Tau and Tauopathy

1. Tau Protein — MAPT, Chromosome 17, and Six Isoforms

Tau is a soluble, intrinsically disordered microtubule-associated protein (MAP) discovered in 1975 by Marc Kirschner and Mark Weingarten as a co-purifying factor required for microtubule assembly from porcine brain. It is encoded by a single gene, MAPT, located on chromosome 17q21.31, spanning ~134 kb and 16 exons.

Alternative splicing of exons 2, 3, and 10 generates six tau isoforms in the adult human CNS, ranging from 352 to 441 amino acids. The most consequential splicing event is exon 10:

• 3R tau — exon 10 excluded; three microtubule-binding repeats (R1, R3, R4).
• 4R tau — exon 10 included; four microtubule-binding repeats (R1, R2, R3, R4).

The healthy adult human cortex maintains an approximately 1:1 ratio of 3R:4R. Disturbance of this ratio is itself a tauopathy mechanism: FTDP-17 mutations near the exon 10 splice site shift the ratio toward 4R; Pick disease selectively aggregates 3R; progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) selectively aggregate 4R. AD — uniquely among major tauopathies — aggregates both 3R and 4R isoforms.

The longest isoform (2N4R, 441 residues, often called “htau40”) is the standard reference. Tau’s domain organisation is:

2. Tau in Health

In healthy neurons, tau is enriched in the distal axon at concentrations of ~2 µM, an order of magnitude greater than in the soma or dendrites. It performs several intertwined roles:

1. Microtubule stabilisation. Tau binds the outer surface of the microtubule along the protofilament ridge, contacting α/β-tubulin via its R1–R4 repeats. Each repeat docks one tubulin heterodimer; binding promotes longitudinal MT lattice integrity and slows catastrophe.
2. Axonal transport regulation. Tau modulates the run-length of kinesin-1 and dynein motors. At high local densities, tau acts as a soft roadblock that preferentially deflects kinesin-1 (anterograde), permitting selective cargo control.
3. Splicing and DNA protection. A nuclear pool of tau associates with pericentromeric heterochromatin and protects genomic DNA from oxidative damage; tau loss correlates with heterochromatin relaxation.
4. Signalling scaffold. The N-terminus binds Src-family kinases (Fyn) at the dendritic membrane and contributes to NMDA receptor positioning — a route by which tau participates in synaptic Aβ toxicity (Ittner & Götz, Cell 2010).

Strikingly, tau-knockout mice are viable and develop normally, suggesting redundancy with MAP1A/MAP1B. This provides a permissive ceiling for tau-lowering therapies: even substantial reduction of tau may be tolerated.

3. Tau Hyperphosphorylation in AD

Tau bears 85 potential phosphorylation sites (45 Ser, 35 Thr, 5 Tyr) on the longest isoform. Healthy tau is modestly phosphorylated (~2–3 mol Pi / mol tau). In AD brain tau is hyperphosphorylated to ~8 mol Pi / mol tau — a ~3- to 4-fold increase — that detaches tau from microtubules and seeds aggregation (Iqbal & Grundke-Iqbal, 1986+).

Kinases that phosphorylate tau

• GSK3β — glycogen synthase kinase 3 beta; the dominant proline-directed tau kinase. Phosphorylates Ser199, Ser202, Thr205, Thr231, Ser396, Ser404. Constitutively active; inhibited by Akt.
• CDK5 — cyclin-dependent kinase 5 with regulator p25 (cleaved from p35 by calpain in stressed neurons). Phosphorylates many of the same proline-directed sites.
• MARK1–4 — microtubule affinity-regulating kinases (PAR-1 family). Phosphorylate the KXGS motifs within the MTBR (Ser262, Ser293, Ser324, Ser356) — this releases tau from the microtubule.
• DYRK1A — encoded on chromosome 21; overexpressed in Down syndrome and links the trisomy-21 AD risk to tau hyperphosphorylation.
• PKA, CaMKII, casein kinase 1δ/ε, ERK1/2, JNK, p38 — secondary tau kinases.

Phosphatases that dephosphorylate tau

• PP2A (protein phosphatase 2A) — the dominant tau phosphatase, accounting for ~70% of tau dephosphorylation activity in human brain. PP2A activity falls ~50% in AD frontal cortex due to upregulation of inhibitors I1_PP2A and I2_PP2A (SET) — a major contributor to the imbalance.
• PP1, PP5, PP2B (calcineurin) — secondary contributors.

