Hypothesis

In aged brains, localized presynaptic ATP deficiency acts as a permissive signal that enables C1q tagging of synapses, converting metabolic stress into complement-driven elimination[1]. When presynaptic mitochondria fail to maintain ATP above a critical threshold, phosphatidylserine is externalized and phosphatidic acid accumulates, providing a lipid ligand that stabilizes C1q binding independent of classical pathway activation[2]. Simultaneously, the fall in extracellular adenosine (generated from ATP breakdown) reduces tonic inhibition of microglial A2A receptors, lifting a brake on phagocytosis[4]. Thus, the brain couples an energy-sensing checkpoint to the developmental pruning cascade, removing synapses that are too costly to sustain.

Mechanistic Model

1. Presynaptic ATP sensor – Declining oxidative phosphorylation lowers intravesicular ATP, leading to impaired vesicle cycling and increased cytosolic Ca2+.
2. Lipid remodeling – Ca2-dependent activation of phospholipase D converts phosphatidylcholine to phosphatidic acid; calcium-dependent scramblases expose phosphatidylserine on the outer leaflet.
3. C1q docking – Phosphatidic acid and exposed phosphatidylserine create a high-affinity docking platform for the globular heads of C1q, allowing C1q to bind without requiring IgM or IgG antibodies[1].
4. Microglial gating – Extracellular ATP is hydrolyzed to adenosine by CD39/CD73; adenosine engages microglial A2A receptors, sustaining high cAMP that suppresses CR3-mediated engulfment[4]. As ATP falls, adenosine declines, disinhibiting microglia and permitting C3b-opsonized synapses to be phagocytosed[3].
5. Feedback loop – Removed synapses reduce local excitatory drive, further lowering presynaptic activity and ATP consumption, sharpening the elimination of the least efficient connections.

Predictions and Experiments

• Prediction 1: In aged mouse hippocampus, synaptosomal ATP levels will correlate inversely with C1q intensity; regions with low ATP will show highest C1b deposition[2]. Test: Measure ATP using a luciferase-based assay on isolated synaptosomes from young vs. old mice; perform immunoblot for C1q and quantify correlation.
• Prediction 2: Boosting presynaptic ATP with nicotinamide riboside (NR) will reduce C1q tagging and preserve synaptic density without affecting neuronal counts. Test: Treat aged mice with NR for 8 weeks, assess synaptosomal ATP, C1b staining, synapse number (via synaptophysin/PSD-95), and NeuN+ cell counts.
• Prediction 3: Pharmacological blockade of microglial A2A receptors will exacerbate complement-mediated pruning even when ATP is normal[4]. Test: Administer an A2A antagonist (e.g., SCH-58261) to young mice, then challenge with low-dose Aβ oligomers; evaluate C1q deposition and microglial phagocytosis ex vivo.
• Prediction 4: Genetic ablation of presynaptic phospholipase D (PLD1) will diminish phosphatidic acid accumulation and render synapses resistant to C1q binding despite low ATP[2]. Test: Use conditional PLD1 knockout in excitatory neurons; repeat aging cohort analyses for C1q tagging and synapse loss.

If any of these predictions fail—e.g., NR does not lower C1q tagging, or A2A blockade does not increase pruning—the hypothesis would be falsified, indicating that ATP-sensing lipids or adenosine gating are not central to the complement-driven elimination process[3].