Introduction

Kainate receptors (KARs) are tetrameric ion channels that belong to the family of ionotropic glutamate receptors (iGluRs)1. Functional KARs are assembled from subunits GluK1-5, encoded by GRIK1-5 genes2,3,4. GluK1-3 are considered ‘primary’ subunits as they can form functional homomeric and heteromeric channels. On the other hand, the ‘secondary’ GluK4-5 subunits only form functional heterotetramers with GluK1-3 subunits5. KARs are expressed ubiquitously throughout the central nervous system (CNS) and are distributed both pre- and post-synaptically, playing a crucial role in regulating neuronal activity6. Consequently, activation of KARs elicits a diverse array of pre- and post-synaptic effects on glutamatergic and gamma-aminobutyric acid (GABA)ergic synaptic transmission7,8. Whereas presynaptic KARs modulate neurotransmitter release and postsynaptic KARs mediate excitatory neurotransmission, extrasynaptic KARs are involved in controlling neuronal excitability2,9,10,11,12,13. The functional diversity of KARs, including ion selectivity, ligand specificity, and kinetics, is further expanded through RNA editing, alternative splicing of subunit transcripts, and interactions with auxiliary subunits14,15,16,17,18,19. Thus far, KARs have been found to associate with Neto-type auxiliary subunits, Neto1-2, which increase current amplitude, alter agonist efficacy, slow the deactivation and desensitization rates, increase the channel open probability, and modulate neuronal localization10,20,21,22,23.

Previous studies demonstrated that perturbations in KAR function are implicated in neurological and neurodegenerative disorders, including epilepsy, ischemia, stress, anxiety, intellectual disability, and pain2,24,25,26,27,28,29. Small molecule inhibitors of KARs might therefore have therapeutic potential. One of the prospective classes of KAR inhibitors are ion channel blockers that act on unedited receptors which have glutamine instead of arginine at the functionally important Q/R-site that determines calcium permeability and plays a critical role in glutamate-induced neurotoxicity30,31. Two types of KAR channel blockers have been identified: polyamines and polyamine-containing toxins or toxin-like molecules32.

Polyamines are ubiquitous in bacterial, plant, and animal cells and often serve as blockers of cation-selective ion channels33,34,35. Polyamine block of ion channels, primarily by endogenous spermine (SPM) and spermidine, was first discovered for potassium channels36 but subsequently revealed for iGluRs37, nicotinic acetylcholine receptors38, and cyclic nucleotide-gated (CNG) ion channels39. Due to their cationic nature, polyamines bind to negatively charged regions of biomolecules, including electronegative pores of cation-selective voltage- and ligand-gated ion channels, often with micromolar affinity35,38,39,40,41,42,43. At glutamatergic synapses of the developing and mature CNS, polyamines act as permeant ion channel blockers of both α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs), another member of the iGluR family, and KARs35,44,45. Accordingly, cytoplasmic polyamines have been recognized as important determinants of neuronal signaling that regulate action potential firing rates46 as well as the strength of neurotransmission47,48.

The second category of ion channel blockers that act at both AMPARs and KARs includes extracellularly-applied exogenous polyamine- or acylpolyamine-containing toxins, such as argiopin, also known as argiotoxin-636 (AgTx-636)49,50,51, from the spider Argiope lobata, Joro spider toxin JSTX-352,53 from Nephilia clavata, philanthotoxin-433 (PhTx-433) from the wasp Philanthus triangulum54,55, and Nephila spider toxin-8 (NpTx-8)51. Neuroprotective properties demonstrated by some of these toxins56,57,58 inspired the development of synthetic analogs, including PhTx-433 derivatives56,59,60,61,62,63,64 like PhTx-343, PhTx-56, and PhTx-74, the Joro spider toxin analog 1-naphthyl acetyl spermine (NASPM)65,66,67, which reduces oxidative stress and protects neurons after ischemia68,69, and IEM-1460 adamantane derivative67,70 that attenuates epileptic seizures71,72 and reduces visceral pain73,74 and receptor activity related to schizophrenia75.

Previously, the molecular mechanism of ion channel block was characterized structurally for exogenous AMPAR channel blockers AgTx-636, IEM-1460, and NASPM67. More recently, synaptic complexes of GluA2 homotetramer with auxiliary subunits γ5 and CNIH2 were shown to be blocked by the endogenous polyamine spermidine76. However, the structural basis of KAR channel block by polyamines and polyamine-related molecules has remained elusive.

Here, we embark on studies of KAR channel block using single-particle cryo-EM, electrophysiology, and MD simulations. We solve cryo-EM structures of homotetrameric GluK2 KAR in complex with PhTx-74, NpTx-8, SPM, and bioactive compound Kukoamine A (KukoA) (Fig. 1a), a spermine alkaloid originally isolated from Lycium chinense (goji berry), which is known for its hypotensive effect77. It has diverse biological activities, including anticancer, neuroprotective, and anti-inflammatory properties78,79. We show that PhTx-74, NpTx-8, SPM, and KukoA act as trapping blockers. We also perform MD simulations of our structures in near physiological conditions, including room temperature and the presence of electrolytes and lipid bilayers, to determine specific interactions of blocker molecules with KARs and to investigate the blocker pose stability and binding site variability. In addition, we compute the pore dimensions and electrostatic profiles for the KAR-blocker complexes to decipher the molecular determinants of blocker binding in the ion channel pore.

