Skip to content

Tau biology and spread

TL;DR — Tau is the pathology that tracks the syndrome. Braak's examination of 83 brains found that amyloid distribution and packing density were of limited staging value, whereas neurofibrillary tangles and neuropil threads followed a stereotyped transentorhinal → limbic → isocortical sequence that defines six stages (Braak 1991, PMID 1759558), and clinicopathological synthesis concludes that cognitive severity correlates best with neocortical tangle burden (Nelson 2012, PMID 22487856). Quantitative work makes this sharper: among 61 people all at Braak stage V, neocortical p-tau burden varied from 0.2% to 53.7%, and only p-tau burden and microinfarcts independently predicted cognitive decline — not Aβ, LATE-NC, Lewy bodies or other cerebrovascular measures (Richardson 2026, PMID 42184025). The stereotyped sequence is nevertheless an average: tau-PET in 1,612 individuals resolved four spatiotemporal trajectories with prevalences of 18–33%, stable longitudinally and replicated with a second tracer (Vogel 2021, PMID 33927414). Tau-PET topography predicts where atrophy will occur at the single-patient level over 15 months, whereas amyloid-PET does not (La Joie 2020, PMID 31894103). Therapeutically, tau has so far delivered target engagement without clinical benefit: the antisense oligonucleotide BIIB080 reduced CSF t-tau by 56% (95% CI 50–62) and lowered tau-PET signal, while four anti-tau monoclonal antibodies produced no convincing clinical effect in a network meta-analysis of six RCTs (Edwards 2023, PMID 37902726; Cai 2024, PMID 39945003).

What tau is and what goes wrong

Tau is a microtubule-associated protein encoded by MAPT, expressed in the adult human CNS as six isoforms differing in the number of N-terminal inserts and in whether three or four microtubule-binding repeats are present (3R/4R). In Alzheimer's disease both 3R and 4R tau aggregate — a feature that distinguishes AD from the 4R tauopathies (progressive supranuclear palsy, corticobasal degeneration) and 3R tauopathy (Pick disease). Aggregation is preceded by hyperphosphorylation, which reduces microtubule affinity and increases the free cytoplasmic pool available for self-assembly.

Cryo-EM of filaments extracted from an AD brain resolved paired helical and straight filaments at 3.4–3.5 Å: both are built from two identical protofilaments comprising tau residues 306–378 in a combined cross-β/β-helix fold, and the two filament types differ only in inter-protofilament packing — they are ultrastructural polymorphs of the same fold (Fitzpatrick 2017, PMID 28678775). The practical consequence is that "the AD tau fold" is a defined structural entity, which is why tau-PET ligands and antibodies raised against one conformation may not transfer across tauopathies.

Tau is sufficient to cause neurodegeneration, but MAPT mutations do not cause AD

MAPT mutations cause frontotemporal dementia with parkinsonism linked to chromosome 17 — the clearest evidence that tauopathy is causal in neurodegeneration rather than merely a marker. Hutton sequenced MAPT in FTDP-17 families and found three missense mutations (G272V, P301L, R406W) and three exon-10 5'-splice-site mutations that destabilise a stem-loop regulating alternative splicing of the microtubule-binding repeat, shifting the 3R/4R ratio (Hutton 1998, PMID 9641683; Strang 2019, PMID 30742061). But they cause FTD, not Alzheimer's disease. This is a load-bearing asymmetry for the amyloid cascade: APP and presenilin mutations produce full AD with tangles; MAPT mutations produce tauopathy without AD-type amyloid. Tau is downstream of amyloid in the disease AD, and independently sufficient for a different disease.

Braak staging

Stage Distribution Typical clinical correlate
I–II (transentorhinal) Mild or severe alteration of transentorhinal layer Pre-α Usually clinically silent
III–IV (limbic) Pre-α affected in transentorhinal and entorhinal cortex; mild CA1 involvement Incipient to mild impairment
V–VI (isocortical) Destruction of virtually all isocortical association areas Dementia in most, but not all

Braak's own finding — that amyloid distribution and density were "of limited significance for differentiation of neuropathological stages" while tangles and neuropil threads permitted six-stage resolution, and that only a few key preparations were needed to assign a stage (PMID 1759558) — is the reason tau, not amyloid, became the pathological staging axis. Compare Thal's five amyloid phases, which also follow a hierarchy but map less tightly onto symptoms (Thal 2002, PMID 12084879).

