Subunit gating resulting from individual protonation events in Kir2 channels

Inwardly rectifying potassium (Kir) channels play a critical role in stabilizing the membrane potential, thus controlling numerous physiological phenomena in multiple tissues. Channel conductance is activated by cytoplasmic modulators that open the channel at the 'helix bundle crossing' (HBC), formed by the coming together of the M2 helices from each of the four subunits, at the cytoplasmic end of the transmembrane pore. We introduced a negative charge at the bundle crossing region (G178D) in classical inward rectifier Kir2.2 channel subunits that forces channel opening, allowing pore wetting and free movement of permeant ions between the cytoplasm and the inner cavity. Single-channel recordings reveal a striking pH-dependent subconductance behavior in G178D (or G178E and equivalent Kir2.1[G177E]) mutant channels that reflects individual subunit events. These subconductance levels are well resolved temporally and occur independently, with no evidence of cooperativity. Decreasing cytoplasmic pH shifts the probability towards lower conductance levels, and molecular dynamics simulations show how protonation of Kir2.2[G178D] and, additionally, the rectification controller (D173) pore-lining residues leads to changes in pore solvation, K+ ion occupancy, and ultimately K+ conductance. While subconductance gating has long been discussed, resolution and explanation have been lacking. The present data reveals how individual protonation events change the electrostatic microenvironment of the pore, resulting in distinct, uncoordinated, and relatively long-lasting conductance states, which depend on levels of ion pooling in the pore and the maintenance of pore wetting. Gating and conductance are classically understood as separate processes in ion channels. The remarkable sub-state gating behavior of these channels reveals how intimately connected ‘gating’ and ‘conductance’ are in reality.


72
provides a major hydrophobic bottleneck that blocks ion conduction [5,9]. This 73 implicates expansion of the HBC as a critical step in Kir channel opening, as seen in a 74 recent Kir6.2 structure [11]. We reported the crystal structure of a chicken Kir2.2[G178D] 75 mutant channel [12], in which the introduced G178D mutations at the HBC functionally 76 stabilize the open conformation, a strategy used previously to obtain an open crystal 77 structure of a bacterial homolog KirBac3.1 [13]. In the G178D structure, the HBC gate is 78 slightly wider than in previous structures, and molecular dynamics (MD) simulations 79 demonstrate rapid wetting of the G178D pore at the HBC region, followed by further 80 expansion and K + conductance through the channel.

81
In the present study, we have carried out detailed single-channel analyses of 82 cKir2.2[G178D] and hKir2.1[G177E] channels that reveal very striking sub-state gating 83 behavior. These channels show multiple, very consistent subconductances, that are not 84 observed in wild type channels, and that are quite different from occasional 85 subconductances reported for Kir2.1 or Kir2.2 [14,15]. While subconductance gating 4 has been sporadically observed in many related and unrelated channels [14][15][16][17][18] (Fig. 1A). This mutant demonstrated activity in the 100 absence of the normally obligatory ligand PI(4,5)P2, as well as an increased unitary 101 conductance (~60 pS vs. ~46 pS for WT) in excised patches [12]. Closer analysis 102 reveals subconductances within the single channel current (Fig. 1B) that are completely 103 absent in WT Kir2.2 currents ( Fig. 2B and S1). Under symmetric pH 7.4 buffer conditions 104 we identified four distinct conducting states (labeled O1-O4) with relative amplitudes of 105 0.45, 0.74, 0.92 and 1.00 (Fig. 1B). Although the fully open state still dominated at 106 physiological pH (P O4 = 0.59), the other sub-states contributed significantly: P O3 = 0.26, 107 P O3 = 0.07 and P O1 = 0.02 (Table S1). The vast majority of transitions occurred between 108 adjacent conducting levels, although transitions to and from the zero-conductance level 109 (C) were more frequent than transitions between non-adjacent conducting levels ( Fig.   110 1C, Table S1). The similar mutation G178E also demonstrated increased unitary 111 conductance and four open states (Fig. 1D) that were almost identical to the G178D 112 conducting states, indicating that the appearance of this sub-state gating was directly 113 caused by replacement of glycine at position 178 with an acidic residue.

115
Selectivity filter gating is unaffected by the G178D mutation

116
We further analyzed the brief intra-burst closures (burst closure cutoff <100 msecs) that 117 occur in both WT and G178D channels, at pH 7.4. In the case of the mutant, only 118 closures after which the channel returned to the sub-state that it closed from (i.e., Oi -> C 119 -> Oi) were used for analysis. The frequency of occurrence was slightly lower for G178D 120 (3.7 s -1 ) than for WT (5.9 s -1 ), potentially due to exclusion of Oi -> C events that 121 apparently returned to states other than Oi in G178D, but C state dwell times (6.4 ± 0.5 122 ms and 7.1 ± 0.5 ms, mean ± S.E.) were not significantly different between WT and 123 G178D (p = 0.343, unpaired T-test). Furthermore, the fraction of transitions between the 124 fully closed state (C) and any of the conducting sub-states (O1 to O4) in G178D (Fig. 1C) 125 followed the probability distribution of these sub-states (Table S1). Therefore, we 126 suggest the transitions to the C state are distinct closures (potentially at the selectivity 127 filter) that occur independently of the G178D-driven, stepwise sub-state transitions. We 128 further suggest that the rare transitions between non-adjacent conducting states in Fig.