The pathogenic shift is not driven by any single kinase but by a kinase/phosphatase imbalance — with PP2A loss the largest single contributor. The relevant equation is simply:

In AD k_kinase rises (GSK3β, CDK5/p25, DYRK1A) and k_phos falls (PP2A inhibition), so steady-state [p-tau] climbs until tau detaches from microtubules and assembles into insoluble filaments.

4. Paired Helical Filaments and Neurofibrillary Tangles

The histological signature of tauopathy is the neurofibrillary tangle (NFT)— an intracellular bundle of fibrillar tau within the soma and dendrites of affected neurons. Under the electron microscope, the constituent fibrils show two morphologies:

• Paired helical filaments (PHFs) — the classical AD fibril; two strands twisting around each other with a crossover repeat of ~80 nm and a maximum diameter of ~22 nm.
• Straight filaments (SFs) — minority species; same fold but a different inter-protofilament packing.

Both were first described by Michael Kidd in 1963 (Nature 197:192) using negative-stain EM of AD cortex, eight years before tubulin had been purified and twelve years before tau was identified. Kidd’s drawings of the “paired helical filament” remain one of the founding observations of structural neuropathology.

Tangles evolve through three histological stages:

The structural core of every tauopathy filament is the microtubule-binding region (MTBR); the N-terminal projection and proline-rich region remain disordered and form a flexible “fuzzy coat” that limits cryo-EM resolution beyond the rigid core.

Hexapeptide nucleating motifs — PHF6 and PHF6*

David Eisenberg’s lab pinpointed two short steric-zipper motifs that drive tau aggregation: PHF6 = VQIVYK (residues 306–311 in repeat R3) and PHF6* = VQIINK (275–280 in R2, 4R-only). These sequences alone form fibrils in vitro; mutating VQIVYK abolishes fibril formation. They are the seeds of every tauopathy fold.

5. The Six Tau Folds — Cryo-EM and Disease Specificity

The single most consequential development in tau biology since the protein was identified came from the Goedert & Scheres groups at the LMB Cambridge between 2017 and 2021. Using cryo-electron microscopy at near-atomic resolution on ex vivo filaments extracted directly from human autopsy brain, they showed that each major tauopathy has a distinct, reproducible filament fold — and that these folds are diagnostic.

This was a paradigm shift. Until 2017 it was widely assumed all tau filaments shared a common cross-β architecture varying only by isoform. In fact, the same protein folds differently in different diseases, and the fold predicts the clinical phenotype. The framework now spans:

5.1 The Alzheimer fold — Fitzpatrick & Scheres, Nature 2017

The first ex vivo amyloid filament solved by cryo-EM. Anthony Fitzpatrick, Benjamin Falcon, and colleagues purified PHFs and SFs from the temporal cortex of an AD donor and reconstructed them at 3.4–3.5 Å resolution. The ordered core comprises residues V306–F378 — the second half of R3 through R4 and ~10 residues of the C-terminal flank — folded into a characteristic C-shape with eight β-sheets per protofilament. PHFs and SFs share an identical protofilament fold; they differ only in protofilament-protofilament interface (C2 vs C2¹).

5.2 The straight filament — same fold, different packing

The straight filament shares the AD protofilament fold but uses an alternate dimer interface (two protofilaments related by C2¹ rotational symmetry rather than C2), giving the filament a non-twisting, straight EM appearance. Both filament types coexist in every AD case; PHFs are far more abundant.

5.3 Why this matters — the “tau strain” concept

The fold is reproducible across patients: every confirmed AD case has the AD fold; every CTE case has the CTE fold; PSP and CBD differ despite both being 4R. This converts tau pathology from a single proteinopathy into a family of conformational strains — reminiscent of prion strains (PrP^Sc) where one primary sequence supports many self-templating folds with distinct disease phenotypes (Goedert, Spillantini, Hasegawa & Scheres, Annu Rev Biochem 2017+).

