Trans-synaptic molecular context of NMDA receptor nanodomains

Abstract

Tight coordination of spatial relationships between protein complexes is required for cellular function. In neuronal synapses, proteins responsible for neurotransmission form subsynaptic nanoclusters whose trans-cellular alignment modulates synaptic signal propagation. However, the spatial relationships between these proteins and NMDA receptors (NMDARs), which are required for learning and memory, remain undefined. Here, we mapped key NMDAR subunits relative to active zone and post-synaptic density reference proteins using multiplexed super-resolution DNA-PAINT microscopy in rat hippocampal neurons. GluN2A and GluN2B subunits formed diverse nanoclusters that, surprisingly, were not localized near presynaptic vesicle release sites marked by Munc13-1. However, a subset of release sites was enriched with NMDARs, and modeling indicated this nanotopography promotes NMDAR activation. These enriched sites were internally denser with Munc13-1, aligned with PSD-95, and closely associated with specific NMDAR nanodomains. NMDAR activation rapidly reorganized this relationship, suggesting a structural mechanism for tuning NMDAR-mediated synaptic transmission. These findings suggest synaptic functional architecture depends on assembly of and trans-cellular spatial relationships between multiprotein nanodomains.

Fig. 1. Mapping endogenous NMDA receptor organization with DNA-PAINT.

a Schematic of a primary antibody labeled with a DNA-PAINT docking strand-conjugated secondary nanobody and imaged with fluorescent imager strands. Red star indicates fluorophore. b Schematic of four-target DNA-PAINT labeling using primary antibodies preincubated with secondary nanobodies. c DNA-PAINT rendering (10 nm pixels) of myristoylated-EGFP cell fill, surface expressed GluN2A, PSD-95, and Bassoon demonstrating four-target, synaptic DNA-PAINT. Scale bar 2 µm. d Boxed region from top, including widefield image of myr-EGFP. Scale bar 500 nm. e Confocal image of EGFP-GluN2B CRISPR knock-in cell. Scale bar 20 µm. f Boxed region from left, showing surface expressed GluN2A, surface expressed GluN2B (EGFP knock-in), and PSD-95 labeling colocalized at synapses. Scale bar 4 µm. g DNA-PAINT rendering (10 nm pixels) of endogenous, surface expressed GluN2A and GluN2B (EGFP knock-in), PSD-95, and Munc13-1. Scale bar 1 µm. h, i Zoom-in on two representative synapses showing nanoclusters of each protein and their co-organization. Scale bar 200 nm. Data in c–f are from single exemplar experiments. Data in g-i are from an exemplary region analyzed in later figures; see Statistics and reproducibility.

Fig. 2. Endogenous GluN2 subunits form diverse nanodomain types.

a Example synapse of DNA-PAINT localizations of GluN2A and GluN2B showing GluN2 NCs have diverse co-organization. Each point is a localization, and its heat map codes normalized local density. NCs are indicated by dash-bordered areas. Peaks of the NCs of the opposing protein are indicated by colored x’s. Gray outline indicates PSD border, defined by PSD-95 localizations. b GluN2 subunit autocorrelations decayed faster than PSD-95 and plateaued below one, indicating small NCs with few localizations between them. c GluN2B NCs were more numerous than GluN2A or PSD-95 NCs (Kruskal-Wallis test P = 0.004; post-hoc Dunn’s test for multiple comparisons: GluN2A vs GluN2B P = 0.003, GluN2B vs PSD-95 P = 0.0014). d Both GluN2A and GluN2B NCs were smaller than PSD-95 NCs (Kruskal-Wallis test P < 0.0001; post-hoc Dunn’s test for multiple comparisons: GluN2A vs GluN2B P = 0.0004, GluN2A vs PSD-95 and GluN2B vs PSD-95 P < 0.0001). e GluN2 subunit NCs were distributed slightly less centrally than PSD-95 NCs. f Extrasynaptic GluN2 tended to be within ~200 nm of the PSD edge. g, h Extrasynaptic GluN2 clusters were on average smaller than synaptic GluN2 NCs (unpaired t-tests, GluN2A P = 0.0673, GluN2B P = 0.0025). i Cross-correlation and j, k cross-enrichments indicated strong overlap of GluN2A and GluN2B densities at short distances. l, m 52.6% of GluN2A and 42.0% of GluN2B NCs were significantly enriched with the opposite subunit. n 36.7% of GluN2A and 22.6% of GluN2B NC peaks were located within 30 nm of an opposite GluN2 NC peak. o, p 39.0% of GluN2A and 23.4% of GluN2B NC areas spatially overlapped with the opposite subunit. Data in b and i–k are means ± SEM shading. Throughout figures, gray shading in cross-enrichment graphs (such as j, k) indicates enrichments at distances <55 nm used to calculate enriched and de-enriched populations (such as l, m). Points in c are individual synapses and points in d and g, h individual NCs. Lines in c, d, g and h are means ± SEM. Data in (n) are shown as frequency histograms (20 nm bins), with dashed line indicating the division summarized in inset pie charts. n = 173 GluN2A, 274 GluN2B, 134 PSD-95 NCs from 74 synapses throughout synaptic data; n = 49 GluN2A and 119 GluN2B extrasynaptic NCs from 74 synapses. Specific n per data point in (b) and (I) varies from 74 due to data processing, see methods for details. **P < 0.01, ***P < 0.001, ****P < 0.0001 from Dunn’s test in c and d and two-tailed unpaired t-test in (h).

