Structural and Functional Role of the Extracellular S5-P Linker in the HERG Potassium Channel

Human ether-a-go-go–related gene (HERG)^* encodes the pore-forming subunit of the rapid delayed rectifier potassium (I_Kr) channel (Sanguinetti et al., 1995), an important determinant of action potential (AP) configuration and duration in human heart (Tseng, 2001). Suppression of I_Kr function due to inherited mutations in HERG, or more importantly due to adverse drug effects, has been linked to congenital or acquired long QT syndrome (Splawski et al., 2000), indicating an important role of HERG/I_Kr in maintaining the electrical stability of the heart. Under physiological conditions, the HERG/I_Kr channel passes little outward current at depolarized voltages (e.g., during phase 2 of cardiac AP) (Hancox et al., 1998; Zhou et al., 1998). This helps reduce the amount of inward current needed to maintain the plateau phase. Upon membrane repolarization (phase 3 of cardiac AP), there is a resurgence of outward current through the HERG/I_Kr channel that ensures proper return of voltage to the resting membrane potential (Hancox et al., 1998; Zhou et al., 1998). Such a “rectification” property of HERG/I_Kr is due to a fast inactivation process that can be turned on (by membrane depolarization) or off (by repolarization) in a matter of milliseconds (Spector et al., 1996).

The fast inactivation process of HERG/I_Kr shares important features with those of “C-type inactivation” in the Shaker channel (Hoshi et al., 1991; Yellen et al., 1994; Smith et al., 1996). These include the sensitivity to mutations introduced to the outer mouth region of the channel, slowing of the inactivation process by an outer mouth blocker (e.g., TEA), or by an increase in pore occupancy by permeant ions (e.g., K⁺) (Hoshi et al., 1991; Yellen et al., 1994; Smith et al., 1996). Shaker's C-type inactivation has been well studied. It is due to conformational changes around the outer mouth, that constrict the pore or change the pore's selectivity in such a way that K⁺ flux is prevented (Yellen et al., 1994; Liu et al., 1996; Starkus et al., 1997). The C-type inactivation process in the Shaker channel and in other voltage-gated K (Kv) channels for which C-type inactivation has been described (Kv1.3 [Panyi et al., 1995], Kv1.4 [Rasmusson et al., 1995], Kv1.5 [Fedida et al., 1999], and Kv2.1 [Klemic et al., 1998]) share two common features: slow kinetics and a lack of intrinsic voltage dependence. The slow kinetics is at least partially attributed to hydrogen bonds formed around the outer mouth of the pore (Doyle et al., 1998; Larsson and Elinder, 2000; Ortega-Saenz et al., 2000). These hydrogen bonds stabilize the outer mouth in the open state, resisting collapse during C-type inactivation (Larsson and Elinder, 2000; Ortega-Saenz et al., 2000). The trigger for C-type inactivation in Shaker and related Kv channels is believed to be the S4 movements induced by membrane depolarization (Gandhi et al., 2000). There-fore, C-type inactivation in these channels is not intrinsically voltage sensitive, but derives its voltage sensitivity from a coupling to the voltage-dependent activation process. C-type inactivation in the HERG channel is extremely fast, and is intrinsically voltage-sensitive: it can occur at a time scale and in a voltage range with little or no channel activation (Sanguinetti et al., 1995; Rasmusson et al., 1998). What is the structural basis for these unique features of C-type inactivation in the HERG channel?

Fig. 1 shows an alignment of amino acid sequences that line the outer vestibule (S5-P and P-S6 linkers) and the outer portion of the pore (pore or P-loop) of HERG, KcsA (the bacterial proton-gated K channel whose crystal structure has been solved [Doyle et al., 1998]), and major classes of Kv channel. This comparison reveals two unique features of the HERG sequence. First, amino acids that are implicated in forming hydrogen bonds that stabilize the Shaker outer mouth in the open state (E418, W434, and Y445, highlighted by gray shade in Fig. 1) are well conserved among Kv channels as well as in KcsA, but are missing in HERG. This reduced hydrogen bonding capability in HERG may make its outer mouth more flexible, easier to collapse by conformational changes triggered by membrane depolarization. This can contribute to the uniquely fast C-type inactivation kinetics in the HERG channel. Second, the S5-P linker in the HERG channel is much longer than those of KcsA and other Kv channels (43 aa vs. 14–23 aa). This long S5-P linker may play a unique role in the structure and function of the HERG channel.

Previously, we have shown that replacing the histidine at position 587 in the middle of HERG's S5-P linker (circled in the HERG sequence in Fig. 1) to proline or cysteine in disulfide-bonded state has profound effects on three aspects of channel function: (a) disrupting C-type inactivation, (b) reducing K⁺ selectivity, (c) accelerating activation, and causing a hyperpolarizing shift in the voltage dependence of activation (Dun et al., 1999a). On the other hand, substituting H587 with cysteine in reduced state has little or no impact on the above channel function (Dun et al., 1999a). Since position 587 is far from the P-loop in a one-dimensional sequence, how do mutations here affect the outer mouth properties (C-type inactivation and K⁺ selectivity)? How is this position involved in the activation gating process? We hypothesized that the long S5-P linker of the HERG channel does not assume a random coil conformation, as is the case for the much shorter “turret” (the S5-P equivalent) in the KcsA crystal structure (Doyle et al., 1998). Instead, this linker may assume certain secondary structure around position 587 and make contacts with other domains of the channel. Therefore, perturbing the conformation here by replacing H587 with proline or by disulfide formation in H587C can affect several aspects of channel function.

