Kainate receptor subunits are differentially and spatially regulated in vertebrates. We analyzed publicly available human transcriptomic data to determine whether the exon encoding the GluK1 ATD splice (exon 9) was differentially expressed in the human brain. RNASeq data analysis of GRIK1 (Ensembl ID: ENSG00000171189) collected from BrainSpan ATLAS indicated that exon 9 is present in multiple areas of the brain that also show prominent GluK1 expression, such as the cerebellar cortex (Figure 1; Figure 1-figure supplement 1). Variable expression of exon 9 in different areas of the brain prompted us to delve deeper into the molecular mechanisms and perform a structure-function analysis of this GluK1 splice variant.

Previous studies on functional analysis of GluK1 receptors have primarily focused on GluK1-2 isoform that lacks the ATD splice insert. Therefore, we carried out extensive electrophysiological investigation to assay the functional differences between GluK1-1a and GluK1-2a. The only difference between these two variants is the presence of 15 amino acid (KASGEVSKHLYKVWK) insertion in the R2 subdomain (lower lobe) of the GluK1-1a ATD. We evaluated multiple gating properties of the two variants by whole-cell patch-clamp electrophysiology. Interestingly, we observed that GluK1-1a receptors desensitize significantly slowly when compared to GluK1-2a (Ʈ_Des, GluK1-1a: 5.21 ± 0.50 ms; GluK1-2a: 3.55 ± 0.23 ms **P=0.0092) on prolonged treatment with 10 mM glutamate (Figure 2A; Table 1). We also tested receptor desensitization on application of saturating concentrations of kainate (1mM). Since the receptors displayed very slow desensitizing kainate currents, instead of the rate (Ʈ_Des), we calculated the % desensitization values at 1 s for comparison. We observed that the % desensitization with kainate was distinct for both ATD splice variants with significantly slower desensitization in the variant with the splice insert (GluK1-1a: 72.06 ± 2.33 %, GluK1-2a: 93.2 ± 0.55 ***P=0.0006) consistent with the glutamate evoked currents (Figure 2B; Table 1).

Next, we evaluated glutamate sensitivity for the two variants. Dose-response experiments (glutamate, GluK1-1a: 0.1-2 mM, and GluK1-2a: 0.01-3 mM), showed a significantly lower potency of glutamate for GluK1-1a variant compared to the non-spliced form (EC₅₀ glutamate, GluK1-1a: 379.3 ± 52 µM, GluK1-2a: 187.7 ± 33 µM *P=0.0129 (Figure 2C Table 1). However, for high-affinity agonist kainate, the dose-response curves for both variants were similar likely indicating differences in the stability of glutamate versus the kainate-bound states in GluK1-1a and GluK1-2a (Figure 2D; Figure 2-figure supplement 1). Thus, the potency of kainate (1mM) versus glutamate (10 mM) (I_K/I_G ratio) is significantly higher for GluK1-1a compared to GluK1-2a (GluK1-1a: 1.51 ± 0.13, GluK1-2a: 0.56 ± 0.4 ****P< 0.0001) (Figure 2D; Table 1).

Further, we investigated the voltage-dependent polyamine block and found significant differences between the two variants. The presence of splice residues seemed to enhance outward currents at positive potentials in GluK1-1a compared to GluK1-2a (rectification index, +90mV /-90 mV; GluK1-1a = 0.96 ± 0.11; GluK1-2a = 0.61 ± 0.10 *P=0.0385) without affecting the reversal potential (Figure 2E; Table 1). It’s unclear how splice residues situated ∼30 Å away from the TM domain affect the pore properties. This property is mainly affected by the TM2 region of the KAR (Bowie and Mayer, 1995), although a recent report in which the ATD of GluK2 was deleted showed enhanced rectification (Li et al., 2019). It’s likely that the presence of splice residues alters pore structure thereby affecting rectification (Perrais et al., 2009). We also carried out outside-out patch recordings to examine GluK1-1a receptors. However, we observed extremely weak electrical currents with low amplitudes when GluK1-1a was expressed in isolation. As a consequence, the data we obtained from these recordings did not provide reliable results for curve fitting or thorough analysis.

