residues in the β-roll cavity of agERG channels (Fig. 2D). Y727 is located where the purine ring of cAMP is located in the CNBD of HCN2 channels, whereas M729 is located where the cyclic phosphate of cAMP is in the HCN2 structure.

Effect of Mutations in the Intrinsic Ligand on hERG Channel Gating.

To investigate whether the intrinsic ligand regulates the gating of ERG channels, we mutated the intrinsic ligand and measured the effects on channel function. Expression of agERG channels in Xenopus oocytes did not produce detectable voltage-activated K⁺ currents so we mutated the intrinsic ligand in hERG channels. In hERG channels, the intrinsic ligand includes conserved residues F860, N861, and L862 (Fig. S1). The mutant channels were expressed in Xenopus oocytes, and currents were recorded by using two-electrode voltage clamp. All of the electrophysiological studies in this section were performed in the background of S620T mutation in the pore region that removes the C-type inactivation in hERG channels (47).

To explore the role of the intrinsic ligand in the function of hERG channels, we examined the effect of mutating the conserved residues to alanine. Currents were elicited by a series of voltage pulses from −100 to 40 mV in 20-mV increments followed by a pulse to −60 mV (or −100 mV; Fig. 3 A–D, Inset). All three mutant hERG channels (F860A, N861A, and L862A) gave rise to robust voltage-activated K⁺ currents in Xenopus oocytes (Fig. 3 A–D). The conductance–voltage relationships for the wild-type (WT) channels and the three mutant channels were generated from the tail currents at −100 mV (Fig. 3 A–D, Inset) and were not significantly different (P > 0.05, ANOVA) from one another [WT: the half-maximal activation voltage (V_1/2) = −18.3 ± 3.2 mV, the slope of the relation (s) = 12.4 ± 2.1 mV; F860A: V_1/2 = −22.5 ± 2.2 mV, s = 11.3 ± 2.4 mV; N861I: V_1/2 = −22.1 ± 1.7 mV, s = 10.5 ± 0 0.9 mV, and L862A: V_1/2 = −20.0 ± 0.7 mV, s = 9.9 ± 0.5 mV) (Fig. 3E).

Fig. 3.

Mutations in the intrinsic ligand have no measurable effect on the conductance–voltage relationship of hERG channels. (A–D) Representative currents recorded from whole oocytes expressing WT (A), F860A (B), N861A (C), and L862A (D) hERG channels elicited by a series of voltage pulses from −100 to 40 mV in 20-mV increments followed by a pulse to −60mV (or −100 mV; Inset). (E) Conductance–voltage relations for WT (filled squares), F860A (open squares), N861A (open triangles), and L862A (open circles) hERG channels measured from the tail currents at −100 mV. Data points were fit with a Boltzmann equation to determine the V_1/2 and s values. The number of cells n ≥ 3 for each of the channels. (Scale bars: 1 μA and 1 s; Inset, 0.5 μA and 0.1 s.)

To investigate the effect of the mutations in the intrinsic ligand on the deactivation kinetics, we examined tail currents (Fig. 4). The tail currents were elicited by first activating the channels with a pulse to 20 mV and then applying a voltage pulse to −100 mV and −120 mV. The time constant of deactivation was determined by fitting the tail currents with a single exponential function (Fig. 4 A–D). The F860A and L862A mutations caused an increase in the rate of deactivation, whereas N861A had a deactivation time constant (τ) that was statistically similar to the WT channels (at −100 mV, WT: τ = 108 ± 13 ms; F860A: τ = 49 ± 5 ms; N861A: τ = 94 ± 5 ms; L862A: τ = 66 ± 3 ms. At −120 mV, WT: τ = 45 ± 4 ms; F860A: τ = 26 ± 4 ms; N861A: τ = 40 ± 2 ms; L862A: τ = 31 ± 2 ms) (Fig. 4E). Residues F860 and L862 in hERG correspond to residues Y727 and M729 in the agERG structure that form multiple direct polar and nonpolar interactions with the residues in the β-roll cavity (Fig. 2D). In contrast, residue N861 in hERG corresponds to N728 in agERG that is facing away from the β-roll cavity and forms virtually no direct interactions with residues on the β-roll. These results indicate that the intrinsic ligand stabilizes the open state of hERG channels, similar to an agonist of a ligand-gated channel. Interestingly, these effects of mutations of the intrinsic ligand are similar to the effects of many LQTS mutations in the amino-terminal PAS domain (32, 35, 36), suggesting that the PAS domain might regulate the channel via the CNBHD and perhaps through the intrinsic ligand.

