| Tau Protein | |
|---|---|
| Gene | MAPT |
| UniProt | P10636 |
| PDB | 2MZ7, 5O3L, 5O3T, 6QJH |
| Mol. Weight | 45–65 kDa (isoform-dependent, 0N3R to 2N4R) |
| Localization | Microtubule-associated, axonal |
| Family | Microtubule-associated protein family |
| Diseases | Alzheimer's Disease, Frontotemporal Dementia, PSP, CBD, Pick's Disease, CTE |
Tau Protein is an important component in the neurobiology of neurodegenerative diseases. This page provides detailed information about its structure, function, and role in disease processes.
Tau Protein is a microtubule-associated protein encoded by the [MAPT--TEMP--/genes)--FIX-- gene on chromosome 17q21.31[1]. It belongs to the family of intrinsically disordered proteins that undergo pathological aggregation in a group of neurodegenerative disorders collectively termed tauopathies[2]. Tau plays essential roles in neuronal physiology under normal conditions, including microtubule stabilization, axonal transport regulation, and synaptic function maintenance[3]. However, in disease states, tau becomes hyperphosphorylated, misfolds, and forms toxic oligomers and filamentous aggregates that drive neurotoxicity[4].
The tau protein has attracted intense research attention since its identification as a major component of the neurofibrillary tangles (NFTs) that characterize [Alzheimer's Disease--TEMP--/diseases)--FIX-- pathology[5]. The density of tau pathology in the brain correlates strongly with cognitive decline in AD, making tau a critical therapeutic target[6].
Human tau exists as six isoforms ranging from 352 to 441 amino acids in length, generated by alternative splicing of exon 2, exon 3, and exon 10 of the MAPT gene[7]. These isoforms differ in the presence of:
The isoforms are designated as:
In the adult human brain, approximately equal amounts of 3R and 4R tau are present[8]. The balance between 3R and 4R tau is crucial, as dysregulation of this ratio leads to specific tauopathies—for example, Pick's disease shows predominantly 3R tau, while PSP and CBD show elevated 4R tau[9].
Tau protein consists of several distinct domains:
The microtubule-binding repeats (MTBRs) are the most conserved region and are essential for tau's physiological function[10]. Each repeat is approximately 31 amino acids long and contains a conserved Lys-Ser-Pro (KSP) motif that is heavily phosphorylated in disease states[11].
The primary physiological function of tau is to bind and stabilize microtubules, which are essential for axonal transport[12]. Tau promotes microtubule assembly and prevents microtubule disassembly by binding to the microtubule surface through its repeat domains[13]. One tau molecule can bind to multiple microtubules, and a single microtubule can be stabilized by multiple tau molecules along its length[14].
The binding affinity of tau for microtubules is regulated by its phosphorylation state. Under normal conditions, tau is lightly phosphorylated, which allows dynamic binding and release from microtubules, enabling neuronal plasticity[15].
Tau regulates axonal transport by modulating the function of motor proteins kinesin and dynein[16]. Through interactions with the scaffolding proteins in the motor complex, tau influences the cargo-carrying capacity and velocity of axonal transport[17]. This function is critical for delivering synaptic vesicles, organelles, and signaling molecules from the cell body to synaptic terminals[18].
Emerging evidence indicates that tau plays roles in synaptic function beyond its well-characterized axonal functions[19]. Tau interacts with postsynaptic proteins including PSD-95 and NMDA receptors, influencing synaptic plasticity and signaling[20]. The presence of tau at synapses suggests it may participate in activity-dependent processes and memory formation[21].
During neuronal development, tau expression patterns differ from adults. Embryonic neurons express predominantly fetal tau isoforms (0N3R), which are later replaced by adult isoforms[22]. This developmental switch coincides with the maturation of axonal transport systems and establishment of neuronal polarity[23].
The conversion of tau from a normal, functional protein to a pathological aggregator begins with aberrant hyperphosphorylation[24]. In AD and other tauopathies, tau is phosphorylated at over 40 serine, threonine, and tyrosine residues[25]. Key pathological phosphorylation sites include:
Multiple kinases phosphorylate tau in disease states, including[26]:
The balance between kinases and phosphatases—particularly PP2A, which accounts for approximately 70% of tau phosphatase activity—is disrupted in disease[27]. PP2A activity is reduced in AD brain, contributing to hyperphosphorylation[28].
Hyperphosphylated tau undergoes a conformational change that exposes the MTBRs, enabling pathological tau-tau interactions[29]. The aggregation proceeds through:
The oligomeric forms of tau are increasingly recognized as the most toxic species[30]. Tau oligomers:
Tau pathology propagates through neural circuits in a pattern consistent with trans-synaptic spread[32]. Externalized tau oligomers can be taken up by neighboring neurons, where they serve as seeds that induce endogenous tau aggregation[33]. This prion-like mechanism explains the stereotypical progression of tau pathology through connected brain regions in AD[34].
The spread of tau pathology follows vulnerable neural networks:
As tau pathology accumulates, neurons lose the normal functions of tau protein[36]:
This loss-of-function component compounds the toxic gain-of-function from aggregates, accelerating neuronal demise[37].
In [Alzheimer's Disease--TEMP--/diseases)--FIX--, tau pathology progresses in a predictable staging system[38]:
The progression of tau pathology correlates better with cognitive decline than amyloid-beta burden[39]. Tau PET imaging using radioligands like [^18F]Flortaucipir (AV-1451) enables visualization of tau pathology in living patients and is now used for AD diagnosis and monitoring[40].
The FTD spectrum includes several tauopathies with distinct pathological features[41]:
Over 50 pathogenic MAPT mutations have been identified, causing familial FTD through mechanisms including[42]:
[CTE--TEMP--/diseases)--FIX-- results from repetitive traumatic brain injury and is characterized by perivascular tau pathology in cortical sulci[43]. CTE shows a unique pattern of tau pathology that differs from AD, with early involvement of subcortical structures and progressive accumulation in frontal and temporal cortices[44].
Multiple therapeutic approaches are being developed to target tau pathology[45]:
Current clinical trials targeting tau include[46]:
The study of Tau Protein has evolved significantly over the past decades. Research in this area has revealed important insights into the underlying mechanisms of neurodegeneration and continues to drive therapeutic development.
Historical context and key discoveries in this field have shaped our current understanding and will continue to guide future research directions.
Racial disparities in AD are observed in neuropsychological testing performance with African Americans performing more poorly on executive function and visuospatial tests. Mixed findings across longitudinal studies on cognitive decline rates. Neuropathological studies show mixed pathologies more common in African Americans (AD + Lewy bodies, AD + infarcts). CSF biomarker studies show lower phosphorylated tau181 and total tau in African Americans vs white individuals, but no differences in plasma or PET tau biomarkers. Higher amyloid PET burden found in African Americans in one study.
Model System: Human observational studies (cross-sectional and longitudinal cohorts) and postmortem brain studies
Statistical Significance: Various p-values reported in referenced studies; specific values not provided in this review
Identified three major molecular subtypes of AD with distinct dysregulated pathways: tau-mediated neurodegeneration, amyloid-β neuroinflammation, synaptic signalling, immune activity, mitochondrial organisation, and myelination.
Model System: Human AD brain tissue
Statistical Significance: Not applicable
Awuah Wireko Andrew et al., (2023)
Protocol for quantifying soluble and insoluble levels of cortical and hippocampal Ab1-40 and Ab1-42, and phospho-tau levels to establish baseline for preclinical studies
Model System: Transgenic AD mouse models
Statistical Significance: Not applicable - methodology proposal
CSF biomarkers become abnormal first, followed by atrophy rates, then cognitive test scores, and finally regional brain volumes. In amyloid+ and APOE+ subjects, CSF biomarkers follow the sequence: amyloid-b1-42, phosphorylated tau, total tau. In the broader population, total tau and phosphorylated tau appear before amyloid-b1-42 with higher uncertainty.
Model System: Human subjects from ADNI (285 subjects: 92 CN, 129 MCI, 64 AD)
Statistical Significance: MCI to AD conversion: P=2.06e-7; CN to MCI conversion: P=0.033
F-18 TKH5105 selectively binds to pathological PHF tau deposition in living AD patients and differentiates diseased brains from healthy controls. AD patients showed high retention in temporal cortex (known high density of neurofibrillary tangles) compared to cerebellum. Healthy controls' uptake in inferior temporal cortex was identical to cerebellum activity. Also showed in vitro binding to glial tau pathology in corticobasal degeneration and PSP. Rapid entry into gray matter areas, no toxic events reported.
Model System: Human subjects - patients with Alzheimer's disease (n=not specified) and healthy controls
Statistical Significance: Not reported
18F-AV-1451 demonstrates high selective binding and affinity to tau protein aggregates (15 nM Kd in brain slices, 0.7 nM in purified PHF-tau) with relatively spare binding to normal monomeric tau. Showed approximately 29 times greater binding to tau aggregates compared to beta-amyloid in AD brain gray matter. 'Tau-rich' brains (significant tau tangles with beta-amyloid) showed increased uptake in gray matter; 'tau-poor' brains and healthy controls did not. Normal brain showed only diffuse background activity. In 40 subjects, worse memory performance was associated with greater PET tau binding in entorhinal cortex. No significant serious safety concerns reported.
Model System: Human brain tissue (ex vivo), human subjects (AD patients and controls)
Statistical Significance: Correlation between memory performance and entorhinal cortex tau binding noted
FDDNP demonstrates binding to both beta-amyloid and tau pathology (was not designed as specific tau tracer)
Model System: Human subjects
Statistical Significance: Not reported
Significant negative associations found between [18F]THK5317 regional retention (mainly temporal regions: posterior parahippocampal gyrus, fusiform gyrus, inferior/middle/superior temporal gyri) and both episodic memory (RAVL learning, Rey delayed recall) and global cognition (FSIQ, MMSE). Association with FSIQ was more extensive than with MMSE. After FDR correction, inferior temporal gyrus retention remained significantly negatively associated with FSIQ.
Model System: Human patients with probable AD (9 AD dementia + 11 prodromal AD/MCI, PiB-positive)
Statistical Significance: p < 0.05 (uncorrected); inferior temporal gyrus-FSIQ association survived FDR correction
First-generation tracers (THK5117, THK5351, PBB3, AV1451) bind to NFTs, ghost tangles, and neuritic plaques. THK5117 and THK5351 showed binding to pretangles and PHF tau. Contradictory findings for binding to pretangles between studies. Multiple binding sites identified with varying affinities.
Model System: Postmortem human AD brain tissue (frozen and paraffin sections)
Statistical Significance: Ki values ranging from 0.001 pM to 800 nM depending on binding site
Tau retention largely restricted to medial temporal lobe (MTL) in CN elderly. This pattern consistent with primary age-related tauopathy (PART). Variable and relatively low or absent cortical findings.
