Molecular Analysis of a Binding Site for Quinidine in a Human Cardiac Delayed Rectifier K+ Channel
Role of S6 in Antiarrhythmic Drug Binding
Abstract The antiarrhythmic agent quinidine blocks the human cardiac hKv1.5 channel expressed in mammalian cells at therapeutically relevant concentrations (EC50, 6.2 μmol/L). Mechanistic analysis has suggested that quinidine acts as a cationic open-channel blocker at a site in the internal mouth of the ionic pore and that binding is stabilized by hydrophobic interactions. We tested these hypotheses using site-directed mutagenesis of residues proposed to line the internal mouth of the channel or of nearby residues. Amino acid substitutions in the midsection of S6 (T505I, T505V, T505S, and V512A) reduced the dissociation rate for quinidine, increased the affinity (0.7, 1.5, 3.4, and 1.4 μmol/L, respectively), and preserved both the voltage-dependent open channel–block mechanism and the electrical binding distance (0.19 to 0.22). In contrast, smaller or nonsignificant effects were observed for: deletion of the intracellular C-terminal domain, charge neutralizations in the region immediately C-terminal to S6, elimination of aromatic residues in S6, and mutations at the putative internal turn of the P loop, at the external entrance of the pore, and at sites in the S4S5 linker. The ≈10-fold increase in affinity with T505I and the reduction of the dissociation rate constant with the mutations that increased affinity are consistent with a hydrophobic stabilization of binding. Moreover, the T505 and V512 residues align on the same side of the putative α-helical S6 segment. Taken together, these results localize the hydrophobic binding site for this antiarrhythmic drug in the internal mouth of this human K+ channel and provide molecular support for the open channel–block model and the role of S6 in contributing to the inner pore.
Cardiac voltage-gated K+ channels are integral membrane proteins that open a highly selective K+ permeation pathway in response to changes in the transmembrane potential. The resulting K+ efflux counteracts, and eventually overcomes, the inward current that sustains the plateau phase of the long cardiac action potential. As such, these channels contribute to the control of the action potential duration and are molecular targets for the class of antiarrhythmic drugs that act by prolonging action potential duration (class III action). Quinidine is a widely used antiarrhythmic agent that blocks the cloned human K+ channel hKv1.5 at pharmacologically relevant concentrations.1 The hKv1.5 protein is expressed in human heart,2 and a current with similar biophysical and pharmacological properties has been identified in human atrial myocytes.3 4 Thus, hKv1.5 may form an important molecular target for treatment of atrial tachyarrhythmias, which represent a major clinical problem with serious morbidity.5
Quinidine block of hKv1.5 is characterized by a time-dependent decline of open-channel current and a subsequent biphasic time course of tail currents with an initial rising phase followed by a decay that is slower compared with control values.1 The voltage dependence of block displays a steep phase coincident with the voltage dependence of channel activation, but it continues to show an increase in block at more positive potentials, where the channel is fully activated. The latter more shallow voltage dependence is consistent with a binding reaction sensing ≈20% of the applied transmembrane electrical field and indicates that quinidine acts in its cationic form, which predominates at physiological pH (>95%, pKa 8.9). Although quinidine block of hKv1.5 resembles open-channel block by QA derivatives,6 7 the drug also contains an aromatic moiety that could be involved in binding, and its hydrophobic moiety (quinucleidine ring) differs substantially from the alkyl side chains of TEA and its QA derivatives.
hKv1.5 belongs to a class of voltage-gated K+ channels thought to consist of four subunits, each containing six transmembrane segments (S1 to S6) (for a review, see Reference 88 ). The segment between S5 and S6 forms the external part of the ion conduction pathway (deep pore or P region). The flanking S5 and S6 segments may contribute to the presumably wider intracellular mouth of the ion channel. To identify putative binding sites for quinidine in the hKv1.5 pore region, we tested for potential electrostatic interaction by neutralizing acidic residues or by introducing charged residues in S6 and in the S4S5 linker. Since the cluster of aromatic residues at the cytoplasmic end of S6 could participate in π-electron interaction with quinidine’s positive charge,9 we also changed the aromatic character of this region. Finally, we modified residues in the S6 segment and the deep pore region proposed to line the ion conducting pore.10 Previous studies using two-electrode voltage clamp of Kv1.4 and Kv1.5 channels expressed in Xenopus oocytes have suggested that these cloned channels are only blocked with high concentrations of quinidine (EC50, 700 to 1200 μmol/L for Kv1.4) (References 11 and 1211 12 and authors’ unpublished data, 1996). Such concentrations would be highly toxic to humans, whereas the value of 6 μmol/L observed with expression in mammalian cells is in the therapeutic range. Therefore, we used mammalian cell lines to assess the effects of the mutations reported in the present study.