To explore the role of the S5-P linker in HERG channel function, and to test whether there is a secondary structure in this extracellular domain, we apply cysteine-scanning mutagenesis to positions 571–613 of HERG. We also include the P-S6 linker (positions 631–638) because, as suggested by the crystal structure of KcsA and by functional studies on the Shaker channel, it interacts with the S5-P linker during gating processes (Doyle et al., 1998; Gandhi et al., 2000; Larsson and Elinder, 2000; Ortega-Saenz et al., 2000). The wild-type (WT) HERG channel has 12 native cysteines that are functionally “silent”: channel function is not affected by treatment with DTT, H₂O₂, or extracellular MTS reagents (Dun et al., 1999a; Fan et al., 1999). Therefore, cysteine substitutions are made in the WT HERG background. We examine: (a) channel behavior before and after DTT treatment, and (b) effects of extracellular MTSET or MTSES modification of introduced cysteine side chains on channel function. Based on these data, we propose a structural model for the outer vestibule of the HERG channel, in which the middle part of the S5-P linker assumes an α-helical structure and serves as a bridge of communication between the outer mouth and the voltage-sensing domain. Therefore, the S-P linker in HERG plays a central role in determining the voltage-sensitivity of the C-type inactivation process.

HERG in a vector, pGH19 (a gift from Dr. Gail A. Robertson, University of Wisconsin-Madison, Madison, WI), was subcloned into the KpnI/XbaI site of a vector, pAlterMax, and subjected to cysteine scanning mutagenesis using the oligonucleotide-directed method (Altered site II in vitro mutagenesis system; Promega). Mutations were confirmed by direct DNA sequencing around the mutation sites. In most cases, two independent colonies were picked and used for cRNA transcription and oocyte expression. No differences in the channel phenotype were noticed between two colonies from the same mutant. For transcription, plasmids were linearized by NotI and transcribed using T7 RNA polymerase using a commercial kit (Mmessage Mmachine; Ambion). The quality and quantity of cRNA products were evaluated using densitometry (ChemiImager model 4400; α-Innotech Corp.).

Oocytes from Xenopus laevis were isolated as described before (Tseng-Crank et al., 1990), and incubated in an ND96-based medium (in mM: NaCl 96, KCl 2, CaCl₂ 1.8, MgCl₂ 1, HEPES 5, Na-pyruvate 2.5, pH 7.5, supplemented with 4% horse serum and penicillin/streptomycin) at 16°C. Five to 12 h after isolation, each oocyte was injected with 40 nl of cRNA solution (containing 4–18 ng cRNA), using a Drummond digital microdispenser. Oocytes were incubated in the above medium at 16°C, and studied 2–4 d after cRNA injection.

Membrane currents were recorded from whole oocytes using the “2-cushion pipette” voltage clamp method (Schreibmayer et al., 1994). Both the current-passing and voltage-recording pipettes were filled with 3 M KCl and the tips were plugged with 1% agar in 3 M KCl (tip resistance 0.1–0.3 MΩ). During recordings, the oocyte was continuously superfused with a low Cl ND96 solution to reduce interference from endogenous Cl channels (in mM: NaOH 96, KOH 2, CaCl₂ 1.8, MgSO₄ 1, HEPES 5, Na-pyruvate 2.5, titrated to pH 7.5 with methanesulfonic acid). The grounding electrodes (containing Ag/AgCl pellets) are filled with 3 M KCl and connected to the bath solution via salt bridges made of 1% agar in 3 M KCl. In some experiments on cysteine-substituted mutants that were poorly expressed, [K⁺]_o was elevated from 2 to 98 mM ([Na⁺]_o removed to maintain the osmolality) to increase current amplitude and facilitate current measurements. This will be specified in the text or figure legends. Voltage clamp was done at room temperature (24–26°C) with OC-725B or OC-725C amplifier (Warner Instruments). Voltage clamp protocol generation and data acquisition were controlled by pClamp5.5 via a 12-bit D/A and A/D converter (DMA; Axon Instruments, Inc.). Current data were low-pass filtered at 1 kHz (Frequency Devices) and stored on disks for off-line analysis.

Most of the voltage clamp protocols and methods of data analysis are described in the text or figure legends. The following software was used for data analysis: pClamp6 or 8, EXCEL (Microsoft), SigmaPlot, SigmaStat, and PeakFit (SPSS).

To explore α-helical periodicity in mutation-induced perturbations of channel function, we used a Fourier transform method as described (Cornette et al., 1987; Li-Smerin et al., 2000). Assuming that channels are distributed between only two states, closed and open, the free energy needed (ΔG_o) when channels are moved from closed to open states at 0 mV is:

where R is the universal gas constant and T is absolute temperature. V_0.5 is the half-maximum activation voltage and k is the slope factor obtained from fitting the activation curve with a Boltzmann function (Fig. 5 legend, see equation). In cases when a double Boltzmann function is required to fit the activation curve, V_0.5 and k are the parameter values from the Boltzmann component in the more negative voltage range (see Fig. 6). Mutation-induced perturbation in ΔG_o is calculated as:

where ΔG_o^MUT and ΔG_o^WT are ΔG_o values for mutant and WT HERG channels, respectively. A Fourier transform method was used to evaluate the periodicity of ΔΔG_o, according to:

