The aging brain accumulates dysfunctional neurons not because it actively prunes inefficient cells, but because senescent glia lose the ability to recognize a metabolic “eat‑me” signal on stressed neurons. We propose that neurons facing energy deficit expose surface phosphatidylserine (PS) in proportion to their ATP/AMP ratio, a signal normally sensed by glial TREM2 and CX3CR1 receptors to trigger phagocytosis. In young tissue, this coupling ensures rapid removal of metabolically compromised neurons, preserving network efficiency. With age, glial senescence downregulates TREM2/CX3CR1 expression and impairs lysosomal acidification, uncoupling PS exposure from engulfment. Consequently, neurons that would be cleared persist, release SASP, and fuel a vicious cycle of glial reactivity and further neuronal stress.

This hypothesis generates clear, testable predictions. First, pharmacological elevation of neuronal AMP‑activated protein kinase (AMPK) should increase surface PS on neurons in vitro, measurable by Annexin V‑flow cytometry. Second, blocking TREM2 or CX3CR1 on microglia will abolish the phagocytic response to AMPK‑activated neurons despite high PS exposure. Third, in aged mice, senolytic clearance of p16^high^ glia followed by an AMPK activator (e.g., AICAR) will restore glial phagocytic flux, reduce Annexin V^+^ neuron burden, and improve cognitive performance relative to senolytics alone. Fourth, sex‑dependent efficacy observed with ABT263 may stem from baseline differences in glial AMPK activity; measuring p‑AMPK in isolated microglia should predict senolytic response magnitude.

To falsify the model, we would need to show that manipulating neuronal metabolic state does not alter PS exposure, or that enhancing glial TREM2/CX3CR1 does not rescue phagocytosis of metabolically stressed neurons in senescent glia backgrounds. Likewise, if senolytic plus AMPK activation fails to lower Annexin V^+^ neuron numbers or improve behavior despite confirmed drug target engagement, the core premise—that a metabolic tag governs glial clearance and its loss drives pathology—would be undermined.

Experimental design: primary neuron‑glia cocultures from young and aged mice, treated with oligomycin to induce ATP drop, will be imaged for live‑cell PS labeling and microglial phagocytic cups (Lifeact‑GFP). In vivo, 20‑month‑old male and female APP/PS1 mice receive intermittent ABT263 (to clear senescent glia) followed by two weeks of AICAR in drinking water. Endpoints include flow cytometry of brain‑isolated neurons for Annexin V, Iba1^+^ microglial phagocytic index (pHrodo‑labeled neuronal debris), SASP cytokine panels, and Morris water maze performance. Statistical power set at n=10 per sex/group to detect 20% changes with α=0.05.

If validated, this framework shifts therapeutic focus from broadly inhibiting neuronal death to reinstating the metabolic surveillance link between neurons and glia, offering a combinatorial strategy: senolytics to remove faulty surveillants, then metabolic agonists to restore their phagocytic competence.