TUNING AND TESTING DEVELOPING NEURAL NETWORKS

In vitro neural networks are powerful model systems that can help bridge the gap between single-cell properties and emergent network behaviors. This allows for the linking of cell-scale properties to network behavior, providing an important step in connecting individual cellular actions to the coordination at the level of the brain. This dissertation investigates how different interventions — global, local, and spatiotemporally localized — alter neural network dynamics, primarily using calcium as a unifying readout. Specifically, by manipulating the structural, oxidative, and electrophysiological pathways, this work elucidates multimodal impacts on calcium dynamics across the network. We study how the network-level calcium signal is impacted not only across scales but also through multiple pathways. To investigate structural interventions, we examined how global tuning of the actin cytoskeleton modulates human neural progenitor cell (hNPC) calcium signals. We first characterized actin dynamics and rhythms in cells, finding that while actin remains dynamic throughout differentiation, its scale of activity shifts significantly. Global pharmacological arrest of the actin cytoskeleton revealed a dual effect: while it did not increase the proportion of active cells, it significantly increased the frequency of activity within those cells. Furthermore, when the network was chemically primed for excitability, actin arrest significantly increased the proportion of active cells. Collectively, these results suggest that actin may act as a structural dampener of hNPC calcium dynamics, regulating the threshold for network excitability. To investigate oxidative modulation, we studied the impacts of local, non-invasive photobiomdulation on early differentiated hNPC using 370 nm light. This work confirmed that 370 nm light induces reactive oxygen species (ROS) in a well-characterized, dose-dependent manner governed by light intensity and dwell time. We demonstrated that ROS induction is tightly localized and confined to stimulated regions. Although we observed no major changes in calcium activity patterns, there was a notable increase in basal calcium levels in stimulated cells. These findings suggest that 370 nm light is a viable tool for non-invasive modulation of neural cells. This modest impact on basal calcium establishes a foundation for future work exploring how light-mediated oxidative shifts can influence long-term neural processes, such as differentiation or synaptic communication. Finally, we addressed electrophysiological modulation through local targeted spatiotemporal intervention. By stimulating primary neurons using optogenetics within a single field view, we compared how uniformly stimulated cells were contextually impacted by neighboring cells stimulated with varied stimulation patterns. This approach allowed us to determine how the context of localized activity in one regions of cells influences signal processing and propagation of network wide-dynamics. We found that patterned spatiotemporal stimulation could influence uniformly stimulated cells. These results highlight how the timing and geometry of electrical and stimulated inputs serve as critical determinants of population-level coordination. Together, these studies characterize three distinct regulatory axes that tune the behavior of neural networks: cytoskeletal, oxidative, and electrophysiological. By defining these pathways, this work provides a framework for understanding how structural and oxidative dynamics, often overlooked in traditional models, can contribute to network level functional changes. In addition to medical applications and research on neural modulation, these findings reveal how diverse modalities of information are coupled within neural populations. This has far reaching potential to diversify and inspire novel computing and bio-inspired architectures.

Department dc:contributor.department

Biology

Year dc:date.issued

2026

Author dc:creator

• Gates III, Sylvester James

Advisor dc:contributor.advisor

• Losert, Wolfgang

Language dc:language.iso

en

Identifier

https://doi.org/10.13016/u7fo-wyzv

OAI identifier oai:identifier

oai:drum.lib.umd.edu:1903/35537