Mg²⁺ Modulates Voltage-Dependent Activation in Ether-à-Go-Go Potassium Channels by Binding between Transmembrane Segments S2 and S3

The Drosophila ether-à-go-go (eag) gene and its homologues encode a distinct subfamily of voltage-gated K⁺ channels (Warmke et al. 1991; Brüggemann et al. 1993; Warmke and Ganetzky 1994; Wei et al. 1996). In most members of the eag family, activation kinetics are dramatically regulated by extracellular Mg²⁺, an effect that is not seen in other types of voltage-gated K⁺ channels (Terlau et al. 1996; Frings et al. 1998; Schönherr et al. 1999; Tang et al. 2000). Analysis of eag ionic and gating currents indicates that Mg²⁺ directly modulates the process of voltage-dependent gating, presumably by binding to a site in or near the voltage sensor (Terlau et al. 1996, Tang et al. 2000). To investigate the mechanism by which Mg²⁺ regulates voltage-dependent activation and to gain novel insights into the structure and function of the voltage sensor, we have identified the Mg²⁺-binding site in eag channels.

In proteins, bound Mg²⁺ ions are often coordinated by the acidic side chains of aspartate and glutamate residues (da Silva and Williams 1991). Therefore, our analysis focused on acidic amino acids found in and near transmembrane segments S2 through S4, a region that includes essential components of the voltage sensor in K⁺ channels (Liman et al. 1991; Papazian et al. 1991, Papazian et al. 1995; Perozo et al. 1994; Planells-Cases et al. 1995; Aggarwal and MacKinnon 1996; Mannuzzu et al. 1996; Seoh et al. 1996; Cha and Bezanilla 1997; Cha et al. 1999; Glauner et al. 1999). Segments S2 and S3 contain three acidic residues that are conserved among all subfamilies of voltage-gated K⁺ channels (Warmke and Ganetzky 1994; Chandy and Gutman 1995). Previous work indicates that these residues make important contributions to the biogenesis, structure, and function of the voltage sensor in Shaker K⁺ channels (Papazian et al. 1995; Planells-Cases et al. 1995; Seoh et al. 1996; Tiwari-Woodruff et al. 1997, Tiwari-Woodruff et al. 2000). In the Drosophila eag channel, these residues correspond to D274 and D284 in S2 and D319 in S3 (see Fig. 1 A). In addition to these highly conserved positions, segments S2 and S3 contain two acidic amino acids, D278 in S2 and D327 in S3, that are conserved only among members of the eag subfamily (see Fig. 1 B) (Warmke and Ganetzky 1994; Chandy and Gutman 1995). Because these residues are restricted to the eag family, they are prime candidates to contribute to the Mg²⁺-binding site.

The mechanism of voltage-dependent activation has been extensively studied in K⁺ channels, notably Shaker channels (Bezanilla et al. 1994; Hoshi et al. 1994; Zagotta et al. 1994b; Larsson et al. 1996; Mannuzzu et al. 1996; Schoppa and Sigworth 1998a,Schoppa and Sigworth 1998b; Smith-Maxwell et al. 1998; Cha et al. 1999; Glauner et al. 1999; Ledwell and Aldrich 1999). Membrane depolarization initiates a series of voltage-dependent conformational changes that increase the probability of pore opening. Entry into the open state involves a less voltage-dependent and highly cooperative transition (Hoshi et al. 1994; Zagotta et al. 1994a,Zagotta et al. 1994b; Ledwell and Aldrich 1999). A conserved negatively charged residue in S2 and several positively charged residues in S4 contribute to the protein's ability to detect and respond to changes in membrane potential (Aggarwal and MacKinnon 1996; Seoh et al. 1996). During activation, the S4 residues traverse much or all of the transmembrane electric field (Aggarwal and MacKinnon 1996; Larsson et al. 1996; Seoh et al. 1996; Starace et al. 1997; Baker et al. 1998). Fluorescent measurements suggest that the S2 segment is involved in conformational changes that precede S4 movement during gating (Cha and Bezanilla 1997).

The voltage sensor is contained in a domain of the channel that includes S2 through S4 at a minimum, with a possible structural contribution from segment S1 (Papazian et al. 1995; Tiwari-Woodruff et al. 1997; Hong and Miller 2000; Li-Smerin et al. 2000a,Li-Smerin et al. 2000b). Evidence for tertiary structural interactions between charged residues in S2, S3, and S4 has been obtained using second site suppressor analysis (Papazian et al. 1995; Tiwari-Woodruff et al. 1997). At least some of these interactions are state dependent, occurring in a subset of the conformations along the activation pathway, consistent with the dynamic role of the voltage sensor in channel gating (Tiwari-Woodruff et al. 2000).

