Allosteric coupling of the inner activation gate to the outer pore of a potassium channel

In potassium channels, functional coupling of the inner and outer pore gates may result from energetic interactions between residues and conformational rearrangements that occur along a structural path between them. Here, we show that conservative mutations of a residue near the inner activation gate of the Shaker potassium channel (I470) modify the rate of C-type inactivation at the outer pore, pointing to this residue as part of a pathway that couples inner gate opening to changes in outer pore structure and reduction of ion flow. Because they remain equally sensitive to rises in extracellular potassium, altered inactivation rates of the mutant channels are not secondary to modified binding of potassium to the outer pore. Conservative mutations of I470 also influence the interaction of the Shaker N-terminus with the inner gate, which separately affects the outer pore.

In potassium channels, functional coupling of the inner and outer pore gates may result from energetic interactions between residues and conformational rearrangements that occur along a structural path between them. Here, we show that conservative mutations of a residue near the inner activation gate of the Shaker potassium channel (I470) modify the rate of C-type inactivation at the outer pore, pointing to this residue as part of a pathway that couples inner gate opening to changes in outer pore structure and reduction of ion flow. Because they remain equally sensitive to rises in extracellular potassium, altered inactivation rates of the mutant channels are not secondary to modified binding of potassium to the outer pore. Conservative mutations of I470 also influence the interaction of the Shaker N-terminus with the inner gate, which separately affects the outer pore. I on flow through voltage-gated (Kv) potassium channels is limited by at least two mechanisms taking place at opposite ends of the ion permeation pore. An inner gate is formed by the intracellular portion of the S6 helices of the four subunits coming together in a ''bundle crossing'', which opens in response to membrane depolarizations to allow ions to flow [1][2][3][4] . Another gate is found at the outer end of the pore, which switches between a conducting and non-conducting conformation. Rearrangement of the outer gate, also known as the selectivity filter for its role in accommodating potassium and promoting its throughput over sodium, takes place following opening of the inner gate, and the two processes are functionally coupled; the closed conformation of the intracellular pore stabilizes the conducting conformation of the selectivity filter whereas an open inner conformation favors a modified filter structure that is accompanied by a decay of current, referred to as C-type inactivation [5][6][7][8] . Furthermore, inner gate opening is promoted by the non-conducting conformation of the selectivity filter 9 .
Energetic interactions between residues and associated conformational rearrangements along a structural path between the inner and outer pore may underlie their coupling [10][11][12] and has been proposed to explain a set of high resolution crystal structures obtained from the potassium channel KcsA, from Streptomyces lividans 13 . KcsA is gated mainly by protons, rather than by voltage, but the proton-activated currents decay over time in a manner that is similar to C-type inactivation in Kv channels. A continuum of intermediate conformations of the aperture of the inner gate is correlated with reorientation and a decrease in potassium occupancy of the selectivity filter 13,14 . Also identified was a residue that lies between the inner pore lining helix and the backbone amino acids of the selectivity filter 15 . Snapshots of this residue, F103, show that it is rotated as the gate widens, and lies closer to the selectivity filter, which is concomitantly reoriented into a conformation with a lower capacity for potassium ions. One appealing interpretation of the correlated differences in structure and loss of potassium binding in the KcsA channel is that opening of the inner gate re-orients F103, which then physically interacts with the selectivity filter to modify it and limit conduction. Such an interpretation of the observed structures is supported by conservative mutations of F103 that predictably modify the decay of KcsA-mediated current. Cuello et al 15 also provide two pieces of evidence that suggest a similar sequence of molecular events upon opening and role for a F103equivalent residue, an isoleucine in Shaker-related channels. First, using the Kv1.2 crystal structure as a starting point, they show strong theoretical energetic interactions of I402 with a similar subset of residues in the selectivity filter (see Figure 1). Second, they show that alanine substitution of the F103-equivalent residue, I470, in the Ntype inactivation-removed Shaker (Shaker-IR) reduces the rate at which current decays; this is consistent with previous studies showing that the I470C mutation in Shaker-IR can result in a strong reduction in current decay 16,17 .
