Structural Determinants Mediating Tertiapin Block of Neuronal Kir3.2 Channels

Abstract

Tertiapin (TPN) is a 21 amino acid venom peptide from Apis mellifera that inhibits certain members of the inward rectifier potassium (Kir) channel family at a nanomolar affinity with limited specificity. Structure-based computational simulations predict that TPN behaves as a pore blocker; however, the molecular determinants mediating block of neuronal Kir3 channels have been inconclusive and unvalidated. Here, using molecular docking and molecular dynamics (MD) simulations with ‘potential of mean force’ (PMF) calculations, we investigated the energetically most favored interaction of TPN with several Kir3.x channel structures. The resulting binding model for Kir3.2-TPN complexes was then tested by targeted mutagenesis of the predicted contact sites, and their impact on the functional channel block was measured electrophysiologically. Together, our findings indicate that a high-affinity TPN block of Kir3.2 channels involves a pore-inserting lysine side chain requiring (1) hydrophobic interactions at a phenylalanine ring surrounding the channel pore and (2) electrostatic interactions with two adjacent Kir3.2 turret regions. Together, these interactions collectively stabilize high-affinity toxin binding to the Kir3.2 outer vestibule, which orients the ε-amino group of TPN-K21 to occupy the outermost K⁺ binding site of the selectivity filter. The structural determinants for the TPN block described here also revealed a favored subunit arrangement for assembled Kir3.x heteromeric channels, in addition to a multimodal binding capacity of TPN variants consistent with the functional dyad model for polybasic peptide pore blockers. These novel findings will aid efforts in re-engineering the TPN pharmacophore to develop peptide variants having unique and distinct Kir channel blocking properties.

Graphical Abstract

Venom-derived peptides serve as effective tools for interrogating both the physiological and pathophysio-logical role of their molecular targets.¹ Sculpted over millennia by natural selection, these small peptides produced by a variety of venomous species display high affinity and target specificity that make them ideal tools for preclinical research that assesses druggable targets.^2,3 Moreover, advances in peptide therapeutics are creating new opportunities for engineering venom peptides for clinical applications that include immune suppression for rheumatoid arthritis and management of chronic pain.^4–7

Inwardly rectifying potassium (Kir) channels are ubiquitously involved in the regulation of K⁺ transport in excitable and nonexcitable cells, with several implicated in a variety of human diseases.⁸ Currently, there are relatively few pharmacological agents with high affinity and target selectivity available to assess the role of specific Kir channels and their varied functions.^8,9 One notable exception, however, is tertiapin (TPN), a 21 amino acid peptide present in the venom of the European honey bee.¹⁰ TPN exhibits nanomolar potency but limited specificity, inhibiting multiple Kir channel isoforms (Kir1.1 and Kir3.x) as well as K_Ca channels.^11,12 We previously reported structure-based computational docking methods to explore the molecular basis for the high-affinity TPN block of Kir1.1 channels.¹³ These studies produced an in silico model of the Kir1.1-TPN complex that agreed with alanine-scanning and site-specific mutagenesis studies identifying key determinates, mediating nanomolar binding and the functional Kir1.1 channel block.^14,15

Here we explored the molecular basis for TPN block of neuronal G protein-gated inwardly rectifying potassium channels (Kir3.x or GIRK) that mediate slow inhibitory postsynaptic potentials in the nervous system following activation of Gi/o-coupled neurotransmitter receptors.^16,17 Neuronal Kir3.x channels assemble primarily as heteromeric assemblies of Kir3.x subunits, where gene ablation studies in mice highlight a significant role of the Kir3.2 subunit in several neural functions including seizure propensity, motor activity/coordination, respiratory control, and sensory processing.^18–24,24 We applied structure-based molecular docking with molecular dynamics (MD) simulations to identify the energetically favored contacts mediating the Kir3.x channel block by TPN and to calculate its binding free energy. The predicted molecular determinants were then assessed in vitro to validate their functional impact on the Kir3.2 channel block by TPN. The findings reported here will help guide future re-engineering efforts to design TPN-inspired variant peptides exhibiting improved potency and selectivity that target heteromeric assembled Kir channels.

MATERIALS AND METHODS

Modeling the TPN Peptide Structure.

The NMR solution structure of TPN, including all 21 lowest energy conformers (PDB ID: 1TER), was used for structure-based docking to Kir3 channels. The tertiary TPN peptide structure is stabilized by two disulfide bonds formed between C5–C18 and C3–C14²⁵ as shown in Figure 1A. There are five basic residues with solvent-exposed side chains (4 lysine and 1 arginine), having a net charge of 5e at pH 7.0. Replacement of the C-terminal K21 residue with an alanine residue, TPN-K21A, was performed in silico using the MODELLER program.²⁶ TPN residues 4–7 form a type-I reverse turn, and residues 12–19 form an α-helix. The methionine residue (M13) has previously been shown to be oxidized in the air resulting in a 4- to 5-fold decrease in the inhibitory action of TPN on heteromeric Kir3.1/Kir3.4 channels.²⁷ Substitution with glutamine, TPN-M13Q (TPN_Q), precludes oxidation, and the TPN_Q peptide variant has an IC₅₀ value similar to the nonoxidized TPN peptide.²⁷ For our computational simulations, we examined the binding energetics of TPN (non-oxidized) and not TPN_Q.

Figure 1.

Molecular and structural features of the TPN peptide and Kir3.2 channel. (A) Amino acid sequence and NMR structure of TPN (PDB ID: 1TER). The side chains of the 5 basic residues are shown in cyan sticks. The two disulfide bonds C3–C14 and C5–C18 are shown as yellow sticks. (B) Sequence alignment of mouse Kir3.x subunit amino acids forming the outer vestibule of Kir3.x channels. The structure of the Kir3.2 homotetramer is depicted (PDB ID: 3SYA), including a surface rendering of the outer vestibule region that serves as the target for TPN interaction and binding (individual channel subunits are color-coded).

Modeling Kir3 Channel Structures.

The crystal structure of the mouse homotetrameric Kir3.2 channel (PDB ID: 3SYA) served as the structural template for homology modeling of Kir3 channels.²⁸ Because the Gβγ-bound crystal structure of Kir3.2 (PDB ID: 4KFM) yields similar rigid-body TPN docking results as the unbound structure,²⁹ and studies of the native cardiac K_ACh channel (i.e., Kir3.1/Kir3.4 channel) indicate channel opening by receptor-activated Gβγ dimer interactions is not required for TPN to block the channel,³⁰ we limited the Kir3 homology modeling to the unbound Kir3.2 channel conformation (i.e., 3SYA).

