Mapping the Binding Site of a Humanether-a-go-go-related Gene-specific Peptide Toxin (ErgTx) to the Channel's Outer Vestibule*

The goals of this study are to investigate the mechanism and site of action whereby a humanether-a-go-go-related gene (HERG)-specific scorpion peptide toxin, ErgTx, suppresses HERG current. We apply cysteine-scanning mutagenesis to the S5-P and P-S6 linkers of HERG and examine the resulting changes in ErgTx potency. Data are compared with the characteristics of charybdotoxin (ChTx, or its analogs) binding to the Shaker channel. ErgTx binds to the outer vestibule of HERG but may not physically occlude the pore. In contrast to ChTx·Shaker interaction, elevating [K] o (from 2 to 98 mm) does not affect ErgTx potency, and through-solution electrostatic forces only play a minor role in influencing ErgTx·HERG interaction. Cysteine mutations of three positions in S5-P linker (Trp-585, Gly-590, and Ile-593) and 1 position in P-S6 linker (Pro-632) induce profound changes in ErgTx binding (ΔΔG > 2 kcal/mol). We propose that the long S5-P linker of the HERG channel forms an amphipathic α-helix that, together with the P-S6 linker, forms a hydrophobic ErgTx binding site. This study paves the way for future mutant cycle analysis of interacting residues in the ErgTx·HERG complex, which, in conjunction with NMR determination of the ErgTx solution structure, will yield information about the topology of HERG's outer vestibule.

The rapid delayed rectifier (I Kr ) 1 is an important repolarizing current in many regions of the heart (1). The major subunit that forms the I Kr channel in human heart is HERG (2). Inherited mutations in HERG have been identified and linked to congenital long QT syndrome (LQT2) (2,3). Furthermore, drugs that suppress I Kr have been linked to "acquired LQT" (3).
These all point to the critical role played by I Kr in maintaining cardiac electrical stability. Therefore, information about the structure of the I Kr /HERG channel is important. Such information will help in rational drug design for agents that are useful for combating some forms of LQT syndrome (I Kr /HERG agonists), or for agents with zero or low risk for inducing acquired LQT syndrome (no effects on I Kr /HERG).
One of the most critical factors in determining I Kr /HERG function is its C-type inactivation process. This is the basis for the "inward rectification" property of I Kr /HERG, which is important for shaping the action potentials in the heart. Inward rectification dictates that there is little or no outward I Kr at positive plateau voltages (important for maintaining the plateau phase) but large outward I Kr during phase 3 (ensuring efficient repolarization back to the resting membrane potential). Furthermore, the C-type inactivation process in the I Kr / HERG channel appears to be intimately related to channel sensitivity to many different drugs (4 -6). C-type inactivation results from conformational changes in the outer mouth region of the channel, which prevent current flow through the pore (7). Therefore, an important target for structural analysis is the outer mouth region of the I Kr /HERG channel.
There is a general consensus that the outer vestibules of various potassium channels share a common architecture (8 -10). Indeed, homology modeling of outer vestibules of mammalian voltage-gated K ϩ (Kv) channels based on the crystal structure of KcsA (9), a bacterial proton-gated potassium channel, has been successful in several cases (10). However, such a strategy may not be applicable to the HERG channel. First, the C-type inactivation in the I Kr /HERG channel is unique among all potassium channels that have this gating process: C-type inactivation in I Kr /HERG is extremely fast in onset and in reversal and is strongly voltage-dependent (11). These unique kinetic features suggest that conformational changes in the outer mouth region of HERG during membrane depolarization may be different from those of the other channels. Second, HERG has an unusually long S5-P linker (43 amino acids), much longer than the 14-to 18-amino acid S5-P linker seen in most other Kv channels, as well as the so-called "turret" of KcsA (see Fig. 1A). We have shown that mutations in the middle of HERG's S5-P linker, far away from the P (pore) loop in the one-dimensional sequence, can have profound effects on the outer mouth properties of the channel (12). There have been no reports on similar effects of mutations in the middle of S5-P linker for other Kv channels. Thus, the S5-P linker of HERG may play a unique structural and functional role. Understanding such a role may provide answers to questions such as why the C-type inactivation in the HERG channel has such unique kinetic properties and how C-type inactivation is re-lated to the channel's sensitivity to various drugs.
