Bioengineered Neural Networks as Computational Systems: Designing Anatomically Inspired Microcircuits to Decode Neural Function and Dysfunction

Abstract

Engineered neural networks provide invaluable platforms for studying neural function and dysfunction within controlled microenvironments. Recent advancements in engineered interfaces, such as multinodal microfluidic platforms and 3D scaffolding technologies, have opened a broad avenue for manipulating the physical dimensionality of such networks to support the emergence of circuit motifs akin to those seen in vivo. However, the impact of these topologies on the functional dynamics of the networks remains largely elusive. In the work included in this thesis, we aimed to study how principles of neural design could be applied to engineer more anatomically relevant neural network models, and assess the impact of such architectures on the networks’ computational capacity. In Paper I, we introduce highly nanoporous microelectrodes optimized for biocompatibility, electrochemical and electrophysiological performance, which significantly enhanced the quality of data acquisition in subsequent papers. Paper II presents a twonodal microfluidic device with microfluidic tunnel geometries inspired by the Tesla valve, facilitating selective control of afferent-efferent connectivity between distinct neural populations. In Paper III, we engineer feedforward cortical-hippocampal neural networks in four-nodal microfluidic devices to recapitulate the cortical-hippocampal loop. By applying amyloid beta to the first node in the networks, we furthermore demonstrate the utility of these devices for preclinical Alzheimer’s disease modelling. Paper IV showcases a 12-nodal microfluidic platform for recapitulating the laminar structure of the neocortex. By applying hypoxia to a hub node in the network, we furthermore demonstrate the utility of the device for studying neural adaptation and degeneracy in response to localized perturbations. Finally, in Paper V, we demonstrate fabrication and utilization of biocompatible 3D interfaces of the polymer SU-8 for supporting 3D network growth, readily compatible with the PDMS-based microfluidics and microelectrode arrays. Using electron microscopy and immunocytochemistry, we show that neurons utilize these scaffolds for self-assembly into complex multi-layered networks. Electrophysiological characterization furthermore indicates a rich dynamical repertoire, with distinct differences compared to networks established on planar interfaces. Collectively, our findings demonstrate that hierarchical, multi-nodal configurations enhance the computational complexity and dynamic repertoire of engineered neural cortical and hippocampal networks. Furthermore, three-dimensionality may enhance functional complexity by contributing to more sparsely connected networks. The presented neuroengineering solutions offer versatile tools for investigating neural microcircuitry, shedding light on the relationship between network topology and function, and significantly broadening the scope of advanced preclinical neural network models.