Materials and Methods
The PCR-based site-directed mutagenesis used two complementary primers containing the desired mutation and two flanking PCR primers (P1 and P4) encompassing two unique restriction sites (Pst I and BamHI) enclosing the S4 to S6 region of hKv1.5. Two parallel PCR reactions, each with a flanking primer and a primer containing the desired mutation, generated two DNA fragments with overlapping ends. After gel purification, both fragments were annealed in a third reaction, and a fragment containing all the sequence between P1 and P4 was amplified with these primers. The final PCR product was digested with Pst I and BamHI, gel-purified, and ligated into the pBSKS+ vector containing the coding region of hKv1.5 from which the Pst I and BamHI segments had been removed. All PCR-generated sequence was verified by double-stranded sequencing. Once the desired sequence was confirmed, the mutant hKv1.5 was released from pBS with Xba I and Xho I and subcloned into the pMSVneo expression vector.13 Plasmid DNA was purified by sedimentation on CsCl gradients and transfected into Ltk− cells as described previously.13 For each mutation, at least six foci were collected and grown. The cell lines were checked for hKv1.5-specific mRNA expression (usually five of the six foci were positive) and tested for functional expression. Cell culture and cell preparation for experimental use were as reported previously.13 If a mutation yielded functional channels, then all mRNA-positive lines expressed the channel, albeit at different levels. Of the expressing cell lines, those with currents between 1 and 4 nA at +40 mV were used preferentially to ensure adequate voltage-clamp control. A mutation was considered nonfunctional when no voltage-gated currents were observed after two or three different inductions of at least two mRNA-positive cell lines. The absence of cell surface protein was confirmed by immunohistochemistry in a few cases (I513A and F517V) using hKv1.5-specific antibodies as described elsewhere.2
Recordings were made with an Axopatch-1 or Axopatch-200A patch-clamp amplifier (Axon Instruments) using the whole-cell configuration of the patch-clamp technique.14 Currents were recorded at room temperature (21°C to 23°C) and were sampled at 1 to 10 kHz after anti-alias filtering at 0.5 to 5 kHz. Data acquisition and command potentials were controlled by a custom-made programmable stimulator.13 Micropipettes with DC resistance of <2.5 MΩ were made from starbore borosilicate glass (Radnoti Glass Co). After establishing the whole-cell configuration, the capacitive transients elicited by 10-mV voltage-clamp steps from −80 mV were recorded at 50 kHz to obtain cell capacitance and access resistance. Although we selected cell lines for current levels in the range of 1 to 4 nA (at +60 mV), some mutations consistently expressed larger currents. To ensure adequate voltage-clamp control, we calculated the residual access resistance (total access resistance minus amount of analogue compensation) for each experiment individually (range, 0.3 to 3 MΩ) and excluded cells in which the series resistance error exceeded 5 mV.
The intracellular pipette filling solution contained (mmol/L) KCl 110, HEPES 10, K4BAPTA 5, K2ATP 5, and MgCl2 1 and was adjusted to pH 7.2 with KOH, yielding a final intracellular K+ concentration of ≈145 mmol/L. The bath solution contained (mmol/L) NaCl 140, KCl 4, CaCl2 1.8, MgCl2 1, HEPES 10, and glucose 10 and was adjusted to pH 7.35 with NaOH. Quinidine (quinidine gluconate, Eli Lilly) was added from a 10 mmol/L aqueous stock solution. All other chemical compounds were obtained from Sigma Chemical Co.
The holding potential was −80 mV unless indicated otherwise. For V512A, we used a holding potential of −100 mV, because the voltage dependence of activation occurred at more negative voltages compared with WT. The effects of drug exposure or removal were monitored with test pulses to +50 mV applied every 20 seconds until steady state was obtained (10 to 15 minutes). The interpulse interval was 15 or 20 seconds for all protocols to prevent accumulation of block or slow inactivation.15 For each mutation, we determined the apparent affinity, electrical binding distance δ, and binding kinetics, similar to the approach used previously.1 16 Data were corrected for passive linear leak, and raw tracings shown in this article were digitally filtered at 1 kHz in the frequency domain after Fourier transformation. The time courses of activation and of drug-induced kinetic changes were fitted with a sum of exponentials. The voltage dependence of channel activation was fit with a Boltzmann equation: y=1/(1+exp[−(E−Eh)/k]), in which k represents the slope factor, E the imposed voltage, and Eh the voltage at which 50% of the channels are activated. The curve-fitting procedures used a nonlinear least-squares (Gauss-Newton) algorithm, and the results were displayed in linear and semilogarithmic format together with a plot of the residual deviations of the data from the fitted curve. Goodness of the fit and the required number of exponential components were judged by comparing χ2 values statistically (F test) and by inspection for systematic nonrandom trends in the residual deviations. Results are expressed as mean±SEM. ANOVA with appropriate post hoc comparisons was used to compare the differences in mean values; P<.05 was considered significant.
