Tau pathway mapping and therapeutic audit

What this research found

Tau does two opposite jobs in the brain: a transient phosphorylation at Thr-205 helps consolidate long-term memories, while multi-site hyper-phosphorylation drives the tangles of Alzheimer's disease. This audit maps where those two fates diverge and then screens the clinical-stage anti-tau pipeline for drugs whose mechanism cannot tell them apart, flagging three of ten programmes as carrying the clearest theoretical risk to memory consolidation. It is a mechanism-and-pipeline review rather than a new experiment, and its verdict is that a well-designed anti-tau drug removes disease tau selectively instead of depleting all tau.

• The split between tau's useful and harmful roles is set by which sites get phosphorylated. A learning signal activates p38-gamma kinase, producing transient phospho-tau at Thr-205 that tunes hippocampal engram cells, whereas disease kinases such as GSK-3-beta and CDK5 drive hyper-phosphorylation at sites including pS396 and pS404, detaching tau into oligomers, filaments and a spreading seeding core.
• Three of ten clinical-stage programmes were flagged as highest risk because their mechanism cannot distinguish physiological from disease tau: the antisense oligonucleotide BIIB080 (diranersen), which silences MAPT messenger RNA and lowers cerebrospinal-fluid total tau by 50 to 65%, and the N-terminal antibodies gosuranemab and semorinemab, which bind all six tau isoforms regardless of phospho-state.
• Five programmes were rated low risk because they engage only the pathological form: bepranemab, which blocks cell-to-cell seed spreading; E2814 (etalanetug), which neutralises seeds and lowers eMTBR-tau243 by about 89%; the conformational antibody zagotenemab, which spares monomer; the AADvac1 vaccine; and the aggregation inhibitor HMTM/LMTM. Tilavonemab and the ACI-35 vaccine sit in between at moderate risk.
• The concern is time-asymmetric. Any deficit would fall on forming new long-term memories rather than on losing existing ones, and no trial has reported a memory-worsening safety signal to date.
• Five experiments are proposed to settle the question: whether a total-tau drug spares the physiological Thr-205 pool in cerebrospinal fluid, whether animals on each drug still form new remote memories, whether engram sparsity and recruitment are disturbed, what depth of hippocampal tau-lowering is efficacious short of the roughly 68% that impairs memory in mice, and how delayed recall tracks lowering in trial arms.

How it was done

The work is a mechanism map paired with a pipeline audit rather than an original analysis. It first traced the two divergent fates of native tau from the MAPT gene — the physiological learning-induced Thr-205 phosphorylation route through hippocampal engram cells to memory consolidation, and the disease route through multi-site hyper-phosphorylation, misfolding, oligomerisation and tangle seeding to synapse failure. Each clinical-stage anti-tau therapy was then classified by what its mechanism actually binds or suppresses, from broad targets such as total tau and pan-tau epitopes through to narrow ones such as disease conformers and aggregates, and assigned a memory-risk flag on that basis alongside its development stage. The output is a single-page decision brief for clinicians and informed patients, with a plain-language verdict and a list of confirmatory experiments.

Data sources

• Nature Communications (2026) — tau Thr-205 phosphorylation by p38-gamma in hippocampal engram cells is required for remote-memory consolidation
• ALZFORUM therapeutics database and company disclosures, 2024–2026 — ten clinical-stage anti-tau programmes