Graph Theoretic and Motif Analyses of the Hippocampal Neuron Type Potential Connectome

Abstract

We computed the potential connectivity map of all known neuron types in the rodent hippocampal formation by supplementing scantly available synaptic data with spatial distributions of axons and dendrites from the open-access knowledge base Hippocampome.org. The network that results from this endeavor, the broadest and most complete for a mammalian cortical region at the neuron-type level to date, contains more than 3200 connections among 122 neuron types across six subregions. Analyses of these data using graph theory metrics unveil the fundamental architectural principles of the hippocampal circuit. Globally, we identify a highly specialized topology minimizing communication cost; a modular structure underscoring the prominence of the trisynaptic loop; a core set of neuron types serving as information-processing hubs as well as a distinct group of particular antihub neurons; a nested, two-tier rich club managing much of the network traffic; and an innate resilience to random perturbations. At the local level, we uncover the basic building blocks, or connectivity patterns, that combine to produce complex global functionality, and we benchmark their utilization in the circuit relative to random networks. Taken together, these results provide a comprehensive connectivity profile of the hippocampus, yielding novel insights on its functional operations at the computationally crucial level of neuron types.

Weighing the Evidence in Peters’ Rule: Does Neuronal Morphology Predict Connectivity?

“…This resource tracks the axonal and dendritic presence of 122 neuron types across 26 subregions and layers of the rodent hippocampal formation (HF) (http://hippocampome.org/morphology), including the dentate gyrus (DG), CA3/CA2/CA1, the subiculum (SUB), and the entorhinal cortex (EC). Applying Peters’ rule at Level 1 for the 122 × 122 pairs of neuron types yields a potential connectome of >3300 potential synapses by identifying all axonal–dendritic colocations in any anatomical parcel [40]. In other words, Peters’ rule is useful in excluding approximately 80% of all-to-all type–level connectivity.…”

“…To a first approximation, certain connections are based on the axonal–dendritic coincidence [53]. Describing the network in this fashion is useful to expose the backbone architecture underlying the main signal propagation paths and local processing unit blocks [40]. Next, accounting for cell-type-specific targeting (e.g., axo-axonic and interneuron specific cells in the hippocampus) in Level 1 violation of Peters’ rule may reveal precise functional ramifications, including direct inhibition or disinhibition of key players within the larger network [54].…”

A comprehensive knowledge base of synaptic electrophysiology in the rodent hippocampal formation

“…This knowledge base identifies 122 neuron types across the dentate gyrus, areas CA3, CA2, CA1, subiculum, and entorhinal cortex based on their main neurotransmitter released (glutamate or GABA), the laminar distribution of axons and dendrites, and converging molecular and electrophysiological evidence. This framework conveniently extends to the notion of potential connections, defined as directional pairs of presynaptic and postsynaptic neuron types (Rees et al, ). Based on available evidence regarding neuronal morphology and known targeting specificities (Rees, Moradi, & Ascoli, ), we estimate the existence of at least 3,120 potential connections in the hippocampal formation (http://hippocampome.org/connectivity).…”

Simple models of quantitative firing phenotypes in hippocampal neurons: comprehensive coverage of intrinsic diversity

“…Future studies should aim to identify a family of models for an experimentally known 525 network-level property within the anatomical constraints of connectivity among hippocampal neuron types [55] using the sampling regions for those types created in this study. Then, the identified family of models should be evaluated for their predictive power, or one could investigate how the predictive abilities increase by scaling up the network, or by adding more mechanisms such as synaptic plasticity and spatial context for synaptic integration.…”