On the structural basis of modal gating behavior in K(+) channels.

Modal-gating shifts represent an effective regulatory mechanism by which ion channels control the extent and time course of ionic fluxes. Under steady-state conditions, the K(+) channel KcsA shows three distinct gating modes, high-P(o), low-P(o) and a high-frequency flicker mode, each with about an order of magnitude difference in their mean open times. Here we show that in the absence of C-type inactivation, mutations at the pore-helix position Glu71 unmask a series of kinetically distinct modes of gating in a side chain-specific way. These gating modes mirror those seen in wild-type channels and suggest that specific interactions in the side chain network surrounding the selectivity filter, in concert with ion occupancy, alter the relative stability of pre-existing conformational states of the pore. The present results highlight the key role of the selectivity filter in regulating modal gating behavior in K(+) channels.

INTRODUCTION

Potassium channels are ubiquitous membrane proteins with a fundamental role in generation and modulation of the electrical excitability in cells1. Channel function is finely controlled by the interplay between activation gating at the stimulus-driven bundle crossing, and C-type inactivation gating at the selectivity filter2. Recent high resolution crystallographic analyses have provided atomic level details of K+ channels trapped with the activation gate in the closed3-5 or open conformation6-9 and the inactivation gate in the conductive or non-conductive conformations5,9.

A series of crystal structures of KcsA trapped in various degrees of gate opening and ion occupancy9 have shown that entry to the C-type inactivated state is associated with a sequential reduction in ion occupancy at the S2 and S3 binding sites correlated to the extent of opening at the inner bundle gate. Although these structures provide unique insights into the basic structural transitions underlying the K+ channel gating cycle, a cursory look at most single-channel recordings reveal that even the simplest ion channels exhibit kinetically complex behaviors beyond the present set of structures. This functional heterogeneity involves conductive and non conductive states10-13 as well as a variety of sub conductance levels12,14-17 for which there are no current structural correlates. We have provided evidence showing that under saturating stimulus conditions and at steady-state (when the activation gate is in its fully-open conformation), most of these gating fluctuations arise from conformational changes at the selectivity filter11,12,18,19. Not unexpectedly, the nature of the permeant ion20-24 as well as a variety of mutations near the filter17,25-28 have been shown to dramatically modulate the frequency and lifetimes of these gating events. In many cases, the structural consequences of these perturbations are reflected in changes in the ionic occupancy at the filter, which partly explains the divergence from normal functional behavior26,29.

Modal gating appears to be a characteristic feature of many K+ channels, where time-dependent single-channel activity can switch abruptly between periods of high and low open probability under fixed experimental conditions30-32. In several channels, differential inactivation rates underlie some of these gating regimes31-34. The KcsA selectivity filter and adjacent regions display a considerable amount of conformational flexibility, as revealed from a comparison of existing KcsA structures in high and low K+5, partial and fully-open states9, in the presence of blockers35 and in the (so-called) “flipped” structure observed in the E71A mutant11. It is easy to speculate that this intrinsic structural flexibility might underlie some of the heterogeneous functional behavior of the selectivity filter that leads to multiple gating modes 12.

Here, we have probed the functional and structural origins of modal gating in KcsA by studying a series of side-chain substitution at Glu71 position. These mutations sharply reduce entry into the C-type inactivated state while stabilizing three kinetically defined gating modes, depending on the type of side chain at position 71. These gating modes are reminiscent of those seen in wt KcsA, based on their distinct intra-burst open probability (P), and were named high-Po mode (for mutants E71A/G/C/V/S/T), low-Po mode (E71I) and Flickery mode (E71Q). High-resolution closed-state crystal structures of some of these mutations, together with molecular dynamics (MD) simulations reveal changes in the ion profiles and water occupancy in and around the selectivity filter. These observations provide an initial rational to the origins of the conformational fluctuations occurring at the selectivity filter of the open-conductive channel.

