Structure of the HERG K+ channel S5P extracellular linker: role of an amphipathic alpha-helix in C-type inactivation.

The HERG K+ channel has very unusual kinetic behavior that includes slow activation but rapid inactivation. These features are critical for normal cardiac repolarization as well as in preventing lethal ventricular arrhythmias. Mutagenesis studies have shown that the extracellular peptide linker joining the fifth transmembrane domain to the pore helix is critical for rapid inactivation of the HERG K+ channel. This peptide linker is also considerably longer in HERG K+ channels, 40 amino acids, than in most other voltage-gated K+ channels. In this study we show that a synthetic 42-residue peptide corresponding to this linker region of the HERG K+ channel does not have defined structural elements in aqueous solution; however, it displays two well defined helical regions when in the presence of SDS micelles. The helices correspond to Trp585-Ile593 and Gly604-Tyr611 of the channel. The Trp585-Ile593 helix has distinct hydrophilic and hydrophobic surfaces. The Gly604-Tyr611 helix corresponds to an N-terminal extension of the pore helix. Electrophysiological studies of HERG currents following application of exogenous S5P peptides show that the amphipathic helix in the S5P linker interacts with the pore region of the channel in a voltage-dependent manner.

HERG (human ether-a-go-go related gene) encodes the poreforming ␣-subunit of the rapid delayed rectifier potassium channel, I Kr (1). The channel is an important contributor to repolarization of the cardiac action potential (1)(2)(3). Furthermore, mutations in HERG cause congenital long QT syndrome type 2 (4), which results in a markedly increased risk of ventricular arrhythmias and sudden cardiac death (5,6). A wide range of drugs that block the HERG K ϩ channel also result in drug-induced long QT syndrome, the most common cause of serious drug-induced arrhythmia and death (7,8). Therefore, there is considerable interest in gaining a better understanding of the structure and structure-function relationships in the HERG K ϩ channel.
HERG is a member of the family of voltage-gated K ϩ channels (VGK) 1 that contain six transmembrane domains, denoted by S1-S6, and a pore helix that is interposed between S5 and S6. The positively charged S4 acts as the voltage sensor for activation (9). Unlike other members of the VGK family, the HERG channel also undergoes very rapid voltage-dependent inactivation and recovery from inactivation (1, 10 -12). Consequently the HERG K ϩ channel functions as an inward rectifier, i.e. it passes little current at depolarized potentials but large currents during the terminal repolarization phase of the cardiac action potential (13,14). This rapid inactivation is also critical for the role of the channel in suppressing arrhythmias initiated by ectopic electrical excitation (11,15).
Inactivation of the HERG K ϩ channel results from conformational changes in the outer pore region of the channel (11, 12, 16 -19) and involves so-called "collapse of the pore" (20). The pore region of HERG, including the pore helix, a selectivity filter, and S6, is highly homologous to that of other members of the VGK family (8,9) as well as to the bacterial K ϩ channel KcsA (21,22). This feature has enabled homology models of HERG K ϩ channels to be constructed based on the KcsA structure (21). However, the extracellular loop connecting the pore helix to the top of S5 (S5P loop) in HERG is very different from that in other VGK family members. First, the S5P loop in HERG is about 40 amino acids long, compared with 10 -15 in most other members of the VGK family (9, 23) (also see Fig. 1). Second, many mutations in the S5P loop disrupt the inactivation process in HERG (19,23,24). Third, toxins that bind to the S5P loop of other VGK channels, e.g. agitoxin and charybdotoxin (25), do not bind to HERG. Conversely, toxins that bind to this region of HERG, e.g. ErgToxin and BeKm-1, do not bind to other members of the VGK family (26,27). The S5P loop of HERG therefore appears to be a critical region of the protein, but at present there is little specific information known about its three-dimensional structure.
Following the determination of the structure of a number of prokaryotic K ϩ channels (22, 28 -30), great progress has been made in the understanding of the spatial arrangement of amino acid side chains in the pore region of K ϩ channels. Nevertheless, the crystallization and determination of the structure of ion channels remains a tremendously difficult task (31), and to date no structures of mammalian ion channels have been determined (32). For this reason, CD spectropolarimetry (33,34) and NMR spectroscopy (35) have been used to gather information on the structure of ion channels and/or domains. Although NMR spectroscopy has limitations in terms of the size of proteins whose structure can be determined, it has the advantage of permitting the determination of structures in a lipid environment (36), thereby eliminating the problems associated with crystallization of membrane proteins (31).
In this study we have used a combination of CD spectropolarimetry, two-dimensional 1 H NMR spectroscopy, and electrophysiology to investigate the structure and function of the S5P linker of the HERG K ϩ channel. Our findings show that the S5P linker contains an amphipathic ␣-helix. Exogenous application of the S5P peptide fragment or a peptide corresponding to the amphipathic ␣-helix results in altered ionic selectivity and disruption of inactivation of the HERG K ϩ channel. These results suggest that the amphipathic ␣-helix in S5P is critical for inactivation of HERG K ϩ channels.

EXPERIMENTAL PROCEDURES
Peptide Preparation-The peptides were synthesized on a 0.50-mmol scale using O-(Benzotriazol-1-yl)-N,N,NЈ,NЈ-tetramethyluronium hexafluorophosphate activation of Boc-amino acids with in situ neutralization chemistry, as previously described (37). The syntheses were performed on Boc-Tyr(2BrZ)-OCH2-Pam resin using standard amino acid side chain protection, except that methionine residues at positions 5 and 10 were replaced by the isosteric norleucine residue to prevent adventitious oxidation of the peptide. This step is necessary to stabilize the synthetic peptide and is not expected to affect the peptide conformation (37). Each residue was reacted for 10 min, and coupling efficiencies were determined by the quantitative ninhydrin reaction. Prior to a standard HF cleavage (10 ml of p-cresol:HF 1:9, 0°C, 60 min) and workup, the N-terminal Boc protecting group was removed (100% trifluoroacetic acid), followed by formyl group removal (1.5 ml of ethanolamine in 25 ml of N,N-dimethylformamide and 5% water, twice for 30 min).
Three peptides were synthesized in this study: a 42-residue peptide corresponding to the S5P linker of HERG (residues Ala 570 -Tyr 611 ), to which we refer as the S5P peptide; a 42-residue peptide in which the putative amphipathic helix corresponding to residues Gly 584 -Lys 595 (23) was replaced with a GGGSGGGSGGGS linker, to which we refer as the del-helix peptide; and finally a 19-residue peptide corresponding to the putative amphipathic helix and four residues at each end (i.e. Ser 581 -Ser 599 of wild type HERG), to which we refer as the helix peptide.