Based on the aforementioned evidence, Cuello et al 13,15 suggested that opening-induced perturbations in structure observed for KcsA promote inactivation in all Kv channels, but this provocative proposal remains uncertain for several reasons. First, it is not known if the array of solved structures of the KcsA channel identified is truly representative of different stages of the inactivation process or, even if it is, whether such a structural series of events is common among related channels 18 . Second, the slowed inactivation of Shaker-IR produced by alanine substitution of I470 may be secondary to altered coordination of potassium as this cation is known to bind to the outer pore and slow C-type inactivation 19,20 . For example, a mutation in the pore of Shaker, A463C, has been suggested to slow C-type inactivation by enhancing the occupancy of potassium in the outer pore 21,22 . Third, the events proposed by Cuello et al may not apply to the full-length Shaker channel, where ion flow is reduced rapidly and dramatically upon opening and the inner gate is impacted by its own association with the full N-terminus 23,24 . The N-terminus blocks current by binding within the inner pore cavity, which accelerates inactivation; this may be due to a reduction of potassium ions residing in the selectivity filter and to allosteric signalling from an inner gate that has difficulty in closing, both of which promote the non-conductive conformation of the selectivity filter 6 .
To test the hypothesis that the archetypal potassium channel, Shaker, undergoes C-type inactivation as proposed for KcsA, we combined mutagenesis, recording of ion flux by two microelectrode voltage clamp and tracking of conformational changes in the selectivity filter by monitoring a fluorescent reporter attached to an outer pore residue. By examining the impact of conservative mutations of * * * * * I470 on slow inactivation and showing that the mutant channels remain sensitive to extracellular potassium, we provide critical support for a model in which this residue undergoes a re-orientation as the inner gate opens and widens, physically impinging on the selectivity filter to change its conformation and limit conduction, as has been suggested for F103 in the KcsA channel. We also suggest that I470 interacts functionally with the long N-terminus to control affinity of its interaction with the inner side of the channel and speed Ctype inactivation.

Results
A set of conserved residues identifies a putative coupling pathway that is common among Shaker-related potassium channels. Previously, calculations of interaction energies were carried out showing that a set of residues in Kv1.2 forms a pathway that couples the inner and outer pore gates as in KcsA (Figure 1) 15 . Isoleucine 402 (F103 in KcsA) was shown to make strong van der Waals contacts with threonine 373 and threonine 374 (T74, T75 in KcsA) in the P-helix portion of the outer pore. Additionally, phenylalanine 103 of KcsA makes a more subtle interaction with methionine 96, which corresponds to alanine 396 in Kv1.2, and a strong interaction with isoleucine 100, which corresponds to valine 399 in Kv1.2. The residues of Kv1.2 mentioned above are completely conserved with those of Shaker ( Figure 1).
Conservative substitutions at site I470 of Shaker modify C-type inactivation rate but maintain its sensitivity to extracellular potassium. To examine the coupling of the inner gate to the outer pore, we carried out mutagenesis of the I470 residue of the Shaker potassium channel in combination with electrophysiological approaches to quantify the rates of C-type inactivation. We began by studying the inactivation-removed Shaker channel, -IR, in which a portion of the distal N-terminus is removed; this eliminates a fast component of current decay and induced effects on C-type inactivation and the outer pore. Single conservative substitutions of I470 in Shaker-IR were introduced and the rates of current inactivation were determined. Inactivation was measured by pulsing oocytes expressing Shaker-IR from a holding potential of -80 mV to a test potential of 160 mV ( Figure 2A); this test potential was chosen to maximally activate the channels and prevent any effect on current decay from the voltage-dependence of activation. That all of the mutant channels were not activated at 280 mV and maximally activated at 160 was confirmed from conductancevoltage (G-V) relationships that were determined for all constructs used in this study (data not shown). When oocytes injected with constructs with conservative mutations of I470 to leucine ( Figure 2B), valine (2C), cysteine (2D) and phenylalanine (2E) were given long depolarizing pulses, all showed altered rates of inactivation compared to the wild type construct (Figure 2A). Shaker-IR I470V inactivation traces required a double exponential function for adequate fitting. We treated the faster, K 1dependent time constant as the likely kinetic correlate of the process observed with the remaining constructs and used it for calculations and analysis. As found for KcsA 15 , smaller side chains generally showed greater impairment of inactivation.
