Importance of lipid–pore loop interface for potassium channel structure and function

We used an ssNMR-based hybrid strategy to establish 3D structural views of Chim before and after C-type inactivation. First, we constructed a homology model of the closed conductive state by using the crystal structure of KcsA (25) [Protein Data Bank (PDB) ID code 3EFF] and constructed a tetramer by using high ambiguity driven docking (HADDOCK) (26) (SI Methods). This model then served to predict and evaluate experimental CC and CHHC correlation experiments as the basis of 3D molecular structures (e.g., ref. 21). Experimental spectra were recorded on fully ¹³C,¹⁵N-labeled (Fig. S1) and on fractionally deuterated Chim (27). Resonance assignments were largely taken from published work (22). The monomeric structure of the closed pore domain (residues Chim 22–115), reconstituted in asolectin liposomes, was then calculated in Crystallography and NMR System (CNS) software (28). Distance restraints were mostly derived by comparing the experimental CHHC spectra acquired with three different mixing times (50, 250, and 500 µs) along with the CC restraints obtained from fractional deuteration studies. Overall, approximately 70% of all expected correlations were visible in CHHC spectra (SI Methods). Additionally, torsion angle restraints obtained from earlier chemical-shift assignments (22) were supplemented during structure calculation in CNS. The final structure, compatible with all experimental data, was characterized by more than 2,000 restraints, including 339 long-range distances (Table S1 and Fig. S2). Note that the pore loop region (residues Chim 44–90) was characterized by 156 long-range, 169 medium-range, and 148 sequential distance restraints.

The 3D structural backbone model of the closed channel state is depicted in Fig. 1 (Left), together with a selected set of CHHC distance restraints (Fig. 1, dashed lines). The overall channel architecture was in good agreement with the WT KcsA crystal structure (25). In detail, however, we observed significant differences with respect to the TM1 (S5) pore-loop region. First, TM1 is extended until residue Chim Asp53 and thus contains an additional turn as diagnosed by secondary chemical shifts seen earlier (8) and CHHC restraints obtained here (Fig. 1, Right). Second, the turret structure connecting TM1 and pore helix significantly differs from the one of the WT KcsA X-ray structure (Fig. 1, Right). The structural changes in this TM1-helix and turret region (henceforth “TM1T region”) along with intermolecular contacts detected in our study led to a widening of the pore (Fig. S3A). This observation supports our previous data indicating that reduced steric hindrance in a widened Chim pore facilitates binding of the scorpion toxin Kaliotoxin (KTX) at the extracellular mouth of the channel (29).

Next, we applied a similar ssNMR-based protocol to investigate structural details of open-inactivated state of the Chim channel in the lipid bilayer. Similar to WT KcsA (30), this state can be induced for the Chim channel by triggering inner gate bundle opening at low pH and reducing K⁺ concentrations to less than 20 mM (8, 9). Notably, further reduction to pH 3.7 did not change our Chim ssNMR spectra, even though macroscopic and single channel measurements suggest that the apparent activation pK_a is closer to pH 4.2 for WT KcsA (31). For the analysis of the ssNMR data, an initial 3D structural model was constructed by using torsion angle restraints in channel regions known to undergo conformational changes after inactivation (8, 9). The resulting monomer fold was, in the second stage, evaluated by using long-range correlations detected in the CHHC data. The final structure (Table S1 and Fig. S2B provide a summary of structural restraints) revealed distinct structural rearrangements in several domains of the channel compared with the closed-conductive state (Fig. 2A). First, we confirmed earlier backbone chemical-shift changes in the selectivity filter residues (8, 9). In line with X-ray results on constitutively open truncated KcsA mutants (10, 23), the selectivity filter in its inactivated state is reminiscent of that of a filter under low ionic conditions [PDB ID code IK4D (30)]. Second, the absence of the observed intermolecular correlation Ile100CB–Thr74CB observed in the closed conductive state (Fig. 2B) and the distinct chemical-shift changes in the transmembrane helix 2 (TM2)—together with earlier results using water-edited ssNMR spectroscopy (32)—suggest that the TM2 helix rotates during K⁺ channel opening, as proposed from KcsA X-ray structural analyses (23). As detailed in Fig. S3B, further analysis of our data suggests an opening of the activation gate by approximately 23 Å, as defined by the Cα–Cα intersubunit distance at positions WT/Chim Thr112 (23), and close to the physiologically relevant opening in the full-length channel (33).