The NMR spectroscopy sample was prepared by dissolving 2.6 mg of the S5P peptide in ϳ400 l of 90% H 2 O, 10% D 2 O (v/v) containing 12 mg of SDS-d25 (ϳ100 mM; the critical micellar concentration for SDS is 8 mM) in a 5-mm outer diameter susceptibility-matched microcell (Shigemi, Tokyo, Japan). This resulted in a peptide concentration of 1.4 mM and pH 3.3.
NMR Spectroscopy-NMR spectroscopy experiments were performed on a Bruker Avance-600 DRX spectrometer with a 5-mm 1 H inverse probe with operating temperatures of 20, 30, and 37°C. The homonuclear two-dimensional experiments that were performed included double quantum-filtered correlation spectroscopy (38) with a phase cycle modified for fast recycle times (38), total correlation spectroscopy (39) with MLEV spin-lock periods of 35 and 90 ms, and nuclear Overhauser enhancement spectroscopy (NOESY) (40) with mixing times of 200 and 300 ms. All two-dimensional spectra were acquired using time proportional phase detection (41). In double quantum-filtered correlation spectroscopy and NOESY experiments, water signal suppression was achieved by low power irradiation at the water resonance during the relaxation delay (1.3 s) and during the mixing period in NOESY experiments. In total correlation spectroscopy experiments, the water signal was suppressed using the WATERGATE gradient module (42). All of the spectra were processed using XWIN-NMR software (Bruker, Karlsrü he, Germany).
Structure Calculations-Analyses of two-dimensional spectra were carried out using the XEASY program (43). Distance constraints were obtained from cross-peak volumes in the NOESY spectra recorded at 30°C with a mixing time of 200 ms. This yielded 416 nonredundant upper distance constraints. An additional 14 distant constraints for hydrogen bonding were obtained from a hydrogen-deuterium exchange experiment. No dihedral angle constraints were used in the structure calculations because backbone amide peaks were too broad, probably because of conformational averaging of the peptide structure when in the presence of SDS micelles. The simulated annealing protocol in the torsion angle dynamics program DYANA (44) was used to obtain preliminary three-dimensional structures prior to refinement. Of the 1600 structures generated in DYANA, 40 of the "best" structures, with the lowest NOE violations, were chosen for refinement using the standard simulated annealing script in CNS (45). In this refinement process, the high temperature dynamics and cooling cycle were performed in Cartesian space. Analysis of the ensemble of S5P structures was also carried out using PROMOTIF, a program that identifies structural motifs in proteins (46). The figures were generated using MOLMOL (47).
Circular Dichroism Spectropolarimetry-CD spectropolarimetry spectra were recorded on a Jasco J-720 spectropolarimeter equipped with a Neslab RTE-111 temperature controller. CD spectropolarimetry data were collected using a 1-mm cuvette over the wavelength range 190 -250 nm and with a resolution of 0.5 nm, a bandwidth of 1 nm, and a response time of 1 s. Final spectra were the sums of three scans accumulated at a speed of 20 nm/min and were base line-corrected. Peptide concentration was 0.45 mg/ml for the S5P peptide or 0.15 mg/ml for the del-helix and helix peptides, in 10 mM sodium phosphate, pH 3.0 and 7.0, with or without the addition of 100 mM SDS. The data are presented as molar ellipticity [], where [] ϭ /(10 ϫ c ϫ l) and is ellipticity (mdeg), c is the molar concentration of the sample (mol/liter), and l is the pathlength in cm.
FIG. 1. A, alignment of the S5, pore helix, and selectivity filter regions of Shaker, KcsA and HERG. Gray shading indicates S5 (Shaker and HERG) or outer helix (KcsA) and the pore helices. The asterisks indicate methionine residues that were replaced with norleucine in the synthesized peptide to simplify structure assignment (see text for details). Residues in HERG shown in bold type indicate the peptide that was synthesized in this study. The inset shows a cartoon of a HERG K ϩ channel subunit, with the region of interest highlighted by gray shading and bold lines. B, alignment of the region surrounding the amphipathic helix (shaded region) of HERG with other members of the ERG/EAG/ELK family of K ϩ channels. Residues in white with black shading indicate hydrophobic residues conserved in all eight members of the ERG/EAG/ELK family.
Electrophysiology-Chinese hamster ovary (CHO) cells stably transfected with HERG K ϩ channels were maintained in culture using Dulbecco's modified Eagle's medium/Ham's F-12 medium (Invitrogen) with 5% fetal bovine serum, as previously described (48). CHO cells were plated on 13-mm glass coverslips 48 -72 h prior to patch clamp analysis. The coverslips were then placed in a 0.5-ml perfusion chamber mounted on the stage of a Nikon Eclipse TE200 inverted microscope. The cells were superfused with normal Tyrode solution that contained 130 mM NaCl, 4.8 mM KCl, 0.3 mM KH 2 PO 4 , 0.35 mM NaH 2 PO 4 , 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , 10 mM HEPES, pH adjusted to 7.4 with NaOH. In the Na ϩ -free external solution NaCl was replaced with N-methyl-D-glucamine (NMDG)-Cl, and the NaH 2 PO 4 was omitted. The cells were patched using micropipettes fabricated from thin walled borosilicate glass (Vitrex Microhematocrit tubes, Modulohm I/S, Denmark) with a horizontal pipette puller (model P-87; Sutter Instrument Co.). The pipette solution contained 120 mM potassium gluconate, 20 mM KCl, 1.5 mM MgATP, 5 mM EGTA, 10 mM HEPES, pH adjusted to 7.4 with KOH. The permeability ratio of K ϩ /Na ϩ , ␣, was calculated from the reversal potential measured in standard Tyrode solution using the following constant field equation, and assuming that [K ϩ ] i ϭ 145 mM and [Na ϩ ] i ϭ 5 mM. Conventional whole cell voltage clamp recordings were performed using an Axopatch 200B amplifier interfaced with a Digidata 1200 A/D converter operated using pClamp software (Axon Instruments, Foster City, CA). All of the experiments were performed at room temperature. Whole cell capacitance was determined from capacitative transient decay in current recordings following voltage steps Ϯ 10 mV from the holding potential, and at least 80% series resistance compensation was achieved in all of the reported experiments. The protocols used in the specific experiments are described in the figure legends. In all protocols a 20-ms duration 20 mV step from the holding potential of Ϫ80 to Ϫ100 mV was applied at the start of each sweep to enable off-line leak correction. We assumed that the leak was linear in the voltage range Ϫ120 to ϩ40 mV. Data analysis was performed using the Clampfit module of the PClamp software. The data are expressed as the means Ϯ S.E. for n experiments, and analysis of variance was performed using Microsoft Excel. A p value of Ͻ0.05 was considered significant.
