Unraveling Autophagy Regulation: Roles of V-ATPase, Translational Control, and Machine Learning-Based Genomic Prediction

Abstract

Macroautophagy/autophagy is a vital cellular process responsible for directing intracellular components and damaged organelles to the lysosome for degradation and recycling. This mechanism is crucial in various human diseases, as it maintains cell survival, bioenergetic stability, organismal development, and controlled cell death. This thesis seeks to elucidate the regulatory mechanisms of autophagy through three key approaches: (a) examining the role of the vacuolar-type H+-translocating ATPase (V-ATPase) in the regulation of autophagy within the context of follicular lymphoma, (b) analyzing how upstream open reading frames (uORFs) influence the translational control of autophagy, and (c) leveraging machine learning and artificial intelligence to identify novel autophagy-related genes. The V-ATPase is the major proton pump for intra-organellar acidification. Therefore, the integrity of the V-ATPase is closely associated with cellular homeostasis. The identification of recurrent mutations in subunits and regulators of the V-ATPase in follicular lymphoma underscores the involvement of autophagy and energy-sensing pathways in the pathogenesis of this disease. In chapter 2, we report a new hotspot mutation in the ER-resident V-ATPase chaperone VMA21, leading to VMA21 mislocalization to lysosomes. This mislocalization impairs V-ATPase function, preventing full lysosomal acidification and functionality. Consequently, this dysfunction leads to amino acid depletion in the cytoplasm and triggers compensatory autophagy, as confirmed by multiple assays in human and yeast cells. In chapter 3, with the aim of studying V-ATPase-related mutations using the yeast model system, we report that Big1 is another homolog of ATP6AP1 in yeast cells, and we characterize the role of Big1 in V-ATPase assembly. In addition to its role in acidifying the vacuole or lysosome, our data support the concept that the V-ATPase may function as part of a signaling pathway to regulate autophagy through a mechanism that is independent from Tor/MTOR. While autophagy regulation has been extensively studied at the transcriptional level, relatively little is understood about its translational control. In chapter 4, we present evidence of upstream open reading frame-mediated translational regulation of multiple Atg proteins in Saccharomyces cerevisiae and human cells. In yeast, the translation of several key autophagy regulators, including Atg13, is repressed by canonical uORFs under nutrient-rich conditions and is activated during nitrogen-starvation. Similarly, we found that the predicted non-canonical uORFs in human ATG4B and ATG12 suppress the translation of downstream coding sequences. These findings indicate that uORF-mediated translational control is a conserved mechanism across ATG genes from yeast to human, suggesting a model in which certain ATG genes circumvent the general translational repression occurring under stress conditions to sustain appropriate autophagy levels. Data-driven approaches for genomic discovery have advanced significantly with the rise of machine learning techniques. However, the inherent complexity of genomic data, with its intricate relationships, poses challenges in effective modeling and analysis. In chapter 5, we introduce a recommender framework inspired by an industrial recommendation system specifically designed to predict and identify autophagy-related genes by leveraging information from diverse genomic databases. This framework constructs a graph where nodes represent genes with embedding vectors derived from a knowledge graph of genomic databases, and edges capture protein-protein interactions. By filtering critical features and calculating relevancy scores for genes known to be associated or unassociated with autophagy, we can recommend unlabeled gene candidates that are likely involved in autophagy. We highlight its potential as a robust tool for discovering novel autophagy-related genes and accelerating insights in genomic research.