Fig. 1: Functional characterization of GluK2 channel block by NpTx-8, PhTx-74, kukoamine A and spermine and cryo-EM.
figure 1

a Chemical structures of NpTx-8, PhTx-74, KukoA, and SPM. b Whole-cell patch-clamp currents recorded at –60 mV membrane potential from HEK293 cells expressing GluK2 in response to application of KA alone (black traces) or with 2.7 μM NpTx-8, 75 μM PhTx-74, 3 mM KukoA or 20 mM of SPM (red traces). c Concentration-dependencies of GluK2 block by NpTx-8 (blue), PhTx-74 (pink), KukoA (green) and SPM (orange) fitted to the logistic equation with the parameters: IC50 = 0.51 ± 0.01 µM and nHill = 1.14 ± 0.01 (n = 9) for NpTx-8, IC50 = 7.53 ± 0.35 µM and nHill = 0.99 ± 0.04 (n = 10) for PhTx-74, IC50 = 704 ± 31 µM and nHill = 1.45 ± 0.05 (n = 7) for KukoA, and IC50 = 7.77 ± 0.51 mM and nHill = 0.89 ± 0.04 (n = 8) for SPM. Data are mean ± SEM. Source data are provided. d Voltage-dependence of the whole-cell current recorded from an HEK293 cell expressing GluK2 in response to a step of –80 mV and then a –80 to +90 mV voltage ramp during application of KA alone (black trace) or with 20 mM SPM (red trace). Vertical blue lines indicate –60 mV and +60 mV voltages, at which rectification index measurements were taken. e Rectification index, (IBlock / IControl)–60mV / (IBlock / IControl)+60mV, calculated at 1.5 μM NpTx-8, 30 μM PhTx-74, 2 mM KukoA and 20 mM of SPM. Data are mean ± SEM. The number of biologically independent measurements, n = 9 for NpTx-8, n = 9 for PhTx-74, n = 9 for KukoA, and n = 9 for SPM. Data are representative of two independent experiments. f Cryo-EM density for full-length GluK2PhTx74, with GluK2 subunits colored green, pink, blue and orange. The red contour envelopes the LBD-TMD region. g Cryo-EM density for the LBD-TMD region with the micelle and ATD signals masked out (see the red contour in f).

Full size image

Results

Functional characterization of KAR channel block

We used whole-cell patch-clamp recordings to study the function of calcium-permeable homotetrameric GluK2 receptors transiently expressed in HEK 293 cells. In the presence of the positive allosteric modulator Concanavalin A (ConA), which dramatically reduces KAR desensitization, the agonist kainic acid (KA) elicited large inward currents with a steady-state value of IControl (Fig. 1b, black traces). In the presence of the ion channel blockers NpTx-8, PhTx-74, KukoA, or SPM, the current amplitude was reduced and characterized by a smaller steady-state value, IBlock (Fig. 1b, red traces).

NpTx-8, PhTx-74, KukoA, and SPM inhibited GluK2-mediated currents in a concentration-dependent manner, with stronger inhibition observed at higher blocker concentrations (Fig. 1c). The potency of the blockers varied widely, with the half-maximal inhibitory concentration, IC50 = 0.51 ± 0.01 µM (n = 9) for NpTx-8, IC50 = 7.53 ± 0.35 µM (n = 10) for PhTx-74, IC50 = 704 ± 31 µM (n = 7) for KukoA, and IC50 = 7.77 ± 0.51 mM (n = 8) for SPM. Such dramatic differences in blocker potency are typically associated with faster dissociation kinetics for low-affinity blockers and slower ones for high-affinity blockers. When the blocker dissociation kinetics are faster than channel closure, simultaneous termination of the blocker and agonist application results in the appearance of resurgent, or “hook”, currents carried by channels that have been relieved from the block but are still awaiting pore closure preceding agonist dissociation80,81,82. Accordingly, the highest amplitude hook currents were observed for the lowest-affinity blocker SPM. In contrast, no hook currents were observed for the highest-affinity blocker NpTx-8 (Fig. 1b). The intermediate affinity blockers, PhTx-74 and KukoA, showed hook currents of intermediate amplitude (Fig. 1b).

While the presence of the hook currents is strongly dependent on the blocker dissociation rate, their amplitude and kinetics are also determined by the interaction of the blocker with channel gating80,81,82,83. If a blocker prevents channel closure or desensitization, the hook current may have a higher amplitude and decay much slower than the control tail current. On the contrary, for a blocker that does not interfere with channel gating (trapping blocker), the hook current amplitude does not exceed the amplitude of the control tail current, and the current decay kinetics are similar. Therefore, the hook currents observed for PhTx-74, KukoA, and SPM belong to the second category, indicating that these molecules represent trapping blockers of the GluK2 channels.

Inhibition of GluK2-mediated currents by the positively charged molecules of NpTx-8, PhTx-74, KukoA, and SPM was also voltage-dependent, consistent with the mechanism of channel block. In the absence of blockers, the GluK2-mediated currents showed strong inward rectification due to the block by intracellular polyamines (Fig. 1d). In the presence of NpTx-8, PhTx-74, KukoA, or SPM, the current amplitude was reduced but the extent of current reduction, represented by IBlock / IControl, was much stronger at negative compared to positive membrane voltages, reflected in the values of the current rectification index, (IBlock / IControl)–60mV / (IBlock / IControl)+60mV, being much smaller than 1 (Fig. 1e). The similar values of the current rectification index for NpTx-8, PhTx-74, KukoA, and SPM suggest that all four blockers likely bind to approximately the same or overlapping binding sites inside the GluK2 ion channel pore.