For electrophysiology experiments the peptides were prepared as 0.5 mM stock solutions in either normal Tyrode or Na ϩ -free external solution, and aliquots were stored at Ϫ20°C. Once thawed aliquots were stored at 4°C for up to 2 weeks. The aliquots were diluted as required on the day of experiment. The peptides were applied to cells using a picospritzer II (Intracell, Cambridge, UK) to ensure rapid application (typically less that 20 ms) and minimize the amount of peptide used in each experiment. A new coverslip was used for each experiment to ensure no residual contamination of cells with peptides.
Analysis of Peptide Binding Data-Apparent on rates, () and off rates (k Ϫ1 ) for peptide binding were obtained by fitting single exponential functions to the data for onset of current block and recovery of current following washout of the peptide. The on rate (k ϩ1 ) was calculated using the formula, To obtain the time constant for the apparent on rate of peptide binding at negative potentials, it was necessary to correct for channel deactivation (see Fig. 9). We assumed that the rate of binding of the peptide to the channel was independent of the rate of deactivation, and therefore, where A and C are constants, obs is the observed single exponential time constant measured from the rate of change in current following addition of the peptide, deact is the time constant of deactivation (estimated from the single exponential fit to the current recorded in the absence of peptide), and on is the apparent time constant for peptide binding, i.e. on ϭ 1/.

Circular Dichroism
Spectropolarimetry-Far-UV CD spectropolarimetry spectra of the S5P HERG peptide in 10 mM sodium phosphate buffer with and without 100 mM SDS at 20°C and pH 3.0 are shown in Fig. 2. The large minimum between 195 and 200 nm and ellipticity close to zero at 222 nm for the S5P peptide in aqueous solution (thin line) indicates that the peptide does not have a well defined secondary structure under these conditions. This outcome prompted our investigation of the structural properties of the peptide under membrane-like conditions. A dramatic change in the CD spectropolarimetry spectral profile was observed upon addition of 100 mM SDS (Fig. 2, thick line); positive ellipticity was observed between 190 and 195 nm, the position of the minimum shifted to 205 nm, and a marked shoulder was present at 222 nm. This indicated that the S5P peptide contains helical elements in the micellar environment at pH 3.0. Similar CD spectropolarimetry profiles were obtained at pH 7.0 (also see Fig. 10) and over the temperature range 20 -30°C. NMR spectroscopy experiments were therefore performed at a pH level of ϳ3 to facilitate data collection.
NMR Spectroscopy-The two-dimensional NOESY spectrum of the S5P peptide in aqueous solution at 25°C had only a few very weak cross-peaks, suggesting that the peptide had a flexible, predominantly random coil structure (data not shown) that was consistent with the CD spectropolarimetry data (Fig.  2). The number and intensity of cross-peaks in the two-dimensional NOESY spectrum improved markedly upon the addition of 100 mM SDS. The appearance of amide-amide cross-peaks indicated that the peptide contained some helical structure when in the presence of SDS micelles (Fig. 3), in accordance with the CD spectropolarimetry results.
Assignments of proton resonances from the S5P peptide were made using standard methods (49) of analyzing two-dimensional total correlation spectroscopy and NOESY spectra. The cross-peaks were generally broader than those in aqueous solution, presumably because of the increased correlation time of the peptide and possible conformational averaging and slow exchange between conformations. At 20°C, NOE cross-peaks were too broad to analyze and in several instances overlapped with other cross-peaks, especially in the "fingerprint" region. Increasing the temperature brought about significant narrowing of the cross-peaks; however, some NOE cross-peaks became less intense, and a few disappeared. The NOESY spectrum obtained at 30°C was seen to be the best for resonance assignment and structural calculations, but where possible, spectra obtained at 20°C or 37°C were used to resolve peaks that overlapped at 30°C (see Table S1 in the supplementary data).
Before doing the structure calculations, the chemical shift values obtained for C␣ protons were analyzed to provide information on possible secondary structure present in the peptide. For this we used the chemical shift index method of Wishart et al. (50). A prominent grouping of chemical shift deviations below Ϫ0.1 ppm for most residues between 16 and 22 implied a prominent helix in this part of the molecule (Fig. 4). It is also possible that a short helix is present in the region formed by residues 39 -41 because they form a small cluster of three with low chemical shift deviations. The absence of a group of residues with chemical shift deviations greater than ϩ0.1 ppm indicated that no ␤-sheet structure is expected for the S5P HERG peptide. The rest of the S5P HERG peptide, residues 1-15 and 23-38, is likely to be flexible or in random conformations.
In addition to the presence of significant amide-amide NOE connectivities, the two-dimensional NOESY spectrum of S5P showed many medium range connectivities between residues 14 -24 and 38 -42 ( Fig. 4). This supported the proposal that these regions formed helices as indicated by the CD spectropolarimetry spectra (Fig. 2) and chemical shift indices (Fig. 4). No long range NOEs were detected, implying that the S5P peptide does not have a well defined tertiary structure. A total of 416 distance (upper limit) constraints (of which 74 were medium range) were obtained from the 200-ms NOESY spectrum (see Table S2 in the supplementary data); the constraints were used in structure calculations using DYANA and CNS (see "Experimental Procedures"). The best 20 structures, namely those with the lowest penalty values, were considered to be representative of the structure and were therefore used in the figures shown here.
Structural Fold-Comparison of the 20 best S5P structures showed that S5P did not have a distinct three-dimensional fold in SDS micelles. However, two regions in the peptide were well defined; this was indicated by their relatively low root mean square deviation values, with respect to the local mean structure. These were regions defined by residues 14 -24 and 35-42 that had mean backbone root mean square deviations of 0.31 and 0.23 Å, respectively (see Table S2 in the supplementary data). These two well defined regions encompassed all residues displaying medium range NOE values and were predicted to be helical by the CSI analysis (Fig. 4). Also, PROMOTIF analysis (46) of the ensemble of S5P structures predicts that the regions defined by residues 16 -25 and 35-42 form ␣-helices in the presence of SDS micelles. Fig. 5A shows the superpositions of 20 structures on to the backbone of the well defined region defined by residues 14 -24. The rest of the molecule, except for the region defined by residues 35-42, appears disordered ( Fig.  5B). It should also be noted that because the region defined by residues 24 -35 is disordered (represented by a dashed line in Fig. 5B), it is not possible to define the spatial relationship between the two helical regions. The conclusion that the regions defined by residues 1-14 and 24 -35 lacked defined structures in solution was based on the absence of medium and long range NOEs.
In the KcsA structure (22), the ends of the extracellular loop that corresponds to S5P in HERG, (i.e. Ala 50 and Tyr 62 in KcsA) are close together, being separated by only ϳ11 Å. We attempted to constrain the ends of the S5P peptide by introducing a "dummy" distance constraint of 10.9 Å between the backbone C␣ of Ala 1 and that of Tyr 42 in the NMR spectroscopy structure calculation. This could have provided useful insights into the three-dimensional fold of the peptide and also information about the relative orientation of the amphipathic ␣-helix (see below) with respect to rest of the molecule. However, the calculation failed to indicate a distinct structural fold.
