Positive surface charge of GluN1 N-terminus mediates the direct interaction with EphB2 and NMDAR mobility

Abstract

Localization of the N-methyl-D-aspartate type glutamate receptor (NMDAR) to dendritic spines is essential for excitatory synaptic transmission and plasticity. Rather than remaining trapped at synaptic sites, NMDA receptors undergo constant cycling into and out of the postsynaptic density. Receptor movement is constrained by protein-protein interactions with both the intracellular and extracellular domains of the NMDAR. The role of extracellular interactions on the mobility of the NMDAR is poorly understood. Here we demonstrate that the positive surface charge of the hinge region of the N-terminal domain in the GluN1 subunit of the NMDAR is required to maintain NMDARs at dendritic spine synapses and mediates the direct extracellular interaction with a negatively charged phospho-tyrosine on the receptor tyrosine kinase EphB2. Loss of the EphB-NMDAR interaction by either mutating GluN1 or knocking down endogenous EphB2 increases NMDAR mobility. These findings begin to define a mechanism for extracellular interactions mediated by charged domains.

Fig. 1. GluN1 interacts with EphB2 at synapses.

a Representative images of synaptosomes immunostained for GluN1 in green, EphB2 in red, and vGlut1 in cyan. Far right panel shows merged image. Scale bar = 1.5 µm. b Quantification of the percentage of vGlut1⁺ synaptosomes that colocalize with either EphB2 only, GluN1 only, or both EphB2 and GluN1 (green dots represent n = 612 synaptosomes, five fields). c Schematic of proximity ligation assay shows EphB2 in red and NMDAR in green. Rolling circle amplification of fluorescently tagged oligonucleotides (represented in magenta) occurs after primary and secondary probes bind. Representative images of synaptosomes immunostained for vGlut1 in cyan (middle panel). PLA between EphB2 and GluN1 is shown in magenta (left panel). Right panel shows merged image. Scale bar = 1.5 µm. d Quantification of the percentage of vGlut1⁺ synaptosomes that colocalize with PLA puncta (PLA + ICC). The control condition (PLA Ctrl) was performed without EphB2 primary antibody (****p < 0.0001, unpaired t-test; green dots represent PLA Ctrl n = 569, 6 fields; PLA + ICC n = 421 synaptosomes, 11 fields).

Fig. 2. GluN1 hinge mutants disrupt the EphB2–GluN1 interaction in HEK293T cells.

a A model of domain organization of the GluN1 and the GluN2B subunits of the NMDA receptor. The red box indicates the N-terminal domain (NTD) of GluN1. b The crystal structure of the NTD of GluN1 (PDB: 4PE5). The red box indicates the hinge region within the NTD with specific amino acids highlighted in red. c The structure of the GluN1 NTD hinge region represented in cartoon form with the location of six amino acids highlighted in red (I272, N273, T335, G336, R337, N350). d The surface representation of the structure of the GluN1 NTD hinge region with the location of the six amino acids in c highlighted in red. e Charge map of the GluN1 NTD hinge region. The charge map was generated using the adaptive Poisson–Boltzmann solver (APBS) plugin in PyMOL. f Charge maps of GluN1 NTD hinge region mutants (WT, Quintuple, Sextuple). Yellow outline indicates the location of the six key hinge region amino acids in WT GluN1. g Representative images of the PLA assay results in HEK293T cells. HEK293T cells were transfected with either WT Myc-GluN1, Quintuple mutant Myc-GluN1 (I272A/N273A/T335A/G336A/R337A), or Sextuple mutant Myc-GluN1 (I272A/N273A/T335A/G336A/R337A/350Q), and GluN2B, FLAG-tagged-EphB2, and EGFP. The upper panels show PLA signal alone. The lower panels are the merged images with EGFP in green and PLA signal in magenta. Scale bar = 10 µm. h Quantification of the effects of GluN1 mutants on PLA puncta number. PLA puncta number are quantified by counting the number of puncta per 100 µm² in EGFP⁺ cells and normalizing to the WT condition (***p < 0.005, ANOVA; green dots represent n = 30 cells for each condition).