Quinidine blocks hKv1.5 with an affinity of 6 μmol/L,1 and preliminary experiments indicated affinities between 3.7 and 22 μmol/L for apparent open-channel block of other Kv channels expressed in L cells: Kv1.1 (22±4 μmol/L, n=4), Kv1.2 (19±2 μmol/L, n=2), Kv2.1 (4.3±0.5 μmol/L, n=4), Kv4.2 (3.7±1 μmol/L, n=3), and Kv1.4 (47±5% suppression of peak current with 20 μmol/L quinidine, n=4). This general similarity suggested a binding site in the pore region. Fig 1⇓ illustrates the high degree of conservation in the S6 sequences of these cloned Shaker-related channels. The α-helical wheel projection of this sequence reveals a clustering, on two thirds of the circumference, of hydrophobic residues that are absolutely conserved among the four mammalian Kv families (and Shaker). The variable sites populate the remainder of this putative helical surface, suggesting that these hydrophobic residues may be exposed to the aqueous environment. Indeed, the functional effects of mutations of these nonconserved sites in Shaker and Kv3.1 have been interpreted to indicate that these residues line part of the ion-conducting pore.10 17 A short stretch of aromatic residues is conserved near the intracellular end of S6 (Fig 1A⇓), and the Kv1 family also displays a conserved stretch of negatively charged residues immediately C-terminal to S6. Since open-channel block by quinidine appears to involve electrostatic and hydrophobic components,1 18 we targeted the variant aromatic, charged, and hydrophobic residues in S6. Fig 2⇓ shows that the basic phenotype of hKv1.5 was preserved for 11 functional mutations in this region, as well as for 2 mutations each in the P loop and S4S5 linker. In each case, the current activated rapidly, with time constants between 1 and 4 milliseconds at +60 mV, followed by a limited degree of slow inactivation during the remainder of the 250-millisecond depolarization. Therefore, these mutations had minimal impact on channel behavior in the voltage range used to assess the affinity for open-channel block by quinidine. The effects of quinidine on the various mutations illustrated in Fig 2⇓ are analyzed in detail below.
Mutations in P Loop and in the Midsection of S6
Since quinidine block of hKv1.5 resembled internal QA binding, we tested whether the pore mutation T477S affected quinidine binding. This mutation is equivalent to the T441S mutation that reduced internal TEA affinity 10-fold in Shaker.18 Fig 2⇑ shows that 6 μmol/L quinidine blocked T477S by 40%, consistent with a modest reduction (<2-fold) of the quinidine affinity (Table⇓). The L508 residue has been implicated in the differential affinity of Kv2.1 and Kv3.1 for internal TEA.10 20 Mutation of L508 into methionine (the corresponding residue in Kv3.1, see Fig 1A⇑) had minimal effects on both affinity and kinetics of block (Fig 2⇑, Table⇓). Therefore, residues that modify internal TEA affinity 10-fold in the Shaker, Kv2.1, or Kv3.1 background have minimal effects on quinidine binding in the hKv1.5 channel.
This finding raised the possibility that quinidine binding may be similar to binding of more hydrophobic TEA derivatives, which are similarly less affected by the T441S mutation in Shaker.7 Therefore, we mutated other putative pore-lining residues in the midsection of S6 (Fig 1B⇑). Mutation T505S preserves the polar hydroxyl group, and mutation T505V retains the geometry but is nonpolar. In addition, valine is the corresponding residue in the Kv2 and Kv4 subfamilies (Fig 1A⇑). Furthermore, we replaced T505 with leucine and isoleucine to increase the size and hydrophobic character of the side chain. The latter mutations have been shown to increase C8-triethylammonium block in Shaker channels.7 These four mutations yielded functional channels, but low current levels precluded study of the T505L mutation. Fig 3A⇓ shows the effects of 0.6, 2, and 6 μmol/L quinidine on T505I and the subsequent washout. Quinidine induced a reversible and time-dependent relaxation of the open-channel current. Both the extent and apparent rate of block increased in a concentration-dependent manner. Since block had reached steady state in 250 milliseconds, we used the relative reduction of current (at +50 mV) compared with the control value as an index of channel block. The concentration dependence of block derived from 19 such observations (Fig 3C⇓) was fitted with a standard binding isotherm and yielded an EC50 of 0.71±0.08 μmol/L. The substitution with the smaller hydrophobic residue valine (T505V) enhanced the affinity to a lesser extent (EC50, 1.4 μmol/L; Table⇑) compared with the order of magnitude increase with T505I. The more conservative mutation T505S also enhanced the affinity, but to a lesser degree (EC50, 3.4 μmol/L; Table⇑).