P(ω) is the Fourier transform power spectrum as a function of angular frequency (ω), “n” is the number of residues within the segment under analysis, Vj is |ΔΔG_o| at a given position “j + 1” in the segment, and <V> is the average value of |ΔΔG_o| of the segment. To evaluate the α-helical characteristics of a power spectrum, an α-periodicity index (α-PI) is calculated as:

DTT stock solution (0.5 M in deionized water) was frozen at −20°C in small aliquots. An aliquot was thawed and diluted to 5 mM in bath solution immediately before use. MTSET or MTSES powder (Toronto Research Chemicals) was dissolved in deionized water at 0.1 M shortly before experiments. The stock solution was stored on ice and used within 2 h. After control data were obtained, the MTSET or MTSES stock solution was diluted with bath solution to 1 mM and applied immediately to the oocyte. After treatment with DTT, MTSET, or MTSES, effects on the function of cysteine-substituted channels were evaluated after a 10–15 min washout period to ascertain that the effects resulted from covalent modifications.

Of the 51 cysteine-substituted HERG mutants we made, three (N573C, K595C and K638C) do not produce functional channels and are not included in the experiments reported here. For the remaining 48 mutants, we characterize their channel phenotypes before and after DTT treatment by measuring three parameters of channel function: (a) degree of C-type inactivation, (b) selectivity for K⁺ over Na⁺ ions, and (c) voltage dependence of activation. We also examine the effects of two thiol-modifying methanethiosulfonate (MTS) reagents, MTSET and MTSES, on these cysteine-substituted mutants. Modifications of thiol side chains by both MTS reagents increase the side chain volume, but MTSET adds a positive charge while MTSES adds a negative charge. A comparison of their effects on channel function may provide clues as to the location or environment of the thiol side chains. For example, modification of a side chain inside a narrow pore or crevice by both MTS reagents may cause a similar effect due to steric hindrance of ion permeation or protein conformational changes. On the other hand, modification of a side chain on the channel surface close to the pore entrance or to the voltage sensor by the two MTS reagents likely will lead to different or opposite changes, due to electrostatic effects on the local ion concentrations around the pore or on the local electrical field sensed by the voltage sensor.

In the following sections, we will first use data from two mutants (I583C and W585C), along with data from the WT HERG, to illustrate the voltage clamp protocols and data analysis used to characterize the channel phenotypes. We then summarize data from all 48 cysteine mutants. The data allow us to deduce positions in the S5-P and P-S6 linkers where side chain properties are critical for the HERG channel function. Finally, we will focus on two specific issues: (a) periodicity of mutation-induced perturbations of channel function, and (b) effects of MTS modification of pore-entrance residues. These data reveal better defined features in the outer vestibule structure of the HERG channel.

In Fig. 2, data for each of the three channels (WT HERG, I583C, and W585C) are taken from the same oocyte before and after DTT treatment, a reducing agent whose only action is to break disulfide bonds. The voltage clamp protocol is diagrammed on top of column B: from a holding voltage (V_h) of −80 mV, prepulses to 60 mV for 0.2 s are used to fully activate the channels and, if the C-type inactivation is present, to cause channel inactivation. These prepulses are followed by repolarizing steps to V_r ranging from 30 to −120 mV. During the repolarizing steps, channels rapidly (τ of a few ms) reach new, lower levels of C-type inactivation determined by V_r. At V_r −40 mV or more negative, recovery from C-type inactivation is followed by deactivation. The tail current (I_tail) amplitudes are measured from the plateau level (at V_r positive to −40 mV) or the peak level (at V_r −40 mV or more negative). To pool data from different cells, I_tail amplitudes of each cell are normalized by the maximal outward tail current from the same cell. For WT HERG and WT-like mutants (e.g., I583C after DTT treatment, see below), the maximal outward currents occur at V_r −60 mV. For mutants with altered behavior (e.g., I583C before DTT), the maximal outward tail currents are taken from those at V_r 30 mV.

The amplitudes of these tail currents are determined by two factors: (a) steady-state level of C-type inactivation at V_r, and (b) driving force for currents through the channels (difference between V_r and the reversal potential, E_rev). For WT HERG (Fig. 2 A), there is little outward current at 60 mV or at V_r > 0 mV due to a strong C-type inactivation process. As V_r goes from 0 to −60 mV, I_tail becomes more outward due to a voltage-dependent recovery from C-type inactivation. At more negative V_r, the tail current decreases in amplitude because the decrease in driving force outweighs the decrease in the level of C-type inactivation. Therefore, outward I_tail peaks at −60 mV and the tail I-V relationship has a negative slope in the V_r range of 30 to −60 mV. The tail current reverses at an E_rev of −100 mV. Based on the constant field equation (Fig. 6 legend), the E_rev corresponds to a high permeability ratio of K⁺ to Na⁺ (P_K/P_Na = 481 ± 129, n = 8). DTT treatment does not induce any significant changes in the kinetics or tail I-V relationship of WT HERG, confirming that the native cysteine side chains in the HERG channel do not engage in disulfide formation.