Peptide toxins have been very useful tools in structural analysis of potassium channels, well before the data of KscA crystal structure were available (13,14). Many short peptide scorpion toxins (␣-KTx) are available that can block various potassium channels with differences in specificity and potency (13)(14)(15). Many ␣-KTx toxins have six cysteines, forming three disulfide bridges (15). They have a well-conserved positive residue (Lys-27 in ChTx) that serves to plug the channel pore (Fig. 1B) (16). ErgTx is purified from Centruroides noxius Hoffmann (GenBank TM accession number CnErg1). It has 42 amino acids and 4 disulfide bonds (17). Its amino acid sequence and disulfide bond pattern are not homologous to those of representative ␣-KTx toxins shown in Fig. 1B. Previous work has shown that ErgTx is a potent blocker of HERG expressed in mammalian cells and native I Kr in guinea pig ventricular myocytes (IC 50 in the low nanomolar range) (18). It does not block I Ks , I K, ATP , I RK1 , or EAG at 1 M (18). Recently, we showed that ErgTx sensitivity may be determined by the S5-P linker of the HERG channel (19).
Due to the four disulfide bonds, ErgTx should have a compact and rigid structure, amenable to NMR analysis of its solution FIG. 1. A, alignment of amino acid sequences in the outer vestibule region of representative potassium channels with that of HERG. The end of S5, P-loop and the beginning of S6 are underlined; S5-P linker (corresponding to the "turret" of KcsA) and P-S6 linker are marked. Shaded residues signify the "GXG" motif of potassium channel signature sequence. Note that, compared with the lengths of S5-P linker in these representative potassium channels (14 or 18 residues), the S5-P linker in HERG is much longer (43 residues). A putative amphipathic ␣-helix (positions 583-595) is shown as an insert. Positions in HERG examined in this study are those from 571 to 613 and from 631 to 638 (numbers marked at top). B, alignment of amino acid sequences of representative ␣-KTx with that of ErgTx. Abbreviations are: charybdotoxin (ChTx or ␣-KTx1.1), noxiustoxin (Ntx or ␣-KTx2.1), and agiotoxin 2 (AgTx2 or ␣-KTx3.2). Cysteines are boxed and connected to illustrate disulfide bridges. The well-conserved lysine residues in ␣-KTx (Lys-27 in ChTx) crucial for the pore blockade activity are highlighted by the shaded area. The histidine (His-29, H) in ErgTx is also highlighted. In both panels, alignment was performed using the ClustalV method. Gaps (.) are inserted to enhance alignment.
FIG. 2. Effects of mutations in the S5-P and P-S6 linkers on outer mouth properties of HERG and on TEA blockade. Top: Partial HERG amino acid sequence. Positions examined here (587 and 631) are marked. A, replacing His-587 with lysine (H587K) destroys C-type inactivation and K ϩ selectivity. Panels a and b depict current traces of WT and H587K HERG channels elicited by the voltage clamp protocol shown at the top. Panel c shows the complete tail I-V relationships for WT and H587K (n ϭ 15 and 6). Panel d shows that removing extracellular Na ϩ ions shifts the reversal potential (E rev ) of H587K from Ϫ15 to Ϫ60 mV. Inset: H587K current traces before and after Na o ϩ removal (V r Ϫ60 mV). B, replacing Ser-631 with tyrosine (S631Y), glutamate (S631E), or lysine (S631K) reduces HERG sensitivity to extracellular TEA. Panels a-d show current traces elicited by the voltage clamp protocol depicted at the top before (thin traces) and after (thick traces) adding TEA (50 mM) to the bath solution (with equimolar [Na] o reduction). The channel types are marked. Panel e summarizes the fractions of current not suppressed by TEA o (I TEA /I C , ratio of peak tail current amplitudes in the presence and absence of TEA). The numbers of measurements are shown in parentheses. Experiments were conducted in 2 mM [K] o . structure. Our ultimate goal is to map the ErgTx binding site on HERG and to identify amino acid pairs that interact with each other across the toxin-channel interface (20,21). In this way, we hope to obtain a "footprint" of the interaction surface of ErgTx on HERG and to derive the three-dimensional arrangement of relevant residues on the outer vestibule of this channel. This study is a first step toward that goal. We applied cysteinescanning mutagenesis, replacing all residues in the outer vestibule region of HERG by cysteine one at a time, and studied how these mutations affect ErgTx binding to the channel. We chose cysteine, instead of alanine (22) or lysine (23), for two reasons. First, the cysteine side chain is small, hydrophobic, and usually well tolerated. Thus, it increases our chance of studying more mutant channels. Second, cysteine side chains can be specifically modified by methanethiosulfonate reagents. This will allow us to use different strategies to estimate the distances between channel residues of interests and the toxin binding site (21,24). Based on these data, we propose that the central region of HERG's S5-P linker forms an amphipathic ␣-helix that interacts with the pore of the channel to form the ErgTx binding site.