The amino acid sequence identity for residues forming the outer vestibule of Mus musculus and Homo sapiens Kir3 channels is 100%, so the mouse channels are expected to recapitulate their human channel orthologs structurally. The sequence alignment of Kir3.x subunit residues forming the outer toxin-binding vestibule is shown in Figure 1B, which includes the channel turret and pore entry regions. There is greater than 70% amino acid homology between Kir3.x subunits, and thus, reliable models of Kir3.x isoforms can be generated using the Kir3.2 structure as a template. Homology models for homotetrameric Kir3.1 and Kir3.3 channels were generated using MODELLER by threading the aligned Kir3.1 and Kir3.3 sequence on the corresponding sequence of the Kir3.2 crystal structure.²⁶

To validate the stability of the Kir3.x homology models, MD simulations were performed in a lipid bilayer. Briefly, the crystal structure of Kir3.2 and the homology modeled homomeric channels for Kir3.1 and Kir3.3 were placed in a lipid bilayer composed of 228 POPE (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) molecules in the x–y plane. All three Kir3 channel systems (Kir3.1, Kir3.2, Kir3.3 homotetrameric channels) were then solvated and ionized with a 150 mM KCl aqueous solution. Three K⁺ ions were placed in the selectivity filter and pore cavity, as observed in the original Kir3.2 crystal structure. The Kir3.x systems were prepared using the VMD software with a box size of 93 × 92 × 107 ų and equilibrated with pressure coupling until the correct water and lipid densities were obtained.³¹ The x and y dimensions were then fixed, and pressure coupling was applied only in the z-direction. Next, the restraints on the protein atoms were reduced gradually by first reducing those on the side-chain atoms from k = 30 kcal/mol/Ų to 0 kcal/mol/Ų in 3 ns. Then, the backbone atoms were relaxed in a similar manner. All three Kir3 channel systems were simulated for 50 ns to check the overall stability of the models. The RMSD of the backbone atoms of the homomeric Kir3.1, Kir3.2, and Kir3.3 channel structures showed a plateau after the first few nanoseconds and remained stable throughout the MD simulations, confirming their structural stability.

Heterotetrameric Kir3.1/Kir.3.2 channels were assembled using the modeled Kir3.1 subunit and Kir3.2 subunit from the crystal structure with a 2:2 subunit stoichiometry. Two different heterotetramers were constructed based on two different subunit arrangements around the central pore, Kir3 [1–1–2–2] and Kir3[1–2–1–2]. Both tetramers were then equilibrated in the lipid bilayer by MD, as described for the homotetramers.

Modeling the Kir3.x-TPN Complex.

The primary binding pose of TPN with the Kir3.1, Kir3.2, and Kir3.3 homomeric channel systems was first generated by molecular docking using the HADDOCK program.^32,33 HADDOCK has previously been used successfully to study the binding of various toxins to sodium and potassium channels.^34–39 All 21 lowest energy NMR conformers of TPN were used in ensemble docking to sample the side-chain orientations properly, as individual TPN conformers have previously been shown to influence rigid-body docking to Kir1.1 channels.¹³ We first performed unbiased “blind” docking on Kir3.2 to determine whether a basic residue inserted into the selectivity filter to potentially block the pore (i.e., pore-inserting lysine). From the “greedy” cluster analysis of binding poses, the strongly interacting residues were identified. These pairwise interactions were then used for the next stage in restraint-based docking.

For the proof-of-concept of the unbiased blind docking, and for comparison with the previous computational studies of TPN binding to Kir3.2,^40,41 we also performed four separate restraint-based docking calculations with the Kir3.2 homotetramer, where each of the four different lysine residues of TPN was constrained as the pore-inserting lysine. In all of the docking calculations, 10 000 complexes were generated and the top 200 selected based on their docking score for further analysis of pairwise interactions. After the pore-inserting lysine was identified with docking and MD simulations of the Kir3.2-TPN complex, restraint docking was also performed in docking TPN to the Kir3.1 and Kir3.3 homomeric channels and then heteromeric Kir3.1/Kir3.2 channels. In each case, the best complex was selected according to the energy score, the number of pairwise contacts made, and the blocking of the channel by TPN.

For MD refinement of each complex, the Kir3.x-TPN structure obtained from the docking was aligned with the corresponding channel model in the membrane, and the coordinates of the toxin were transferred to the channel model. All Kir3.x-TPN complex structures were then equilibrated using the same protocols used in the relaxation of the channel protein in the membrane. After equilibration, each system was simulated for 50 ns to check the stability of the complex as well as the binding interactions through the analysis of the trajectory data. During the MD simulations of the complexes, a small restraint of 0.1 kcal/mol/Ų was applied to the Cα atoms of the channel residues to preserve the integrity of the channel structure. Two K⁺ ions were placed in the selectivity filter and remained in the S2 and S4 K⁺ binding sites during the MD simulation of the complexes. The trajectory data generated from the MD simulation were then used in characterizing the TPN-bound complexes.

MD Simulations and PMF Calculations with Umbrella Sampling.

MD simulations were performed using the NAMD code (version 2.11) with the CHARMM36 force field.^42,43 The periodic boundary conditions and an NPT ensemble were used in the simulations. The pressure was kept at 1 atm and temperature at 300 K using Langevin coupling with damping coefficients of 5 ps⁻¹. Lennard-Jones interactions were switched off smoothly within distances of 10–13.5 Å. Electrostatic interactions were computed without truncation using the particle-mesh Ewald algorithm. A time step of 2 fs was employed in all MD simulations, where the trajectory data was saved at 1 ps intervals, and the reaction coordinates were recorded at every time step in the umbrella sampling simulations.

The PMF for dissociation of TPN from the Kir3.2 homotetramer was constructed using umbrella sampling MD simulations. The method for the PMF calculations was described in detail previously and has been successfully applied to the study of other ion channel toxins.^7,34–36,39 Briefly, the reaction coordinate is chosen as the distance between the center of masses of the channel protein and the toxin along the channel axis. Initially, 30 umbrella windows are created along the channel axis at 0.5 Å intervals using steered MD with a force constant k = 30 kcal/mol/Ų and a pulling velocity of 5 Å/ns. Six more windows were then subsequently added to ensure that the toxin had reached the bulk region indicated by a steady-state PMF value. The same force constant of k = 30 kcal/mol/Ų was used in umbrella sampling simulations, which is found to be optimal for toxins of similar size. Because the overlapping density distributions between the neighboring windows should be >5% to avoid numerical instabilities in construction of the PMF, two additional windows were included where the overlap was less than 5%. The reaction coordinates collected from the simulations were unbiased using the weighted histogram analysis method (WHAM), which combines the individual PMFs optimally.⁴⁴ The umbrella sampling simulations were continued until the convergence of the PMF was achieved from the block-data analysis of the PMF data.

The TPN binding constant (K_eq) was then calculated by integrating the PMF, W(z) along the z-axis from the binding site to the bulk solution, using eq 1:

$$K_{\text{eq}}=\pi R^2{\int }_{\text{site}}^{\text{bulk}}{\text{e}}^{-W\left(z\right)/k_{\text{B}}T}\text{d}z$$ (1)

The πR² value provides a measure of the binding pocket cross-sectional area determined from the MD simulations, and k_B is the Boltzmann constant.^38,45 The free energy (ΔG_b) for the TPN-bound Kir3.2 channel complex was then determined using eq 2:

$$\Delta G_{\text{b}}=-k_{\text{B}}T\text{ln}\left(K_{\text{eq}}{C}_0\right)$$ (2)

where C₀ is the standard concentration of 1 M.