The Amphipathic Helix-Detailed analysis of the first well defined region showed that it was an amphipathic helix, with the side chains of Trp 16 , Leu 17 , Leu 20 , and Ile 24 forming the hydrophobic face and the side chains of His 18 , Asn 19 , Asp 22 , and Gln 23 defining a hydrophilic edge. (Fig. 5). The side chains of residues 16 -20 were well defined, with root mean square deviation values of 0.23 Å when their side chain heavy atoms were superimposed on each other (see Table S2 in the supplementary data). There were numerous medium range NOEs between the Trp 16 aromatic side chains and those of Ile 14 , Leu 17 , and Leu 20 ; this helped to define the conformation of the hydrophobic region of the molecule, placing Trp 16 as the center of this site.
Role of Amphipathic Helix in Inactivation-The unusual kinetics of HERG K ϩ channels, viz. slow activation/deactivation and rapid inactivation/recovery from inactivation, are illustrated in Fig. 6A. Depolarizations to ϩ40 mV cause HERG channels to open slowly and then inactivate rapidly, resulting in a relatively small current flow. Subsequent repolarization to Ϫ120 mV results in rapid recovery from inactivation and hence a large increase in inward current followed by a slower decay in the current as the channels deactivate. To investigate whether exogenous application of the S5P peptide would have any effect on full-length HERG channels, we superfused cells with 1 M S5P peptide, and the membrane was depolarized from Ϫ80 to ϩ40 mV for 0.5 s followed by a hyperpolarization to Ϫ120 mV for 1 s, and the protocol repeated every 5 s. Application of the 42-residue S5P peptide caused a reversible suppression of HERG current (Fig. 6, B and C). Application of 1 M S5P peptide to CHO cells transfected with HEAG or rELK2 channels resulted in no reduction in current (data not shown).
The apparent on rates and off rates for peptide binding to the channels were obtained by fitting single exponential functions to the data for onset of current block and recovery of current following washout of the peptide. In the example illustrated in Fig. 6C, the time constants for peptide binding and wash-off were 17.8 s ( ϭ 0.056 s Ϫ1 ) and 53.7 s (k Ϫ1 ϭ 0.019 s Ϫ1 ), respectively. The mean values obtained from four separate cells were ϭ 0.045 Ϯ 0.007 s Ϫ1 and k Ϫ1 ϭ 0.024 Ϯ 0.005 s Ϫ1 (n ϭ 4). Substituting values for and k Ϫ1 ϭ into Equation 2 gave an on rate (k ϩ1 ) of 2.1 Ϯ 0.4 ϫ 10 4 M Ϫ1 s Ϫ1 , and the dissociation constant for peptide binding ( To investigate further the binding of the S5P peptide to HERG channels, we used a picospritzer to rapidly apply peptides and minimize the amount of peptide used in each experiment. First we investigated the dose response of the effect of S5P peptide by applying the peptide at doses ranging from 0.1 to 100 M (Fig. 6D) using the same voltage protocol as used in Fig. 6A. The IC 50 for inhibition of the tail current at Ϫ120 mV measured from the dose-response curve was 1.9 M (Fig. 6D). This value is similar to that measured from the on and off rates To investigate the voltage dependence of the effect of S5P peptide on HERG channels, we used a voltage protocol where within each sweep channels were first activated by stepping to ϩ40 mV for 500 ms and then stepped to a test voltage in the range ϩ30 to Ϫ110 mV followed by a Ϫ120 mV step for 800 ms to fully deactivate the channels. The protocol was then repeated with peptide rapidly applied for 1 s during the test voltage step (Fig. 7A). Based on the on rate for S5P peptide binding to HERG K ϩ channels calculated above, we used 100 M S5P peptide in these experiments so as to achieve Ͼ90% binding within 1 s of application.
The effect of rapid application of 100 M S5P peptide was voltage-dependent. For example, application of S5P peptide caused an increase in outward current at ϩ30 mV (Fig. 7A, solid arrows) but a significant decrease in the current at Ϫ50 mV (Fig. 7A, dotted arrows). In the protocol illustrated in Fig.  7A, we waited 60 s between sweeps to ensure that the effect of the peptide had washed out (also see Fig. 7B). The traces in Fig.  7B show the average (Ϯ S.E.) current recorded during five successive sweeps at 60-s intervals where the test potential was Ϫ70 mV for all five sweeps. Both the control currents (Fig.  7B, thin line) and currents recorded during application of the S5P peptide (Fig. 7B, thick line) were superimposable. This illustrates two important points. First, during the 60-s intersweep interval, the effect of the peptide was completely washed out, and second, the effect of the peptide was very reproducible. Also of note from Fig. 7B is that in the presence of the S5P peptide the current recorded at Ϫ70 mV reversed from an outward current to a small inward current, suggesting that application of the peptide altered the selectivity of the channel (see below).
Examples of currents recorded at voltages in the range ϩ30 to Ϫ110 mV (in 10 mV steps) using the voltage protocol illustrated in Fig. 7A are illustrated in Fig. 8. At positive potentials (panels a-c) application of the peptide resulted in an increase in current that was maintained for the duration of application of the peptide. At voltages in the range 0 to Ϫ20 mV (panels d-f), there was an initial increase in current followed by a slower decline in current during application of the peptide. In the voltage range Ϫ30 to Ϫ60 mV (panels g-j), there was a rapid decrease in current during application of the peptide. The most dramatic effects, however, occurred at Ϫ70 (panel k) and Ϫ80 mV (panel l) where the peptide caused a reversal in current flow. This suggests that the channel has lost its selectivity for K ϩ over Na ϩ (the only other cation present at significant concentrations in the extracellular solution). At progressively more negative potentials it became difficult to discern an effect of the peptide. From the data in Fig. 6, however, we know that the peptide can inhibit the channel at Ϫ120 mV; therefore the lack of apparent effect at the most negative voltages in Fig. 8 is presumably because the rate of channel deactivation is as fast if not faster than the rate of peptide binding to the channel at these voltages.
To quantify the effect of the S5P peptide on HERG channels, we analyzed the rates of change in current magnitude following the addition of the peptide (i.e. the apparent on rates; Fig. 9, A and B), the maximum increase (or decrease) in current during the 1-s application of the peptide (Fig. 9, A and C), and the change in reversal potential caused by application of the peptide (Fig. 9D). In a separate series of experiments we also measured the off rate for peptide dissociation from the channels at ϩ30 mV (Fig. 9E).