If the S6 segment adopts an α-helical secondary structure in this region, then V512 should align better with T505 than L508 (Fig 1B⇑). We replaced V512 by isoleucine (found in Kv2.1) and by alanine to increase or decrease the size of the hydrophobic side chain, but V512I was poorly expressed. Fig 2⇑ shows that 2 μmol/L quinidine reduced currents from the V512A mutation by ≈75% in a time-dependent manner. From the concentration dependence of block (Fig 3C⇑), we derived an EC50 of 1.4 μmol/L, ie, about fivefold enhancement of the affinity.
Time and Voltage Dependence of Block
For a bimolecular reaction, the time course of block proceeds in a monoexponential manner, with a time constant τ as follows: τ=1/(k[D]+l), where k and l are the apparent rate constants for binding and dissociation, respectively, and [D] is drug concentration. Fig 3⇑, panels D and E, illustrate that the apparent rate of block (1/τ) displayed a linear concentration dependence for mutations T505I and V512A, similar to previous observations for block of hKv1.5 by quinidine and bupivacaine enantiomers.1 16 From the linear fit to these data, we obtained values for the apparent association and dissociation rate constants (Table⇑). An alternate approach is to combine the above equation with l/k=EC50, which allows extraction of rate constants k and l for each experiment separately. The Table⇑ shows that the values obtained with both independent methods were in close agreement.
The voltage dependence of hKv1.5 block by quinidine displays a biphasic voltage dependence with a steep phase coincident with the voltage dependence of channel opening and a shallow phase at potentials of >0 mV.1 By use of the Woodhull formalism,21 the latter can be used to gauge the electrical binding distance δ, ie, the fraction of the transmembrane electrical field sensed at the binding site. The preservation of the electrical binding distance was of particular interest in the T505I and V512A mutations, which significantly increased the affinity. Activation kinetics of T505I are similar to that of WT (Fig 3A⇑), but the currents deactivate more slowly (Fig 3F⇑), whereas V512A opens at more negative voltages (activation midpoint Eh, −42 mV). This allowed us to measure the fractional electrical distance over an extended voltage range (down to −70 mV for T505I), as illustrated in Fig 3G⇑ and 3H⇑. Positive to −20 mV, we used the fractional suppression of current during depolarization. In a second protocol, we first induced a block during a prepulse to +50 mV, followed by repolarization to potentials between −70 and −20 mV. Fig 3F⇑ shows that the tail current in the presence of 2 μmol/L quinidine displayed an initial rising phase, which settled in a quasi-steady current at −30 mV. Since the channel stayed open longer, the rising phase should reflect the relaxation to the new but lower level of block expected from the intrinsic voltage dependence of block. Fig 3G⇑ shows that the data obtained in this way could indeed be described by the same fractional distance over this extended voltage range.
Mutations C-Terminal to S6
Although the results for the mutations of threonine 505 are consistent with a hydrophobic QA-type block, the V512A mutation suggests that other residues may also be involved. We extended the mutagenesis to aromatic and charged residues in and just beyond the C-terminal end of S6. A deletion of the 57 carboxyl terminal amino acids (▵C57) resulted in a channel with functional properties similar to WT.22 This deletion did not abolish quinidine binding, as shown by the time-dependent decline of current induced by 6 μmol/L quinidine (Fig 2⇑). In fact, the apparent affinity was increased about twofold. Analysis of the binding kinetics indicated that the increased affinity resulted from an increased association rate constant, suggesting that the removal of this peptide segment improves diffusional access to the binding site (Table⇑). The topology of the highly charged cytoplasmic segment immediately distal to the S6 helix (Fig 1⇑) is unknown. It could potentially fold back in the pore or otherwise form a ring of localized negatively charged residues near the inner entrance of the pore and thereby influence the binding of cationic drugs. Therefore, we examined the effects of charge neutralizations E524Q+D526N and E528Q+E529Q. Both yielded functional channels without significant effects on the affinity for quinidine (Fig 2⇑, Table⇑), especially if one considers that these dual mutations each eliminate eight negative charges in the tetrameric channel. Similarly, the mutation H522G, which could eliminate a repulsive cationic interaction, had minimal effects on the affinity (Fig 2⇑, Table⇑). The potential contribution of the aromatic cluster at the cytoplasmic end of S6 (Fig 1A⇑) to a quinidine binding site was analyzed by eliminating aromatic residues individually or in combination. No functional expression was obtained with F517V, Y521N, or Y519N+F520V, although Y519N alone yielded functional hKv1.5-like current. Consistent with the poor tolerance for mutations in this region, the quadruple mutant FnYFY to VnNVN of the sequence 517 to 521 was nonfunctional as well. In the case of F517V, we confirmed the absence of functional protein with immunohistochemistry (see “Materials and Methods”). Block of the functional mutation Y519N by 6 μmol/L quinidine was similar to WT, with an EC50 of 7.3 μmol/L (Fig 2⇑, Table⇑).