Briefly, crude venom was dissolved in water and centrifuged at 10,000 ϫ g for 15 min. The supernatant was lyophilized and kept at Ϫ20°C until final purification. After Sephadex G-50 gel filtration, fraction II was directly applied to a semi-preparative C18 reverse-phase column (Vydac, Hesperia, CA) and eluted with a linear gradient from 5% of solvent A (0.12% trifluoroacetic acid in water) to 60% solvent B (0.10% trifluoroacetic acid in acetonitrile) over 90 min. The component eluted at ϳ30 min was further chromatographed by high performance liquid chromatography using an analytical C18 reverse-phase column. This gave a major pure component (ErgTx), whose primary structure was obtained by direct amino acid sequencing and by mass spectrometry ( Fig. 1B) (17).
Molecular Biology-HERG in a vector pGH19 was a kind gift of Dr. Gail Robertson (University of Wisconsin-Madison). We subcloned the HERG cDNA sequence (GenBank TM accession number U04270) into the KpnI/XbaI site of a vector, pAlterMax, which was required for the oligonucleotide-directed mutagenesis procedure with a commercial kit (Altered Site II in vitro mutagenesis system, Promega). Cysteine substitution mutations were confirmed by direct DNA sequencing around the mutation sites. In most cases, two separate colonies from each mutant were used for cRNA transcription and oocyte expression. No differences were seen in the phenotype of channels translated from the two cRNAs. For transcription, wild-type (WT) and mutant HERG sequences in the pAlterMax vector were linearized with NotI and transcribed using the T7 RNA polymerase and a commercial kit (mMES-SAGE mMACHINE, Ambion, Austin, TX). All cRNA samples were quantified by denaturing RNA gels using densitometry (ChemiImager model 4400, ␣-Innotech Corp.).
Oocyte Preparation and Injection-Oocyte isolation and cRNA injection were as described previously (25). Briefly, stage V oocytes were isolated from follicular cell layer after mild collagenase digestion and injected with cRNA solution using a Drummond digital microdispenser. The injection volume was ϳ40 nl/oocyte, equivalent to cRNA of 12-18 ng/oocyte.
Electrophysiological Experiments-Three to five days after cRNA injection, channel function was studied using the two-microelectrode voltage clamp method as previously described (26). During recordings, oocytes were superfused with a low-Cl ND96 solution at room temperature. For experiments shown in Fig. 5, oocytes were not superfused but were placed in the bath solution of a fixed volume (1 ml). ErgTx was dissolved in sterile bovine serum albumin (0.1 mg/ml) solution at 2 M and frozen in small aliquots. An aliquot was thawed and used for experiments in Ͻ2 days without refreezing. After control data were obtained, 5 l of the ErgTx stock solution was added to the bath solution (1 ml) to reach a final concentration of 10 nM. Repetitive pipetting was needed to ensure complete equilibration of ErgTx in the bath solution. Toxin effects were evaluated when steady state was reached (4 -10 min). Voltage clamp protocol generation and data acquisition were controlled by pClamp 5.5 (Axon Instruments). Data analysis was performed with pClamp 6 or 8, Excel (Microsoft) and PeakFit (Jandel Scientific). Specific protocols and methods of data analysis are described in the figure legends. Where appropriate, data are presented as mean Ϯ S.E. Statistical analysis was performed with one-way analysis of variance, followed by Dunn's test (SigmaStat 2.0, SPSS). Fig.  2A (panels a and c) shows the hallmark of wild-type (WT) HERG currents: strong inward rectification due to rapid onset and reversal of C-type inactivation in a voltage-dependent manner. A depolarization pulse to ϩ60 mV elicited little outward current because of C-type inactivation. Subsequent repolarization to ϩ40 to Ϫ60 mV induced outward tail currents with a distinct rising phase (recovery from C-type inactivation). This led to a prominent negative slope in the tail I-V relationship (ϩ60 to Ϫ60 mV, Fig. 2A, panel c). At more negative voltages, tail currents became less outward and reversed at Ϫ100 mV (reversal potential or E rev ), close to the Nernst K ϩ equilibrium potential (E K ϳ Ϫ105 mV in 2 mM [K] o ).