Voltage-dependent activation in eag can be described reasonably well by a sequential model in which each subunit undergoes two voltage-dependent transitions that prime the channel for opening, followed by a cooperative and less voltage-dependent step that opens the pore (Tang et al. 2000). One notable feature of eag activation is that voltage-dependent transitions that occur at more hyperpolarized potentials are rate limiting for pore opening, which is not true in Shaker channels (Bezanilla et al. 1994; Zagotta et al. 1994b; Terlau et al. 1996; Schönherr et al. 1999; Tang et al. 2000). Analysis of eag ionic and gating currents suggests that extracellular Mg²⁺ modulates rate limiting transitions that occur at hyperpolarized potentials, and voltage-dependent transitions close to the open state (Tang et al. 2000). In the activation model, the effects of Mg²⁺ can be simulated by reducing the forward and backward rates of the first voltage-dependent transition, and to a lesser extent, the forward rate of the second voltage-dependent transition (Tang et al. 2000).

To test the hypothesis that eag-specific acidic residues in S2 and S3 contribute to the Mg²⁺-binding site, these residues and others were individually mutated to determine the effect on sensitivity to Mg²⁺. The results indicate that Mg²⁺ binds between segments S2 and S3, and thereby modulates activation gating in eag channels.

Mutations in eag were generated by PCR using a three or four primer strategy (Landt et al. 1990; Sarkar and Sommer 1990). PCR products were digested with appropriate restriction enzymes and transferred into a pGEMHE subclone of wild-type eag (Tang and Papazian 1997). Transferred regions containing mutations were verified by dideoxy sequencing.

Constructs were linearized with NotI before in vitro transcription using the mMessage mMachine kit (Ambion). RNA encoding wild-type and mutated eag channels was injected into Xenopus oocytes for electrophysiological analysis (Timpe et al. 1988; Tang and Papazian 1997).

Ionic currents from wild-type and mutant eag channels were recorded at room temperature (20–22°C) using a two electrode voltage clamp as described previously (Timpe et al. 1988; Papazian et al. 1991). Oocytes were bathed in normal Ringer's solution, containing 118 mM NaCl, 1.8 mM CaCl₂, 10 mM HEPES, pH 7.2. The bath solution also contained from 0 to 20 mM MgCl₂, as indicated. To record eag tail currents, KCl was substituted for an equal concentration of NaCl, as noted. Electrodes contained 3 M KCl and had resistances of 0.3–1.0 MΩ.

Voltage pulse protocols were applied and data were acquired using pClamp v5.5.1 software and a TL-1 Labmaster interface (Axon Instruments, Inc.). The sampling rate was varied as needed to resolve activation kinetics. Data were filtered at 0.2 to 2 kHz using an 8-pole Bessel filter (Frequency Devices), and digitized at a frequency five times higher than the filter frequency. Linear capacitive and leak currents were subtracted using a P/−4 protocol (Bezanilla and Armstrong 1977).

Activation kinetics were compared in the absence and presence of MgCl₂, using concentrations up to 20 mM. Ionic currents were evoked by pulsing from a holding potential of −80 or −90 mV to voltages ranging from −60 mV to +60 or +100 mV in 20 mV increments. Activation kinetics were quantified in two ways as previously described (Tang et al. 2000). The late rising phase of the ionic current was fitted with a single exponential component to derive a time constant for activation, τ. Alternatively, the time to half maximal current amplitude was measured to obtain a value for t_1/2. These methods gave compatible results. Values of τ and t_1/2 are provided as mean ± SEM.

The probability of opening as a function of voltage was determined by measuring the amplitude of isochronal tail currents in 89 mM KCl bath solution. Ionic currents were evoked by pulsing from a holding potential of −90 mV to voltages ranging from −70 mV to +80 mV in 10 mV increments. Tail currents were recorded upon return to −90 mV. Tail current amplitudes were determined at isochronal points ranging from 4 to 10 ms following repolarization, normalized to the maximum value, and plotted versus test potential. The data were fitted with a single Boltzmann equation using Origin 5.0 software (Microcal) to obtain a slope factor and midpoint potential.

The effect of prepulse hyperpolarization on activation kinetics was investigated in the presence and absence of Mg²⁺ (Cole and Moore 1960). From a holding potential of −90 mV, wild-type channels were subjected to 50 ms prepulses to voltages ranging from −150 mV to −50 mV in 10 mV increments, followed by a test pulse to +60 mV. For mutant channels, the prepulse duration was increased to 250 ms and the range was extended in the depolarized direction, as noted.