RESULTS

Variable modal kinetic behavior of wt KcsA

At steady-state, and under saturating proton concentrations, KcsA predominantly resides in the non-conductive C-type inactivated state11,12. These long silent periods are interrupted by brief sojourns into the conductive conformation, before transitioning back to the non-conductive inactivated state11 (Fig. 1a). KcsA has been reported to exhibit a highly variable single channel kinetic behavior, displaying at least three distinct patterns or “modes” of gating12 (Fig. 1b). These modes are characterized by set variations in mean open and mean closed times: a high-Po mode with long open times τo ~100 ms, a low-Po mode with intermediate open times τo ~ 10 ms and a Flickery mode characterized by very short open times τo < 1 ms. The distribution of these modes is found to be random in nature with no obvious evidence for pH or voltage dependence. The proportion of time spent in the individual modes varies greatly from one patch to another; however the predominant mode is that of the high-Po. Steady-state single-channel recordings show that transitions between modes can take place within the same burst (Fig. 1c), supporting the view that modal behavior in KcsA results from the conformational heterogeneity of individual channels. However, these transitions are infrequent and modal changes are typically observed after sojourns into the C-type inactivated state.

Figure 1

Modal gating behavior of wt KcsA. (a) A continuous recording of KcsA single-channel currents measured under steady-state conditions at pH 3.0 and + 150 mV in 200 mM symmetric K+ solutions. (b) KcsA displays a highly variable kinetic behavior which arises from a combination of three distinct modes of channel activity, the high-Po, low-Po and the flickery mode (Left). Histograms show a distribution of open times within bursts for each of the three modes of channel activity with mean open times indicated in parenthesis (Right). (c) Channels occasionally switch between modes within a burst of activity, suggesting that modes arise from a homogenous population of channels.

Mutations at position 71 stabilize diverse gating regimes

While in voltage-dependent channels modal gating has been associated with specific biochemical modifications36,37, the origin of KcsA’s kinetic heterogeneity and sub-conductance levels has remained unclear. Mutations near the selectivity filter have shown to substantially reduce this variability. In particular, the C-type inactivation-removing mutation E71A also unmasks a kinetically homogeneous high-Po behavior, indistinguishable from the high-Po gating mode seen in wt KcsA11,12. We therefore carried out an in-depth analysis of the role of different side-chain substitutions at position 71 on the steady-state single channel kinetics of KcsA. Out of a total fifteen substitutions, nine mutations: Ala, Cys, Thr, Ser, Val, Ile, Gln, His and Gly were well tolerated. Mutations to Arg, Lys, Leu, Asp, Asn and Phe severely compromised channel folding and stability and were not analyzed further.

Analysis of macroscopic currents from the nine functional Glu71 mutants revealed that while there were no major effects on the time course of activation gating, all mutations directly affect the stability of the C-type inactivated state (Fig. 2a). As shown earlier11,38, E71A eliminates C-type inactivation, E71H severely enhanced it, and E71S stabilized an intermediate level. Mutants E71C, E71I, E71V, E71T and E71Q also slow down C-type inactivation with steady-state Po larger than 0.5 (Fig. 2b). The steady-state single channel activity faithfully reflected each of the mutants macroscopic behavior, where the long silent periods characteristic of wt KcsA recordings (>100 ms), were mostly absent (Fig. 2c).

Figure 2

Glu71 mutants stabilize individual gating modes in a side-chain specific way (a) Macroscopic responses of wt and various Glu71 mutants elicited by pH jumps from 8.0 to 4.0 using a rapid solution exchanger in the presence of 200 mM KCl and the membrane potential held at +150 mV. The current trace for the E71G mutant is shown at a relative amplitude, compared with the other traces, the inset shows the same trace expanded in the current axis. (b) A plot of Isteady/ Ipeak for various Glu71 mutants (n>5) (c) Single-channel currents were recorded under steady-state conditions at pH 4.0 and +150 mV in 200 mM symmetric K+ solutions. Grey box highlights mutants that are focused in this study. (d) Selectivity versus Na+ estimated from single-channel I-V ramps under bi-ionic conditions. No detectable Na+ currents were seen in any of the mutants. Eapparent is the potential at which K+ currents can last be resolved. Error bars show s.d (n>5)