Fig. 4.

Mutations in the intrinsic ligand accelerate deactivation of hERG channels. Tail currents for WT (A), F860A (B), N861A (C), and L862A (D) hERG channels were elicited by first activating the channels with a pulse to 20 mV and then applying a voltage pulse to −100 mV and −120 mV. (E) Box plot of deactivation time constants at −100 mV and −120 mV for WT, F860A, N861A, and L862A hERG channels. The time constants were derived by fitting the tail currents at −100 mV and −120 mV with a single exponential function as depicted by the overlaid gray traces in A–D. The median is represented by the middle line, the bottom and top lines are the 25th and 75th percentiles, and the vertical lines are the 10th and 90th percentiles. **P < 0.01, *P < 0.05, by ANOVA. n ≥ 4 for each of the channels. (Scale bar: 2 μA and 0.1 s.)

These effects of mutations of the intrinsic ligand are somewhat different from the effects previously seen in zELK and EAG1 channels. For zELK channels, mutation of the tyrosine residue in the intrinsic ligand to alanine or deletion of the conserved YNL tripeptide shift the conductance–voltage relationship to more depolarized voltages (18). Surprisingly, for EAG1 channels, mutation of the conserved residues Y672 and L674 to alanine in the intrinsic ligand shift the conductance–voltage relationship to more hyperpolarized voltages (22, 48). Therefore, although all KCNH channels studied so far are regulated by the intrinsic ligand, the manner in which they are regulated differs for different channels.

LQTS Mutations in the Intrinsic Ligand.

In hERG channels, the intrinsic ligand harbors hereditary mutations N861I and N861H, which were associated with LQTS (11, 31). To investigate the effect of the N861H and N861I mutations on channel functional properties, mutant hERG channels were transiently expressed in mammalian human embryonic kidney 293 (HEK293) cells and currents were recorded in the whole-cell patch-clamp configuration. Whereas we measured robust currents from cells with WT hERG channels, no measurable currents were recorded from cells with N861H or N861I mutant hERG channels (Fig. 5 A–D). Western blot analysis showed that all of the channels had a robust band at ∼135 kDa, which represents an immature form of the protein, but only WT hERG channels had a band at ∼155 kDa, which represents the mature (N-glycosylated) form of the protein (Fig. 5E). Our results indicate that N861H and N861I channels had a defect in maturation, similar to what was shown for LQTS mutations in the CNBHD, including N861I, which was proposed to be retained in the endoplasmic reticulum (41–43). The lack of hERG currents is consistent with an LQTS phenotype. Taken together, our findings suggest that the intrinsic ligand regulates the deactivation rate and surface expression of hERG channels.

Fig. 5.

LQT2 mutations in the intrinsic ligand prevent functional expression of hERG channels. Whole-cell patch-clamp recordings from cells transfected with WT hERG (A), hERG N861H (B), and hERG N861I (C) channels by using the indicated voltage protocol. (D) Current–voltage relationship indicating robust current from WT hERG (squares), but no measurable currents from hERG N861H (circles) or hERG N861I (triangle). (Scale bar: 1 s and 0.5 nA.) n ≥ 3 for each of the channels. (E) Representative immunoblot of whole-cell lysates from control (untransfected) HEK293 cells or cells expressing WT hERG, hERG N861H, or N861I, as indicated. Similar results were obtained in three independent experiments.

Conformations of the C-Linker Region.

Interestingly, the crystal structures of the C-linker/CNBHDs of agERG, zELK, and mHCN2 channels have different oligomeric assembly. The C-linker/CNBHD of agERG channels crystallized as a monomer, zELK channels crystallized as a dimer (18), and mHCN2 channels crystallized as a tetramer (21) (Fig. 6A). The differences in the oligomeric assembly stem from differences in the folding and orientation of the C-linker regions of the three structures. Structural alignment of the entire C-linker/CNBHDs of agERG with zELK and mHCN2 gives an rmsd of 13 Å and 12 Å, respectively, and alignment of just the C-linkers gives an rmsd of 17 Å and 13 Å, respectively. Alignment of the β-rolls of the C-linker/CNBHDs of agERG, zELK, and mHCN2 reveals that the C-linkers are in dramatically different conformations in the three structures (Fig. 6B). The αA′ helix of agERG is rotated by ∼180° and shifted by ∼30 Å relative to the αA′ helix in mHCN2. The αA′ helix of zELK is rotated by ∼55° relative to its orientation in mHCN2. These results reveal that the C-linker can adopt different conformations in the agERG, zELK, and mHCN2 channel fragments. Whether these configurations correspond precisely to native conformations of the channel is unknown, but it seems likely that they reflect a dynamic nature of the C-linker region, consistent with its proposed role in channel gating (49).