Neto 1 and Neto 2 proteins significantly influence the surface expression, synaptic localization, and functional properties of the GluK1-2a receptors (Copits et al., 2011; Palacios-Filardo et al., 2016; Sheng et al., 2015). It has been demonstrated that desensitization of GluK1-2a is accelerated by Neto1 but delayed by Neto2 (Sheng et al., 2015; Copits et al., 2011; Palacios-Filardo et al., 2016). Hence, to understand the influence of these KAR auxiliary proteins on GluK1-1a, we performed an electrophysiological analysis of GluK1-1a and GluK1-2a receptors co-expressed with either Neto1 or Neto2. Co-expression of Neto1 hastened desensitization of GluK1-1a significantly but not of GluK1-2a at saturating glutamate concentrations (Ʈ_Des, GluK1-1a +Neto1: 3.56 ± 0.22 ms **P=0.0090; GluK1-2a +Neto1: 4.32 ± 0.34 P=0.1495). On the other hand, Neto2 led to a ∼ 13.4-fold decrease in the desensitization rate of GluK1-1a (Ʈ_Des, GluK1-1a: 5.21 ± 0.50 ms, +Neto2: 69.62 ± 9.98 ms **P=0.0024) while the desensitization rate of GluK1-2a was delayed only by ∼ 6.1-fold (Ʈ_Des, GluK1-2a: 3.55 ± 0.23 ms, +Neto2: 21.68 ± 2.64 ms **P=0.0017). Thus, while Neto1 accelerated the desensitization of GluK1-1a, Neto2 significantly slowed it at saturating glutamate concentrations. When both GluK1-1a and GluK1-2a were coexpressed with Neto2, the rate of desensitization for GluK1-1a was approximately 3.2 times slower compared to GluK1-2a (***P=0.0009) (Figure 3A; Table 1. Thus, the presence of splice insert leads to differential modulation of GluK1 desensitization by Neto proteins.

Moreover, Neto1 also enhanced the recovery from desensitization for both the variants (Ʈ_Recovery, GluK1-1a: 3.53 ± 0.81 s, +Neto1: 0.68 ± 0.07 s *P=0.0390; GluK1-2a: 5.31 ± 0.50 s +Neto1: 1.15 ± 0.12 s ***P=0.0002). GluK1-1a recovers ∼ 1.7 times faster than GluK1-2a (*P=0.0125) when co-expressed with Neto1. Neto 2, on the other hand, slowed recovery for both variants to a similar extent (Ʈ_Recovery, GluK1-1a: 3.53 ± 0.81 s, +Neto2: 8.32 ± 0.81 s **P=0.0044; GluK1-2a: 5.31 ± 0.50 s, +Neto2: 7.91 ± 0.71 s *P=0.0430) and did not show differential modulation (Figure 3B; Table 1). In addition, both Neto1 and Neto2 increased the potency of glutamate for GluK1-1a by 9.7-fold (39 ± 10 µM) and 11.2-fold (34 ± 8 µM), respectively (Figure 3C; Figure 3-figure supplement 1A). This is similar to the effect observed in GluK1-2 receptors whereby the glutamate EC₅₀ was shown to increase by Neto proteins [Neto1: 34-fold and Neto2: 7.5-fold (Palacios-Filardo et al., 2016) and Neto1/2: 10-30X (Fisher, 2015)].

Since we observed a significant increase in the I_K/I_G ratios in GluK1 due to the presence of splice insert, we next aimed to determine the influence of co-expressing Neto proteins on this parameter. Interestingly, we observed that while Neto 1 and Neto2 reduced the I_K/I_G ratios in GluK1-1a (GluK1-1a: 1.51 ± 0.13, +Neto1: 1.25 ± 0.04 P=0.1500, +Neto2: 1.0 ± 0.07 **P=0.0090), they increased it for the non-spliced variant (GluK1-2a: 0.56 ± 0.04, +Neto1: 1.37 ± 0.12 ****P<0.0001, +Neto2: 1.16 ± 0.04 ****P<0.0001) (Figure 3D; Figure 3-figure supplement 1B; Table 1) (Zhang et al., 2009) highlighting the effect of splice residues. Interestingly, this differential modulation of the two variants by Neto proteins resulted in comparable I_K/I_G ratios.

Next, we investigated the effects of Neto1 and Neto2 on the voltage dependent endogenous polyamine block since presence of splice insert had enhanced the outward rectification of GluK1. We observed that both Neto1 and Neto2 significantly enhanced the outward rectification of GluK1-2a (GluK1-2a: 0.61 ± 0.10, +Neto1: 1.14 ± 0.14 *P=0.0338, +Neto2: 1.37 ± 0.10 **P= 0.0022). However, they did not significantly increase it for GluK1-1a (GluK1-1a: 0.96 ± 0.11, +Neto1: 1.16 ± 0.09 P=0.1880, +Neto2: 0.80 ± 0.05 P=0.2401) and did not show differential modulation (Figure 3E & 3F; Table 1).

