Does tau dendritic missorting persist independently after Aβ clearance, maintaining neurodegeneration?

arXiv Preprint NeurIPS Nature Methods PLOS ONE

---

🌐 Open in Overleaf →

Title: Fyn-anchored dendritic tau becomes self-sustaining after transient Aβ exposure

Mechanism: Aβ oligomers drive tau missorting from axon to dendritic spines, where tau binds FYN and stabilizes an NMDA receptor-associated excitotoxic signaling complex. Once established, this tau-Fyn-PSD95/NMDAR scaffold may persist without continued Aβ, maintaining calcium dysregulation, spine loss, and downstream degeneration. Target gene/protein/pathway: MAPT (tau), FYN, PSD95/DLG4, NMDAR/SRC-family signaling Supporting evidence: Strong prior literature links dendritic tau to Fyn-dependent excitotoxicity in hippocampal pyramidal neurons; Aβ is upstream in many models, but the unresolved point is persistence after Aβ withdrawal. Falsifiable experiment: In primary mouse or human iPSC-derived cortical neurons, expose cultures briefly to Aβ oligomers, then immunodeplete/remove Aβ completely. Track dendritic tau, Fyn localization, spine calcium, and cell viability over 2-4 weeks. Test whether delayed Fyn inhibition or tau reduction reverses pathology after Aβ removal. Persistence despite Aβ clearance but reversal with Fyn/tau intervention would support the hypothesis. Confidence: 0.76

Title: Local dendritic tau hyperphosphorylation locks in missorting through a GSK3β-CDK5 feedback loop

Mechanism: Aβ initiates tau entry into dendrites, but persistence is maintained by local kinase activation, especially GSK3B and CDK5/p25, which continually phosphorylate tau at missorting-associated epitopes and reduce its axonal microtubule affinity. This creates a cell-autonomous phospho-tau maintenance state independent of ongoing extracellular Aβ. Target gene/protein/pathway: MAPT, GSK3B, CDK5, CAPN1, p25 Supporting evidence: Tau phosphorylation by GSK3β/CDK5 is well established in AD-relevant systems, and calpain-p25 signaling can remain active after an initiating insult. Falsifiable experiment: Use compartmentalized microfluidic neuron cultures with transient Aβ exposure confined to dendrites. After washout, quantify dendritic phospho-tau and kinase activity. Apply delayed GSK3β inhibitors, CDK5 blockade, or calpain inhibitors only after Aβ removal. If phospho-tau missorting persists and is selectively reversed by kinase suppression, the hypothesis is supported. Confidence: 0.72

Title: Dendritic tau missorting persists because it seeds local proteostatic failure in spines and dendrites

Mechanism: Mislocalized tau impairs local autophagy-lysosome and endosomal trafficking in dendrites, causing accumulation of tau species that are no longer dependent on Aβ input. Once dendritic clearance capacity fails, tau remains trapped in the somatodendritic compartment and continues synaptotoxic signaling. Target gene/protein/pathway: MAPT, RAB5, RAB7, LAMP1, TFEB, autophagy-lysosome pathway Supporting evidence: AD models show early endosomal/autophagic dysfunction and tau accumulation in neurites; a plausible extension is that transient Aβ exposure pushes dendrites into a self-maintaining clearance-deficient state. Falsifiable experiment: After transient Aβ treatment and complete washout, measure dendritic endosome/lysosome flux, tau turnover, and spine density. Rescue with TFEB activation, lysosomal acidification enhancers, or endosomal trafficking correction. If tau missorting persists after Aβ clearance but normalizes when dendritic proteostasis is restored, this hypothesis gains support. Confidence: 0.68

Title: Microglia sustain Aβ-independent dendritic tau toxicity through complement-mediated synapse pruning