Fig. 6.

The C-linker adopts different orientations in the structures of the C-linker/CNBHDs of different KCNH and HCN channels. (A) A monomer of the C-linker/CNBHD of agERG (red; Left); a dimer formed by the C-linker/CNBHDs of zELK channels (18) viewed parallel to the twofold axis (blue and yellow; Center); and a tetramer formed by the C-linker/CNBDs of HCN2 channels (21) viewed parallel to the fourfold axis (green and gray; Right). (B) The C-linkers of agERG (red), zELK (yellow), and HCN2 (green) with superimposed α-carbons of the corresponding β-rolls. The CNBHDs were removed for clarity. (C) Superposition of the elbow-on-the-shoulder interface of agERG (red), zELK (yellow), and HCN2 (green) channels shown from two different perspectives rotated by 90°.

The C-linker contains an “elbow” and “shoulder” region. The elbow is formed by the αA′ and αB′-helices, whereas the shoulder is formed by the αC′ and αD′-helices (Fig. 1C). Despite the structural differences in the C-linkers of agERG, zELK, and mHCN2, in all three structures the elbows rest on the shoulder, and the elbow-on-the-shoulder interface is very similar (Fig. 6C). For the zELK and mHCN2 structures, the elbow-on-the-shoulder interface contains the primary intersubunit interactions in the C-linker/CNBD fragment and is formed by the elbow of one subunit resting on the shoulder of the neighboring subunit. In the zELK structure, the subunits forming the elbow-on-the-shoulder interface are related by a twofold symmetry (18), whereas in HCN2, the subunits are related by a fourfold symmetry (21) (Fig. 6A).

Surprisingly, however, in the agERG structure, the elbow is resting on its own shoulder (Fig. 1C and Fig. S2). This is because the loop between the αB′ and αC′-helices is much longer in the agERG structure, 12 amino acids in agERG compared with just 2 in zELK and 6 in mHCN2, and the αD′-helix is also much longer in the agERG structure, 18 amino acids in agERG compared with 8 in zELK and 11 in mHCN2. As a result, despite the length and sequence similarity to the C-linkers of zELK and mHCN2 channels, the C-linker of agERG channels contains only four α-helices, as opposed to six in the C-linkers of zELK and mHCN2. The manifestation of these differences in the fold of the C-linker is that the conserved elbow-on-the-shoulder interface, which is the primary region of intersubunit interactions in the zELK and mHCN2 structures, is instead an intrasubunit interface in the agERG structure.

Discussion

With this X-ray crystal structure of the C-linker/CNBHD of agERG channels, there is now a structure of the C-terminal region of channels from each of the three KCNH subfamilies: ELK (zELK; ref. 18), EAG; (mEAG; ref. 22), and ERG (agERG). These structures share some features in common that are unique to the KCNH family. In each structure, the β-roll cavity is negatively charged in contrast to the positively charged cavity, where the cyclic nucleotides bind in canonical cyclic nucleotide-binding proteins. In addition, each structure contains an intrinsic ligand after the C-helix that occupies the β-roll cavity at the position where cyclic nucleotides would normally bind. These signature features of KCNH channels reveal why the KCNH channels are not regulated by the direct binding of cyclic nucleotides (23–26).

Mutations of residues in the intrinsic ligand affect gating of channels in all three KCNH subfamilies, but the effects are different depending on the subfamily (refs. 18, 22, and 48 and Fig. 4). There are several plausible ways in which the intrinsic ligand could be dynamically regulated in the cell. Phosphorylation, Ca²⁺-calmodulin, and PIP₂ binding sites have all been reported to occur near the intrinsic ligand of different KCNH channels (50–54). Another possibility is that the intrinsic ligand might compete with an unknown extrinsic ligand for the binding site on the CNBHD. Therefore, the binding of the extrinsic ligand would displace the intrinsic ligand from the site. Because the intrinsic ligand is covalently attached to the αC-helix, displacing it could move the αC-helix. This mechanism is reminiscent of what happens in CNG and HCN channels when cyclic nucleotides bind to the CNBD and draw in the αC-helix (49). Regulation of mEAG1 channels by flavonoids was recently shown to occur by this mechanism (48).