We also calculated desensitization and deactivation kinetics in excised outside-out patches. However, GluK1-1a receptors when expressed alone, exhibited low peak amplitudes in outside-out recordings that prevented reliable calculation of gating kinetics. Hence, we only compared properties of receptors co-expressed with Neto proteins and the results obtained were consistent with those obtained from whole-cell recordings. Neto 2 slowed down desensitization of GluK1-1a by ∼ 1.5 times compared to GluK1-2a (Ʈ_Des, GluK1-1a +Neto2: 31.89 ± 4.08 ms; GluK1-2a +Neto2: 20.91 ± 2.11; P=0.0665) (Figure 3G, Table 2). The desensitization rates for GluK1-1a and GluK1-2a receptors coexpressed with Neto1 was also significantly altered (Ʈ_Des, GluK1-1a +Neto1: 4.83 ± 0.46 ms; GluK1-2a +Neto2: 2.84 ± 0.35; *P=0.0310) (Figure G, Table 2. Faster solution exchange times in excised patch recordings allowed us to also measure deactivation kinetics using 1 ms application of 10 mM glutamate. Surprisingly, unlike desensitization, for receptors coexpressed with Neto2, the deactivation rate of GluK1-1a is significantly faster compared to that of GluK1-2a (Ʈ_Dea, GluK1-1a +Neto2: 5.18 ± 0.65 ms; GluK1-2a +Neto2: 10.74 ± 1.48; **P=0.0077). In contrast, Neto1 did not significantly alter deactivation kinetics of both GluK1 variant receptors (Ʈ_Dea, GluK1-1a +Neto2: 2.83 ± 0.20 ms; GluK1-2a +Neto2: 2.14 ± 0.37; P) (Figure 3H, Table 2). Thus, both whole-cell and excised patch recordings confirm the unique functional properties and differential modulation by Neto proteins due to the presence of fifteen amino acid insert in GluK1-1a receptors.

Since our electrophysiological analysis revealed functional differences between the two splice variants and their modulation by Neto proteins, we generated receptors with splice residue mutants and performed functional assays to identify key residues. The splice insert (KASGEVSKHLYKVWK) is dominated by positively charged residues and contains four lysins and one histidine. We hypothesized that these charged residues might affect the interactions at the ATD-LBD interface and influence receptor functions. To investigate this, we prepared charge-reversal (K/H to E) and charge neutral mutants (K/H to A) (Figure 4-table supplement 1) and carried out functional assays. Our cell surface biotinylation assay showed that all mutants (glutamate or alanine) reached the cell surface efficiently (data not shown). However, the charge neutral mutants (K/H to A) gave either very low peak amplitudes (<40 pA) or were not functional, and hence, were not included in the study. The charge reversal mutants K₃₆₈-E, and K_375/379/382H₃₇₆-E revealed fascinating insights into the role of splice residues in altering GluK1 receptor properties (Figure 4A). Interestingly, both mutants, K₃₆₈-E and K_375/379/382H₃₇₆-E exhibited a significantly slower rate of glutamate-evoked desensitization compared to wild-type GluK1-1a (Ʈ_Des, GluK1-1a: 5.21 ± 0.50 ms; K₃₆₈-E: 7.89 ± 1.14 ms *P=0.0334; K_375/379/382H₃₇₆-E: 9.61 ± 1.47 *P=0.0290) (Figure 4B; Table 1. We also observed a significant delay in the recovery from desensitized state for K₃₆₈-E mutant (K₃₆₈: 5.61 ± 0.95 s *P=0.0417) compared to wild-type GluK1-1a. In addition, K_375/379/382H₃₇₆-E mutant also exhibited a slowdown in the recovery (K_375/379/382H₃₇₆-E: 4.83 ± 0.31 s P=0.2774) (Figure 4C; Table 1). Our investigations of glutamate-and kainate-evoked responses for wild-type and mutant receptors considering their peak amplitudes (I_K/I_G) revealed a significant decrease for the K₃₆₈-E mutant (GluK1-1a: 1.51 ± 0.13; K₃₆₈-E: 0.81 ± 0.13 **P=0.0037) and a reduction was observed for K_375/379/382H₃₇₆-E receptors (1.17 ± 0.28 P=0.3733) compared to wild-type (Figure 4D; Table 1). This observation is reciprocal to the effect of splice when compared to the non-spliced form indicating importance of these residues in influencing receptor desensitization and recovery. A similar reversal in trend was also observed for the measurements of rectification index for these mutants at positive and negative potentials (+90 mV and -90 mV). The rectification index was significantly reduced in the case of mutant K_375/379/382H₃₇₆-E (K_375/379/382H₃₇₆-E: 0.62 ± 0.14 *P=0.0499). Surprisingly, no outward rectification was observed for the K₃₆₈-E mutant and must be further investigated to fully understand the reasons (Figure 4E; Figure 5-figure supplement 2C; Table 1).

Next, we investigated the effects of splice mutants on receptor modulation by Neto proteins. To test whether mutations in splice residues disrupted interactions with Neto proteins, we performed receptor pull-downs using an antibody against the His-tag of the receptor. Our results showed that the mutant receptors could efficiently pull down Neto1 (detected using the Neto1 antibody) and Neto2 (detected using the GFP antibody) (Figure 5-figure supplement 1) suggesting that mutants don’t completely abolish GluK1-1a and Neto interactions.

