The N–Terminal Tail of hERG Contains an Amphipathic α–Helix That Regulates Channel Deactivation

The cytoplasmic N–terminal domain of the human ether–a–go–go related gene (hERG) K+ channel is critical for the slow deactivation kinetics of the channel. However, the mechanism(s) by which the N–terminal domain regulates deactivation remains to be determined. Here we show that the solution NMR structure of the N–terminal 135 residues of hERG contains a previously described Per–Arnt–Sim (PAS) domain (residues 26–135) as well as an amphipathic α–helix (residues 13–23) and an initial unstructured segment (residues 2–9). Deletion of residues 2–25, only the unstructured segment (residues 2–9) or replacement of the α–helix with a flexible linker all result in enhanced rates of deactivation. Thus, both the initial flexible segment and the α–helix are required but neither is sufficient to confer slow deactivation kinetics. Alanine scanning mutagenesis identified R5 and G6 in the initial flexible segment as critical for slow deactivation. Alanine mutants in the helical region had less dramatic phenotypes. We propose that the PAS domain is bound close to the central core of the channel and that the N–terminal α–helix ensures that the flexible tail is correctly orientated for interaction with the activation gating machinery to stabilize the open state of the channel.


Introduction
The human ether-a-go-go related gene (hERG) encodes K v 11.1, the pore forming subunit of the rapidly activating delayed rectifier K + channel (I Kr ) [1]. Reduction of hERG channel activity either by drugs [2] or genetically inherited mutations [3] results in prolongation of the QT interval on the surface electrocardiogram and a markedly increased risk of arrhythmias and sudden cardiac death [4]. hERG channels are tetrameric with each subunit containing cytoplasmic Nand C-terminal domains and six transmembrane domains. The fifth and sixth transmembrane domains along with an intervening pore helix from each of the four subunits surrounds the ion conducting pore [5]. In addition, a cyclic nucleotide binding domain (cNBD) immediately C-terminal to the pore domain is thought to contribute to the stabilization of the tetrameric structure [6]. Conversely, the cytoplasmic Nterminus of each subunit contains a Per-Arnt-Sim (PAS) domain (residues   [7,8] that is stable as a monomer and interacts with the remainder of the channel thereby regulating the kinetics of channel opening and closing [7,9,10]. In hERG, N-terminal deletions that remove the PAS domain (D2-373 [11], D2-354 [12,13], and D2-138 [7]) significantly enhance the rate of deactivation of the channel. Further, the N-terminal domain (residues 1-136) is able to restore deactivation gating in N-terminally truncated hERG [7,9]. However, deletions within the short N-terminal tail that precedes the PAS domain (residues 1-25) also result in significantly faster rates of deactivation [7,12,13]. Moreover, application of a peptide corresponding to the N-terminal 16 residues can slow the deactivation kinetics of channels with most of the N-terminus deleted (D2-354, [12]).
To clarify the role of the N-terminal tail domain in hERG K + channel deactivation, we determined the solution state structure of a construct encompassing both the PAS domain and the Nterminal tail. We show that the tail contains an amphipathic ahelical region from T13 to E23. Deletion of either the initial unstructured segment (D2-9) or replacement of the amphipathic a-helical region with a flexible linker resulted in faster deactivation, suggesting that both regions are necessary but neither is sufficient to permit normal deactivation. Alanine scanning of the N-terminal tail indicated that residues R5 and G6 are involved in critical interactions that stabilize the open state of the channel. Although the majority of alanine mutations in the amphipathic ahelical region did not have a significant effect on the rate of deactivation, both I19A and R20A showed an enhanced rate of deactivation. This suggests that the a-helix may act as a spacer, rather than being involved in critical specific interactions with other domains of the channel.

Protein expression
The PAS domain (residues 1 to 135) was expressed as an Nterminal cleavable glutathione-S-transferase (GST) fusion in E. coli C41 strain after overnight induction at 20uC. A freeze/thaw method was used to lyse the cells with 20 mM Tris buffer containing 5 mM 2-mercaptoethanol, 0.1% v/v Tween 20, and 150 mM NaCl. The lysate was incubated with glutathione beads (GE Healthcare, Amersham, UK) for 3 h and the protein eluted by TEV protease digestion overnight at 4uC. The protein was concentrated and passed through a Superdex 75 column (GE Healthcare), equilibrated in 10 mM HEPES pH 6.9, 150 mM NaCl, 5 mM N-octyl-D-glucoside (OG) (Anatrace Inc., Maumee, OH, USA) and 3 mM tris(2-carboxyethyl)phosphine (TCEP). The purified PAS domain eluted as a single peak at the expected molecular size for a monomer (as previously described [7]). The 13 C/ 15 N double-labelled PAS domain was produced by substituting the nitrogen and carbon sources in bacterial growth medium with 15 N-enriched NH 4 Cl and 13 C-enriched glucose, respectively.