The rate of C-type inactivation in Kv channels such as Shaker is sensitive to extracellular potassium, which occupies the outer pore and reduces this rate 19,20 . It is possible that the effects of the mutations to I470 on C-type inactivation are secondary to modified occupancy of potassium to the outer pore. To examine this possibility, we determined the rates of inactivation using an extracellular solution containing 99 mM potassium rather than 3 mM, in the same oocytes. With the higher concentration of extracellular potassium, the time course of inactivation was slowed in all I470 mutant constructs as compared to that made in the normal extracellular potassium concentration. When inactivation rates in normal and elevated extracellular potassium were plotted against each other, regression analysis demonstrated a linear relationship among the channels used indicating that the effect of raising extracellular potassium from 3 mM to 99 mM rates is preserved among them ( Figure 2F).
The effects of I470 mutations can be compared and contrasted with those of A463C Shaker-IR, a mutation of alanine 463 in the pore (Figure 1) that also modifies (reduces) the rate of C-type inactivation 21,22 . However, unlike those for the I470 mutants, the tendency of C-type inactivation to be slowed by increasing the extracellular potassium concentration was not preserved for  Shaker-IR A463C across a wide range of tested concentrations (Supplementary Figure 1), and in fact appeared to be slightly accelerated in the higher potassium concentration. The lack of sensitivity to extracellular potassium of this mutant is consistent with previous findings suggesting that slowed rate of C-type inactivation is due to elevated potassium occupancy of the outer pore, which we observed even with an extracellular solution containing a normal level of potassium, rather than to an effect on the conformational change that leads to C-type inactivation 21,22 .
The relative effects of conservative substitutions of I470 in fulllength Shaker are similar to those in Shaker-IR and remain sensitive to potassium. We next examined the full-length Shaker channel to determine if the interaction between I470 and the outer pore is retained even with the long N-terminus and the very fast decay in ion flow that occurs upon depolarization of the membrane potential. Unlike in Shaker-IR, where C-type inactivation can be observed directly from current decay, full-length Shaker exhibits more complex current decay due to the co-existence of fast N-terminal block and slow pore rearrangement. Therefore, in addition to extracting the rate of slow inactivation from fitting of complex ionic current decay, we used full-length Shaker S424C, and performed voltage-clamp fluorimetry in conjunction with two electrode voltage clamp to more easily and accurately track conformational changes in the outer pore that follow C-type inactivation 25 . When current decay and fluorescence traces from Shaker-FL S424C were fit to exponential functions, the time constant of fluorescence decay corresponded well with a slower time constant of current decay ( Figure 3A).
Next, the residue T449, located just C-terminal to the GYG triplet of the selectivity filter and positioned extracellularly in the Kv1.2 3D structure (see alignment of Shaker-related channels in Figure 1), was mutated to valine (T449V) in Shaker-FL S424C. Mutation of T449 in Shaker antagonizes outer pore rearrangement and, like A463C, is thought to enhance potassium occupancy in the outer pore 20 . When Shaker-FL S424C T449V was depolarized, rapid N-type inactivation signal was still observed, but current decay was far less profound and the rapid transient was followed by a much more slowly decaying second phase ( Figure 3B), compared to Shaker-FL S424C. Fluorescence signals recorded from Shaker-FL S424C during depolarization were almost completely abolished in Shaker-FL S424C T449V. Taken together, these results show a dramatic reduction of outer pore rearrangement by T449V and show that this alteration in outer pore structure is tracked accurately by S424C fluorescence.
The rates of change in fluorescence were measured in channels containing the same series of conservative mutations at site I470 shown in Figure 2 and were slowed by valine, cysteine and phenylalanine and sped up by leucine ( Figure 3C-F). When the same experiments were conducted with an elevated level of extracellular potassium, a slowing of S424C fluorescence kinetics corresponding in time with C-type inactivation was observed for all mutant channels. As with wild type, current decay was fit with multiple exponentials, and compared to the fluorescence traces; time constants from slow current decay rates and fluorescence decay rates corresponded well ( Figure 3G). The current decay of Shaker-FL I470L was rapid and profound ( Figure 3C) and it was not possible to reliably extract rates of current decay that corresponded to a separate slow exponential process. Nevertheless, the data show that the changes in fluorescence, and the underlying conformational changes in the outer pore, are tightly associated with slow inactivation of ionic current.
