Binary and analog variation of synapses between cortical pyramidal neurons

Abstract

Learning from experience depends at least in part on changes in neuronal connections. We present the largest map of connectivity to date between cortical neurons of a defined type (L2/3 pyramidal cells), which was enabled by automated analysis of serial section electron microscopy images with improved handling of image defects. We used the map to identify constraints on the learning algorithms employed by the cortex. Previous cortical studies modeled a continuum of synapse sizes (Arellano et al. , 2007) by a log-normal distribution (Loewenstein, Kuras and Rumpel, 2011;de Vivo et al. , 2017;Santuy et al. , 2018) . A continuum is consistent with most neural network models of learning, in which synaptic strength is a continuously graded analog variable. Here we show that synapse size, when restricted to synapses between L2/3 pyramidal cells, is well-modeled by the sum of a binary variable and an analog variable drawn from a log-normal distribution. Two synapses sharing the same presynaptic and postsynaptic cells are known to be correlated in size (Sorra and Harris, 1993;Koester and Johnston, 2005;Bartol et al. , 2015;Kasthuri et al. , 2015;Dvorkin and Ziv, 2016;Bloss et al. , 2018;Motta et al. , 2019) . We show that the binary variables of the two synapses are highly correlated, while the analog variables are not. Binary variation could be the outcome of a Hebbian or other synaptic plasticity rule depending on activity signals that are relatively uniform across neuronal arbors, while analog variation may be dominated by other influences. We discuss the implications for the stability-plasticity dilemma. †Correspondence to svenmd@princeton.edu ,

Graph Properties of the AdultDrosophilaCentral Brain

“…Assuming the strength of a connection is proportional to the number of synapses in parallel, we can plot the distribution of connection strengths, summing over the whole central brain, as shown in Figure 5. We find a nearly pure power law with an exponential cutoff, very different from the log-normal distribution of strengths found by Song [7] in pyramidal cells in the rat cortex, or the bimodal distribution found for pyramidal cells in the mouse by Dorkenwald [8]. However, we caution that these analyses are not strictly comparable.…”

An anatomical substrate of credit assignment in reinforcement learning

“…For LMAN-to-MSN-spine synapses (n = 443 pairs) a marginally significant effect was observed (shuffle control p-value = 0.045; rand control: p = 0.025; sA_dD control: p = 0.098), while no significant differences from randomized control pairs was seen for shaft synapses of both axon types (all p-values > 0.05; HVC n = 128; LMAN n = 215). Notably, while the enrichment of similarly sized pairs observed in Area X spine synapses is smaller than that observed in hippocampus 8,41 , it is comparable to the enrichment reported for neocortex 9,38,40 . To quantitatively assess the ability of the observed neural architecture to correctly implement credit assignment in the context of vocal learning, we constructed a numerical simulation of the model consistent with our anatomical findings.…”

“…2e, Extended Data Fig. 5), reminiscent of a similar finding of bimodal synapse sizes in the mammalian cortex 38 . In fact, the larger size of HVC spine synapses, compared to those from LMAN, is consistent with their hypothesized role in driving temporally specific MSN activity 39 and appears to be largely explained by a more robust tendency of HVC synapses to occupy the larger synapse state (HVC-spine: 81% µ S = 0.18 µm 2 ; 19% µ L =0.71 µm 2 ; LMAN-spine: 89% µ S = 0.14 µm 2 ; 11% µ L = 0.54 µm 2 ).…”

Chandelier cell anatomy and function reveal a variably distributed but common signal

“…Such circuit reconstruction at scale is intractable without intensive computational processing (Berning et al, 2015;Dorkenwald et al, 2017;Jain et al, 2010;Januszewski et al, 2018). We used a series of novel machine learning based methods to perform high-quality image alignment, automated segmentation, and synapse detection for the volume (Dorkenwald et al, 2019). Nonetheless, proofreading is still necessary for precise measurements of anatomy and connectivity.…”

“…The initial segmentation identified small supervoxels that were agglomerated into cells, and we built a novel cloud-based proofreading system to edit the agglomerations to perform targeted error correction. As a basis to begin analysis, the 458 cell bodies in the volume were manually classified as excitatory, inhibitory, or glia based on morphology and ultrastructural features and all 364 PyCs were proofread to correct segmentation errors (Dorkenwald et al, 2019).…”