Mechanism: Aβ may initiate tau missorting, but activated microglia then maintain degeneration via C1q/C3-CR3 complement signaling, selectively pruning tau-burdened synapses even after plaques or soluble Aβ are reduced. In this model, the persistent driver is microglial state, not Aβ itself. Target gene/protein/pathway: C1QA/C1QB/C1QC, C3, ITGAM/CR3, TREM2, TYROBP; microglia-complement pathway Supporting evidence: Complement-mediated synapse loss and TREM2-dependent microglial remodeling are strongly implicated in AD; this offers a concrete post-Aβ persistence mechanism. Falsifiable experiment: In an Aβ-bearing tau model, clear Aβ pharmacologically or genetically, then longitudinally assay dendritic tau, synapse number, and microglial engulfment. Add delayed C1q, C3, or CR3 blockade after Aβ clearance. If neurodegeneration continues after Aβ removal but is reduced by complement inhibition, this supports the model. Confidence: 0.74

Title: Astrocytic glutamate-handling failure maintains tau-mislocalized dendritic stress after Aβ clearance

Mechanism: Aβ exposure induces a persistent reactive astrocyte state with reduced EAAT2/SLC1A2-mediated glutamate uptake, increasing extrasynaptic NMDA receptor activation. Dendritic tau then continues to amplify excitotoxic signaling even when Aβ is gone. The persistent lesion is a neuron-astrocyte circuit state. Target gene/protein/pathway: SLC1A2/EAAT2, GRIN2B-containing NMDARs, MAPT, astrocyte reactivity pathways Supporting evidence: Astrocyte glutamate dysregulation and extrasynaptic NMDAR toxicity are well supported in neurodegeneration, and tau is known to modulate excitotoxic vulnerability. Falsifiable experiment: In neuron-astrocyte co-cultures or iPSC tri-cultures, apply transient Aβ, remove it, then measure astrocytic glutamate uptake, extrasynaptic NMDAR signaling, dendritic tau, and neuron survival. Test delayed EAAT2 upregulation or NR2B-selective antagonism after Aβ washout. Rescue despite absent Aβ would support this mechanism. Confidence: 0.64

Title: Tau missorting becomes Aβ-independent when it transitions into locally seeded oligomer propagation

Mechanism: Aβ acts as an initiator, but once dendritic tau oligomers/seeds form, they template further local tau misfolding and spread trans-synaptically between connected excitatory neurons. Continued degeneration is then driven by tau seeding kinetics rather than Aβ presence. Target gene/protein/pathway: MAPT oligomerization/seeding pathway; heparan sulfate proteoglycan-dependent uptake Supporting evidence: Tau seeding and trans-neuronal propagation are established in multiple tauopathy models. The specific hypothesis here is that dendritic missorting is the transitional state that enables persistent local seeding after an Aβ trigger. Falsifiable experiment: Induce transient Aβ exposure in a connected neuronal network, then remove Aβ and assay tau seeding with biosensor cells or seeding assays from isolated dendritic fractions. Apply delayed anti-tau seed antibodies or block tau uptake after Aβ clearance. Continued spread after Aβ removal, blocked by anti-seeding interventions, would support the hypothesis. Confidence: 0.79

Title: BIN1-dependent trafficking failure stabilizes somatodendritic tau after Aβ has been cleared

Mechanism: Aβ may initially perturb tau localization, but persistence depends on altered BIN1-regulated membrane trafficking and endocytic sorting that prevents tau re-entry into the axon and favors dendritic retention. This would connect AD genetic risk directly to the persistence problem. Target gene/protein/pathway: BIN1, MAPT, endocytosis/membrane curvature trafficking pathways Supporting evidence: BIN1 is a strong AD risk gene and has been linked to tau biology, trafficking, and neuronal membrane dynamics, making it a plausible determinant of whether missorting resolves or persists. Falsifiable experiment: In human iPSC neurons with BIN1 knockdown, overexpression, or AD-risk alleles, trigger transient Aβ exposure and then remove Aβ. Quantify recovery versus persistence of tau polarity over time. If BIN1 perturbation specifically controls post-Aβ persistence of dendritic tau, the hypothesis is supported. Confidence: 0.61

If useful, I can convert these into SciDEX-ready hypothesis records with shorter fields and a ranking by therapeutic tractability.