Effects of Mutations in the Extracellular Linkers Suggest a Unique Outer Vestibule Structure in the HERG Channel-
Replacing histidine at position 587 to lysine (H587K, Fig. 2, top), disrupted both the C-type inactivation process and the K ϩ selectivity of the pore. Current traces from H587K were elicited by the same voltage clamp protocol as that used for WT HERG ( Fig. 2A, panel b). The step to ϩ60 mV induced a prominent outward current. Subsequent repolarization steps elicited smaller outward currents that reversed between Ϫ10 and Ϫ20 mV. The I-V relationship of H587K was almost linear in the voltage range between Ϫ140 and 0 mV, with an upward turn at more positive voltages ( Fig. 2A, panel c). Therefore, the C-type inactivation process was disrupted in H587K. Furthermore, removing extracellular Na ϩ ions shifted the E rev of H587K in the negative direction to about Ϫ60 mV ( Fig. 2A, panel d), indicating that external Na ϩ ions contributed significantly to currents through the H587K channel pore. The calculated K ϩ to Na ϩ permeability ratio (P K :P Na ) for H587K was 1.5 Ϯ 0.1, much lower than that of the WT HERG (191 Ϯ 91). These changes in H587K were not due to the added permanent positive charge, because increasing the protonation of H587 in WT HERG (by changing the extracellular pH from 8.5 to 6.5) did not affect the C-type inactivation process or the K:Na selectivity of the pore (27). Furthermore, replacing His-587 with proline creates the same phenotype as H587K (12). These observations indicate that the extracellular S5-P linker of HERG, or at least the middle of this linker including position 587, participates in conformational changes that determine the channel's outer mouth properties. There has been no report on similar effects of mutations made in the S5-P linker of the Shaker channel on the channel's outer mouth properties. Fig. 2B illustrates another example of differences between HERG and Shaker in their outer mouth properties. Position 449 in the Shaker channel is located at the external entrance to the pore, corresponding to position 631 of the HERG channel (Fig. 1A). It is an important determinant of channel sensitivity to external TEA (28). Replacing the threonine residue at this position with an aromatic residue (T449Y and T449F) greatly enhances TEA binding (due to a stabilizing interaction between positively charged TEA and -electrons of the aromatic ring at 449), while replacing Thr-449 with a positively charged residue (T449K) has the opposite effect (due to electrostatic repulsion) (28). We tested the effects of mutating the equivalent residue in the HERG channel (Ser-631) on TEA potency (29). TEA blocked the outer mouth of the WT HERG channel with an IC 50 of ϳ50 mM. This is reflected by the decrease of outward tail current at Ϫ80 mV in Fig. 2B (panel a). The appearance of a prominent outward peak and the higher level of outward current at ϩ20 mV in the presence of TEA was due to an interference of C-type inactivation by TEA bound to the outer mouth of the channel (7). Surprisingly, replacing Ser-631 with an aromatic residue (S631Y) reduced the sensitivity to external TEA (Fig. 2B, panel  b). Furthermore, replacing Ser-631 with a negatively charged residue (S631E) or a positively charged residue (S631K) reduced the sensitivity to TEA. There is no difference in TEA sensitivity between these two mutants. These data suggest that the outer mouth configuration in the HERG channel differs significantly from that of the Shaker channel so that the side chain at position 631 is shielded from bound TEA.
ErgTx Selectivity- Fig. 3 shows that ErgTx potently suppressed the HERG current amplitude. The suppressing effect could be detected at 1 nM ErgTx. However, ErgTx did not affect Kv1.4, Kv4.3, Kv2.1, or KvLQT1 even at 50 nM. This apparent selectivity, by itself, is not conclusive evidence for a unique outer vestibule structure in the HERG channel among the potassium channels examined here. Therefore, we further investigated the mechanism by which ErgTx suppresses HERG current, and positions in HERG that are important for ErgTx binding. These features are compared with those of ChTx (or its analogs) blockade of the Shaker (or Shaker-like) channel. In the latter cases, the mechanism and site of action of the toxins have been well characterized (16,(22)(23)(24). Such a comparison will help us deduce how and where ErgTx binds to the HERG channel and reduces the current.
Concentration Dependence of ErgTx Suppression of HERG- Fig. 4A illustrates a representative time course of changes in HERG current amplitude when the oocyte was exposed to increasing concentrations of ErgTx (1-100 nM) and after wash out of the toxin. Currents were elicited by repetitive depolarization pulses from V h Ϫ80 to ϩ20 mV for 1 s applied once every minute. The peak tail current amplitudes before ErgTx application (I c ) and that at the steady state of ErgTx effect (I tx ) were used to measure the fraction of unblocked channels (I tx / I c ). ErgTx suppressed HERG current in a concentration-dependent manner. The effect was totally reversible. The data points can be well fit with Equation 1, where A max is the fraction of current sensitive to ErgTx (93 Ϯ 3%), and K d is the dissociation constant (6.45 Ϯ 1.03 nM).