However, the key observation from this set of mutants is that, besides slowing down C-type inactivation, these side-chain substitutions displayed unique intra-burst kinetic patterns arising from differences in the duration of opening and closing dwell times. Substitution to Ala, Gly, Cys, Thr, Val and Ser resulted in a common phenotype with long opening bursts and few short intra-burst closures. Substitution to Ile displays very homogenous kinetics with similar mean open and closed times, while Gln leads to a sustained, high frequency flickering behavior with very short open and closed sojourns. Interestingly, substitution to Gly also reduced the single channel conductance by factor of 10 (to 1.2 pA), a feature that was also reflected in the small amplitude of its macroscopic currents.

Overall, mutations that slowed down inactivation displayed a very homogenous kinetic behavior (Supplementary Fig. 1). All of the tested Glu71 mutants were fully selective to K+ against Na+ under bi-ionic conditions (Fig. 2d). Given their kinetic homogeneity and the obvious similarities to the pre-existing gating modes in wt-KcsA, we chose the mutants that best represented each type of kinetic mode for further analysis: E71A for the high-Po mode, E71I for the low-Po mode and E71Q for the Flickery mode.

Kinetic analysis of Glu71 mutants

The high open probability of E71A, E71I and E71Q (>0.7) allowed us to easily target recordings arising from a single active channel. This way, experiments carried out at different proton concentrations confirmed that the effect of increasing pH lies mostly in decreasing the burst length, as a result of the closure of the activation gate, with no major effects on the behavior of the burst itself. This is a clear indication that the transitions within the burst fully reflect the conformational fluctuations at the selectivity filter (Supplementary Fig. 2).

The rate constants to and from the conductive state for the various gating mode mutants were determined by fitting open and closed dwell times distributions to a model with one conductive and three nonconductive states (Fig. 3). These three non conductive states were defined as the “Slow” inactive state (Is), with dwell times (τc) around 100 ms; an Intermediate Inactive state (Ii) having dwell times between 1-10 ms; and the Flicker (F) state with dwell times in the 0.1-0.5 ms range. This classification is based on the lifetimes from single wt KcsA patches11,12. The behavior of the severely C-type inactivated mutant E71H was also analyzed to set a kinetic baseline for the transitions from a fully inactivated state. In Figure 3 (right panels) the kinetic schemes and the shaded region show the predominant transition in each of the mutants. Our key observation is that each mutation is associated with changes in the rate constants governing transitions to and from each of the states that define modal gating in KcsA. Thus, while the E71Q mutation mostly affects transitions into and out of the F state, E71I favors transitions to Ii and E71H is fundamentally biased towards Is. As expected, most Glu71 mutants had a profound influence on the lifetime of Is, a fact that might suggest that destabilization of Is is a precondition to the stabilization of Ii or F.

Figure 3

Kinetic behavior of Glu71 mutants. Representative single-channel activity for Glu71 mutants (Left). Histograms show a distribution of closed and open channel lifetimes for the entire recordings (Middle). Single-channel current recordings were best fit by three closed and one open state for Glu. The closed states were defined as F, Ii and Is based on their lifetimes (Right). Rate constants of recovery from Is for wt and E71H are over estimated due to low Po and uncertainty in the number of channels in the patch.

Individual rate constants derived from the fitted experimental dwell time distributions in the context of our basic four state model (Table 1) were used to simulate single-channel and macroscopic ensemble currents. Since these mutants did not alter activation, we used rate constants previously determined for wt KcsA to describe the proton-dependent transitions12. As expected, the models reproduced all of the unitary characteristics of the different gating modes as well as the time course for the macroscopic responses to pH pulses (Supplementary Fig. 3).