Model System: Human cognitively normal (CN) elderly subjects
Statistical Significance: MTL binding pattern consistent with neuropathological literature
AD patients have significantly higher tau tracer retention than CN individuals. Binding in inferior lateral temporal, posterior cingulate, and lateral parietal regions matches known regional deposition of tau pathology. In atypical AD presentations, spatial pattern of retention matches underlying clinical phenotypes.
Model System: AD patients (prodromal and dementia stages)
Statistical Significance: p<0.05 for group differences between AD and CN
Differences in binding restricted to MTL regions in MCI vs CN. Lateral temporal and parietal areas show differences when examining only amyloid-beta-positive MCI.
Model System: MCI patients (amyloid-beta positive and negative)
Statistical Significance: Significant differences in amyloid-beta-positive MCI
Regional pattern of tau pathology expected in these diseases with relatively good discrimination from healthy volunteers. However, many ROIs coincide with areas showing off-target binding to MAO-B in basal ganglia, creating overlap across diagnostic groups. Longitudinal imaging shows increase in tracer binding with disease progression.
Model System: Patients with clinical diagnoses of corticobasal syndrome (CBS) and progressive supranuclear palsy (PSP)
Statistical Significance: Tracer binding correlates with clinical scores of functional impairment in PSP
In DS with amyloid-beta positivity, binding pattern resembles AD and increases with disease progression. In PD and DLB, binding largely variable with inconsistencies between studies and overlap with CN controls.
Model System: Parkinson's disease (PD), Dementia with Lewy bodies (DLB), Down's syndrome (DS) patients
Statistical Significance: Variable results across studies
AV1451 showed low/no affinity to straight tau filaments of CBS and PSP in vitro. Mixed results for correlation between in vivo AV1451 binding and postmortem tau load. THK tracers and PBB3 showed better ability to bind to non-AD tau deposits.
Model System: Patients with CBS or PSP who had in vivo PET imaging and subsequent neuropathological assessment
Statistical Significance: Variable correlation between in vivo and postmortem findings
THK family and AV1451 show similar and highly correlated regional distribution patterns. [11C]PBB3 shows different regional distribution with no correlation with THK tracers, indicating potentially different molecular targets.
Model System: AD patients
Statistical Significance: High correlation between THK and AV1451; no correlation between PBB3 and others
Moderate to strong positive correlation between tau PET and CSF p-tau across CN and patient groups. Discordance patterns: prodromal AD shows more isolated tau PET positivity while AD dementia shows more CSF p-tau elevation. [18F]AV1451 superior to CSF p-tau for diagnosing AD dementia.
Model System: CN subjects, MCI and AD patients, non-AD disorders
Statistical Significance: p<0.05 for correlations in combined analyses
Tau PET closely corresponds to hypometabolism across neocortical regions in AD. Tau binding more closely related to hypometabolism than to MRI results or amyloid-beta levels. Local and distant negative correlations between tau and MRI measures. Tau levels most strongly related to rates of neurodegenerative change.
Model System: CN subjects and AD patients
Statistical Significance: Significant correlations between tau and hypometabolism
Tau tracer binding associated with cognitive measures in prodromal and dementia-stage AD. Episodic memory associated with temporal lobe binding; executive/global cognitive function associated with neocortical binding. Tau binding in MTL associated with episodic memory even in CN individuals.
Model System: CN elderly, MCI, and AD patients
Statistical Significance: Significant associations across multiple studies
Observable increases in tau PET over time. In largest study, tau accumulation rates were rather uniform across brain regions, arguing against stepwise progression suggested by histopathology. Significant tau accumulation over 1 year in amyloid-beta-positive CN individuals.
Model System: CN elderly, MCI, and AD patients
Statistical Significance: Significant longitudinal increases observed
[18F]RO-948 elevated in medial temporal areas and broadly throughout cortex including precuneus, lateral parietal, occipital, and prefrontal cortex in AD vs controls. [18F]MK-6240 binding patterns consistent with Braak staging. Lower off-target binding to MAO-B rich areas compared to first-generation.
Model System: AD patients and healthy controls
Statistical Significance: Significant differences between AD and controls
Four different high-affinity binding sites identified on tau protofibril. AV1451 and MK-6240 bind to site 1 with higher affinity. THK5351 binds strongly to sites 1 and 3. PBB3 binds to all 4 sites with similar affinity.
Model System: Computational modeling using cryo-EM tau filament structures
Statistical Significance: Predicted binding affinities in nanomolar range consistent with experimental data
Excellent correspondence between FTP retention and neurofibrillary tangle distribution in primary AD. All 8 AD patients (Braak VI) showed markedly elevated SUVR in Braak stage regions. Non-Alzheimer tauopathies showed lower, more variable retention than AD but higher than controls. FTLD-TDP and FTLD-FUS showed tracer retention in tau-negative regions. Quantification detected advanced (Braak VI) but not early (Braak I-IV) NFT pathology. Subcortical SUVRs were similar across PSP, CBD, AD, and other pathologies.
Model System: Human postmortem brain tissue with antemortem PET imaging; 20 patients with diverse neurodegenerative diagnoses
Statistical Significance: Mean + 2 SD threshold used to define elevated SUVR above controls; specific p-values not provided
Soleimani-Meigooni et al., (2020)
All patients with PSP and CBD showed elevated SUVR in putamen, globus pallidus, and subthalamic nucleus above control threshold. However, SUVR levels were similar in AD, other non-Alzheimer tauopathies, and FTLD-TDP. Precentral/postcentral gyrus SUVRs were highest in AD. Dentate nucleus SUVR was similar across tauopathies and controls.
Model System: Human patients with PSP, CBD, AD, and other neurodegenerative diseases
Statistical Significance: Mean + 2 SD threshold used
Soleimani-Meigooni et al., (2020)
All patients with Braak VI pathology (n=8) showed elevated SUVR above control threshold. Patients with Braak I-III incidental pathology did not show elevated SUVR in Braak ROIs. PVC recovered signal in some patients with Braak III/IV pathology but also produced false positives in non-Alzheimer tauopathies.
Model System: Human patients with varying Braak stages (0-VI)
Statistical Significance: Mean + 2 SD threshold used
Soleimani-Meigooni et al., (2020)
Five clusters formed capturing tau tangle strain polymorphism across diseases; fAD PSEN1 and fAD APP share tau cluster; sAD and DS partially overlap; PiD tau deposits separate from all other diseases; clear difference observed between fAD and sAD tau tangles despite similar cryo-EM structures
Model System: Human tau tangles from sAD, fAD, DS, and Pick's disease brain samples
Statistical Significance: Inter-disease cluster separation
EMBER distinctly discriminated tau deposits localized to neurons, astrocytes, and oligodendrocytes in PiD; neurons, astrocytes, and oligodendrocytes have distinct conformational strains in same brain; spatial resolution revealed cell-type specific strain variation
Model System: Pick's disease brain sections with tau inclusions in neurons, astrocytes, and oligodendrocytes
Statistical Significance: Cell-type specific cluster separation
Glaucoma was associated with 36% higher PHF-tau levels in occipital cortex but was not statistically significant (OR = 1.36, 95% CI 0.91-2.03, p = 0.13). Weighted back to full cohort, OR was 1.30 (0.81-2.01).
Model System: Human postmortem occipital cortex tissue from ACT cohort (subset)
Statistical Significance: p = 0.13 (not significant)
Participants with Braak stages of IV or greater had elevated FTP-PET signal. FTP-PET was elevated in participants with Alzheimer's disease. Optimal meta-ROI SUVr cut-point of 1.29 was determined with 87% sensitivity and 82% specificity for identifying AD-spectrum diagnoses. Participants with PART and hippocampal sclerosis did not show elevated FTP-PET signal. Secondary AD neuropathology in LBD participants showed borderline elevated FTP-PET signal.
Model System: Human participants from Mayo Clinic Study of Aging (MCSA) and Mayo Clinic Alzheimer's Disease Research Center (ADRC)
Statistical Significance: p=0.008 for AD vs LBD SUVr comparison; rho's from 0.61-0.70, p ≤ 0.002 for IHC correlation
Quantitative measurements of hippocampal and temporal lobe tau burden were highly correlated to FTP-PET signal. Moderately strong correlations observed with rho's from 0.61-0.70 (p ≤ 0.002). Several tissue samples had tau burden in range of 5% or less, showing limitations of biomarker correlations for low levels of tau neuropathology.
Model System: Human postmortem brain tissue from 26 study participants
Statistical Significance: rho's from 0.61-0.70, p ≤ 0.002
Normal aging leads to increases in H4K16ac; AD associated with H4K16ac loss in lateral temporal lobe including near AD susceptibility loci. Tau pathology correlates with H3K9ac dysregulation in up to 23% of all H3K9ac domains. Tau-related H3K9ac alterations cluster in large genomic segments covering several megabase pairs.
Model System: Human AD postmortem brain (lateral temporal lobe, entorhinal cortex, dorsolateral prefrontal cortex)
Statistical Significance: Not specified
Cláudio Gouveia Roque et al., (2024)
Combined set of three miRNAs (miR-27a-3p, miR-30a-5p, miR-34c) and three piRNAs adequately detects AD and predicts MCI to AD conversion. When combined with phospho-tau and Ab42/40 ratio, achieved AUC of 0.98.
Model System: Human CSF samples
Statistical Significance: Not specified
Cláudio Gouveia Roque et al., (2024)
AD-Mimics had more Braak stages III-IV (vs V-VI in Confirmed-AD), more FTLD-tau (18.6% vs 1.6%), FTLD-TDP-43 (21.2% vs 5.0%), gross infarcts (39.2% vs 26.3%), microinfarcts (35.0% vs 26.5%), and hippocampal sclerosis (22.9% vs 12.3%). Unidentified-AD had more neocortical Lewy bodies (32.4% vs 13.7%), FTLD-tau (7.6% vs 1.6%), and FTLD-TDP-43 (9.1% vs 5.0%) but less frequent neuritic plaques (65.6% vs 76.4%).
Model System: Human postmortem brain tissue
Statistical Significance: Most comparisons p<0.05; some became insignificant after Bonferroni correction
Gauthreaux et al., (2020)
FTLD-tau (OR=17.92, p<0.001) and FTLD-TDP-43/other (OR=6.61, p<0.001) were associated with increased odds of AD-Mimics. Lewy body pathology (OR=0.45, p=0.004) was associated with decreased odds. Age at death also increased odds (OR=1.05).