It is important to note that there is a residual current (on average ϳ10% of control) not suppressed by high concentrations of ErgTx. This is clearly shown in Fig. 4A: increasing [ErgTx] from 50 to 100 nM induced little further suppression of the current. In the presence of 100 nM ErgTx, the current's waveform resembled that of the control current (showing Ctype inactivation and a high K ϩ selectivity, trace d of Fig. 4A). Therefore, this cannot be due to an ErgTx-insensitive background or "leak" conductance. Instead, it suggests that ErgTx did not totally occlude the HERG pore. This is different from ChTx blockade of the Shaker channel: Lys-27 of ChTx binds and plugs the channel pore completely (22,23). The difference between ErgTx suppression of HERG and ChTx blockade of Shaker is consistent with the lack of a "Lys-27-equivalent" in the ErgTx sequence (Fig. 1B).

Lack of [K] o Sensitivity in ErgTx⅐HERG Interaction-Fig.
4B also shows that ErgTx potency was not affected by elevating [K] o from 2 to 98 mM. This is distinctly different from the situation of ChTx blockade of the Shaker channel. In this case, Lys-27 of ChTx is critical for K o sensitivity (16,22). It is suggested that Lys-27 of ChTx plugs the potassium channel pore by binding to a site close to K ϩ binding site within the pore (16,22). Elevating [K] o increases K ϩ occupancy inside the pore, and the resulting electrostatic repulsion between K ϩ ions and Lys-27 of ChTx destabilizes toxin binding. The lack of K o sensitivity in ErgTx suppression of HERG is again consistent with the lack of a Lys-27-equivalent residue in ErgTx and with the suggestion that ErgTx does not physically plug the HERG pore.
Cysteine Scanning Mutagenesis Detects Positions Critical for ErgTx Binding to HERG-Studies of toxin binding to potassium channels have implicated the S5-P linker (turret of KcsA, Fig. 1A) and P-S6 linker as important components of toxin binding site (23,24). Therefore, we performed cysteine-scanning mutagenesis by replacing all residues in the S5-P linker (positions 571-613, 43 residues) and the P-S6 linker (positions 631 to 638, 8 residues) of HERG with cysteine one at a time and studied the resulting effects on current suppression by ErgTx.

FIG. 5. Effects of cysteine substitutions in the S5-P and P-S6 linkers of HERG on the potency of ErgTx blockade.
In all cases (except WT) the oocytes were treated with dithiothreitol (5 mM, 0.25-4 h) before recordings. Recordings were conducted in 98 mM [K] o . Channels were activated by 1-s depolarization pulses to ϩ20 mV applied once every min, and peak amplitudes of tail currents at Ϫ80 mV were used to monitor effects of ErgTx (10 nM). Before toxin application, currents were recorded under the control conditions for 4 -9 min to establish a baseline (I c ). Toxin was then added and mixed with the bath solution by repetitive pipetting. The toxin effects reached a steady state in 6 -10 min, when I tx was measured. The fraction of remaining current in the presence of 10 nM ErgTx (I tx /I c ) was used to estimate the dissociation constant (K d ) based on modified Equation 1 (assuming that the maximal toxin effect is 90% reduction of the current, see . Plotted are ⌬⌬G values (means and S.E. bars, n ϭ 3-7 each) against channel types along the abscissa. The asterisks denote mutants whose expression level was too low for measuring ErgTx effects (N573C, K595C, P605C, N633C, E637C, and K638C). White bars represent data not different from WT. Gray and black bars represent data different from WT at p Ͻ 0.05 and p Ͻ 0.01, respectively. W585C, G590C, and P632C were not suppressed even in the presence of 100 nM ErgTx (estimated ⌬⌬G Ͼ 3 kcal/mol). Shown at the top is alignment of partial amino acid sequences of Shaker and HERG. The putative amphipathic ␣-helix in the S5-P linker of HERG-(583-595) is shown as an insert and boxed. The corresponding sequence along the abscissa is also boxed. Shaded residues in the Shaker sequence are those important for binding of ChTx or analog. Shaded residues in the HERG sequence are those important for ErgTx binding (⌬⌬G Ͼ 2 kcal/mol). K638C). Of the remaining 45 mutants, the ErgTx potency was evaluated in 98 mM [K] o using the same voltage clamp protocol as shown in Fig. 4A. One ErgTx concentration (10 nM) was used in all measurements, and the K d values were calculated using the following modified Equation 2, The assumption is that the maximal effects in all cases were a 90% suppression (Fig. 4). This is valid because WT HERG has an IC 50 of 7.2 Ϯ 1.1 nM when estimated using Equation 2 based on data obtained with 10 nM ErgTx, very close to the IC 50 determined from the complete concentration-response relationship (6.5 Ϯ 1.0 nM, Fig. 4). The data are summarized in Fig. 5. Most mutants (30 out of 45) showed little or no change in ErgTx binding (changes in binding free energy, ⌬⌬G, less than 0.5 kcal/mol). Of the remaining 15 mutants, three in the S5-P linker (W585C, G590C, and I593C), and one in the P-S6 linker (P632C) caused outstanding changes in the binding free energy (Ͼ2 kcal/mol). For the remaining 11 mutants, the changes in binding free energy were