Sample preparation and NMR spectroscopy
The NMR sample consisted of 0.21 mM 13 C/ 15 N PAS domain protein in solution containing 10 mM HEPES, 3 mM TCEP, 5 mM OG, and 7% D 2 O, at pH 6.9. All NMR experiments were performed on a Bruker Avance II 900 MHz NMR spectrometer at 298 K. 2D 1 H-15 N HSQC, 3D 1 H-15 N NOESY and 3D 1 H-13 C NOESY data were acquired using traditional methods while 3D HNCO, HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, C(CO)NH, H(CCO)NH and HCCH-TOCSY data were acquired using a non-uniform sampling method and maximum entropy reconstruction [14]. The sample was buffer-exchanged in D 2 O before acquiring 3D HCCH-TOCSY and 1 H-13 C NOESY NMR spectra.

NMR chemical shift assignment and structure calculations
All NMR spectra were analysed using XEASY3 [15]. Sequence-specific backbone assignments were made using 3D HNCO, HNCA, HN(CO)CA, HNCACB and CBCA(CO)NH data. Side-chain chemical shift assignments were made using 3D (H)CC(CO)NH-TOCSY, H(CC)(CO)NH-TOCSY and HCCH-TOCSY data. A total of 2634 distance constraints were derived from 3D 1 H-15 N and 1 H-13 C NOESY data, 24 hydrogen bond constraints were derived from the 1 H-13 C NOESY data (based on amide protons that were still observable afte exchange of the sample into D 2 O buffer), and 178 dihedral angle constraints (w,y) were derived from TALOS [16]. The error range used in the structure calculations were set to twice the standard deviation estimated by the program. Automated NOE assignment and structure calculations were performed using the program CYANA v2.1 [17]. An ensemble of the 20 structures with the lowest target function values was chosen to represent the solution structure of the protein. Energy minimization of these structures was performed using the program AMBER 10 [18]. The generalized Born (GB) solvent model was used for the final energy minimization using the distance constraints from the CYANA calculation. The energy-minimized structures were validated using the PSVS server [19] and deposited in the PDB [20] under the accession code 2L0W. Chemical shift assignments were also deposited in the BioMagResBank under accession code 17066. Secondary structure elements were predicted using Talos+ [21].
Electrophysiology HERG cDNA (a gift from Dr Gail Robertson, University of Wisconsin) was subcloned into a pBluescript vector containing the 59 untranslated region (UTR) and 39 UTR of the Xenopus laevis bglobin gene (a gift from Dr Robert Vandenberg, University of Sydney). Mutagenesis was carried out using the Quickchange mutagenesis method (Agilent Technologies, CA, USA) and confirmed by DNA sequencing. Wild-type (WT) and mutant channel cDNAs were linearized with BamHI and cRNA transcribed with T7 RNA polymerase using the mMessage mMachine kit (Ambion, city, TX, USA).
Xenopus laevis oocytes were prepared as previously described [22]. Stage V and VI oocytes were isolated, stored in tissue culture dishes containing ND96 (in mM: KCl 2.0, NaCl 96.0, CaCl 2 1.8, MgCl 2 1.0 and HEPES 5.0) supplemented with 2.5 mM sodium pyruvate, 0.5 mM theophylline and 10 mg mL 21 gentamicin, adjusted to pH 7.5 with NaOH and incubated at 18uC. All experiments were approved by the Garvan/St Vincent's Animal Ethics Committee (Approval ID 08/34).
