Identification of PUFA interaction sites on the cardiac potassium channel KCNQ1

Figure 1.. Effect of PUFA on KCNQ1 channel. (A) Representative example of the effect of PUFA on the KCNQ1 channel, showing a shift in the voltage dependence for channel opening toward negative voltages (ΔV50) and an increase in the maximal conductance (ΔGmax). Curves represent Boltzmann fits to control (black line) and 20 µM of the PUFA LIN (blue). (B) Top view of the xKCNQ1 structure with the pore domain in gray and the voltage-sensing domain in black. xKCNQ1 residues R218 and K316 (corresponding to R228 and K326, respectively, in hKCNQ1) are highlighted in brown and purple, respectively. Previous studies identifying these residues as important for PUFA effects used N-AT and DHA-Gly (Liin et al., 2015; Liin et al., 2018). (C and D) Effect of LIN or STE at indicated free FA (FFA) concentrations on V50 (C) or Gmax (D) of WT hKCNQ1. Data shown as mean ± SEM; n = 5–8 per data point. Concentration-response curves for LIN were fitted using Eq. 2 with the Hill coefficient constrained to either 1 or −1 (see Materials and methods for details). ΔV50,max was constrained to −25 mV to make the fit more robust. ΔV50,max denotes maximal effect in ΔV50; Rel., relative.

Figure 2.. LIN and STE distribution around KCNQ1. (A and D) Left: 2-D structure of LIN (A), an 18-carbon chain with two cis double bonds, and STE (D), a fully saturated 18-carbon chain, overlaid with the Martini CG models of LIN and STE, respectively. Right: Running average over 500 ns of the number of LIN (A) and STE (D) molecules within 6 Å of the channel in the upper bilayer. (B and E) 2-D occupancy maps of LIN (B) and STE (E) head groups around KCNQ1 shown as mean occupancies during the interval of 35–50 µs of the simulation. 0.03 in color scale corresponds to a 3% occupancy (i.e., LIN head group present in 3% of all frames in B). Top-down view of the channel in ribbon representation to clarify LIN (B) and STE (E) enrichment regions. Densities at different layers along the z axis (red line in channel side view below the density maps). Arrows highlight LIN or STE enrichment at regions I and II. (C and F) 3-D contour map of the occupancy of LIN (C) and STE (F) around KCNQ1. Threshold values are set at 15% (mauve or yellow) and 20% (blue or purple) of the total time (a LIN molecule is present in C at least 15% and 20% of the frames, respectively). LIN enrichment regions I and II are…

Figure S1.. Lipid distribution around KCNQ1 in different systems. (A) An overview of the CG system offering both extracellular (left) and intracellular (right) views of the KCNQ1 channel embedded in the multicomponent membrane bilayer. The protein is displayed by the gray tube representation, and each lipid component is colored as detailed. (B) Contour maps of POPE (gray) and POPG (green), with threshold values at 26% (top row) and CHOL with a threshold value at 20% (bottom row) for the systems containing LIN or STE and the FA-free (FA-FREE) system. PIP2 is not present in the outer leaflet and was therefore not included in this analysis. (C) Radial distribution function of LIN (blue) and STE (purple) head groups around the KCNQ1 channel during the last 45 µs of the simulation. The radial distribution function g(r) allows a look into the distribution of particles by describing how the density varies as a function of distance from a reference particle. We observed LIN head groups peaking at ∼15 Å and 30 Å from the center of the channel, with each peak corresponding to regions I and II. The LIN and STE head group peaks differ by a factor of two or three.

Figure S2.. Channel stability in AA simulations. Backbone RMSD of the KCNQ1 channel relative to the initial structure for the four different AA simulations of the LIN and STE systems. ΔV50,max denotes maximal effect in ΔV50.