Kir3 Channel Expression in Xenopus Oocytes.

The in silico structural predictions were tested in vitro by measuring TPN_Q inhibition of Kir3 channels heterologously expressed in Xenopus oocytes.^46,47 TPN_Q or TPN_Q variant peptides were tested on heteromeric Kir3.1/Kir3.2 channels because they represent the predominant neuronal Kir3 channel expressed in the mammalian nervous system.^17,48 Moreover, the homomeric Kir3.2 channel (e.g., in the crystal structure) poorly couples to GPCR’s and yields relatively small basal and receptor-activated Kir3 currents.^49–51 Activation of the Kir3.2 homomeric channel requires Gβγ overexpression that produces single-channel properties that are unlike native neuronal Kir3 channels and the Kir3.1/Kir3.2 heterotetramer.^52,53 Due to these confounding issues with Kir3.2 homotetramer, we focused our structure–function comparisons on the Kir3.1/Kir3.2 heteromer.

Ovaries from Xenopus laevis were shipped overnight on ice (Xenopus 1, Dexter, MI) and then collagenase-treated at room temperature to obtain individual oocytes as described previously.⁴⁶ Isolated stage V–VI oocytes were selected and maintained for up to 7 days at 17–18 °C in the following Oocyte Ringer’s medium (in mM): 82.5 NaCl, 2.5 KCl, 1.0 CaCl₂, 1.0 MgCl₂, 1.0 NaHPO₄, 5.0 HEPES, 2.5 Na pyruvate, pH 7.5 (NaOH), with 2% heat-inactivated horse serum.

Oocytes were injected (50 nl) with a mixture of cRNAs encoding the rat Kir3.1 subunit (UniProtKB P63251) (0.5 ng/oocyte) and the mouse Kir3.2a subunit (UniProtKB P48542) (0.5 ng/oocyte). Site-directed mutagenesis of the mouse Kir3.2a subunit cDNA (E127K, E127R, and Y159A sub-stitutions) was performed by oligonucleotide-directed PCR and confirmed by DNA sequencing (Mutagenex, Columbus, OH). For some experiments, cRNA encoding the human muscarinic m2 receptor (UniProtKB P08172) was included (0.5 ng/oocyte) to test for receptor-activated mutant Kir3 channel currents via bath application of acetylcholine (ACh). For other experiments, cRNAs encoding the bovine Gβ1 subunit (UniProtKB P62871) (5 ng/oocyte) and bovine Gγ2 subunit (UniProtKB P63212) (5 ng/oocyte) were included instead of the m2 receptor. Gβ1γ2 coexpression provided direct constitutive channel activation and thus bypassed m2 receptor desensitization processes during the TPN_Q concentration–response experiments.

The cRNA-injected oocytes were incubated for 3–5 days at 17–19 °C prior to electrophysiological recording. All cRNAs were synthesized in vitro by T7 or T3 RNA polymerase from the linearized cDNAs (mMessage mMachine, Ambion, Austin, TX).

Two-Electrode Voltage-Clamp Electrophysiology.

Kir3.1/Kir3.2 channel currents were recorded from oocytes using the two-electrode voltage-clamp technique.⁴⁶ Oocytes were initially superfused with ND98 solution (in mM): 98 NaCl, 1 MgCl₂, and 5 HEPES at pH 7.5 (NaOH). Glass electrodes having tip resistances of 0.8–1.0 MΩ were used to voltage-clamp oocytes at a holding membrane potential of −40 mV (GeneClamp 500B, Axon Instruments). A voltage protocol was evoked every 1 or 5 s and consisted of a step change to −80 mV (50 ms in duration), followed by a ramp from −80 to +20 mV (200 ms in duration).

After establishing baseline currents during the first few minutes of recording, the bath solution was changed to a “high K⁺ solution” that consisted of (in mM) 20 KCl, 78 NaCl, 1 MgCl₂, and 5 HEPES at pH 7.5 (NaOH). Large inwardly rectifying K⁺ currents were associated with the transition to high K⁺ solution attributed predominantly to constitutively active “basal” Kir3.1/Kir3.2a channel currents. Application and washout of ACh and the various TPN_Q peptide concentrations in the 20 mM K⁺ solution were performed by a perfusion barrel positioned next to the recorded oocyte for rapid solution exchange.⁴⁶ All recordings were performed at room temperature (21–23 °C). The membrane currents were digitized using an A/D acquisition board (Digidata 1440 acquisition system, Axon Instruments) and then later analyzed using pCLAMP software (version 10.6).

TPN_Q and Variant Peptides.

TPN_Q (>99% purity) (TPN UniProtKB P56587) was obtained from Tocris Bioscience (Ellisville, MO). TPN_Q variant peptides were custom synthesized in a similar manner by solid-phase peptide synthesis (SPPS) (Peptides International, Louisville, KY). All TPN_Q peptides incorporated the two disulfide bridges using thiol-protecting cysteine N^α−9-fluorenylmethyloxycarbonyl (Fmoc) chemistry. The folded peptides were purified (>95% purity) by reversed-phase high-performance liquid chromatography (HPLC), with molecular mass confirmed by mass spectroscopy. Purified peptides were shipped lyophilized in trifluoroacetate (TFA) salt, then dissolved in water, and stored at −23 °C as concentrated stock solutions before use. Final peptide concentrations were determined by UV spectrophotometry, measuring sample absorbance at a 280 nm wavelength (Nanodrop 2000, ThermoScientific) and an extinction coefficient with the molecular weight calculated for each peptide (ProtParam tool, Swiss Institute of Bioinformatics).

Electrophysiological Data and Statistical Analysis.

The TPN_Q peptide concentration producing 50% inhibition of the G_β1γ2-activated Kir3.1/3.2 current (IC₅₀) was obtained by fitting the fractional inhibition (I_K/I_K,max) associated with increasing peptide concentrations, using the following Hill function (eq 3);

$$I_{\text{K}}/I_{\text{K},\text{max}}=1+{\left(\left[{\text{TPN}}_{\text{Q}}\right]/{\text{IC}}_{50}\right)}^h$$ (3)

where h is the Hill coefficient reflecting the slope of the concentration–response curve. I_K,max represents the current amplitude at −80 mV evoked by 20 mM K⁺ in the absence of peptide, and I_K represents the steady-state current amplitude at −80 mV in the presence of the stated TPN_Q peptide concentration. The contribution of endogenous “background” currents evoked by 20 mM K⁺ (measured in uninjected oocytes) was significantly less than the much larger G_β1γ2-activated Kir3.1/3.2 current amplitudes and was not adjusted for in the analysis.

Statistical comparisons between the various experimental groups were performed by one-way ANOVA, where P < 0.05 was considered significant. Experiments were each replicated in 7–9 oocytes from two or more frogs.