Practical consequences:

• Diagnosis from a fibril. A cryo-EM map of an autopsy filament now defines the tauopathy — arguably with more certainty than histology.
• Drug binding pockets are fold-specific. The AD fold contains a hydrophobic cavity that the CTE fold widens; the CBD fold has its own non-protein density. Tau-PET ligands (flortaucipir / AV-1451) bind the AD fold but barely the PSP/CBD folds — explaining why first-generation tau-PET is essentially an AD-tau tracer.
• Therapeutic antibodies must be conformation-aware. An antibody targeting the AD fold will not recognise PSP or CBD tau and vice versa.

6. Tau Spreading — Prion-Like Propagation

Heiko Braak’s 1991 staging map showed AD tau pathology spreading along a stereotyped trajectory: transentorhinal cortex → hippocampus → limbic cortex → temporal/frontal/parietal isocortex → primary sensory cortex. The trajectory exactly tracks synaptically connected anatomical circuits, not vascular territories. The mechanistic explanation came later.

Bess Frost, Marc Diamond and colleagues (J Biol Chem 2009; Nat Neurosci 2010+) showed that fibrillar tau, applied extracellularly, is taken up by neighbouring cells and seeds the misfolding of soluble endogenous tau — a templated conversion formally analogous to prion propagation. Karen Duff and colleagues then showed in transgenic mice that tau pathology, once seeded in the entorhinal cortex, spreads transsynaptically (de Calignon, Polydoro & Hyman, Neuron 2012; Liu & Duff, Neuron 2012).

Routes of cell-to-cell tau transfer

• Small extracellular vesicles (exosomes). Polanco, Götz and colleagues showed tau seeds packaged in CD81⁺/CD9⁺ exosomes propagate efficiently between neurons.
• Free release at the synapse. Activity-dependent release of soluble tau from presynaptic terminals; Holtzman and colleagues showed neuronal activity drives tau release into the ISF.
• Receptor-mediated uptake. Heparan sulphate proteoglycans (HSPGs), LRP1, and α/β/γ-LRP family members internalise extracellular tau into recipient neurons.
• Tunnelling nanotubes. Direct cytoplasmic continuity between neurons via F-actin protrusions (Tardivel & Neumann).

The seeding activity is fold-specific: AD-tau seeds template AD-tau folds in recipient cells; CTE-tau seeds template CTE folds (Hasegawa, Sanders & Diamond). Anti-tau antibodies that block extracellular tau or neutralise vesicle-bound seeds are therefore plausible disease-modifying agents (Section 10). MAPT antisense oligonucleotides (BIIB080) lower total tau and the seed pool simultaneously.

Mathematically, network-based diffusion models — treating tau spread as anisotropic diffusion along a connectome graph — reproduce the observed Braak trajectory remarkably well (Raj, Kuceyeski & Weiner, Neuron 2012; Vogel et al., Nat Med 2020). Tau-PET (flortaucipir, MK-6240) in vivo in living patients now confirms this trajectory and ties it to cognitive decline.

7. Why Tau Tracks Cognition Better Than Amyloid

A central tension in modern AD neurobiology is that amyloid initiates, but tau executes. Neuropathological and biomarker correlations consistently show:

• Plaque burden correlates only modestly with cognition (typical r ≈ 0.3–0.4); many cognitively normal elders have heavy plaque load (“non-demented with AD pathology”).
• NFT burden correlates strongly with cognitive impairment (r ≈ 0.6–0.8); Braak stage at autopsy predicts antemortem MMSE.
• Tau-PET signal in entorhinal/temporal cortex predicts subsequent cognitive decline more accurately than amyloid-PET (Ossenkoppele et al., JAMA Neurol 2016, 2020+).

Mechanistically this makes sense:

1. Spatial overlap with degeneration. NFT topography matches regional atrophy and FDG hypometabolism almost 1:1; amyloid topography (default-mode network) only partially overlaps.
2. Synapse loss. Tau accumulates in pre- and post-synaptic terminals before frank neuronal death; synaptic dysfunction is the proximal cause of symptoms (Spires-Jones & Hyman, Neuron 2014).
3. Network failure. Tau spread along functionally coupled networks degrades the network itself; default-mode connectivity collapses with tau, not amyloid.
4. Time course. Amyloid accumulation saturates ~5–10 years before symptoms; tau accumulation is ongoing during the symptomatic phase, so tau dose tracks symptom dose more directly.