Finally, we found that several medium- and long-range correlations observed for the turret region in the closed conductive state vanished after inactivation, e.g., the one comprising Chim Asp53 to Chim Thr85 (Fig. 2B), whereas new correlations appeared, e.g., the one comprising Chim Thr56 to Chim Pro83 (Fig. S4A). In the structural ensemble, these alterations in distance and dihedral angle restraints are characteristic of a destabilization and outward rotation of the extended TM1 helix accompanied by a loss of the multipoint hydrogen-bond network of Chim Glu71, Chim Asp80, and Chim Trp67 (Fig. 2A). Although a distinct set of residues in the turret and the pore helix were absent in our multidimensional correlation spectra (Chim Ala50, Glu51, Asp64, and Tyr82), the results clearly indicate that conformational changes in the turret region accompany Chim inactivation. Importantly, back-titration to pH 7 fully restores the closed conductive state (Fig. S4B).

Previously, we have shown that Chim Gly58 in the turret region is a key residue for C-type inactivation-sensitive binding of KTX (21, 29). The conformational changes we observed between the closed and the open-inactivated state suggested that bulky residues at position 58 not only interfere with KTX binding, but also with C-type inactivation. As the Kv1.3 channel exhibits a pronounced C-type inactivation, we tested the effect of bulky side chains at position G377 (equivalent to Chim Gly58) on C-type inactivation in the Kv1.3 channel. We mutated Kv1.3 residue Gly377 to alanine, valine, and phenylalanine, respectively (Fig. 2C) and recorded macroscopic currents in Xenopus oocytes with a two-electrode voltage clamp. Although current–voltage relations, time rise to peak, and deactivation kinetics of WT and mutant Kv1.3 channels showed no significant differences (Table S2), mutating Kv1.3 G377V and especially Kv1.3 G377F markedly slowed the time course (τ_inact) of Kv1.3 C-type inactivation [Kv1.3, τ_inact = 0.66 ± 0.02 s (SEM); Kv1.3 G377A, τ_inact = 0.82 ± 0.01 s; Kv1.3 G377V, τ_inact = 3.61 ± 0.20 s; Kv1.3 G377F, τ_inact = 28.6 ± 0.8 s; n = 3–5; Fig. 2C). Combining the structural and functional data suggests that bulky side chains at turret residue Chim Gly58 affect the closure of the inactivation gate at the selectivity filter.

The sequences of WT and Chim differ only by 11 aa in the turret region (Fig. 1). Structural changes observed between Chim in lipid bilayers and the crystal structure of WT KcsA could therefore be induced by changes in protein sequence and/or the presence of the lipid bilayer. To examine the influence of protein sequence, we investigated the structure of WT KcsA by ssNMR in lipid bilayers. Similar to the chimeric channel, we conducted a series of multidimensional ssNMR experiments to obtain resonance assignments and structural restraints for the WT pore loop region. We could readily assign turret residues specific for WT (Fig. S5) and generate a 3D model of WT in lipid bilayers by using a series of CHHC experiments (Fig. S6A). The resulting structure (Table S3) is shown in Fig. 3A in red, with specific medium to long-range restraints annotated in cyan and dotted lines. We compared our results vs. the 3D structure of Chim (Fig. 3A, Right, blue) and the X-ray structure of WT (Fig. 3A, Right, gray). Structures of the selectivity filter and TM1/TM2 helices are in good agreement. In contrast, we observed significant structural differences in the turret region, including an elongation of the TM1 helix for bilayer embedded WT and Chim channels. Notably, our ssNMR results are in good agreement with earlier electron paramagnetic resonance (EPR) work on WT KcsA in asolectin liposomes (31). The EPR data indicated that TM1 helix extends until WT Arg52.

Previous work has shown that mutations in the pore loop region can have a profound influence on inactivation gating of K⁺ channels (13, 34–36). We investigated the influence of mutations at the pore-helix position Glu71 on the pore loop structure by comparing ssNMR data on WT KcsA (Fig. 3B, red) to spectra obtained on E71Q (Fig. 3B, green) and E71A (Fig. 3B, black) mutants (i.e., WTom; full data in Fig. S6B). As expected (36), we observed chemical-shift variations at residue WTom Asp80 that is part of a multipoint hydrogen-bond network including Glu71 in WT (35) for WTom E71Q and E71A. Interestingly, we also observed a disappearance of the extended α-helix (as diagnosed by WTom Ala50) for both gating mode mutants. Additional turret residues including WTom Ile60 and WTom Pro55 exhibited chemical shift changes and peak doubling, respectively (indicated in orange on the 3D structure in Fig. 3B). These spectral variations extended to the pore helix (WTom Thr72) and residues in the selectivity filter (WTom Thr74/Thr75). Cross peaks for selectivity filter residues differed by approximately 2 ppm compared with correlations seen for WT KcsA in the conductive state (Fig. 3B, red). These observations would be compatible with increased molecular mobility around WTom Thr74/WTom Thr75. Indeed, previous MD simulations showed a considerable increase in the frequency and lifetime of WT Val76 reorientation in the WT E71Q mutant correlated with short-lived flicker states in WT E71Q single-channel recordings (36). Taken together, our results on KcsA are consistent with the view that the pore loop region functions as a functional unit that includes the extended TM1 helix coupled via the turret to the inactivation gate and extending all the way to the pore–TM2 interface. This view is in remarkable agreement with previous mutational studies in Shaker Kv channels (13).