RESULTS

MD Binding Modes of TPN-Kir3 Channel Complexes.

From blind docking of TPN to the Kir3.2 crystal structure and RMSD greedy clustering analysis of the poses, TPN K21 was identified as the favored pore-inserting lysine whose side chain interacted with the carbonyl oxygens of Kir3.2 Y157 in the selectivity filter formed by the TIGYG motif. The interaction of TPN-K21 with Y157 obtained from the initial TPN-Kir3.2 docking was used as a restraint in subsequent docking calculations to refine and improve the binding mode. To test the uniqueness of this TPN-Kir3.2 channel complex and compare it with others, three additional docking simulations were performed restraining either K16, K17, or K20 as the pore-inserting lysine. The dominant pose was similarly selected from the HADDOCK clustering analysis, which was then refined by MD simulations for 50 ns as described in the Methods. A quantitative analysis of the protein–protein interface interactions in each binding mode is presented in Table 1, where the average distances between pairs of interacting residues are given for all four complexes. The four different TPN-Kir3.2 channel complexes were named according to the assigned pore-inserting TPN lysine: K16, K17, K20, and K21.

Table 1.

MD Identified Pairwise Contacts in the Kir3.2-TPN Complexes^a

Kir3.2	K21	MD	K16	MD	K17	MD	K20	MD
Y157–O(ABC)	K21–NZ	2.85	K16–NZ	2.89	K17–NZ	2.99	K20–NZ	2.96
E127–O1–2(A)	K16–NZ	3.50	K21–NZ	2.68	K21–NZ	2.57
E127–O1(B)							K17–NZ	2.55
E127–O1–2(D)	R7–N1–2	2.65
Y159–Cδ(A)	W15–CZ	3.44
Y159–Cε(C)	I9–CZ	4.33
Y159–Cδ(C)	I9–Cγ	4.84

^aThe individual Kir3.2 subunits are indicated by A, B, C, and D in parentheses. The four restrained docking results are given in columns K21, K16, K17, and K20, followed by the MD distance results in each case. The average N–O and C–C distances obtained from the MD simulations are given for each TPN complex in Å.

As indicated in Table 1 and illustrated in Figure 2, the K21 binding mode complex had the most interaction contacts between TPN and Kir3.2 when compared to the K16, K17, and K20 complexes. In all four TPN-Kir3.2 complexes derived from the restraint docking, the amide group of the pore-inserting TPN lysine interacted with three of the four Kir3.2 Y157 carbonyl oxygens (Kir3.2 subunits A, B, and C). Another shared interaction with all four docked complexes was the formation of a salt bridge between one of the Kir3.2 turret E127 side chains, and a TPN lysine side chain (either K16, K17, or K21). For the K16-, K17-, and K20-inserting binding modes, no other electrostatic or hydrophobic interactions were present (Figure 2A–C). This was in contrast to the K21-inserting complex shown in Figure 2D, where there was an additional salt bridge contact between an adjacent Kir3.2 turret, providing two separate turret contacts (E127 subunit A with TPN K16, and E127 subunit D with TPN R7).

Figure 2.

Binding modes of TPN with the Kir3.2 channel for the four MD-derived complexes. (A) K16 binding mode. (B) K17 binding mode. (C) K20 binding mode. (D) K21 binding mode. For the K16, K17, and K20 binding modes, only one cross section (i.e., 2 channel subunits) is shown as there were no interactions in the other Kir3.2 subunits. For the K21 binding mode of TPN with the Kir3.2 channel, the hydrophobic interactions with Kir3.2 subunits A and C are indicated with a circle in the upper panel. Additional interactions with Kir3.2 subunits B and D are then illustrated in the lower panel. The respective pairwise binding interactions in each binding mode are shown with green sticks. Two K⁺ ions are indicated with golden spheres and located in the K⁺ selectivity filter at positions S2 and S4.

In addition, hydrophobic contacts were evident and involved the pore entry tyrosine “ring”, where Y159 side chains from Kir3.2 subunits A and C formed interactions with the hydrophobic TPN I8, I9, and W15 residues (Figure 2D). In total, these additional TPN-Kir3.2 subunit contacts in the K21 complex involved three of the four Kir3.2 subunits and produced the most stable binding mode compared to either of the K16-, K17-, and K20-inserting complexes. The K21 binding mode was also in agreement with the previously reported interaction of TPN with rat homotetrameric Kir1.1 channels that exhibit nanomolar binding affinity.¹³

After identifying the K21 binding mode as the most energetically favored Kir3.2-TPN complex, we next performed restraint docking of TPN with the homology modeled Kir3.1 and Kir3.3 channels using K21 as the pore-inserting lysine. For homotetrameric Kir3.1 channels, the MD simulations indicated that TPN initially had interactions only with the selectivity filter, i.e., the side chain of K21 with the carbonyl oxygens of Kir3.1-Y148 and dissociated from the channel within 10 ns (data not shown). In contrast to Kir3.2, there are no acidic residues present in the Kir3.1 turret region to form interactions with TPN basic residues (cf. Figure 1B). Thus, the absence of any other interactions with the turret region is responsible for the instability of the Kir3.1-TPN complex.

For homotetrameric Kir3.3 channels, docking of TPN did not produce any binding poses with any of the TPN lysines as the pore-inserting residue. Comparison of the sequence of Kir3.2 versus Kir3.3 indicated Kir3.3 has a GYGH motif instead of a GYGY motif at the pore entry site and yet does possess an equivalent glutamic acid residue in the turret region as found in the Kir3.2 and Kir3.4 subunits for electrostatic TPN interactions (cf. Figure 1B). In Kir3.2 channels, because the Y159 side chain is involved in hydrophobic interactions with the TPN I8, I9, and W15 side chains, the equivalent Kir3.3 H125 site may not support similar hydrophobic interactions, and/or the Kir3.3 turret E93 residue may not be available as in Kir3.2 due to a nearby proline residue in the Kir3.2 turret region. These MD findings together indicate that the Kir3.2 subunit is essential for high-affinity TPN binding to neuronal heteromeric Kir3 channels, analogous to the Kir3.4 subunit in cardiac heteromeric Kir3.1/Kir3.4 channels.⁵⁴

Binding Free Energy of the Kir3.2-TPN Complex.

The binding free energy of the Kir3.2-TPN complex was next calculated to check the reliability and accuracy of the proposed K21 binding mode. The PMF for the dissociation of TPN from the homotetrameric Kir3.2 channel was constructed using umbrella sampling simulations as described in the Methods. A major caveat in PMF calculations is ligand distortion during pulling, which occurs mostly with flexible ligands. The calculated bulk RMSD of TPN was 2.4 Å, which is higher than typical RMSDs of toxins, and occurs as a result of the flexibility of the loop conformations in the residues 8–11.¹³ Thus, it was important to check for any distortions of TPN during umbrella sampling simulations. The average backbone RMSD of TPN was therefore calculated at each window and plotted against the reaction coordinate (see Figure 3A). The RMSD plot shows that the flexibility of TPN is reduced upon binding to Kir3.2 due to suppression of loop fluctuations by the previously described interactions of I8 and I9 with Y159 (see Table 1 and Figure 2). After TPN dissociates, its RMSD reverts back to the bulk baseline level (indicated by the red line in Figure 3A), indicating no artificial distortion of TPN occurs during the umbrella sampling MD simulations.

