NMDA receptors are selectively partitioned into complexes and supercomplexes during synapse maturation

Abstract

How neuronal proteomes self-organize is poorly understood because of their inherent molecular and cellular complexity. Here, focusing on mammalian synapses we use blue-native PAGE and 'gene-tagging' of GluN1 to report the first biochemical purification of endogenous NMDA receptors (NMDARs) directly from adult mouse brain. We show that NMDARs partition between two discrete populations of receptor complexes and ∼1.5 MDa supercomplexes. We tested the assembly mechanism with six mouse mutants, which indicates a tripartite requirement of GluN2B, PSD93 and PSD95 gate the incorporation of receptors into ∼1.5 MDa supercomplexes, independent of either canonical PDZ-ligands or GluN2A. Supporting the essential role of GluN2B, quantitative gene-tagging revealed a fourfold molar excess of GluN2B over GluN2A in adult forebrain. NMDAR supercomplexes are assembled late in postnatal development and triggered by synapse maturation involving epigenetic and activity-dependent mechanisms. Finally, screening the quaternary organization of 60 native proteins identified numerous discrete supercomplexes that populate the mammalian synapse.

Figure 1. Native supermolecular assembly of NMDAR subunits in the mammalian brain.

(a) Schematic showing the higher-order assembly from individual proteins to complexes and supercomplexes (complexes of complexes). (b) Native protein complexes of GluN1, GluN2A and GluN2B detected in BNP immunoblot screen of mouse forebrain extracts. Approximately 0.8 and ∼1.5 MDa complexes indicated by filled and open arrowheads, respectively (hereafter used to label all figures). The expected size of each protein in monomeric form indicated with pink rectangle. On left side, non-denaturing molecular mass indicated in mega-Daltons (MDa). (c) BNP GluN1 immunoblot of fractions from glycerol gradient (10–30%) ultracentrifugation. ‘IN', forebrain extract supernatant. On right side, non-denaturing molecular mass indicated in MDa. (d) BNP GluN1 immunoblot of fresh human cortical biopsy samples from the inferior frontal (inf. front.), inferior temporal (inf. temp.) and superior frontal (sup. front.) lobes. Mouse forebrain extract supernatant shown for comparison. These data show the ∼0.8 and ∼1.5 MDa NMDA receptor complexes (1.5-NR and 0.8-NR) were conserved between mouse and humans.

Figure 2. Generation of TAP-tagged GluN1 (Glun1^TAP/TAP) knock-in mice for purification of native NMDARs.

(a) Genetic engineering of TAP-tags into GluN1. Left, schematic shows tetrameric NMDAR with two GluN1 subunits (grey) engineered with tandem affinity peptide (TAP-tags, magenta) on their extracellular N-termini. GluN2 subunits (GluN2A, cyan; GluN2B, orange) shown with C-terminal cytoplasmic ESDV motifs/PDZ-ligands. TAP-tag encodes 3xFlag and His_x6. Right, schematic shows gene-targeting vector carrying TAP-tag sequence in Glun1 exon1, 5′ and 3′ regions of homology with genome and resultant targeted allele below. The neomycin selection cassette was subsequently deleted using Cre/loxP. Grey-filled boxes, exons; magenta, TAP cassette; brown box, neo neomycin resistance cassette; brown triangle, loxP site; dotted lines, homology arms. (b) Purification of native NMDARs from Glun1^TAP/TAP mouse forebrains. Left, schematic of purification steps and samples corresponding to right, immunoblots of NMDAR subunits, PSD95, actin. Dissected mouse forebrains were homogenized and fractionated. Crude membrane fraction was solubilized (membrane extract) and separated by centrifugation (supernatant and pellet). Supernatant was incubated with Flag-affinity resin capturing TAP-tagged receptors with some residual Flow-through. High yields of native receptor were released (eluate) by peptide-antigen exchange. Following elution no material remained (exchanged resin). Right, purification from Glun1^TAP/TAP (TAP) and WT control mouse shows receptor subunits (GluN1, GluN2A, GluN2B), PSD95, and actin detected by SDS–PAGE immunoblots from indicated fractions. Note the higher molecular weight of TAP-engineered GluN1 compared with WT. IB, immunoblotting antibody; MW, molecular weight markers. SDS, SDS–PAGE. (c) Coomassie-stained BNP of TAP-purified NMDAR and PSD95 complexes from Glun1^TAP/TAP and Psd95^TAP/TAP mice, respectively. TAP-purified complexes isolated from forebrain extracts from control (WT), Glun1^TAP/TAP and Psd95^TAP/TAP mice were separated on BNP gel and Coomassie stained. Supplementary Fig. 2a shows excised bands used in native proteomic analysis (TAP-BNP-MS). Filled arrow indicates ∼1.5 MDa complexes and open arrow indicates 0.8-NR. Molecular weight in MDa shown on left and in stained ladder in first and last gel lanes. (d) Coomassie-stained SDS–PAGE of TAP-purified NMDARs from Glun1^TAP/TAP and control (WT) mice. Abundant constituents of bands 1–4 were identified by MALDI-MS: GluN2A/B, TAP-GluN1, PSD93 and PSD95, respectively. MW shown in kDa on left.