Cryo-EM reconstruction of GluK2 in the presence of ion channel blockers

To uncover the molecular details of NpTx-8, PhTx-74, KukoA, and SPM binding inside the ion channel pore, we subjected GluK2 to single-particle cryo-EM in the presence of each ion channel blocker individually and positive allosteric modulator BPAM344 (BPAM), which stabilizes the upper D1-D1 ligand-binding domain (LBD) dimer interface (Supplementary Figs. 1–5, Table 1). A typical 3D reconstruction of the full-length GluK2 showed 3-layer architecture, characteristic of AMPARs and KARs3,84,85,86,87,88,89, with the amino terminal domain (ATD) at the top, LBD in the middle, and the transmembrane domain (TMD) at the bottom (Fig. 1f). The ATD and LBD layers, connected by flexible linkers, showed a high degree of relative movement, signified by blurriness of the ATD layer compared to the LBD-TMD in 2D class averages (Supplementary Figs. 1–4). To improve the quality of 3D reconstructions, we therefore performed classification and refinement by focusing on the LBD-TMD region only (Fig. 1f, g). Indeed, cryo-EM reconstruction of the LBD-TMD region alone resulted in improved map quality, with well-resolved densities for BPAM, carbohydrates, and annular lipids (Fig. 1g; Supplementary Fig. 5a–c).

Table 1 Cryo-EM data collection, refinement, and validation statistics
Full size table

Overall, the 2.81–3.75 Å reconstructions of GluK2 in the presence of BPAM and ion channel blockers were similar to one another and to the reconstruction of GluK2 in the presence of BPAM only, which yielded the closed-state structure, GluK2closed (PDB ID: 8FWS), published before86. The biggest difference between the reconstructions was the presence of variably shaped densities inside the pore of the TMD (Fig. 2). While no density was observed in GluK2closed, reconstructions for the samples made in the presence of the blockers revealed densities in the middle of the pore that were roughly matching the chemical structure of the corresponding blocker (Fig. 1a; Supplementary Fig. 5d–f). Each density had an approximate 4-fold rotational symmetry that characterizes the channel of kainate receptor at this location and can fit the blockers in four equivalent poses, different by 90-degree rotation around the axis of local symmetry (Supplementary Fig. 5e, f). Unfortunately, symmetry expansion as well as focused classification and refinement techniques did not improve these densities, most probably because of the very small size of blocker molecules. The heterogeneity of blocker binding poses due to the local 4-fold rotational symmetry as well as the dynamic nature of binding (see MD simulations section below) are the likely reasons why the blocker densities are somewhat weaker than the density for the surrounding protein.

Fig. 2: TMD density in the absence and presence of ion channel blockers.
figure 2

Coronal (a) and transverse (b) plane views of cryo-EM density for the TMD in GluK2closed, GluK2NpTx8, GluK2PhTx74, GluK2KukoA, and GluK2SPM, with GluK2 subunits colored green, pink, blue and orange, and blockers red. The dashed black line in a shows where the density is cut away in b, while the dashed black line in b shows where the density is cut away in a.

Full size image

Reconstruction in the presence of SPM showed a narrow density matching the short and thin chemical structure of this polyamine. For KukoA, the density had a dumbbell shape, consistent with the two-headed structure of this molecule. For NpTx-8 and PhTx-74, the extracellular end of the density had a bulky cap, which likely represents averaged conformations of the blockers hydrophobic head, while its intracellular region included four prongs, likely representing four different orientations of the positively charged polyamine tail (Fig. 2). No other densities apparent in the presence but not in the absence of ion channel blockers were found in our cryo-EM reconstructions, strongly suggesting that NpTx-8, PhTx-74, KukoA, and SPM binding occurs in the ion channel pore only.

Structures of GluK2 and blocker binding site

We built atomic models of the GluK2 LBD-TMD region for the cryo-EM reconstructions in the presence of the different ion channel blockers. Guided by the cryo-EM density (Supplementary Fig. 5), we built LBDs, LBD-TMD linkers, and TMDs, including the intracellular cap and the C-terminal helix, often absent in KAR reconstructions, as well as lipids, carbohydrates, and ion channel blockers in the middle of the ion channel pore (Fig. 3a–c). Each LBD exhibits a bilobed architecture comprising the upper (D1) and lower (D2) lobes, revealed by crystal structures of the isolated LBD90,91 and by cryo-EM structures of full-length KARs3,85,86,87,88,89. The LBD layer has a dimer of A/D and B/C dimers arrangement, observed previously in AMPAR structures84,92,93, and structures of KARs in the closed3,85,86,88 and open94 states.

Fig. 3: GluK2PhTx74 structure and PhTx-74 binding site in the channel pore.
figure 3

GluK2PhTx74 structure viewed parallel to the membrane (a, b) and intracellularly (c). The molecules of BPAM and PhTx-74 are shown as space-filling models (yellow), carbohydrates (green) and lipids (purple) in sticks. d Closeup view of the PhTx-74 binding site in the channel pore. Only two (A and C) of four subunits are shown, with the front and back subunits (B and D) removed for clarity. The molecule of PhTx-74 (yellow) and residues involved in its binding (blue and green) are shown in sticks.

