1. Introduction
One of the most intriguing features of potassium channels is their high selectivity for the larger K
+ ion versus the smaller Na
+ ion, while maintaining a high ion flux [
1]. MacKinnon’s [
2] and Perozo’s [
3] groups have contributed the most on this matter through the study of the structure of a potassium channel from
Streptomyces lividans, KcsA. This channel is a homotetrameric membrane protein, where each monomer is conformed by two α-helical transmembrane segments (TM1 and TM2) and a C-terminal cytoplasmic end. The channel pore is found between TM1 and TM2, contributed by a short tilted helix and the selectivity filter [
2], which also operates as a gate (outer gate). The four C-terminal ends arrange as a helical bundle with a conformation that is sensitive to pH, being another pore gate (inner gate) [
3] (
Figure 1). Importantly, the signature sequence of the selectivity filter of KcsA (TVGYG) is homologous to that of the eukaryotic K
+ channels. The backbone carbonyls of these residues conform four K
+ binding sites (sites S1–S4, from the extra to the intracellular side), as seen by crystallography [
2,
4], which can adopt different conformations at high or low K
+ concentrations [
4,
5,
6,
7,
8]. Nuclear magnetic resonance (NMR) studies have also reported such changes [
9,
10,
11,
12,
13,
14]. The conformation at low K
+ concentration shows no ions at the center of the selectivity filter (sites S2 and S3), thus adopting a “collapsed” structure which impedes ion flow through it. K
+ binds to S1 and S4 with an average occupancy of just one ion distributed between those two sites. However, at high K
+ concentrations, a conformational change is induced whereby a second K
+ enters the filter, with an average occupancy of two K
+ ions per channel, either at the S1–S3 or the S2–S4 positions, thus enabling ion conduction [
4,
7]. The structure of KcsA in the presence of other permeating ions such as Tl
+, Rb
+ or Cs
+ has also been studied through X-ray crystallography. In the case of the first ion, the occupation is identical to K
+, while the other two ions are not able to bind the S2 site, although they still induce a conductive conformation with an average of two ions per channel [
8]. The nonpermeant Na
+, on the other hand, is not able to induce such a conformational change and shows an average occupancy of one ion per channel at the S1 and S4 sites [
15].
In terms of functional activity, KcsA is described through a cycle including four different states. At neutral pH, the channel is in a closed/conductive resting state, whereby the cytoplasmic helical bundle (the inner gate) impedes ion flow, while the selectivity filter (the outer gate) displays a conductive form. At acidic pH, the inner gate opens, enabling the open/conductive state which allows ion flow, making KcsA a proton-activated channel [
16,
17,
18]. However, this is not a stable state and the outer gate adopts a conformation reminiscent of C-type inactivation in eukaryotic K
+ channels [
19,
20,
21], which impedes ion flow within a few seconds [
20,
21,
22] in the open/inactivated form. The cycle is completed when the pH returns to neutrality, which induces the closure of the inner gate and thus the transient closed/inactivated state, which rapidly evolves to the initial closed/conductive resting state [
23]. This cycle reveals the concerted action of the two channel gates.
The elucidation of the molecular reasons for why the inactivated channel is nonconductive is of great interest to the mechanisms of ion conduction but is currently subject to controversy. Some authors relate the nonconductive, “collapsed” conformation of the selectivity filter seen in X-ray experiments at low K
+ concentrations [
4] (PDB entry: 1K4D) with the inactivated conformation [
15], such that the access of K
+ to the most internal S2 and S3 K
+ binding sites would not be permitted [
24]. In contrast, other authors have found only modest conformational changes in the G77 residue during inactivation [
25] compared to the resting state, concluding that the selectivity filter remains “resting-like” upon inactivation, with all four K
+ binding sites accessible to cations. Our goal in this paper is to contribute to this issue by studying ion binding under equilibrium conditions, as a tool to explore the access of cations to the ion binding sites within the selectivity filter of KcsA in the inactivated state, and to compare such accessibility with that which is exhibited by the channel in the resting state. Unfortunately, there is not a single consensus model for the inactivated state of KcsA and therefore, we used several “open/inactivated” channel models in our study. The first of such models is the wild-type (WT) KcsA at pH 4, which is the standard experimental system used to measure channel inactivation rates upon a pH jump from 7 to 4, which is known to maintain the opening the inner gate as long as the acidic condition is imposed [
26,
27]. Our second channel model is the deletion 1–125 KcsA that lacks the cytoplasmic 125 to 160 residues at the C-terminal helical bundle. Such deletion destabilizes the channel’s inner gate, and although it appears closed in the X-ray structure [
2,
4] (PDB entry: 1K4C), electron paramagnetic resonance data shows that in detergent solution at neutral pH it remains widely open [
28]. Third and last, we use a full-length version of the so called H25R, R117Q, E120Q, R121Q, R122Q and H124Q sextuple mutant channel (OPEN) KcsA, in which a number of mutations, mostly at the C-terminal region (see Methods), result also in a permanently open inner gate, even at neutral pH [
29].