Although these mutations did not modify the affinity for quinidine to a large extent, it was important to determine whether the open channel–block mechanism was preserved. Therefore, we routinely measured the apparent electrical binding distance δ, as illustrated for three mutations in Fig 4⇓. The Table⇑ shows that the values clustered in a narrow range (δ=0.18 to 0.23), similar to that of WT. Fig 2⇑ shows that the time dependence of block was preserved for most mutations; the Table⇑ lists the derived values for association and dissociation rate constants. If quinidine acts as an open-channel blocker, then the blocking reaction will compete with channel closure. Depending on the relative kinetics of both processes, this may result in an initial rising phase followed by a slowed decline of the deactivating tail current. Superposition with the current in control leads to a “crossover” phenomenon, which has been observed for hKv1.5 block by several drugs.1 10 16 Fig 5⇓ illustrates this effect for T505S, L508M, and Y519N. Similar results were obtained for H522G, S515E, T477S, and R485Y (not shown). With V512A the crossover was not observed, but the slow deactivation revealed the time-dependent relaxation to a lower level of block directly, as was the case for T505I (see Fig 3F⇑).
Mutations in the S4S5 Linker
The N-terminal “inactivation ball” in Shaker channels behaves like an open-channel blocker, presumably by occluding the internal mouth of the channel.23 The S4S5 linker has been proposed as its putative receptor site.24 Since the conservative mutation E395D in Shaker virtually abolished N-type inactivation,24 we examined whether the same substitution affected quinidine binding in hKv1.5. For E431D, we found a small reduction in the estimated affinity (EC50, 9.5±2 μmol/L; n=3), ie, a less than twofold change. If the interaction of quinidine with S6 residues involves hydrophobic interactions, then the cationic quinucleidine nitrogen might interact with residues in the S4S5 linker. Unfortunately, the charge reversal E431K to test this hypothesis led to a nonfunctional channel. Although the E431 residue is highly conserved between Kv1 isoforms, various residues are tolerated at the equivalent of Q426. The mutation Q426E did not significantly affect quinidine binding (Table⇑, Fig 2⇑), suggesting that the charged nitrogen is not in the vicinity of this residue.
Charged Residues Are Poorly Tolerated in S6
In a further attempt to influence the local electrostatic potential in or near the ion conducting pore, we made the following mutations: T505D, T505K, V512E, V512D, and S515E. These mutations should introduce a local electrostatic field that may be strong enough to eliminate permeation in a narrow pore.25 Indeed, no measurable current was detected with T505D, T505K, V512E, or V512D. It appears that hKv1.5 is intolerant to charged residues in this region, at least at these variable sites where uncharged residues are found in all Kv family channels cloned to date.8 Interestingly, S515E exhibited hKv1.5-type currents, activating at potentials ≈10 mV more negative than WT. As for WT, quinidine block of S515E was time and voltage dependent (Fig 2⇑), with an EC50 of 3.9 μmol/L. However, the time course of block at +60 mV was faster with time constants of 5 and 2.7 milliseconds for 6 and 20 μmol/L quinidine, respectively (n=2 each), resulting in larger values for both association and dissociation rate constants (Table⇑).
Lack of Effect of the Endogenous Kvβ2.1 Subunit
While this study was in progress, we discovered that the L cells express an endogenous Kvβ subunit that represents a mouse Kvβ2.1 homologue.26 This subunit does not induce fast N-type inactivation but has more subtle effects. In HEK293 cells, which do not express Kvβ subunits,26 hKv1.5 activates at potentials 10 to 15 mV more positive than those in L cells. Coexpression of hKv1.5 with mKvβ2.1 in the Kvβ-free HEK293 cells resulted in a hyperpolarizing shift of the activation curve, essentially reconstituting the L-cell phenotype.26 Since the presumably Kvβ-free oocyte-expressed Kv1.4 and Kv1.5 are fairly insensitive to quinidine (EC50, >200 μmol/L) (Reference 1212 and authors’ unpublished data, 1996), we wondered whether the high quinidine affinity of L-cell–expressed hKv1.5 could be due to the associated mKvβ2.1 subunit. Therefore, we tested the effects of quinidine on hKv1.5 stably expressed in HEK293 cells. Quinidine (6 μmol/L) acted as an open-channel blocker with a similar electrical distance (δ=0.21±0.02), similar kinetics (16±3 milliseconds at +60 mV), and similar levels of block (68±8% at +60 mV, n=3). These results demonstrate that hKv1.5 is sensitive to micromolar quinidine concentrations, irrespective of the absence or presence of the Kvβ2.1 subunit.