Full size image

Four TMDs form a cation-selective ion channel, composed of the M1, M3, and M4 transmembrane helices and a re-entrant M2-loop between M1 and M3 (Fig. 3c, d). M1 and M4 form the ion channel periphery, while the extended region of M2 and the C-terminal half of M3 line the ion channel pore. All four ion channel blockers fit the section of the ion channel pore intracellular to the M3 helices bundle crossing, which includes the central cavity lined by side chains of L645, I648, S649, and T652, and the selectivity filter contributed by side chains Q621 and E625 and backbone carbonyls of Q621, Q622, G623, S624, and E625 (Figs. 3d and 4a–d). The bulky hydrophobic heads of NpTx-8 and PhTx-74 occupy the central cavity, while their polyamine tails stretch nearly the entire length of the selectivity filter (Figs. 3d and 4a, b). Based on the four-pronged cryo-EM densities corresponding to NpTx-8 and PhTx-74 (Fig. 2), the polyamine tails of these blockers do not align with the pore axis but instead lean towards the selectivity filter walls adopting one of four nearly identical positions in the pseudo-4-fold symmetrical ion channel. KukoA has two identical bulky heads at the polyamine ends that each have equal possibility to fit the central cavity, while the other interacts with E625 at the intracellular pore entrance (Fig. 4c). Two upper thirds of the SPM molecule fit the extracellular part of the selectivity filter, with the corresponding density aligned with the pore axis (Fig. 2). The density is much weaker for the lower third of SPM, which likely adopts numerous conformations in the intracellular region of the selectivity filter (Fig. 4d).

Fig. 4: Comparison of the blocker binding region in different structures.
figure 4

Blocker binding region formed by M2 and M3 in GluK2NpTx8 (a, blue), GluK2PhTx74 (b, green), GluK2KukoA (c, violet), GluK2SPM (d, orange), GluK2Closed (e, gray, PDB ID: 8FWS), and GluK2Open (g, red, PDB ID: 9B35), with residues contributing to blocker binding shown in sticks. Only two (B and D) of four subunits are shown, with the front and back (A and C) subunits removed for clarity. The gating hinges are indicated by blue arrows in the GluK2Open structure. f, h Superposition of the blocker-bound structures from a–d with GluK2Closed (f) or GluK2Open (h).

Full size image

When superposed with the closed state structure GluK2closed (Fig. 4e), all four blocker-bound structures show nearly identical conformation of the channel pore (Fig. 4f, Supplementary Fig. 6), indicating that the structures were solved in the blocker-bound closed state. Notably, when the open-state structure GluK2open (PDB ID: 9B35; Fig. 4g) is superposed with the blocker-bound structures (Fig. 4h), the blocker binding region appears the same as well because the differences associated with channel opening start extracellularly to the blocker binding site with M3 helices bending at the gating hinge. To further validate our models of blocker binding in the pore region and identify the critical interactions between the blockers and the channel, we performed MD simulations.

Molecular dynamics simulations and blocker-channel interactions

The cryo-EM structures GluK2NpTx8, GluK2PhTx74, GluK2KukoA, and GluK2SPM were used as initial models for MD simulations. For each structure, we built a simulated system (Fig. 5) and used MD protocols similar to those used recently for simulating the GluK2closed and GluK2open structures86,94 (see Methods, Supplementary Tables 1–3). All systems were stable throughout the entirety of the MD simulations (Supplementary Fig. 7). Short equilibrium simulations were designed to refine the blocker binding poses and explore the dynamics of these small molecules inside the ion channel pore. Some mobility of the blocker molecules during MD simulations was expected due to a crude and averaged appearance of the cryo-EM density for the blocker molecules (Fig. 2). While all blocker molecules remained in the region determined by the corresponding structures, they exhibited different dynamics and varying extents of binding pose alteration (Fig. 6a).

Fig. 5: MD simulation system.
figure 5

A representation of the MD simulated GluK2PhTx74 system with the receptor subunits A (green), B (pink), C (blue) and D (orange) shown as ribbons, lipid bilayer acyl chains in white and hydrophilic head groups as sticks, BPAM and PhTx-74 in yellow space-filling models, cholesterol as violet sticks, water as white continuum, and K+ and Cl− ions as green and pink spheres, respectively.

Full size image
Fig. 6: Representative structures and water occupancy from MD simulations.
figure 6

a Close-up view of the blocker binding sites after the MD simulations of GluK2NpTx8 (blue), GluK2PhTx74 (green), GluK2KukoA (violet), and GluK2SPM (orange), with blocker molecules (yellow) and residues involved in their binding shown in stick representation. The red dashed lines show the hydrogen bonds between the blockers and the binding site residues, the blue dashed lines show the hydrophobic contacts between the blockers and the hydrophobic side chains of the protein residues. Only two (A and C) of four subunits are shown, with the front and back subunits (B and D) removed for clarity. For GluK2KukoA, two representative structures (Pose 1 and Pose 2) from different parts of simulation are shown. b Similar views as in a, but with water molecules shown as red and white sticks.