Table 1

Kinetic parameters of the model

	F → O	F ← O	Ii → O	Ii ← O	Is → O	Is ← O	N
Wt	7983 ± 1739	5.6 ± 1.34	526 ± 30	0.9 ± 0.3	0.09 ± 0.01	0.45 ± 0.07	3
E71A	6339 ± 268	10.9 ± 4.8	1055 ± 291	2.2 ± 0.9	14.0 ± 3.4	0.22 ± 0.10	3
E71I	6814 ± 287	21 ± 11.0	152 ± 59	151 ± 58	2.1 ± 1.9	0.33 ± 0.17	5
E71Q	1972 ± 115	1342 ±187	97 ± 66	3.5 ± 1.6	1.86 ± 1.2	0.34 ± 0.34	5
E71H	7906 ± 1853	325 ± 128	526 ± 30	44.2 ± 11.8	0.18 ± 0.02	24.0 ± 1.0	3

Crystal structures reveal subtle changes in ion occupancy

The discovery of individual mutants that greatly stabilized each of the major gating modes in KcsA provides a unique opportunity to probe the structural basis of each gating mode. As the E71A structure is already available in the putative conductive conformation (closed inner gate)11, we focused on mutations E71I, that favor the Ii state, and E71Q, favoring the F state. Crystals of mutants E71I and E71Q were obtained as Fab complexes, diffracting at resolutions of 2.3 Å and 2.7 Å, respectively, and were solved by molecular replacement methods5 using the structure of wt KcsA (1K4C) as search model.

Experimental 2Fo – Fc electron density maps corresponding to the filter region and Fo – Fc omit maps for the ion distribution profile are shown in Figures 4 for E71I and 5 for E71Q. Overall, the selectivity filter structures of all gating mode mutants (Figs. 4 and 5) showed no major changes in backbone conformation and thus, corresponds to the conductive conformation of the filter observed in most closed KcsA structures (RMSD with respect to 1K4C is 0.25 Å for E71I and 0.12 Å for E71Q). This result was not surprising, given that these mutations decrease the rate and extent of C-type inactivation and should indeed stabilize the conductive conformation. However, close observation of the electron density maps around the filter region revealed interesting differences in terms of the relative ion occupancy and the number of water molecules behind the filter (Figs. 4 and 5).

Figure 4

Crystal structure of E71I. (a) Single-subunit line representation of the P-loop of E71I overlaid onto the wt structure5 (PDB entry 1K4C) highlights the conductive conformation of the selectivity filter backbone. (b) One-dimensional electron density profiles along the central symmetry (z) axis is shown. S1-S4 denotes the K+ binding sites. Gray peaks in the background correspond to one-dimensional electro density profile of the wt structure. (c) Electron density map of residues 60–84 from two diagonally symmetric subunits. Sticks, polypeptide chain; blue mesh, 2σ-contour of the 2Fo – Fc electron density map for the protein; magenta mesh, 6σ-contour of the Fo – Fc omit map for the ions ; red mesh, 4σ-contour of the Fo – Fc omit map for the waters. (d) A single-subunit P-loop is shown with side chains at Glu71 and Asp80 in stick representation. The H-bond interaction between the three crystallographic water molecules within the cavity behind the filter and the rest of the protein are represented by black dotted lines.

Figure 5

Crystal structure of E71Q. (a) Single-subunit line representation of the P-loop of E71Q overlaid onto the wt structure5 (PDB entry 1K4C) highlights the conductive conformation of the selectivity filter backbone. (b) One-dimensional electron density profiles along the central symmetry (z) axis is shown. S1-S4 denotes the K+ binding sites. Gray peaks in the background correspond to one-dimensional electro density profile of the wt structure. (c) Electron density map of residues 60–84 from two diagonally symmetric subunits. Sticks, polypeptide chain; blue mesh, 2.5σ-contour of the 2Fo – Fc electron density map for the protein; magenta mesh, (4-6)σ-contour of the Fo – Fc omit map for the ions ; red mesh, 5σ-contour of the Fo – Fc omit map for the waters. (d) A single-subunit P-loop is shown with side chains at Glu71 and Asp80 in stick representation. At 2.7 Å resolution we observe no crystallographic waters within the cavity behind the filter.