Xenopus laevis oocytes were injected with 5-10 ng cRNA and incubated at 18uC for 24-48 h prior to electrophysiological recordings. All experiments were undertaken at room temperature (21-22uC). Two-electrode, voltage-clamp experiments were performed using a Geneclamp 500B amplifier (Molecular Devices Corp, Sunnyvale, CA, USA). Glass microelectrodes had tip resistances of 0.3-1.0 MV when filled with 3 M KCl. Oocytes were perfused with ND96 solution (see above). In all protocols a step depolarization of +20 mV from the holding potential of 290 mV was applied at the start of each sweep to enable off-line leak-current subtraction. We assumed that the current leakage was linear in the voltage range 2160 to +40 mV. Data acquisition and analysis were performed using pCLAMP software (Version 9.2, Molecular Devices Corp, Sunnyvale, CA, USA) and Excel software (Microsoft, Seattle, WA, USA). All parameter values were estimated as mean 6 standard error of the mean (SEM) for n experiments, where n denotes the number of different oocytes studied for each construct. Isochronal activation curves were measured using standard tail current analysis [1]. Cells at a holding potential of 290 mV were subjected to 4-s depolarizing steps to voltages in the range 270 to +50 mV before stepping the voltage to 270 mV where tail currents were recorded. Tail current data were normalized to the maximum current value (I max ) and fitted with a Boltzmann expression:  where g/g max is the relative conductance, V 0.5 is the half-activation voltage, V t is the test potential and k is the 'slope factor'. Alternatively, the data were fitted with the thermodynamic form of the Boltzmann expression: where DG 0 is the work done at 0 mV, z g is the effective number of gating charges moving across the membrane electric field, E. F, is Faraday's constant, R is the universal gas constant and T is the absolute temperature. Equations (1) and (2) are equivalent, however from Equation (2) we can calculate the effect of mutations on changes in the chemical potential (DG 0 ) and electrostatic potential (-z g EF) that drives activation and deactivation of the channel.
To measure rates of deactivation, oocytes were depolarized from a holding potential of 290 mV to +40 mV for 1 s to fully activate the channels; they were then repolarized to potentials in the range 250 to 2160 mV. A double exponential function was fitted to the decaying portion of tail currents. In order to compare rates of deactivation for different mutants at comparable driving forces, the voltages in the range 260 to 2160 mV were converted to electrochemical driving forces (-z g EF) as defined in Equation (2).

Results
The solution structure of hERG 1-135 domain was well folded with dispersed peaks in the 1 H-15 N-HSQC (Fig. S1). More than 97% of the backbone amides were assigned, the exceptions being A40 and R77. 75% of the sidechain N-H from arginine, asparagine and glutamine were assigned; the exceptions were R4, R5, R73, R76, R77 and Q84. All resonances of Ha and Hb atoms were assigned except for Hb of R4 and R5. All Ca, Cb and C were assigned except for the carbonyls of R76 and V113. In addition, more than 90% of the remaining sidechain proton and carbon resonances were assigned. These assignments allowed a total of 2634 distance constraints (Table 1) to be unambiguously derived and these were used in combination with the dihedral angle and hydrogen-bond constraints to calculate the solution structure of the N-terminal 135 residues of the hERG K + channel (Fig. 1). The RMSDs of all backbone and heavy atoms excluding . Both the unstructured tail and the a-helix of the Nterminal domain are required to slow deactivation kinetics of WT hERG channels. Sequence of the N-terminal domain in WT hERG1 channels is shown in panel A. Mutant constructs were designed to examine the effect of removing the unstructured tail (D2-9) or the entire N-terminal domain (D2-25), or the effect of disrupting the Nterminal a-helix (GGS) by replacing residues L15 and T17-E23 with P_GGSGGSG as shown in red. (B) Typical rates of deactivation observed in tail currents recorded at 2120 mV following a step to +40 mV. To aid comparison current traces were normalized to peak tail current. Constructs D2-9 (blue), D2-25 (green) and GGS (red) all produced channels with faster deactivation rates than WT hERG (black). (C and D) Mean 6 SEM for deactivation rates recorded over a range of voltages from 250 mV to 2160 mV. Current decay associated with channel deactivation was best fitted to two exponentials, generating t fast (C) and t slow (D) for WT (open circles, n = 11), D2-9 (blue squares, n = 14), D2-25 (green triangles, n = 5) and GGS (red inverted triangles, n = 10) channels. All mutant channels enhanced the rates of both the fast and slow components of deactivation over the entire voltage range. doi:10.1371/journal.pone.0016191.g003 the N-terminal tail (S26 to K135) were 0.40 and 1.10 Å , respectively.