Model System: Human postmortem data from NACC
Statistical Significance: FTLD-tau: p<0.001; FTLD-TDP-43: p<0.001; LBD: p=0.004; age: p<0.001
Gauthreaux et al., (2020)
Lewy body pathology (OR=1.94, p<0.001), FTLD-tau (OR=4.28, p=0.003), and FTLD-TDP-43/other (OR=2.51, p=0.014) were associated with increased odds of Unidentified-AD. Age at death decreased odds (OR=0.96).
Model System: Human postmortem data from NACC
Statistical Significance: LBD: p<0.001; FTLD-tau: p=0.003; FTLD-TDP-43: p=0.014; age: p<0.001
Gauthreaux et al., (2020)
Highly significant correlation between neocortical flortaucipir SUVr and Ptau tissue concentration (Pearson r = 0.81; p < 0.0001). Correlation preserved after adjustment for Aβ (Pearson r = 0.73, p < 0.0001). Modest correlation in limbic regions (Pearson r = 0.52; p = 0.080). No correlation in basal ganglia (Pearson r = 0.27; p = 0.280). Apparent SUVr-Aβ correlation (r = 0.61) disappeared when adjusted for Ptau.
Model System: Human postmortem brain tissue from 3 subjects with dementia (2 with clinical AD diagnosis, 1 with dementia of unknown origin)
Statistical Significance: Neocortical: p < 0.0001; Limbic: p = 0.080; Basal ganglia: p = 0.280
All three subjects had intermediate to high AD neuropathological change. Braak stages: Case 1 (VI/V), Case 2 (IV/V), Case 3 (V/V). Thal phases: Cases 1 and 3 = 4 (high), Case 2 = 3 (intermediate). CERAD: Cases 1 and 3 frequent, Case 2 moderate. No atypical tau pathology identified in any case.
Model System: Human postmortem brain tissue
Statistical Significance: N/A - descriptive neuropathology
All three cases visually interpreted as having AD pattern (τAD) of pathologic tau. Case 2 showed least retention (limited to mesial and lateral temporal). Case 3 showed most intense signal (lateral temporal, occipital, parietal, right-lateralized frontal). Case 1 showed moderate retention (frontal, lateral temporal, parietal).
Model System: Human in vivo PET imaging
Statistical Significance: N/A - qualitative assessment
Strong evidence that tau PET tracers (18F-flortaucipir, 18F-MK6240, 18F-RO948, 18F-PI2620) bind AD tau aggregates in advanced Braak stages (>IV). Accuracy for detecting tau load in Braak V-VI was 87.5% (95% CI, 77.2%-93.5%). Strong correlations (R2 range, 0.66-0.76) between tau PET levels and quantitative neuropathologic tau burden. Tracer binding weaker in non-AD tauopathies and overlaps with off-target regions.
Model System: Human postmortem brain tissue
Statistical Significance: R2 range 0.66-0.76 for AD tau correlations; 87.5% accuracy for Braak V-VI detection
Both increased amyloid and tau PET levels associated with cognitive decline, but relationship predominantly driven by tau. A-positive, T-positive participants demonstrated cognitive decline over 2-year follow-up; one met MCI criteria at 2-year follow-up. A-negative, T-positive participant remained cognitively stable.
Model System: Cognitively unimpaired human participants (PREVENT-AD cohort)
Statistical Significance: When both amyloid and tau included in model, only tau remained significant
Elevated baseline tau PET levels strongly associated with more rapid cognitive decline. Tau PET outperformed amyloid PET and structural MRI in head-to-head comparisons. Visually determined positive 18F-flortaucipir PET scan associated with increased risk for future cognitive decline (MMSE HR 1.68, 95% CI 1.22-2.32) and functional decline (CDR sum of boxes HR 1.40, 95% CI 1.11-1.76) after 18 months. Both intensity and extent of baseline tau PET predictive for future brain atrophy rates.
Model System: Human participants with MCI and AD dementia
Statistical Significance: MMSE hazard ratio 1.68 (95% CI, 1.22-2.32); CDR sum of boxes hazard ratio 1.40 (95% CI, 1.11-1.76)
Tau PET tracers demonstrated excellent diagnostic performance for distinguishing AD dementia from non-AD neurodegenerative disorders with sensitivity and specificity above 90%. Tau PET superior to structural MRI, amyloid PET, and most biofluid markers. Utility at individual-patient level limited for non-AD tauopathies (PSP, CBD, Pick disease). Estimated capability to detect up to 70% of neurodegenerative dementias when correctly implemented.
Model System: Human participants across diagnostic groups (AD dementia, FTD, PSP, CBS, MCI, CU)
Statistical Significance: Sensitivity and specificity above 90% for distinguishing AD dementia from non-AD disorders
Off-target binding profiles vary across tracers. 18F-flortaucipir shows off-target binding in basal ganglia, substantia nigra, longitudinal sinuses, pituitary, choroid plexus. 18F-RO948 and 18F-MK6240 show greater binding to meninges and skull. 18F-PI2620 shows promise for 4R tauopathies with lower off-target binding in basal ganglia.
Model System: Human participants
Statistical Significance: Not specified
PVC improved discriminative accuracy between cognitively impaired and unimpaired individuals cross-sectionally but resulted in less robust longitudinal changes. PVC modestly restores hippocampal signal and correlation between hippocampal signal and clinical symptoms.
Model System: Tau PET data from multiple studies
Statistical Significance: Not specified in detail
Inferior cerebellar gray matter most widely used reference region. Inferior cerebellar reference provided most sensitive measure for cross-sectional group differences. Eroded white matter or eroded white matter cerebellar composite with longitudinal processing pipeline most suitable for longitudinal analyses.
Model System: Tau PET data from multiple cohorts
Statistical Significance: Not specified
Flortaucipir PET predicted B3-level tau pathology with sensitivity 92.3-100% and specificity 52.0-92.0%. Predicted high ADNC with sensitivity 94.7-100% and specificity 50.0-92.3%. Majority read analysis showed 92.3% sensitivity and 80.0% specificity for B3, 94.7% sensitivity and 80.8% specificity for high ADNC. Inter-rater reliability high (Fleiss k=0.74, P<.001). SUVR cutpoint >1.113 yielded 84.2% sensitivity for B3 and 86.5% for high ADNC with 100% specificity.
Model System: Human participants with terminal illness (n=64 in primary cohort; 156 enrolled total; 67 autopsied)
Statistical Significance: P<.001 for inter-rater reliability; 95% CI ranges provided for all sensitivity/specificity values
Addition of supplemental data improved specificity and maintained overall accuracy. Of 26 impaired participants with non-AD clinical diagnosis, 19 had less than high ADNC at autopsy; 16 of these had flortaucipir PET accurately interpreted as not AD pattern.
Model System: Historically collected autopsy cases from academic centers (n=16)
Statistical Significance: Not specified
ESM explained 70.2% of overall spatial variance in tau distribution and 50.9% within individual subjects. Entorhinal cortex confirmed as best epicenter, corroborating autopsy findings. Model predicted spatial patterns of tau regardless of amyloid status.
Model System: Human participants (312 individuals along AD continuum)
Statistical Significance: Global pattern r2 = 0.702 (p < 0.01), individual mean r2 = 0.509, SD = 21.8% (p < 0.01)
Model performance remained high even among Aβ- individuals despite low overall tau burden. Tau patterns resembled early Braak staging even in subjects without elevated amyloid and in APOE4-negative cognitively normal individuals.
Model System: Amyloid-negative human participants
Statistical Significance: High model fit maintained in Aβ- subjects
Underestimated regions (more tau than predicted) had significantly greater amyloid burden. Significant correlation between regional model residuals and regional amyloid levels.
Model System: Human brain regions (66 cortical regions)
Statistical Significance: t = 2.9, p = 0.004; correlation p < 0.001
No population-level hemisphere preference. Individual epicenter hemisphere associated with tau deposition asymmetry. Left-limbic epicenters showed greater left temporoparietal binding; right-limbic epicenters decreased with disease progression. Right-limbic subjects were older.
Model System: Individual human subjects
Statistical Significance: Epicenter hemisphere associated with asymmetry p < 0.001; age effect p = 0.01
Abnormally high cortical 18F T807 binding in MCI and AD patients compared to CN controls. Elevated neocortical 18F T807 binding particularly in inferior temporal gyrus was associated with clinical impairment. The association of cognitive impairment was stronger with inferior temporal 18F T807 than with mean cortical 11C PiB. Regional 18F T807 was correlated with mean cortical 11C PiB among both impaired and control subjects. Patterns corresponded to Braak staging scheme.
Model System: Human participants: clinically normal older adults (N=56), mild cognitive impairment (N=13), and mild AD dementia (N=6)
Statistical Significance: MCI/AD vs CN: p<0.05 to p<0.001 for multiple ROIs; Inferior temporal: rho=-0.83 (p<10^-4) for MMSE in MCI/AD; rho=0.75 (p=0.0002) for CDR-SB in MCI/AD
Gene expression-pseudotime significantly associated with: tau positivity (F=17.64, P<0.001), amyloid positivity (F=28.22, P<0.001), tau-amyloid co-morbidity (F=9.58, P<0.001), brain infarcts (F=5.32, P<0.05), tau-amyloid-infarct co-morbidity (F=7.49, P<0.001), clinical diagnosis (F=56.72, P<0.001), clinical conversion (F=56.61, P<0.001), executive function (R=0.23, P<0.001), memory performance (R=0.27, P<0.001). Subjects with higher pseudotime progressed to advanced disease in average 3.18 years (SD 2.33).
Model System: In vivo blood samples (ADNI database)
Statistical Significance: All P-values FWE-corrected; clinical conversion: P<0.001
Lecanemab demonstrated 27% slowing of disease progression compared to placebo (-0.45 [95%CI: -0.67 to -0.23] CDR-SB) after 18 months. Showed substantial reductions in amyloid plaques and attenuation of insoluble tau deposition. Plasma P-tau181 and GFAP were significantly reduced.
Model System: Human clinical trial - patients with prodromal to mild AD
Statistical Significance: p < 0.001 for primary endpoint
Donanemab showed 29% reduction in disease progression in combined population (-0.7 [95%CI, -0.95 to -0.45] CDR-SB). In individuals with low/intermediate tau, a larger 40% effect was observed (-0.67 [95%CI, -0.95 to -0.40]). 52% of participants stopped treatment when amyloid plaque clearance was achieved.
Model System: Human clinical trial - patients with early symptomatic AD
Statistical Significance: p < 0.001 for primary endpoint
Donanemab demonstrated dramatic reductions in amyloid PET signal in early-stage trials. Used 'Goldilocks' tau PET criterion - requiring enough medial temporal lobe uptake to confirm AD but not too much tau to be past responsiveness to amyloid removal.