At potentials in the range Ϫ20 mV to ϩ30 mV, we fitted a single exponential to the increase in current to obtain apparent on rates (Fig. 9A, panels a and b). At Ϫ20 mV there were two phases to the change in current during application of S5P peptide, i.e. an initial increase in current followed by a slower decrease in current. We fitted single exponential functions to both phases; in the example shown in Fig. 9A (panel b), the time constant for the initial increase was 89 ms, and the time constant for the subsequent decrease was 1.4 s. To obtain the time constant for the apparent on rate at more negative potentials, it was necessary to correct for channel deactivation (see "Experimental Procedures"). In the example illustrated in Fig.  9A (panel c), the time constant of deactivation in the control trace (no peptide present) was 223 ms, and the time constant of current decline following application of the S5P peptide was 65 ms. Substituting these values into Equation 3 gives an apparent time constant for peptide binding of 92 ms, which is very similar to that obtained by dividing the two current traces (88 ms; as shown in Fig. 9A, panel c, inset). For voltages in the range ϩ30 to Ϫ10 mV, the apparent time constant for peptide binding, ϳ100 ms, did not vary with voltage (Fig. 9B). However, in the voltage range Ϫ20 to Ϫ80 mV, the apparent time constant for peptide binding decreased significantly as the voltage became more negative (decreasing from 905 Ϯ 59 ms at Ϫ30 mV to 47 Ϯ 8 ms at Ϫ80 mV).
We measured the maximum increase (at potentials Ն Ϫ20 mV) and decrease (at potentials Յ Ϫ20 mV) in current during application of the S5P peptide from the single exponential curves fitted to the data after correcting for channel deactivation if appropriate (Fig. 9A, panel c, inset). The maximum change in current during a 1-s application of 100 M S5P peptide varied significantly with voltage. Note that the decrease at Ϫ70 mV (113 Ϯ 8%) was over 100%, indicating that the current had reversed from an outward to an inward current.
To estimate the change in reversal potential, we fitted single exponentials to the decaying phase of currents recorded in the voltage range Ϫ60 to Ϫ90 mV and extrapolated the fits back to the start of the test voltage step (Fig. 9D). In the example illustrated in Fig. 9D, the reversal potential changed from Ϫ80 to Ϫ64 mV. The mean change in reversal potential was from Ϫ83 Ϯ 1.5 mV (normal Tyrode solution) to Ϫ68 Ϯ 2.2 mV (in the presence of S5P peptide, n ϭ 6). From the change in reversal potential following application of S5P peptide, we estimate, using Equation 1, that the permeability ratio for K ϩ over Na ϩ (P K /P Na ) was reduced from 188 Ϯ 38 to 27 Ϯ 5 (n ϭ 6).
To estimate the off rate for peptide dissociation from the channels at ϩ30 mV, we depolarized cells to ϩ30 mV, and after 1 s (when a steady-state current had been reached) we spritzed on the peptide for 1 s and then maintained the cell at ϩ30 mV for a further 3 s (Fig. 9E). In the example illustrated in Fig. 9E, the time constant for the increase in current in the presence of S5P peptide was 123 ms, and the time constant of the decrease in current following wash-off of the peptide was 2.26 s. The time constant for wash-off, 2.06 Ϯ 0.56 s (n ϭ 4), was consideffect of S5P peptide ( ϭ 17.8 s, ϭ 0.056 s Ϫ1 ) and the wash off of the S5P peptide effect ( ϭ 53.7 s, k Ϫ1 ϭ 0.019 s Ϫ1 ). D, dose dependence of effects of S5P peptide on currents recorded at Ϫ120 mV (n ϭ 3-4 at each dose), using the protocol shown in A. The IC 50 was obtained by fitting a Hill function to the mean data.
FIG. 6. A, typical example of ion currents recorded from a CHO-HERG cell during a step depolarization to ϩ40 mV for 0.5 s followed by a step to Ϫ120 mV. At a holding potential of Ϫ80 mV, channels are in the closed state (C). During the step to ϩ40 mV channels open (O) slowly but inactivate (I) rapidly resulting in little current flow (thin line). During the step to Ϫ120 mV, the channels rapidly recover from inactivation (on the time scale of ms), resulting in a large increase in current (thick line), and then deactivate slowly (on the time scale of tens to hundreds of ms), resulting in a decay in current (dashed line) back to the zero current level (indicated by dotted horizontal line at the left of the trace). The dotted box on the voltage protocol shows the region of the traces depicted in B. B, typical example of currents recorded from a CHO-HERG cell using the voltage protocol shown in A before (i), after 1 min superfusion with 1 M S5P peptide (ii), and 6 min after washout of the peptide (iii). The horizontal dotted line shows the zero current level. The dotted vertical arrow indicates the point at which tail currents at Ϫ120 mV were measured (see C). C, changes in peak tail current recorded at Ϫ120 mV prior to superfusion, during superfusion, and following washout of 1 M S5P peptide. The peptide was applied for 1 min (shown by the solid bar at the top of the graph). The labels (i), (ii), and (iii) indicate the time points for which traces are shown in B. The thin lines show single exponential fits to the data for the onset of the erably shorter at ϩ30 mV than was seen using the protocol in Fig. 6, where the time constant for wash-off was 42 Ϯ 9 s (n ϭ 4; see above). This suggests that dissociation of the peptide from the channels is voltage-dependent. Substitution of the time constants measured at ϩ30 mV into Equation 2 yielded an on rate of 5.5 ϫ 10 4 M Ϫ1 s Ϫ1 and a peptide affinity of 8.9 Ϯ 2.2 M (n ϭ 4).
The above results, most notably the altered ionic selectivity, suggest that the exogenously applied S5P peptide is interacting either directly with the selectivity filter of the channel or binding to the channel in such a way as to cause a change in the conformation of the selectivity filter. To determine which region of the S5P peptide was responsible for the interaction with the pore region of the HERG channel, we synthesized two additional peptides. In the first peptide the amphipathic helix was deleted and replaced with a (GGGS) 3 linker (del-helix peptide), and in the second the N-and C-terminal regions were deleted, leaving just the amphipathic helix with four residues at either side (helix peptide). All three peptides, S5P, del-helix, and helix peptides, adopted a random coil conformation in aqueous solution at pH 7.0 (Fig. 10A). In the presence of SDS micelles, the del-helix peptide remained in a random coil conformation, and the S5P peptide contained some ␣-helical ele-ments (the spectra obtained at pH 3.0 (Fig. 2) and pH 7.0 ( Fig.  10A) were indistinguishable), and the helix peptide adopted an almost pure ␣-helix, as expected. Typical examples of HERG tail currents recorded at Ϫ80 mV during control sweeps or during application of the S5P, del-helix, or helix peptides are shown in Fig. 10B. Application of the del-helix peptide had no significant effect on HERG currents. The helix peptide, however, had a very similar effect to that observed with the fulllength S5P peptide (also see below) .