Figure 3.

Potential of mean force (PMF) plot for the K21 binding mode of the TPN-Kir3.2 channel complex. (A) Average RMSD of the TPN backbone atoms with respect to the NMR structure during the dissociation of TPN from the Kir3.2 binding pocket at each window position during umbrella sampling MD simulations (black curve). The red line indicates the RMSD of TPN backbone atoms in bulk solution. (B) Convergence study of the Kir3.2-TPN PMFs. A 2 ns sampling size is shown for overlapping time windows, moved in 1 ns steps over the range of the entire 6 ns data. The final PMF over the 1–6 ns range was used for calculating the binding free energy.

A confounding issue with PMF calculations is the convergence of the results and exclusion of equilibration data from the production data. Therefore, the block data were analyzed on full simulation trajectories to check the convergence and to separate and identify the equilibration data.

Shown in Figure 3B are the individual PMFs built using 2 ns blocks of data, which were slid in 1 ns steps over the collected 6 ns of data for each window. The large drop of the PMF after the first block (0–2 ns data) indicates the system is not equilibrated, but after that, the PMFs fluctuate around a common baseline, which indicates that the system is equilibrated. The TPN binding constant was, therefore, determined by integrating the final PMF (1–6 ns data) shown in Figure 3B using eq 1 (cf. Methods section).

The PMF integration yielded a TPN binding free energy of −11.7 kcal/mol, where R = 0.69 Å was used for the cross-sectional area of the TPN binding. This derived binding free energy value is in close agreement with the experimentally measured value of −11.1 kcal/mol for heteromeric Kir3.1/Kir3.2 channels (described below) and therefore supports the K21 binding mode being the most energetically favored Kir3.2-TPN structural complex.

Binding Strength of Kir3.2-TPN Pairwise Interactions.

The trajectory data from the umbrella sampling simulations provide additional information on the binding mechanism of TPN, as well as the binding strength of the pairwise interactions. A very useful measure in this regard is the persistence length of the interacting pairs, which roughly gives the distance of the toxin from the binding pocket at which contact is broken. The persistence length is directly related to the interaction strength; hence, it provides complementary information on the relative strength of individual interactions in a binding mode. To this end, we calculated the average N–O and C–C distances for each pair using the data from umbrella windows. The calculated distances are plotted as a function of the window position and shown in Figure 4.

Figure 4.

Persistence length of interacting pairs for the K21 binding mode of the Kir3.2-TPN complex. For each pairwise interaction indicated in Table 1, the average distance of TPN from the binding pocket at which the contact is broken is shown along the z-axis. The trajectory data are taken from the umbrella sampling simulations used to construct the final PMF shown in Figure 3. Error bars are not included to avoid cluttering but were ~1 Å for most data points.

From these results, the pore-inserting lysine K21 keeps contact with Y157 in the selectivity filter the longest, for about 5 Å, which is rarely seen among toxin blockers of Kv channels. The sharp rise in the PMF up to the z-axis position at 34 Å is associated with the persistence of the contact pairs. After that, all contacts are broken, and the PMF rises more gradually due to the screened Coulomb interactions.

Among the other charge interactions, Kir3.2-E127(subunit D)-TPN-R7 has a smaller contact distance than Kir3.2-E127(subunit A)-TPN-K16, suggesting that the former may be stronger. However, the first interaction breaks down after only 2 Å pulling, while the latter contact persists for 4 Å. It appears that the R7 side chain is in a strained conformation in the binding mode and is the first to detach after a short pulling of TPN.

Using the persistence distances, we can also comment on the relative strength of the hydrophobic interactions of TPN with the two Y159 side chains. The Kir3.2-Y159(subunit A) TPN-W15 contact persists longer, indicative of being the strongest of the hydrophobic contacts. Because I8 and I9 are on the flexible loop of TPN, backbone fluctuations may facilitate their quicker detachment from Kir3.2-Y159(subunit C).

The sequence and order for detachment of the interacting TPN contacts from Kir3.2 can be summarized as follows (first to last): R7–N_1–2, I8–C_z, I9–C_γ, K16–N_z, W15–C_z, K21–N_z. A holistic inspection of the topology of the binding surfaces also reveals that the Kir3.2 outer vestibule has a wider turret span compared to Kv channels, which allows TPN to rotate within the vestibule. The sudden changes in the pair distances in Figure 4 are due to such rotations. Because K21 is the C-terminal residue, it also has more freedom to rotate and make contacts being the last contact to detach.

Comparative MD Binding Modes of TPN to Heteromeric Kir3 Channel Complexes.

Since native Kir3.2 channels are predominantly expressed as heteromeric channels assembled with other Kir3.x subunits, we performed restrained docking of TPN to homology modeled Kir3.1/Kir3.2 channels having a 2:2 subunit stoichiometry. The asymmetric Kir3.2 subunit interactions of TPN with the homotetrameric channel shown previously indicate a favored interaction with two Kir3.2 subunits that are adjacent to one another. We, therefore, compared TPN interactions with two Kir3.1/Kir3.2 channels having different subunit arrangements around the central pore, namely, Kir3 [1–1–2–2] (having adjacent Kir3.2 subunits) and Kir3[1–2–1–2] (having diagonally arranged Kir3.2 subunits).

As shown in Table 2, the TPN binding mode interactions for the Kir3 [1–1–2–2] heteromer channel retained all of the interactions of the Kir3.2 homotetramer, where the adjacent electrostatic Kir3.2 E127 turret interactions with TPN R7 and K16 were conserved, as well as the Kir3.1 and Kir3.2 pore region hydrophobic interactions with TPN I8, I9, and W15 (Figure 5A). All contacts were retained during 50 ns of MD simulations, as observed with the homotetramer.

Table 2.

Pairwise Contacts between TPN and 2 Distinct Kir3.1/Kir3.2 Hetero-Tetrameric Channels Assembled with Different Subunit Arrangements^a

	Binding mode of Kir3 [1-1-2-2] heteromer
	Kir3 subunit	Tertiapin	MD mean distance (Å)
	(A) Y146-O, (C,D)Y 157-O	K21-NZ	2.92
	(A) Y 148-Cε	I8-Cγ	4.40
	(A) Y 148-Cε	I9-Cγ	4.40
	(C) E127-O1–2	K16-NZ	3.12
	(C)V161-Cγ	W15-CZ	4.24
	(D) E127-O1–2	R7-N1–2	3.28
	(D)V161-Cγ	19-Cγ	4.40
	(D) Y159-Cε	W15-CZ	4.32
	Binding mode of Kir3 [1 −2–1 −2] heteromer
	Kir3 subunit	Tertiapin	MD mean distance (Å)
	(C) Y146–0, (B,D) Y157-O	K21-NZ	2.98
	(A) G147-O	K16-NZ	-
	(A) Y 148-Cε	H12-Cδ	4.52
	(A) Y 148-Cε	W15-CZ	4.52
	(B) E127-O1–2	R7-N1–2	-
	(C) Y 148-Cε	I8-Cγ	4.35
	(C) Y 148-Cε	I9-Cγ	4.35
	(D)G 158-O	K16-NZ	-

^aThe individual Kir3 subunits are illustrated by A, B, C, and D as they are arranged around the central channel pore. Restrained docking was performed using the K21 binding mode. The average N–O and C–C distances obtained from 50 ns of MD simulations are shown for each pair of contacts.