Figure 3.. Contact analysis for LIN and STE in CG simulations. (A) Histogram of total interaction time of each residue in KCNQ1 for LIN. The gray highlighted boxes in the background indicate the position of the transmembrane helices. The total interaction times reported cover the simulation period of 5–50 µs, since convergence of the number of LIN molecules within 6 Å of the channel is achieved after 5 µs into the simulation. The total interaction time for each residue has been averaged across the four channel subunits to account for the fourfold symmetry of the channel. (B) Box plots showing the total interaction time for the residues, with those that fall above Q3 + 1.5 interquartile range in the total interaction time (i.e., high-contact residues) labeled. (C) Heat maps of total interaction times in the 3-D structure of the channel shown both from a top and a side view. The unit for the color scale is in microseconds. (D–F) Same as in A–C but for longest lifetime of each residue. As with the total interaction time calculations, the longest lifetime measurements are based on the last 45 µs of the simulation and are averaged across the four channel subunits. (G–L) Same as in A–F but for…

Figure S3.. Contact analysis for inner leaflet LIN and STE in CG simulations. (A) Histogram of total interaction time of each residue in KCNQ1 for intracellular LIN. The gray highlighted boxes in the background indicate the position of the transmembrane helices. The residues identified as interacting with the LIN head group (i.e., R239 and R249) are colored according to E and F onto the histograms. (B) Same as in A but for longest lifetime of each residue. (C and D) Same as in A and B but for STE. (E and F) Side view of a LIN binder interacting with residues R249 and R239 located on opposite ends of the S4–S5 linker. (G) Experimental mutation of predicted binder residues does not impair the effect of LIN. Data shown as mean ± SEM; n = 5 per data point. Concentration-response curves were fitted using Eq. 2 with the Hill coefficient constrained to −1 (see Materials and methods for details). ΔV50,max was constrained to shared values between WT and mutant to make the fits more robust, and data are shown as normalized effect (100% was defined as the maximal effect from the fit). For 70 µM LIN, ΔV50 and ΔGmax were within ±1.2 mV and ±9%, respectively (P > 0.05 with one-way ANOVA followed by…

Figure 4.. LIN binds at interaction sites through a network of charged and hydrophobic residues. (A and B) 3-D contour maps of the occupancy of a LIN constant binder at site Iα (A) and site II (B) are shown from side view (left) and top view (right). Threshold values are set at 10% (blue) and 30% (pink) of the total time. PH denotes pore-helix. Potassium ions in the permeation pathway are indicated as green spheres. (C and D) Side view of a LIN constant binder interacting with the voltage-sensing domain at site Iα. (E) Overlay of all LIN constant binder poses at site Iα. The LIN head group is shown in pink and the LIN tail in blue. (F and G) Side view of a LIN constant binder interacting with the pore domain at site II. (H) Overlay of all LIN binder poses at site II. The occupancy of the membrane lipid head groups is shown for visual purposes to denote the position of the membrane bilayer. (I and J) Mutation of predicted binder residues impairs the effect of LIN. (I) Concentration dependence of the LIN effect on V50 of indicated hKCNQ1 site I mutants. Data shown as mean ± SEM; n = 4–8 per data point. Concentration-response curves were fitted using Eq. 2 with the Hill coefficient constrained…

Figure S4.. Contact analysis for AA LIN simulations. The LIN molecules that had a total interaction time in the upper quartile (UQ) were selected for the analysis and referred to in the text as the “LIN UQ binders.” The focus on LIN UQ binders enables direct elucidation of the long-term interactions and exclusion of random and short-lived interactions. There were in total 17 LIN UQ binders in each system. The total interaction time and longest lifetime measurements reported are averaged across the four channel subunits. We also tracked the “constant binder” LIN molecules for each specific site during the entire 5-µs simulation. There were in total two and three LIN constant binders in the LIN site I and LIN site II systems, respectively. (A) Left top: Histogram of total interaction time of each residue in KCNQ1 for LIN UQ binders in the LIN site I system. The gray highlighted boxes in the background indicate the position of the transmembrane helices. Left bottom: Heat maps of total interaction times in the 3-D structure of the channel shown both from top and side views. Right: Same as in A and B but for longest lifetime of each residue. The unit for the color scale is in microseconds. The…