Typical examples of the effects of the helix peptide on HERG tail currents recorded at 0, Ϫ40, and Ϫ80 mV (using the same voltage protocol as shown in Fig. 7A) are illustrated in Fig. 11. Application of 100 M helix peptide resulted in increased current at 0 mV, a decreased current at Ϫ40 mV, and a clear reversal of current flow at Ϫ80 mV (Fig. 11A). The apparent time constants for helix peptide binding to the HERG channel are plotted in Fig. 11B, and the maximal increase (at potentials Ͼ Ϫ20 mV) and decrease (at potentials Ͻ-20 mV) in current observed during 1 s of application of 100 M helix peptide are shown in Fig. 11C. Qualitatively, the effects seen with application of the helix peptide are very similar to those seen with the S5P peptide (compare Figs. 9 and 11). There were, however, some significant differences. The apparent time FIG. 7. A, a typical example of a family of currents recorded during a double pulse protocol to examine the voltage dependence of S5P peptide effects on HERG channels. The cell was depolarized from a resting membrane potential of Ϫ80 to ϩ40 mV for 500 ms and then stepped to potentials in the range ϩ30 to Ϫ110 mV for 2 s. The channels were deactivated by hyperpolarizing the cell to Ϫ120 mV for 800 ms, and the protocol was repeated except that during the second test pulse 100 M S5P peptide was applied for 1 s (as indicated by the solid bar). The cell membrane potential was then held at Ϫ80 mV for 53 s between each sweep. The thick lines show the current traces recorded for the ϩ30 mV test pulse (solid arrows) and Ϫ50 mV test pulse (dotted arrows). The addition of peptide caused a significant increase in current at ϩ30 mV but a marked suppression of current at Ϫ50 mV. The horizontal dotted line indicates zero current. B, a typical example of the mean Ϯ S.E. of the current recorded during five successive sweeps at 60-s intervals where the test step was Ϫ70 mV for all five sweeps. The voltage protocol is illustrated above the current trace, and the dotted boxes (thin, control; thick, with S5P peptide) indicate the regions that are plotted. 100 M S5P peptide was applied for 1 s as indicated by the thick line in the voltage protocol. The effect of the peptide was reproducible (as indicated by the small error bars) and the effect of the peptide washed off within 60 s. The horizontal dotted line indicates the zero current level. It is worth noting that the current recorded at Ϫ70 mV in the presence of the peptide reversed from an outward current to an inward current. constants for peptide binding were faster for the helix peptide than for the S5P peptide (compare Figs. 9B and 11B), the off rate was also slightly faster for the helix (time constant for dissociation at ϩ30 mV was 0.85 Ϯ 0.15 s (n ϭ 5) for the helix peptide compared with 2.06 Ϯ 0.56 s (n ϭ 4) for the S5P peptide), and the maximum increase in current seen at positive potentials was less for the helix peptide than for S5P peptide (Table I). However, the shift in the reversal potential (14 Ϯ 2 mV, n ϭ 4) following addition of the helix peptide was similar to that seen with the S5P peptide (16 Ϯ 2 mV, n ϭ 6). The del-helix peptide, 100 M, had no effect on the reversal potential of HERG currents and caused no increase in current at positive potentials; rather it caused a very modest decrease in current at all potentials (Table I).
The shift in reversal potential following application of either the 42-residue S5P peptide or the 19-residue helix peptide was consistent with a decreased selectivity for K ϩ over Na ϩ . To further investigate this possibility we looked at the effects of both peptides when all external Na ϩ was replaced with NMDG ϩ . In the absence of external Na ϩ , acute application of either the S5P peptide or the helix peptide caused only modest decreases in current at all test potentials (Fig. 12A). Furthermore there was no current reversal at Ϫ80 mV following application of either peptide in NMDG ϩ solutions. It should be noted, however, that both the S5P and helix peptides caused inhibition of current at Ϫ120 mV when applied for 1 min in NMDG ϩ external solutions (Fig. 12B). This confirms that the peptide was still active, but either its apparent affinity was much reduced or the effect it has on the channel was substantially altered in the absence of external Na ϩ .

DISCUSSION
The S5P Linker Contains an Amphipathic Helix-In this work we have identified an amphipathic ␣-helix that is present in the extracellular linker connecting the outer ends of S5 and the pore helix of the HERG K ϩ channel. The 42-residue peptide spanning this region is unstructured in aqueous medium but contains ␣-helical elements when in the presence of SDS micelles. The major helical element extends from Trp 585 to Ile 593 (WLHNLGDQI; Fig. 5) and has a well defined hydrophobic face. Interestingly, this helix contains a glycine, Gly 590 , which in the center of a helix usually has a destabilizing effect (51). It is possible that the presence of the glycine is the reason that the S5P peptide is unstructured in water. It appears that an interaction between the hydrophobic surface of the helix and the hydrocarbon chains of the SDS stabilizes the helix in vitro. In vivo and in the native protein, the stabilizing force may be provided by the helix interacting with the membrane or with other parts of the protein.
The amphipathic ␣-helix we have identified (Trp 585 -Ile 593 ) corresponds closely with the predictions originally made by Pardo-Lopez et al. (23), based on the effects of cysteine scan-  (I S5P ) by the control current (I con ; see inset; also see text for details of method used to deconvolve time constants). The inset also highlights how we calculated the maximum decrease in current for test potentials in the range Ϫ30 to Ϫ70 mV. In this example the maximum decrease was 140%, indicating that the current had reversed from outward to inward. B, the means Ϯ S.E. for the time constants measured from the ascending phase (open symbols, n ϭ 4) and descending phase (after correction for deactivation; closed symbols, n ϭ 5) of currents following addition of S5P peptide. There were no significant differences between the time constants measured at positive potentials (analysis of varianne). Time constants at voltages in the range Ϫ70 to Ϫ30 mV were significantly different from each other. C, summary of the maximum change (increase at positive potentials (open symbols) or decrease at potentials Յ Ϫ20 mV (closed symbols)) in current following application of 100 M S5P peptide (mean Ϯ S.E., n ϭ 4 -5). Note there are two points for Ϫ20 mV reflecting the fact that the current initially increased but subsequently decreased in the presence of the peptide at this voltage. D, typical examples of current traces recorded at Ϫ70, Ϫ80, and Ϫ90 mV during control conditions and Ϫ60, Ϫ70, and Ϫ80 mV in the presence of S5P peptide. The reversal potential was calculated from the extrapolated current at the start of the test pulse. Extrapolations were obtained by fitting a single exponential to the slow decaying phase of the current. In this example the reversal potential was shifted from 80 to 64 mV during addition of the S5P peptide. E, typical examples of currents recorded at ϩ30 mV during application and wash-off of 100 M S5P peptide (indicated by a thick bar above the voltage protocol). Thin lines show single exponential fits used to obtain time constants for the apparent binding (123 ms) and wash-off (2.26 s) of the peptide. ning mutagenesis on ErgToxin binding to the HERG K ϩ channel. Therefore, the amphipathic helix we have identified in the isolated peptide is almost certainly present in the intact protein as well.
