Metabolic reprogramming and dysfunction resulting from amyloid ß42-induced oxidative stress is emerging as
an important mechanism at the onset of the pathophysiological changes that ultimately lead to Alzheimer’s
disease (AD). Cellular metabolic function and interactions are highly dynamic and heterogeneous and thus
require tools that are non-destructive and yield measurements with high spatial resolution. Label-free, two photon
excited fluorescence (TPEF) imaging has been used by a number of groups to identify metabolic functional
changes associated with AD and senile plaques. We propose to build on our established and validated multi-
parametric optical metabolic readouts of mitochondrial dynamics and redox state, based on extraction and
analysis of NAD(P)H and FAD TPEF images, to characterize the spatiotemporal dynamics of metabolic
reprogramming that takes place at the very early stages of plaque formation in AD. To achieve this, we will rely
on a novel 3D brain-like tissue model of HSV-1-induced AD. The tissues consist of a hybrid silk-collagen hydrogel
embedded with human induced neural stem cells (hiNSCs) that are infected with HSV-1. Over the span of a
week, the tissues develop several physiologically relevant AD traits, including neuronal loss, gliosis,
neuroinflammation, synaptic dysfunction, and multicellular Aß fibrillar plaque like formations (PLFs). These
tissues are easily accessible for repeated label-free TPEF imaging over reasonable time scales to enable
dynamic monitoring of the metabolic changes that occur in the immediate milieu of PLFs as they are forming. To
ensure accurate monitoring of metabolic function, free of interference from other endogenous fluorophores, we
aim to perform spectral imaging studies to characterize the excitation/emission profiles of key fluorescent
contributors (Aim 1). This information will be essential for establishing and validating an efficient image
acquisition protocol that relies on a limited number of images, which can be analyzed without compromising the
potential to extract quantitative assessments of NAD(P)H, FAD (and corresponding redox ratio and mitochondrial
clustering metrics), lipofuscin and plaque associated fluorescence. In addition, we propose to enhance the
efficiency of image acquisition and analysis using deep learning approaches that combine denoising and object
classification steps (Aim 2). We will demonstrate the potential of this label-free, multiparametric, TPEF imaging
approach to yield unique insights on the dynamics of metabolic reprogramming events that are associated with
AD development through repeated imaging of the HSV-1 treated hiNSCs in the absence or presence of anti-viral
or metformin treatment (Aim 3). We expect our findings to provide a foundation for pursuing the use of this
combined tissue model/label-free, metabolic imaging platform to address important mechanistic questions
regarding AD etiology and the potential success of novel related treatments. The imaging and image analysis
approaches we establish will be broadly applicable to the study of brain tissues and could be extended to probe-
based minimally invasive measurements in animals and possibly humans.