Full size image

In general, all blockers formed several stable hydrogen bonds in the narrow region of the selectivity filter via the nitrogen atoms of their polyamine tails, arranged in a pattern similar to that observed for SPM. In addition, the hydrophobic bulky groups of NpTx-8, PhTx-74, and KukoA formed several stable hydrogen bonds and hydrophobic contacts in the channel central cavity. Specifically, NpTx-8 formed stable hydrogen bonds in the selectivity filter with the backbone carbonyl of S624, the side chain of Q622, and the carbonyl oxygen of Q621 in an asymmetric fashion. NpTx-8 also retained a hydrogen bond with the E625 side chain at the intracellular pore entrance. Its headgroup formed a stable hydrophobic contact with T652, which is located at the extracellular boundary of the central cavity, as well as hydrogen bonds and hydrophobic contacts with L645, I648, and S649 of one subunit, and simultaneously, with the identical residues of the diagonally opposing subunit (Fig. 6a). Contacts of the NpTx-8 indole group with the protein excluded water from the central cavity and thus enhanced the hydrophobic effect (Fig. 6b). The pose of NpTx-8 was asymmetric, favoring the entropy of binding to the highly degenerate binding site in the ~4-fold symmetrical homotetrameric ion channel. In longer simulations, the NpTx-8 pose will likely switch between similar residues in other subunits, which is entropically favorable without compromising binding. Among the four simulated structures, NpTx-8 formed the most extensive network of contacts with GluK2, corroborating its highest potency.

Compared to the starting pose in the cryo-EM model, PhTx-74 dramatically changed its pose during the MD simulations to maximize its interactions with the channel pore and formed several hydrogen bonds and hydrophobic interactions with the residues lining the central cavity (Fig. 6a). The resulting pose formed contacts across the channel pore. At the same time, water molecules were excluded from the central cavity (Fig. 6b). The tail of the blocker formed hydrogen bonds with the backbone carbonyls of Q621 and Q622, which were not present in the cryo-EM structural model. Overall, PhTx-74 formed fewer interactions with the channel pore than NpTx-8, explaining its somewhat weaker potency and stronger discrepancy with the cryo-EM model. This example of discrepancy between cryo-EM and MD results illustrates limitations of both methodologies and inherent differences in conditions, including temperature and averaging due to local symmetry. Thus, MD simulations naturally drive each individual ligand and protein residue towards the nearest local energy minimum, resulting in breaking the local symmetry due to entropic preferences, while the local symmetry constraints in cryo-EM structures may be pulling the system out of the local energy minima.

KukoA exhibited two semi-stable poses during simulations (Fig. 6a). The first pose remained almost the same as in the corresponding cryo-EM structure, where KukoA formed hydrogen bonds with the selectivity filter, similar to SPM. In this pose, the KukoA headgroup formed stable hydrophobic contacts with I648 across the central cavity. Water was mostly excluded from this region of the pore (Pose 1 in Fig. 6b). The second pose (Pose 2 in Fig. 6b) was characterized by shifting the blocker molecule extracellularly towards the pore center and letting both headgroups form hydrogen bonds with the protein in a more asymmetric manner. Also, a more compact conformation is generally more favorable than an extended conformation. In our simulations, both poses existed for similar periods of time. However, more extended simulations are needed to determine which one of the two poses is more energetically stable. Overall, the enhanced dynamic behavior of KukoA at the pore binding site and fewer interactions with the pore-lining residues correlate well with its lower potency compared to NpTx-8 and PhTx-74.

The least potent blocker SPM maintained the same pose throughout the entire MD simulation, which was very similar to its initial position in the corresponding cryo-EM structure (Fig. 6a). SPM formed stable hydrogen bonds with Q621 and Q622, as well as the side chains of E625 at the intracellular channel entrance. However, hydrogen bonds formed with surface protein groups solvated in water are typically weak, contributing less than 1 kcal/mol to the binding energy. In addition, the absence of a bulky head group in the SPM molecule allowed numerous water molecules to occupy the central cavity (Fig. 6b). Availability of water molecules to readily solvate polar atoms of SPM suggests a low free energy barrier for blocker unbinding. Once the channel is in the open conformation, the corresponding fast dissociation of the blocker is manifested by the appearance of pronounced hook currents (Fig. 1b).

Discussion

In this study, we investigated the trapping of four channel blockers of GluK2 KARs, NpTx-8, PhTx-74, KukoA, and SPM, which demonstrate a broad range of inhibitory potencies (Fig. 1). While all blocker molecules formed hydrogen bonds with the Q621 and Q622 residues (either with their backbone or side-chain oxygens), the more potent blockers NpTx-8 and PhTx-74 appear to do so in an asymmetric fashion, which may contribute to their stronger binding to the pore via increased entropy of their asymmetric interactions with identical 4-fold degenerate hydrogen bond acceptor sites. The weaker binder SPM retains its symmetric starting pose. In contrast, the second weakest binder, KukoA, adopted an alternative pose that retained the symmetry of the persistent hydrogen bonds of the tail.

Despite the four blockers displaying varying potencies and kinetics of KAR inhibition, they all interact with the channel gating machinery according to a trapping mechanism. The molecular basis of the trapping mechanism becomes apparent when comparing the structures of KAR channel in the closed blocked and open states86,94 (Fig. 7). Similar to toxins and toxin-like molecules in AMPARs67, the KAR channel blockers appear to fit the channel pore as keys fit a lock, with their positively-charged polyamine tails filling up the narrow, negatively-charged selectivity filter and bulky heads occupying the wider central cavity. In comparison to the open-state structure, it is easy to imagine how each one of these molecules can reach their binding site by entering the pore from extracellular space, with their positively charged polyamine tails moving forward toward the intracellular space, driven by the negative membrane voltage (Fig. 7). It is also clear that the negatively charged and narrow selectivity filter creates a barrier for the neutral bulky heads of NpTx-8, PhTx-74, and KukoA to traverse entirely through the pore into the intracellular space. Upon closure of the extracellular gate at the bundle crossing of M3 helices, all four blockers also lose the ability to exit back to the extracellular space, thus being trapped inside the ion channel pore (Fig. 7).