Figure 3. Mutant mouse screen of NMDAR supermolecular assembly.

(a) Left, BNP GluN1 immunoblot of forebrain extracts WT (lanes 1 and 2, duplicates), Psd95^−/− (lanes 3 and 4), Psd93^−/− (lanes 5 and 6). Right panel, quantification shows relative to WT (n=6) 1.5-NR decreased to 17% (P<0.001) and 15% (P<0.001) in Psd95^−/− (n=6) and Psd93^−/− (n=6) mice, respectively. (b) Assembly of 1.5-NR does not require conserved PDZ-ligands. Left, schematic of NMDAR in NR-ΔPDZlig mice (Glun2b^dV/dV/Glun2a^dC/dC/Glun1^TAP/TAP). The receptor lacks the GluN2A CTD and terminal valine of GluN2B ESDV motif and contains the TAP-tagged GluN1. GluN1 subunit, grey; TAP-tags, magenta; GluN2A subunit, cyan; GluN2B subunit, orange. Right, BNP immunoblot of NR-ΔPDZlig forebrain extracts (lanes 3 and 4) and control (Glun1^TAP/TAP; lanes 1 and 2) mice with Flag antibody detecting TAP-GluN1. Sample load was normalized by total NMDAR concentration (Supplementary Fig. 4b). (c) PSD95 and PSD93 assemble with NMDARs in 1.5-NR from NR-ΔPDZlig mice. 1.5-NR was TAP-purified NR-ΔPDZlig mice (shown as duplicate lanes labelled 1 and 2) and BNPs immunoblotted with antibodies to PSD95 (left panel) and PSD93 (right panel). (d) Assembly of 1.5-NR requires GluN2B CTD and does not require GluN2A CTD. Left, BNP GluN1 immunoblot of forebrain extracts from WT (lanes 1 and 2, duplicates), Glun2b^{2A(CTR)/2A(CTR)} (lanes 3 and 4), Glun2a^{2B(CTR)/2B(CTR)} (lanes 5 and 6) mice. Cyan and orange labels indicate chimeric Glun2a and Glun2b knock-in mutations, respectively. Right, quantification shows in Glun2b^{2A(CTR)/2A(CTR)} mice 1.5-NR decreased to 6% (t-test, P<0.001) of WT (n=6). (e) Immuno-depletion of GluN2B or PSD95 removes all 1.5-NR. Extracts from Glun1^TAP/TAP/Psd95^EGFP/EGFP double knock-in mice were subunit-depleted with antibodies (shown in lanes) then separated on BNP for immunoblotting with Flag antibody to detect NMDARs. Lanes; input, total extract; immuno-depleting antibodies (from left to right): non-specific IgG, GluN2A, GluN2B and GFP. See Supplementary Fig. 4c for controls. (f) Schematic summary of tripartite genetic requirements of Glun2b, Psd95 and Psd92 for the assembly of ∼1.5 MDa NMDAR supercomplexes. GluN2A and GluN2B subunits assemble into three 0.8-NR subtypes (complexes of GluN2A di-heteromers, GluN2A/GluN2B tri-heteromers, GluN2B di-tetramers). GluN1 subunit, grey; GluN2A subunit, cyan; GluN2B subunit, orange. BNP, blue-native PAGE. Molecular weight in MDa shown on right. Error bars indicate s.e.m. Representative results from triplicate experiments shown. IB, immunoblot.

Figure 4. The GluN2A/GluN2B relative molar abundance using mouse genetics and quantitative tagging.

(a) SDS–PAGE immunoblot of WT, Glun2b^{2A(CTR)/2A(CTR)} and Glun2a^{2B(CTR)/2B(CTR)} forebrain extract. Top panel, detected the GluN2A C-terminal domain (CTD). Second panel down, GluN2B CTD. Glun2a^{2B(CTR)/2B(CTR)} and Glun2b^{2A(CTR)/2A(CTR)} provided negative controls for the specificities of the GluN2A CTD and GluN2B CTD antibodies, respectively. Third, fourth and fifth panels down, detected GluN2A NTD, GluN2B NTD subunits and GluN1, respectively. Blue and orange labels indicate chimeric Glun2a and Glun2b knock-in mutations, respectively. (b) Dilution series of Glun2a^{2B(CTR)/2B(CTR)} forebrain extract into that of Glun2b^{2A(CTR)/2A(CTR)} indicated sensitivity of quantification. Immunoblots detected GluN2A CTD (upper panel) and GluN2B CTD (lower panel). (c) Quantification of GluN2A CTD and GluN2B CTD immunoblots. Measurements from Glun2b^{2A(CTR)/2A(CTR)} (P<0.001) and Glun2a^{2B(CTR)/2B(CTR)} (t-test, P<0.02) were normalized to WT. (d) SDS–PAGE immunoblot of WT, Glun2b^{2A(CTR)/2A(CTR)} extracts from left to right: cortex, hippocampus and cerebellum. Top panels, detected the GluN2A C-terminal domain (CTD). Lower panels, detected GluN2A NTD. (e) Quantification of GluN2A relative to GluN2B in eight different brain regions from top to bottom: olfactory bulb, caudate putamen, hippocampus, cortex, colliculus, cerebellum, thalamus and hindbrain. GluN2A was measured using GluN2A CTD antibodies against WT extracts (n=3). GluN2B was measured by measuring GluN2A CTD antibodies against Glun2b^{2A(CTR)/2A(CTR)} extracts (n=3) and subtracting the intensity from that of WT. Error bars indicate 1 s.d.