The one-dimensional electron density profile along the pore axis of E71I points to a clear loss of ion occupancy at the S2 binding site, together with an apparent decrease in the occupancy at S1 (Fig. 4b). On the other hand, the E71Q mutant shows a modest increase in the occupancy at the S2 binding site in comparison to wt KcsA (Fig. 5b). Besides these changes, the most striking difference among the mutants is the number of water molecules buried between the pore-helix and the selectivity filter, bridging the filter region to the rest of the protein. In wt KcsA, a water molecule is coordinated by hydrogen bond interactions with Glu71, Asp80 and the backbone of Tyr78 in the selectivity filter. In the E71I structure, the “cavity” formed by the reduced side chain volume is filled by 3 crystallographic waters interacting with the filter through a network of hydrogen bonds connecting the backbone of Gly79, Asp80, Leu81 and Tyr78 and the carboxyl group in Asp 80 (Fig 4c, d). There was no observable coordinated water in this region in E71Q (Fig 5c, d) but this is likely due resolution issues (at 2.7 Å).

Glu71 mutants affect filter conformational dynamics

With the availability of individual crystal structures underlying each of the major gating modes, we then addressed the question of whether increased stability of Ii and F in the E71I and E71Q mutants might be reflected in the conformational fluctuations of the filter. Molecular dynamics (MD) simulation studies suggest that water molecules behind the selectivity filter can affect the filter’s conformational flexibility during ion permeation39. The timeframes of the single channel transitions observed for Ii and F (and of course, Is) can be orders of magnitude away from the dynamic window available to straightforward MD simulations, however the present structures offer a unique opportunity to evaluate short timeframe differences that might point to subsequent events directly linked to these gating transitions. We carried out a series of 20 ns MD simulations for mutants E71A, E71I, E71Q and E71G along with wt KcsA, in a fully explicit system embedded in a lipid bilayer.

From these MD runs, we observed two events of particular interest: First, the amide plane at Val76-Gly77 in the selectivity filter (in one of the subunits at a given time) underwent a 180° reorientation, pointing the backbone carbonyl oxygen of Val76 away from the conduction pathway (Fig. 6a). This type of transition has been observed in a number of MD simulations40-43 and leads to the loss of a favorable interaction at the S3 binding site, potentially affecting the free energy barrier associated to the translocation of K+ from S3 to S244. A plot of the backbone torsion angle as a function of the simulation time shows that the outward-facing (away from the conduction pathway) Val76 carbonyl is greatly stabilized in E71Q, where this conformation occurs ~40% of the total simulation time at least in a single subunit (Fig. 6b). In contrast, in wt KcsA the Val76 carbonyl remained outward-facing some 13% of the time, while for the E71I mutant it was only 3% (Fig. 6b). At the functional level, the dwell time ratio between the F and O states in the E71Q mutation was approximately 40:60 (Fig 3), essentially mirroring the outward-facing/straight conformations in the simulation above.

Figure 6

Underlying conformational dynamics of the selectivity filter and the fast gating kinetics. (a) Structural snapshots of outward-facing carbonyl conformations. (b) Dynamics of carbonyl reorientation in KcsA. Time traces of the Val76 carbonyl dihedral angle (N-CA-C-O) during 20 ns molecular dynamics trajectories. Different color lines correspond to different subunits. Potassium ions were initially placed in the cavity and sites S1 and S3. (c) Distribution of Glu71-Asp80 Cα-Cα distances. The green and magenta fits correspond to populations with Asp80 facing-down (centered at ~10.3Å) and “flipped”-outward (centered at ~11.1Å) respectively.

Given the known differences in the timescales of single channel kinetic and MD simulations, this comparison between relative populations of states in simulations with the single channel kinetics only represent part of a global sequence of events and at best, should be considered only as a qualitative piece of the filter dynamics puzzle. Still, we draw attention to the fact that out of five MD runs (wt, E71A, -I, -Q and -G), only the flicker-prone E71Q mutant showed a considerable increase in the frequency and lifetime of Val76 reorientation (Fig. 6 and Supplementary Fig. 4). Therefore, we would like to suggest that the reorientation of the Val76 carbonyl might indeed be associated to the conformational changes that eventually lead to short-lived flicker states in single-channel records.