The hERG channel, however, is specialized for its role in repolarizing the cardiac action potential. This specialization includes a slow rate of channel deactivation, allowing the channels to remain open during the repolarization of the cardiac action potential. Here, we show that the slow deactivation arises, in part, from the intrinsic ligand. We found that, unlike in ELK (18) and EAG (22) channels, in hERG channels mutations in the intrinsic ligand did not measurably affect the conductance–voltage relationship. Instead, these mutations accelerated the deactivation rate of hERG channels, similar to the effects of N-terminal mutations that cause LQTS, a condition characterized by a delayed repolarization of the ventricular action potential (32).

hERG channels harbor hereditary mutations in the intrinsic ligand that are associated with LQTS (11, 31). We found that LQTS mutations N861I and N861H drastically decreased currents from and altered trafficking of hERG channels. LQTS-associated mutations in the C-linker/CNBHD have been shown to abolish hERG currents by interfering with the hERG channel expression at the cell surface (41–43). Together these findings indicate that the intrinsic ligand is an important structural element for a LQTS phenotype.

The structure of agERG also revealed another difference from the structures of other related channels: The C-linker region is in a dramatically different conformation compared with the structures of the ELK and HCN channels. Although the elbow-on-the-shoulder interface is conserved in all three structures, the interface is intrasubunit in agERG and intersubunit in ELK and HCN despite the high degree of sequence similarity between these channels. It remains unclear whether the different conformations and oligomeric assembly seen in the structures from different channels reflect differences in conformational state, differences in channel type, or arise from working with channel fragments. However, these differences almost certainly reflect that the C-linker is a dynamic structure. In CNG and HCN channels, the C-linker couples the binding of cyclic nucleotide to the opening of the pore, a process thought to involve a rearrangement of the elbow-on-the-shoulder interface (49). It seems likely that the C-linker is also involved in coupling the intrinsic ligand of KCNH channels to opening (or closing) of the pore.

Materials and Methods

FSEC.

The C-linker/CNBHD of agERG channels [gene identifier (GI): 158285159, amino acids S535–Q734] was covalently fused to a C-terminal GFP in the pCGFP-BC bacterial expression vector kindly provided by T. Kawate and E. Gouaux, Vollum Institute, Portland, OR (44). The constructs were transformed into BL21 (DE3) cells. The cell cultures were induced with IPTG and harvested by centrifugation. The cell pellets were resuspended in a lysis buffer and sonicated. Insoluble protein was separated by centrifugation, and the supernatant was analyzed with FSEC on a Superdex 200 10/300 GL column (GE Healthcare).

Scale-Up Protein Purification.

The C-linker/CNBHD of agERG (amino acids S535–Q734) was subcloned into a modified pMALc2T vector (New England Biolabs) containing an N-terminal maltose-binding protein (MBP) affinity tag followed by a thrombin cleavage site. The protein was expressed in BL21 (DE3) Escherichia coli cells as described (23). The cells were harvested by centrifugation, resuspended in a lysis buffer [500 mM KCl, 1 mM Tris(2-carboxyethyl)phosphine (TCEP), 30 mM Hepes, 1 mM PMSF, and 2.5 mg/mL DNase at pH 8.0] and lysed in an Emulsiflex-C5 (Avestin). Insoluble protein was separated by centrifugation. The C-linker/CNBHD of agERG was purified on an amylose affinity column and then was loaded on a HiTrap SP FF ion-exchange column following an overnight cleavage with thrombin at 4 °C. The protein was eluted with a linear KCl gradient and concentrated to 5–10 mg/mL for crystallization.

Crystallization and Structure Determination.

Crystals were grown at 20 °C by using the sitting-drop vapor diffusion method. One hundred fifty nanoliter drops of the concentrated protein and reservoir solution were mixed 1:1 by a Mosquito (TTP LABTECH). The final protein solution contained the following: ∼250 mM KCl, 1 mM TCEP, and 30 mM Hepes at pH 7.0. The reservoir solution contained the following: 9 mM urea, 18.2% (wt/vol) PEG 3350, and 7.3% (vol/vol) tacsimate at pH 7.0. The crystals were cryoprotected in reservoir solution supplemented with 25% (vol/vol) glycerol before being flash frozen in liquid nitrogen.

