Human hippocampal CA3 uses specific functional connectivity rules for efficient associative memory

Abstract

The human brain has remarkable computational power. It generates sophisticated behavioral sequences, stores engrams over an individual's lifetime, and produces higher cognitive functions up to the level of consciousness. However, so little of our neuroscience knowledge covers the human brain, and it remains unknown whether this organ is truly unique, or is a scaled version of the extensively studied rodent brain. To address this fundamental question, we determined the cellular, synaptic, and connectivity rules of the hippocampal CA3 recurrent circuit using multicellular patch clamp-recording. This circuit is the largest autoassociative network in the brain, and plays a key role in memory and higher-order computations such as pattern separation and pattern completion. We demonstrate that human hippocampal CA3 employs sparse connectivity, in stark contrast to neocortical recurrent networks. Connectivity sparsifies from rodents to humans, providing a circuit architecture that maximizes associational power. Unitary synaptic events at human CA3–CA3 synapses showed both distinct species-specific and circuit-dependent properties, with high reliability, unique amplitude precision, and long integration times. We also identify differential scaling rules between hippocampal pathways from rodents to humans, with a moderate increase in the convergence of CA3 inputs per cell, but a marked increase in human mossy fiber innervation. Anatomically guided full-scale modeling suggests that the human brain's sparse connectivity, expanded neuronal number, and reliable synaptic signaling combine to enhance the associative memory storage capacity of CA3. Together, our results reveal unique rules of connectivity and synaptic signaling in the human hippocampus, demonstrating the absolute necessity of human brain research and beginning to unravel the remarkable performance of our autoassociative memory circuits.

Cell-specific wiring routes information flow through hippocampal CA3

“…2B ). No intersomatic distance dependence of connectivity was observed either, in line with other reports in the CA3 of rats, mice, and humans (Guzman et al, 2016; Sammons et al, 2024; Watson et al, 2024) ( Fig. 2E, Supplementary Fig.…”

“…We next examined excitatory postsynaptic potential (EPSP) properties. As previously observed in multiple species, (Guzman et al, 2016;Miles and Wong, 1986;Sammons et al, 2024;Watson et al, 2024), CA3-CA3 synapses are relatively weak, with mean unitary depolarisations across all subclasses of just 0.94 ± 0.10 mV (Fig. 3E).…”

“…After PN subclassification, our dataset assessed 3114 putative connections between identified CA3 PNs, observing 85 monosynaptic connections. Due to slicing, this is likely an underestimate of synaptic connectivity in the intact brain, which we have previously estimated to be at least 1.4-fold higher than recorded in slices without further corrections (Watson et al, 2024). Connectivity was no different between sexes or hemispheres ( Supplementary Fig.…”

“…A subset of recorded CA3 PNs from this dataset (1272 tested connections) was previously included in a manuscript for cross species connectivity comparison (Watson et al, 2024).…”