129
1C, are likely to represent combinations of sequential transitions between adjacent 6 conducting levels that are not temporally resolved at the experimental data acquisition 131 and filter frequencies (3 kHz / 1 kHz, see Methods for details).

134
While the pKa of solvent-exposed aspartate side chains is very acidic (~pKa 3), it can be 135 strongly modulated by the protein and membrane microenvironment [19][20][21]. Thus, it is 136 conceivable that the effective pKa may shift within the channel pore [22], and that the 137 sub-states are a consequence of G178D side chain protonation. We therefore further 138 investigated the pH-dependence of Kir2.2[G178D] sub-state behavior. While the single 139 conductance of WT Kir2.2 was independent of pH ( Fig. 2B), the sub-state occupancy 140 was strikingly pH-dependent in G178D channels ( Fig. 2A). The same four conducting 141 states were detectable at each pH ( Fig. 2A,C) with pH-independent sub-state amplitudes 142 (Table S2, Fig. 2D), but there was a clear pH dependence to the sub-state occupancy, 143 with a gradual shift from the lowest-conductance sub-state (O1) being predominant at pH 144 6.2, to the highest conductance (O4) being almost exclusively present at pH 8 (Fig. 2C).

145
Such dependence on pH supports independence of the full closing events, and points to 146 a role of the protonation state of the introduced G178D residues in occurrence and 147 conductance of the sub-states.

148
Over the negative membrane potentials at which each conductance state was 149 experimentally resolvable (from -160 to -60 mV) the subconductance states were clearly

210
This unpredicted finding, i.e., that g-E, containing only two introduced G177E residues 211 per channel, still generates four distinct conducting sub-states, indicates a more complex 212 mechanism of sub-state generation, potentially involving additional ionizable groups.

214
Additional ionizable residues contribute to sub-state generation

215
As discussed above, for protonation of the introduced acidic residues to be involved in 216 pH-dependent sub-states, the effective pKa must be shifted to a much higher pH than 217 that of solvent-exposed aspartate or glutamate side chains. To estimate pKa values for 218 these aspartates and other potentially ionizable residues in the pore, we used a  (Table S3). This suggests that the titration of both residues G178D and D173 225 may contribute to sub-state generation.

226
To examine the contribution of the protonation state of the rectification controller,  Table 1. In the first system, none of the D173 or G178D residues were 254 protonated, i.e., each side chain was negatively charged, resulting in a system (termed 255 Qcav -8) with a net charge of -8 in the central cavity. Assuming that individual residues 256 would be protonated in a stepwise manner with decreasing pH (corresponding to 257 acidification of the medium), we protonated one each of the four side chains of D173 and 258 G178D in the second (Qcav -6) system. In the third system (Qcav -4), two opposing D173 259 and G178D each were protonated. We also introduced a WT-Qcav -2 system, in which 260 residue 178 was the wild type G178, and only two opposing D173 were protonated.

261
As in the electrophysiological experiments, Qcav -8 (mimicking conditions at high 262 pH) was the most conductive system with an average of 12.7 (S.D. 1.2) K + ions being 263 conducted in 1 µs. As shown in Table 1  The flux of single K + ions contributing to the conductance rates summarized in Table 1 is 282 visualized in Fig. 6 (top panels), Fig. S2, and Movies S1 and S2. As is clearly seen, the 283 number of negatively charged residues in the pore essentially determines the K + 284 occupancy in the pore. Between the selectivity filter (SF) and M308 of the G-loop, the 285 pore pools an average of 9.3 K + ions in the Qcav -8 system, while K + occupancy steadily 286 decreases to 6.3 (in Qcav -6) to 5.6 (in Qcav -4), as protonation increases (Fig. S3). The 287 lack of any G178D aspartates in the WT-Qcav -2 system further reduces the occupancy 288 to 3.6 K + ions.

289
As previously demonstrated, the introduction of negative charges at residue 178 290 in the bundle crossing region increases the diameter at the HBC gate [12]. Thus, the 291 position of the hydrophobic isoleucine residue 177 within the Kir2.2 HBC (Fig. 1A), is 292 strongly affected by changes of the protonation state of both G178D and D173 (Fig. 7).