Furthermore, we conducted experiments using electrophysiology to investigate whether coexpression of Neto proteins can restore the functionality and influence the functions of splice mutants. We observed that mutants K₃₆₈-E and K_{368/375/379/382}H₃₇₆-E, desensitize significantly slower in the presence of Neto1 (Ʈ_Des, GluK1-1a +Neto1: 3.56 ± 0.22 ms; K₃₆₈-E +Neto1: 4.37 ± 0.25 ms *P=0.0393; K_375/379/382H₃₇₆-E +Neto1: 3.74 ± 0.20 ms P=0.5390; K_{368/375/379/382}H₃₇₆-E +Neto1: 4.86 ± 0.67 ms P=0.1242) (Figure 5A; Figure 5-figure supplement 2A; Table 1). This observation suggests that GluK1-1a modulation by Neto1 is influenced by the splice residues. On the other hand, Neto2 does not have a significant effect on the desensitization of these mutant receptors compared to wild-type-Neto2 (Figure 5A; Figure 5-figure supplement 2A; Table 1).

Similarly, K₃₆₈-E and K_{368/375/379/382}H₃₇₆-E mutants recover significantly faster from the desensitized state when coexpressed with Neto1. On the other hand, the recovery from the desensitized state is slower on coexpression of mutants with Neto2. Thus, the pattern of mutant receptor recovery from desensitized state is comparable to that of wild-type receptors, indicating that K₃₆₈-E and K_375/379/382H₃₇₆-E mutants do not influence receptor recovery (Figure 5B; Table 1).

Furthermore, to determine whether the agonist efficacy of mutant receptors changed in the presence of Neto proteins, I_K/I_G ratios were measured. Neto1 increased the agonist efficacy for the K₃₆₈-E and K_375/379/382H₃₇₆-E mutants (K₃₆₈-E +Neto1: 6.99 ± 0.47 ***P=0.0002; K_375/379/382H₃₇₆-E +Neto1: 1.69 ± 0.15 *P=0.0326; K_{368/375/379/382}H₃₇₆-E +Neto1: 1.15 ± 0.20 P=0.6495) (Figure 5C; Figure 5-figure supplement 2B; Table 1). Similarly, Neto2 also considerably increased the I_K/I_G values for mutant K₃₆₈-E, rescuing the kainate efficacy (K₃₆₈-E +Neto2: 2.13 ± 0.13 ***P=0.0007), while it does not seem to have any significant effect on the K_375/379/382H₃₇₆-E mutant (K_375/379/382H₃₇₆-E +Neto2: 0.86 ± 0.08 P=0.2045) (Figure 5C; Figure 5-figure supplement 2B; Table 1).

Consistent with the observation above, our examination of the rectification index revealed a significant increase in outward current for K₃₆₈-E mutant (K₃₆₈-E +Neto1: 2.84 ± 0.16 **P=0.0021). However, K_375/379/382H₃₇₆-E did not show any difference compared to wild-type receptors, while K_{368/375/379/382}H₃₇₆-E displayed a decrease in the outward current (K_375/379/382H₃₇₆-E +Neto1: 1.23 ± 0.07 P=0.5351; K_{368/375/379/382}H₃₇₆-E +Neto1: 0.59 ± 0.13 **P=0.0081) (Figure 5D; Figure 5-figure supplement 2C; Table 1). Interestingly, similar to Neto1, Neto2 was also able to rescue the outward rectification of the K₃₆₈-E mutant (K₃₆₈-E +Neto2: 0.76 ± 0.26), suggesting the importance of this residue in receptor modulation by Neto proteins (Figure 5D; Figure 5-figure supplement 2C; Table 1).

Thus, our analysis of the gating properties of GluK1-1a mutants co-expressed with Neto proteins suggests that positively charged residues at positions 368, 375, 376, 379, and 382 in the splice insert influence receptor modulation by Neto proteins. While Neto1 could primarily influence receptors with K/H to E splice mutations, Neto 2 did not show such differential effects and could not influence the gating properties of the mutants studied here significantly. This suggests that while the modulation of the receptor by Neto 1 is affected by mutations in splice insert, the modulation by Neto 2 remains largely unaffected.

To evaluate the effects of splice residues on domain organization and structure of GluK1-1a receptors we pursued its structure determination via single particle cryo-EM. Construct optimization was carried out to improve the expression and stability of purified protein. Briefly, the free cysteines in the TM1 region were mutated (C₅₅₂Y, C₅₅₇V) based on the sequence analysis with kainate and AMPA receptors, and this construct was named as GluK1-1a_EM (Figure 6-figure supplement 1 and 2). The whole-cell patch clamp showed that GluK1-1a_EM was functional (Figure 6-figure supplement 3).

The structures of GluK1-1a_EM were determined using single-particle cryo-EM. The receptors were either DDM solubilized or reconstituted in lipid nanodiscs and trapped in a desensitized state using 2 mM of high-affinity agonist 2S, 4R-4-methyl glutamate (SYM2081) (Figure 6-figure supplement 4). A resolution of ∼5.2 Å was achieved for the receptors in lipid nanodiscs, but the transmembrane region was not resolved due to an orientation bias (Figure 6; Figure 6-figure supplement 5; Figure 6-table supplement 1). The full-length receptor in detergent micelles had a resolution of 8.2 Å, including the transmembrane region, which was ∼8 Å for the extracellular domain (Figure 6-figure supplement 5; Figure 6-table supplement 1).