Statistics highlighting the extremely high precision and stereochemical quality of the ensemble of hERG PAS domain and N-terminal tail structures are shown in Table 1. The average MolProbity score of 1.55 places the ensemble in the 94th percentile relative to all other structures ranked by MolProbity [23]. The high stereochemical quality of the ensemble stems from a complete absence of bad close contacts, high Ramachandran plot quality (95% of residues in the most favored region), and a very low percentage of unfavorable sidechain rotamers. During the automated NOESY assignment/structure calculation process the CANDID module of CYANA assigned 86% of all NOESY crosspeaks to give an average of 19 NOE constraints per residue.
The NMR solution structure for the segment from residues S26 to K135 was almost identical to the crystal structure of the same region (Fig. 1B). The average RMSD of this NMR ensemble relative to the crystal structure (1BYW) was less than 1 Å . The only significant discrepancy was observed for the loop between Hb and Ib, a region previously shown to be highly dynamic in a molecular dynamics study [24].
In addition to the structure of the PAS domain (residues S26 to K135), the NMR solution structure revealed a randomly distributed a-helical tail (blue lines in Fig. 1A), in particular residues T13 to E23 ( Fig. 2A) with a backbone RMSD of 0.31 Å . Despite not being able to determine a fixed conformation of the N-terminal tail, it was consistent with the secondary structure prediction using the NMR chemical shifts and inter-residue NOE constraints (Fig. S2).
Closer examination of the N-terminal helix (T13 to E23) revealed that it was amphipathic with positively charged residues (R20 and K21), negatively charged residues (D16 and E23), and polar residues (T13 and T17), located on the same side of the helix (Fig. 2B), while non-polar residues (F14, L15, I18, I19 and F22 but not I19) were located on the opposite face of the helix. Residues 1-9 are disordered with no clearly defined structure, while residues 10-12 adopted a turn conformation.
To study the functional significance of the a-helix in the Nterminal tail, we compared the effects of deleting only the initial portion of the N-terminus (D2-9), with deletion of the entire Nterminal tail (D2-25), or replacement of the a-helix with a flexible linker (denoted GGS mutant, see Fig. 3A). Typical examples of tail currents for D2-9, D2-25 and GGSmut channels (recorded at -120 mV and normalized to the peak inward current amplitude) are shown in Fig. 3B. All mutant channels showed significant enhancement of both the fast (Fig. 3C) and slow (Fig. 3D) components of deactivation over the entire voltage range studied. None of the mutants affected the relative amplitudes of the fast and slow components of deactivation at the most negative potentials. However, at less negative potentials where the fast component became less dominant, the amplitude of the fast component relative to the slow component was greater in all three mutant channels compared to WT hERG (Fig. S3).
When comparing the rates of deactivation for WT and mutant channels at a single voltage, it is important to consider that changes to steady-state activation can affect the electrochemical potential for deactivation. Steady-state activation properties were determined by fitting, with a single Boltzmann expression (Equation 1), the I-V relationship of peak tail currents at 270 mV plotted against the preceding voltage step (Fig. 4). The resulting half-maximal voltage for activation (V 0.5 ) of D2-9 channels (25.660.9 mV, n = 14) was shifted in the depolarizing direction compared to WT hERG (223.260.8 mV, n = 11, ANOVA p,0.05), without any change in slope. A small but statistically significant shift in activation V 0.5 was also observed for GGSmut (218.560.6 mV, n = 10) channels, while D2-25 channels were similar to WT hERG (Fig. 4A). The chemical (DG 0 ) and electrostatic (-z g EF) potential that drives activation was calculated by fitting the activation data with a Boltzmann function in the form of Equation 2 (data summarized in Table S1). Since changes in DG 0 parallel changes in activation V 0.5 , D2-9 channels had a significantly smaller chemical potential for activation (-1.960.3 kJ mol -1 ) than WT hERG (26.860.2 kJ mol 21 , ANOVA p,0.05). To compensate for changes in electrochemical driving force, the rates of deactivation calculated at voltages between 250 mV to 2160 mV were plotted against the electrochemical potential for deactivation (Fig. 4B). After correction, D2-9, D2-25 and GGSmut channels all had enhanced rates of deactivation compared with WT hERG (ANOVA p,0.05). Thus, alterations to the electrochemical driving force for deactivation could not explain the enhanced deactivation rates seen with these mutant channels.