Model System: Human clinical trial - early AD patients
Statistical Significance: Not fully specified
Blood tests showed high accuracy in differential diagnosis of AD vs. other dementias. Plasma P-tau181 and P-tau217 are exquisitely sensitive to treatment effects of anti-amyloid therapies, with large reductions reaching steady state 6-12 months after treatment initiation. However, reductions occurred even in trials without clinical efficacy, suggesting sensitivity to small, non-clinically meaningful effects.
Model System: Human subjects and clinical trial participants
Statistical Significance: High discriminative accuracy reported in validation studies
Discovery of quinoline and benzimidazole lead compounds (BF-126, BF-158, BF-170) with tau binding capability but relatively low selectivity over amyloid plaques. Brain uptake was good (11.3% ID/g at 2 min) but clearance was slow (3.1% ID/g at 30 min). EC50 values >200 nM with high nonspecific binding.
Model System: Synthetic heparin-induced tau polymers (HITPs) and postmortem AD human brain homogenates (AD-PHFs)
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
THK-523 showed higher in vivo retention in tau transgenic mice vs amyloid model mice. In AD patients, higher cortical retention vs healthy controls with distribution matching PHF histopathology. However, high white matter retention hindered clear visualization and precluded further development.
Model System: Synthetic HITP tau fibrils, AD brain homogenates, rTg4510 tau transgenic mice, APP/PS1 amyloid mice
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Both compounds showed higher tau affinities than THK-523 with Kd values of 2.6 nM (THK-5105) and 5.2 nM (THK-5117) for AD-PHF. Higher Ab/tau selectivity. Distribution in mesial temporal sections coincided with Gallyas-Braak staining and tau immunostaining but not PiB. THK-5117 showed faster kinetics and better SNR than THK-5105. Tracer retention associated with dementia severity and brain atrophy.
Model System: Synthetic HITP tau, human AD-PHF tau aggregates, human AD brain sections
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
High binding affinity for tau in AD brain homogenates (Kd = 2.9 nM) with faster kinetics and lower white matter retention. Peak SUVR achieved within 5 min post-injection, exiting brain by 90 min. SUVR stabilized at ~1.5 in controls and ~2.5 in AD patients at 50 min. Distinguished AD from controls. Also tested in PSP patient showing uptake in midbrain.
Model System: Human AD brain homogenates, human AD brain sections, 16 healthy controls, 5 MCI subjects, 13 AD patients
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Good tau affinity (Kd [HITP] = 2.55 nM) and 50-fold selectivity for tau over Ab. Greater accumulation in medial temporal region of AD patients vs controls. Captured spreading of neurofibrillary tau from limbic system to neocortical areas. Distribution different from 11C-PiB. Tested in CBD patient showing tracer retention in basal ganglia.
Model System: HITPs, PS19 and rTg4510 transgenic mice, human AD brain sections, human subjects
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Both tracers showed high affinity for tau (Kd = 14.6 nM for T807, 22 nM for T808), >25-fold selectivity for AD-PHF over Ab, lack of white matter binding. Good correlation with tau immunostaining but not Ab. Linear correlation between NFT loads and T807 autoradiography intensity across 26 human donors. T807 showed rapid brain uptake (4.16% ID/g at 2 min) and clearance. Clinical PET showed accumulation matching PHF distribution with increasing neocortical spread associated with dementia severity.
Model System: Human AD brain sections, APPSWE-PS19 transgenic mice, human subjects
Statistical Significance: Linear correlation between NFT burden and cognitive scores confirmed
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Tau pathology appears in hippocampus, parahippocampus, and entorhinal cortex in early dementia stages. Increased tau in inferior temporal lobe associated with worse memory. CSF tau levels correlated with tau imaging in 6 brain regions consistent with Braak staging. Test/retest reproducibility ~4-5%. ~10% year-over-year increase in mean cortical SUVR in high amyloid burden subjects. Increased tau accompanied by lower MMSE performance.
Model System: Human subjects from Harvard Aging Brain Study (75 older subjects)
Statistical Significance: Statistically significant correlations between CSF tau and tau PET in entorhinal/parahippocampal regions, inferior temporal, middle temporal, and superior temporal cortices
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Regional tau PET patterns agreed with regions underlying clinical symptoms in atypical AD. FDG PET abnormally low where tau high. In PPA, tau signal spreading through language networks. In PSP, positive signals in substantia nigra, globus pallidus, and subthalamic nucleus, but some overlap with controls limiting early-stage detection.
Model System: PSP patients, primary progressive aphasia patients, atypical AD patients
Statistical Significance: p < 0.001 uncorrected for PSP vs controls in specific brain regions
Hartmuth C. Kolb, José Ignacio Andrés (2017)
T807 perfectly colocalized with tau-containing neurons and dystrophic neurites. Confirmed strong T807 binding to tau pathology in AD but not to cerebral amyloid, DLB, MSA, or TDP-43. Tangles and dystrophic neurites account for most of the in vivo T807 signal.
Model System: Human AD brain tissue
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
All three compounds displayed high-affinity binding to tau in sub-10-nM range with strong selectivity over Ab in human brain tissue. They displaced T808 binding in human brain slices at Braak stages V/VI. Binding colocalized with aggregated tau IHC but not Ab. Baboon PET showed good brain permeability, washout, and low white matter binding. Phase I trial ongoing in HCs and AD patients.
Model System: Human brain tissue, baboon PET studies
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Amygdala shows high tau-PET signal in early stages. Earliest tau-PET uptake in amygdala ~10 years before AD diagnosis. Amygdala, ERC, and hippocampus show largest annual increase in tau-PET uptake in Aβ-positive cognitively unimpaired individuals.
Model System: Human subjects (cognitively unimpaired, MCI, AD patients)
Statistical Significance: Not specified in detail in this review
Cases showed: (1) pathology often exists around the periphery of amygdalae near meninges and/or lateral ventricle; (2) peri-amygdaloid grey matter including entorhinal cortex frequently shows pathologies; (3) cortical and transitional regions are vulnerable; (4) phospho-Tau pathology is constant in all aged individuals. Cases 3-6 showed comorbid Ab, Tau, a-synuclein, and TDP-43 pathologies.
Model System: Human postmortem amygdala tissue (UK-ADC autopsy cohort)
Statistical Significance: Not applicable (descriptive case series)
Nelson et al., (2018)
TEBM identified biomarker event sequence over mean period of 17.3 years (95% CI: 11.4-27.1 years). Order: CSF/PET Aβ abnormality first (~0.03 years), followed by tau markers (~2.65-2.7 years), hippocampal volume (~5.1 years), entorhinal volume (~5.7 years), first cognitive changes (~7.5 years), mid-temporal volume (~10.4 years), and ventricular abnormality. Validated with OASIS dataset showing same ordering within 95% CIs.
Model System: Human data from ADNI (Alzheimer's Disease Neuroimaging Initiative) and OASIS (Open Access Series of Imaging Studies)
Statistical Significance: 95% confidence intervals reported for all event timings
Successfully estimated subject-specific brain-wide influences due to Aβ, tau and Aβ∙tau on neuronal activity; quantified group differences (AD vs CU) via rank sum tests
Model System: Human subjects from TRIAD cohort (47 cognitively unimpaired, 16 AD patients)
Statistical Significance: Non-parametric rank sum tests used
Sanchez-Rodriguez et al., (2023)
Identified 756 genes associated with Aβ effects, 650 genes for tau effects, and 1987 genes for Aβ∙tau synergistic effects on neuronal activity; included AD-risk genes (CD33, ADAM10, SNCA, CLU)
Model System: Human postmortem brain tissue from Allen Human Brain Atlas (6 neurotypical adult brains)
Statistical Significance: 99% bootstrap CI not including zero
Sanchez-Rodriguez et al., (2023)
Top enriched pathways: neuroinflammation/immune system processes, developmental processes (sensory organ development, tissue morphogenesis), cell communication/transport mechanisms (regulation of secretion, vesicle-mediated transport); immune pathways particularly overrepresented in Aβ∙tau set
Model System: Gene sets from transcriptomic analysis
Statistical Significance: q < 0.05, FDR-corrected
Sanchez-Rodriguez et al., (2023)
Aβ molecular associates enriched in pyramidal cells (q=0.002, δ=3.804) and endothelial-mural cells (q=0.008, δ=2.950); tau associates enriched in interneurons (q=0.021, δ=2.834); Aβ∙tau signature highly enriched in microglia (q<0.001, δ=10.425)
Model System: Gene sets from transcriptomic analysis with single-cell data from mouse somatosensory cortex and hippocampus CA1
Statistical Significance: q < 0.05
Sanchez-Rodriguez et al., (2023)
rs2288911 is a microglia-specific eQTL for APOE expression (P=1.1x10^-13). Associated with cerebral amyloid angiopathy (CAA) (P=1.18x10^-7) but not AD pathology or amyloid/tau proteinopathy. Association with CAA persists after adjusting for APOE4 (P=9.9x10^-6). No interaction with APOE4 on CAA.
Model System: Human DLPFC microglia from ROS/MAP participants
Statistical Significance: P=1.1x10^-13 for eQTL; P=1.18x10^-7 for CAA association
The DPS model explains 62% of total variance, significantly outperforming single biomarkers (ADAS: 49.4%, MMSE: 46%, p<0.01). RAVLT30 (Rey Auditory Verbal Learning Test) was identified as the earliest biomarker to become abnormal. Hippocampal volume (HIPPO), ABETA, and TAU precede cognitive tests (MMSE and ADAS). Mean ADPS at baseline: N=-0.03, MCI=2.85, AD=6.49. Rate of progression increases with disease severity: N=-0.08/year, MCI=0.76/year, AD=1.46/year.