I583C before DTT treatment has a very different tail I-V relationship from that of WT HERG (Fig. 2 B, a). There is a prominent outward current at 60 mV. Repolarizing steps elicit smaller outward tail currents that reverse at about −20 mV. Therefore, there is a positive slope in the tail I-V relationship in the V_r range of 30 and −60 mV (panel c, open circles). This reflects the change in driving force for current through the channels. Therefore, the C-type inactivation process is disrupted in I583C under these conditions. Furthermore, the E_rev (−21.1 ± 6.9 mV) is much more positive than that of WT HERG. Removing extracellular Na⁺ ions shifts the E_rev in the negative direction toward E_K, indicating that Na⁺ ions can permeate through the I583C channels under these conditions. The calculated P_K/P_Na is 2.3 ± 1.0 (n = 6).

Fig. 2 B, b and c (closed circles), shows that DTT treatment converts the I583C channel phenotype from the mutant behavior to a “WT-like” behavior. DTT-treated I583C shows a prominent negative slope in the tail I-V relationship, indicating a restoration of the C-type inactivation process. DTT treatment also causes a negative shift in E_rev (to −93.4 ± 2.6 mV), indicating an increase in P_K/P_Na (to 117.6 ± 32.4). These observations indicate that the introduced cysteine side chains in I583C are disulfide bonded before DTT treatment. Under these conditions, the channel function (C-type inactivation and K⁺ selectivity) is disrupted. However, when the cysteine side chains are reduced to the free thiol state, channel function is well maintained.

W585C manifests a mutant behavior similar to that of I583C before DTT treatment: a positive slope in the tail I-V relationship in the V_r range of 30 to −60 mV and an E_rev of −23.3 ± 5.5 mV (P_K/P_Na = 2.1 ± 0.5, n = 5) (Fig. 2 C, a and c). However, unlike I583C, DTT treatment did not convert the mutant behavior to a WT-like behavior (although the W585C current amplitude is modestly increased by DTT treatment; see also of Fig. 5 C, a and b). Relatively aggressive DTT treatment (5 mM for 4 h) cannot convert the W585C phenotype to the WT-like behavior. There are two possible explanations for this apparent resistance of W585C's mutant phenotype to DTT treatment. First, the introduced cysteine side chains in W585C form very stable disulfide bonds that cannot be reduced under our experimental conditions. Second, the tryptophan side chain at 585 is critical for channel function so that replacing it with cysteine, even in the reduced free thiol state, disrupts the C-type inactivation and K⁺ selectivity.

To distinguish between these two possibilities, we test the effects of MTS reagents (MTSET and MTSES) on W585C before and after DTT treatment. MTS reagents can modify free thiol side chains, but not thiol side chains engaged in stable disulfide bonds. Therefore, if DTT-treated W585C has free thiol groups despite the mutant behavior, extracellular MTS should be able to modify the side chains and alter channel function. Furthermore, if there is disulfide formation in W585C before DTT treatment, the MTS effects should be enhanced by DTT treatment.

Figs. 3 A and 4 A show that MTSET or MTSES treatment has no effects on WT HERG. These confirm that native cysteines in HERG are not accessible, or not reactive, to extracellular MTS reagents. Therefore, we are confident that the effects of MTS treatment on cysteine-substituted mutants are due to modification of the introduced cysteine side chains. Fig. 3 B shows that the effect of MTSET on I583C is enhanced by DTT treatment. This is consistent with the notion that the introduced cysteine side chains in I583C are disulfide bonded before DTT treatment. This causes a disruption of channel function (C-type inactivation and K⁺ selectivity) and a low MTSET sensitivity. Breaking the disulfide bonds by DTT restores the channel function and increases MTSET sensitivity.

Fig. 3 C shows that before DTT treatment MTSET has a very small suppressing effect on W585C. DTT treatment markedly enhances MTSET effect on W585C. These data confirm the second possibility listed above: the introduced cysteine side chains in W585C can form disulfides. DTT breaks the disulfides and enhances MTSET sensitivity, but does not restore the WT-like behavior. Therefore, the tryptophan residue at 585 is critical for HERG channel function. In addition, the increase in W585C current amplitude after DTT treatment indicates that disulfide formation in W585C suppresses current amplitude.

Fig. 4, B and C, illustrate the effects of MTSES on I583C and W585C before and after DTT treatment. Consistent with data shown in Fig. 3, without DTT treatment MTSES has little effects on either channels. After DTT treatment, MTSES markedly suppresses the I583C current amplitude but modestly increases the W585C current amplitude. Therefore, MTSET and MTSES have qualitatively the same effect on I583C, but different effects on W585C. The implication of these observations will be addressed in the context of the proposed structural model for HERG's outer vestibule (see discussion).

Fig. 5 illustrates the voltage clamp protocol and data analysis used to characterize the voltage dependence of activation of WT HERG and the two mutant channels. The protocol is diagrammed in the middle of column A. From V_h −80 mV, 1-s test pulses to V_t from −70 to 60 mV are applied at 10-mV increments. The peak amplitudes of tail currents are measured and normalized by the maximum tail current following V_t to 60 mV. This gives an estimate of the fraction of channels activated at the end of the previous test pulse to V_t. For WT HERG and WT-like mutants (e.g., I583C after DTT), the tail currents are measured at −80 mV at which the currents are outward and large (Fig. 5, A and B, b). For mutants with altered behavior (e.g., I583C before DTT), tail currents at −80 mV are inward and small due to changes in ion permeation properties. Therefore, a more hyperpolarized voltage (−120 mV) is used to increase the amplitude of inward tail currents and facilitate measurement.