Figure S5.. Pattern of LIN interaction with KCNQ1 in AA simulations is preserved in longer and initial seed unbiased simulations. (A) To assess whether the AA simulations we performed have allowed sufficient sampling of binding events, we continued the LIN site I system for an additional 10 µs for better sampling of the interaction sites. Contact analysis for the 15-µs AA LIN site I simulation. Histogram of total interaction time (left) and longest lifetime (right) of each residue in the transmembrane segment of KCNQ1 for LIN. The gray highlighted boxes in the background indicate the position of the transmembrane helices. The contact analysis performed for the entire 15-µs AA simulation follows the exact same trend as the initial 5 µs with the same identified constant binders keeping their close interaction with the channel. (B and C) To evaluate and rule out any potential convergence/initial seed system bias in AA MD simulations, a system in which no LIN molecules were in close proximity to sites I and II in the CG simulations, called LIN-dissociated, was selected as a seed system for one production run. In this control AA MD simulation, LIN molecules were allowed to diffuse in the…

Figure S6.. Contact analysis for LIN constant binders in AA simulations. (A) Histograms of total interaction time and longest lifetime of each residue within 6 Å of the LIN constant binders at site Iα, site Iβ, and site II. The identified binder residues (i.e., those residues that are in direct interaction with the LIN constant binder) are colored according to Fig. 4 onto the histograms. (B) Comparison of AA contact analysis between LIN and STE. Histograms of total interactions time and longest lifetime of each residue within 6 Å of the LIN and STE UQ binders at the identified interaction sites Iα, Iβ, and II. The identified binder residues are colored according to Fig. 4 onto the histograms. We distinctly see how the STE UQ binders fail to interact with the channel at site II, where no residues with high interaction time and lifetime were found. We do, however, observe STE to bind to both sites Iα and Iβ, albeit with notable differences in longest lifetime and total interaction time with residues from S5 compared with LIN.

Figure S7.. Binder residues at sites Iα, Iβ, and II. (A) Distances between the binder residues and the LIN (blue) and STE (purple) constant binders at site Iα are displayed across the 5-µs simulation period. (B and C) Same as in A but for site Iβ and site II, respectively. Variations in fluctuations across the distance measurements of the residues interacting with the STE and LIN constant binders are especially evident for the hydrophobic residues stabilizing the FA tails. In particular, the STE constant binders are unable to maintain a close interaction with the S5 residues Y268 and L272 in the site Iα (A). Overall, the lack of tail flexibility in STE and the stiffness of its tail prevents this saturated FA from interacting as efficiently as the more mobile unsaturated tail of LIN with the network of hydrophobic residues at site I or to fit into the more restricted cavity at site II.

Figure S8.. Representative pose of a LIN constant binder at sites Iα and Iβ. The second largest cluster at site Iα and the largest clusters at site Iβ identify overall comparable interactions between the LIN head group and R218 on S4 and a tail that sits either along S4 or more toward S3. Altogether, at either site Iα or Iβ, we observe how the LIN constant binder interacts with its head group at the top of S4, and its rather mobile tail swinging between S3, S4, and S5 results in a wide range of conformations. (A and B) Side view of a LIN constant binder interacting with the voltage-sensing domain at site Iα. This LIN pose is the representative pose from the second-largest cluster, representing 15% of all conformations. The LIN head group is in the same position as in the largest cluster, but with a tail that sits further along S4. (C and D) Side view of a LIN constant binder interacting with the voltage-sensing domain at site Iβ. This LIN pose is the representative pose from the largest cluster, containing 67% of all conformations. The binder residues at site Iβ include the basic residue R218 and the polar residue S215 on S4, which interact via salt-bridge formation and hydrogen bond with…