modest although statistically significant (0.5-1 kcal/mol). Fig. 5, top, shows a sequence alignment between Shaker and HERG in the outer vestibule region. We compare the positions known to be important for ChTx (or its analog) binding to Shaker or Shaker-like channels (20, 22-24, 30, 31), with those important for ErgTx binding to HERG. There are two differences. First, in the Shaker channel, the positions important for ChTx binding cluster close to the P-loop. In the HERG channel, the positions in the S5-P linker important for ErgTx binding are farther away from the P-loop. Second, charge neutralization in the Shaker channel can have profound effects on ChTx binding (32). However, this is not the case for ErgTx binding to HERG. Therefore, neutralizing positively charge residues in the S5-P linker (R582C, K608C, and K610C) did not enhance ErgTx binding, and neutralizing negatively charged residues here (E575C, D580C, D591C, and D609C) had little or only modest effects. Residues that are critical for ErgTx⅐HERG interaction are all uncharged. Another important point is that the three positions in the S5-P linker (Trp-585, Gly-590, and Ile-593) all fall within the putative amphipathic ␣-helix (boxed in both the HERG sequence and the abscissa of Fig. 5). The implications of these findings will be addressed under "Discussion." Do Cysteine Mutations Disrupt ErgTx Binding by Inducing Global Changes in the HERG Outer Vestibule?-Some of the cysteine mutations caused a disruption of the C-type inactivation process and the K ϩ selectivity of the pore, similar to the phenotype of H587K shown in Fig. 2A. One such example is shown in Fig. 6B. G590C did not C-type inactivate or select K ϩ over Na ϩ (E rev Ϫ15 mV) (left of Fig. 6B). It could not be suppressed by 10 (or even 100) nM ErgTx (Fig. 6B, right). This is in sharp contrast to the behavior of WT HERG, which showed a strong C-type inactivation and K:Na selectivity (Fig.  6A, left) and was strongly suppressed by ErgTx (10 nM suppressed current by Ͼ50%, Fig. 6A, right). These observations call into question whether the changes in ErgTx binding resulted from global changes in the outer vestibule conformation. Fig. 6 (C and D) suggests that this is not the case. Although D591C could not C-type inactivate or select K ϩ over Na ϩ ions (E rev Ϫ20 mV), it retained a high ErgTx sensitivity (Fig. 6C). On the other hand, Q592C retained C-type inactivation and a high K:Na selectivity (E rev Ϫ100 mV) but had a significantly lowered ErgTx sensitivity than WT (Fig. 6D).
ErgTx Binds to the Outer Mouth of HERG and Electrostatic Forces Are Involved in Toxin Binding-We suggested above that ErgTx does not plug the HERG channel pore, as is the case for ChTx blockade of the Shaker channel. This, in conjunction with the pattern of positions involved in ErgTx binding to HERG, raises the following question: Does ErgTx suppress HERG current by binding to the outer mouth region and hindering current flow through the pore or by binding away from the pore and modifying other channel function, such as gating? In Shaker or Shaker-like channels, the external TEA binding site overlaps with that of ␣-KTx (16). Therefore, TEA binding and ␣-KTx binding are mutually exclusive. Fig. 7A shows that application of 50 mM TEA (ϳIC 50 , Fig. 2B) reduced the degree of WT HERG current suppression by ErgTx markedly (p Ͻ 0.001). This observation supports an outer mouth binding site for ErgTx.
Do electrostatic forces play any role in ErgTx binding to the HERG channel? ErgTx has a pI of 7.88 and carries Ͻ2 positive charges at pH 7. Fig. 5 further shows that positions in the HERG channel important for ErgTx binding are all uncharged. Charge neutralization in the neighboring region had modest or no effect on ErgTx binding. To explore whether electrostatic forces play any role in ErgTx⅐HERG interaction, we altered the net charge on ErgTx by changing the extracellular solution pH from 7.5 to 8.5 or to 6.5 and examined the resulting changes in toxin potency. Fig. 7B shows that, for WT HERG, changing pH o from 7.5 to 8.5 reduced ErgTx potency. This suggests that net negative charges on ErgTx hindered ErgTx⅐HERG binding. Shifting pH o from 7.5 to 6.5 did not affect ErgTx binding to WT HERG. However, removing either of the two histidines in the S5-P linker (H578C or H587C) made ErgTx more potent at pH o 6.5. This indicates that protonation of these histidine residues in the WT HERG channel negated the increased binding affinity of positively charged ErgTx at pH o 6.5 (probably by increasing the protonation of H29 of ErgTx, Fig. 1B). On the other hand, putting a permanent positive charge at position 578 (H578K) reduced ErgTx potency at pH o 6.5 and 7.5 (when ErgTx should be positively charged) but not at pH o 8.5 (when ErgTx should be negatively charged). This is consistent with an electrostatic repulsion between positive charge at 578 and positive charge on ErgTx at the lower pH o range that hindered toxin binding. Therefore, ErgTx⅐HERG interaction can be influenced by electrostatic forces. However, the degree of influence is much less for ErgTx⅐HERG binding than for ChTx⅐Shaker binding (32).