The second observation involves the outward “flipping” of the Asp80 side-chain relative to its position in wt KcsA, a movement reminiscent of the conformation observed in one of the crystal forms of E71A11 (the so-called “flipped” structure) and in the selectivity filter of Kir 3.145. Monitoring the Cα-Cα distance between the Glu71 and Asp80 reveals a very narrow distribution in wt KcsA, indicative of a strong interaction between these two residues (Fig. 6c, top). On the other hand, E71Q and E71I display a broader distribution with a distinct second population that corresponds to channels with “flipped” Asp80 (Fig. 6c, bottom two). In fact, this dual-population behavior is observed in all gating mode mutants at position 71, suggesting that the overall mobility of the Asp80 side chain is enhanced in the absence of an interaction with Glu71 (Fig. 6 and Supplementary Fig. 5).

DISCUSSION

K+ channel stationary gating is known to involve non-conductive kinetic states with lifetimes ranging from sub millisecond to several seconds. Transitions between these and conductive states define burst properties at the single-channel level, while changes in the equilibrium between these gating events lead to gating mode-shifts. Modal gating is a common feature in a wide range of channels, particularly Kv30-32, Nav34, Cav46-48, BK49,50 channels as well as the AChR51-53, and NMDA receptors54,55. In some channels, the mechanism for modal inter-conversion is subject to cellular control via phosphorylation and other post-translational modifications30,31,36,37,56, but the molecular underpinning of these events has remained unknown in the majority of channels. In this study we show that kinetically diverse conformational states that give rise to modal gating shifts can also originate from conformational fluctuations at or near the filter. Conveniently, in KcsA each of the naturally occurring modes can be individually stabilized, depending on the nature of the side-chain at position 71 in the pore-helix. Given that a wide range of mutations converged on a limited set of gating modes, we suggest that the various side-chain substitutions at position 71 do not introduce new kinetic behaviors, but actually modulate the relative stability of pre-existing conformational states that are intrinsic to wt KcsA. These states were defined by their intrinsic dwell time distribution (each roughly tenfold faster) as the slow (or deep) inactivated state Is, and intermediate inactivated conformation (Ii) and the highly fluctuating F or flicker state.

The question still remains as to why the wt channel exhibits gating heterogeneity while mutants reveal a more homogenous behavior. One common feature among these mutations is a substantial loss of C-type inactivation and given that modes arise predominantly as channels recover from this inactivated state, it is likely that modes switches are associated with channels transitioning between the deep inactivated and a series of open conductive conformations. These transitions might have slightly different energy paths, therefore ensuing a heterogeneous behavior in wt KcsA.

The structural snapshots that underlie the molecular events leading to C-type inactivation (Is) have recently been defined crystallographically9. These include a sequential loss of ion binding sites S2 and S3, a pinching of the permeation pathway at Gly77 and a compression of the filter along the fourfold symmetry axis. The crystal structures and MD simulations of the E71I and E71Q mutations offer insights into the short-lived Ii and F states, respectively.

Ion occupancy in the E71I mutant filter is almost fully lost at S2 and partially decreased at S1, an ion profile that is distinctly different to that seen in either the fully conductive or the C-type inactivated filter9. We propose that this conformation of the filter is related to the intermediate Ii state for two reasons: First, even with the loss of external ion binding sites the backbone conformation is essentially that of the fully conductive filter. In comparison, loss of S2 and a partial loss of S3 in a incompletely C-type inactivated filter leads to obvious changes in the filter backbone9. Second, recovery from C-type inactivation has been shown to be sensitive to external permeant ion concentration and to the ability of the ion to move from one site to another57. Given its easy access to the external bulk K+, ion rebinding at S1 is expected to be more favorable than at S2 and S3. We therefore suggest that during permeation in wt KcsA, loss of ions at the external binding sites lead to the intermediate Ii states while sequential vacancies at the deeper S2 and S3 sites lead to the more stable, fully developed C-type inactivated (Is) state.