Diffraction data were collected at the ALS (beamline 8.2.1) at Lawrence Berkeley National Laboratory in Berkeley. Data were analyzed with Mosfilm (55) and HKL2000 (56) software. The structure of the agERG C-linker/CNBHD was solved by molecular replacement using the structure of the CNBHD of zELK channels (PDB ID code 3UKN) (18) as a search model. The molecular replacement was carried out by using Phaser in PHENIX (57) followed by the energy- and electron density-guided structure optimization with Rosetta (45). The quality of the model was improved by numerous cycles of refinement in PHENIX and manual model building in Coot (58). The asymmetric unit contained one molecule in the P3₂21 space group. The crystallographic data and refinement statistics are summarized in Table S1. Analysis with molprobity of the final model indicated 1.5% ramachandran outliers. Figures were made using PyMOL. The electrostatic potential surface calculations were carried out by using the APBS plugin for PyMOL and colored from red (−3 kT/e) to blue (+3 kT/e).

Channel Expression in Oocytes and Cultured Cells.

The full-length agERG (GI: 158285159) with a C-terminal FLAG epitope was generated by Bio Basic and subcloned into the pGEMHE oocyte expression vector (a gift of E. Liman, University of Southern California, Los Angeles, CA). The hERG1 expression clone (GI: 487737) containing the S620T mutation to remove C-type inactivation (47) was fused at position 1159 to a monomeric citrine fluorescent protein (gift from R. Y. Tsien, University of California, San Diego, CA) and subcloned into the pGH19 oocyte expression vector as described (37). The cRNA was transcribed by using the mMessage mMachine kit (Ambion). Xenopus laevis oocytes were defolliculated and injected with 50 nL of agERG or hERG cRNA per oocyte. The oocytes were incubated for 3–10 d at 16 °C in ND96 with 50 μg/mL gentamicin.

For electrophysiological recordings in HEK293 cells, WT and mutant (N861H and N861I) hERG expression clones were subcloned into the pcDNA3.1 mammalian expression vector. HEK293 cells were cultured and transiently transfected with hERG as described (36). The cells were incubated for 24–48 h before analysis.

Electrophysiology and Data Analysis.

Currents from whole oocytes or HEK293 cells expressing the WT or mutant hERG channels were recorded with a two-electrode voltage-clamp (OC-725C; Warner Instruments) connected to an analog to digital converter (ITC-18; Instrutech) or a patch-clamp (EPC10; HEKA). Data were recorded by using Patchmaster software (HEKA) and analyzed using Igor Pro software (WaveMetrics). The solutions for electrophysiological recordings from oocytes (37) and HEK293 cells (36) were described.

Currents were elicited by a series of 1-s voltage pulses from −100 to 40 mV in 20-mV increments followed by a 3-s repolarizing pulse to −60 mV. For more detailed investigation of the tail currents, the channels were first activated with a pulse to 20 mV followed by the application of voltage pulses between −120 and 20 mV in 20-mV increments. The holding potential was −80 mV for all experiments. Experiments were performed at room temperature. Error bars indicate the SEM. Statistical analysis was performed by using one-way ANOVA. A value P < 0.05 was considered statistically significant. n represents the number of recordings.

To obtain conductance–voltage relationships, peak tail current amplitudes at −100 mV (Fig. 3 A–D, Insets) were normalized to the largest peak conductance amplitude, which followed a step to 40 mV. These normalized data were then plotted against the test voltage and were fit with a Boltzmann equation. Deactivating currents were fit with a single-exponential equation.

Cell Lysis and Western Blot Analysis.

Cells were homogenized and lysed in lysis buffer as described (36). Lysates were cleared by centrifugation. The protein concentration of the supernatant was quantified by using a Bradford assay (Pierce). Protein samples (20 μg) were incubated with equal amounts of Laemmli sample buffer with 5% (vol/vol) β-mercaptoethanol (Bio-Rad) for 30 min at room temperature, subjected to 7.5% SDS/PAGE, and electrophoretically transferred onto nitrocellulose membranes. Membranes were immunoblotted with an anti–hERG-KA antibody followed by an HRP-linked secondary antibody (Jackson Labs) and developed by using an ECL detection kit (Thermo).