Population coupling predicted input connections in two mouse V1 EM volumes as in slices (odds ratio 1.25–1.36 per SD), but within layer 2/3 of MICrONS the ratio was only 1.03–1.09

Question: Okun et al. (2015) reported that neurons with high population coupling receive more synaptic connections from neighbours (slice odds ratio about 1.36 per SD of coupling) but do not send more. We tested this claim with EM connectomes and calcium imaging of the same neurons. Approach: We calculated coupling from the full traces of 168,971 ROIs in 19 MICrONS scans (33,767 EM-matched nuclei) and 4,579 ROIs in 3 V1 Deep Dive (V1DD) sessions (180 proofread matched cells). We fitted logistic models of local connections with one intercept per presynaptic cell, in three designs: pooled pairs within 100 um without an overlap term (as in slices), layer 2/3 pairs only, and pairs conditioned on axon-dendrite overlap. We also correlated coupling with EM input synapse counts. Findings: Pooled over cell types, the odds ratio per SD of target coupling was 1.25 (95% CI 1.22-1.28; 1,705 presynaptic cells, 30,042 targets) in MICrONS and 1.36 (1.21-1.52) in V1DD, as in slices. In MICrONS, nucleus volume alone reduced it to 1.09, because large cells had both high coupling and many inputs. For layer 2/3 to layer 2/3 pairs it was 1.09 (1.06-1.12), and with axon-dendrite co-travel as opportunity it was 1.03 (0.99-1.07) for layer 2/3 targets; both intervals exclude 1.36. V1DD layer 2/3 gave 1.14 (0.96-1.37) for 60 targets. Outputs did not follow coupling (MICrONS 0.99). Coupling correlated with input synapse counts (partial Spearman 0.27 in 483 layer 2/3 cells with proofread dendrites; 0.44 in 60 V1DD layer 2/3 cells). Coupling was reliable within a scan (0.90) and across days (0.54), but agreed only partly between videos and darkness (mean 0.38). Conclusion: In these volumes, choristers receive more input mainly because they are larger cells with more synapses and more axon-dendrite overlap; extra connection probability per unit of overlap is small. Limitations: calcium coupling in awake mice; one mouse per dataset; partial proofreading and automatic coregistration (87% agreement with manual matches) can bias estimates toward 1; V1DD has few layer 2/3 targets.

population coupling · choristers and soloists · connectomics · MICrONS · V1 Deep Dive · mouse visual cortex · calcium imaging · synaptic connectivity · electron microscopy · replication · axon-dendrite overlap · layer 2/3