WT HERG has the same voltage dependence of activation before and after DTT treatment (Fig. 5 A). The activation curve is well described by a single Boltzmann function, with an average half-maximum activation voltage (V_0.5) of −17.0 ± 1.5 mV (n = 14). On the other hand, without DTT treatment I583C manifests a more negative Boltzmann component (V_0.5 = −40.2 ± 0.9 mV, n = 6, Fig. 5 B, c, open circles) than that of WT HERG. Furthermore, there appears a slowly rising phase of channel activation in the positive voltage range. Therefore, the activation curve requires a second Boltzmann component for a good fit (Fig. 5 legend). DTT treatment of I583C restores the WT-like behavior: the activation curve now follows a single Boltzmann function. The V_0.5 value is similar to that of WT HERG (−13.4 ± 2.4 mV, n = 5, Fig. 5 B, c, closed circles). W585C manifests the mutant behavior before and after DTT treatment: its activation curves under both conditions require a double Boltzmann function for a good fit, and the V_0.5 values of the negative Boltzmann component are more negative than the V_0.5 of WT HERG (−31.5 ± 3.7 mV, n = 5, and −27.4 ± 3.0 mV, n = 7, before and after DTT, respectively).

We apply the above voltage clamp protocols and data analysis to the other 46 cysteine-substituted mutant channels made in the S5-P linker (positions 571–613) and in the P-S6 linker (positions 631–637). Results are summarized in Fig. 6. Fig. 6 A summarizes “rectification factors” (see equation in Fig. 6 legend), which indicate whether the C-type inactivation process is maintained (values < 0 or upward histogram bars) or interrupted (values > 0 or downward histogram bars). Fig. 6 B summarizes the permeability ratios of K⁺ to Na⁺ (P_K/P_Na). Fig. 6 C summarizes the half-maximum activation voltages (V_0.5) of the (major) Boltzmann component used to fit the activation curves (see equation, Fig. 5 legend). Fig. 6, D and E, summarize the ratios of currents after MTSET or MTSES treatment to control (I_MTS/I_C, as described in Figs. 3 and 4).

For WT HERG and mutants with WT-like behavior (prominent C-type inactivation and strong K⁺ selectivity), the channel phenotypes are the same before and after DTT treatment. This indicates that the introduced cysteine side chains are mostly in the reduced state without DTT treatment. Therefore, data from both before and after DTT treatment are combined and shown as white histogram bars in Fig. 6. For mutant channels with altered behavior before DTT treatment, data obtained under these conditions are shown as open circles, while data obtained after DTT treatment are shown as gray histogram bars.

Based on Fig. 6, we can divide the degrees of impact on channel function of side chain modification into three groups: high-impact, intermediate-impact, and low-impact. High-impact positions are those where cysteine substitution disrupts the channel function even when the thiol groups are in the reduced state. This is revealed by a marked MTSET sensitivity with or without DTT treatment, despite the maintained mutant behavior. The three nonfunctional mutants (N573C, K595C, and K638C) are also included in this group. There are a total of 15 high-impact positions: three at the outer end of S5 (I571, G572, and N573), eight in the 583–597 segment (W585, L586, N588, L589, G590, D591, I593, and K595), one between the 583–597 segment and the P-loop (G604), and three in the P-S6 linker (P632, T634, and K638).

Intermediate-impact positions are those where cysteine substitution in free thiol state is well tolerated (WT-like behavior, some requiring DTT treatment due to disulfide formation), but further MTSET modification produces a significant impact on channel function. There are 10 intermediate-impact positions: seven in the 583–597 segment (I583, G584, H587, Q592, G594, P596, and Y597), one between the 583–597 segment and the P-loop (D609), and two at the two ends of the P-loop (T613 and S631).

There are 22 positions where introduced cysteine side chains are well tolerated (WT-like phenotype) and do not form disulfide bonds (DTT has no further effects). Importantly, MTSET treatment does not have significant irreversible effects on these cysteine-substituted channels. Since the WT residues at these positions are mostly hydrophilic or even charged, it is likely that the insensitivity to MTSET is not due to a lack of accessibility, but due to a lack of impact of side chain modification on channel function. Therefore, these are low-impact positions where side chains are most likely solvent (extracellular aqueous phase) exposed: 574–582, 598–603, 606–608, 610, 612, 635, and 636.

Finally, there are four positions (P605, Y611, N633, and E637) where cysteine substitution leads to a DTT-resistant mutant phenotype. These cysteine-substituted channels are insensitive to MTSET or MTSES even after aggressive DTT treatment. Therefore, it is unclear whether the introduced thiol side chains are in the reduced state but are not accessible or not reactive to extracellular MTSET or MTSES, or the introduced thiol side chains are in very stable disulfide bonds that cannot be broken by DTT under our experimental conditions. The low expression level of all four mutants suggests that the second possibility may be more likely.