The Drosophila eag K⁺ channel was expressed in Xenopus oocytes. Ionic currents were recorded using a two electrode voltage clamp in the presence and absence of 10 mM extracellular Mg²⁺, a concentration that produced a near maximal slowing of activation kinetics (see Fig. 4 B). Eag kinetics were quantified in two ways (Tang et al. 2000). A single exponential component was fitted to the late phase of activation to estimate a time constant (τ) for the process (Fig. 2 A). However, this τ value does not adequately describe the initial time course of the ionic current, which displays sigmoid kinetics (Ludwig et al. 1994; Terlau et al. 1996; Tang et al. 2000). Therefore, the time to half maximal current amplitude (t_1/2) was also measured (Fig. 2 B) (Tang et al. 2000). The t_1/2 value reflects changes in both the delay and time course of the ionic current.

Mg²⁺ slowed eag kinetics throughout the activation voltage range, with a larger effect after smaller depolarizations (Fig. 2) (Tang et al. 2000). Mg²⁺ decelerated activation by ∼10-fold at 0 mV, compared with 4-fold at +100 mV. To investigate whether eag activation kinetics become Mg²⁺-independent at very depolarized voltages, plots of τ and t_1/2 versus test potential were extrapolated beyond the experimentally accessible voltage range. Both plots were well fitted by single exponential functions (Fig. 2), which were used to predict limit values for τ and t_1/2 at infinite positive voltage. This procedure resulted in estimates of 6 ms for τ_lim in the presence of Mg²⁺, compared with ∼2 ms in its absence. Similarly, t_1/2lim was estimated to be 12 ms in the presence and ∼4 ms in the absence of Mg²⁺. Thus, Mg²⁺ is predicted to slow activation kinetics by about threefold at infinite positive voltage.

As previously described, hyperpolarizing prepulses also decelerate eag activation, increasing the delay and decreasing the rate of onset of ionic currents evoked by a subsequent test pulse (Ludwig et al. 1994; Terlau et al. 1996; Schönherr et al. 1999; Tang et al. 2000). In combination with hyperpolarizing prepulses, 10 mM Mg²⁺ produced a longer delay before the onset of ionic currents and further slowed activation kinetics (Fig. 3 A). These effects were quantified by measuring τ and t_1/2 as a function of prepulse potential in the presence and absence of Mg²⁺ (Fig. 3 B). The data reveal the existence of rate-limiting, Mg²⁺-sensitive transitions that occur at hyperpolarized potentials during the activation of wild-type eag channels (Tang et al. 2000).

Eag-specific acidic residues in S2 and S3 are good candidates to bind Mg²⁺ because of their location in the voltage sensing domain and their conservation in the eag subfamily. Neutralization of D278, the eag-specific acidic residue in S2, in the mutant D278V eliminated Mg²⁺ modulation of activation kinetics (Fig. 4). Fig. 4 A (left) shows currents evoked by depolarizing from −80 mV to +60 mV in the presence and absence of 10 mM Mg²⁺. Whereas wild-type eag showed measurable changes in activation kinetics upon exposure to 1 mM Mg²⁺, D278V was insensitive to Mg²⁺ concentrations as high as 20 mM (Fig. 4 B). Ionic current kinetics in D278V were unaffected by extracellular Mg²⁺ throughout the voltage range for activation (data not shown).

Interestingly, the conservative mutation D278E also had Mg²⁺-insensitive activation kinetics (Fig. 4 A, right, and B). This is significant because the glutamate substitution in D278E retains a carboxylic acid side chain, but is longer by one methylene group than the original aspartate residue. These data suggest that the mutation prevents Mg²⁺ from binding because the binding site is sterically constrained in the vicinity of D278.

Activation kinetics in D278V and D278E were quantified as a function of prepulse potential (Fig. 4 C). In these mutants, rate-limiting transitions were accessed at hyperpolarized potentials, as indicated by an increase in τ after hyperpolarizing prepulses (Fig. 4 C, left). Values for t_1/2 showed a similar trend (Fig. 4 C, right). Extracellular Mg²⁺ did not modulate the kinetics of these transitions, however (Fig. 4 C). An extended range of prepulses was used for D278V because this mutation shifted the probability of opening versus voltage (Po–V) curve in the depolarized direction compared with wild-type eag and the mutant D278E (Fig. 4 D). The transition to slower activation kinetics occurred at more depolarized potentials in D278V than in wild-type or D278E channels (Fig. 4 C).