To study ion binding to KcsA, we used a previously reported assay based on the intrinsic fluorescence from residues W67 and W68 of the channel, located at the short pore helix (
Figure 1B), which are very sensitive to the selectivity filter conformation [
2,
6]. This fluorescent signal allows for the monitoring of the thermal denaturation of KcsA, which includes the dissociation of the tetrameric protein into its subunits and their partial unfolding [
30,
31,
32], a process found to be dependent on the binding of ions to the selectivity filter [
30]. The experimental observable in this assay, the apparent t
m (midpoint denaturation temperature in degrees Celsius) of the native to denatured thermal transition, is ideally suited to study cation binding [
30,
32], since it exhibits an extraordinarily wide range of cation-dependent changes. This allows the determination of apparent dissociation constants for the binding processes ranging from 10
−2 to 10
−9 M in the previously characterized closed state at pH 7.0 [
30,
33]. From these studies, it has been established that there is a single set of sites with low affinity (millimolar K
D values) for the nonpermeant cations such as Na
+ or Li
+, which have been associated to the crystallographic S1 and S4 sites. However, two different sets of sites have been found for the permeant K
+, Rb
+, Tl
+, and even Cs
+ ions, as their concentration increases. Previous studies by crystallography also evidence the ability of these cations to induce concentration-dependent transitions between nonconductive and conductive conformations of the selectivity filter. Based on this analogy, the high affinity set (micromolar K
D values) has been assigned also to the crystal S1 and S4 sites, thus securing displacement of competing nonpermeant cations. The second set of sites shows low affinity (millimolar K
D values), thus favoring cation dissociation and permeation, and is contributed by all S1–S4 crystallographic sites.
3. Discussion
The accessibility of different cations to the stack of K
+ binding sites in the selectivity filter of the KcsA channel has been previously documented by X-ray crystallography and by a variety of other techniques [
2,
4,
9,
11,
14,
15,
28,
36]. Regardless of the use of different KcsA proteins (full length, mutations or partial deletions of the channel), most of this information has been obtained under equilibrium conditions, in detergent solution and at neutral pH, at which the channel is in the resting state when in the presence of adequate cations, i.e., with a closed inner gate and a conductive selectivity filter.
Figure 9 summarizes such potential occupations of the selectivity filter by cations in an idealized manner. The nonpermeant Na
+ binds with low affinity to the S1 and S4 sites of a nonconductive filter. Likewise, the strong potassium channel blocker Ba
2+ binds with an extremely high affinity to the S2 and S4 sites. On the other hand, the permeant K
+ at low concentration binds with fairly high affinity to the S1 and S4 sites of a nonconductive, “collapsed” filter. Blocking the access to the S4 site by TBA
+ at these low concentrations of K
+ allows for K
+ binding to the S1 site only. Finally, at higher K
+ concentrations, there is a cation-induced change in the conformation of the selectivity filter to a conductive state, by which K
+ gain access with roughly equal probability to all S1 to S4 sites through a permeation-allowing binding event, characterized now by a lower affinity. Likewise, Cs
+ or Rb
+ are both permeant species which behave similarly to K
+, with the exception that at high concentrations these two cations bind to the S1, S3 and S4 sites only, but cannot access the S2 site (
Figure 9).
In this report we have used a thermal denaturation assay to study the equilibrium binding of all the above cations to three different models of KcsA in the inactivated state. Our goal was to explore possible changes in the accessibility of the cations to the stack of K
+ binding sites in the selectivity filter as a consequence of channel inactivation. Two major hypotheses on this matter are currently entertained. The first hypothesis postulates that the inactivated state of the selectivity filter corresponds to a “collapsed” conformation, similar to that seen in presence of low K
+ concentration [
15] (PDB entry: 1K4D), in which access to the most internal S2 and S3 sites is not permitted [
24]. On the contrary, the second hypothesis claims that the filter remains in a “resting-like” conformation upon inactivation, with all the K
+ binding sites similarly accessible to cations [
28]. In any case, either hypothesis should be able to explain why the inactivated channel is nonconductive. Our results show that qualitatively, all the binding sites within the inactivated selectivity filter, probed either individually or collectively though the binding of the different cations, remain as accessible as in the resting state. Therefore, we have no evidence to support a “collapsed” filter conformation with inaccessible sites in the inactivated state and must conclude that the inactivated filter is “resting-like” in terms of cation accessibility under equilibrium conditions. In spite of such a similarity, the quantitative comparison between the equilibrium binding constants derived from the resting channel state and from the three inactivated models points out differences which refer mainly to a decreased affinity in the binding of permeant cations to the inactivated channels (
Figure 10). Such a decreased affinity seems more clearly observed for K
+ binding, especially in the first binding event. This implies that the initial binding of K
+ to the S1 and S4 sites, which occurs at a low cation concentration and is proposed to be the basis for ion selectivity [
30], is altered in the inactivated versus the resting state. In the case of the pH 4 inactivated channel, such an affinity is nearly 50-fold lower than that in the resting state. As indicated above, the results from the pH 4 model should be taken with caution, because of the additional effect of pH on the thermal stabilization of the channel protein. Nonetheless, the 1–125 and OPEN channel models also detect a decrease in the affinity for binding of permeant species, and therefore, we conclude that this is a common qualitative feature in all three inactivated models.