To study the molecular details that determine the influence of lipids on the pore loop structure, we resorted to MD simulations for WT and Chim (Fig. 4) embedded in lipid bilayers. In the case of WT (Fig. 4A), we observed the formation of an intermolecular hydrogen bond between the WT Arg52 amidine group and WT I60CO (Fig. 4, orange lines) as well as intermolecular protein–lipid interactions (Fig. 4, magenta lines) that we suggest stabilize the helical elongation of the TM1 helix up to residue WT Gly53. The spatial proximity of the WT Arg52 amidine group to WT Ile60 was verified (Fig. 4, red lines) by a tailored ssNMR rotational resonance experiment (37) (Fig. S7). During the MD trajectory (Fig. 4B), we observed a remarkable correlation between helical propensity of residues WT Ala50 to WT Gly53 (Fig. 4, green) and the presence of hydrogen bonding between the WT Arg52 amidine group and WT Ile60CO (Fig. 4, orange). This view was supported by ssNMR data on the WTom E71A/Q mutants lacking the TM1T helix and exhibiting chemical-shift changes at WTom Ile60 (Fig. 3B). In the case of Chim, in which residues 53 and 54 are mutated (Figs. 1 and 4 C and D), we again observed intermolecular protein–lipid electrostatic interactions, namely between negatively charged Chim Asp53 and Chim Asp54 side chains and positively charged phosphatidyl-choline headgroups (Fig. 4, yellow lines) that stabilize the helical elongation of the TM1 helix up to residue Chim Asp53. Neutralizing Asp-53 and 54 (D53A, D53AD54A; Fig. 4D) strongly reduced α-helical propensity of residues 50 to 53, compared with Chim (Fig. 4D, green columns).

Expression, purification, and reconstitution into liposomes was done as described previously (21, 24). Reconstitution was performed at a 100/1 lipid/channel molar ratio. pH titrations were performed by thoroughly washing the proteoliposome pellet with the desired phosphate or citrate buffers, followed by 30 min ultracentrifugation at 125000 × g and +4 °C (9).

ssNMR experiments were conducted by using 3.2-mm or 4-mm triple-resonance (¹H,¹³C,¹⁵N) magic-angle-spinning (MAS) probe heads at static magnetic fields of 14.1, 16.5 and 18.8 T corresponding to proton resonance frequencies between 600 and 800 MHz (BrukerBiospin). CHHC (e.g., ref. 21) experiments were recorded at an effective sample temperature of 243 K. The CC correlation experiments were performed with an effective sample temperature varying between 273 K and 280 K. MAS frequencies used were 9.375 kHz (at 600 MHz and 700 MHz) and 12 kHz (800 MHz) for CHHC spectra and 10.92 kHz (700 MHz) for (¹³C,¹³C) proton-driven spin diffusion (PDSD) and phase-alternated recoupling irradiation scheme spin diffusion experiments performed under weak coupling conditions (43). A ¹H field strength of 83.3 kHz was used for 90° pulses and small phase incremental alternation 64 (44) decoupling during evolution and acquisition. CHHC experiments were acquired with ¹H-¹H mixing times of 50, 250, and 500 µs by using an initial cross-polarization (CP) step of 700 µs and CP steps of 80 µs enclosing the proton mixing step to select for one bond (C-H) magnetization transfer. Mixing times of 20 ms or 30 ms were used to obtain intraresidue (¹³C,¹³C) correlations in PDSD experiments.

MD simulations were carried out using the Groningen Machine for Chemical Simulations (GROMACS) simulations package, version 4.5.3 (45), with the Groningen Molecular Simulation computer program package (GROMOS53a6) force field (46). The simulation systems were represented by the atomic models of Chim and WT KcsA, embedded in a bilayer in an aqueous solution of KCl. Potassium and chloride ions were added to electrostatically neutralize the systems and to mimic 80 mM KCl solutions. The KcsA starting structure was derived from crystal structure 3EFF (PDB) (25) to probe spontaneous helical elongation of TM1, the Chim starting structure from our ssNMR-based structure model. Mutations in the Chim model were introduced with Pymol (Delano Scientific). All systems were simulation at constant-pressure for at least 25 ns.

The expression of Kv1.3 WT and mutant channels in the Xenopus oocyte expression system was as described previously (31). Activation, deactivation time courses, and voltage-conductance relations were measured with test pulses of 100 ms starting from a holding potential of −100 mV. For measurement of Kv1.3 inactivation, we used long test-pulse durations of 60 s at test potentials of +60 mV, where Kv1.3 channels are completely activated. Inactivation time courses and recovery from inactivation were evaluated by using HEKA-PULSEFIT in combination with KaleidaGraph software. Statistical significance was tested by Student t test.