HCN channels enhance robustness of patterned activity propagation in heterogeneous conductance-based ring networks
Continuous attractor network (CAN) models lend a powerful framework that has provided deep insights about several aspects of brain physiology. However, most CAN models employ homogeneous, rate-based or artificially spiking neurons with precisely structured synaptic connectivity, precluding detailed analyses of the impact of specific neural-circuit components and associated heterogeneities on CAN dynamics. To address this caveat, we built populations of tunable and scalable conductance-based, physiologically constrained, ring network models consisting of distinct rings of excitatory and inhibitory neurons. We assessed the network for its ability to sustain robust propagation of patterned activity across the rings. First, in homogeneous ring networks, we found that robust activity propagation could be sustained through several different combinations of synaptic weights, demonstrating synaptic degeneracy in the emergence of robust activity propagation. We incorporated intrinsic heterogeneity through randomized perturbations to ion channel parameters of all neurons and synaptic heterogeneity by adding jitter to the Mexican-hat connectivity between inhibitory neurons. We found the number of networks exhibiting robust propagation of patterned activity to reduce with increase in the degree of synaptic or intrinsic heterogeneities. Motivated by the ability of intrinsic neuronal resonance to stabilize heterogeneous rate-based CAN models, we hypothesized that increasing HCN-channel (a resonating conductance) density would stabilize activity propagation in heterogeneous ring networks. Strikingly, we observed that increases in HCN-channel density resulted in a pronounced increase in the proportion of heterogeneous networks that exhibited robust activity propagation, across multiple trials and across three degrees of either form of heterogeneity. Together, heterogeneous networks made of neurons with disparate intrinsic properties and variable HCN channel densities yielded robust activity propagation, demonstrating intrinsic degeneracy in the emergence of robust activity propagation. Finally, as HCN channels also contribute to changes in excitability, we performed excitability-matched controls with fast HCN channels that do not introduce resonance. We found that fast HCN channels did not stabilize heterogeneous network dynamics over a wide range of conductance values, suggesting that the slow negative feedback loop introduced by HCN channels is a critical requirement for network stabilization. Together, our results unveil a cascade of degeneracy in ring-network physiology, spanning the molecular-cellular-network scales. These results also demonstrate a critical role for the widely expressed HCN channels in enhancing the robustness of heterogeneous neural circuits by implementing a slow negative feedback loop at the cellular scale.