A tetrameric receptor model was built based on the crystal structures of GluK1-1a ATD, GluK1 LBD (kainate-bound state; PDB:3C32), and the TM domain based on a highly identical GluK2_EM (PDB:5KUF). Our cryo-EM map represented ATD residues-1-398, but the density corresponding to the ATD splice (368-382) was poorly resolved. The ATD-LBD linkers were resolved for all subunits (A to D) in both structures, the S1 and S2 domains were built entirely and the TMD (TM1, TM3 and TM4) was built only for detergent-solubilized receptors (Figure 6; Figure 6-figure supplement 6 and 7). For receptors reconstituted in lipid nanodiscs, we observed only the TM3 bundle. TM2 was not resolved in either dataset (Figure 6; Figure 6-figure supplement 7). Extra densities were observed in the ECD layer of the GluK1-1aEM ND map, which coincided with potential N-linked glycosylation sites (Figure 6-figure supplement 8). GluK1-1a_EM maps showed general conservation of the architecture of kainate receptors captured in the desensitized state (Khanra et al., 2021; Kumari et al., 2019; Meyerson et al., 2016; Selvakumar et al., 2021). Consistent with earlier studies on homomeric and heteromeric KARs in the desensitized state, GluK1-1a exhibited a modular organization with three layers, namely, the two-fold symmetric ATD, quasi-four-fold symmetric (LBD and four-fold symmetric TMD. The presence of the ATD splice insert in GluK1-1a did not affect the arrangement of the receptor domains in the desensitized state (Figure 6; Figure 6-figure supplement 7).

To understand the functional differences between GluK1-1a and GluK1-2a, as well as their modulation by Neto proteins, a whole-cell patch-clamp analysis was performed. HEK293 WT mammalian cells were seeded on siliconized glass coverslips in 35 mm dishes using Dulbecco’s Modification of Eagle’s Medium (DMEM) containing 10 % fetal bovine serum (FBS), 2 mM glutamine, and 10 units/mL Penicillin-Streptomycin. Cells were transfected with GRIK1-1a or GRIK1-2a cloned in pRK7 (in the presence or absence of rNeto1/rNeto2 cloned in IE-pRK8 as indicated) with EGFP expressing plasmid using Xfect (Clontech) according to manufacturer’s instructions. Similar protocols were followed to test the functionality of the GRIK1-1a_EM and GRIK1-1a splice mutant constructs. The whole-cell patch-clamp recording was performed with an EPC 10 USB amplifier (HEKA) 24-48 h post-transfection. Cells were lifted using 1.5 mm diameter thin wall glass capillary tubes (30-0066 Harvard Apparatus), pulled to a fine tip with a Sutter P-1000 micropipette puller (Sutter Instruments, Novato, CA) containing internal solution (10 mM HEPES pH 7.4, 100 mM CsF, 30 mM CsCl, 4 mM NaCl, 0.5 mM CaCl₂, 5 mM EGTA and 300 mOsm). Cells were continuously perfused with external/bath solution (10 mM HEPES pH 7.4, 150 mM NaCl, 2.8 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂ and 300 mOsm) through a peristaltic perfusion system (Multichannel systems). Current was measured by holding the membrane at -60 mV, using a 2-kHz low-pass filter. Ligands (10 mM glutamate, 1 mM kainate, 2 mM SYM2081, or 10 µM UBP301) were applied through a double-barrelled theta pipette connected with an ultra-fast piezo-based perfusion system (Multichannel systems) via v8 Perfusion Fast-Step System (VC-77SP). Recordings were controlled and measured using Patchmaster-v2x90.2 (Heka Elektronik). The raw files were analyzed using Clampfit 11.2 and

Fitmaster-v2x90.4. All traces were normalized before being used for calculations. The exponential fit was used to estimate the rate of desensitization (Ʈ_Des) for 100 ms glutamate application, measured as the decline of the current from 80% of its peak amplitude. Mean-weighted Tau (τ_Des) values were determined using the single exponential, two-term fitting (Levenberg-Marquardt). To compute the recovery from desensitization, a paired-pulse experiment was performed, where the amplitudes of the test pulse were normalized to that of the desensitizing pulse (calculated as relative amplitude); and plotted in comparison to the time between the desensitizing pulse and test pulse in seconds. The rate of recovery (Ʈ_Recovery) was obtained by fitting the one-phase association exponential function. Dose-response experiments were performed for GluK1-1a (co-expressed with or without Neto proteins) and GluK1-2a with different concentrations of glutamate or kainate in the range of 1 µM to 3 mM or 0.5 µM to 600 µM, respectively. Dose-response values were calculated as the percentage of maximum response against log[agonist] concentrations and fitted using variable slope (Hill’s equation) in GraphPad Prism 8.0.1. For calculating kainate efficacy (I_K/I_G), the ratio of peak amplitudes evoked by kainate versus glutamate was employed. For the calculation of the rectification index, a ratio of peak amplitudes obtained at +90/-90 mV was utilized. Current-voltage (IV) plots were prepared using the voltage ramp from -90 to + 90 mV with a 10 mV increment step, and the current amplitude was normalized to that obtained at -90 mV. For kainate-evoked currents, the percentage desensitization was calculated. The steady-state current measured at the end of 1 s kainate application was divided by peak current. This ratio was then subtracted from 1 and multiplied by 100 to give percentage (%) desensitization (Fisher and Fisher, 2014).