293
As protonation increases and the net charge drops within the cavity, the pore diameter at 294 this location narrows, on average, from 8.2 Å (in Qcav -8) to 6.1 Å (in Qcav -4). Similarly, 295 although M181 side chain flexibility is greater than that of the hydrophobic I177, 296 increasing protonation also results in a narrowing of the most frequently sampled M181 297 minimum distances from 10.6 Å (Qcav -8) to 8.9 Å (Qcav -4), with an additional peak at 6.2 298 11 Å (Fig. 7). Minimum distances in the HBC gate of the WT-Qcav -2 simulation are similar to 299 values of the Qcav -4 system. Additional pore-constricting residues in the inward direction 300 are M308 and M302 of the G-loop. Interestingly, the presence of the G178D mutant, 301 whether protonated or not, has a pronounced effect on M308, leading to widening in the 302 mutant channel compared to WT-Qcav -2 simulations. The most frequent minimum 303 distances range from 9.4 Å (Qcav -8) to 8.6 Å (Qcav -4), but much narrower M308 residues 304 (≤ 5.4 Å) in a considerable fraction of WT-Qcav -2 simulations. M302 is less influenced by 305 the G178D mutation, constricting the G-loop to minimum distances around ~6.0 Å in all 306 systems. The constriction of these gates (Fig. 7) is accompanied by temporarily 307 desolvated periods (Fig. 6 and Fig. S2, bottom panels). Thus, in addition to direct effects 308 on K + occupancy, increasing protonation also decreases pore solvation in the HBC and  Since the first recordings of recognizably individual ion channels, it has been evident that 317 most ion channels exhibit stereotypically on-off behavior, with two experimentally 318 measurable current levels, one that is indistinguishable from zero and one that is 319 characteristic of the specific channel conductance under a given ionic condition. This

346
'opening' is that the step from zero to full conductance is the last and unified step that 347 follows hidden conformational changes resulting from the gating ligands or sensors, and 13 kinetic analyses have provided some very compelling kinetic models that fit this 349 paradigm [36][37][38][39][40]. But there is still a lingering discomfort with this framework, since a 350 final single opening step cannot easily be accomplished in multimeric (e.g., tetrameric 351 Kir) channels. One notion is that channel opening actually occurs in multiple steps, but 352 that these are so highly cooperative that each step cannot be experimentally resolved.

353
Another possibility is that there is cooperativity in the effects of each step on 354 conductance, which becomes maximal with the first subunit 'opening', or which is not

374
and assuming that the protonation rate is a diffusion-limited association (kon ~ 10 10 M -375 1 s -1 ), the effective on-rate (and off-rate at pH 3) would therefore be immeasurably fast 376 (~10 7 s -1 ). The apparent pKa for sub-state gating is ~7 (Fig. 3). An upshift of the 377 aspartate pKa from 3 to 7 within the constricted space that could prevail within the inner 378 cavity would therefore slow on-and off-rates by 10 4 (i.e., ~10,000-fold), bringing them 379 both into the millisecond range at neutral pH. Thus, the striking conclusion that the One obvious and straightforward conclusion arising from our current and previous [12] 385 analyses is that increase of the net negative charge of the conductive pore leads to 386 higher K + currents through it (although this may be challenged by recent MD simulations 387 suggesting that conductance is inhibited if all four Kir2.2 D173 residues are ionized at 388 lower membrane potentials [29]). Our simulations suggest that the mechanism of pore 389 negativity-driven increases in conductance primarily relies on increased overall K + 390 occupancy of the pore, but also on consequent pore widening and increased solvation. this local re-arrangement is sufficient to slightly increase wetting at the HBC constriction.

396
By attributing fixed protonation states to the two key pore-lining residues (Kir2.2 mutant 397 G178D and wild type D173) with predicted pKa values closest to cytoplasmic pH (Table   398 S3), we could model K + transport through the channel under conditions mimicking 399 different intracellular pH. In four systems containing various net negative charges 400 located within the inner cavity, we observed gradual expansion of the HBC region with 401 increasing net negative charge. Expansion was significant at the hydrophobic I177 402 residue (Fig. 7), resulting in higher K + occupancies at the HBC and in the inner cavity 403 (Fig. S3).

404
The recognition that titration of both G177E and D172 residues contributes to the 405 sub-state behavior in the highly homologous and archetypal Kir2.1 channel, shows that 406 this sub-state gating is a generalizable phenomenon, dependent on the presence of 407 titratable residues in the inner cavity. Clearly, the contribution of each residue is not 408 equal; this is most evidenced by there being still (at least) 4 distinct sub-states in g-E 409 channels (with 4 D172 and 2 G177E residues), but only three very distinct sub-states in 410 ng-nE (with 4 D172N and only 2 G177E titratable residues). One possibility to reconcile 411 these seemingly disparate behaviors is that the major contribution comes from the 412 G177E residues, and that the contribution of D173 residues is minimal, and may depend 413 on the presence or absence of G177E residues. Thus, four levels in E-E E-E channels 414 would be due to 0, 1, 2, or 3 (or 4) G178D residues being protonated, with the 415 conductance difference between 3 and 4 protonated G178D side chains being too small 15 to measure experimentally. The three conductance levels in ng-nE ng-nE channels 417 would then be due to 0, 1, or both G178D residues being protonated, and additional sub-418 states would emerge in the g-E g-E channels as D172 residues are additionally titrated.

421
We report striking sub-state gating behavior as a result of introduction of acidic residues

503
18 Minimum distances were measured between opposing protein subunits for all residues, 504 thus resulting in two distance pairs (chains 1-3 and 2-4). The corresponding distance-505 versus-time plots can be found in Figure S4.       MovieS1md16m844xD1734xG178Dcharged1usdf.mp4