Model System: Human ADNI cohort (ages 55-90): 687 subjects with 3658 visits total - 1103 Normal, 1513 MCI, 1010 AD diagnoses
Statistical Significance: p<0.01 for DPS vs ADAS comparison; p<0.01 for RAVLT30 preceding all other biomarkers; p<0.02 for HIPPO, ABETA, TAU preceding MMSE and ADAS
Model successfully reproduced Braak stages of Alzheimer's disease progression: initial seeding in entorhinal cortex (stages I-II), progression to hippocampus and brain stem, then to limbic system (stages III-IV), and finally to cortical association areas and primary sensory-motor regions (stages V-VI)
Model System: 2D finite element model from T2-weighted MRI of 32-year-old male brain (sagittal slice)
Statistical Significance: N/A - computational model
Model reproduced characteristic progression patterns for: (a) amyloid-β in Alzheimer's disease, (b) tau inclusions in Alzheimer's disease, (c) α-synuclein in Parkinson's disease, (d) TDP-43 in ALS. All showed generic progression trends: fast progression through limbic system once near ventricles, cortical invasion, and late parietal lobe involvement
Model System: 3D finite element whole brain model from T2-weighted MRI
Statistical Significance: N/A - computational model
LC neuronal loss averages 63% in AD; LC volume decreases by 8.4% per Braak stage increase; 8% of LC neurons are p-tau-positive at Braak stage 0, doubling by Braak stage I, reaching 100% by Braak stage VI; rostral portion affected more severely (83% loss) compared to middle (23%) and caudal (15%) parts
Model System: Human postmortem brain tissue
Statistical Significance: Significant correlation between LC volume loss and Braak stage (p<0.05)
p-tau begins to accumulate in LC early in life, in some cases as young as 10 years of age; 90% of individuals have some tau pathology in LC by age 30; 72% of individuals aged 31-40 years have tau lesions; 94% of individuals aged 41-50 years have tau lesions
Model System: Human postmortem brain tissue
Statistical Significance: Not reported
Four significant function clusters: (1) APP metabolism/Aβ formation (P=4.56e-7); (2) Tau protein binding (P=3.19e-5); (3) Lipid metabolism (P=1.45e-7); (4) Immune response (P=6.32e-5). Highly significant correlation between common and rare variant results (P=1.32e-7)
Model System: Gene-set analysis from GWAS summary statistics
Statistical Significance: q <= 0.05 after FDR correction
Amyloid-beta positive participants had higher tau levels in both amygdala and entorhinal cortex (p < 0.001); they also had smaller amygdala volumes than Aß- participants (p < 0.05). Within Aß+ groups, elevated amygdala tau was associated with lower amygdala volumes (ß=-445.198, p=0.01). Both amygdala (ß=-2.908, p=0.02) and entorhinal tau (ß=-3.347, p=0.02) were significant predictors of MMSE scores, but only in Aß+ participants. Amygdala volume was a significant predictor of memory among Aß+ participants (ß=0.001, p=0.01).
Model System: Human participants - cognitively unimpaired older adults from ADNI cohort (Aß- = 145, Aß+ = 81)
Statistical Significance: p < 0.001 for tau levels in Aß+; p < 0.05 for amygdala volume; p = 0.01 for amygdala tau-volume relationship; p = 0.02 for tau predicting MMSE; p = 0.01 for volume predicting memory
Racial disparities in AD are observed in neuropsychological testing performance with African Americans performing more poorly on executive function and visuospatial tests. Mixed findings across longitudinal studies on cognitive decline rates. Neuropathological studies show mixed pathologies more common in African Americans (AD + Lewy bodies, AD + infarcts). CSF biomarker studies show lower phosphorylated tau181 and total tau in African Americans vs white individuals, but no differences in plasma or PET tau biomarkers. Higher amyloid PET burden found in African Americans in one study.
Model System: Human observational studies (cross-sectional and longitudinal cohorts) and postmortem brain studies
Statistical Significance: Various p-values reported in referenced studies; specific values not provided in this review
Identified three major molecular subtypes of AD with distinct dysregulated pathways: tau-mediated neurodegeneration, amyloid-β neuroinflammation, synaptic signalling, immune activity, mitochondrial organisation, and myelination.
Model System: Human AD brain tissue
Statistical Significance: Not applicable
Awuah Wireko Andrew et al., (2023)
Protocol for quantifying soluble and insoluble levels of cortical and hippocampal Ab1-40 and Ab1-42, and phospho-tau levels to establish baseline for preclinical studies
Model System: Transgenic AD mouse models
Statistical Significance: Not applicable - methodology proposal
CSF biomarkers become abnormal first, followed by atrophy rates, then cognitive test scores, and finally regional brain volumes. In amyloid+ and APOE+ subjects, CSF biomarkers follow the sequence: amyloid-b1-42, phosphorylated tau, total tau. In the broader population, total tau and phosphorylated tau appear before amyloid-b1-42 with higher uncertainty.
Model System: Human subjects from ADNI (285 subjects: 92 CN, 129 MCI, 64 AD)
Statistical Significance: MCI to AD conversion: P=2.06e-7; CN to MCI conversion: P=0.033
F-18 TKH5105 selectively binds to pathological PHF tau deposition in living AD patients and differentiates diseased brains from healthy controls. AD patients showed high retention in temporal cortex (known high density of neurofibrillary tangles) compared to cerebellum. Healthy controls' uptake in inferior temporal cortex was identical to cerebellum activity. Also showed in vitro binding to glial tau pathology in corticobasal degeneration and PSP. Rapid entry into gray matter areas, no toxic events reported.
Model System: Human subjects - patients with Alzheimer's disease (n=not specified) and healthy controls
Statistical Significance: Not reported
18F-AV-1451 demonstrates high selective binding and affinity to tau protein aggregates (15 nM Kd in brain slices, 0.7 nM in purified PHF-tau) with relatively spare binding to normal monomeric tau. Showed approximately 29 times greater binding to tau aggregates compared to beta-amyloid in AD brain gray matter. 'Tau-rich' brains (significant tau tangles with beta-amyloid) showed increased uptake in gray matter; 'tau-poor' brains and healthy controls did not. Normal brain showed only diffuse background activity. In 40 subjects, worse memory performance was associated with greater PET tau binding in entorhinal cortex. No significant serious safety concerns reported.
Model System: Human brain tissue (ex vivo), human subjects (AD patients and controls)
Statistical Significance: Correlation between memory performance and entorhinal cortex tau binding noted
FDDNP demonstrates binding to both beta-amyloid and tau pathology (was not designed as specific tau tracer)
Model System: Human subjects
Statistical Significance: Not reported
Significant negative associations found between [18F]THK5317 regional retention (mainly temporal regions: posterior parahippocampal gyrus, fusiform gyrus, inferior/middle/superior temporal gyri) and both episodic memory (RAVL learning, Rey delayed recall) and global cognition (FSIQ, MMSE). Association with FSIQ was more extensive than with MMSE. After FDR correction, inferior temporal gyrus retention remained significantly negatively associated with FSIQ.
Model System: Human patients with probable AD (9 AD dementia + 11 prodromal AD/MCI, PiB-positive)
Statistical Significance: p < 0.05 (uncorrected); inferior temporal gyrus-FSIQ association survived FDR correction
First-generation tracers (THK5117, THK5351, PBB3, AV1451) bind to NFTs, ghost tangles, and neuritic plaques. THK5117 and THK5351 showed binding to pretangles and PHF tau. Contradictory findings for binding to pretangles between studies. Multiple binding sites identified with varying affinities.
Model System: Postmortem human AD brain tissue (frozen and paraffin sections)
Statistical Significance: Ki values ranging from 0.001 pM to 800 nM depending on binding site
Tau retention largely restricted to medial temporal lobe (MTL) in CN elderly. This pattern consistent with primary age-related tauopathy (PART). Variable and relatively low or absent cortical findings.
Model System: Human cognitively normal (CN) elderly subjects
Statistical Significance: MTL binding pattern consistent with neuropathological literature
AD patients have significantly higher tau tracer retention than CN individuals. Binding in inferior lateral temporal, posterior cingulate, and lateral parietal regions matches known regional deposition of tau pathology. In atypical AD presentations, spatial pattern of retention matches underlying clinical phenotypes.
Model System: AD patients (prodromal and dementia stages)
Statistical Significance: p<0.05 for group differences between AD and CN
Differences in binding restricted to MTL regions in MCI vs CN. Lateral temporal and parietal areas show differences when examining only amyloid-beta-positive MCI.
Model System: MCI patients (amyloid-beta positive and negative)
Statistical Significance: Significant differences in amyloid-beta-positive MCI
Regional pattern of tau pathology expected in these diseases with relatively good discrimination from healthy volunteers. However, many ROIs coincide with areas showing off-target binding to MAO-B in basal ganglia, creating overlap across diagnostic groups. Longitudinal imaging shows increase in tracer binding with disease progression.
Model System: Patients with clinical diagnoses of corticobasal syndrome (CBS) and progressive supranuclear palsy (PSP)
Statistical Significance: Tracer binding correlates with clinical scores of functional impairment in PSP
In DS with amyloid-beta positivity, binding pattern resembles AD and increases with disease progression. In PD and DLB, binding largely variable with inconsistencies between studies and overlap with CN controls.
Model System: Parkinson's disease (PD), Dementia with Lewy bodies (DLB), Down's syndrome (DS) patients
Statistical Significance: Variable results across studies
AV1451 showed low/no affinity to straight tau filaments of CBS and PSP in vitro. Mixed results for correlation between in vivo AV1451 binding and postmortem tau load. THK tracers and PBB3 showed better ability to bind to non-AD tau deposits.
Model System: Patients with CBS or PSP who had in vivo PET imaging and subsequent neuropathological assessment
Statistical Significance: Variable correlation between in vivo and postmortem findings
THK family and AV1451 show similar and highly correlated regional distribution patterns. [11C]PBB3 shows different regional distribution with no correlation with THK tracers, indicating potentially different molecular targets.
Model System: AD patients
Statistical Significance: High correlation between THK and AV1451; no correlation between PBB3 and others
Moderate to strong positive correlation between tau PET and CSF p-tau across CN and patient groups. Discordance patterns: prodromal AD shows more isolated tau PET positivity while AD dementia shows more CSF p-tau elevation. [18F]AV1451 superior to CSF p-tau for diagnosing AD dementia.
Model System: CN subjects, MCI and AD patients, non-AD disorders
Statistical Significance: p<0.05 for correlations in combined analyses
Tau PET closely corresponds to hypometabolism across neocortical regions in AD. Tau binding more closely related to hypometabolism than to MRI results or amyloid-beta levels. Local and distant negative correlations between tau and MRI measures. Tau levels most strongly related to rates of neurodegenerative change.
Model System: CN subjects and AD patients
Statistical Significance: Significant correlations between tau and hypometabolism
Tau tracer binding associated with cognitive measures in prodromal and dementia-stage AD. Episodic memory associated with temporal lobe binding; executive/global cognitive function associated with neocortical binding. Tau binding in MTL associated with episodic memory even in CN individuals.
Model System: CN elderly, MCI, and AD patients
Statistical Significance: Significant associations across multiple studies
Observable increases in tau PET over time. In largest study, tau accumulation rates were rather uniform across brain regions, arguing against stepwise progression suggested by histopathology. Significant tau accumulation over 1 year in amyloid-beta-positive CN individuals.
Model System: CN elderly, MCI, and AD patients
Statistical Significance: Significant longitudinal increases observed
[18F]RO-948 elevated in medial temporal areas and broadly throughout cortex including precuneus, lateral parietal, occipital, and prefrontal cortex in AD vs controls. [18F]MK-6240 binding patterns consistent with Braak staging. Lower off-target binding to MAO-B rich areas compared to first-generation.
Model System: AD patients and healthy controls
Statistical Significance: Significant differences between AD and controls
Four different high-affinity binding sites identified on tau protofibril. AV1451 and MK-6240 bind to site 1 with higher affinity. THK5351 binds strongly to sites 1 and 3. PBB3 binds to all 4 sites with similar affinity.