Figure 5.

Preferential binding of TPN to Kir3 [1–1–2–2] heteromers. (A) Stable binding of TPN to the Kir3 [1–1–2–2] heteromer. Upper and lower images display contacts between TPN and the Kir3.1 and Kir3.2 subunits as indicated. Hydrophobic contacts are encircled. (B) Breaking of the electrostatic contacts in the Kir3 [1–2–1–2] heteromer during MD simulations of the complex. The upper and lower panels show the time series for the G158-K16 and E127-R7 distances, respectively. The breakage times of the contacts are indicated by the red arrows in each figure.

In contrast, the TPN binding mode for the Kir3[1–2–1–2] heteromer retained only one R7-E127 turret interaction after docking, the other significant turret interaction, K16-E127, was lost (cf. Table 2). TPN-K16 instead formed weaker contacts with glycine residues located near the pore region (Kir3.2-G158 and Kir3.1-G147). The TPN hydrophobic interactions, however, were retained via contacts with Kir3.1-Y148, which is structurally equivalent to Kir3.2-Y157. However, as shown in Figure 5B, during the 50 ns of MD simulations, two of the primary electrostatic interactions were rapidly broken: TPN-R7 with Kir3.2-E127 and TPN-K16 with Kir3.2-G158 and Kir3.1-G147. Thus, based on these comparative TPN interactions, the potent nanomolar TPN binding affinity would be less likely for the Kir3[1–2–1–2] heteromer channel that has the diagonally arranged Kir3.2 subunits and fewer contacts.

Functional Assessment of the Predicted Pore-Inserting TPN-K21 Side Chain.

The TPN-Kir3.x contact sites identified in silico were next targeted by site-specific amino acid mutagenesis to assess their functional impact on TPN_Q-mediated inhibition of Kir3.1/Kir3.2 channels. Since previous computational simulations placed TPN-K17 as the pore-inserting lysine in Kir3.2,^40,41 whereas our MD calculations did not assign a significant role for K17, we initially compared the inhibitory activity of TPN_Q-K17A and TPN_Q-K21A to TPN_Q inhibition of functional Kir3.1/3.2 channel currents recorded from Xenopus oocytes. The two TPN_Q variant peptides, where the lysine residue was replaced with an alanine residue, effectively removes the interacting basic ε-amino group. TPN_Q inhibited G_β1γ2-activated Kir3.1/3.2 channel currents with an IC₅₀ value of 22 ± 4 nM (mean ± SEM, n = 9), where 300 nM inhibited ~98% of the K⁺ current. This result is comparable to the potency reported for the TPN_Q block of native neuronal Kir3 channels containing the Kir3.2 subunit^55,56 and ~3-fold more potent than that reported for Kir3.2-expressing AtT20 cells (a pituitary cell line, IC₅₀ value = 60 nM).⁵⁷

As shown in Figure 6, both TPN_Q-K17A and TPN_Q-K21A each effectively inhibited G_β1γ2-activated Kir3.1/3.2 channels. The potency for TPN_Q-K17A inhibition (IC₅₀ value = 28 ± 3 nM, mean ± SEM, n = 8) was not significantly different from TPN_Q. Whereas the IC₅₀ for TPN_Q-K21A (34 ± 2 nM, mean ± SEM, n = 8) indicated a modestly reduced potency when compared to TPN_Q (Figure 6B). The derived Hill coefficient values were not significantly different among all three peptides (TPN_Q −0.94 ± 0.40; TPN_Q-K17A −1.20 ± 0.13; TPN_Q-K21A −1.09 ± 0.09). The subtle decrease in TPN_Q-K21A potency was evident when comparing the degree of inhibition produced by 10 nM of each TPN_Q variant peptide to the same oocyte (cf. Figure 6C,D). Although the difference in blocking efficacy for TPN_Q-K21A compared to TPN_Q was consistent with the MD binding mode simulations that favored K21 over K17, the persistent channel block by TPN_Q-K21A is not consistent with the behavior of classic Kv channel pore blockers where neutralizing the pore-inserting residue reduces toxin potency by 2–3 orders of magnitude.^58–60 It is noteworthy that Jin and colleagues similarly observed modest IC₅₀ changes with single alanine substitution variants and a block of rat Kir1.1 channels.¹⁴

Figure 6.

Assessing the functional effects of TPN_Q-K21 on Kir3.1/Kir3.2 channels. (A) Surface rendering of the TPN peptide structure, highlighting the K21 and K17 basic residues that were mutated individually to alanine residues. (B) Concentration–response curves for TPN_Q (black symbols), TPN_Q-K17A (red symbols), and TPN_Q-K21A (green symbols) are shown for their inhibitory activity on G_β1γ2-activated Kir3.1/Kir3.2 currents recorded from Xenopus oocytes (mean ± SEM, n = 8–9 oocytes for each). Solid curves represent a Hill function fit to the mean data values. The respective IC₅₀ values derived from individual oocyte concentration–response experiments, for each peptide, are provided in the inset table (mean ± SEM, *P < 0.05). (C) Time-course plot for peak current amplitudes at −80 mV, illustrating the comparative inhibitory effects of TPN_Q, TPN_Q-K17A, and TPN_Q-K21A at 10 nM for each on G_β1γ2-activated Kir3.1/Kir3.2 currents recorded from the same oocyte. BaCl₂ at 300 μM was additionally tested for comparison. The oocyte membrane potential was clamped at −40 mV with voltage ramps evoked every second as illustrated in the inset and described in the Methods. (D) Comparison of the inhibitory actions of each TPN_Q peptide at 10 nM on G_β1γ2-activated Kir3.1/Kir3.2 currents. The asterisk denotes P < 0.05.