Fig. 7: Trapping of blockers in the closed pore of GluK2 receptor.
figure 7

Frontal section of the TMD surface in the GluK2PhTx74 (a), GluK2NpTx8 (b), GluK2KukoA (c), GluK2SPM (d) and GluK2Open (e, PDB ID: 9B35) structures, with the blocker molecules shown as space-filling models (green) and residues contributing to the selectivity filter and gate shown in sticks (yellow). The surface is colored by electrostatic potential, blue being positively charged, red negatively charged, and white neutral.

Full size image

Interestingly, all four blocker-bound structures, GluK2NpTx8, GluK2PhTx74, GluK2KukoA, and GluK2SPM, were solved in the absence of an agonist, which would typically open the channel and allow blockers to enter the pore. Therefore, these blockers either entered the pore of a ligand-free receptor or during infrequent pore openings due to contamination of solution with traces of agonist (glutamate). However, the possibility of glutamate contamination is unlikely; the structures obtained in this study are from multiple and independent protein purifications. In addition, we did not see an agonist-bound conformation of the receptor during extensive data processing. Further, binding or unbinding of iGluR blockers in the absence of an agonist were previously reported for both NMDARs95 and AMPARs96 and can be explained by dynamic properties of the protein with a plastic selectivity filter that allows spontaneous or blocker-induced openings of the gate. It may be possible to better understand how exactly blockers enter the channel without an agonist by running long MD simulations of the blocker-bound structures and observing putative ways for the blocker to escape the channel.

Calcium-permeable KARs are major contributors to neuropathologies8,30,31. Previous studies have elucidated the structural basis of competitive and noncompetitive antagonism of KARs86,88. However, the structural basis of how polyamines and polyamine-containing channel blockers inhibit KARs has remained poorly understood. Our study uncovers the structural mechanism of KAR channel block and provides molecular templates for the synthesis of new drug molecules targeting KARs. The broad spectrum of KAR channel blocker potencies demonstrated by NpTx-8, PhTx-74, KukoA, and SPM provides a reference for tuning this parameter. Finding the optimal potency, not necessarily the highest, is critical for drug design. For example, the only FDA-approved drugs targeting iGluRs, the ion channel blocker memantine acting on NMDARs and noncompetitive inhibitor perampanel acting on AMPARs, have rather intermediate potencies76,97,98. Nevertheless, new approaches of peptide-mediated targeting to achieve cell-specific modulation of iGluRs can make use of high-affinity drugs as well, as it was recently demonstrated by targeting high-affinity NMDAR channel blocker MK-801 to GLP-1 receptor-expressing brain regions for safe and effective obesity treatment99.

Methods

Constructs

DNA for the full-length rat GluK2 (rGluK2, GenBank CAA77778.1), with V at position 567, C at position 571 and Q at position 621 (the Q/R site), was cloned into pEG BacMam vector for baculovirus-based protein expression in mammalian cells100, with the C-terminal thrombin cleavage site (LVPRG), followed by eGFP and streptavidin affinity tag (WSHPQFEK).

Protein expression and purification

The rGluK2 bacmid and baculovirus were made using standard methods100. The P1 and P2 viruses were produced in Sf9 cells (GIBCO, 12659017) and added to HEK 293S GnTI– cells (ATCC, CRL-3022) incubated at 37 °C and 5% CO2. 12–15 h post-transduction, cells were supplemented with 10 mM sodium butyrate, and the temperature was changed to 32 °C. Cells were harvested 72 h post-transduction using low-speed centrifugation (5500 g, 10 min), washed using 1X PBS pH 8.0, and centrifuged again (5500 g, 15 min). The cell pellet was resuspended in 50 ml of ice-cold lysis buffer consisting of 150 mM NaCl, 20 mM Tris pH 8.0, 5 mM βME, 0.8 μM aprotinin, 2 μg/ml leupeptin, 2 μM pepstatin A, 1 mM PMSF and 20 µM DNQX. Cells were lysed by three cycles of sonication using a Misonix sonicator with an amplitude of 8 for 3 min total, 15 s on and 15 s off. The cell-free lysate was centrifuged (9900 g, 15 min) to remove cell debris. To isolate cell membranes, the supernatant was subjected to ultracentrifugation (186,000 g, 1 h). Cell membranes were mechanically homogenized and solubilized for 2 h in a buffer containing 150 mM NaCl, 20 mM Tris-HCl pH 8.0, 5 mM βME, 20 µM DNQX, and 1% digitonin (Cayman Chemical Company, 14952). Insoluble material was removed by ultracentrifugation (186,000 g, 1 h). The supernatant was added to the pre-equilibrated streptavidin-linked resin (2 ml resin per 1 L of the initial cell culture). The mixture was rotated for 10–14 h at 4 °C. The protein-bound resin was washed with 25 ml of buffer containing 150 mM NaCl, 20 mM Tris-HCl pH 8.0, 0.05% digitonin, and 5 mM βME, and the protein was eluted in 12–15 mL of the same buffer supplemented with 2.5 mM D-desthiobiotin. To remove eGFP and the streptavadin affinity tag, the eluted protein was concentrated and subjected to thrombin digestion (1:200 w/w) at 22 °C for 90 min. The digest reaction was injected into a Superose 6 10/300 GL size-exclusion column (GE Healthcare) equilibrated with a buffer containing 150 mM NaCl, 20 mM Tris-HCl pH 8.0, 0.05% digitonin, and 5 mM βME. The peak fractions corresponding to tetrameric GluK2 were pooled, concentrated to ~5–6 mg/ml, and used for cryo-EM sample preparation. All steps, unless otherwise noted, were performed at 4 °C.