Fig. 3. GluN1 hinge mutants disrupt the EphB2–GluN1 interaction in neurons.

a Representative images of dendrites of DIV6-9 cortical neurons transfected with EGFP–GluN1 (WT, Quintuple, or Sextuple), mCherry, GluN2B, and CRISPR constructs to knock out endogenous GluN1. The control condition is WT GluN1-transfected treated with control reagents instead of ephrin-B2. The other three conditions were treated with ephrin-B2 for 45 min to induce the interaction between the NMDAR and EphBs. Top row panels show the colocalization between EGFP–GluN1 and EphB2 in white. Second row panels show the high contrasted images of EGFP–GluN1 puncta in green. Third row panels show the high contrasted images of endogenous EphB2 in magenta. Last row panels show mCherry. Outlines of morphology are shown in cyan. Scale bar = 5 µm. b Quantification of EphB2 density (puncta number per 100 µm) in indicated groups (***p < 0.005, ****p < 0.0001, ANOVA followed by Tukey’s; green dots represent Control n = 14; WT n = 21 cells; Quintuple n = 28; Sextuple n = 18). c Quantification of EGFP–GluN1 density (puncta number per 100 µm) in indicated groups (p = 0.5958, ANOVA; green dots represent Control n = 14; WT n = 21 cells; Quintuple n = 28; Sextuple n = 18). d Quantification of the effects of the different GluN1 mutants on colocalization (puncta number per 100 µm) between GluN1 and EphB2 (***p < 0.005, ****p < 0.0001, ANOVA; green dots represent Control n = 14; WT n = 21 cells; Quintuple n = 28; Sextuple n = 18).

Fig. 4. Surface charge of the GluN1 NTD hinge region affects NMDAR mobility in mature neurons.

a Representative FRAP images at different time points of DIV21–23 cortical neurons transfected with EGFP–GluN1 (WT, Quintuple, or Sextuple). Recovery of bleached spine puncta (magenta circle) was monitored for 30 min at 30-s intervals. Scale bar = 2 µm. b Quantification of the recovery curve of different GluN1 mutants in DIV21–23 cortical neurons. Graphs represent mean intensity and show fit (****p < 0.0001, Kolmogorov–Smirnov (KS) nonparametric test). Inset: Quantification of the mobile fraction of EGFP–GluN1 puncta in dendritic spines at 30 min after  photobleaching in EGFP–GluN1 mutant transfected cells compared to WT EGFP–GluN1 in DIV21–23 (*p < 0.05, ANOVA; green dots represent WT n = 14 puncta; Quintuple n = 11; Sextuple n = 10). Error bars show S.E.M.

Fig. 5. Specific hinge region amino acid residues are responsible for the EphB2–GluN1 interaction in HEK293T cells.

a Surface charge maps of GluN1 NTD hinge region of the indicated GluN1 single point mutants in order of increasing negative charged in the hinge region (left (positive) to right (negative)), with blue representing positive charge and red representing negative charge. Yellow outline indicates the location of the six key hinge region amino acids in WT GluN1. b Representative images of PLA results in HEK293T cells. HEK293T cells were transfected with the indicated Myc-GluN1 single point mutants, together with GluN2B, FLAG-EphB2, and EGFP. The upper panels show PLA signal alone. The lower panels are merged images of EGFP in green and PLA signal in magenta. Scale bar = 10 µm. c Quantification of the effects of GluN1 mutants on PLA puncta number. PLA puncta number are quantified by counting the number of puncta per 100 µm² in EGFP⁺ cells and normalizing to the WT condition (****p < 0.0001, ANOVA; green dots represent n = 30 cells for each condition). d Surface charge maps of GluN1 NTD hinge region of the indicated GluN1 double point mutants as in a (left (positive) to right (negative)). Yellow outline indicates the location of the six key hinge region amino acids in WT GluN1. e Representative images of PLA results in HEK293T cells. HEK293T cells were transfected with the indicated Myc-GluN1 double point mutants, together with GluN2B, FLAG-EphB2, and EGFP. The upper panels show PLA signal alone. The lower panels are merged images of EGFP in green and PLA signal in magenta. Scale bar = 10 µm. f Quantification of the effects of GluN1 mutants on PLA puncta number. PLA puncta number are quantified by counting the number of puncta per 100 µm² in EGFP⁺ cells and normalizing to the WT condition (****p < 0.0001, ANOVA; green dots represent n = 30 cells for each condition).