Below the central skeptical point applies to all seven: current literature strongly supports that Aβ can induce dendritic/somatodendritic tau missorting and synaptic toxicity, but it does not cleanly establish that the state becomes truly Aβ-independent after complete Aβ removal. Most cited evidence is either acute Aβ exposure, constitutive transgenic overexpression, or end-stage human tissue, which cannot separate “self-sustaining tau pathology” from “residual upstream injury,” incomplete Aβ clearance, or generic degeneration.

Fyn-anchored dendritic tau self-sustains after transient Aβ

• Weak evidence: Tau is required for Aβ/Fyn/NMDAR toxicity, but the key evidence mainly shows tau enables Aβ toxicity, not that the tau-Fyn scaffold persists after Aβ is gone. The classic mechanistic support is upstream of Aβ exposure, not post-clearance persistence.
• Alternative mechanisms: Persistent calcium dysregulation could reflect irreversible spine injury, residual oligomeric Aβ, or broader membrane/trafficking damage rather than a self-maintained tau-Fyn complex.
• Translational risks: Fyn inhibitors have had limited clinical traction; postsynaptic rescue in rodent neurons may not translate once human circuits have substantial synapse loss.
• Falsifying experiment: Pulse human iPSC cortical neurons with labeled Aβ oligomer, then verify near-zero residual Aβ by orthogonal assays. If dendritic tau/Fyn/PSD95 colocalization and calcium pathology collapse once Aβ is truly removed, the hypothesis fails.

GSK3β-CDK5 feedback locks in missorting

• Weak evidence: This is biologically plausible, but one key primary neuron study found local cdk5 activation with Aβ-induced missorting while GSK3β was not clearly changed in that acute setting, so the proposed dual-kinase maintenance loop may be overstated.
• Alternative mechanisms: MARK/BRSK-driven polarity failure, microtubule loss, or generic stress signaling may explain persistence better than a dedicated GSK3β-CDK5 bistable loop.
• Translational risks: Kinase inhibitors are pleiotropic; rescue could reflect broad suppression of stress responses rather than specific reversal of a tau-maintenance circuit.
• Falsifying experiment: After transient Aβ, remove Aβ and use inducible, compartment-specific suppression of GSK3B and CDK5 separately. If missorting persists despite confirmed kinase shutdown, the maintenance-loop model is weakened.

Proteostatic failure in dendrites traps tau after Aβ

• Weak evidence: Autophagy/endolysosomal dysfunction is common in AD, but direct evidence that transient Aβ exposure alone creates a durable, dendrite-localized clearance defect sufficient to maintain missorting is limited.
• Alternative mechanisms: Proteostasis defects may be downstream consequences of tau accumulation, aging, APOE state, or lysosomal stress, not the primary persistence driver.
• Translational risks: TFEB/autophagy manipulations often improve many proteotoxic states; a positive rescue would not uniquely support this hypothesis.
• Falsifying experiment: Measure tau turnover in dendrites after Aβ washout using pulse-chase labeling plus compartment-resolved lysosomal flux. If tau clearance normalizes while missorting persists, proteostatic failure is not the main maintenance mechanism.

Microglia/complement sustain degeneration after Aβ clearance

• Weak evidence: Strong evidence supports complement-dependent synapse pruning in AD and microglial neurodegeneration in tauopathy, but that does not prove microglia maintain tau missorting itself after Aβ removal. This may explain ongoing synapse loss better than ongoing tau polarity failure.
• Alternative mechanisms: Microglia could simply amplify damage initiated by tau or amyloid, while neuronal intrinsic tau seeding or excitotoxicity remains primary.
• Translational risks: Microglia are state-dependent and can be protective; broad complement blockade may impair host defense and synapse remodeling.
• Falsifying experiment: Clear Aβ, then deplete or reprogram microglia after missorting is established. If dendritic tau polarity remains abnormal despite reduced pruning/inflammation, the hypothesis only explains downstream degeneration, not persistence of missorting.