The current synthesis (Hyman’s “amyloid-facilitated tauopathy”): amyloid creates a permissive environment in which tau pathology, once seeded, can escape the medial temporal lobe and propagate into the neocortex; tau, not amyloid, then drives neurodegeneration and clinical decline. This explains why anti-amyloid drugs slow but do not stop progression once tau is widespread, and why anti-tau strategies (Section 10) are increasingly the therapeutic frontier.

8. Tauopathies Beyond Alzheimer’s Disease

Tau aggregates drive a family of clinically distinct neurodegenerative syndromes that share fibrillar tau but differ in isoform, fold, cellular tropism, and anatomical predilection.

Three tauopathies worth a closer look

• FTDP-17. The 1998 demonstration that MAPT missense and splice-site mutations cause familial frontotemporal dementia with parkinsonism (Hutton, Lendon, Heutink; Spillantini, Goedert) was the proof that tau dysfunction alone— without amyloid — suffices to cause neurodegeneration. This was definitive evidence against a tau-only-as-bystander view.
• PSP and CBD. Pure 4R tauopathies with characteristic glial inclusions (tufted astrocytes in PSP, astrocytic plaques in CBD). Distinct cryo-EM folds (Section 5) finally explained why these clinically different diseases share isoform composition but produce different syndromes.
• CTE. Repetitive head-impact exposure (boxing, American football, military blast) produces a 3R+4R tauopathy with a perivascular distribution at the depths of cortical sulci — the consensus 2013/2021 NINDS criteria. The CTE fold (Falcon, Nature 2019) contains a hydrophobic cavity with a non-protein density and is structurally distinct from the AD fold.

9. Plasma p-tau Biomarkers — the Diagnostic Revolution

The single most important practical advance in AD diagnosis since amyloid-PET has been the maturation of blood-based phospho-tau assays. Between 2018 and 2024 the field progressed from research-grade CSF assays to clinically deployable plasma tests with diagnostic performance rivalling CSF and amyloid-PET.

9.1 The three plasma p-tau assays

9.2 Why p-tau217 is now dominant

p-tau217 outperforms p-tau181 on essentially every metric:

• Larger fold-change. ~7× higher in AD vs controls (vs ~2× for p-tau181).
• Higher AUC. 0.93–0.96 against amyloid-PET; comparable to CSF.
• Earlier rise. Detectable above threshold ~20 years before symptom onset in DIAN cohort (Barthélemy et al., Nat Med 2020).
• Correlates with both A and T. Reflects the brain’s integrated amyloid–tau response, not tau alone.
• Robust to pre-analytical variation (centrifugation, freeze-thaw); easier to deploy in primary care.

9.3 The %p-tau217 ratio

Mass-spectrometry assays (C2N PrecivityAD2; Washington University) report %p-tau217 = phosphorylated p-tau217 / non-phosphorylated tau217, a normalisation that controls for total tau and gives discriminative AUCs of 0.94–0.97 against amyloid-PET in memory clinic populations (Barthélemy & Bateman, Nat Med 2024). The combined Aβ42/40 + %p-tau217 score is now the highest-performing blood test for AD.

9.4 Clinical implications

For the first time, AD can be diagnosed by a venous blood draw with accuracy approaching CSF or PET. This collapses cost, broadens access, and enables population-scale screening — essential as anti-amyloid therapy (lecanemab, donanemab) becomes available and requires biomarker confirmation. The 2024 NIA-AA revised criteria explicitly endorse plasma p-tau217 (and its ratio) as a Core 1 biomarker for AD diagnosis. The full biomarker landscape, including CSF and PET, is covered in Part VI.

10. Anti-Tau Therapy

Tau lowering and tau immunotherapy have been pursued for two decades; the first generation of trials largely failed, but mechanistic clarity from cryo-EM and seeding biology has produced a much more rational second generation now in late-phase trials. The full therapeutic landscape is in Part VII.