Model System: Computational modeling using cryo-EM tau filament structures
Statistical Significance: Predicted binding affinities in nanomolar range consistent with experimental data
Excellent correspondence between FTP retention and neurofibrillary tangle distribution in primary AD. All 8 AD patients (Braak VI) showed markedly elevated SUVR in Braak stage regions. Non-Alzheimer tauopathies showed lower, more variable retention than AD but higher than controls. FTLD-TDP and FTLD-FUS showed tracer retention in tau-negative regions. Quantification detected advanced (Braak VI) but not early (Braak I-IV) NFT pathology. Subcortical SUVRs were similar across PSP, CBD, AD, and other pathologies.
Model System: Human postmortem brain tissue with antemortem PET imaging; 20 patients with diverse neurodegenerative diagnoses
Statistical Significance: Mean + 2 SD threshold used to define elevated SUVR above controls; specific p-values not provided
Soleimani-Meigooni et al., (2020)
All patients with PSP and CBD showed elevated SUVR in putamen, globus pallidus, and subthalamic nucleus above control threshold. However, SUVR levels were similar in AD, other non-Alzheimer tauopathies, and FTLD-TDP. Precentral/postcentral gyrus SUVRs were highest in AD. Dentate nucleus SUVR was similar across tauopathies and controls.
Model System: Human patients with PSP, CBD, AD, and other neurodegenerative diseases
Statistical Significance: Mean + 2 SD threshold used
Soleimani-Meigooni et al., (2020)
All patients with Braak VI pathology (n=8) showed elevated SUVR above control threshold. Patients with Braak I-III incidental pathology did not show elevated SUVR in Braak ROIs. PVC recovered signal in some patients with Braak III/IV pathology but also produced false positives in non-Alzheimer tauopathies.
Model System: Human patients with varying Braak stages (0-VI)
Statistical Significance: Mean + 2 SD threshold used
Soleimani-Meigooni et al., (2020)
Five clusters formed capturing tau tangle strain polymorphism across diseases; fAD PSEN1 and fAD APP share tau cluster; sAD and DS partially overlap; PiD tau deposits separate from all other diseases; clear difference observed between fAD and sAD tau tangles despite similar cryo-EM structures
Model System: Human tau tangles from sAD, fAD, DS, and Pick's disease brain samples
Statistical Significance: Inter-disease cluster separation
EMBER distinctly discriminated tau deposits localized to neurons, astrocytes, and oligodendrocytes in PiD; neurons, astrocytes, and oligodendrocytes have distinct conformational strains in same brain; spatial resolution revealed cell-type specific strain variation
Model System: Pick's disease brain sections with tau inclusions in neurons, astrocytes, and oligodendrocytes
Statistical Significance: Cell-type specific cluster separation
Glaucoma was associated with 36% higher PHF-tau levels in occipital cortex but was not statistically significant (OR = 1.36, 95% CI 0.91-2.03, p = 0.13). Weighted back to full cohort, OR was 1.30 (0.81-2.01).
Model System: Human postmortem occipital cortex tissue from ACT cohort (subset)
Statistical Significance: p = 0.13 (not significant)
Participants with Braak stages of IV or greater had elevated FTP-PET signal. FTP-PET was elevated in participants with Alzheimer's disease. Optimal meta-ROI SUVr cut-point of 1.29 was determined with 87% sensitivity and 82% specificity for identifying AD-spectrum diagnoses. Participants with PART and hippocampal sclerosis did not show elevated FTP-PET signal. Secondary AD neuropathology in LBD participants showed borderline elevated FTP-PET signal.
Model System: Human participants from Mayo Clinic Study of Aging (MCSA) and Mayo Clinic Alzheimer's Disease Research Center (ADRC)
Statistical Significance: p=0.008 for AD vs LBD SUVr comparison; rho's from 0.61-0.70, p ≤ 0.002 for IHC correlation
Quantitative measurements of hippocampal and temporal lobe tau burden were highly correlated to FTP-PET signal. Moderately strong correlations observed with rho's from 0.61-0.70 (p ≤ 0.002). Several tissue samples had tau burden in range of 5% or less, showing limitations of biomarker correlations for low levels of tau neuropathology.
Model System: Human postmortem brain tissue from 26 study participants
Statistical Significance: rho's from 0.61-0.70, p ≤ 0.002
Normal aging leads to increases in H4K16ac; AD associated with H4K16ac loss in lateral temporal lobe including near AD susceptibility loci. Tau pathology correlates with H3K9ac dysregulation in up to 23% of all H3K9ac domains. Tau-related H3K9ac alterations cluster in large genomic segments covering several megabase pairs.
Model System: Human AD postmortem brain (lateral temporal lobe, entorhinal cortex, dorsolateral prefrontal cortex)
Statistical Significance: Not specified
Cláudio Gouveia Roque et al., (2024)
Combined set of three miRNAs (miR-27a-3p, miR-30a-5p, miR-34c) and three piRNAs adequately detects AD and predicts MCI to AD conversion. When combined with phospho-tau and Ab42/40 ratio, achieved AUC of 0.98.
Model System: Human CSF samples
Statistical Significance: Not specified
Cláudio Gouveia Roque et al., (2024)
AD-Mimics had more Braak stages III-IV (vs V-VI in Confirmed-AD), more FTLD-tau (18.6% vs 1.6%), FTLD-TDP-43 (21.2% vs 5.0%), gross infarcts (39.2% vs 26.3%), microinfarcts (35.0% vs 26.5%), and hippocampal sclerosis (22.9% vs 12.3%). Unidentified-AD had more neocortical Lewy bodies (32.4% vs 13.7%), FTLD-tau (7.6% vs 1.6%), and FTLD-TDP-43 (9.1% vs 5.0%) but less frequent neuritic plaques (65.6% vs 76.4%).
Model System: Human postmortem brain tissue
Statistical Significance: Most comparisons p<0.05; some became insignificant after Bonferroni correction
Gauthreaux et al., (2020)
FTLD-tau (OR=17.92, p<0.001) and FTLD-TDP-43/other (OR=6.61, p<0.001) were associated with increased odds of AD-Mimics. Lewy body pathology (OR=0.45, p=0.004) was associated with decreased odds. Age at death also increased odds (OR=1.05).
Model System: Human postmortem data from NACC
Statistical Significance: FTLD-tau: p<0.001; FTLD-TDP-43: p<0.001; LBD: p=0.004; age: p<0.001
Gauthreaux et al., (2020)
Lewy body pathology (OR=1.94, p<0.001), FTLD-tau (OR=4.28, p=0.003), and FTLD-TDP-43/other (OR=2.51, p=0.014) were associated with increased odds of Unidentified-AD. Age at death decreased odds (OR=0.96).
Model System: Human postmortem data from NACC
Statistical Significance: LBD: p<0.001; FTLD-tau: p=0.003; FTLD-TDP-43: p=0.014; age: p<0.001
Gauthreaux et al., (2020)
Highly significant correlation between neocortical flortaucipir SUVr and Ptau tissue concentration (Pearson r = 0.81; p < 0.0001). Correlation preserved after adjustment for Aβ (Pearson r = 0.73, p < 0.0001). Modest correlation in limbic regions (Pearson r = 0.52; p = 0.080). No correlation in basal ganglia (Pearson r = 0.27; p = 0.280). Apparent SUVr-Aβ correlation (r = 0.61) disappeared when adjusted for Ptau.
Model System: Human postmortem brain tissue from 3 subjects with dementia (2 with clinical AD diagnosis, 1 with dementia of unknown origin)
Statistical Significance: Neocortical: p < 0.0001; Limbic: p = 0.080; Basal ganglia: p = 0.280
All three subjects had intermediate to high AD neuropathological change. Braak stages: Case 1 (VI/V), Case 2 (IV/V), Case 3 (V/V). Thal phases: Cases 1 and 3 = 4 (high), Case 2 = 3 (intermediate). CERAD: Cases 1 and 3 frequent, Case 2 moderate. No atypical tau pathology identified in any case.
Model System: Human postmortem brain tissue
Statistical Significance: N/A - descriptive neuropathology
All three cases visually interpreted as having AD pattern (τAD) of pathologic tau. Case 2 showed least retention (limited to mesial and lateral temporal). Case 3 showed most intense signal (lateral temporal, occipital, parietal, right-lateralized frontal). Case 1 showed moderate retention (frontal, lateral temporal, parietal).
Model System: Human in vivo PET imaging
Statistical Significance: N/A - qualitative assessment
Strong evidence that tau PET tracers (18F-flortaucipir, 18F-MK6240, 18F-RO948, 18F-PI2620) bind AD tau aggregates in advanced Braak stages (>IV). Accuracy for detecting tau load in Braak V-VI was 87.5% (95% CI, 77.2%-93.5%). Strong correlations (R2 range, 0.66-0.76) between tau PET levels and quantitative neuropathologic tau burden. Tracer binding weaker in non-AD tauopathies and overlaps with off-target regions.
Model System: Human postmortem brain tissue
Statistical Significance: R2 range 0.66-0.76 for AD tau correlations; 87.5% accuracy for Braak V-VI detection
Both increased amyloid and tau PET levels associated with cognitive decline, but relationship predominantly driven by tau. A-positive, T-positive participants demonstrated cognitive decline over 2-year follow-up; one met MCI criteria at 2-year follow-up. A-negative, T-positive participant remained cognitively stable.
Model System: Cognitively unimpaired human participants (PREVENT-AD cohort)
Statistical Significance: When both amyloid and tau included in model, only tau remained significant
Elevated baseline tau PET levels strongly associated with more rapid cognitive decline. Tau PET outperformed amyloid PET and structural MRI in head-to-head comparisons. Visually determined positive 18F-flortaucipir PET scan associated with increased risk for future cognitive decline (MMSE HR 1.68, 95% CI 1.22-2.32) and functional decline (CDR sum of boxes HR 1.40, 95% CI 1.11-1.76) after 18 months. Both intensity and extent of baseline tau PET predictive for future brain atrophy rates.
Model System: Human participants with MCI and AD dementia
Statistical Significance: MMSE hazard ratio 1.68 (95% CI, 1.22-2.32); CDR sum of boxes hazard ratio 1.40 (95% CI, 1.11-1.76)
Tau PET tracers demonstrated excellent diagnostic performance for distinguishing AD dementia from non-AD neurodegenerative disorders with sensitivity and specificity above 90%. Tau PET superior to structural MRI, amyloid PET, and most biofluid markers. Utility at individual-patient level limited for non-AD tauopathies (PSP, CBD, Pick disease). Estimated capability to detect up to 70% of neurodegenerative dementias when correctly implemented.