HEK 293-T/17 cells were used for outside-out patch experiments 48-96 h post transfection. Cells were transfected with GRIK1-1a or GRIK1-2a cloned in pRK7 (in the presence or absence of rNeto1/rNeto2 cloned in IE-pRK8 as indicated) with EGFP expressing plasmid at DNA ratios of 3:0.5 for receptor alone and 2:8:0.5 with Netos respectively using TransIT®-LT1 Transfection Reagent according to manufacturer’s instructions.

The extracellular solution used for the experiment contained 10 mM HEPES pH 7.3, 150 mM NaCl, 2.8 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂ and 10 mM glucose, The intracellular solution contained 10 mM HEPES pH 7.3, 110 mM CsF, 30 mM CsCl, 5 mM EGTA, 4 mM NaCl, and 0.5 mM CaCl2.

Patch micropipettes were pulled using a P-97 Sutter puller, and the resistance was maintained at 2.3-2.6 megaohms. After forming the whole-cell configuration, the patch was pulled away from the cell to facilitate an outside-out patch. The cells were voltage-clamped at -70 mV, and a saturated concentration of glutamate (10 mM) was rapidly applied using a three-barrelled theta glass attached to a Siskiyou MXPZT-300 solution switcher. Glutamate applications of 1 ms were used to determine deactivation, while applications of 100 or 1000 ms were used to determine desensitization.

Comparisons between wild-type receptors, EM, or mutant constructs were obtained using an unpaired t-test (two-tailed, with or without the Welch test) or by Brown-Forsythe and Welch ANOVA followed by Dunnett’s multiple comparisons. Statistical analysis was carried out in GraphPad Prism, version 8.0.1. P values < 0.05 were considered statistically significant and are reported (* = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001).

Based on our electrophysiology analysis of the wild-type GluK1-1a/GluK1-2a receptors, structural analysis of GluK1-1a_EM, and recent reports that suggest that the presence of the positive patches in GluK2 ATD affects the interaction with Neto proteins (He et al., 2021b; Li et al., 2019; Vinnakota et al., 2021), we performed SDM to understand the role of splice residues in the receptor kinetics. All splice mutations were introduced in the wild-type GRIK1-1a pRK7 construct for electrophysiology, as well as the GRIK1-1a_EM-EGFP-His₈-pEGBacMam construct for surface expression and pull-downs using the ligation-free cloning approach (Zhang et al., 2017). Clones were confirmed by sequencing. Initially, the positively charged and other residues of the splice were substituted with alanine. Later, charge reversal mutants (K and/or H to E) were also prepared to understand their role in the receptor kinetics as well as interaction with Neto proteins. The mutants prepared are summarized in Figure 4-table supplement 1.

To understand the effect of mutations on GluK1-1a and Neto1/2 interaction, co-immunoprecipitation was carried out using an anti-His monoclonal antibody against the receptor. Similar constructs and transfection protocols were followed for surface expression analysis. Cells were pelleted 65-70 h post-transfection and washed with TBS (20 mM Tris pH 8, 150 mM NaCl). These cells were sonicated and solubilized in 500 uL lysis buffer (20 mM Tris pH 8, 150 mM NaCl, 1% glycerol, protease inhibitor cocktail, 30 mM DDM). Post solubilization, debris was removed by centrifuging at 17,000 x g for 45 minutes at 4 °C. The supernatant was incubated for pre-clearing with 20 µL of pre-equilibrated Protein A agarose beads (Thermo Scientific) for 1 h on a rotator at 4 °C. Post-pre-clearing, ∼ 10% sample was saved as input, and the rest was used for pull-downs. Simultaneously, an antibody (anti-His, Cell Signaling Technology) was added to 40 µL pre-equilibrated Protein A agarose (Thermo Scientific) and incubated for 1 h at 4 °C, followed by the addition of pre-cleared lysate. It was further incubated at 4 °C overnight (14-16 h). The unbound fraction was removed and washed four times with 500 µL wash buffer (20 mM Tris pH 8, 150 mM NaCl, 1% glycerol, 0.75 mM DDM) to remove non-specific interactions. Protein was eluted in 30 µL elution buffer (100 mM Tris pH 6.8, 12% glycerol, 4% SDS, 10 mM DTT, 2% β-mercaptoethanol) by heating at 95 ° C for 10 minutes. Rabbit IgG controls were set up to confirm the validity of the experiment. To analyze the pull-down, 8% SDS-PAGE followed by western transfer was performed. The immunoblots were probed using anti-actin (Sigma), anti-His (Cell Signaling Technology), and anti-Neto1 (Sigma)/anti-GFP (Sigma) to detect the internal control (actin), the receptors and the co-immunoprecipitated Neto proteins respectively.

