Conformational heterogeneity in closed and open states of the KcsA potassium channel in lipid bicelles

Abstract

The process of ion channel gating-opening and closing-involves local and global structural changes in the channel in response to external stimuli. Conformational changes depend on the energetic landscape that underlies the transition between closed and open states, which plays a key role in ion channel gating. For the prokaryotic, pH-gated potassium channel KcsA, closed and open states have been extensively studied using structural and functional methods, but the dynamics within each of these functional states as well as the transition between them is not as well understood. In this study, we used solution nuclear magnetic resonance (NMR) spectroscopy to investigate the conformational transitions within specific functional states of KcsA. We incorporated KcsA channels into lipid bicelles and stabilized them into a closed state by using either phosphatidylcholine lipids, known to favor the closed channel, or mutations designed to trap the channel shut by disulfide cross-linking. A distinct state, consistent with an open channel, was uncovered by the addition of cardiolipin lipids. Using selective amino acid labeling at locations within the channel that are known to move during gating, we observed at least two different slowly interconverting conformational states for both closed and open channels. The pH dependence of these conformations and the predictable disruptions to this dependence observed in mutant channels with altered pH sensing highlight the importance of conformational heterogeneity for KcsA gating.

Figure 1.

Histidine residues in KcsA structure. The KcsA crystal structure (PDB no. 1K4C) is shown highlighting the histidine residues labeled with ¹⁵N in the NMR experiments. H20 (not observed in the structure and indicated by asterisks) lies on the N-terminal amphiphathic helix, H25 is the pH sensor and is located on TM1, and H124 resides on TM2 at the C-terminal end of the bundle crossing (side chains not modeled).

Figure 2.

Assignments of H124 and H20 peaks in the [¹H,¹⁵N]HSQC spectrum of WT KcsA ΔC. (A) The [¹⁵N]HSQC NMR spectrum of [¹⁵N]histidine-labeled WT KcsA ΔC at pH 4.5 shows two intense peaks. (B) Lowering the contour levels of the spectrum reveals multiple peaks of varying intensity in the range for backbone amides. Resonance assignments were achieved by collecting similar [¹H,¹⁵N]HSQC on KcsA histidine mutants. (C) Superimposition of the HSQC spectrum of H124R KcsA ΔC (green) with the WT spectrum (purple) at pH 4.5, showing that four peaks correspond to H124. (D) Superimposition of the [¹⁵N]HSQC spectrum of H20A KcsA ΔC (red) with the WT spectrum (purple) at pH 4.5, showing that four peaks correspond to H20.

Figure 3.

Histidine peaks change with pH in the [¹H,¹⁵N]HSQC spectrum of WT KcsA ΔC. (A) pH-dependent chemical shift changes of the C-terminal H124 peaks (labeled peaks A and B) at pH 4.5 (blue), 5.5 (orange), 6.5 (green), and 7.5 (pink). Peaks A and B are observed through the pH range measured (pH 3–8). (B) Raw intensities plotted for C-terminal H124 peaks A and B as a function of pH. (C) pH-dependent chemical shift changes of the primary H124 peaks (labeled C1 and C2) at pH 3.0 (yellow), 4.5 (blue), 5.5 (orange), 6.5 (green), and 7.5 (pink). C1 and C2 peaks are marked with an arrow for pH 3.0 and 4.5. At pH 5.5, 6.5, and 7.5 only C2 is visible. (D) Raw intensities plotted for primary H124 peaks C1 and C2 as a function of pH. (E) pH-dependent chemical shift changes of the H20 peaks (labeled peaks D–F) at pH 3.7 (red), 4.0 (blue), 5.0 (black), and 5.5 (pink) showing their pH-dependent changes in chemical shift. (F) Raw intensities plotted for H20 peaks D–F.

Figure 4.

pH sensor KcsA mutants alter the pH dependence of the H124 peak intensities. Intensity ratios for C-terminal H124 peaks A/B (calculated from data shown in Fig. 3 B) are shown as a function of pH for WT, H25R, E118A, and E118A/E120A.

Figure 5.