Fig. 6. EphB2–GluN1 interaction in HEK293T cells is charge-dependent.

a Representative images of PLA results in HEK293T cells. HEK293T cells were transfected with the indicated FLAG-EphB2 single point mutants (WT, Y504E, or Y504F), together with Myc-GluN1 WT, GluN2B, and EGFP. The upper panels show PLA signal alone. The lower panels are merged images of EGFP in green and PLA signal in magenta. Scale bar = 10 µm. b Quantification of the effects of EphB2 mutants on PLA puncta number. PLA puncta number are quantified by counting the number of puncta per 100 µm² in EGFP⁺ cells and normalizing to the WT condition. (****p < 0.0001, ANOVA; n = 30 cells for each condition). c Representative images of PLA results in HEK293T cells. HEK293T cells were transfected with the indicated Myc-GluN1 single point mutants, together with GluN2B, FLAG-EphB2 Y504E, and EGFP. The upper panels show PLA signal alone. The lower panels are merged images of EGFP in green and PLA signal in magenta. Scale bar = 10 µm. d Quantification of the effects of GluN1 mutants on PLA puncta number. PLA puncta number are quantified by counting the number of puncta per 100 µm² in EGFP⁺ cells and normalizing to the WT condition. (*p < 0.05, ***p < 0.005, ANOVA; n = 30 cells for each condition). e Representative images of PLA results in HEK293T cells. HEK293T cells were transfected with the indicated Myc-GluN1 double point mutants, together with GluN2B, FLAG-EphB2 Y504E, and EGFP. The upper panels show PLA signal alone. The lower panels are merged images of EGFP in green and PLA signal in magenta. Scale bar = 10 µm. f Quantification of the effects of GluN1 mutants on PLA puncta number (*p = 0.0219, ****p < 0.0001, ANOVA; green dots represent n = 30 cells for each condition).

Fig. 7. GluN1 charge mutants show disrupted EphB–NMDAR interaction in neurons.

a Representative images of dendrites of DIV6-9 cortical neurons transfected with EGFP–GluN1 (WT, I272A, or N273A/R337D), mCherry, GluN2B, and CRISPR constructs to knock out endogenous GluN1. The control condition is WT GluN1-transfected treated with control reagents instead of ephrin-B2. The other three conditions were treated with ephrin-B2 for 45 min to induce the interaction between the NMDAR and EphBs. Top row panels show the colocalization between EGFP–GluN1 and EphB2 in white (arrows indicate examples of colocalized puncta). Second row panels show the high contrasted images of EGFP–GluN1 in green. Third row panels show endogenous EphB2 in magenta. Last row panels show mCherry. Outlines of morphology in cyan. Scale bar = 5 µm. b Quantification of EphB2 density (puncta number per 100 µm) in indicated groups (*p < 0.05, ANOVA; green dots represent n = 30 cells for each condition). c Quantification of EGFP–GluN1 density (puncta number per 100 µm) in indicated groups (p = 0.0128, ANOVA followed by Tukey’s; comparisons showed no significant differences; green dots represent n = 30 cells for each condition). d Quantification of the effects of the different GluN1 mutants on colocalization (puncta number per 100 µm) between GluN1 and EphB2 (*p < 0.05, ANOVA; green dots represent n = 30 cells for each condition).

Fig. 8. GluN1 charge mutants show increased mobility due to disrupted EphB–NMDAR interaction.