Figure 5. 1.5-NR and 1.5-PSD95 assemble late in postnatal development.

(a) BNP immunoblots show WT mouse forebrain extracts at developmental time points probed with antibodies against GluN1, GluN2A, GluN2B, PSD95 and PSD93. Filled arrow, ∼1.5 MDa complexes that only assembly from ∼P16 onwards. Sample load was normalized by forebrain mass. Molecular weight in MDa shown on right. (b) Quantification of changing abundance of 0.8-NR and 1.5-NR from P12 (n=3) to P16 (n=3) using BNP immunoblots of GluN1. 0.8-NR increased by 25%, whereas 1.5-NR increased by 445%; t-test, P>0.01; Error bars indicate s.e.m.

Figure 6. 1.5-NR and 1.5-PSD95 assembly is correlated with synapse maturation.

(a) HDAC1/2 inhibition with TSA causes increased assembly if 1.5-PSD95. Cultured primary cortical neurons (DIV7) were treated with DMSO (control; lanes 1 and 2) 0.25 μM TSA (lanes 3 and 4) or for 18–24 h before cells were harvested and analysed by BNP PSD95 immunoblot. (b) Quantification shows TSA treatment increased 1.5-PSD95 to 166% (n=6) of control. t-test, P<0.001; Error bars indicate s.e.m. (c) HDAC1/2 inhibition causes decreased and increased assembly of 0.8-NR and 1.5-NR, respectively. Cultured primary cortical neurons (DIV7) were treated with DMSO (control; lanes 1 and 2) 0.25 μM TSA (lanes 3 and 4) or for 18–24 h before cells were extracted and analysed by BNP GluN2B immunoblot. (d) Left, quantification shows TSA treatment decreased 0.8-NR to 28% (n=6) of control. P<0.001; Error bars indicate s.e.m. Right, TSA treatment increased 1.5-NR to 251% (n=6) of controls. P<0.05; error bars indicate s.e.m. (e) Inhibition of palmitoylation with 2-bromopalmidate (2-BP) decreased 1.5-PSD95. Cultured primary cortical neurons (DIV14) were treated with DMSO (control; lanes 1 and 2) 10 μM 2-BP (lanes 3 and 4) for 8 h before cells were extracted and analysed by BNP PSD95 immunoblot. (f) Quantification shows 2-BP treatment decreased 1.5-PSD95 to 10% (n=6) of control. P<0.001; error bars indicated s.e.m. (g) Inhibition of palmitoylation with 2-BP does not change the expression levels of PSD95 or GluN2B. (h) 2BP-mediated disassembly of 1.5-PSD95 is activity dependent. Cultured primary cortical neurons (DIV14) were treated with DMSO (control; lanes 1, 2 and 3) 10 μM 2-BP (lanes 4, 5 and 6), 10 μM 2-BP with 1 μM TTX (lanes 7, 8 and 9), 10 μM 2-BP with 50 μM APV (lanes 10, 11 and 12) for 8 h before cells were extracted and analysed by BNP PSD95 immunoblot. Filled arrow indicates 1.5-PSD95. Molecular weight in MDa shown on right. (i) Quantification shows 2-BP treatment decreased 1.5-PSD95 to 14% of control (n=6, P<0.001). TTX rescued 2-BP-dependent disassembly (n=6, P<0.001) to 74% of control. AP5 rescued 2-BP-dependent disassembly (n=6, P<0.001) to 69% of control; error bars indicated s.e.m.

Figure 7. Screen of native protein assemblies of 60 brain proteins detected by BNP (blue non-denaturing PAGE) immunoblot of mouse forebrain extracts.

Complexes were extracted with a panel of five different buffers. For most candidate proteins, different buffers extracted complexes of a similar size (see Methods section for details and Supplementary Data 1 for list of all complexes). Novel and previously described complexes indicated by filled and open arrowheads, respectively. The expected size of each protein in monomeric form indicated with pink rectangle. Non-denaturing molecular mass indicated in mega-Daltons (MDa).