Significantly, the D278V and D278E mutations had differential effects on eag activation kinetics. Although both mutants were insensitive to extracellular Mg²⁺, the activation kinetics of D278E, which retains a negatively charged side chain, were rapid, resembling those of the wild-type channel in the absence of Mg²⁺ (Fig. 4A and Fig. C). In contrast, the kinetics of D278V were slow, similar to those of the wild-type channel in the presence of Mg²⁺ (Fig. 4A and Fig. C). These results suggest that Mg²⁺ slows activation kinetics in wild-type eag in part by screening the negative charge at position D278. The neutralization mutation D278V mimics the Mg²⁺-bound state of wild-type eag, in which the charge at D278 is shielded by the ion, whereas D278E resembles the Mg²⁺-free state of wild-type eag, in which the charge is unshielded. These data support the conclusion that D278 in S2 contributes to the binding site by coordinating the Mg²⁺ ion.

D327, the eag-specific acidic residue in S3, was replaced by alanine or phenylalanine to generate the mutants D327A and D327F. Phenylalanine is found in the analogous position in Shaker channels (Fig. 1 B).

Extracellular Mg²⁺ concentrations up to 20 mM had no significant effect on activation kinetics in D327A channels (Fig. 5A and Fig. B). In this mutant, the late phase of activation was not well fitted by a single exponential component (data not shown), so activation kinetics were estimated by measuring t_1/2 (Fig. 5 C, right). Mg²⁺ had no significant effect on t_1/2 over the tested prepulse range (Fig. 5 C), which was extended to −40 mV because D327A shifted the Po–V curve in the depolarized direction (Fig. 5 D).

In contrast, Mg²⁺ slightly slowed the kinetics of D327F channels (Fig. 5 C). Values of τ and t_1/2 were determined as a function of prepulse potential in the presence and absence of Mg²⁺ (Fig. 5 C). In D327F, the value of t_1/2 increased from ∼33 ms in the absence of Mg²⁺ to ∼53 ms in the presence of 20 mM Mg²⁺, a difference that is statistically significant (P < 0.05, ANOVA) (Fig. 5 B). A small residual sensitivity to Mg²⁺ was also apparent when τ was plotted as a function of prepulse potential (Fig. 5 C). One interpretation of these results is that the negatively charged τ electrons associated with the phenyl ring in phenylalanine may be able to interact weakly with Mg²⁺ in D327F (Dougherty 1996; Williams et al. 1998; Wouters 1998).

D327F and D327A channels exhibited slow activation kinetics whether Mg²⁺ was present or absent (Fig. 5 A). In D327F, values for τ and t_1/2 resembled those of wild-type eag measured in the presence of Mg²⁺, whereas t_1/2 values in D327A were even longer (Fig. 5 C). Thus, neutralization of D327 resulted in activation kinetics that were qualitatively similar to those of the Mg²⁺-bound state of wild-type eag. These data suggest that Mg²⁺ slows activation kinetics in the wild-type channel in part by screening the charge at D327. The results support the conclusion that D327, like D278, contributes to the Mg²⁺-binding site by coordinating the ion.

Activation kinetics of the conservative mutant, D327E, were sensitive to extracellular Mg²⁺ (data not shown), in contrast to the results obtained with D278E (Fig. 4). Values for τ and t_1/2 measured either in the presence or absence of Mg²⁺ resembled those of the wild-type channel (data not shown). Mg²⁺ sensitivity of D327E suggests that the ion-binding site may be less sterically constrained near D327 than near D278.

D274 in S2 corresponds to an acidic residue that is widely conserved among voltage-dependent K⁺ channels (Chandy and Gutman 1995; Warmke and Ganetzky 1994) (Fig. 1 B). In the Shaker K⁺ channel, the D274 homologue is E283. Previous results strongly suggest that E283 lies near the extracellular surface of the Shaker protein, probably in a water-filled pocket or cavity (Tiwari-Woodruff et al. 2000). Upon Shaker activation, two S4 residues, R368 and R371, move into the vicinity of E283 (Tiwari-Woodruff et al. 2000). If S2 adopts an α-helical conformation, as suggested by perturbation analysis of Shaker and Kv2.1 K⁺ channels (Monks et al. 1999; Hong and Miller 2000; Li-Smerin et al. 2000a), in eag D274 and D278 will reside on the same face of the helix. In this case, D274 in eag may be located in a water-filled pocket that contains the Mg²⁺-binding site. To test this hypothesis, we investigated whether D274 in S2 contributes to the Mg²⁺-binding site in eag channels.