Model System: Human participants across diagnostic groups (AD dementia, FTD, PSP, CBS, MCI, CU)
Statistical Significance: Sensitivity and specificity above 90% for distinguishing AD dementia from non-AD disorders
Off-target binding profiles vary across tracers. 18F-flortaucipir shows off-target binding in basal ganglia, substantia nigra, longitudinal sinuses, pituitary, choroid plexus. 18F-RO948 and 18F-MK6240 show greater binding to meninges and skull. 18F-PI2620 shows promise for 4R tauopathies with lower off-target binding in basal ganglia.
Model System: Human participants
Statistical Significance: Not specified
PVC improved discriminative accuracy between cognitively impaired and unimpaired individuals cross-sectionally but resulted in less robust longitudinal changes. PVC modestly restores hippocampal signal and correlation between hippocampal signal and clinical symptoms.
Model System: Tau PET data from multiple studies
Statistical Significance: Not specified in detail
Inferior cerebellar gray matter most widely used reference region. Inferior cerebellar reference provided most sensitive measure for cross-sectional group differences. Eroded white matter or eroded white matter cerebellar composite with longitudinal processing pipeline most suitable for longitudinal analyses.
Model System: Tau PET data from multiple cohorts
Statistical Significance: Not specified
Flortaucipir PET predicted B3-level tau pathology with sensitivity 92.3-100% and specificity 52.0-92.0%. Predicted high ADNC with sensitivity 94.7-100% and specificity 50.0-92.3%. Majority read analysis showed 92.3% sensitivity and 80.0% specificity for B3, 94.7% sensitivity and 80.8% specificity for high ADNC. Inter-rater reliability high (Fleiss k=0.74, P<.001). SUVR cutpoint >1.113 yielded 84.2% sensitivity for B3 and 86.5% for high ADNC with 100% specificity.
Model System: Human participants with terminal illness (n=64 in primary cohort; 156 enrolled total; 67 autopsied)
Statistical Significance: P<.001 for inter-rater reliability; 95% CI ranges provided for all sensitivity/specificity values
Addition of supplemental data improved specificity and maintained overall accuracy. Of 26 impaired participants with non-AD clinical diagnosis, 19 had less than high ADNC at autopsy; 16 of these had flortaucipir PET accurately interpreted as not AD pattern.
Model System: Historically collected autopsy cases from academic centers (n=16)
Statistical Significance: Not specified
ESM explained 70.2% of overall spatial variance in tau distribution and 50.9% within individual subjects. Entorhinal cortex confirmed as best epicenter, corroborating autopsy findings. Model predicted spatial patterns of tau regardless of amyloid status.
Model System: Human participants (312 individuals along AD continuum)
Statistical Significance: Global pattern r2 = 0.702 (p < 0.01), individual mean r2 = 0.509, SD = 21.8% (p < 0.01)
Model performance remained high even among Aβ- individuals despite low overall tau burden. Tau patterns resembled early Braak staging even in subjects without elevated amyloid and in APOE4-negative cognitively normal individuals.
Model System: Amyloid-negative human participants
Statistical Significance: High model fit maintained in Aβ- subjects
Underestimated regions (more tau than predicted) had significantly greater amyloid burden. Significant correlation between regional model residuals and regional amyloid levels.
Model System: Human brain regions (66 cortical regions)
Statistical Significance: t = 2.9, p = 0.004; correlation p < 0.001
No population-level hemisphere preference. Individual epicenter hemisphere associated with tau deposition asymmetry. Left-limbic epicenters showed greater left temporoparietal binding; right-limbic epicenters decreased with disease progression. Right-limbic subjects were older.
Model System: Individual human subjects
Statistical Significance: Epicenter hemisphere associated with asymmetry p < 0.001; age effect p = 0.01
Abnormally high cortical 18F T807 binding in MCI and AD patients compared to CN controls. Elevated neocortical 18F T807 binding particularly in inferior temporal gyrus was associated with clinical impairment. The association of cognitive impairment was stronger with inferior temporal 18F T807 than with mean cortical 11C PiB. Regional 18F T807 was correlated with mean cortical 11C PiB among both impaired and control subjects. Patterns corresponded to Braak staging scheme.
Model System: Human participants: clinically normal older adults (N=56), mild cognitive impairment (N=13), and mild AD dementia (N=6)
Statistical Significance: MCI/AD vs CN: p<0.05 to p<0.001 for multiple ROIs; Inferior temporal: rho=-0.83 (p<10^-4) for MMSE in MCI/AD; rho=0.75 (p=0.0002) for CDR-SB in MCI/AD
Gene expression-pseudotime significantly associated with: tau positivity (F=17.64, P<0.001), amyloid positivity (F=28.22, P<0.001), tau-amyloid co-morbidity (F=9.58, P<0.001), brain infarcts (F=5.32, P<0.05), tau-amyloid-infarct co-morbidity (F=7.49, P<0.001), clinical diagnosis (F=56.72, P<0.001), clinical conversion (F=56.61, P<0.001), executive function (R=0.23, P<0.001), memory performance (R=0.27, P<0.001). Subjects with higher pseudotime progressed to advanced disease in average 3.18 years (SD 2.33).
Model System: In vivo blood samples (ADNI database)
Statistical Significance: All P-values FWE-corrected; clinical conversion: P<0.001
Lecanemab demonstrated 27% slowing of disease progression compared to placebo (-0.45 [95%CI: -0.67 to -0.23] CDR-SB) after 18 months. Showed substantial reductions in amyloid plaques and attenuation of insoluble tau deposition. Plasma P-tau181 and GFAP were significantly reduced.
Model System: Human clinical trial - patients with prodromal to mild AD
Statistical Significance: p < 0.001 for primary endpoint
Donanemab showed 29% reduction in disease progression in combined population (-0.7 [95%CI, -0.95 to -0.45] CDR-SB). In individuals with low/intermediate tau, a larger 40% effect was observed (-0.67 [95%CI, -0.95 to -0.40]). 52% of participants stopped treatment when amyloid plaque clearance was achieved.
Model System: Human clinical trial - patients with early symptomatic AD
Statistical Significance: p < 0.001 for primary endpoint
Donanemab demonstrated dramatic reductions in amyloid PET signal in early-stage trials. Used 'Goldilocks' tau PET criterion - requiring enough medial temporal lobe uptake to confirm AD but not too much tau to be past responsiveness to amyloid removal.
Model System: Human clinical trial - early AD patients
Statistical Significance: Not fully specified
Blood tests showed high accuracy in differential diagnosis of AD vs. other dementias. Plasma P-tau181 and P-tau217 are exquisitely sensitive to treatment effects of anti-amyloid therapies, with large reductions reaching steady state 6-12 months after treatment initiation. However, reductions occurred even in trials without clinical efficacy, suggesting sensitivity to small, non-clinically meaningful effects.
Model System: Human subjects and clinical trial participants
Statistical Significance: High discriminative accuracy reported in validation studies
Discovery of quinoline and benzimidazole lead compounds (BF-126, BF-158, BF-170) with tau binding capability but relatively low selectivity over amyloid plaques. Brain uptake was good (11.3% ID/g at 2 min) but clearance was slow (3.1% ID/g at 30 min). EC50 values >200 nM with high nonspecific binding.
Model System: Synthetic heparin-induced tau polymers (HITPs) and postmortem AD human brain homogenates (AD-PHFs)
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
THK-523 showed higher in vivo retention in tau transgenic mice vs amyloid model mice. In AD patients, higher cortical retention vs healthy controls with distribution matching PHF histopathology. However, high white matter retention hindered clear visualization and precluded further development.
Model System: Synthetic HITP tau fibrils, AD brain homogenates, rTg4510 tau transgenic mice, APP/PS1 amyloid mice
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Both compounds showed higher tau affinities than THK-523 with Kd values of 2.6 nM (THK-5105) and 5.2 nM (THK-5117) for AD-PHF. Higher Ab/tau selectivity. Distribution in mesial temporal sections coincided with Gallyas-Braak staining and tau immunostaining but not PiB. THK-5117 showed faster kinetics and better SNR than THK-5105. Tracer retention associated with dementia severity and brain atrophy.
Model System: Synthetic HITP tau, human AD-PHF tau aggregates, human AD brain sections
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
High binding affinity for tau in AD brain homogenates (Kd = 2.9 nM) with faster kinetics and lower white matter retention. Peak SUVR achieved within 5 min post-injection, exiting brain by 90 min. SUVR stabilized at ~1.5 in controls and ~2.5 in AD patients at 50 min. Distinguished AD from controls. Also tested in PSP patient showing uptake in midbrain.
Model System: Human AD brain homogenates, human AD brain sections, 16 healthy controls, 5 MCI subjects, 13 AD patients
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Good tau affinity (Kd [HITP] = 2.55 nM) and 50-fold selectivity for tau over Ab. Greater accumulation in medial temporal region of AD patients vs controls. Captured spreading of neurofibrillary tau from limbic system to neocortical areas. Distribution different from 11C-PiB. Tested in CBD patient showing tracer retention in basal ganglia.
Model System: HITPs, PS19 and rTg4510 transgenic mice, human AD brain sections, human subjects
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Both tracers showed high affinity for tau (Kd = 14.6 nM for T807, 22 nM for T808), >25-fold selectivity for AD-PHF over Ab, lack of white matter binding. Good correlation with tau immunostaining but not Ab. Linear correlation between NFT loads and T807 autoradiography intensity across 26 human donors. T807 showed rapid brain uptake (4.16% ID/g at 2 min) and clearance. Clinical PET showed accumulation matching PHF distribution with increasing neocortical spread associated with dementia severity.
Model System: Human AD brain sections, APPSWE-PS19 transgenic mice, human subjects
Statistical Significance: Linear correlation between NFT burden and cognitive scores confirmed
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Tau pathology appears in hippocampus, parahippocampus, and entorhinal cortex in early dementia stages. Increased tau in inferior temporal lobe associated with worse memory. CSF tau levels correlated with tau imaging in 6 brain regions consistent with Braak staging. Test/retest reproducibility ~4-5%. ~10% year-over-year increase in mean cortical SUVR in high amyloid burden subjects. Increased tau accompanied by lower MMSE performance.
Model System: Human subjects from Harvard Aging Brain Study (75 older subjects)
Statistical Significance: Statistically significant correlations between CSF tau and tau PET in entorhinal/parahippocampal regions, inferior temporal, middle temporal, and superior temporal cortices
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Regional tau PET patterns agreed with regions underlying clinical symptoms in atypical AD. FDG PET abnormally low where tau high. In PPA, tau signal spreading through language networks. In PSP, positive signals in substantia nigra, globus pallidus, and subthalamic nucleus, but some overlap with controls limiting early-stage detection.
