Paired recordings in mouse V1 found most excitatory but less than half of the inhibitory connections that MICrONS electron microscopy predicted for 17 matched cell classes

Question: Does a dense electron microscopy (EM) connectome predict which cell pairs paired recordings find connected, for each cortical cell class? Data: MICrONS minnie65 v1507 (1,524 V1 neurons with proofread axons; 40.0 million candidate pairs within 400 µm; 667,937 connected pairs) and the Allen Synaptic Physiology multipatch data r2.1 (16,978 probed mouse V1 pairs; 1,317 connections). We used the 17 classes of Campagnola et al. 2022 (L2/3, L4, L5 ET, L5 IT, L6 CT pyramidal cells; Pv, Sst and Vip cells in four layers). Method: For each probed multipatch pair, we took the EM connection probability of the same class pair at the same soma distance. EM distances were scaled by 1.22 (cortical thickness 893 vs 730 µm). We compared expected and detected counts with a cluster bootstrap over EM cells and multipatch slices. Results: EM gave the correct order of connection probability across 72 class pairs (Spearman 0.64; 0.74 with expert EM labels). Multipatch detected 0.84 (95% CI 0.73-0.95) of the EM-expected connections from pyramidal cells but only 0.44 (0.40-0.49) from interneurons (ratio 1.89, 1.61-2.21). The inhibitory gap depended on the pathway: 0.63-0.95 for Pv to Pyr, Pv to Pv, Sst to Vip and Vip to Sst (0.52), but 0.15-0.21 for Pv to Sst, Pv to Vip, Sst to Sst and Vip to Pyr, Pv and Vip. The E/I ratio stayed between 1.46 and 2.04 with stricter proofreading, expert or agreeing labels, synapse size cutoffs, 1.3 or 2 mM Ca, animal age and distance scale. A model in which inhibitory EM connections need about 7,100 cleft voxels (about two median synapses) while almost all excitatory connections are found gave the best held-out fit, and agrees with smaller inhibitory quantal amplitudes (0.105 vs 0.219 mV). But its gain over separate E and I scale factors was not significant per class pair (z = 0.54), it left the rare inhibitory motifs under-detected, and it predicts a distance trend that the data do not show. A simulated 350 µm slice removed only 1-5% of EM connections, more for excitatory cells. Conclusion: for excitatory cells, EM synapse presence predicts paired-recording connectivity within about 20%. For interneurons, EM shows about two times more connections, mainly in pathways that physiology calls rare. Main limitations: one EM animal of a different age, morphological EM classes vs Cre-line classes, and partial axon proofreading.

connectomics · MICrONS · multipatch · synaptic physiology · mouse visual cortex · connection probability · interneurons · electron microscopy · structure-function · cell types · Allen Institute · inhibitory connectivity