Two pieces of evidence support the suggestion that the amphipathic ␣-helix interacts with another part of the HERG protein. First, in the ensemble of the 20 lowest energy structures, the hydrophobic residues Trp 585 , Leu 586 , and Leu 589 are contained within a region that is very tightly constrained (root mean square deviation for the side chain heavy atoms of residues WLHNL in the peptide was 0.23 Å). Furthermore these hydrophobic residues are 100% conserved in all members of the ERG/EAG/ELK channel family (Fig. 1), whereas there is much less conservation among the hydrophilic residues. The conser-vation of sequence as well as the tight structural constraints argues in favor of the hydrophobic surface, representing a region of specific protein-protein interaction. Second, application of exogenous S5P peptide to wild type HERG K ϩ channels caused dramatic changes in channel properties (see below).
The Amphipathic Helix Interacts with the Pore Region of the Channel-Exogenous application of the S5P peptide had at least two effects on HERG currents. First it suppressed the current (Figs. 6 and 12B). Second, it appeared to disrupt inactivation (e.g. it caused an increased current flow at positive potentials; Fig. 8) and reduced the selectivity for K ϩ over Na ϩ (Fig. 9D). We have classed the last two effects together because it is likely that they are related (see below). Furthermore, these effects were also seen following superfusion of CHO-HERG cells with the helix peptide that corresponded to residues Ser 581 -Ser 599 but not with a peptide in which the amphipathic ␣-helix was replaced with a (GGGS) 3 linker. Thus it seems that the major site of interaction between the S5P peptide and the rest of the channel involves the amphipathic ␣-helix.
Current Suppression-When cells were exposed for prolonged periods to S5P peptide, current was suppressed in a dose-dependent manner with an IC 50 of ϳ1.9 M (Fig. 6D). This inhibitory effect of the S5P peptide was independent of external Na ϩ (Fig. 12B); however, in the absence of external sodium, the potency of the peptide appeared to be much reduced. For example 100 M peptide only caused ϳ60% inhibition after 1 min in cells superfused with Na ϩ -free solution (Fig. 12B), but ϳ93% inhibition after 10 s in cells superfused with Na ϩ -external solution (Fig. 6D).
Disrupted Inactivation and Altered Selectivity-The addition of 100 M S5P peptide to HERG K ϩ channels caused a depolarizing shift in the reversal potential by 16 Ϯ 2 mV (Fig. 9D and Table I). However, when all external Na ϩ was replaced by NMDG ϩ , the S5P peptide no longer caused a shift in reversal potential (Table I). These results indicate that the channels had a decreased selectivity for K ϩ over Na ϩ . If a decrease in selectivity for K ϩ relative to Na ϩ was the only effect of acute application of S5P peptide, then one would not have expected an increase in current at positive voltages (Fig. 8, panels a-c), because the positive shift in reversal potential would result in a smaller driving force for outward current flow. The relatively small current flow through HERG K ϩ channels at positive potentials is due to rapid voltage-dependent inactivation (1, 10 -12). Thus one possibility is that in addition to altering the selectivity of the channels, the peptide has altered inactivation. Such a hypothesis is also consistent with previous reports showing that many of the mutations in the S5P linker that affect inactivation also affect the selectivity for K ϩ over Na ϩ (19). The association between disruption in inactivation and changes in ionic selectivity is one of the strongest pieces of evidence supporting the "collapse of the pore" model of inactivation in HERG K ϩ channels (3). Thus we suggest that the exogenously applied S5P peptide either itself directly binds to the pore region and thereby disrupts inactivation and K ϩ -Na ϩ selectivity, or it binds to the outer pore region in such a way that it induces a conformational change in the selectivity filter region.
The very significant reduction in the effect of acute application of S5P peptide when external Na ϩ was replaced by NMDG ϩ (compare Figs. 9 and 12) suggests that the increase in current seen with application of S5P peptide at positive potentials is Na ϩ -dependent. Previous work from Balser and colleagues (52) has shown that inactivation of HERG K ϩ channels is promoted by extracellular Na ϩ . Thus it may be that the S5P peptide is competing with external Na ϩ for binding to the site that when occupied by Na ϩ promotes inactivation (52). FIG. 10. A, far-UV CD spectropolarimetry spectra for S5P peptides. All three peptides display a predominantly random coil conformation in aqueous environment (thin lines) with little evidence for secondary structure. However, in the presence of SDS micelles (thick lines) both the S5P peptide and the helix peptide display elements of helical structure, whereas the del-helix remains unstructured. Calculation of the mean residue molar ellipticity, which takes into account the number of peptide bonds in the peptide, indicates that the helix peptide is almost completely helical in SDS micelles, with very little of the sequence in alternative conformations, whereas the S5P contains significant random coil content in addition to the helical elements. This is consistent with the experimentally determined structure of the S5P peptide and what would be expected from secondary structure prediction algorithms. B, typical examples of HERG currents recorded at Ϫ80 mV test potentials under control conditions (thin traces) and following addition of peptides (thick traces). S5P peptide causes a shift to inward current, the del-helix peptide has minimal effect on the current and the helix peptide has a similar effect to the full-length peptide. The dotted line in each trace indicates the zero current level.
Is Binding of S5P to the HERG K ϩ Channel Voltage-dependent?-The effect of the acute application of S5P peptide to HERG K ϩ channels was clearly voltage-dependent (Figs. 8 and  9). The voltage dependence of the time constants for change in current following addition of S5P peptide (Fig. 9B) also suggests that there are two components to peptide binding. First, at positive potentials the time constant of current increase (Fig.  9B, open symbols) is rapid and voltage-independent. Second, at potentials in the range Ϫ20 to Ϫ80 mV, the time constant for the decrease in current (Fig. 9B, closed symbols) is voltage-dependent. It should be noted that in our analysis of the apparent on rates at negative voltages, we assumed that the effect of peptide binding to the channels was independent of deactivation (see "Experimental Procedures" and Fig. 9A). This is the simplest model that is consistent with the data that we have; however, we cannot assume that the model is necessarily correct. If our assumption regarding independence of peptide binding and deactivation is incorrect, then it is possible that part of the "voltage dependence" of the time constants in the voltage range Ϫ20 to Ϫ80 mV seen in Fig. 9B may reflect the voltage dependence of the rates of deactivation in this voltage range rather than voltage dependence of binding per se. Conversely, the off rate measured at ϩ30 mV (a voltage at which no deactivation was observed) was considerably faster (0.48 s Ϫ1 ; Table I) than the off rate measured from Fig. 6C (0.024 s Ϫ1 ) when cells were held at Ϫ80 mV between pulses. From these values the calculated on rates were 5.5 ϫ 10 4 M Ϫ1 s Ϫ1 at ϩ30 mV and 2.1 ϫ 10 4 M Ϫ1 s Ϫ1 at ϳ-80 mV. Thus we can conclude that there is a voltage-dependent component to peptide binding and that the off rate appears to be more voltage-sensitive than the on rate. However, the precise influence of voltage on peptide binding in the voltage range Ϫ20 to Ϫ80 mV remains to be determined.