Stage is not enough: density matters too

Assigning Braak stage V does not fix the clinical picture. Quantitative pixel assessment of middle frontal and superior temporal gyri in 61 Braak-V individuals found neocortical p-tau burden ranging from 0.2% to 53.7%. Frontal and temporal burden each correlated with performance in language/semantic memory and attention/working memory. In multivariable analysis only p-tau burden and microinfarcts significantly affected cognitive decline; Aβ, LATE-NC, Lewy body pathology and other cerebrovascular measures did not. People with low mean neocortical burden (≤13%) had significantly better cognitive trajectories over their final 15 years than those with high burden (≥23.5%) (Richardson 2026, PMID 42184025). The authors' conclusion — that density should be incorporated alongside regional location in future staging systems — is a concrete, testable revision to a 35-year-old scheme.

Tau topography explains clinical heterogeneity

Tau-PET made the Braak logic measurable in life, and immediately showed that topography tracks phenotype in a way amyloid does not. In 20 patients with varied phenotypes plus 15 amyloid-negative controls (Ossenkoppele 2016, PMID 26962052):

  • In posterior cortical atrophy (n=7), ¹⁸F-AV-1451 and FDG-PET abnormalities targeted posterior regions specifically, while ¹¹C-PiB bound diffusely across neocortex.
  • Amnestic-predominant patients (n=5) showed highest tau retention in medial temporal and lateral temporoparietal regions.
  • Three of five logopenic-variant patients showed asymmetric left > right tau uptake.
  • Across 30 regions in 16 patients with all three scans, tau and FDG were strongly negatively associated (r = −0.49 ± 0.07, P<0.001), while PiB–FDG (r = 0.16 ± 0.09) and tau–PiB (r = 0.18 ± 0.09) associations were weak.
  • Younger age was associated with greater neocortical tau; older age with more medial temporal tau; APOE ε4 carriers had greater temporal and parietal uptake.
  • Worse domain-specific test performance mapped onto tau in the corresponding region (memory–medial temporal; visuospatial–occipital/right temporoparietal; language–left > right temporoparietal).

This is the imaging counterpart of the autopsy subtypes: hippocampal-sparing (11%), typical (75%) and limbic-predominant (14%) AD, with hippocampal-sparing cases younger and more often male, limbic-predominant older and more often female (Murray 2011, PMID 21802369).

Four trajectories, not one

Applying data-driven modelling to tau-PET from 1,612 individuals identified four distinct spatiotemporal trajectories with prevalences of 18–33%: limbic-predominant and medial-temporal-sparing patterns (replicating the autopsy subtypes) plus posterior and lateral-temporal patterns resembling atypical clinical variants. Subtypes were stable over longitudinal follow-up, replicated in an independent sample with a different radiotracer, differed in demographics, cognition and outcome, and — from network diffusion modelling — appeared to originate and spread through distinct corticolimbic networks (Vogel 2021, PMID 33927414). The authors' framing is deliberately provocative: this "perhaps warrants a re-examination of the notion of 'typical AD' and a revisiting of tau pathological staging".

How tau spreads

The prion-like model holds that misfolded tau templates conversion of native tau in connected neurons, propagating trans-synaptically rather than by proximity. Human evidence is correlational but consistent and quantitative:

Evidence Finding
Connectivity predicts sequence Connectivity of patient-specific tau epicentres predicted the estimated tau spreading sequence cross-sectionally; longitudinally, tau accumulation rates correlated with connectivity strength to those epicentres (Franzmeier 2020, PMID 33246962)
Connectivity models beat Braak staging operationally A connectivity-based, patient-centred model improved assessment of tau accumulation rates versus Braak-stage readouts and reduced simulated trial sample sizes by ~40% (PMID 33246962)
APOE ε4 accelerates amyloid-driven spread Across ADNI (n=237) and Avid-A05 (n=130), amyloid PET mediated the association between ApoE4 status and subsequent tau increase (ADNI b 0.15, 95% CI 0.05–0.28, P=0.001; Avid-A05 b 0.33, 0.14–0.54, P<0.001), and ApoE4 carriers accumulated tau at lower Centiloid thresholds (Steward 2023, PMID 37930695)
Tau, not amyloid, predicts where the brain will shrink In 32 patients at early symptomatic stages, global tau-PET intensity — but not amyloid-PET — predicted the rate of subsequent atrophy independent of baseline cortical thickness, and tau distribution indicated the topography of future atrophy at the single-patient level, most strongly in younger patients (La Joie 2020, PMID 31894103)

An important caveat: functional connectivity is measured in the same individuals whose tau is measured, so "spread along connections" and "tau accumulates where the network is most active/vulnerable" are not separated by these designs. The mouse seeding literature supplies mechanism but not human dose–response; the analogous amyloid seeding experiments show that different aggregate forms drive propagation versus maturation, which complicates any single-conformer therapeutic strategy (Li 2022, PMID 36270003).

Tau positivity in life: frequencies and prognosis

Pooling 21 cohorts (6,514 participants, 13 countries) with visually rated flortaucipir scans indicating advanced (Braak V–VI) tangle pathology (Moscoso 2025, PMID 40522652):

Group Tau-PET positive 5-year progression
Cognitively unimpaired, overall 349/3,487 (9.8%) A+T+ 57% (95% CI 45–71) to MCI/dementia
Cognitively unimpaired, age 60 → 90 3% (2–4) → 19% (16–24) A+T− 17% (13–22); A−T− 6% (5–8)
MCI at age 75 43% (41–46) A+T+ → dementia 70% (59–81)
AD dementia at age 75 79% (77–82) —
Overlap with amyloid 92% of tau-positive individuals were also amyloid-positive —

Two facts follow. Tau positivity is far from negligible in unimpaired older people, so a positive tau scan is not by itself a dementia diagnosis. And the A+T+ combination is the strongest available short-horizon prognostic marker, which is the empirical basis for the 2024 criteria's split between diagnostic Core 1 and prognostic Core 2 biomarkers (see diagnostic criteria).

Structure and network propagation are different levels of mechanism

Cryo-EM shows that tauopathies are not collections of the same generic fibril: AD and primary age-related tauopathy share one filament fold, whereas Pick disease, chronic traumatic encephalopathy, corticobasal degeneration and progressive supranuclear palsy have distinct folds; some inherited MAPT splice mutations reproduce the argyrophilic-grain-disease fold (Shi 2021, PMID 34588692). A therapy that recognises recombinant tau or one conformer may therefore fail because the disease-relevant epitope is structurally inaccessible or belongs to another fold.

At the systems level, multimodal longitudinal imaging provides a testable bridge from amyloid to tau. In 140 amyloid-positive participants across the AD spectrum, amyloid-associated hyperconnectivity from temporal tau epicentres to vulnerable posterior regions mediated faster downstream tau accumulation; the temporal ordering was replicated in 345 people with preclinical AD and low cortical tau signal (Roemer-Cassiano 2025, PMID 39841807). In cognitively unimpaired multicohort data, baseline tau and amyloid or APOE ε4 status modified the association between hippocampal connectivity and later neocortical tau accumulation and memory decline (Ziontz 2024, PMID 39651679). These observational mediation results support activity-dependent spread but do not prove that reducing connectivity will slow disease.