A striking feature of these summary results is the shared common mutant phenotype. Therefore, perturbing the conformation of HERG's outer vestibule, either due to disulfide bond formation involving introduced cysteine side chains (as evidenced by a restoration of WT-like behavior after DTT treatment), or due to a replacement of residues at high-impact positions with cysteine (as evidenced by a prominent MTSET sensitivity after DTT-treatment despite the maintained mutant phenotype), leads to common results: a disruption of C-type inactivation (Fig. 6 A), a decrease in K⁺ selectivity (Fig. 6 B), and a negative shift in V_0.5 of channel activation (Fig. 6 C). These observations support a key role played by HERG's outer vestibule in channel function. Note that the P_K/P_Na values are rough estimates of channel selectivity, because the intracellular K⁺ and Na⁺ concentrations are not directly measured. The estimated P_K/P_Na values may be mutant-dependent, because mutation-induced changes in membrane conductance may affect these intracellular ion concentrations. However, these uncertainties do not alter the conclusions listed above.

Another striking feature is the marked impact on channel function of cysteine substitution at 15 consecutive positions in the middle of the S5-P linker: the 583–597 segment (including 587 studied previously [Dun et al., 1999a]). K595C destroys functional expression. The other 14 channels all manifest the common mutant phenotype before DTT treatment. After DTT, seven assume the WT-like behavior: I583C, G584C, H587C, Q592C, G594C, P596C, and Y597C (rectification factor is switched from positive to negative, P_K/P_Na is increased, and V_0.5 of activation is shifted in the depolarizing direction). The remaining seven maintain the mutant behavior: W585C, L586C, N588C, L589C, G590C, D591C, and I593C. However, the enhancement of their sensitivity to MTSET by DTT treatment (Fig. 6 D) reveals that the introduced cysteine side chains at these positions indeed are reduced, despite the maintained mutant behavior (i.e., high-impact positions).

The distribution of these seven high-impact positions along the 583–597 segment gives a hint of α-helical pattern. To further explore this possibility, mutation-induced perturbations in channel activation gating (after DTT treatment) are used to analyze α-helical periodicity as described (Cornette et al., 1987; Li-Smerin et al., 2000) (see materials and methods). Fig. 7 A illustrates the Fourier transform power spectrum of the 583–597 segment. For K595C that does not produce functional channels, the value <V> of this segment is used (see materialsandmethods; Li-Smerin et al., 2000). The major peak occurs at an angular frequency of 94°, close to the expected peak of an ideal α-helix (100°). The α-helical tendency, as calculated by α-PI, is 1.64. Although the α-PI is less than two, the level considered as a strong indicator of α-helix formation, it nevertheless supports such a helical tendency. We further analyze the α-helix forming tendency along the S5-P linker from positions 571 to 613, using a 15-aa sliding window (Fig. 7 B). The analysis indicates that the middle part of this linker has the strongest likelihood of forming an α-helix. The amino end and especially the carboxyl end have little or no sign for long α-helices.

We have shown previously that MTSET or MTSES modification of S631C in HERG induces a switch of channel phenotype from WT-like to mutant (disrupting C-type inactivation and K⁺ selectivity) (Fan et al., 1999). These are similar to the mutant phenotype created by replacing S631 with a positively charged or a negatively charged residue (S631K and S631E, respectively) (Fan et al., 1999). In one-dimensional sequence, 631 is at the carboxyl end of the P-loop (Fig. 1). In three-dimensional space, position 631 is situated at the external entrance to the pore (Doyle et al., 1998). If the HERG outer mouth has a narrow dimension here, side chains at position 631 from all four subunits can be very close to each other. Adding multiple charges here (in a homotetramer of S631K or S631E, or after simultaneous modification of several S631C side chains by MTSET or MTSES) will lead to an electrical repulsion between like charges. These repulsive forces between 631 side chains around the external pore entrance may interfere with conformational changes necessary for an efficient C-type inactivation and a proper K⁺ selectivity.

T613 is at the amino end of the P-loop in one-dimensional sequence (Fig. 1). If it is also close to the external entrance of the pore, we predict that MTSET or MTSES modification of T613C should induce a phenotype switch, similar to the effects of MTS modification on S631C. Fig. 8 B shows tail I-V relationships of T613C before and after MTSET or MTSES modification. Before MTS treatment, the tail I-V relationships manifest a prominent negative slope in the V_r range of 30 to −60 mV (strong C-type inactivation) and an E_rev close to −100 mV (high P_K/P_Na). After modification by either MTSET or MTSES, the tail I-V relationships have a positive slope in the same V_r range (C-type inactivation disrupted) and the E_rev values are shifted toward 0 mV (P_K/P_Na reduced). These results confirm our prediction.

Among the other eight positions in the outer vestibule region where cysteine substitution sustains the WT-like behavior and MTS modification has a marked impact on channel function (Fig. 6), only one has the same response as those of S631C and T613C. Fig. 8 A shows that MTSET or MTSES modification of G584C also induces a phenotype switch. This position is 30 amino acids away from the pore entrance in the one-dimensional sequence (Fig. 1). However, the similarity in the MTS effects between G584C and the two pore-entrance residues suggests that in three-dimensional space G584 may be also close to the pore entrance.

Fig. 8 C shows that MTSET and especially MTSES cause a marked suppression of S631C channel conductance. This is not seen in either T613C (Fig. 8 B) or G584C (Fig. 8 A). These observations suggest that 631 side chains may be deeper into the pore than those of 613 or 584. Therefore, in addition to the charge effect on conformational changes around the pore entrance, MTS modification also causes a steric hindrance to current flow through the S631C pore.

Although MTSET modification of D609C also induces a phenotype switch, similar to the effects of MTSET on the pore-entrance residues, MTSES has little effect on D609C (Fig. 8 D). Therefore, the MTSET effects on D609C are due to a mechanism different from an induced electrical repulsion between like charges.