To further examine the TPN_Q-K21A response, we performed molecular docking of the modeled TPN-K21A peptide to both the Kir3.2 homomer structure and to the modeled Kir3 [1–1–2–2] heteromer to mimic the in vitro experiment. Interestingly, unbiased blind docking of TPN-K21A revealed a new energetically favored docking pose for both channels, where K16 took over as the pore-inserting lysine, interacting with the Y157 carbonyls. We then performed restrained docking of TPN-K21A with the Kir3 [1–1–2–2] channel pore, where the K16 side chain was restrained as the pore-inserting residue. The resulting “A21 binding mode” yielded an energetically favored binding pose where the hydrophobic interaction previously observed in the K21 binding mode (TPN-W15:Kir3.2-Y159) was maintained in addition to the K16 side chain interacting with the Kir3.2-Y157 carbonyls. Additional electrostatic contacts between TPN-K17 and K20 and channel vestibule residue side chains were also identified that stabilize the TPN-K21A interaction with the Kir3 [1–1–2–2] outer vestibule. Thus, the persistent functional block of Kir3.1/Kir3.2 channels by TPN_Q-K21A in Xenopus oocytes (see Figure 6) is consistent with the behavior of a classic pore blocker where the K21A amino acid change enables an alternative basic TPN residue to assume the role of the pore blocker (see Discussion).

Functional Assessment of Predicted TPN Hydrophobic Contacts with the Kir3.2 Pore Entry Y159 Residue.

To examine directly the functional impact of the predicted TPN_Q hydrophobic interactions at Kir3.2-Y159, the blocking effects of TPN_Q were examined on Kir3 channels, where the Kir3.2 subunit Y159 residue was mutated to an alanine residue to effectively remove the hydrophobic, hydroxylated benzene ring side chain. Since the Kir3.1 subunit also presents a tyrosine residue at the equivalent position (Y148), heteromeric Kir3.1/Kir3.2-Y159 channels would still possess 2 of these 4 tyrosine side chains assuming the 2:2 subunit stoichiometry.

As shown in Figure 7, heteromeric Kir3.1/3.2 channels effectively assembled with the Y159A subunit to form functional m2 receptor-activated, inward rectifying K⁺ currents that were largely indistinguishable from channels assembled with the wild-type (wt) Kir3.2 subunit. The only noted difference was a slightly larger I_K,basal amplitude, with no difference in I_K,ACh amplitude. Notably, however, the sensitivity of Kir3.2-Y159A-containing channels to block by 100 nM TPN_Q was markedly reduced, indicating the Y159 side chain is necessary to confer the high-affinity TPN_Q block.

Figure 7.

Assessing the functional effects of targeted Kir3.2 mutations on TPN_Q inhibition. (A) Surface rendering of the outer vestibule of the heteromeric Kir3.1/3.2 channel, highlighting Kir3.2 residues (light gray subunits) targeted by mutagenesis based on the identified TPN contacts from the K21 binding mode complex. The two adjacent Kir3.2-E127 turret residues are indicated (magenta). The pore entry “ring” (blue residues) formed by the two adjacent Kir3.2-Y159 residues and two adjacent Kir3.1-Y148 residues (dark gray subunits) are also shown. (B) Time-course plots from representative oocyte recordings of ACh-evoked Kir3.1/Kir3.2 channel currents in response to 100 nM TPN_Q. The Kir3.2 wild-type (wt) subunit (black), Kir3.2-Y159A subunit (blue), and Kir3.2-E127R/K subunits (magenta) were individually coexpressed with the Kir3.1 subunit. The shaded (gray) area indicates the solution transition to 20 mM K⁺ (cf. Methods). The time periods for 1 μM ACh application and coapplication with 100 nM TPN_Q are indicated for each. (C) Comparative summary of the inhibitory effects of 100 nM TPN_Q on currents recorded from the Kir3.2 subunit mutations. Bars represent the mean and SEM from 7 to 9 oocytes for each condition. The asterisk denotes P < 0.05.

Functional Assessment of Predicted TPN Contacts with Kir3.2 Turrets.

To interrogate the predicted contact interactions between TPN_Q and the Kir3.2 turret region, we also tested the impact of Kir3 channels assembled with a Kir3.2 subunit having either an E127R or E127K “charge swapping” substitution mutation. Each of these single residue changes was expected to disrupt the two electrostatic side-chain interactions between the adjacent Kir3.2 turret E127 residues and TPN_Q residues R7 and K16.

As shown in Figure 7, both of the Kir3.2-E127 turret point mutations (E127R or E127K) assembled to form inward rectifying Kir3 channels that were activated by coexpressed m2 receptor activation. Their sensitivity to the TPN_Q block (100 nM), however, was also markedly reduced and consistent with the MD simulations (Figure 6C). The impact of the E127R/E127K mutations on 100 nM TPN_Q-mediated inhibition was slightly less than the Y159 mutation when compared to wild-type, indicating a rank order of Y159 > E127 for these two assessed Kir3.2 contact sites.

Lastly, we also assessed the efficacy of a TPN_Q variant peptide, where R7 was replaced with a “charge swapping” glutamic acid residue (TPN_Q-R7E). As shown earlier (cf. Figure 5), the basic TPN_Q-R7 side chain is predicted to form an electrostatic interaction with one of the Kir3.2 turret E127 residues, where the R7 → E7 variant peptide would be expected to disrupt the interaction of TPN_Q with one turret site, without affecting the adjacent turret E127 site that interacts with TPN_Q-K16.

As shown in Figure 8, TPN_Q-R7E (300 nM) had minimal blocking activity (~7% inhibition) on G_β1γ2-activated Kir3.1/3.2 channel currents in comparison to the TPN_Q-mediated channel block (~94% inhibition), indicating the basic TPN_Q-R7 residue is necessary for mediating the high-affinity Kir3.2 channel block.

Figure 8.

Assessing the functional effects of TPN_Q-R7 on Kir3.1/Kir3.2 channels. (A) Surface rendering of the TPN peptide structure, highlighting the R7 basic residue that was mutated to a charge swapping glutamic acid residue, TPN_Q-R7E. (B) Comparative summary of the inhibitory effects of TPN_Q (black) and TPN_Q-R7E (purple) at 300 nM on G_β1γ2-activated Kir3.1/Kir3.2 currents recorded from Xenopus oocytes. Bars represent the mean and SEM from 7 to 9 oocytes for each condition. The asterisk denotes P < 0.05.

DISCUSSION

The results presented here are the first to couple structure-based MD predictions of TPN-Kir3.x channel interactions with functional validation measured in vitro via the TPN_Q block of expressed neuronal Kir3.1/3.2 channels. This combined in silico plus in vitro approach enables iterative examination of the structural and molecular details mediating the high-affinity block of Kir3.2 channels by TPN_Q. The diagram provided in Figure 9 illustrates the key findings from this study and summarizes the main determinants mediating TPN_Q binding and block of heteromeric Kir3.1/3.2 channels. Our findings are consistent with a multipoint interaction model for TPN and the Kir3 ion channel target that includes a functional dyad pore-blocking mechanism.^61–65

Figure 9.