Cryo-EM sample preparation and data collection

Purified rGluK2 at 5–6 mg/ml was incubated with channel blockers, 100 µM NpTx-8, 100 µM PhTX-74, 500 µM KukoA, or 1 mM SPM, for 12–15 h at 4 °C. For 20–30 min before grid preparation, the protein-blocker solutions were supplemented with 500 µM BPAM344. UltrAuFoil R 0.6/1.0, 300 mesh gold grids (EMS, Morrisville, NC) were used for cryo-EM sample preparation. Before sample application, the grids were treated in a PELCO easyGlow cleaning system (Ted Pella, 25 s, 15 mA) to make their surface hydrophilic. Subsequently, 3 µl of the protein sample was applied to each cryo-EM grid. Grids were made using a Vitrobot Mark IV (Thermo Fisher Scientific) set at 100% humidity and 4 °C, blot time of 3 s, blot force of 3, and a wait time of 30 s. The grids were imaged using Leginon 3.5 on a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operating at 300 kV equipped with a post-column GIF Quantum energy filter with slit width set to 20 eV and equipped with either Gatan K3 (Gatan) or Falcon4 (Thermo Fisher Scientific) direct electron detection camera. The images were collected in counting or super-resolution mode across a defocus range of −1.0 to −2.0 µm.

Cryo-EM data processing

Data were processed using cryoSPARC v4.4.1101. Movie frames were aligned using the Patch Motion Correction algorithm. Contrast transfer function (CTF) estimation was performed using the patch CTF estimation. Following CTF estimation, micrographs were manually inspected, and those with outliers in defocus values, ice thickness, and astigmatism, as well as micrographs with lower, predicted CTF-correlated resolution, were excluded from further processing (individually assessed for each parameter relative to the overall distribution). Particles were first picked using a blob picker and then by Topaz102. Junk particles were removed through successive rounds of two-dimensional classification. A set of particles corresponding to tetrameric rGluK2 was used for ab initio reconstruction. Subsequently, particles were cleaned up by successive rounds of heterogeneous refinement. During the processing of all four data sets, we noted that the transmembrane domain (TMD) and ligand-binding domain (LBD) regions of GluK2 displayed excellent density compared to the amino-terminal domain (ATD) due to flexibility in the ATD-LBD linker region relative to the LBD-TMD region, consistent with the previous observations86. Accordingly, the particles were signal subtracted for the ATD and detergent micelle and cleaned up by multiple rounds of ab initio reconstruction and heterogeneous refinement. Finally, homogeneous, non-uniform, and local refinement with a focused mask around the LBD-TMD regions considerably improved the density of the LBD-TMD regions. The unsharpened maps obtained from cryoSPARC were post-processed in Phenix through anisotropic sharpening, significantly improving the density of the LBD-TMD linker regions. Structural biology applications employed in this project adhered to and were configured by SBGrid. Data processing details are summarized in Table 1.

Model building and refinement

The models of blocker-bound GluK2, GluK2NpTx8, GluK2PhTx74, GluK2KukoA, and GluK2SPM were built in Coot103 using the corresponding experimental cryo-EM densities and the GluK2Closed structure (8FWS)86 as a guide. The models were tested for overfitting by shifting their coordinates by 0.5 Å (using shake) in Phenix 1.18104, refining each shaken model against a corresponding unfiltered half map, and generating densities from the resulting models in Chimera. The resulting models were real space refined in Phenix 1.18 and visualized in Chimera105 or Pymol 2.5.2106.

Patch-clamp recordings

DNA encoding rGluK2 (described in the Construct section) was introduced into a pIRES plasmid for expression in eukaryotic cells engineered to produce green fluorescent protein via a downstream internal ribosome entry site107. HEK 293 cells (ATCC, Cat#CRL-1573) grown on glass coverslips in 35 mm dishes were transiently transfected with 1–5 μg of plasmid DNA using Lipofectamine 2000 Reagent (Invitrogen). Recordings were made 24–48 h after transfection at room temperature. Currents from whole cells, typically held at a –60 mV potential, were recorded using Axopatch 200B amplifier (Molecular Devices, LLC), filtered at 5 kHz, and digitized at 10 kHz using low-noise data acquisition system Digidata 1440A and pCLAMP 10.2 software (Molecular Devices, LLC). The external solution contained: 150 mM NaCl, 2.4 mM KCl, 4 mM CaCl2, 4 mM MgCl2, 10 mM HEPES pH 7.3. 7 mM NaCl was added to the extracellular activating solution containing 3 mM Glu to improve visualization of the border between two solutions coming out of a two-barrel theta glass pipette, which allowed its more precise positional adjustment for faster solution exchange. The internal solution contained: 150 mM CsF, 10 mM NaCl, 10 mM EGTA, 20 mM HEPES pH 7.3. Before the experiment, cells were treated in extracellular solution with 0.3 mg/ml ConA (Sigma) for 5–15 min and then transferred to solution without ConA. Rapid solution exchange was achieved with a two-barrel theta glass pipette controlled by a piezoelectric translator. Typical 10–90% rise times were 200–300 µs, as measured from junction potentials at the open tip of the patch pipette after recordings. Data analysis was performed using Origin 2023 software (OriginLab Corporation).