The mutations D274A and D274E were generated and expressed in Xenopus oocytes. In the absence of Mg²⁺, the activation kinetics of D274A and D274E were similar to wild-type eag. And, as in wild-type, activation of D274A and D274E was significantly slowed by Mg²⁺, indicating that D274 is not required for Mg²⁺ binding (Fig. 6 A). In fact, the neutralization mutation D274A was as sensitive to Mg²⁺ as wild-type eag. Values of τ measured as a function of prepulse potential in the presence and absence of Mg²⁺ were similar in wild type and D274A (data not shown). In addition, the change in t_1/2 was virtually identical for Mg²⁺ concentrations up to 20 mM (Fig. 6 B). These results indicate that D274 does not contribute to eag's Mg²⁺-binding affinity, and is therefore unlikely to interact with the bound ion.

Interestingly, two lines of evidence suggest that substituting D274 with the longer acidic amino acid, glutamate, increased the binding affinity for Mg²⁺. First, Mg²⁺ had a larger effect on t_1/2 in D274E than in wild-type eag, with a shift in the dose response curve to lower Mg²⁺ concentrations (Fig. 6 B). Second, after removing Mg²⁺ from the bath solution, activation kinetics recovered more slowly in D274E than in wild-type eag (Fig. 6 C) or other Mg²⁺-sensitive mutants such as D274A (data not shown). This effect was quantified by measuring the time interval between 10 and 90% of the peak current amplitude as a function of perfusion time during wash in and wash out of a bath solution containing 20 mM Mg²⁺ (Fig. 6 C). These results indicate that the mutation D274E decreases the Mg²⁺ dissociation rate, thereby increasing the apparent binding affinity. The data suggest that the longer glutamate residue is able to interact with the bound ion, unlike the original aspartate. We conclude that D274 is close to the binding site, consistent with the idea that D274 is exposed in the same water-filled pocket that contains the ion-binding site.

The position corresponding to D284 in eag is highly conserved among voltage gated K⁺ channels (Fig. 1 B). In Shaker channels, the homologous residue E293 plays a key role in the mechanism of voltage-dependent activation. It contributes to the single channel gating charge, either by moving relative to the transmembrane electric field or by determining the profile of the field traversed by charge-moving residues in S4 (Seoh et al. 1996). In addition, this residue has been implicated in a structural interaction with a positively charged lysine in S4 (K374) in Shaker channels (Papazian et al. 1995; Tiwari-Woodruff et al. 1997).

In the eag mutants D284N and D284A, activation kinetics were fast, similar to wild-type eag, and unlike neutralization mutations at D278 and D327 (Fig. 7; compare to Fig. 4 and Fig. 5). Mg²⁺ had little effect on activation kinetics in D284N and D284A when currents were evoked by depolarizing from a holding potential of −80 mV (data not shown). However, hyperpolarizing prepulses revealed altered kinetics in the presence of Mg²⁺ in D284N (Fig. 7A and Fig. B). A similar trend was evident in D284A channels (Fig. 7A and Fig. C). These results indicate that activation kinetics in D284 mutants remain sensitive to hyperpolarizing prepulses, and that transitions occurring at hyperpolarized voltages retain their sensitivity to Mg²⁺. Therefore, D284 is not required for Mg²⁺ binding.

Interestingly, the results indicate that the sensitivity of activation kinetics to prepulse voltages has been shifted to more hyperpolarized potentials in the D284 mutants, despite the fact that their Po–V curves were shifted in the depolarized direction (Fig. 7 D). The data suggest that mutations at D284 separate the voltage dependence of transitions in the eag activation pathway, shifting rate-determining, Mg²⁺-sensitive transitions in the hyperpolarized direction, and channel opening in the depolarized direction. In the presence of Mg²⁺, these transitions appeared to be faster in the mutants than in wild type (compare Fig. 2 and Fig. 7). This may explain the observation that ionic current kinetics in D284N and D284A were unaffected by Mg²⁺ at voltages where these channels opened. In wild type, in contrast, activation kinetics measured in the presence and absence of Mg²⁺ did not converge even at +100 mV (see Fig. 2). These results are consistent with the idea that D284 makes an important contribution to the gating mechanism, as does its homologue, E293, in Shaker channels (Seoh et al. 1996).

D319, a residue near the intracellular end of S3, is highly conserved in different types of voltage-gated K⁺ channels (Fig. 1 B). In Shaker, the analogous residue, D316, has been implicated in structural interactions with K374 in S4 that are essential for proper folding of the voltage sensor (Papazian et al. 1995; Tiwari-Woodruff et al. 1997).