Summary illustration highlighting the asymmetric Kir3.2 subunit interactions for high-affinity TPN binding and inhibition of Kir3.2 channels. Pairwise contacts were derived from Tables 1 and 2, with TPN residues shown in red and Kir3 residues in black. (A) For the homotetrameric Kir3.2 channel, the GYG selectivity filter Y157 carbonyl groups interact with TPN_Q-K21. TPN hydrophobic residues W15, I8, and I9 interact with multiple subunit Y159 residues that form the pore entry hydrophobic ring (indicated by the dotted white circle). In addition, two E127 turret residues from adjacent subunits make electrostatic contacts with TPN residues R7 and K16. (B) For the heterotetrameric Kir3 [1–1–2–2] channel, the corresponding TPN contact residues are shown, highlighting the conserved role of the two adjacent E127 turret contacts, and multiple hydrophobic pore entry ring interactions. Mutations at either turret or hydrophobic pore entry ring residues result in a significant loss in the TPN_Q-mediated channel block.

Mechanism for High-Affinity TPN Binding. The identified molecular determinants for TPN binding to Kir3.1/3.2 channels are in excellent agreement with those previously reported for the TPN_Q block of epithelial Kir1.1 channels.^14,15 Rat Kir.1.1 channels exhibit a 10-fold higher sensitivity to the TPN_Q block (IC₅₀ = 2 nM) compared to Kir3.1/3.2 channels (IC₅₀ = 22 nM) and is likely attributable to additional interactions between TPN_Q and the homotetrameric Kir1.1 outer vestibule.¹³ The cardiac heteromeric Kir3.1/3.4 channel (TPN_Q IC₅₀ value = 8 nM) and native I_KACh channel (TPN IC₅₀ value = 30 nM) have comparable blocking affinities to the Kir3.1/3.2 channel.^27,30,66 To confer a high-affinity Kir3.1/3.2 channel block by TPN_Q, two mechanistic prerequisites emerged from our study: (1) electrostatic interactions between TPN and multiple Kir3.2 channel turrets and (2) hydrophobic contacts with the hydrophobic “ring” that surrounds the pore entry. These interactions together position the ternary TPN structure to insert its K21 lysine side chain and interfere with K⁺ occupancy at the S1 binding site of the selectivity filter.

Kir channels insensitive to the TPN_Q block notably lack one or more of the two prerequisites. For Kir3 channels, both Kir3.2 and Kir3.4 subunits possess these necessary sites enabling high-affinity TPN block when assembled as heteromers with the TPN-insensitive Kir3.1 subunit. For the Kir3.1 subunit, introducing the Kir3.4 turret contact has previously been shown to confer TPN_Q sensitivity, given it already possesses the other necessary phenylalanine at the pore entry.⁵⁴ For the TPN-insensitive human Kir1.1 isoform, introducing both the turret and pore entry hydrophobic contact residues from the rat isoform is sufficient to recapitulate high-affinity TPN_Q binding (IC₅₀ ~ 1 nM) observed for the rat Kir1.1 isoform.^13,15

Pore Blocking Mechanism.

For TPN, the side chains of K21 and W15 are ~4 Å apart and display the hallmark features of a “functional dyad” motif described for classic Kv channel toxin pore blockers (Figure 10).⁶⁴ On the basis of our findings reported here, we propose that interactions of K21 with the carbonyl oxygen in the GYG selectivity filter, and W15 with the hydrophobic ring formed by Y159 surrounding the pore entry, both work in concert to physically plug the channel pore and block K⁺ conduction. The additional TPN_Q contacts (R7 and K16) contribute to the binding affinity and effectively position and hold the W15:K21 functional dyad in place by interacting via ionic bonds with the two Kir3.2 turret interaction sites (E127).

Figure 10.

Multiple functional dyad options for TPN variant peptides. Shown are the side-chain distances for W15-K21 and W15-K16 in TPN. The W15:K21 dyad is energetically favored with TPN molecular docking. When K21 is replaced with A21, the W15:K16 dyad becomes favored to occupy and block the Kir3.2 channel pore.

Unexpectedly, however, our findings revealed a persistent block by a TPN_Q-K21A variant peptide that initially might indicate S1 occupancy by the K21 side chain is not essential for binding affinity and pore block. However, further molecular docking and examination of the TPN-K21A interaction revealed that a new binding mode is favored with the K21A variant peptide, where the K16 side chain now occupies the selectivity filter. In this new binding mode (A21), an alternative W15:K16 functional dyad plugs the pore. The side chains of the two adjacent residues W15 and K16 also fall within the 6–7 Å spatial criteria of the “functional dyad” motif (Figure 10). This raises the question of why the W15:K16 functional dyad is not favored with TPN_Q but is with the TPN-K21A variant. An examination of the C-terminal position of A21 in the docked complex reveals that the side chain of K21 introduces steric clashes with the channel vestibule and thereby prevents this binding mode from occurring with the native TPN peptide. Only with the K21A substitution does this binding mode become feasible.

These findings indicate the polybasic surface chemistry of TPN (4 Lys and 1 Arg) enables multiple different binding modes when individual sites are neutralized and provides a rational explanation for the persistent block by TPN_Q-K21A reported here for the Kir3.1/3.2 channel, and also for the Kir1.1 channel where alanine substitution at each TPN basic residue similarly conferred modest changes in the binding affinity individually.¹⁴ Exploring in computational space and engineering double mutant TPN_Q variant peptides will provide yet further details on the versatility of the TPN pharmacophore as a classic pore blocking functional dyad, and whether alternative mechanisms may be deployed.⁶⁷ These studies are now feasible, as demonstrated here utilizing this approach.

Implications for Channel Subunit Arrangement in Heteromeric Kir3 Channels.

The computational binding studies of TPN to the outer vestibule region of the homotetrameric Kir3.2 channel, in comparison to TPN-insensitive Kir3.1 and Kir3.3 channels, indicate TPN makes contact with two adjacent subunit turrets at both E127 side chains. Assuming a 2:2 subunit stoichiometry for the TPN_Q-sensitive heteromeric Kir3.1/Kir3.2 channels, the arrangement of the four subunits around the central channel pore would indicate the TPN-sensitive heterotetramer is organized with two adjacent Kir3.1-Kir3.1 subunits and two adjacent Kir3.2-Kir3.2 subunits (versus a diagonal arrangement). The MD simulations of the Kir3 [1–1–2–2] heteromers versus Kir3 [1–1–2–2] heteromers confirmed this favored arrangement (cf. Figure 9). Moreover, since both Kir3.1 and Kir3.3 subunits are predicted to be TPN-insensitive from the MD simulations, Kir3.1/Kir3.3 heterotetramers expressed and assembled in the nervous system would be expected to be TPN_Q-insensitive.^68,69 Previous studies using heterologous expression of concatenated Kir subunits with different subunit arrangements were not conclusive on Kir subunit organization around the pore.^70,71 Based on our results, TPN_Q sensitivity may be a useful “readout” assay to further test these still open questions. Assessing the TPN_Q block of concatenated Kir3 subunit arrangements would be a logical next step to explore further and address this important question.

In summary, our structure- and function-based findings revealed several new TPN-Kir3.x channel interaction details that are necessary for future efforts aimed at re-engineering the TPN_Q scaffold to design new peptide variants with improved potency and Kir channel specificity. Such venom-inspired pharmacological tools can accelerate preclinical drug discovery efforts and serve as the impetus for this work.