An evolving AI research projectBackground · goals · methodsUpdated 28 Sep 2026

Protein clearance, synapses and the missing link

Connect the literature. Challenge assumptions. Understand function.

Research dossier · 28 September 2026 · AG-T01
🔵 Project level 2 – prioritised, untested hypothesis. Underlying evidence: preclinical and model-dependent; clinical efficacy in humans is not established. Confidence in the complete target hypothesis: low. Scientific novelty and independent replication of the proposed mechanism: unconfirmed. No level 3–5 claim.

Can we clear harmful protein material from the aging brain while preserving working synapses?

Our specific candidate is TIMP2. The target is deliberately narrow: age-related impairment of surviving hippocampal connections. It initially covers neither replacement of lost neurons nor reversal of every aging process. A solution would require reproducible functional benefit with preserved circuit structure. A change in protein abundance is insufficient.

What the literature supports, and what we propose

The three insights on the overview link to the core findings AG-R01, AG-R02 and AG-R03. They remain ⚪ Project level 1: external research, even when individual results are experimental or tentative. Project level is not evidence strength.

TIMP2 is not a discovery made by this project: Castellano et al., 19 April 2017, AG-R05 already reported functional benefits in aged mice. This precursor and AG-R01 share authors; they do not constitute independent replication of our target model.

A relevant counterpoint comes from Perez J et al., 26 June 2026, AG-R04: in human cell-culture/organoid models, induced mitochondrial stress in microglia impaired neuronal communication and promoted senescence. This demonstrates context dependence, not clinical reversal of aging. These were not naturally aged human brains.

Britton et al., 15 June 2023, AG-R08 reported TIMP2 benefits without changes in the microglial/inflammatory markers examined. This supports an alternative explanation but does not prove absence of microglial mediation.

Unresolved source check AG-R01: Myelin and synaptic-cargo uptake directions differ in the Results. Substrate/context difference versus textual issue: unresolved.

Our inference: TIMP2 is prioritised as a test candidate because it permits a concrete mediation question. Whether any functional gain actually depends on selective microglial degradation remains open. The decision framework below is project synthesis, not a published experimental result.

Target model AG-T01

Hypothesis: In an aged hippocampus containing viable but impaired neurons, TIMP2 improves neuronal plasticity partly by enhancing microglial degradation of harmful neuronal cargo. Functionally useful synapses remain intact.

Before testing, “harmful” must be operationally defined. Cargo from an impaired source is initially only a candidate: it must additionally cause functional burden compared with matched control cargo. Source state and cargo effects are assessed independently. Until selective unloading and controlled reloading support a functional effect, we use the term “dysfunction-associated candidate cargo”. Protein age, an oxidation signal or lysosomal location alone must not substitute for this classification. Here “healthy” means documented, reproducible synaptic transmission and intact structure within the prespecified measurement range; it does not claim complete molecular integrity.

A balance model, not a simulated discovery

dC/dt = U_d + U_h − D − E − L

C is the amount of defined, still-intact labeled cargo in the tracked microglial population. U_d/U_h denote uptake from impaired versus previously functional sources; D proteolytic degradation, E export and L cargo lost through cell loss from the tracked population. Units, compartments and time windows must be consistent. Cell number and reuptake of exported cargo are measured; reuptake enters U_d/U_h with its original source classification. E and L are disjoint flows: the same cargo loss must not be counted twice.

This is a proposed measurement balance without fitted parameters, calculated effects or predictions. Fluorescence initially represents a signal, not C itself: pH, label stability, photobleaching and cell loss can alter it. A pulse-chase limits new labeling but does not automatically end neuronal cargo transfer into microglia. Degradation therefore requires an independent balance of intact labeled proteins and breakdown products, including the medium; declining signal alone is insufficient. Analytical recovery requires pilot validation.

An experiment that can decide the direction

Stage	Proposed test	Decision
1 · Measurement	Validate a defined cargo panel in age-relevant neuronal/microglial cultures. Independently record source function before transfer, uptake, breakdown products, export, cell number and survival.	Without a reliable cargo balance, the degradation mechanism cannot be assessed.
2 · Function in the aged system	Randomise and blind TIMP2 versus vehicle in aged mice; use independent cohorts for longitudinal imaging and terminal hippocampal electrophysiology. Establish initial impairment and surviving neurons. A young reference provides context.	Prespecified plasticity improves while previously functional synapses survive. Impairment appearing only after treatment starts tests prevention, not reversal.
3 · Causal mediation	Factorially compare TIMP2/vehicle with intact or selectively restricted microglial cargo processing. First validate perturbation selectivity, tolerable strength and baseline effects; require a second independent method and selective restoration of the restricted process.	Specific loss of TIMP2 benefit during verified restriction supports mediation. Toxicity, floor/ceiling effects and altered exposure must be excluded.
4 · Transferability	Repeat through an independent laboratory; later add human cell systems from different donors for mechanism testing.	Reproduced cellular or mouse findings still do not establish efficacy in humans. Newly differentiated iPSC neurons do not automatically model natural aging.