Model System: PSP patients, primary progressive aphasia patients, atypical AD patients
Statistical Significance: p < 0.001 uncorrected for PSP vs controls in specific brain regions
Hartmuth C. Kolb, José Ignacio Andrés (2017)
T807 perfectly colocalized with tau-containing neurons and dystrophic neurites. Confirmed strong T807 binding to tau pathology in AD but not to cerebral amyloid, DLB, MSA, or TDP-43. Tangles and dystrophic neurites account for most of the in vivo T807 signal.
Model System: Human AD brain tissue
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
All three compounds displayed high-affinity binding to tau in sub-10-nM range with strong selectivity over Ab in human brain tissue. They displaced T808 binding in human brain slices at Braak stages V/VI. Binding colocalized with aggregated tau IHC but not Ab. Baboon PET showed good brain permeability, washout, and low white matter binding. Phase I trial ongoing in HCs and AD patients.
Model System: Human brain tissue, baboon PET studies
Statistical Significance: N/A
Hartmuth C. Kolb, José Ignacio Andrés (2017)
Amygdala shows high tau-PET signal in early stages. Earliest tau-PET uptake in amygdala ~10 years before AD diagnosis. Amygdala, ERC, and hippocampus show largest annual increase in tau-PET uptake in Aβ-positive cognitively unimpaired individuals.
Model System: Human subjects (cognitively unimpaired, MCI, AD patients)
Statistical Significance: Not specified in detail in this review
Cases showed: (1) pathology often exists around the periphery of amygdalae near meninges and/or lateral ventricle; (2) peri-amygdaloid grey matter including entorhinal cortex frequently shows pathologies; (3) cortical and transitional regions are vulnerable; (4) phospho-Tau pathology is constant in all aged individuals. Cases 3-6 showed comorbid Ab, Tau, a-synuclein, and TDP-43 pathologies.
Model System: Human postmortem amygdala tissue (UK-ADC autopsy cohort)
Statistical Significance: Not applicable (descriptive case series)
Nelson et al., (2018)
TEBM identified biomarker event sequence over mean period of 17.3 years (95% CI: 11.4-27.1 years). Order: CSF/PET Aβ abnormality first (~0.03 years), followed by tau markers (~2.65-2.7 years), hippocampal volume (~5.1 years), entorhinal volume (~5.7 years), first cognitive changes (~7.5 years), mid-temporal volume (~10.4 years), and ventricular abnormality. Validated with OASIS dataset showing same ordering within 95% CIs.
Model System: Human data from ADNI (Alzheimer's Disease Neuroimaging Initiative) and OASIS (Open Access Series of Imaging Studies)
Statistical Significance: 95% confidence intervals reported for all event timings
Successfully estimated subject-specific brain-wide influences due to Aβ, tau and Aβ∙tau on neuronal activity; quantified group differences (AD vs CU) via rank sum tests
Model System: Human subjects from TRIAD cohort (47 cognitively unimpaired, 16 AD patients)
Statistical Significance: Non-parametric rank sum tests used
Sanchez-Rodriguez et al., (2023)
Identified 756 genes associated with Aβ effects, 650 genes for tau effects, and 1987 genes for Aβ∙tau synergistic effects on neuronal activity; included AD-risk genes (CD33, ADAM10, SNCA, CLU)
Model System: Human postmortem brain tissue from Allen Human Brain Atlas (6 neurotypical adult brains)
Statistical Significance: 99% bootstrap CI not including zero
Sanchez-Rodriguez et al., (2023)
Top enriched pathways: neuroinflammation/immune system processes, developmental processes (sensory organ development, tissue morphogenesis), cell communication/transport mechanisms (regulation of secretion, vesicle-mediated transport); immune pathways particularly overrepresented in Aβ∙tau set
Model System: Gene sets from transcriptomic analysis
Statistical Significance: q < 0.05, FDR-corrected
Sanchez-Rodriguez et al., (2023)
Aβ molecular associates enriched in pyramidal cells (q=0.002, δ=3.804) and endothelial-mural cells (q=0.008, δ=2.950); tau associates enriched in interneurons (q=0.021, δ=2.834); Aβ∙tau signature highly enriched in microglia (q<0.001, δ=10.425)
Model System: Gene sets from transcriptomic analysis with single-cell data from mouse somatosensory cortex and hippocampus CA1
Statistical Significance: q < 0.05
Sanchez-Rodriguez et al., (2023)
rs2288911 is a microglia-specific eQTL for APOE expression (P=1.1x10^-13). Associated with cerebral amyloid angiopathy (CAA) (P=1.18x10^-7) but not AD pathology or amyloid/tau proteinopathy. Association with CAA persists after adjusting for APOE4 (P=9.9x10^-6). No interaction with APOE4 on CAA.
Model System: Human DLPFC microglia from ROS/MAP participants
Statistical Significance: P=1.1x10^-13 for eQTL; P=1.18x10^-7 for CAA association
The DPS model explains 62% of total variance, significantly outperforming single biomarkers (ADAS: 49.4%, MMSE: 46%, p<0.01). RAVLT30 (Rey Auditory Verbal Learning Test) was identified as the earliest biomarker to become abnormal. Hippocampal volume (HIPPO), ABETA, and TAU precede cognitive tests (MMSE and ADAS). Mean ADPS at baseline: N=-0.03, MCI=2.85, AD=6.49. Rate of progression increases with disease severity: N=-0.08/year, MCI=0.76/year, AD=1.46/year.
Model System: Human ADNI cohort (ages 55-90): 687 subjects with 3658 visits total - 1103 Normal, 1513 MCI, 1010 AD diagnoses
Statistical Significance: p<0.01 for DPS vs ADAS comparison; p<0.01 for RAVLT30 preceding all other biomarkers; p<0.02 for HIPPO, ABETA, TAU preceding MMSE and ADAS
Model successfully reproduced Braak stages of Alzheimer's disease progression: initial seeding in entorhinal cortex (stages I-II), progression to hippocampus and brain stem, then to limbic system (stages III-IV), and finally to cortical association areas and primary sensory-motor regions (stages V-VI)
Model System: 2D finite element model from T2-weighted MRI of 32-year-old male brain (sagittal slice)
Statistical Significance: N/A - computational model
Model reproduced characteristic progression patterns for: (a) amyloid-β in Alzheimer's disease, (b) tau inclusions in Alzheimer's disease, (c) α-synuclein in Parkinson's disease, (d) TDP-43 in ALS. All showed generic progression trends: fast progression through limbic system once near ventricles, cortical invasion, and late parietal lobe involvement
Model System: 3D finite element whole brain model from T2-weighted MRI
Statistical Significance: N/A - computational model
LC neuronal loss averages 63% in AD; LC volume decreases by 8.4% per Braak stage increase; 8% of LC neurons are p-tau-positive at Braak stage 0, doubling by Braak stage I, reaching 100% by Braak stage VI; rostral portion affected more severely (83% loss) compared to middle (23%) and caudal (15%) parts
Model System: Human postmortem brain tissue
Statistical Significance: Significant correlation between LC volume loss and Braak stage (p<0.05)
p-tau begins to accumulate in LC early in life, in some cases as young as 10 years of age; 90% of individuals have some tau pathology in LC by age 30; 72% of individuals aged 31-40 years have tau lesions; 94% of individuals aged 41-50 years have tau lesions
Model System: Human postmortem brain tissue
Statistical Significance: Not reported
Four significant function clusters: (1) APP metabolism/Aβ formation (P=4.56e-7); (2) Tau protein binding (P=3.19e-5); (3) Lipid metabolism (P=1.45e-7); (4) Immune response (P=6.32e-5). Highly significant correlation between common and rare variant results (P=1.32e-7)
Model System: Gene-set analysis from GWAS summary statistics
Statistical Significance: q <= 0.05 after FDR correction
Amyloid-beta positive participants had higher tau levels in both amygdala and entorhinal cortex (p < 0.001); they also had smaller amygdala volumes than Aß- participants (p < 0.05). Within Aß+ groups, elevated amygdala tau was associated with lower amygdala volumes (ß=-445.198, p=0.01). Both amygdala (ß=-2.908, p=0.02) and entorhinal tau (ß=-3.347, p=0.02) were significant predictors of MMSE scores, but only in Aß+ participants. Amygdala volume was a significant predictor of memory among Aß+ participants (ß=0.001, p=0.01).
Model System: Human participants - cognitively unimpaired older adults from ADNI cohort (Aß- = 145, Aß+ = 81)
Statistical Significance: p < 0.001 for tau levels in Aß+; p < 0.05 for amygdala volume; p = 0.01 for amygdala tau-volume relationship; p = 0.02 for tau predicting MMSE; p = 0.01 for volume predicting memory
18F-AV-1451 demonstrates high selective binding and affinity to tau protein aggregates (15 nM Kd in brain slices, 0.7 nM in purified PHF-tau) with relatively spare binding to normal monomeric tau. Showed approximately 29 times greater binding to tau aggregates compared to beta-amyloid in AD brain gray matter. 'Tau-rich' brains (significant tau tangles with beta-amyloid) showed increased uptake in gray matter; 'tau-poor' brains and healthy controls did not. Normal brain showed only diffuse background activity. In 40 subjects, worse memory performance was associated with greater PET tau binding in entorhinal cortex. No significant serious safety concerns reported.
Model System: Human brain tissue (ex vivo), human subjects (AD patients and controls)
Statistical Significance: Correlation between memory performance and entorhinal cortex tau binding noted
rs2288911 is a microglia-specific eQTL for APOE expression (P=1.1x10^-13). Associated with cerebral amyloid angiopathy (CAA) (P=1.18x10^-7) but not AD pathology or amyloid/tau proteinopathy. Association with CAA persists after adjusting for APOE4 (P=9.9x10^-6). No interaction with APOE4 on CAA.
Model System: Human DLPFC microglia from ROS/MAP participants
Statistical Significance: P=1.1x10^-13 for eQTL; P=1.18x10^-7 for CAA association
Model successfully reproduced Braak stages of Alzheimer's disease progression: initial seeding in entorhinal cortex (stages I-II), progression to hippocampus and brain stem, then to limbic system (stages III-IV), and finally to cortical association areas and primary sensory-motor regions (stages V-VI)
Model System: 2D finite element model from T2-weighted MRI of 32-year-old male brain (sagittal slice)
Statistical Significance: N/A - computational model
Four significant function clusters: (1) APP metabolism/Aβ formation (P=4.56e-7); (2) Tau protein binding (P=3.19e-5); (3) Lipid metabolism (P=1.45e-7); (4) Immune response (P=6.32e-5). Highly significant correlation between common and rare variant results (P=1.32e-7)
Model System: Gene-set analysis from GWAS summary statistics
Statistical Significance: q <= 0.05 after FDR correction