Another interesting feature of this structure relates to the variation in the number and location of water molecules coordinated in the cavity behind the selectivity filter. In Wt KcsA, a water molecule links the carboxyl groups on Glu71 and Asp80 with the backbone amides of Tyr78 and Gly79 in the conductive conformation, but two water molecules in the collapsed, low K+ structure5 that coordinate these residues. Substitution of Glu71 to Ile results in three water molecules that generate extensive H-bonding interactions with the backbones of residues Gly79, Leu81 and Tyr78, and the Asp80 side-chain. Although the precise effect of substituting the functionally important E71-D80 interaction for the three waters of coordination is as yet unknown, it is tempting to speculate that the resulting H-bond network promotes the stabilization of Ii with a relative destabilization of the conductive conformation of the filter (O).

Our molecular dynamics simulations suggest that the frequency of spontaneous reorientation transitions of the Val76 carbonyl group is greatly enhanced in the E71Q mutation while it is mostly unaffected in the rest of the mutants. However, given the wide differences in time-scale between our experimental and computational data, it is difficult to establish a direct correlation between the dwell-time of the outward-facing carbonyl conformation (ns timescale) and the kinetics of single-channel flicker events (μs timescale). Clearly, it is unlikely that one carbonyl Val76 reorientation corresponds to one flicker event observed in electrophysiology. However, we suggest that an increase in the frequency of flickers with parallel increases in the incidence of Val76 reorientation could accompany ion translocation from one binding site to the next during permeation, also triggering transitions that lead to short-lived flicker states. Thus, reorientation of Val76 might be an “initiating” event that leads to subsequent non-conductive conformations of the selectivity filter. These conformations would be metastable in the μs timescales, as has been suggested on the basis of equivalent MD runs44.

MD runs further revealed a bimodal distribution of Asp80 side-chain positions for all of the Glu71 mutations (except E71H). The existence of these two conformations is not surprising if we consider that the loss of the interaction with Glu71 should enhance the mobility of Asp80. It is interesting to note that the enhanced flexibility of Asp80 is also reflected as an increase in the crystallographic B-factor of this region in the E71I crystal structure (Supplementary Fig. 6). We believe that the additional conformational freedom of Asp80, leads to a decrease in the backbone constrains at Tyr78 and Gly79, with obvious consequences to the overall conformational dynamics of the filter. However, it still remains unclear as to how this motional freedom relates to the different gating forms seen in wt KcsA. Since most voltage-dependent K+ channels have a valine at the position corresponding to Glu71 in KcsA, some of these conformational fluctuations of the filter might play a role in other members of the K+ channel family.

In conclusion, the pore helix, selectivity filter, and external vestibule are dynamic structures where small local conformational changes (that include motions of the carbonyl oxygens, small fluctuations of the filter backbone, or changes in the configuration and occupancy of water molecules behind the filter) can lead to drastic effects on gating. These transitions define the interplay between ions and the filter and thus underlie the diverse gating patterns observed in single channel recordings of most K+ channels. In KcsA, selectivity filter fluctuations are defined by a complex energy landscape that defines three kinetically distinct gating fluctuations. Transitions into each of the different gating modes depend on the relative depth of the energy wells associated with the three of pre-existing selectivity filter conformations.

Table 2

Data collection and refinement statistics

	E71Q	E711
Data collection
Space group	I4	I4
Cell dimensions
a=b, c (Å)	155.7,76.1	156.4, 76.0
α=β=γ(°)	90	90
Resolution (Å)	40-2.7	40-2.3
Rsym or Rmerge (%)	6.9(36.3)	8.0(32.4)
I/σI	17.5(3.7)	20.0(2.7)
Completeness (%)	94.0(93.2)	96.1(94.3)
Redundancy	6.1(3.1)	3.9(3.1)
Refinement
Resolution (Å)	40-2.7	40-2.3
No. reflections	23689	39318
Rwork/Rfree (%)	22.6/26.8	26.4/27.1
No. atoms
Protein	4074	4073
Ligand/ion	7	5
Water	4	7
B-factors
Protein	59.2	45.7
Ligand/ion	41	51
Water	52	82.4
R.m.s deviations
Bond lengths (Å)	0.007	0.005
Bond angles (°)	1.36	1.11