a Representative FRAP images at different time points of DIV21–23 cortical neurons transfected with EGFP–GluN1 (WT, I272A, or N273A/R337D) together with GluN2B, CRISPR construct targeting endogenous GluN1, and mCherry. Recovery of bleached spine puncta (magenta circle) was monitored for 15 min at 10s intervals. Scale bar = 2 µm. b Quantification of the recovery curve of different GluN1 mutants in DIV21–23 cortical neurons. Graphs represent mean intensity and show fit (****p < 0.0001, Kolmogorov–Smirnov (KS) nonparametric test). Inset: Quantification of the mobile fraction of EGFP–GluN1 spine puncta at 15 min after photobleaching in EGFP–GluN1 mutant transfected cells compared to WT EGFP–GluN1 in DIV21–23 cortical neurons (*p < 0.05, ANOVA; green dots represent WT n = 33 puncta; I272A n = 12; N273A/R337D n = 12). Error bars show S.E.M. c Representative FRAP images at different time points of DIV21–23 cortical neurons transfected with EGFP–GluN1 (WT, WT+EphB2 knockdown, N273A/R337D + EphB2 knockdown, WT+EphB2 knockdown rescued by RNAi insensitive EphB2) together with GluN2B, CRISPR construct targeting endogenous GluN1, and mCherry. FRAP was conducted on serveral puncta in different dendritic branches. Recovery of bleached spine puncta (magenta circle) was monitored for 15 min at 10s intervals. Scale bar = 2 µm. d Quantification of the recovery curve of different GluN1 mutants in DIV21–23 cortical neurons. Graphs represent mean intensity and show fit (****p < 0.0001, Kolmogorov–Smirnov (KS) nonparametric test). Inset: Quantification of the mobile fraction of EGFP–GluN1 spine puncta at 15 min after bleaching in EGFP–GluN1 mutant transfected cells compared to WT EGFP–GluN1 in DIV21–23 cortical neurons (*p < 0.05, ANOVA; green dots represent WT n = 33 puncta; WT+EphB2 K.D. n = 12; N273A/R337D+EphB2 K.D. n = 21; rescue n = 22). Error bars show S.E.M.

Fig. 9. pH-sensitive histidine mutants in the hinge region elucidate a charge-dependent mechanism.

a Representative images of PLA results in HEK293T cells. HEK293T cells were transfected with the indicated Myc-GluN1 single point mutants, together with GluN2B, FLAG-EphB2, and EGFP. Cells were treated with media at either pH 5.0 or pH 7.3 for 30 min before PLA. Upper panels: PLA signal alone. Lower panels: merge of EGFP in green and PLA signal in magenta. Scale bar = 10 µm. b Quantification of the effects of GluN1 mutants and pH on PLA puncta number (****p < 0.0005, ANOVA; green dots represent n = 30 cells for each condition). c Surface representation models (top) and charge maps (bottom) of WT GluN1 and R337H GluN1 predicted at neutral pH. d Model of experimental design. The same neurons were imaged for both pH conditions. e Representative FRAP images at different time points of DIV21–23 cortical neurons transfected with EGFP–GluN1 (WT or R337H) together with GluN2B, CRISPR construct targeting endogenous GluN1, and mCherry. FRAP of GluN1 puncta was conducted in the same neurons at pH 5.0 (H-positive) and pH 7.3 (H-neutral) and the order of pH presentation varied. Recovery of bleached spine puncta (magenta circle) was monitored for 15 min at 10s intervals. Scale bar = 2 µm. f Left: Quantification of the recovery curve of GluN1 WT or R337H mutant in ACSF at pH 7.3 and pH 5.0 in DIV21–23 cortical neurons. Graphs represent mean intensity and show fit (****p < 0.0001, Kolmogorov–Smirnov (KS) nonparametric test). Right: Quantification of the mobile fraction of EGFP–GluN1 spine puncta at 15 min after photobleaching (*p = 0.0398, WT-pH7.3 vs. R337H-pH7.3; ANOVA; green dots represent WT-pH5.0 n = 26 puncta; R337H-pH5.0 n = 24; WT-pH7.3 n = 23; R337H-pH7.3 n = 35). Error bars show S.E.M. g Left: Quantification of the recovery curve of GluN1 WT or I272H mutant in ACSF at pH7.3 and pH5.0 in DIV21–23 cortical neurons. Graphs represent mean intensity and show fit. (p = 0.8148, WT-pH7.3 vs. I272H-pH7.3, Kolmogorov–Smirnov (KS) nonparametric test). Right: Quantification of the mobile fraction of EGFP–GluN1 spine puncta at 15 min of FRAP (p = 0.9839; WT-pH7.3 vs. I272H-pH7.3; ANOVA; green dots represent WT-pH5.0 n = 26 puncta; I272H-pH5.0 n = 22; WT-pH7.3 n = 23; I272H-pH7.3 n = 30). Error bars show S.E.M.