Experiments of K
+ binding under blockade of the S4 site by TBA
+ could serve to evaluate whether the S1, S4 or both of these sites are involved in the decreased affinity of the first K
+ binding event exhibited by the inactivated channels. Again, these experiments should take into account the much lower affinity of TBA
+ for binding to the inactivated state versus the resting state of the channel, which might not completely guarantee an efficient TBA
+ blockade of K
+ binding. Furthermore, it is possible that the TBA
+ blockade could additionally differ in terms of efficacy among the three inactivated models. In spite of such potential limitations, the results seem to point out two main findings: First, similarly to the resting state [
33], the inactivated filter remains asymmetric, since the affinity for K
+ binding to the extracellular S1 site is lower, relative to the intracellular S4 site. Second, comparing the resting versus the inactivated states, the changes in the K
D’s for K
+ binding under TBA
+ blockade were not as large as those seen in its absence (
Figure 10). Therefore, it could be concluded that K
+ binding to the S1 site was less affected by inactivation than that in the S4 site. Thus, the S4 site appears to be mainly responsible for the observed loss in affinity during the first event of K
+ binding. This observation seems to be in agreement with the model proposed by Valiyaveetil et al., in which inactivation was suggested to affect mainly the S3 and S4 sites of the selectivity filter [
37].
Thus, in overall terms, the most noticeable change between the resting and the inactivated state of KcsA from our binding experiments is a drop in the affinity for permeant ions, especially notorious for K
+, which is precisely the ion that promotes deeper and faster channel inactivation. As indicated throughout the manuscript, similarly to X-ray crystallography or NMR, our experiments were done under equilibrium conditions. Still, it is tempting to compare our results with those from electrophysiology, although they are under kinetic control and have a voltage imposed to define the direction of the flow of ions. In the latter, inactivation has long been associated to a loss of K
+ at the selectivity filter in potassium channels [
38,
39], which explains why C-type inactivation is favored at low K
+ concentrations in eukaryotic potassium channels [
4,
15,
19,
40], while a higher concentration of the ion prevents it [
27,
40,
41,
42,
43]. These observations led to the “foot in the door” or the “ion depletion of the pore” hypothesis, which basically proposed that the presence of ions inside the selectivity filter is fundamental to stabilize it in the conductive conformation [
44,
45,
46]. The drop in the affinity we detected in our equilibrium binding experiments would increase the probability of K
+ loss from the filter, then hampering ion conduction. As for the collapsed conformation found for the inactivated state in the crystallization experiments [
24], it is conceivable that such a conformation was reached temporarily, favored by the affinity drop, where a kinetic intermediate could perhaps become stabilized under the crystallization conditions. In fact, the very essence of the selectivity filter is to be dynamic, subject to continuous sojourns to different conformations, which explains the discrete closures in the conductive state and the different gating modes of KcsA [
47,
48].
Another consequence from the drop in potassium affinity related to the partial loss of selectivity for K
+ versus Na
+. Since sodium affinity seems to not be significantly altered in the inactivated state, the K
D (Na
+)/(K
D (K
+) ratio diminishes. This parameter is related to ion selectivity only at equilibrium [
30,
32,
33,
49], but could perhaps be used to explain the reported loss of selectivity in several potassium channels, observed electrophysiologically when in the inactivated state [
50,
51,
52]. In extreme cases, such as that in the case of the inactivated M96V-KcsA channel model, the loss of selectivity is so severe that in addition to K
+, the usually nonpermeant Na
+ becomes a permeant species in this mutant channel [
53].
In summary, use of cation binding as a tool to explore the accessibility at equilibrium of permeant and nonpermeant cations to different inactivated channel models shows that the stack of ion binding sites in the inactivated filter qualitatively remain accessible to cations, as in the resting channel state. The inactivated selectivity filter is therefore “resting-like” under such equilibrium conditions. Nonetheless, quantitative differences in the KD’s of the binding processes reveal the affinity for binding of permeant cations, mainly K+, to the inactivated channel models is decreased with respect to the resting channel state. This is likely to cause a loss of K+ from the inactivated filter, and consequently, to promote nonconductive conformations. S4 seems the most affected site by the affinity loss, which is interesting because S4 is the first site to accommodate K+ coming from the channel vestibule under physiological conditions of K+ exit from the cell.