Selective inhibition in CA3: A mechanism for stable pattern completion through heterosynaptic plasticity

Abstract Neural assemblies representing different engrams compete for successful retrieval in the CA3 region of the hippocampus, yet the detailed mechanisms underlying their formation remain elusive. Recent research indicates that hippocampal inhibitory neurons respond selectively to stimuli and exhibit diverse plasticity, suggesting their significant role in engram formation. Conventional attractor network models for CA3 typically employ global inhibition, where inhibitory neurons uniformly suppress the activity of excitatory neurons. However, these models may not fully capture the complex dynamics of competition arising from sparse distributed coding and may not accurately reflect the specific roles of inhibitory neurons in the competition between neural assemblies during memory retrieval. We propose a mechanism for engram formation in CA3 using a spiking neural network model, emphasizing the critical role of the association between excitatory and inhibitory neurons through heterosynaptic plasticity. In our model, inhibitory neurons are associated with specific neural assemblies during encoding and selectively inhibit excitatory neurons involved in competing assemblies during retrieval. With a simplified dentate gyrus (DG) in a feed-forward structure, this proposed mechanism results in sparsely distributed engrams in CA3. The sparse distributed coding in the model allows us to investigate the effects of selective inhibition on pattern completion under various configurations, such as partially overlapping competing engrams. Our results demonstrate that selective inhibition provides more stable pattern completion and enhances retrieval performance compared to global inhibition alone. Furthermore, the observed neural activity in the hippocampal subregions of the model aligns with experimental findings on these regions’ roles in pattern separation and pattern completion. 1 Author Summary We explored how memories are stored and retrieved in the hippocampus by focusing on the CA3 region, a critical component of memory processes. Using a spiking neural network model, we propose a new mechanism in which specific inhibitory neurons selectively control the activity of other neurons during memory retrieval. We found that this selective inhibition can be naturally induced during memory encoding. This selective inhibition offers an alternative to traditional models that assume global suppression and provides a more nuanced understanding of how memories compete for successful retrieval. Our findings suggest that this selective inhibition improves the stability and accuracy of memory recall. The model also aligns with known biological functions of the hippocampus, shedding light on how complex memory processes, such as distinguishing between similar memories and accurately reconstructing past experiences, might be managed in the brain. This research offers new insights into the dynamic roles of inhibitory neurons in balancing memory encoding and retrieval, enhancing our understanding of memory function.