10.1 Passive immunotherapy — anti-tau monoclonal antibodies

• Gosuranemab (BIIB092, Biogen). N-terminal tau (eTau). Phase 2 PASSPORT in PSP and TANGO in AD — both negative.
• Tilavonemab (ABBV-8E12, AbbVie). N-terminal. Phase 2 in PSP and AD — negative.
• Semorinemab (RO7105705, Genentech). N-terminal. TAURIEL (early AD) negative on CDR-SB; LAURIET (mild-moderate AD) showed a signal on ADAS-Cog11 but not co-primary.
• Zagotenemab (LY3303560, Lilly). MTBR-directed, conformational. Phase 2 TRAILRUNNER-ALZ-1 negative.
• Bepranemab (UCB0107). Mid-region (residues 235–250). Phase 2 TOGETHER readout 2024–25 showed modest tau-PET slowing.
• E2814 (Eisai). MTBR seed-blocking antibody. In DIAN-TU primary prevention trial in autosomal-dominant AD.
• JNJ-63733657 (posdinemab) (Janssen). Mid-region p-tau217. Phase 2 AuTonomy in early AD; ongoing.

The pattern across the first generation: N-terminal-targeting antibodies, which engage extracellular tau but do not engage the fibril core, have uniformly failed. Newer antibodies targeting the MTBR (the actual fibril core revealed by cryo-EM) and seed-competent tau species are still being evaluated.

10.2 Active immunisation

AADvac1 (Axon Neuroscience) — a vaccine raising antibodies to a misfolded tau epitope. ADAMANT phase 2 in mild AD (Novak et al., Nat Aging 2021) showed neurofilament light reduction; biological signal without convincing clinical benefit. ACI-35 (AC Immune) — phospho-tau liposomal vaccine; phase 1b/2a.

10.3 Antisense oligonucleotides — lower MAPT itself

BIIB080 (IONIS-MAPTRx, tau-ASO) — an intrathecally delivered antisense oligonucleotide targeting the MAPT 3¹ UTR, reducing total tau mRNA and protein. Phase 1b (Mummery et al., Nat Med 2023) demonstrated dose-dependent CSF total-tau and p-tau reduction of ~30–50% over 24 weeks — the first direct demonstration that human CNS tau is druggable. Phase 2 CELIA trial (NCT05399888) in mild AD with cognitive endpoints; readout 2026. Conceptually parallel to tofersen for SOD1-ALS and tominersen for HD.

10.4 Microtubule stabilisers

The compensatory hypothesis: if hyperphosphorylated tau detaches from microtubules, supplement with brain-penetrant taxane-like stabilisers. Davunetide (NAP) — an octapeptide microtubule stabiliser; phase 3 in PSP (Boxer et al., Lancet Neurol 2014) negative. Epothilone D (BMS-241027) — brain-penetrant taxane; phase 1 only. TPI-287 — abeotaxane; phase 1 in AD/CBD/PSP, mixed tolerability.

10.5 Kinase / phosphatase modulators

• GSK3β inhibitors — Tideglusib (NP12) showed no benefit in PSP (TAUROS) or AD (ARGO); narrow therapeutic window because GSK3β has many substrates beyond tau (β-catenin, glycogen synthase, NFAT).
• CDK5/p25 inhibitors — preclinical; selectively targeting p25-bound CDK5 to spare physiological p35-CDK5 is the design challenge.
• PP2A activators / methylation enhancers — sodium selenate, LB-100, LMTM (the failed methylene blue derivative). LMTM (TauRx) reported ambiguous phase 3 readouts; the LUCIDITY trial (2023) was conducted at lower doses to avoid the methemoglobinemia signal.

10.6 Tau aggregation inhibitors

Methylene blue / LMTM (hydromethylthionine mesylate) — the longest-running tau aggregation inhibitor programme (Wischik, TauRx). TRx-237-005 and TRx-237-015 phase 3 trials negative on co-primary; LUCIDITY (2023) ongoing. Mechanistic concern: the compound is also a redox cycler.

10.7 Where the field is heading

• Combination amyloid + tau therapy. The hypothesis: lower amyloid early, then lower tau. DIAN-TU is testing this in autosomal-dominant AD with anti-amyloid (gantenerumab/lecanemab) plus E2814 anti-tau.
• Fold-specific antibodies. Cryo-EM has revealed fold-specific surfaces; conformational antibodies that engage only AD-fold tau (and not healthy soluble tau) should be more selective and tolerable.
• Tau-targeted PROTACs and molecular glues. Heterobifunctional degraders that recruit E3 ligase to phospho-tau or fibrillar tau (Crews lab and others).
• Genetic medicines. CRISPR-based MAPT lowering and AAV-delivered anti-tau intrabodies are in preclinical development.