When S424C fluorescence rates in high extracellular potassium were plotted against those in normal potassium for all constructs tested, regression analysis once again demonstrated linearity ( Figure 3H), with a slope showing sensitivity to extracellular potassium across all mutants. Superimposition of the line obtained from the same experiments using the N-terminally deleted channel shows that the slopes are close to identical. The rank order of the rates of inactivation for the Shaker-FL mutants is similar, but not identical to the corresponding Shaker-IR mutant channels (compare Figure 2F with Figure 3H). Like the Shaker-IR mutants, Shaker-FL I470C and I470L yield the fastest and slowest rates; however, I470V renders a faster rate than I470F, which is now faster than wild type (I470). Nevertheless, the Shaker-FL channels remain sensitive to potassium, favouring our hypothesis that I470 controls coupling in both short and long forms of the Shaker channel without greatly impacting potassium sensitivity. Thus, while the N-terminus greatly accelerates the overall rates of C-type inactivation 6 , it does not disrupt coupling of the 470 side-chain with the outer pore.
Conservative substitutions of I470 modify the interaction of the N-terminus with the inner pore of the full-length Shaker channel. Although elevating extracellular potassium slowed the rates of Ctype inactivation of Shaker-IR channels and changes in fluorescence of Shaker-FL channel to a roughly equal extent, the latter were always faster than the former from the matching Shaker-IR mutant regardless of extracellular potassium concentration. The faster rates of full-length Shaker are probably due to an effect of the N-terminus as it interacts with the inner pore 6 . The I470 residue is among those that have been proposed to strongly and directly interact with the N-terminus it resides in the mouth of the potassium channel pore 14 .
To compare the degree of acceleration of outer pore rearrangement by the N-terminus, we re-plotted data from Figures 2 and 3 ( Figure 4A) to compare rates with (Y-axis) and without (X-axis) the presence of the N-terminus at normal and high concentrations of extracellular potassium. We noticed that the degree of acceleration induced by the N-terminus was consistent between I470 and I470L (black line), whereas it was greater for I470F (above the black line) and smaller for I470C and I470V (below the black line). While I470F inactivated relatively slowly in the absence of the N-terminus, the intact N-terminus had a stronger effect on inactivation than it did on I470V.
To examine if the effect of the N-terminus was related to the strength of its association with the inner pore, the off-rate of the former was estimated using a two pulse protocol, with the Shaker T449V mutant to hinder conformational changes of the selectivity filter; thus, the main contributor to the rapid decay in current after opening is the N-terminus interaction with the inner pore, from which the off-rate can be reasonably estimated.
Brief depolarizing pulses were applied to cells expressing full length T449V channels with the same substitutions at I470. Subsequently, cells were given hyperpolarizing pulses of increasing duration, followed by a second depolarizing pulse to measure the peak amplitude ( Figure 4B). With increasing inter-pulse durations, the peak amplitudes during the second pulse recovered to the levels of those of the first pulse. The amplitudes of the second peaks for all constructs could all be fit to single exponentials, suggesting that the recovery from block was rate-limited by a single gating transition that reflected the unbinding of the N-terminus; thus, these rates are inversely proportional to the affinity of the peptide for the inner cavity ( Figure 4C).