To probe the role of individual residues within the N-terminal tail, native residues from P2-E23 were individually replaced with alanine, or with valine in the case of A9. Measured rates for fast (t fast ) and slow (t slow ) components of deactivation, in addition to the relative contributions of these components are given in Table  S2. At negative potentials, the fast component accounted for the majority of deactivation (.80%). This parameter was therefore used to compare WT and mutant channels. Several of the mutations introduced small (less than 610 mV), and statistically significant shifts in the voltage dependence of channel activation when compared to WT hERG (Table S1). Accordingly, the effects of each mutation on deactivation rate were compared at an equivalent driving force of 230 kJmol -1 (as indicated in Fig. 5). In Fig. 6, the effect of alanine mutants on deactivation rates are classified into those that were unchanged (grey bars), faster (red bars) and slower (blue bars) compared to WT.
Typical tail currents recorded at 2120 mV, as well as mean t fast values plotted against electrochemical driving force, are shown in Fig. 5A for two mutations (R5A and G6A) located in the unstructured N-terminal tail. Both the R5A and G6A mutants had significantly faster deactivation rates than the WT channel; the t fast at 230 kJmol -1 was 16.561.0 ms (n = 8) for R5A and 12.761.3 ms (n = 7) for G6A compared with 28.562.8 ms (n = 11) for WT (ANOVA p,0.05, Fig. 6). These data suggest that either, or both, of these residues could interact with other parts of the channel protein to slow deactivation rates in WT hERG.
Within the amphipathic N-terminal a-helix (residues T13 to E23), alanine mutations I19A and R20A gave channels with faster rates of deactivation when compared with WT hERG; t fast at 230 kJmol -1 was 19.  Rates of deactivation are represented by the decay in tail currents recorded at 2120 mV following a step to +40 mV. Current traces were normalized to peak tail current to aid comparison. B. Mean 6 SEM rates of deactivation (t fast ) plotted against the total electrochemical driving force -(DG 0 -z g EF) for channel deactivation. When compared at an equivalent driving force of -30 kJmol -1 (dotted line) four mutant channels exhibited altered deactivation rates compared with WT hERG (open squares, n = 11). Two mutations, P10A (green circles, n = 14) and D16A (blue triangles, n = 10), produced channels that were slower than WT (left panel), while R5A (red diamonds, n = 8) and G6A mutant channels (orange inverted triangles, n = 7) deactivated faster than WT hERG (right panel). doi:10.1371/journal.pone.0016191.g005 surprisingly, two mutations in the a -helical region, i.e., T13A (t fast , 38.262.2 ms, n = 15) and D16A (t fast , 54.264.7 ms, n = 10) exhibited slower deactivation rates than WT (ANOVA, p,0.05). Interestingly, the four residues with altered deactivation rates (T13, D16, I19, R20) lay on one face of the amphipathic helix (Fig. 6C). In addition, mutation of one residue (P10A) that lies between the amphipathic a-helix and the D2-9 region, also significantly slowed deactivation (44.462.4 ms, n = 14) compared with WT hERG (28.562.8 ms, ANOVA p,0.05).

Discussion
The solution structure of hERG PAS domain determined in this study was found to be very similar to the crystal structure determined previously by Morais Cabral et al. [7] and the solution structure recently reported by Li and colleagues [25]. Our solution structure of the PAS domain (residues 26-135) superimposed very well with the crystal structure apart from the loop between Hb and Ib that was previously shown to be highly dynamic [24]. The major difference between the X-ray and NMR structures is that in the latter the Nterminal tail contained an amphipathic a-helix from residues T13 to E23. The first nine residues in the NMR structures were unstructured as were residues 24-26 that link the a-helix to the PAS domain. The remaining tail-domain residues (P10, Q11, N12) appeared to adopt a turn conformation but there were insufficient NOE constraints to be able to define it as an extension of the a-helix.
Our functional, electrophysiological data indicated that removal of either the initial unstructured segment (D2-9) or replacement of the a-helix with a flexible linker (GGS mutant) both produced similar phenotypes, i.e. markedly faster rates of deactivation. This is essentially the same phenotype produced by deletion of the entire N-terminal tail (D2-25, Fig. 3A, or D2-23 [7] or D2-26 [7]). These data indicate that the initial unstructured N-terminal segment and the a-helical region (residues 13-23) were both required, but neither alone was sufficient, for the normal slow deactivation kinetics of WT hERG channels.