System preparation for molecular dynamics simulations

Initial atomic coordinates for Molecular Dynamics (MD) simulations were obtained from the GluK2NpTx8, GluK2PhTx74, GluK2KukoA, and GluK2SPM structures. Each structure includes LBD and TMD of the full-length receptor, starting with the N-terminal residue S429 of the LBD and ending with the C-terminal residue R874 of the TMD. For all the simulation systems, protein, blocker molecules (NpTx-8, PhTx-74, KukoA, and SPM), BPAM, cholesterol, and experimental lipids from the cryo-EM were kept, and all the other molecules were removed. The GluK2NpTx8 system did not include cholesterol molecules as it lacks them in cryo-EM density. Each simulation box was constructed in CHARMM-GUI membrane builder108,109 by inserting the protein into a POPC bilayer and solvating it with TIP3P water molecules and 150 mM KCl. The systems were set up for MD simulations using the “tleap” module of the AmberTools20 package110. Parametrization of all the ligands was carried out using the general AMBER force field (GAFF)111. The total number of atoms in the final simulation boxes was 311,702 for GluK2NpTx8, 313,114 for GluK2PhTx74, 312,549 for GluK2KukoA, and 315,207 for GluK2SPM system. Each simulation system comprised approximately 73,000 water molecules, 200 K+ and Cl− ions, and 500 lipid molecules.

Molecular dynamics simulation protocols

“pmemd.cuda” program of the Amber20 molecular dynamics software package was used for all MD simulations110. Amber FF99SB–ILDN112 force field parameters were used for protein and ions, TIP3P model for water, and Lipid14113 force field parameters for lipids. All equilibration and production simulations were performed in an NPT ensemble at 300 K temperature and 1 bar pressure with anisotropic pressure scaling. The Langevin thermostat with a collision frequency of 1 ps−1 was used to control the temperature, and the pressure was maintained using a Berendsen barostat with a relaxation time of 1 ps, as implemented in Amber20. All covalent bonds involving hydrogen atoms were constrained using the SHAKE algorithm114, with the integration time step of 2 fs. The long-range electrostatic interaction calculations were performed using the Particle Mesh Ewald (PME) method115, with a non-bonded interaction cutoff radius of 10 Å. Periodic boundary conditions were applied in all directions.

Each system was minimized prior to MD simulations while keeping restraints on protein Cα and ligand (BPAM and blocker) heavy atoms. Next, water and ions were equilibrated at constant volume MD simulations as the temperature was gradually increased from 0 to 300 K, with all protein, ligand, and lipid heavy atoms harmonically restrained at their energy-minimized positions with the force constant of 40 kcal mol–1 Å−2. The systems were then equilibrated for 100 ns at constant pressure MD simulations, gradually releasing the restraints on the protein and ligands to 0.5 kcal mol–1 Å−2. Production simulations (one run per blocker) were carried out for ~600 ns for GluK2NpTx8, GluK2PhTx74, GluK2KukoA, and ~800 ns for GluK2SPM systems without restraints.

Molecular dynamics trajectory analysis

Post-processing and analysis of the trajectories were carried out using CPPTRAJ116 module of AmberTools20 and VMD 1.9.4117. VMD 1.9.4 was used to visualize trajectories, and PyMOL106 (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.) was used to generate simulation snapshot figures. Heavy-atom contact frequencies were calculated in VMD when any heavy-atom of the protein and heavy-atom of the blocker molecule is within 4 Å. Hydrogen bond and hydrophobic interaction analysis were carried out in CPPTRAJ. To determine the list of potential hydrogen bonds, a geometric criterion where the donor-acceptor distance is within 3.6 Å with the donor-hydrogen-acceptor angle cutoff of 90o was used. Hydrophobic contacts were determined when any two pairs of carbon atoms or any carbon and any sulfur atom were at a distance shorter than 4 Å. The persistence of these contacts was reported with duty fraction as a percentage which is defined as the ratio of the duration of bond on time and the total simulation time. Protein-ligand contacts analysis was performed in VMD. Ligand RMSD was computed for heavy atoms of the ligand with reference to their initial positions, with each frame of the trajectory aligned using the coordinates of protein Cα atoms for residues 631 to 661 in the M3 helix. Protein TMD RMSD was computed for Cα atoms of the M1, M2, M3, and M4 TMD helices. Representative structures were generated by averaging the coordinates over the production simulation and selecting the lowest RMSD structure. For GluK2SPM and GluK2NpTx8 systems, representative structures were calculated from the entire production trajectory. For GluK2KukoA system, the first representative structure (GluK2KukoA-1) was extracted from the first 250 ns of the production trajectory, and the second representative structure (GluK2KukoA-2) was extracted from the 350–500 ns portion of the trajectory in which the blocker molecule adopted a new conformation and stabilized. For GluK2PhTx74 system, the representative structure was extracted from the production trajectory after the blocker conformation is stabilized by excluding the first 120 ns.

Statistics and reproducibility

Statistical analysis (Figs. 1a–c and 4a–b) was performed using Origin 2023. Statistical significance was calculated using One-Way ANOVA followed by Fisher’s least significant difference test. In all figure legends, n represents the number of independent biological replicates. All quantitative data were presented as mean ± SEM.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.