To obtain functional GluK1 receptors, rat GRIK1 with ATD splice insert and the shortest C-terminal domain, GRIK1-1a (1-871 amino acid residues), was cloned in the pEGBacMam vector. The receptor was cloned in-frame with a thrombin recognition site (LVPRGSAAAA), EGFP (A₂₀₇K; non-dimerizing mutant), and His₈ at the C-terminus. The wild-type protein generated very low amounts of the tetramer, as observed in fluorescence-assisted size-exclusion chromatography (FSEC). Therefore, based on the alignments of the sequence with GluK2_EM, GluK3EM and GluA2, we mutated free cysteines in the TM1 region to residues corresponding to those of GluK2_EM or GluA2 [1x Cys (C₅₇₆S), 2x Cys (C₅₅₂Y, C₅₅₇V) and 3x Cys (C₅₅₂Y, C₅₅₇V, C₅₇₆S)] to obtain good yields of the tetrameric receptor. All clones were confirmed by restriction digestion and sequencing. The expression of all the mutants was confirmed by immunoblotting against His-tag and FSEC. GluK1-1a with 2x Cys mutations (C₅₅₂Y, C₅₅₇V) gave us the best receptor quality as observed in FSEC and therefore was used as GluK1-1a_EM for large-scale purification for structural studies.

Three liters of Human embryonic kidney (HEK) 293 GnTI^- suspension-adapted cultures (∼1.5-2.0 x 10⁶ cells/mL) at 0.5 µg/mL were transfected (or infected with P2 virus) with GRIK1-1a_EM plasmid [or virus, prepared in DH10Bac as per established protocol (Goehring et al., 2014)] using polyethyleneimine (PEI-MAX, Polysciences; 1 DNA: 3 PEI w/w) as the transfection agent (or, infected at a multiplicity of infection of ∼2). To boost the protein production, 10 mM sodium butyrate (Sigma) was added 16 h post-transfection/infection, and cultures were incubated at 30°C for protein expression. Cells were harvested 65-70 h after transfection / infection, washed with buffer containing 20 mM Tris pH 8 and 150 mM NaCl, and stored at -80 ° C until further processing. The cell pellet was resuspended (20 mL/L culture volume) in lysis buffer (20 mM Tris pH 8, 150 mM NaCl, and protease inhibitor cocktail, Roche). These resuspended cells were disrupted using sonication (QSonica sonicator, 3 cycles of 90 s; 10 s ON/ 20 s OFF; temperature cut-off: 15 °C). The lysate was clarified using low-speed spin (4307 x g for 20 minutes at 4 °C). The membranes were pelleted using ultracentrifugation (118991 x g for 1 h at 4 °C). Solubilization of membrane fraction was done in buffer containing 20 mM Tris pH 8, 150 mM NaCl, 30 mM DDM (Anatrace, D310LA), 5% glycerol, 10 mM imidazole, and protease inhibitor cocktail for 1 h on rotator at 4 °C. The non-solubilized fraction was separated by centrifugation at 47850 x g for 1 h at 4 °C and 4 mL of cobalt-charged TALON resin (Clontech, Takara) was added to the solubilized fraction for batch binding (3 h at 4 °C). Beads were harvested. The column was packed and washed with wash buffer (20 mM Tris pH 8, 150 mM NaCl, 1 mM DDM, 1% glycerol, 40 mM imidazole) until OD₂₈₀ reached zero. The bound receptor was eluted with elution buffer (20 mM Tris pH 8, 150 mM NaCl, 1 mM DDM, 1% glycerol, 250 mM imidazole). The peak fractions were pooled, concentrated at 1.3 mg / mL, and digested overnight at 4 ° C with thrombin (1:100 w/w). Simultaneously, purified GluK1-1a_EM was also reconstituted in MSP1E3D1 nanodiscs with soybean polar lipids (Avanti Polar Lipids, 541602P) in a 1:2:140 ratio (GluK1-1a: MSP: lipids) following already established protocols (Chen et al., 2017; Gao et al., 2016; Ritchie et al., 2009). The thrombin-digested GluK1-1a_EM receptors in DDM, or nanodisc-reconstituted receptors, were further purified via gel filtration (Superose 6 10/300, GE) in buffer containing 20 mM Tris pH 8, 150 mM NaCl, 0.5% glycerol, and 0.75 mM DDM (no detergent for nanodisc protein). The peak fractions were pooled and concentrated to ∼0.6 mg/mL (final DDM concentration, ∼7.5 mM) or ∼0.9 mg/mL GluK1-1a_EM in nanodisc. All the affinity elution fractions and final purified protein were confirmed for purity and homogeneity using SDS-PAGE and FSEC, respectively. In the subsequent sections, purified GluK1-1a_EM in detergent and nanodisc will be called GluK1-1a_EM DDM and GluK1-1a_EM ND, respectively.

