2025 Theses Doctoral

Connecting Behavior to Brain Networks: Sundowning in Alzheimer’s Disease and Beyond

Jin, Michelle

Understanding how distributed neural ensembles coordinate behavior remains a central question in systems neuroscience. Immediate-early gene (IEG) expression commonly serves as a proxy for recent neuronal activation and can be combined with network analysis to model functional connectivity underlying behavior. Genetic strategies for labelling multiple IEG ensembles have powerfully enabled the study of activity patterns underlying two distinct behavioral experiences. These approaches have been well-harnessed to study the neural correlates of memory encoding and retrieval. However, brain-wide mapping of multiple ensembles and their overlap has been constrained by both technical and analytical limitations. This dissertation develops and applies a scalable workflow for brain-wide multiple IEG mapping and downstream network analysis, linking cellular activity to systems-level patterns underlying behavior.

In the first portion of this work, I introduce SMARTTR, an R-based framework for multi-ensemble registration, quantification, and graph-theory analysis of genetically activity-tagged datasets. SMARTTR integrates atlas alignment, permutation testing, and network visualization, and supports direct import of external brain-mapped data. As a demonstration, I mapped the neural ensembles underlying the acquisition and expression of learned helplessness, a model of stress-related behavior. Network analysis revealed that the substantia nigra pars reticulata (SNr) acts as a highly influential node whose connectivity reverses between helpless and resilient states, and that the agranular insula (AI) emerges as a major hub during expression of helplessness. These findings illustrate how network analysis can identify circuit substrates of affective states.

The following chapter extends network analysis to the study of Alzheimer’s disease (AD), examining sex differences in memory retrieval networks. Mapping c-Fos following contextual fear recall in control and AD model mice, we observed strong sexual dimorphism in network organization despite comparable behavioral performance between sexes. Both male and female wild-type mice exhibited central positioning of the ventral dentate gyrus (vDG) within retrieval networks—a feature that was lost in AD animals of both sexes. These results suggest that AD pathology disrupts a shared hippocampal hub while remodeling network topology in a sex-specific manner.

The next chapter investigates sundowning, the evening worsening of agitation and confusion that can affect over half of AD patients. I establish a preclinical sundowning model using APP/PS1 mice by characterizing age-related changes in sleep–wake rhythms and sleep-wake phase-dependent behavior. Unsupervised analyses of pose-tracked behavior revealed a sundowning-like behavioral fingerprint resembling human symptoms. Using activity-dependent tagging and SMARTTR, I identified a distributed pattern of sensorimotor hyperconnectivity unique to the Sundown phase in AD mice. Global network properties reveal reduced connectivity and efficiency in AD networks relative to controls at Sundown. Complementary analysis of resting-state functional magnetic resonance imaging (rs-fMRI) data from the Alzheimer’s Disease Neuroimaging Initiative uncovered an analogous signature in the Salience network of human AD subjects, and overlapping hubs with those identified in mice in the anterior cingulate and prelimbic cortices. Together, these findings delineate a cross-species correspondence between mesoscale network dysfunction and circadian neuropsychiatric symptoms.

The following chapter broadens the translational scope of this work by evaluating the emergence of disease-modifying anti-amyloid therapies for AD and proposing a new conceptual metric to quantify clinically meaningful benefit as a joint measure of preserved function and time. This framework bridges operational definitions of cognitive improvement with patient-centered therapeutic evaluation.

Collectively, this dissertation advances both methodological and conceptual links between circuit-level mapping, behavior, and clinical translation. By combining whole-brain IEG mapping, graph theory, and quantitative behavioral analysis, I demonstrate a generalizable platform for identifying and modeling functional brain networks that underlie complex behavior and disease—setting the stage for future perturbations and offering new routes from systems neuroscience to therapeutic insight.