We perform cysteine-scanning mutagenesis on 51 residues that line the outer vestibule of the HERG channel (positions 571–613 of the S5-P linker and positions 631–638 of the P-S6 linker). Three of the cysteine-substituted channels do not produce functional expression. Of the other 48, we study the effects of cysteine substitution on channel function before and after DTT treatment and the effects of MTS modification. Fig. 9 A summarizes the major findings reported in this study, including: (a) disulfide-forming capability of introduced cysteine side chains, (b) channel phenotypes when the introduced cysteine side chains are in the reduced (free thiol) or oxidized (disulfide bonded) state, (c) effects of MTSET modification, and (d) impact of modifying side chain properties on the HERG channel function deduced from the above data. Disulfide bond formation before DTT or disulfide reduction by DTT is probably not an “all-or-none” reaction. However, from the pronounced changes in channel behavior and/or from the marked increase in MTSET sensitivity after DTT treatment, we can make clear-cut conclusions as to whether disulfide bond formation or reduction has occurred or not. Based on the results, we divide the degrees of impact of side chain modification into three groups: high-impact (cysteine substitution destroys functional expression, or disrupts channel function even when the thiol groups are in the reduced state); intermediate-impact (cysteine substitution in the free thiol state is well tolerated, but further MTSET modification produces a significant impact on channel function); and low-impact (neither cysteine substitution nor further MTSET modification impacts on channel function).

We proposed previously that there are fundamental differences between HERG and Shaker in the outer vestibule structure and the molecular motions underlying their C-type inactivation processes (Dun et al., 1999a; Fan et al., 1999; Tseng, 2001). Our proposal was based on the following observations: (a) In the Shaker channel, side chains at position 449 (equivalent of 631 in HERG, Fig. 1) interact directly with TEA bound to the outer mouth (Heginbotham and MacKinnon, 1992), whereas in HERG, side chains at 631 are shielded from the electrostatic effects of bound TEA (Pardo-Lopez et al., 2002). (b) T449E and T449K facilitate C-type inactivation in the Shaker (Lopez-Barneo et al., 1993), whereas S631E and S631K disrupt C-type inactivation in HERG (Fan et al., 1999). (c) MTS modification of T449C is strongly facilitated by C-type inactivation (Liu et al., 1996), whereas MTS modification of S631C has no state dependence (Dun et al., 1999b). (d) C-type inactivation in Shaker is associated with a decrease in P_K/P_Na (Starkus et al., 1997), whereas in HERG a disruption of C-type inactivation is associated with a decrease in P_K/P_Na (this report). (e) Two scorpion peptide toxins, charybdotoxin (ChTx) and BeKm-1, share a general three-dimensional structure (Miller, 1995; unpublished data), but use different interaction surfaces and different mechanisms in suppressing their respective target channels, Shaker and HERG (Hidalgo and MacKinnon,1995; unpublished data). The differences between ChTx and BeKm-1 most likely reflect differences in the conformation of the receptor site: the outer vestibule of target channel.

The most striking finding in our study is the critical role in the HERG channel function played by the 583–597 segment in the middle of the S5-P linker. We propose a structural model (Fig. 9 C) in which the 583–597 segment forms an α-helix (583–597 helix), with its amino end sitting at the pore entrance and the whole helix exposed to the extracellular aqueous environment. This is based on the following observations: (a) analysis of side chain properties suggests that this segment can form an amphipathic α-helix (Fig. 9 B), and α-helix periodicity analysis of mutation-induced perturbations of channel function supports such a secondary structure (Fig. 7), (b) MTS modification of G584C induces a phenotype-switch similar to that seen with MTS modification of two pore-entrance residues (T613C and S631C), whereas none of the other cysteine mutants have such a response, (c) all residues in the 583–597 segment are accessible to extracellular MTS reagents (Fig. 6 D), and (d) several resides along this segment (W585, G590, Q592 and I593) are critical for the binding of extracellular peptide toxins, ErgTx (Pardo-Lopez et al., 2002) and BeKm-1 (unpublished data). Such an α-helix will be ∼22 Å in length (Fig. 9 B). When oriented parallel to the plane of the cell membrane with its amino end near the pore entrance (Fig. 9 C), its carboxyl end will be able to reach the voltage-sensing domain. This is based on: (a) the dimension of KcsA (Doyle et al., 1998), and (b) a study using “molecular tapes” to estimate the distance between the pore entrance and the voltage-sensing domain of the Shaker channel (Blaustein et al., 2000). In this way, this 583–597 helix can serve as a bridge of communication between the outer mouth and the voltage-sensing domain. We hypothesize that depolarization-induced motions in the S4 domain (outward and/or rotational movements along with a change in the S4 tilt [Cha et al., 1999, Glauner et al., 1999]) can change the orientation of this 583–597 helix or push it toward the pore. This then pinches off the external entrance and causes C-type inactivation. Upon membrane repolarization, a reversal of the above S4 movements then relieves the force on the 583–597 helix, allowing the pore to reopen (recovery from C-type inactivation). These tightly choreographed molecular motions can account for the high voltage-sensitivity and the rapid onset and recovery kinetics of C-type inactivation in the HERG channel.