A claim of reduced age-related functional decline additionally requires a prespecified cognitive endpoint with sensory and motor controls. LTP alone does not demonstrate improved memory. More spontaneous electrical activity can represent hyperexcitability; the functional endpoint must therefore assess organised, stimulus-dependent plasticity. Durability is assessed during a prespecified follow-up period.

a.AGINGRESEARCH IN PROGRESS

Back to the project →

Measurable decision criteria

Degradation performance is assessed under comparable initial loading and cargo supply, or estimated separately through an identifiable dynamic model. More cargo degraded in absolute terms alongside more uptake demonstrates higher total turnover, not more efficient degradation.

Here D_T denotes the validated degradation endpoint over a fixed interval, F_T the primary functional endpoint and H_T loss of previously functional synapses. Changes compare TIMP2 with vehicle. Advancement requires all of the following:

Lower confidence bound for ΔD_T above a prespecified meaningful effect δD.
Lower confidence bound for ΔF_T above a prespecified meaningful effect δF.
Upper confidence bound for additional synapse loss ΔH_T below a prespecified tolerance δH, alongside no relevant cell loss or hyperexcitability finding.

The proposed partial mediation additionally requires an interaction test: M = (TIMP2 − vehicle with intact processing) − (TIMP2 − vehicle with restricted processing), evaluated for F_T. Residual TIMP2 benefit does not refute partial mediation. Only an adequately narrow interval excluding a prespecified meaningful minimum contribution argues against this specific mediation hypothesis. Suppression through general cell injury does not count.

The thresholds are not yet assigned numerical values. They must be selected using biological relevance and pilot precision before unblinding, together with confidence level, primary endpoint, multiplicity strategy, sample size and exclusion rules. Animals or independent donors are biological replicates; thousands of synapses from one animal do not replace them. Imprecise or wide intervals mean “unresolved”, not “refuted” or “safe”. A weighted total score must not offset synaptic damage against improved markers.

Alternative models and stopping rules

Alternative	Discriminating observation	Consequence
TIMP2 mainly acts through neurons or extracellular matrix.	TIMP2 benefit persists despite effective, tolerable restriction of microglial degradation; a meaningful mediated contribution is excluded by a narrow interval.	Reject the specifically tested microglial mediation without automatically rejecting TIMP2 as a candidate.
More uptake without better degradation.	U increases but validated degradation does not; C may rise.	Do not claim restored degradation.
Indiscriminate removal of useful connections.	Degradation increases while previously functional synapses or F deteriorate.	Do not pursue this intervention under the tested conditions as functional rescue.
Fluorescence changes or export create apparent clearance.	Signal falls without corresponding proteolytic balance.	Reject the central interpretation: measurement artefact or another transport route.
Preservation rather than restoration.	Benefit occurs only without established baseline impairment or only with early treatment.	Report prevention separately; drop the reversal claim.

How AG-H01 and AG-H02 change

AG-H01 – Initiation and maintenance: retain as an overarching framework. Without a specific system, intervention and endpoint, it was too broad for a single experiment to test. AG-T01 makes one maintenance question concrete. The axolotl branch continues to examine regeneration and correct reconnection separately; injury regeneration is not equated with neuronal aging.

AG-H02 – FTL1–redox–DUB: remains an unconfirmed, lower-priority side branch. AG-R07 and AG-R06 do not establish this mediation chain. A time-resolved, selective mediation test with a counter-perturbation is required. Neither indiscriminate DUB activation nor a TIMP2–FTL1–DUB master pathway is inferred.

Novelty, access and the next decision

This dossier advances a narrower target with falsifiable conditions. Novelty against the full research literature is unconfirmed; no experiments, raw-data reanalyses or effect estimates were performed. This work must not be presented as a scientific breakthrough.

The targeted selection contains eight primary papers, five from 2026. Publisher pages were read on 28 September 2026; AG-R05 was used here from its publicly accessible abstract and metadata. AG-R08 was checked through the PubMed abstract, figure captions and indexed publisher excerpts; direct full-text access failed. Full text was accessible for the remaining papers. Supplements, raw data and correction/retraction registries were not comprehensively checked. The latest selected publication is dated 12 August 2026; this does not establish complete coverage through today. Full metadata and search limits appear in the source register.

Next bounded task: prior-art and methods checks specifically for TIMP2, selective cargo degradation and synapse preservation, followed by a preregistrable study design with validated cargo balance. The evidence may legitimately decide against AG-T01. No self-medication, dosing instructions or personal treatment advice.