Hypothesis

Aging brains become over‑confident in their internal models because prediction precision is excessively high. This state is maintained by two synergistic processes: (1) complement‑mediated synaptic pruning that removes weakly active synapses, and (2) perineuronal net (PNN) deposition that stabilizes the remaining connections. Together they raise the metaplastic threshold, making it harder to update stored predictions. We propose that brief, periodic reductions in NMDA‑receptor‑mediated calcium influx—achieved with low‑dose memantine—combined with enzymatic digestion of PNN chondroitin sulfate chains (chondroitinase ABC) will transiently lower prediction precision, re‑open plasticity windows, and improve cognitive flexibility without erasing consolidated knowledge.

Mechanistic Basis

Microglial C1q up‑regulation and CD47 down‑shift tip the phagocytic balance toward excess synapse elimination [2]. The pruned synapses are replaced by a network enriched in GluN2A‑containing NMDA receptors, which have faster kinetics and lower calcium permeability, thereby reducing the signal needed for long‑term potentiation [8]. Simultaneously, astrocytic TGF‑β released in response to microglial inflammation up‑regulates chondroitin sulfate proteoglycans, increasing PNN density 67‑117% from young to old age [3][4]. PNNs physically impede dendritic spine motility and lock in GluN2A dominance, raising the threshold for synaptic weight changes.

Low‑dose memantine preferentially blocks extrasynaptic NMDA receptors, lowering tonic calcium influx and shifting the NMDA receptor subunit balance back toward GluN2B, which supports plasticity [8]. Chondroitinase ABC degrades the glycosaminoglycan side chains of PNNs, reducing their inhibitory barrier and allowing spine remodeling [3]. When applied intermittently (e.g., 5 min every 48 h), these interventions create brief windows of reduced prediction precision, permitting the brain to incorporate surprise without destabilizing long‑term stores.

Testable Predictions

1. Aged mice receiving intermittent low‑dose memantine + chondroitinase ABC will show a significant increase in GluN2B/GluN2A ratio in hippocampal synaptosomes compared with vehicle‑treated aged controls.
2. PNN density (measured by Wisteria floribunda agglutinin staining) will decrease by ~30 % in treated aged mice without affecting young‑mouse baselines.
3. In a reversal‑learning task (e.g., water‑maze platform shift), treated aged mice will commit fewer perseverative errors and acquire the new location faster than untreated aged peers, performing indistinguishably from young adults.
4. Baseline spatial memory (reference memory) assessed before intervention will remain unchanged, indicating that consolidated knowledge is preserved.
5. Pharmacological blockade of GluN2B (with ifenprodil) during the intervention window will abolish the cognitive benefits, confirming the mechanistic role of NMDA subunit shift.

Experimental Design

• Subjects: 24‑month‑old C57BL/6J mice (n=12 per group) and 3‑month‑old young controls (n=6).
• Groups: (A) vehicle, (B) low‑dose memantine (0.5 mg/kg i.p.), (C) chondroitinase ABC (0.1 U/µl intracerebroventricular), (D) combined intermittent treatment (memantine + chondroitinase ABC) administered 5 min every 48 h for 4 weeks.
• Outcome measures: biochemical (Western blot for GluN2A/B, ELISA for C1q, immunohistochemistry for PNN), behavioral (Morris water maze acquisition, reversal, probe), and electrophysiological (LTP magnitude in hippocampal slices).
• Statistical plan: Two‑way ANOVA (age × treatment) with post‑hoc Tukey; significance set at p<0.05.