Tau-directed therapeutics

Approach Agent(s) Result Status
Antisense oligonucleotide reducing tau synthesis BIIB080 (intrathecal) Phase 1b, 46 participants with mild AD: dose-dependent CSF t-tau reduction of 56% (95% CI 50–62) and p-tau181 reduction of 51% (38–63) in the two higher-dose cohorts; reduced tau-PET accumulation vs placebo at week 25 (n=13); at week 100 reductions from baseline across regions, largest in the temporal composite (−0.71 SUVR, 95% CI −1.40 to −0.02); generally well tolerated (Edwards 2023, PMID 37902726) Phase 2 ongoing (NCT03186989 was the phase 1b)
Passive anti-tau immunotherapy Gosuranemab, semorinemab, tilavonemab, zagotenemab Network meta-analysis of 6 RCTs, 2,193 patients: semorinemab better than placebo on MMSE and ADAS-Cog (MDs 0.52–3.21 and 0.17–3.30), but placebo ranked best on CDR-SB and ADCS-ADL (SUCRA 75.7% and 79.5%); most safety comparisons showed no difference (Cai 2024, PMID 39945003) No agent has demonstrated convincing clinical benefit
Gosuranemab (N-terminal tau mAb) — TANGO Phase 2, n=654 randomised / 650 treated, 78 weeks; high dose 2,000 mg q4w (Shulman 2023, PMID 38012285; NCT03352557, terminated) CDR-SB adjusted mean change: placebo 1.85; low dose 2.20; intermediate 2.24; high dose 1.85 — no dose favoured over placebo. All doses reduced CSF unbound N-terminal tau (P<0.0001). SAE incidence 10.3–12.3% vs 12.1% placebo Target engagement without clinical effect. Terminated
Semorinemab (anti-tau mAb) — Tauriel Phase 2, 97 sites, n=422 mITT, prodromal-to-mild AD, 73 weeks (Teng 2022, PMID 35696185; NCT03289143) CDR-SB Δ: placebo 2.19 (1.74–2.63); 1,500 mg 2.36 (1.83–2.89); 4,500 mg 2.36 (1.92–2.79); 8,100 mg 2.41 (1.88–2.94). AE rates 88.8–94.7% vs 93.1% placebo No slowing of clinical progression at any dose

The individual trials make the network-meta-analysis ranking less mysterious. TANGO engaged N-terminal tau in CSF at every dose and moved CDR-SB by exactly the same 1.85 points as placebo at the high dose (PMID 38012285). Tauriel moved CDR-SB by 2.19–2.41 points across four arms with overlapping confidence intervals (PMID 35696185). Target engagement of extracellular N-terminal fragments is not sufficient. Whether anti-tau antibodies failed because they engaged extracellular tau while pathology is intracellular, because they were given too late, or because the target is wrong is unresolved; BIIB080's ability to lower intracellular tau production is the cleanest test of the second and third possibilities and has not yet reported a clinical endpoint. See clinical trials landscape.

Open questions

  • Should staging schemes incorporate tau density as well as regional distribution, given the 0.2%–53.7% range within Braak stage V and its independent effect on decline (Richardson 2026, PMID 42184025)?
  • Do the four tau trajectories require separate outcome measures and separate treatment hypotheses, or are they variations on one process (Vogel 2021, PMID 33927414)?
  • Is connectivity-based spread causal, or does connectivity index regional vulnerability? Existing human designs cannot separate these (Franzmeier 2020, PMID 33246962).
  • Does lowering tau production translate into clinical benefit where antibody-mediated extracellular clearance did not (Edwards 2023, PMID 37902726; Cai 2024, PMID 39945003)?
  • What amyloid threshold should trigger intervention in APOE ε4 carriers, given that tau accumulation accelerates at lower Centiloid values in carriers (Steward 2023, PMID 37930695)?
  • Why are 8% of tau-PET-positive individuals amyloid-negative, and what are they — primary age-related tauopathy, another tauopathy, or tracer off-target binding (Moscoso 2025, PMID 40522652)?
  • Can a tau-PET ligand designed around the AD paired-helical-filament fold be shown to be blind to other tauopathy folds, as the cryo-EM structures imply it should be (Fitzpatrick 2017, PMID 28678775)?