In Fig. 9 C, the 583–597 helix is depicted as being oriented parallel to the plane of the cell membrane. However, its orientation and interactions with other domains of the channel may change during channel gating (H. Robert Guy, personal communication). It is possible that in some gating state, part of the 583–597 helix may become more parallel to the pore axis, with its amino terminus participating in pore lining. This can explain the differential effects of MTS modification of thiol groups introduced to three consecutive positions at the amino terminus of this segment: 583–585. As discussed above, the phenotype-switch induced by either MTSET or MTSES modification of G584C can be explained by proposing that residues 584 lie at the pore entrance, with their introduced thiol side chains pointing toward the pore axis. If this part of the helix becomes somewhat parallel to the pore axis, 583 will be deeper into the pore and 585 will be on the channel surface. Increasing 583C side chain volume by either MTSET or MTSES will cause a steric hindrance to current flow through the pore. This can explain why I583C channel conductance is markedly reduced by either MTS reagent (Figs. 3 B and 4 B). On the other hand, adding positive charges to the 585 thiol side chains by MTSET will cause a positive shift in the local surface potential and reduce [K⁺] around the pore, leading to a decrease in channel conductance. MTSES should have the opposite effects. This may explain why W585C channel conductance is decreased by MTSET but increased by MTSES (Figs. 3 C and 4 C).

The crystal structure of a K channel pore domain, KcsA, shows that the outer end of M1 (S5 equivalent) and P-M2 linker (P-S6 equivalent) are close to each other in three-dimensional space (Doyle et al., 1998). Functional studies on the Shaker channel further suggest that the outer end of S5 (E418) interacts with the P-S6 linker (V451/G452) (Fig. 1) (Gandhi et al., 2000; Larsson and Elinder, 2000; Ortega-Saenz et al., 2000). We identify one group of high-impact positions at the outer end of S5 (571–573) and another group in the P-S6 linker (632–634). It is possible that in the HERG channel, region “571–573” interacts with region “632–634” through their side chains and peptide backbones during channel gating, and these interactions are important for channel function. Further experimentation is needed to confirm these interactions (e.g., by double cysteine substitution and testing for inter- or intrasubunit disulfide formation involving the introduced thiol groups), and to determine their role in channel function.

A string of three charges, two lysines flanking an aspartate (KDK), is located close to the amino end of the P-loop (positions 608–610, Fig. 1). Are these charges important for channel function? For example, can they contribute to the voltage-sensitivity of C-type inactivation in the HERG channel? Do they dominate the surface potential around the pore, so that they can influence the pore conductance via effects on the local K⁺ concentration? Neutralizing any one of these charges (K608C, D609C, or K610C) does not perturb channel function (intact C-type inactivation and strong K⁺ selectivity). Modification of K608C or K610C by MTSET, and more importantly by MTSES, has little impact on channel function or pore conductance. Fig. 8 D shows that although MTSES has little effect on D609C, MTSET modification induces a phenotype switch: C-type inactivation and K⁺ selectivity are disrupted. It is possible that the repulsive forces between the positive charge attached to the thiol side chain at 609 by MTSET and the positive charges on the two flanking lysine side chains can limit the backbone flexibility or perturb the conformation to such an extent that the pore function is disrupted. These observations rule out any major role played by these three charged residues in influencing either the voltage-sensitivity of C-type inactivation, or the local surface potential around the pore.

Conformational flexibility of the outer mouth in the HERG channel is key to its function. In this regard, the reduced hydrogen bonding capability around the outer mouth of the HERG channel, as suggested by the amino acid sequence alignment shown in Fig. 1, is an important contributing factor. In the Shaker channel, oxygen of the carboxylate side chain of E418 can form a hydrogen bond with the peptide backbone nitrogen in the P-S6 linker (451/452) (Larsson and Elinder, 2000; Ortega-Saenz et al., 2000). The E418 equivalent in the HERG channel, I571, does not have the oxygen donor for hydrogen bonding. In the Shaker, the nitrogen of W434 forms a hydrogen bond with the hydoxyl side chain of Y445 (Doyle et al., 1998). Again, in HERG the equivalent residues (Y616 and F627) cannot form a hydrogen bond.

In our structural model for HERG's outer mouth, this flexibility allows the 583–597 helix to assume different orientations and may become intercalated into the pore lining in some gating state. Therefore, rigidifying the outer vestibule structure by disulfide bond formation involving cysteine side chains introduced here is detrimental to C-type inactivation and K⁺ selectivity. Cysteine substitutions of critical residues at high-impact positions, or MTS modifications of pore-entrance residues, may also reduce the conformational flexibility and thus produce the same mutant phenotype. In the HERG mutants examined here, a disruption of pore properties (loss of C-type inactivation and K⁺ selectivity) goes hand-in-hand with a hyperpolarizing shift in the voltage dependence of activation (Fig. 6, A–C). This strong association suggests that rigidifying the outer mouth structure makes conformational changes induced by S4 movements propagate to the outer mouth region more efficiently, thus making the HERG channel open more readily. Therefore, we propose that a “floppy” outer mouth in the WT HERG channel makes it more reluctant to open but easier to close (C-type inactivate) than the Shaker channel.

The authors would like to thank Dr. H. Robert Guy (National Cancer Institute, National Institutes of Health) for helpful discussion.

This study is supported by HL 46451 from the National Heart, Lung, and Blood Institute; National Institutes of Health; and a Grant-in-Aid Award from the American Heart Association/Mid-Atlantic Affiliate (G.N. Tseng).