We also calculated the slopes of regression lines between full length and truncated constructs shown in Figure 4A for the remaining mutants falling above or below the solid line ( Figure 4A). We noticed that the stronger the affinity of the peptide for the inner cavity correlated with a comparatively stronger accelerating effect of the N-terminus on the rate of fluorescence decay and reflected C-type inactivation ( Figure 4D). Specifically, the binding of the Nterminus to I470F was also stronger than for I470V; we suspect that this speeds the rate of slow inactivation to a greater extent in the former, leading to the altered rank order of inactivation rates of the mutant Shaker-FL channels in Figure 3H. Discussion Conservative mutations of Shaker I470 modify the rate of C-type inactivation without affecting its sensitivity to extracellular potassium when its concentration is elevated. The retention of wild type potassium sensitivity amongst all I470 mutants tested suggests that altered inactivation rates of the mutants are likely not secondary to altered coordination of this cation by the outer pore at these concentrations. Together with previous studies 13,15 , our findings support  a model in which I470 constitutes an initial part of an allosteric pathway that undergoes a re-orientation as the inner gate opens and widens, impinging upon the outer pore to change its conformation and limit conduction, as has been suggested for F103 in the KcsA channel based on functional and structural information. Our data are consistent between the full-length and inactivation-removed Shaker channels; thus, the allosteric signal emanating from the inner pore and I470 is also important for conformational rearrangement of the outer pore in the former. The effects of I470 mutants contrast with that of A463C, also near the inner activation gate, which slowed the rate of inactivation in a manner that was insensitive to changes in concentration of extracellular potassium. Insensitivity of A463C to extracellular potassium is consistent with previous findings suggesting that its slowed rate of Ctype inactivation is due to permanent occupancy of the outer pore by potassium and not to an effect of the mutation on the conformational change leading to inactivation 21,22 . The mutation T449V also slows inactivation and removes sensitivity to potassium and, when combined with the mutation I470C, almost completely inhibits slow inactivation 16,17 . Although mutations of I470 may have a more subtle influence on the binding affinity of potassium, they do not alter the maximum effect of this cation, unlike mutations at A463 and T449, supporting a role for the former residue in slow inactivation that does not directly impact potassium occupancy.
Unlike the KcsA channel, the Shaker channel also has a long Nterminus that is responsible for fast inactivation and speeds slow Current and fluorescence decay rates plotted in Figures 2 and 4 are re-plotted against each other to compare inactivation rates in full length Shaker (y-axis) and Shaker-IR (x-axis) with the equivalent I470 mutation in the presence of either 3 mM (squares) or 99 mM extracellular potassium (triangles). The black lines represent regression line fits for I470 and the mutant constructs. (B). Ionic currents measured in response to a two pulse protocol used to calculate the off-rates of the N-terminus from its blocking site in 3 mM (top) and 99 mM K1 (bottom). Following a pulse to 160 mV, cells were returned to -100 mV for an increasing period, then re-pulsed to 160 mV. The amplitudes of the second peaks were fit to a single exponential function (dashed lines). (C). Recovery rates from N-type inactivation were calculated from single exponential fits from panel B using Equation 2, and are plotted for the I470 and the four studied mutants. I470F rates were too slow to be visible on the y-axis time scale (n 5 6 for I470; n 5 5 for I470V, I470C and I470F; n 5 4 for I470L). (D). Regression lines are fit to data from each construct from panel A (see inset). The slopes of these curves are then plotted against rates calculated in panel C for each equivalent I470 mutation in the presence of either 3 mM (squares) or 99 mM extracellular potassium (triangles).
www.nature.com/scientificreports inactivation. A comparison of rates of C-type inactivation between full-length Shaker and Shaker-IR suggests that the N-terminus, as well as inhibiting potassium entry into the pore, interacts with the inner pore to separately promote conformational changes in the selectivity filter 6 . These data suggest that the effect of the Nterminus on slow inactivation is correlated with the strength of its association with the inner pore, which, in turn, is influenced by I470. Thus, this residue plays a dual role in the control of slow inactivation in Kv channels that possess a long N-terminus and fast inactivation. Lastly, in contrast to the KcsA channel, Shaker-related channels possess a voltage-sensing domain that could interact with the outer pore to impact slow inactivation; this is supported by studies showing that changes in voltage sensor and pore conformation by tracking changes in fluorescence correlate in time with C-type inactivation 25 . However, a more recent analysis suggests that voltage-sensor movement is faster than C-type inactivation in Shaker and can be modified independently from the pore; furthermore, a protein from Ciona intestinalis that contains a potassium-like voltage sensing domain but lacks a pore undergoes conformational changes during depolarization that are similar to those of the Shaker voltage sensor 26 . Together, the data suggest that the outer pore and voltage sensing elements of Shaker do not directly interact and influence each other during prolonged depolarization.