Fig. 2: AMPA and NMDA receptors cluster at the periphery and the center of the reconstructed postsynaptic density, respectively.

a Examples of synapses from spin-mill serial block face scanning electron microscopy; each line indicates the extent of the cleft in a single 20-nm-thick 2D profile, each circle indicates the location of a gold particle. b Cumulative relative frequency distribution of lateral distances from gold particles to the center of the postsynaptic density (PSD) from 3D profiles. Distances are normalized to the size of the PSD and corrected as fractional area assuming a circular PSD: a gold particle at 0 would be exactly at the center, at 1 exactly at the edge, and at 0.25 equidistant between center and edge. AMPA receptors are slightly biased toward the edge (median = 0.6, p < 0.001, n = 840 particles, N = 2 cultures), while NMDA receptors are biased toward the center (median = 0.3, p < 0.001, n = 550 particles, N = 2 cultures). c Same as a, except showing the centers of clusters determined by k-means clustering. d Sum of squared differences, calculated from each particle to the centroid of the cluster. Data: the actual distances of gold particles to their cluster center. Randomized: the distances of gold particles to their putative cluster center after randomizing the locations of gold particles at each synapse. Each dot: a synapse. Simple linear regression test: AMPA receptors, data, R² = 0.32, randomized = 0.67, p < 0.001; NMDA receptors, data, R² = 0.63, randomized = 0.62, p = 0.04. e Areas of the reconstructed postsynaptic densities. Each dot: a single reconstructed synapse. Error bars: median and 95% confidence interval, p > 0.6, Mann–Whitney test. f Number of clusters per synapse determined by k-means clustering. Each dot: a single reconstructed synapse. Error bars: median and 95% confidence interval, p = 0.06, Mann–Whitney test. Number of gold particles per cluster (g) and per synapse (h). Bias of particle locations toward the center or edge of the postsynaptic density in (b) was tested by comparing each group to a theoretical median of 0.5 using one-sample two-tailed Wilcoxon signed-rank tests. See Supplementary Data Table 1 for full pairwise comparisons and summary statistics.