Alanine scanning mutagenesis of the initial unstructured Nterminal segment (P2 -N12) identified R5 and G6 as the most critical residues for deactivation. Both R5A and G6A mutants had enhanced rates of deactivation that could easily explain the faster deactivation observed with deletion of residues 2-9. Within the ahelical segment two alanine mutants resulted in enhanced rates of deactivation (I19A and R20A). These mutants resulted in a smaller perturbation to deactivation than R5A and G6A, nevertheless the combined effect of all four residues may explain why deletion of residues 2-25 caused a greater enhancement of the rate of deactivation than did deletion of residues 2-9.
Whilst the deletion mutants and the GGS mutant all had faster deactivation phenotypes, three alanine mutants resulted in slowed deactivation, i.e., T13A and D16A in the a-helix and P10A in the linker between the a-helix and the initial unstructured segment. It is possible that these three residues are important for ensuring that the tail does not bind too tightly to the open state of the channel. It is noted that T13 and D16 lie on the same side of the a-helix as I19 and R20, which are the only other residues in the a-helix that significantly perturbed deactivation. This suggests that this surface of the a-helix is involved in protein-protein interactions that affect the rate of deactivation. However, when we mapped the functional effect of alanine mutants onto the structure of the N-terminal ahelix, the residues with perturbed function did not lie along the entire length of the a-helix (Fig. 6C). We therefore suggest that in addition to providing some specific interactions the a-helix may also serve as a spacer to ensure that the flexible tail is held a predetermined distance from the PAS domain itself. Given that alanine mutants tend to stabilize a-helices it is also possible that the P10A mutant may have stabilized a longer helical domain that results in slower deactivation.
It is important to recall that the structure we have solved is an isolated domain. It is possible that the N-terminal tail structure reported here is more flexible than it would be in the whole, intact channel protein. Conversely, it is clear that the N-terminal tail interacts with another part of the channel protein to regulate deactivation and we suggest that the flexibility of the distal N-tail (residues 1-9) is important for its function and/or regulation.

Model for the structural basis of deactivation gating
Deletion of the N-terminal tail (D2-25, Fig. 3A), the entire PAS domain (D2-138, [7]) or majority of the N-terminus (D2-354 [13], D2-373 [11]) all result in a very similar phenotype, i.e. approximately 5-fold faster rate of deactivation. A plausible hypothesis that explains these observations is that the PAS domain binds (with relatively high affinity) to another domain on the hERG channel and it positions the flexible N-terminal tail region close to the central core of the hERG channel where it binds and unbinds sufficiently rapidly to modulate the rate of deactivation. The region(s) of the channel where the PAS/N-terminal a-helix and the flexible N-terminal domains bind remain to be determined. Two obvious candidates are the S4-S5 linker [25], a part of the channel known to be critical for regulation of deactivation gating [12,13,26] and the C-linker + cyclicnucleotide binding domain, as mutations in this domain modulate the kinetics of deactivation [27,28,29]. Li and colleagues showed that the PAS domain can bind to the S4-S5 linker, however these studies were perfomed with an isolated S4-S5 peptide fragment and need to be confirmed in studies involving either the entire channel protein or at least larger domains. Similarly, testing of the hypothesis that the PAS domain and/or N-terminal a-helix bind to the cyclic-nucleotide binding domain will require expression and purification of the cyclic-nucleotide binding domain.  Figure S2 Amino acid sequence of hERG PAS domain and overview of NMR data. The secondary structure elements are labelled according to the hERG PAS domain PDB structure (2L0W). Hydrogen bonds constraints used in the structure calculation are indicated as black circles. Chemical shift index (CSI) prediction of the secondary structure is shown immediately above the amino acid sequence. Thick and thin bars indicate strong and weak NOE cross-peaks intensities for the sequential proton-proton NOE connectivities (dNN, daN and dbN). The observed medium-range NOEs dNN(i, i+2), daN(i, i+2), daN(i, i+3), dab(i, i+3) and daN(i, i+4) and are indicated by lines connecting the two residues that are related by the NOE. (EPS) Figure S3 Relative amplitudes of tfast and tslow that comprise deactivation rates. Tail currents recorded over a range of potentials (Vm) following a test pulse to +40 mV are fit with a double exponential function (see methods). Relative amounts of tfast and tslow components are then plotted against voltage. At negative potentials, where tfast dominates, there is little difference in relative amplitudes between WT (black) and mutant channels (D2-9: blue; D2-25: green; GGSmut: red). However, at less negative potentials (.290 mV) there is a significant increase in the relative amounts of tfast in mutant channels (D2-9, D2-25, GGSmut) compared to WT hERG. (EPS)