Obesity is increasingly recognized as a disorder involving impaired central endocrine regulation in which hypothalamic nutrient sensing, leptin responsiveness, and intracellular quality control are closely intertwined. Within the arcuate nucleus, autophagy exerts divergent metabolic effects in the two principal appetite-regulating neuronal populations. In pro-opiomelanocortin (POMC) neurons, autophagy supports metabolic homeostasis by promoting alpha-melanocyte-stimulating hormone (α-MSH) production, preserving leptin and insulin sensitivity, and activating peripheral lipophagy; its disruption predisposes to hyperphagia, leptin resistance, and obesity-related metabolic dysfunction. In contrast, autophagy in agouti-related peptide (AgRP) neurons promotes feeding and energy conservation by maintaining AgRP expression, integrating ghrelin signals, and sensing peripheral fatty acids; its inhibition alleviates diet-induced obesity in mice. This review critically evaluates discrepancies among models based on the deletion of autophagy-related gene 5,7, and 12 (Atg5, Atg7, and Atg12), as well as differences arising from developmental versus adult manipulations, dietary paradigms, and the reliance of static autophagy markers. We further discuss how POMC and AgRP neuron autophagy may interact with broader hypothalamic and extra-hypothalamic networks, including basonuclin 2 (BNC2), steroidogenic factor 1 (SF1) neurons in the ventromedial hypothalamus (VMH), and paraventricular hypothalamic nucleus (PVH) pathways, as well as extra-hypothalamic components involving the brainstem, glial and tanycytic cells, and neurovascular interaction. Chaperone-mediated autophagy (CMA) has emerged as a potential pathway potentially linking saturated fatty acid exposure, LAMP2A-dependent proteostasis, and insulin responsiveness in hypothalamic neurons, although direct evidence remains limited. Because direct demonstration of altered autophagic flux in human POMC or AgRP neurons remains lacking, hypothalamic autophagy should currently be viewed as a mechanistic and translational research framework rather than a validated clinical target. Future therapies will likely require pathway-biased, neuron-type-specific, and circuit-aware modulation rather than global activation or inhibition of autophagy.