In the eag mutant, D319N, ionic current kinetics were fast, similar to those of D284 neutralization mutants and wild-type eag in the absence of Mg²⁺ (Fig. 8 A; compare to Fig. 2 and Fig. 7). Mg²⁺ had no significant effect on the rate of activation of ionic currents evoked by depolarizing from −80 mV or after hyperpolarizing prepulses (Fig. 8A and Fig. B). An extended range of prepulses was used because the D319 mutation shifted the Po–V curve in the depolarized direction (Fig. 8 C). In contrast to the wild-type channel, a transition to slower activation kinetics was not detected after hyperpolarized prepulses (Fig. 8 B).

The properties of D319N resemble those of D284N and D284A, except that hyperpolarizing prepulses did not reveal Mg²⁺-sensitive, rate-limiting transitions. Significantly, mutations that neutralized either D284 or D319 resulted in fast kinetics that did not mimic the Mg²⁺-bound state of wild-type eag, in contrast to neutralization mutations of D278 and D327. These results strongly suggest that D284 and D319 do not mediate the slowing of activation caused by Mg²⁺ binding. One possibility is that D319N, like D284N and D284A, primarily affects the eag gating pathway, and shifts the Mg²⁺-sensitive, slow transitions in the hyperpolarized direction. If so, the shift would be greater in D319N than in D284N or D284A. Because D319N was Mg²⁺ insensitive in the experimentally accessible voltage range, however, we cannot rule out the possibility that D319 is required for Mg²⁺ binding.

Extracellular Mg²⁺ dramatically slows voltage-dependent activation in eag K⁺ channels (Terlau et al. 1996; Schönherr et al. 1999; Tang et al. 2000). We have tested the hypothesis that eag-specific acidic residues found in transmembrane segments S2 and S3 contribute to the ion-binding site. Our results indicate that Mg²⁺ binds between S2 and S3 in the eag voltage sensor and thereby modulates voltage dependent activation. Eag-specific residues D278 in S2 and D327 in S3 are likely to coordinate the ion. In contrast, D284, an acidic residue in S2 that is conserved among all families of voltage-dependent K⁺ channels, is not required for ion binding. Mutations at this position significantly altered the voltage dependence of transitions in the eag activation pathway, suggesting a role for S2 in channel gating. Neutralization of D319, a widely conserved acidic residue in S3, also affects gating or ion binding, or both. Interestingly, the data indicate that D274 in S2, another widely conserved acidic residue, is not required for ion binding, but is in the vicinity of the binding site. Identification of the Mg²⁺-binding site in eag constrains the packing arrangement of transmembrane segments in the voltage sensor and provides insight into the molecular mechanism of Mg²⁺ action.

Several lines of evidence support the conclusion that D278 and D327 coordinate the Mg²⁺ ion. Neutralization mutations at these positions eliminated or significantly reduced Mg²⁺ modulation of activation kinetics. Importantly, neutralization of either D278 or D327 dramatically slowed activation, mimicking the kinetics of the Mg²⁺-bound state of wild-type eag. The data suggest that Mg²⁺ slows activation in the wild-type channel by screening these negatively charged residues. The conclusion that D278 and D327 bind the ion is further strengthened by evidence that D274 in S2, one α-helical rung above D278, is unlikely to coordinate Mg²⁺, but is quite close to the binding site. The mutation D274E increased the apparent affinity for Mg²⁺, suggesting that glutamate, which has an acidic side chain ∼1.5 Å longer than that of the original aspartate residue, is able to interact with an ion in the site.

The role of eag-specific residues in forming the binding site explains why modulation of activation kinetics by Mg²⁺ is unique to the eag family. Neutralizing either eag-specific aspartate residue eliminated Mg²⁺ sensitivity. It is worth noting that we have previously shown that a short sequence of primarily acidic residues in the S3–S4 loop is not required for Mg²⁺ binding, consistent with the presence of negatively charged residues at the end of S3 in a variety of K⁺ channels (Tang et al. 2000). Our results can also account for the finding that some members of the eag family, such as elk, are insensitive to Mg²⁺ (Trudeau et al. 1999). Elk contains a glutamate residue at the position equivalent to D278 (Fig. 1 B) (Warmke and Ganetzky 1994; Trudeau et al. 1999). We found that the D278E mutation abolished Mg²⁺ sensitivity of Drosophila eag, presumably due to steric constraints that prevent ion binding.

