Hypothesis

SIRT6 activity establishes a metabolic threshold that, when exceeded by sustained energetic stress, switches the brain from passive degeneration to an active, activity‑dependent culling program that removes weakly firing cortical neurons. This culling relies on microglial phagocytosis of stressed somata after synaptic stripping, a process distinct from developmental pruning but sharing downstream effectors such as complement C1q and cathepsin S. Thus, normal aging shows minimal neuronal loss because SIRT6 keeps stress below the threshold; only when SIRT6 function declines or oxidative load rises does the checkpoint fire, producing the focal neuron loss seen in early Alzheimer’s disease.

Mechanistic basis

• SIRT6–SIRT3–Nrf2 axis maintains mitochondrial ROS low and preserves NAD+ levels, buffering oxidative damage (5).
• When SIRT6 wanes, Nrf2 activity drops, mitochondrial ROS rises, and proteostasis falters (3).
• Elevated ROS triggers neuronal expression of "eat-me" signals (phosphatidylserine exposure, complement C1q deposition) that tag weakly active synapses for microglial removal (2).
• Persistent synaptic stripping leads to calcium overload and calpain activation, producing a caspase‑independent somatic death pathway that requires cathepsin S release from microglia (4).
• The overall effect mirrors developmental pruning logic—expensive, low‑output neurons are eliminated—but the trigger is metabolic insufficiency rather than a genetic program.

Testable predictions

1. In aged mice with neuronal‑specific SIRT6 over‑expression, cortical neuron counts will remain stable despite elevated oxidative markers, whereas SIRT6 knock‑out will show accelerated loss of low‑firing Layer II/III pyramidal cells measured by in‑vivo two‑photon imaging.
2. Blocking complement C1q with antibodies will prevent the neuron loss in SIRT6‑deficient mice without affecting ROS levels, indicating that the active culling step is complement dependent.
3. Pharmacological elevation of NAD+ (e.g., NR supplementation) will raise the SIRT6 threshold and rescue neuronal numbers only when administered before ROS crosses a definable concentration (≈2‑fold baseline), providing a quantitative checkpoint.
4. Single‑cell transcriptomics of aging human cortex will reveal a subpopulation of neurons expressing high SIRT6 targets and low activity genes (Arc, Fos) that are simultaneously enriched for phagocytic markers (CD68, LAMP1) in adjacent microglia, a signature absent in young tissue.

Falsifiability

If neuronal loss in aging proceeds independently of SIRT6 levels, complement activation, or synaptic activity (i.e., loss occurs uniformly across all neuronal classes with no correlation to firing rates or metabolic stress), then the proposed checkpoint model is refuted. Conversely, observing a tight correlation between SIRT6 deficiency, focal complement deposition, and selective disappearance of low‑activity neurons would support the hypothesis.