L24C/R117C KcsA is cross-linked into the closed state and does not exhibit ion flux. (A) The crystal structure of L24C/R117C KcsA was solved to 3.4 Å and was nearly identical to WT in the closed state. Comparison of L24C/R117C (green) and WT (yellow) models superimposed onto the electron density of the cross-linked mutant at the bundle-crossing region is shown. Several residues are shown, including the mutated cysteine residues C24 and C117. The two cysteine sulfur atoms are 2.02 Å apart in the structure, consistent with a disulfide bond. It is possible that the disulfide bond was lost in the structure because of radiation damage and therefore not visible in the electron density. (B) Comparison of L24C/R117C and WT at the selectivity filter, with key residues noted. (C) SDS-PAGE analysis of the cross-linked state of L24C/R117C KcsA in the absence of reducing agent. The channel is a tetramer in the absence of cross-linking agent Cu-P (lane 1), but when boiled for 10 min, half of the sample runs at the monomer size (lane 2), and bands at dimer, trimer, and tetramer size are also observed. After incubation with 2 mM Cu-P for 30 min at room temperature, the channel is cross-linked and runs at the tetramer size (lane 3), remaining stable even after boiling (lane 4). The white line indicates that intervening lanes have been spliced out. (D) ⁸⁶Rb⁺ uptake through E71A/L24C/R117C (cross-linked, see Materials and methods) at pH 4.0 (open green circles) and 7 (open black triangles) compared with that through E71A at pH 4.0 (closed green circles) and 7 (closed black triangles), at the same protein to lipid ratio reconstitutions. Protein-free liposomes showed no activity at either pH 7.0 (open black squares) or 4.0 (open green squares). Because flux through E71A at pH 4.0 is so large, we show a y-axis zoom in Fig. S3 to better evaluate the data. Symbols and error bars represent mean ± SE for experiments repeated in triplicate. (E) A bar graph summarizing the ⁸⁶Rb⁺ uptake from A. Normalized mean Rb⁺ uptake after 5 min is plotted for E71A and E71A/L24C/R117C at pH 4.0 and 7.0. The normalized Rb⁺ uptake at 300 s for E71A is 0.26 ± 0.02 at pH 4.0 and 0.019 ± 0.003 at pH 7, and for E71A/L24C/R117C the normalized uptake at 300 s is 0.0036 ± 3 × 10⁻⁴ at pH 4.0 and 0.0026 ± 2 × 10⁻⁴ at pH 7. Error bars represent SE for three experiments. (F) Representative trace from a single-channel electrophysiology experiment in planar lipid bilayers (3:1 POPE/POPG) with L24C/R117C liposomes reconstituted in the presence of 1 mM TCEP. Experiments were performed at 100 mV and filtered at 1 kHz. See Materials and methods for details.

Figure 6.

L24C/R117C KcsA ΔC is cross-linked into the closed state. [¹H,¹⁵N]HSQC spectrum of L24C/R117C KcsA ΔC (red) superimposed onto the WT spectrum (black), both at pH 4.2, showing that H124 peaks A and C2 and two H20 peaks are missing in the spectrum of the cross-linked construct.

Figure 7.

Cardiolipin reveals open state of KcsA. (A) [¹H,¹⁵N]HSQC of KcsA in PC:CL bicelles. WT KcsA was reconstituted into PC:CL bicelles containing 15% cardiolipin. The [¹⁵N]HSQC spectrum in PC:CL bicelles (orange) reveals similarities and differences compared with the [¹⁵N]HSQC of KcsA in PC-only bicelles (black), both at pH 4.5. Additional peaks are observed in PC:CL conditions, identified as H25 by comparison with the [¹⁵N]HSQC spectrum of the mutant H20A/H124R KcsA (blue) at pH 4.5, which retains H25 as its sole histidine. (B) KcsA mutant that allows visualization of H25 with NMR exhibits a higher open probability (Po) in planar lipid bilayers. Liposomes of H20A/H124R KcsA (in the background of a noninactivating E71A mutant commonly used for electrophysiology) were applied to 3:1 POPE:POPG planar lipid bilayers. Mean Po at 100 mV is shown for pH 4.0, 5.0, and pH 5.5, comparing E71A/H20A/H124R (solid bars) with E71A (checked bars [Posson et al., 2013]). The bars represent the mean ± SEs for n = 3 experiments.