Voltage Sensitivity of ErgTx Binding to the HERG Channel-Binding of ChTx to the Shaker channel is voltage-sensitive: Membrane depolarization destabilizes ChTx binding (16). This voltage sensitivity is exclusively mediated by Lys-27 of ChTx (16). Although ErgTx does not have a Lys-27-equivalent, previous work has shown that depolarization also reduces ErgTx binding to the HERG channel (18). It was suggested that this is due to the C-type inactivation process of the HERG channel at depolarized voltages, which reduces ErgTx binding. We have confirmed this finding (Fig. 8) and further explored the mechanism. Fig. 8B summarizes the data and compares it to the voltage dependence of HERG activation. The steep slope of I tx /I c data points in the voltage range of Ϫ40 to 0 mV coincides with the voltage range of a steep increase in channel activation. This suggests that HERG channel activation hinders or destabilizes ErgTx binding. However, stronger depolarization at the plateau of the activation curve (Ͼ0 mV) further reduced ErgTx binding, although the slope was less steep than in the negative voltage range. This decrease in ErgTx binding was not related to C-type inactivation, because in a mutant that did not C-type inactivate but maintained a high ErgTx sensitivity (G572C) membrane depolarization still reduced toxin potency in this voltage range (data not shown). The decrease in ErgTx binding could not be due to an increase in K ϩ ion efflux, because strong depolarization reduced outward current due to C-type inactivation (Fig. 8A). Finally, this could not be due to effects of voltage on the binding of a charged ErgTx molecule within the transmembrane electrical field. As shown in Fig. 8B, the volt-  (H587C, H578K). I tx /I c in ErgTx (10 nM) was measured as described in Fig. 4, and plotted against pH o (n ϭ 3-6 each). In most cases, each oocyte was tested for two or all three pH o conditions in a random order. Shown at the top is partial HERG amino acid sequence in the outer vestibule region, with shaded areas highlighting positions important for ErgTx binding (see Fig. 5). His-578 and His-587 are marked.
age effect was more prominent at pH o 8.5 than at pH o 6.5 ("effective valence" of voltage-sensing domain 0.36 and 0.2 at pH o 8.5 and 6.5, respectively). This is opposite to the expected effect (e.g. ErgTx should be negatively charged at pH o 8.5 and thus membrane depolarization should enhance ErgTx binding). DISCUSSION Our data suggest that ErgTx binds to the HERG channel with a 1:1 stoichiometry, leading to a suppression of current through the channel pore. Cysteine-scanning mutagenesis experiments suggest that three positions in the S5-P linker (Trp-585, Gly-590, and Ile-593) and one position in the P-S6 linker (Pro-632) are critical determinants of ErgTx binding. Below we compare the behavior of ErgTx binding to the HERG channel with that of ChTx (or an analog) binding to the Shaker or Shaker-like channels. Based on the comparison and what is known about the mechanism and site of action of ChTx, we further deduce the mechanism by which ErgTx suppresses HERG currents and the possible location of ErgTx receptor site on the HERG channel.