In the structures of various Mg²⁺-binding sites, four, six, or eight ligands coordinate the ion (da Silva and Williams 1991; Williams 1993). The carboxylate side chains of D278 and D327 could provide at most four oxygen atoms to coordinate the bound Mg²⁺. If these are the only ligands contributed by the eag protein, up to four water molecules ions may remain associated with the bound ion, depending on the geometry of the site. Other water molecules from the hydration shell are presumably lost upon binding.

Interestingly, a variety of cations slow activation in eag (Terlau et al. 1996). Given their qualitatively similar effects on ionic current kinetics, it is reasonable to propose that these different ions bind to the same site as Mg²⁺. The extent to which an ion slows activation is correlated with its enthalpy of hydration (Terlau et al. 1996), consistent with our results, which suggest that an ion undergoes partial dehydration upon binding to eag. Presumably, the binding site must have some structural flexibility if it is able to accommodate different ions. Our results suggest the Mg²⁺-binding site is sterically constrained near D278, but may be more flexible near D327. This may contribute to the ability of the site to bind a variety of ions.

Residues that coordinate the same Mg²⁺ ion must be within atomic distance of one another. Therefore, our results suggest an important structural constraint for the packing arrangement of transmembrane segments in the eag voltage sensor. In a variety of high resolution protein structures containing bound Mg²⁺ ions, the center-to-center distance between Mg²⁺ and coordinating oxygen atoms is approximately 2.2 Å (Kankare et al. 1996; Andersson et al. 1997; Harutyunyan et al. 1997; Larsen et al. 1997; Sliz et al. 1997; Stec et al. 1998; Goldgur et al. 1999; Nichols et al. 1999). In these structures, the carboxylate groups of acidic residues that coordinate Mg²⁺ are within 2.7 to 4.3 Å of each other, providing an estimate of the likely distance between D278 in S2 and D327 in S3 when Mg²⁺ is bound.

Structural proximity between D278 and D327 is compatible with previously inferred structural constraints between S2, S3, and S4 in the Shaker voltage sensor (Papazian et al. 1995; Tiwari-Woodruff et al. 1997, Tiwari-Woodruff et al. 2000). Assuming structural conservation between Shaker and eag, we have combined our results from these channels to derive a model for the packing arrangement of segments S2, S3, and S4 in voltage-dependent K⁺ channels (Fig. 9) (Papazian et al. 1995; Tiwari-Woodruff et al. 1997, Tiwari-Woodruff et al. 2000). This model represents the activated conformation of the voltage sensor, in which E283 in S2 (the Shaker homologue of D274) achieves its closest proximity with R371, the fourth basic residue in the Shaker S4 segment (Tiwari-Woodruff et al. 2000). Importantly, the eag results have been used to constrain the relative positions of segments S2 and S3, which cannot be determined from the Shaker results alone (Tiwari-Woodruff et al. 1997). We have assumed that S2, S3, and S4 are α-helical, consistent with the results of perturbation analysis in Shaker and Kv2.1 channels (Monks et al. 1999; Hong and Miller 2000; Li-Smerin et al. 2000a).

In Fig. 9, the S4 segment is shown tilted relative to S2 and S3, a feature introduced in our earlier model to satisfy the structural constraints inferred for Shaker (Tiwari-Woodruff et al. 1997). Interestingly, Li-Smerin et al. 2000b have recently reported the results of perturbation analysis on the putative outer surface of the Shaker pore domain. Their data suggest that the voltage sensor packs against the pore via a localized interaction surface. Significantly, this surface is tilted relative to the center axis of the pore at an angle very similar to that of the S4 segment in our model (Tiwari-Woodruff et al. 1997). Thus, results from three independent approaches, second site suppressor analysis in Shaker (Papazian et al. 1995; Tiwari-Woodruff et al. 1997, Tiwari-Woodruff et al. 2000), characterization of the Mg²⁺-binding site in eag (this study), and perturbation analysis of the pore domain in Shaker (Li-Smerin et al. 2000b) support the packing model for the voltage sensor proposed in Fig. 9.