References

  1. Braak H, et al. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 1991;82:239-59. PMID 1759558.
  2. Nelson PT, et al. Correlation of Alzheimer disease neuropathologic changes with cognitive status: a review of the literature. J Neuropathol Exp Neurol. 2012;71:362-81. PMID 22487856.
  3. Richardson TE, et al. Neocortical tau burden determines the degree of cognitive impairment in individuals with Braak stage V neurofibrillary degeneration. Acta Neuropathol. 2026;151. PMID 42184025.
  4. Fitzpatrick AWP, et al. Cryo-EM structures of tau filaments from Alzheimer's disease. Nature. 2017;547:185-190. PMID 28678775.
  5. Strang KH, et al. MAPT mutations, tauopathy, and mechanisms of neurodegeneration. Lab Invest. 2019;99:912-928. PMID 30742061.
  6. Ossenkoppele R, et al. Tau PET patterns mirror clinical and neuroanatomical variability in Alzheimer's disease. Brain. 2016;139:1551-67. PMID 26962052.
  7. Vogel JW, et al. Four distinct trajectories of tau deposition identified in Alzheimer's disease. Nat Med. 2021;27:871-881. PMID 33927414.
  8. La Joie R, et al. Prospective longitudinal atrophy in Alzheimer's disease correlates with the intensity and topography of baseline tau-PET. Sci Transl Med. 2020;12. PMID 31894103.
  9. Franzmeier N, et al. Patient-centered connectivity-based prediction of tau pathology spread in Alzheimer's disease. Sci Adv. 2020;6. PMID 33246962.
  10. Steward A, et al. ApoE4 and connectivity-mediated spreading of tau pathology at lower amyloid levels. JAMA Neurol. 2023;80:1295-1306. PMID 37930695.
  11. Moscoso A, et al. Frequency and clinical outcomes associated with tau positron emission tomography positivity. JAMA. 2025;334:229-242. PMID 40522652.
  12. Thal DR, et al. Phases of Aβ-deposition in the human brain and its relevance for the development of AD. Neurology. 2002;58:1791-800. PMID 12084879.
  13. Edwards AL, et al. Exploratory tau biomarker results from a multiple ascending-dose study of BIIB080 in Alzheimer disease: a randomized clinical trial. JAMA Neurol. 2023;80:1344-1352. PMID 37902726.
  14. Cai W, et al. Comparing the efficacy and safety of gosuranemab, semorinemab, tilavonemab, and zagotenemab in patients with Alzheimer's disease: a systematic review and network meta-analysis of randomized controlled trials. Front Aging Neurosci. 2024;16:1465871. PMID 39945003.
  15. Murray ME, et al. Neuropathologically defined subtypes of Alzheimer's disease with distinct clinical characteristics: a retrospective study. Lancet Neurol. 2011;10:785-96. PMID 21802369.
  16. Li X, et al. Seeding, maturation and propagation of amyloid β-peptide aggregates in Alzheimer's disease. Brain. 2022;145:3558-3570. PMID 36270003.
  17. Hutton M, et al. Association of missense and 5'-splice-site mutations in tau with the inherited dementia FTDP-17. Nature. 1998;393:702-5. PMID 9641683.
  18. Shulman M, et al. TANGO: a placebo-controlled randomized phase 2 study of efficacy and safety of the anti-tau monoclonal antibody gosuranemab in early Alzheimer's disease. Nat Aging. 2023;3:1591-1601. PMID 38012285.
  19. Teng E, et al. Safety and efficacy of semorinemab in individuals with prodromal to mild Alzheimer disease: a randomized clinical trial. JAMA Neurol. 2022;79:758-767. PMID 35696185.
  20. Shi Y, et al. Structure-based classification of tauopathies. Nature. 2021;598:359-363. PMID 34588692.
  21. Roemer-Cassiano SN, et al. Amyloid-associated hyperconnectivity drives tau spread across connected brain regions in Alzheimer's disease. Sci Transl Med. 2025;17:eadp2564. PMID 39841807.
  22. Ziontz J, et al. Connectivity, pathology, and ApoE4 interactions predict longitudinal tau spatial progression and memory. Hum Brain Mapp. 2024;45:e70083. PMID 39651679.