The augmentation of current at positive potentials but inhibition at negative potentials following application of the S5P peptide (Fig. 9C) is reminiscent of the effect of Ba 2ϩ on HERG K ϩ channels (53), i.e. 2 mM Ba 2ϩ blocks HERG currents at negative voltages but increases the current at positive voltages, an effect that has been attributed to a voltage-dependent competition between Ba 2ϩ and Na ϩ for binding to an outer porebinding site (52). Although the voltage protocols used in the earlier Ba 2ϩ studies (52,53) and this study are different, it is possible that the S5P peptide, like Ba 2ϩ , may be competing with Na ϩ for binding to an outer pore-binding site and that this competition is voltage-dependent. The significant change in the effect of the peptides on HERG channels when external Na ϩ was replaced with NMDG ϩ also supports the hypothesis that the peptide is competing with external Na ϩ for binding to an outer pore-binding site.
Comparison of the Effects of the S5P Peptide and the Helix Peptide-The effects of adding the S5P peptide to HERG channels could be reproduced, qualitatively, with a 19-residue peptide containing only the amphipathic helix and four flanking residues on either side. There were, however, some quantitative differences in the effects of the two peptides on HERG channels. First, the amphipathic helix appears to bind more rapidly. For example the time constant for binding of 100 M peptide at 0 mV was 145 Ϯ 36 ms (n ϭ 5) for S5P peptide compared with 36 Ϯ 7 ms (n ϭ 3) for the helix peptide (Table I). Second, the maximum increase in current seen during peptide application at positive potentials was greater for the S5P peptide than for the helix peptide. For example the maximum increase at ϩ30 mV was 56 Ϯ 22% (n ϭ 4) for the S5P peptide and 9 Ϯ 3% (n ϭ 3) for the helix peptide (Figs. 9C and 11C and Table I). These data suggest that a region of the S5P peptide not present in the helix peptide, for example the charged motif  11. A, typical examples of the effect of 100 m helix peptide applied for 1 s (denoted by solid bar on voltage protocol) on HERG tail currents recorded during voltage steps to 0, Ϫ40, and Ϫ80 mV. The control traces are shown using thin lines, and the peptide traces are shown using thick lines. The horizontal dotted lines indicate the zero current level. B, summary of the time constants for increases in current (open symbols) and decreases in current (closed symbols) following application of the peptide (mean Ϯ S.E., n ϭ 4) analyzed as described in the legend to Fig. 9B. C, maximal percentage change in current (open symbols, increase in current for voltage range ϩ30 to Ϫ20 mV; closed symbols, decrease in current for voltage range Ϫ20 to Ϫ70 mV) following the addition of 100 M helix peptide (mean Ϯ S.E., n ϭ 4), analyzed as described in the legend to Fig. 9C. The error bars are in general smaller than the size of the symbols.
at the C-terminal end, inhibits access of the amphipathic helix to a binding site on the channel. However, once it has bound the S5P peptide is more effective than the helix peptide at altering channel function. Whether such differences are of physiological significance, however, are debatable given that in the native channel movement of the amphipathic helix is likely to be constrained by its attachments to the S5 and pore helices.
How Does the S5P Peptide Interact with the Rest of the HERG Channel?-Given that the S5P linker is critical for inactivation of HERG K ϩ channels and inactivation in HERG K ϩ channels is voltage-dependent (1, 10 -12), it is possible that voltage-dependent binding of the S5P linker to the outer pore of the channel could contribute to the voltage-dependent inactivation of HERG K ϩ channels. There are a number of charged residues in the S5P linker including an aspartate (Asp 591 ) in the amphipathic ␣-helix (Fig. 1) that in theory could contribute to voltage-dependent binding of the S5P linker to the rest of the channel. The observation that mutation of Asp 591 to a cysteine disrupts inactivation of HERG K ϩ channels (19) suggests that this residue may be important for binding of the amphipathic helix to the rest of the channel, but whether it could contribute to voltage-dependent binding of the amphipathic helix to the pore region of the channel has not been tested.
We were not able to define a unique three-dimensional fold for the S5P loop of the HERG K ϩ channel, despite constraining the ends of the peptide to the homologous residues in the KcsA structure (22). This result is consistent with the NMR spectroscopy data, however, showing that significant stretches of the linker, Ala 571 -Gly 584 and Gly 594 -Gly 603 , may be highly flexible. Flexibility in this region could be advantageous in that it may allow significant movement of the amphipathic ␣-helix, which has been suggested to occur during inactivation (19). To further address this issue will require additional experimental data to help constrain the possible orientations of the S5P linker and to identify the specific site(s) of interaction between the S5P linker and the rest of the channel. Such data could be provided by mutant cycle analysis experiments analogous to those used to identify the sites of interaction between the scorpion toxin BeKm-1 and the outer vestibule of the HERG K ϩ channel (54).
In summary, we have provided the first structural information on the outer pore region of the HERG K ϩ channel, a region that is critical for the rapid inactivation of the channel (19,24). This unique feature of the HERG K ϩ channel is essential for its function in normal cardiac repolarization. We have also shown that the S5P peptide is able to interact with the rest of the HERG channel, resulting in suppression of current, an altered selectivity for K ϩ over Na ϩ , and partial disruption of inactivation. Furthermore, we have shown that it is the amphipathic ␣-helix region that is critical for the effects on ion selectivity and inactivation, because when the amphipathic ␣-helix region of the S5P peptide is replaced with a random coil linker, the peptide no longer affects ion selectivity or inactivation ( Fig.  10B and Table I), whereas a peptide containing only the helix region has very similar effects on the HERG channels as the full-length peptide (compare Figs. 9 and 11). Interestingly, the S5P peptide and the helix peptide are able to affect HERG channels despite being applied in normal Tyrode solution where in isolation they would be expected to have a random coil conformation. Presumably, however, the peptides are able to form helices in the context of the HERG K ϩ channels and the associated membrane environment.