Our results indicate that mutations in the S6 segment of this Shaker-related cardiac K+ channel affect its affinity for quinidine. The fact that preliminary experiments with mammalian Kv channels of several Kv families had not revealed markedly distinct affinities for quinidine (at least when expressed in mammalian cells) suggested that the binding site would be contained in the common S1 to S6 segments. This was further supported by the fact that the deletion of over two thirds of the carboxyl-terminal end of hKv1.5 did not abolish quinidine binding, excluding this section as the binding site. These considerations, and the functional similarity between quinidine binding and that of internal TEA or other QA binding, led us to focus on the S4S5, P, and S6 segments. Taking the similarity of quinidine block to internal QA block as a starting point, we focused on residues in hKv1.5 that were equivalent to those that have been implicated in lining the open-channel pore, in binding of biophysical probes such as QAs or 4-AP, or in the binding site for the inactivating gate.7 17 24 27
Comparison of Quinidine Block With Internal TEA and QA Block
Although block of hKv1.5 by quinidine functionally resembles internal QA block, our results indicate that several Shaker mutations that affect internal TEA block do not affect quinidine block of hKv1.5. The L508M equivalent mutation in Shaker (L472M) and in Kv2.1 (L430M) reduced the affinity for internal TEA >10-fold and could account for the lower internal TEA affinity of Kv3.1, in which a methionine is found at this position.17 20 28 However, the quinidine affinity of hKv1.5 was not significantly affected by either mutation. On the other hand, Choi et al7 studied the effect of mutations of the T505 equivalent site in Shaker (T469) on the affinity for TEA and long-chain alkyl TEA derivatives. The conservative serine substitution at this site reduced TEA affinity 7-fold but reduced that of C8-TEA and C10-TEA ≈2-fold. We observed a minor increase in quinidine affinity with T505S. The internal TEA affinity in Shaker was minimally affected by valine or isoleucine substitution, but these mutations greatly enhanced C8-TEA and C10-TEA block (200-fold and 30-fold, respectively). The pattern of increased quinidine affinity of T505 mutations in hKv1.5 again indicates that quinidine binding does not mimic binding by TEA in the inner mouth of the channel. Quinidine binding is enhanced with increased hydrophobicity at this position, similar to the pattern observed for the highly hydrophobic long-chain TEA derivatives.7 The increase in quinidine affinity of hKv1.5 with the increased hydrophobicity at this position (Fig 6⇓) is consistent with the proposal that quinidine binding is stabilized by hydrophobic interactions.18 However, substitution of these hydrophobic residues enhanced quinidine binding less than that of the hydrophobic TEA derivatives. A possible explanation would be that interaction with other residues buffers the impact of mutations at the T505 site. Although we observed a 10-fold increase in affinity with T505I, none of the mutations abolished the binding of quinidine. This is similar to the results obtained for hydrophobic TEA analogues in Shaker7 and may be due to our inability to determine the affinity of nonfunctional channels with our electrophysiological approach. Since it is impractical to test large numbers of mutations at a substantial number of residues, the present results could be used to guide further attempts to reduce the quinidine affinity by focusing on the T505 position.
Potential Topology of the Binding Site
The alanine substitution for V512 increased the quinidine affinity fourfold. If the S6 segment is indeed a helix, then the 505 and 512 residues would be separated by ≈11 Å (two turns of the helix, Fig 6⇑), which would allow quinidine (12 to 14 Å) to interact with both residues. The limited effect of S515E on the affinity is consistent with such a location, since the distance between positions 505 and 515 would be too large for simultaneous interaction. In addition, if quinidine interacts with the 505 residue, then its positively charged quinucleidine nitrogen is apparently far enough removed from the 515 residue, since the S515E mutation had minimal effect on quinidine binding. Nevertheless, the increase in both binding and unbinding rate constants in the S515E mutation may indicate that the presence of the negative charge (presumably a ring of four charges) introduces an electrostatic field that would increase the local concentration but destabilize binding as well, presumably through long-range electrostatic interactions. Further attempts to localize a residue that would specifically interact with the charged nitrogen were not successful, since introduction of basic and acidic residues at the 512 and 505 positions eliminated functional expression.
