The Woodruff School

SUMMARY

Alzheimer’s disease (AD) is a progressive neurodegenerative disease, accounting for about two thirds of dementia cases. Despite significant efforts to diagnose and cure AD there are still no effective therapeutics to halt disease progression. While the conventional understanding attributes memory loss to the buildup of amyloid and tau proteins, emerging evidence suggests that cognitive decline in AD may stem from neuronal circuit dysregulation rather than protein aggregation. Specifically, alterations in the excitability of inhibitory interneurons may contribute to circuit dysfunction, although the evolution of this dysregulation across brain regions and over time remains poorly understood. To address this gap, this thesis systematically investigated the emergence of parvalbumin interneuron dysfunction in AD, confirming their early involvement in vulnerable brain regions. To study these PV interneurons at the single cell level, with sufficient spatial and temporal resolution, this thesis will utilize patch clamp electrophysiology. The patch clamp technique is remains necessary for fully elucidating cell-type-specific behavior, although it is difficult and time-intensive. While patch clamp systems have emerged that automate certain aspects of the procedure, there remain challenges that can be remedied with improved automation techniques. To overcome these obstacles, several strategies have been developed to improve the whole-cell success rates and facilitate the execution of automated, high-throughput investigations. In the initial identification of cells within acute brain slices, a deep learning methodology automatically nominate neurons for subsequent automated experiments. Addressing concerns regarding pipette localization errors, a convolutional neural network, specifically ResNet101, has been adapted and trained to autonomously detect and rectify the misplacement of pipette tips during automated in vitro patch clamp experiments. Furthermore, to facilitate investigations into synaptic connections between neurons, a method named patch-walking was demonstrated in brain slices, enabling efficient finding of synaptic connections.

SUBJECT:	Ph.D. Dissertation Defense

BY:	Mercedes Gonzalez

TIME:	Wednesday, April 24, 2024, 9:15 a.m.

PLACE:	Whitaker Ford Building, Room 1214, https://gatech.zoom.us/j/92212674946

TITLE:	Automated cellular electrophysiology to investigate the role of interneurons in Alzheimer’s disease

COMMITTEE:	Craig Forest, Co-Chair (ME) Matthew Rowan, Co-Chair (Biology, Emory) Annabelle Singer (BME) Bilal Haider (BME) Christopher Rozell (ECE)

SUMMARY Alzheimer’s disease (AD) is a progressive neurodegenerative disease, accounting for about two thirds of dementia cases. Despite significant efforts to diagnose and cure AD there are still no effective therapeutics to halt disease progression. While the conventional understanding attributes memory loss to the buildup of amyloid and tau proteins, emerging evidence suggests that cognitive decline in AD may stem from neuronal circuit dysregulation rather than protein aggregation. Specifically, alterations in the excitability of inhibitory interneurons may contribute to circuit dysfunction, although the evolution of this dysregulation across brain regions and over time remains poorly understood. To address this gap, this thesis systematically investigated the emergence of parvalbumin interneuron dysfunction in AD, confirming their early involvement in vulnerable brain regions. To study these PV interneurons at the single cell level, with sufficient spatial and temporal resolution, this thesis will utilize patch clamp electrophysiology. The patch clamp technique is remains necessary for fully elucidating cell-type-specific behavior, although it is difficult and time-intensive. While patch clamp systems have emerged that automate certain aspects of the procedure, there remain challenges that can be remedied with improved automation techniques. To overcome these obstacles, several strategies have been developed to improve the whole-cell success rates and facilitate the execution of automated, high-throughput investigations. In the initial identification of cells within acute brain slices, a deep learning methodology automatically nominate neurons for subsequent automated experiments. Addressing concerns regarding pipette localization errors, a convolutional neural network, specifically ResNet101, has been adapted and trained to autonomously detect and rectify the misplacement of pipette tips during automated in vitro patch clamp experiments. Furthermore, to facilitate investigations into synaptic connections between neurons, a method named patch-walking was demonstrated in brain slices, enabling efficient finding of synaptic connections.