Hypothesis

Age-related cognitive decline stems not from a loss of plasticity per se but from excessive microglial‑mediated synaptic pruning that reduces neuronal noise, making brain circuits overly confident in existing predictions. Restoring a modest level of synaptic uncertainty with a low‑dose NMDA receptor antagonist (e.g., memantine) should re‑engage exploratory updating without causing further synapse loss.

Mechanistic Rationale

Microglial synaptic pruning is driven by elevated C1q "eat me" signals and reduced CD47 "don't eat me" protection, leading to elimination of functional synapses 1 2. This pruning removes sources of spontaneous firing variability, decreasing the signal‑to‑noise ratio that Bayesian models rely on for surprise detection. Lower NMDA receptor GluN2B expression and function further diminish calcium‑dependent plasticity mechanisms 3 4, while paradoxically higher residual NMDA activity correlates with poorer performance, indicating dysfunctional receptors rather than excess stabilization. Introducing a low‑dose NMDA antagonist increases neuronal firing variance by partially blocking excitatory transmission, thereby reinstating the exploration‑exploitation balance needed for model updating. This manipulation mimics the effect of controlled environmental uncertainty, encouraging the brain to weaken over‑consolidated connections and strengthen adaptive ones without triggering additional phagocytosis.

Experimental Design

Use aged (20‑month) C1q‑overexpressing mice that exhibit heightened microglial pruning 1. Treat one cohort with low‑dose memantine (0.1 mg/kg/day, sub‑sedentary) for 8 weeks; a control cohort receives vehicle. Assess:

• Synapse density via immunostaining for presynaptic (Synaptophysin) and postsynaptic (PSD-95) markers in hippocampus and prefrontal cortex.
• Microglial phagocytic activity using Iba1 and CD68 co‑labeling and C1q/CD47 ratio.
• NMDA receptor subunit composition (GluN2B vs GluN2A) by Western blot.
• In vivo two‑photon imaging of dendritic spine turnover.
• Behavioral flexibility measured in a reversal learning water maze.
• Electrophysiological LTP magnitude and neuronal noise (coefficient of variation of spontaneous firing) in hippocampal slices.

Predictions and Falsifiability

If the hypothesis is correct, memantine‑treated aged mice will show: (1) preserved or modestly increased synapse density relative to vehicle‑treated aged controls; (2) reduced microglial phagocytic markers and normalized C1q/CD47 ratio; (3) restored GluN2B/GluN2A ratio toward youthful levels; (4) increased spine turnover and LTP magnitude; (5) improved reversal learning performance without impairing baseline memory. Conversely, if memantine fails to improve any of these metrics or exacerbates synapse loss and cognitive deficits, the hypothesis that controlled synaptic noise rescues over‑pruning‑induced rigidity is falsified.