Comparison With the Local Anesthetic Binding Site in Na+ Channels
In Na+ channels, the S6 segment of domain IV has been implicated in the binding of local anesthetics (lidocaine and etidocaine).29 Quinidine also acts as an open-channel blocker of cardiac Na+ channels,30 as do several QA derivatives.31 32 Residues at positions 1764 and 1771 (separated by two helical turns) have been proposed to form molecular determinants for local anesthetic binding in the Na+ channel. If the VIxxNF sequence (512 to 517) is used as a reference for alignment, then the Na+ channel residues 1764 and 1771 would correspond to hKv1.5 residues 502 and 508, ie, shifted one turn on the putative helix (Fig 6⇑). This apparently deeper binding in the Na+ channel pore is consistent with the larger fractional electrical distance (δ, ≈0.5 to 0.7) observed for local anesthetic binding in the Na+ channel compared with values obtained in Shaker-related channels (δ, ≈0.2).
The Difference in Quinidine Affinity Between Various Expression Systems Does Not Involve the β2 Subunit
We used a mammalian cell line to assess the effects of the mutations, since previous studies using two-electrode voltage clamp of Kv1.4 channels expressed in oocytes suggested that these channels poorly sensitize to this drug.11 12 The reason for the discrepancy between mammalian and amphibian expression systems is unclear but does not appear to be related to the absence or presence of a Kvβ subunit, because the affinity for quinidine of hKv1.5 expressed in HEK293 cells was similar to that observed in L cells. This result is also consistent with the fact that terfenadine block of hKv1.5 is similar whether it is expressed in HEK293 cells or L cells.33 34 Importantly, the sensitivity of Kv1.5 expressed in L cells corresponds closely to that of its putative homologue in native atrial myocytes (EC50, ≈5 μmol/L).35 Therefore, it seems logical to use the mammalian expression systems for mutagenesis studies involving these clinically used drugs to gain insight into the molecular determinants of block that might lead to further rational drug development.
A potential pitfall of site-directed mutagenesis is that the observed results may be derived from more generalized perturbations of the protein. We did notice differences in deactivation kinetics in several mutations (Figs 2⇑, 3⇑, and 5⇑), which in part resulted from shifts in the voltage dependence of channel opening.36 However, we could not discern a relationship between altered deactivation gating and the affinity for open-channel block at depolarized potentials; eg, V512A slowed deactivation more than T505I did but affected the (open-channel) affinity less. Nevertheless, we not only determined the apparent affinity but also (1) derived binding parameters, (2) determined the electrical binding distance δ, and (3) checked for drug-modified tail current kinetics as a qualitative indicator for open-channel block. The preserved open channel–block mechanism and the similarity of the binding distance suggest that no major conformational changes were introduced to the binding site. The similarity of the binding distances also indicates that the T505I mutation does not introduce a novel binding domain at a site that is measurably removed from the WT receptor. Analysis of the kinetics further indicates that in general the mutations that affected the affinity less than twofold (ie, change in binding energy, <300 cal/mol) did not affect binding or unbinding parameters (Table⇑). The exceptions were the S515E mutation, in which both were enhanced, and the ▵C57 mutation, in which the twofold affinity change resulted from an isolated increase in the on rate. The latter would suggest that removal of this section removes a diffusional restriction, allowing enhanced access to the receptor. A common pattern for the mutations that did increase the quinidine affinity significantly was that the dissociation rate constant was reduced consistently (T505S, T505V, T505I, and V512A). Furthermore, the increase in affinity was paralleled by the increased hydrophobicity at the 505 position. Both observations are in agreement with a local hydrophobic interaction at this residue, similar to the specific interaction proposed for C8-QA at the equivalent site in Shaker. Finally, the observation that the T505I mutation abolishes the stereoselective block of WT hKv1.5 by bupivacaine enantiomers is also consistent with a specific interaction of the T505 residue with these open-channel blockers.37
Taken together, these results localize the hydrophobic binding site for this antiarrthythmic drug in the internal mouth of this human K+ channel and provide molecular support for the open channel–block model of quinidine and the role of S6 in contributing to the inner pore.
Selected Abbreviations and Acronyms
|PCR||=||polymerase chain reaction|
This study was supported by National Institutes of Health grants HL-47599 and HL-46681. The authors thank Dr Dan Roden for review of the manuscript and Dr Archana Chaudhary, Holly Shear, Debbie Mays, and Ian Hopkirk for technical support.
Previously published as preliminary results in abstract form (Biophys J. 1996;70:A398).
- Received March 8, 1996.