There are some apparent similarities in the behavior of binding of ErgTx and ChTx to their target channels: 1) TEA, an outer mouth blocker, can antagonize toxin binding in both cases. This supports the notion that ErgTx binds to the outer mouth of HERG, and the binding site overlaps with that of TEA. 2) Electrostatic forces are involved in toxin binding in both cases, although to different degrees. In the case of ChTx binding to the Shaker channel, electrostatic forces play a major role in toxin binding and pore blockade. ChTx has a pI value of 9.03 and carries about six positive charges at pH 7. Negative charges around the receptor site help orient the toxin in binding to the receptor (32). Furthermore, the lysine at position 27 of ChTx (Lys-27) binds within the pore and occludes current through the channel (16,22). The pH o experiments shown in Fig. 7B suggest that charge-charge interactions also matter in ErgTx binding to the HERG channel. However, such chargecharge interactions are not a major factor in ErgTx binding to HERG. This is not surprising, because ErgTx has a pI value of 7.88 and carries less than two positive charges at pH 7. Furthermore, ErgTx does not have a positive charge equivalent to Lys-27 in ChTx. 3) Membrane depolarization reduces toxin binding in both cases but likely by different mechanisms. Membrane depolarization destabilizes ChTx binding by two mechanisms, both of which are mediated by Lys-27. First, depolarization enhances K ϩ ion occupancy inside the pore by promoting K i ϩ ion efflux. This can dislodge ChTx bound to the pore through electrostatic repulsion between K ϩ ions and Lys-27. Second, Lys-27 of bound ChTx senses ϳ20% of the transmembrane electrical field. Therefore, membrane depolarization can have a direct effect on Lys-27 and thus ChTx binding. ErgTx does not have a Lys-27-equivalent. Furthermore, our data in Fig. 8 show that the destabilization effect of membrane depolarization was more pronounced at pH o 8.5 (ErgTx negatively charged) than at pH o 6.5 (ErgTx positively charged). This observation rules out the possibility that membrane depolarization destabilizes ErgTx binding by a direct effect on a charged ErgTx molecule bound within the transmembrane electrical field. We propose that conformational changes in the S5-P linker during strong membrane depolarization hinder or destabilize ErgTx binding.
There are distinct differences in the behavior of toxin/channel interactions: 1) Elevating [K] o destabilizes ChTx binding to Shaker (16,22), but does no affect ErgTx binding to HERG. The effect of changing [K] o on ChTx⅐Shaker interaction is exclusively mediated by Lys-27 (16,22). The insensitivity of ErgTx⅐HERG interaction to elevating [K] o is consistent with the notion that ErgTx does not have an Lys-27-equivalent. 2) The pattern of positions important for toxin-channel interaction differs between the two (Fig. 5, top). Positions important for ChTx binding to the Shaker channel are those flanking the pore loop, with positions farther away from the pore loop having decreasing importance in influencing toxin binding (22,23). For ErgTx⅐HERG, the positions in the S5-P linker important for toxin binding are far away from the pore loop in onedimensional sequence. Charge mutations in Shaker have marked effects on ChTx binding (31, 32) but little or no effects on ErgTx binding to the HERG channel.
What can we conclude about the mechanism by which ErgTx suppresses the HERG current? We can conclude that ErgTx binds to the outer vestibule of HERG, but it probably does not plug the pore with a positive charge as is the case for Lys-27 in ChTx. The maximal effect of ErgTx is ϳ90% suppression of the HERG current, not 100%. Again, this is consistent with the notion that ErgTx is not a "molecular plug" of the HERG pore. This situation is similar to ␦-dendrotoxin suppression of ShaKv1.1: ␦-Dendrotoxin does not physically plug the pore but binds in an "off-center" position in the outer vestibule. This leads to a reduction, but not a total occlusion, of current through the pore (33).
What can we learn about the ErgTx receptor site on the HERG channel? The cysteine-scanning mutagenesis data suggest that S5-P and P-S6 linkers are both involved. Although these two domains are not contiguous in one-dimensional sequence, in three-dimensional space, the long (43 amino acids) S5-P linker of HERG may come close to the P-S6 linker and thus to the channel pore. Our working hypothesis is illustrated by the schematic in Fig. 9A. An analysis of possible secondary structures in the S5-P linker using the program, Protean, in LaserGene (34) suggests that positions 583-594 may form an amphipathic ␣-helix. The helical wheel plot in Fig. 9B shows that hydrophobic residues is this region cluster to one face of the ␣-helix, among which Trp-585, Gly-590, and Ile-593 may form contact points with bound ErgTx. The other face of the ␣-helix has mainly hydrophilic residues. This face of the ␣-helix is not involved in ErgTx binding, because neutralizing the negative charge here, D591C, has no effects on toxin binding (Figs. 5 and 6).
In summary, we show that the long S5-P linker of the HERG channel contributes importantly to the outer mouth properties of this channel, supporting the conclusion from a previous report (19). We propose that this linker can engage in intimate interactions with the pore's entryway and participates in conformational changes important for the C-type inactivation FIG. 9. A proposed outer vestibule structure for the HERG channel. A, two-dimensional schematic of an HERG subunit, showing the putative ␣-helix formed by residues 583-594 in the S5-P linker that comes in close contact with the P-S6 linker. B, helical wheel plot of residues 583-594, viewed from the N-terminal end. Five residue numbers are labeled for reference. Note that positions important for ErgTx binding cluster to the hydrophobic face of the amphipathic ␣-helix. process and for K:Na selectivity of the pore. Future work will be focused on identifying toxin and channel residues interacting across the toxin-channel interface. This, in conjunction with ErgTx's solution structure obtained by the NMR technique, will yield a three-dimensional structure of the outer vestibule of the HERG channel.