The purified protein was incubated with 2 mM 2S, 4R-4-methyl glutamate (SYM2081) before grid preparation to capture the receptor in the desensitized state. The concentration of SYM2081 was confirmed by electrophysiology, and the stability of the receptor-SYM2081 complex was tested using FSEC.

GluK1-1a_EM DDM: Double application of 3 µL protein (∼0.6 mg/mL) SYM2081 complex was carried out on glow-discharged gold grids (R 0.6/1, 300 mesh, Quantifoil). The grids were blotted for 4.5 s and 3.5 s, respectively (0 blot force). Vitrobot temperature was maintained at 12°C with 100% humidity, and the sample was vitrified in liquid ethane. The clipped grids were loaded into a 300 keV Titan Krios microscope equipped with a Falcon III direct detector camera (4k x 4k). Movies were recorded in counting mode with a nominal magnification of 59,000X (pixel size: 1.38 Å), and the defocus range of −2.0 to −3.2 µm increased in steps of 0.3. Each movie comprised 25 frames with a total exposure of 60 s. A dose rate of 0.78 e^-/frame was applied, with the total dose being 19.5 e^-/Ų.

GluK1-1a_EM ND: Double application of 2 µL protein (∼0.9 mg/mL) SYM2081 complex was carried out on glow-discharged gold grids (1.2/1.3, 200 mesh, Quantifoil). Grids were blotted for 3 s and 10 s, respectively, followed by vitrification. The clipped grids were loaded into a 300 keV Titan Krios microscope equipped with a K2 direct-detector camera (Gatan). The movies were recorded in super-resolution mode with an energy filter (20eV slit) and a pixel size of 1.41 Å. Each movie was composed of 30 frames with a total exposure of 12 s. A dose rate of 1.36 e^-/frame was applied, with a total dose of 40.8 e^-/Ų.

All movies (GluK1-1a_EM DDM-SYM: 1100, GluK1-1a_EM ND-SYM: 1535) were motion corrected using UCSF Motioncor2 (Zheng et al., 2017). Bad micrographs were removed manually post contrast transfer function (CTF) estimation using CTFFIND4 (Rohou and Grigorieff, 2015). Manually picked particles (∼1000) were 2D classified and used as templates for autopicking in RELION or cryoSPARCv 3. For GluK1-1a_EM DDM-SYM and GluK1-1a_EM ND-SYM, the autopicked particles in cryoSPARCv3 were cleaned up with multiple rounds of 2D classification. Particles in the best classes were used for training TOPAZ for automated particle picking. For GluK1-1a_EM DDM-SYM, initially, 13,750 particles were picked; post iterative rounds of 2D and 3D classification, 5372 particles were used for final 3D reconstruction. In the case of GluK1-1a_EM ND-SYM, initially 1,97,908 particles were picked. Multiple rounds of clean-up using 2D and 3D classification gave 24,531 particles that were used for the final 3D reconstruction.

The final particles were corrected for local motion in cryoSPARCv3, followed by refinement using C1 symmetry for both GluK1-1a_EM DDM-SYM and GluK1-1a_EM ND-SYM. Since we had less information on TMD in both the detergent and nanodisc forms, as observed in 2D classes, we performed local refinement for ATD and LBD using an ECD mask for GluK1-1a_EM DDM-SYM and GluK1-1a_EM ND-SYM, which improved map resolution to 8.01 Å and 5.2 Å (0.143 FSC) respectively. For GluK1-1a_EM DDM-SYM, local refinement using a mask for full-length structure yielded a resolution of 8.2 Å. The 3D maps were sharpened via cryoSPARCv3 (Punjani et al., 2017) or DeepEMhancer (Sanchez-Garcia et al., 2021) using a mask from the cryoSPARCv3 refinement output. Local Resolution was estimated using BlocRes (Cardone et al., 2013) in cryoSPARCv3 or Phenix1.19.2 (Afonine et al., 2018a).

Tetrameric assembly for the GluK1-1a_EM ND-SYM model was built using crystal structures of individual domains (ATD, not yet published, and LBD, PDB: 3C32, GluK1 LBD crystal structure with kainate) fitted into the EM map (5.2 Å) in UCSF Chimera (Pettersen et al., 2004). Further, for building GluK1-1a_EM ND transmembrane domain 3 and GluK1-1a_EM DDM trans-membrane domain, GluK2 TMD (PDB:5KUF) was used due to high identity (∼95%) for fitting into EM density. Phenix1.19.2 (Afonine et al., 2018b) and Namdinator (Kidmose et al., 2019) were used to improve the model via the rigid body and Molecular Dynamics based flexible fitting, respectively. The final models fit well into the EM map of GluK1-1a_EM ND-SYM, GluK1-1a_EM DDM-SYM ECD, and FL, respectively, with a definite density for ATD, LBD, and TMD (pre-M1, M1, M3, and M4). Coot 0.9.4 (Emsley and Cowtan, 2004) was used to analyze the final models, and chimera/chimeraX (Pettersen et al., 2021, 2004) was used for figure preparation.