Results obtained with both Shaker and eag strongly suggest that there is an externally facing, water-filled pocket next to S2. In eag, extracellular Mg²⁺ is able to access a site formed by D278 in S2 and D327 in S3, which would project into the pocket to coordinate the ion. In an α-helix, D274 in S2, located one rung above D278, would also be exposed in the same cavity. Significantly, previous work in Shaker channels indicates that E283, the S2 residue analogous to D274 in eag, is located in a cavity near the extracellular surface of the protein (Tiwari-Woodruff et al. 2000). Upon activation of Shaker channels, two S4 residues, first R368 and then R371, move into the vicinity of E283 (Tiwari-Woodruff et al. 2000). Taken together, the eag and Shaker results suggest that voltage-dependent K⁺ channels contain a water-filled pocket lined by S2 and S3. S4 residues also project into this space, at least in the activated conformation. An important difference between the pockets in Shaker and eag channels determines their differential sensitivity to extracellular Mg²⁺. In eag, acidic residues contribute to a Mg²⁺-binding site at the bottom of the cavity. In Shaker, in contrast, the cavity may be shallower than in eag due to the presence of hydrophobic residues at the positions analogous to D278 and D327 (Fig. 1 B). We conclude that voltage-dependent K⁺ channels contain an externally facing, water-filled pocket between S2, S3, and S4.

Interestingly, this pocket may be nearly isopotential with the extracellular solution. In Shaker, a cysteine substituted for E283 in S2 shows very similar reactivity at depolarized and hyperpolarized potentials (Tiwari-Woodruff et al. 2000). In eag, Mg²⁺ affects activation kinetics at +100 mV (Fig. 2), indicating that Mg²⁺ is not expelled from its binding site by depolarizing the cell.

Results obtained in a variety of laboratories suggest that activation of K⁺ channels involves at least two voltage-dependent conformational changes per subunit (Bezanilla et al. 1994; Perozo et al. 1994; Sigg et al. 1994; Stefani et al. 1994; Zagotta et al. 1994a,Zagotta et al. 1994b; Cha and Bezanilla 1997; Baker et al. 1998; Schoppa and Sigworth 1998a,Schoppa and Sigworth 1998b). Both S2 and S4 are likely to participate in conformational changes during the first phase of activation, which occurs at hyperpolarized potentials (Cha and Bezanilla 1997; Baker et al. 1998). Evidence for involvement of S2 in this phase includes site-directed fluorescent measurements in Shaker (Cha and Bezanilla 1997) and functional analysis of eag mutants (this study). Involvement of the S4 segment in the first phase is likely because intermediate conformations of S4, which do not correspond to either the fully resting or the fully activated conformation, have been detected (Baker et al. 1998; Tiwari-Woodruff et al. 2000). The second phase of activation occurs at more depolarized potentials than the first, involves an additional movement of S4, and is closely correlated with channel opening (Bezanilla et al. 1994; Sigg et al. 1994; Stefani et al. 1994; Zagotta et al. 1994a,Zagotta et al. 1994b; Mannuzzu et al. 1996; Cha and Bezanilla 1997; Baker et al. 1998). In Shaker, R371 in S4 moves into proximity with E283 in S2 during the second phase of the mechanism (Tiwari-Woodruff et al. 2000). Although this picture of activation is undoubtedly oversimplified, it provides a compatible synthesis of results obtained using several approaches.

Based on this picture of activation, we propose a speculative model of the mechanism by which Mg²⁺ modulates voltage-dependent gating in eag. We previously suggested that Mg²⁺ modulates two different voltage-dependent transitions during eag activation, corresponding to the two phases described above (Tang et al. 2000). We have now located the Mg²⁺-binding site within the water-filled pocket between S2 and S3, an ideal location for modulating the kinetics of both phases of activation. Mg²⁺ could modulate the kinetics of the first phase of activation by forming a bridge between S2 and S3, and thereby slowing conformational changes of S2 relative to S3. During both phases of activation, bound Mg²⁺ could slow the rate at which positively charged S4 residues (the eag homologues of R368 and R371 in Shaker) move into the pocket to interact with D274. Although this model is speculative, it is compatible with the emerging picture of the physical mechanism of voltage-dependent gating in K⁺ channels and with our previous characterization of Mg²⁺ action on ionic and gating currents in eag.

Our results are consistent with the expectation that the overall structure and function of the voltage sensor is conserved in K⁺ channels that have a wide range of kinetics and differential regulation. Presumably, the detailed kinetic properties of activation in individual types of K⁺ channels are conferred by sequence differences that result in subtle structural changes despite a conserved architecture.

We are grateful to Dr. James Bowie for helpful discussions and to members of the Papazian lab for their comments on the manuscript.

This work was supported by grants to D.M. Papazian from the National Institutes of Health (NIH; GM43459); the American Heart Association, Western States Affiliate; and the Laubisch fund for cardiovascular research at UCLA. W.R. Silverman was partially supported by an NIH predoctoral training grant (GM08496).