Fig. 3. Only a subpopulation of release sites is enriched with GluN2 subunits.

a Example synapse of DNA-PAINT localizations of PSD-95, GluN2A, GluN2B, and Munc13-1 showing arrangement of receptor subunits relative to release sites. Markers as described in Fig. 2a. b Munc13-1 autocorrelation decayed rapidly and plateaued below one, consistent with its c numerous and small detected Munc13-1 NCs. d, e Munc13-1 NCs were, on average, de-enriched with GluN2A and GluN2B at distances <55 nm (shading). f Schematic indicates possible configurations of GluN2 and Munc13-1 NCs. g Schematic of NC pairing to reveal stereotyped distances of closely-associated NCs. h Paired Munc13-1 NCs were enriched with GluN2A as well as i GluN2B within 55 nm of their peak. j, k 20.5% and 24.1% of Munc13-1 NCs were statistically enriched with GluN2A or GluN2B. l 26.7% and 32.7% of Munc13-1 NCs had a GluN2A or GluN2B NC peak within 60 nm. m, n 10.9% and 10.6% of Munc13-1 NCs spatially overlapped with GluN2A or GluN2B NCs. Data in (b, d, e, and h, i) are means ± SEM shading. Points in c (left) are synapses and (right) NCs, with lines at mean ± SEM. Data in (l) are shown as frequency histograms (20 nm bins), with dashed line indicating the division summarized in inset pie charts. n = 173 GluN2A, 274 GluN2B, 478 Munc13-1 NCs from 74 synapses throughout, except h, i, where n = 113 Munc13-1 NCs when paired with GluN2A and 152 when paired with GluN2B. Specific n per data point in b varies from 74 due to data processing, see methods for details.

Fig. 4. Simulations reveal that NMDAR positioning near a subset of release sites may regulate receptor activation probability.

a Schematic of receptor NC and release site positions in receptor activation simulations; synapse shown is exemplar used in (b–d). Magenta and green circles indicate GluN2A and GluN2B localizations within NCs, respectively, blue x’s indicate release site positions, gray outline indicates PSD border, and dashed line represents minimal bounding ellipse used in modeling (see Methods). Location of release sites and NCs shown in c, d insets are indicated with labeled arrows. b The simulated open fraction of all NMDARs at the synapse varies depending on where glutamate is released. c, d The simulated open fraction of NMDARs varies by GluN2 NC for both GluN2A and GluN2B. Insets for each show two example NCs with differing P_o depending on which release site is active. e Schematic of receptor and release site positions compared in receptor activation modeling in f, g. Gray outline indicates PSD border, green circles indicate GluN2 localization positions, and blue x’s indicate release site positions. f, g Simulated peak synaptic GluN2A and GluN2B open probabilities (P_o) were significantly greater when release sites were in their real positions vs uniformly distributed. h Schematic of receptor and release site positions compared in receptor activation modeling in (I, j). Symbols as in (e). i, j Simulated peak synaptic GluN2A and Glun2B P_o were significantly greater when receptors were in their real positions vs uniformly distributed. n = 8 synapses in (e–j) (data in (a–d) show one exemplar synapse from this dataset). Exact P values: f P = 0.0003; g P = 0.0032; i P = 0.0065; j P = 0.0149; *P < 0.05, **P < 0.01, ***P < 0.001 from two-tailed paired t-test.

Fig. 5. A subset of structurally unique Munc13-1 NCs is enriched with PSD-95 and in the nanocolumn.

a Example synapse of DNA-PAINT localizations of Munc13-1 and PSD-95, showing variable position of Munc13-1 NCs relative to PSD-95 density. Markers as in Fig. 2a. b On average, Munc13-1 and PSD-95 were weakly enriched with one another. c After pairing (as in 3 g), both PSD-95 and Munc13-1 were significantly enriched with one another. d, e 24.4% of Munc13-1 and 40.3% of PSD-95 NCs were statistically enriched with the other protein. f 20.5% of Munc13-1 and 53.3% of PSD-95 NCs had a NC peak of the other protein within 60 nm. g, h14.8% of Munc13-1 and 42.6% of PSD-95 NCs were spatially overlapped. i, j Cross-enrichments of PSD-95 NCs with Munc13-1 or vice versa demonstrate strong cross-enrichment of these proteins when subset by the data in d, e. k, l Munc13-1 and PSD-95 NCs were denser when in the nanocolumn than when outside it (in k: two-tailed Mann-Whitney test, P = 0.0030; in l: two-tailed unpaired t-test, P = 0.0115). Data in (b, c) and (I, j) are means ± SEM shading. Points in (k, l) are NCs with lines at mean ± SEM. n = 174 PSD-95, 478 Munc13-1 NCs from 74 synapses in (b, d–h). n = 98 paired PSD-95 and Munc13-1 NCs in c. n = 117 in nanocol, 84 not in nanocol Munc13-1 NCs; 54 in nanocol, 41 not in nanocol PSD-95 NCs from 74 synapses in (k, l). *P < 0.05 from two-tailed unpaired t-test, ***P < 0.0001 from two-tailed Mann-Whitney test.