- Accepted April 12, 1996.
- © 1996 American Heart Association, Inc.
Snyders DJ, Knoth KM, Roberds SL, Tamkun MM. Time-, voltage-, and state-dependent block by quinidine of a cloned human cardiac potassium channel. Mol Pharmacol. 1992;41:322-330.
Mays DJ, Foose JM, Philipson LH, Tamkun MM. Localization of the Kv1.5 K+ channel protein in explanted cardiac tissue. J Clin Invest. 1995;96:282-292.
Wang Z, Fermini B, Nattel S. Sustained depolarization-induced outward current in human atrial myocytes: evidence for a novel delayed rectifier K+ current similar to Kv1.5 cloned channel currents. Circ Res. 1993;73:1061-1076.
Fedida D, Wible B, Wang Z, Fermini B, Faust F, Nattel S, Brown AM. Identity of a novel delayed rectifier current from human heart with a cloned K+ channel current. Circ Res. 1993;73:210-216.
Armstrong CM. Interaction of tetraethylammonium ion derivatives with the potassium channels of giant axons. J Gen Physiol. 1971;58:413-437.
Chandy KG, Gutman GA. Voltage-gated potassium channel genes. In: Peroutka SJ, ed. Handbook of Receptors and Channels. Boca Raton, Fla: CRC Press; 1994:1-71.
Dougerthy DA. Cation-π interactions in chemistry and biology: a new view of benzene, Phe, Tyr and Trp. Science. 1996;271:163-168.
Yatani A, Wakamori M, Mikala G, Bahinski A. Block of transient outward–type cloned cardiac K+ channel currents by quinidine. Circ Res. 1993;73:351-359.
Snyders DJ, Tamkun MM, Bennett PB. A rapidly activating and slowly inactivating potassium channel cloned from human heart: functional analysis after stable mammalian cell culture expression. J Gen Physiol. 1993;101:513-543.
Baukrowitz T, Yellen G. Use-dependent blockers and exit rate of the last ion from the multi-ion pore of a K+ channel. Science. 1996;271:653-656.
Snyders DJ, Yeola SW. Determinants of antiarrhythmic drug action: electrostatic and hydrophobic components of block of the human cardiac hKv1.5 channel. Circ Res. 1995;77:575-583.
Yellen G, Jurman ME, Abramson T, MacKinnon R. Mutations affecting internal TEA blockade identify the probable pore-forming region of a K+ channel. Science. 1991;251:939-942.
Woodhull AM. Ionic blockade of sodium channels in nerve. J Gen Physiol. 1973;61:687-708.
Hille B. Ionic Channels of Excitable Membranes. 2nd ed. Sunderland, Mass: Sinauer Associates; 1991.
Uebele VN, England SK, Chaudhary AC, Tamkun MM, Snyders DJ. Functional differences in Kv1.5 currents expressed in mammalian cell lines are due to the presence of endogenous Kvβ2.1 subunits. J Biol Chem. 1996;271:2406-2412.
Hartmann HA, Kirsch GE, Drewe JA, Taglialatela M, Joho RH, Brown AM. Exchange of conduction pathways between two related K+ channels. Science. 1991;251:942-944.
Ragsdale DS, McPhee JC, Scheuer T, Catterall WA. Molecular determinants of state-dependent block of Na+ channels by local anesthetics. Science. 1994;265:1724-1728.
Snyders DJ, Hondeghem LM. Effects of quinidine on the sodium current of ventricular guinea pig myocytes: evidence for a drug-associated rested state with altered kinetics. Circ Res. 1990;66:565-579.
O’Leary ME, Horn R. Internal block of human heart sodium channels by symmetrical tetra-alkylammoniums. J Gen Physiol. 1994;104:507-522.
Rampe D, Wible B, Brown AM, Dage RC. Effects of terfenadine and its metabolites on a delayed rectifier K+ channel cloned from human heart. Mol Pharmacol. 1993;44:1240-1245.
Wang Z, Fermini B, Nattel S. Effects of flecainide, quinidine, and 4-aminopyridine on transient outward and ultrarapid delayed rectifier currents in human atrial myocytes. J Pharmacol Exp Ther. 1995;272:184-196.
Snyders DJ, Rich TC, Mays DJ, Tamkun MM, Yeola SW. Involvement of S6 in voltage-dependent deactivation gating of hKv1.5 potassium channels. Biophys J. 1996;70:A144. Abstract.
Franqueza L, Delpón E, Gay P, Snyders DJ, Tamargo J, Valenzuela C. Molecular determinants for stereoselective bupivacaine binding in hKv1.5 channels. Biophys J. 1996;70:A400. Abstract.