Fig. 6. Subunit-specific NMDAR nanodomains are organized with distinct trans-synaptic molecular contexts.

In each section separated by dashed lines, the schematic (a, f, i, l) indicates the conditional comparison being made, and cross-enrichment plots (means ± SEM shading) show the measurements made with respect to GluN2A (b, g, j, m) and GluN2B (c, h, k, n). a–c Munc13-1 NCs in the nanocolumn were significantly more enriched with both GluN2A and GluN2B than Munc13-1 NCs outside the nanocolumn. n = 117 in nanocol, 84 not in nanocol Munc13-1 NCs from 74 synapses. d, e Receptor activation simulations showed glutamate release from nanocolumnar release sites trended toward greater synaptic activation of both GluN2A and GluN2B-containing NMDARs vs release from non-nanocolumnar sites. n = 7 synapses. f–h GluN2A and GluN2B NCs in the nanocolumn were denser than those outside the nanocolumn. n = 27 in nanocol, 21 not in nanocol GluN2A NCs; 42 in nanocol, 26 not in nanocol GluN2B NCs from 74 synapses. i–k GluN2A and GluN2B NCs were more cross-enriched with one another when in the nanocolumn. n as in (g, h). l–n Munc13-1 NC enrichment with GluN2A and GluN2B can predict Munc13-1 NC enrichment with PSD-95. n = 96 Munc13-1 NCs with GluN2A, 119 without GluN2A, 115 with GluN2B, 156 without GluN2B NCs from 74 synapses. o Model: NMDAR distribution in synapses is governed by nanodomains with distinct trans-synaptic molecular contexts. Active zones contain molecularly diverse release sites likely with differing vesicle priming and release properties, and whose functional impact depends in part on differential trans-synaptic alignment. Receptor nanodomains near nanocolumn release sites contain GluN2A and GluN2B subunits; NMDARs accumulated near release sites will be activated more efficiently and potentially transduce unique signaling due to the presence of a mixed population of intracellular C-termini.

Fig. 7. Acute NMDAR activation drives rapid reorganization of release site/receptor relationship.

a–c DNA-PAINT renderings of GluN2A, Munc13-1, and PSD-95 at synapses in neurons treated with vehicle, NMDA, or NMDA + APV. Scale bar 2 µm. d–f Zoom-in of boxed synapses in (a–c). Scale bar 200 nm. g–i Autocorrelations of GluN2A, Munc13-1, and PSD-95 are not significantly altered by NMDA or NMDA + APV treatment, relative to vehicle. j The number of Munc13-1 NCs per synapse is slightly increased by NMDA treatment (Kruskal-Wallis test P = 0.0277; post-hoc Dunn’s test for multiple comparisons: vehicle vs NMDA P = 0.0229, vehicle vs NMDA + APV P = 0.7344, NMDA vs NMDA + APV P = 0.3760). n = 76 vehicle-treated, 80 NMDA-treated, and 82 NMDA + APV-treated synapses in (g–j). k–m Subsetting of the cross-enrichment of Munc13-1 NCs with PSD-95 shows the percent of Munc13-1 NCs enriched with PSD-95 (i.e., in the nanocolumn) is not changed by NMDA or NMDA + APV treatment. n–p The percent of Munc13-1 NCs significantly enriched with GluN2A is not changed by NMDA or NMDA + APV treatment. q NMDA treatment selectively increases the average cross-enrichment of Munc13-1 NCs with GluN2A vs vehicle. This effect is rescued by co-application of APV. n = 508 vehicle-treated, 667 NMDA-treated, and 611 NMDA + APV-treated Munc13-1 NCs in (n–q). r Schematic showing subsetting of Munc13-1 NC cross-enrichments with GluN2A by whether they are in the nancolumn or not. s, t NMDA treatment selectively increases the enrichment of Munc13-1 NCs with GluN2A within, but not outside, the nanocolumn. n = 137 vehicle-treated, 179 NMDA-treated, and 146 NMDA + APV treated nanocolumnar Munc13-1 NCs in (s), and n = 121 vehicle-treated, 156 NMDA-treated, and 152 NMDA + APV-treated non-nanocolumnar Munc13-1 NCs in t. Data in (g–i, q), and s, t are means ± SEM shading, and specific n per data point varies due to data processing (see Methods for details). Points in (j) are individual synapses with line at mean ± SEM. *P < 0.05 from Dunn’s test.