Astrocytic EAAT2 failure maintains post-Aβ dendritic stress

• Weak evidence: Astrocyte glutamate dysregulation is credible, but the direct chain “transient Aβ -> lasting astrocyte EAAT2 failure -> persistent dendritic tau missorting” is not well demonstrated. Much of the EAAT2 literature is correlational or end-stage tissue.
• Alternative mechanisms: Extrasynaptic NMDAR stress could be secondary to neuronal tau pathology, microglial cytokines, seizure-like network activity, or general astrocyte reactivity rather than a specific Aβ-imprinted astrocyte state.
• Translational risks: EAAT2 upregulation and NR2B antagonism are systemically messy and may not reverse established structural degeneration.
• Falsifying experiment: In tri-culture, replace astrocytes after Aβ washout with naive astrocytes while keeping neurons constant. If tau missorting persists unchanged, astrocytic persistence is not required.

Tau missorting transitions into local tau seeding/propagation

• Weak evidence: This is one of the more plausible persistence mechanisms because tau seeding is robustly supported, but the specific bridge from dendritic missorting to an autonomous seeded state after a transient Aβ trigger is still mostly inferential.
• Alternative mechanisms: Continued spread could still depend on low residual Aβ, inflammatory mediators, or overexpression artifacts in tau-seeding systems.
• Translational risks: Anti-seeding therapies may need very early timing and may not rescue neurons already committed to degeneration.
• Falsifying experiment: Use endogenous-tau neurons, transient Aβ exposure, rigorous Aβ depletion, then test whether lysates from post-washout neurons seed biosensor cells. If seeding activity disappears when Aβ is removed or when new tau synthesis is blocked, the “autonomous seeded state” claim weakens.

BIN1-dependent trafficking failure stabilizes somatodendritic tau

• Weak evidence: This is the weakest mechanistically. BIN1 is a strong AD risk locus and linked to tau biology, but direct evidence that BIN1 governs resolution versus persistence of post-Aβ tau missorting is sparse.
• Alternative mechanisms: BIN1 effects may be indirect through endocytosis, membrane curvature, myelination, or cell-type-specific functions unrelated to dendritic tau retention.
• Translational risks: Human BIN1 biology is isoform- and cell-type-specific; iPSC perturbations may not model the relevant neuronal context.
• Falsifying experiment: Perform allele-edited isogenic iPSC neuron comparisons with endogenous tau and transient Aβ. If BIN1 genotype changes baseline trafficking but not the rate of tau re-polarization after complete Aβ removal, this hypothesis fails.

Bottom line
The most defensible persistence candidates are tau seeding/propagation (6) and microglia-mediated downstream degeneration (4), with Fyn/NMDAR (1) and kinase maintenance (2) as plausible but not yet cleanly demonstrated post-Aβ states. BIN1 (7) is the most speculative. The main experimental trap across all seven is failing to prove true Aβ absence; without that, “Aβ-independent persistence” remains unproven.

Primary sources

• Zempel et al., 2010, Aβ oligomers induce tau missorting, local Ca2+ rise, spine loss: https://pubmed.ncbi.nlm.nih.gov/20826658/
• Ittner et al., 2010, dendritic tau mediates Aβ toxicity via Fyn/NMDAR pathway: https://pubmed.ncbi.nlm.nih.gov/20655099/
• Ripoli et al., 2014, soluble Aβ oligomers drive tau mislocalization to spines and receptor signaling deficits: https://pubmed.ncbi.nlm.nih.gov/24713000/
• Noble et al., 2003, Cdk5 as a driver of tau aggregation in vivo: https://pubmed.ncbi.nlm.nih.gov/12765608/
• Rajendran and Paolicelli, 2018, microglia-mediated synapse loss in AD: https://pubmed.ncbi.nlm.nih.gov/29563239/
• Shi et al., 2019, microglia drive APOE-dependent neurodegeneration in tauopathy: https://pubmed.ncbi.nlm.nih.gov/31601677/
• Nogueras-Ortiz et al., 2023, tau-oligomer-containing synapse elimination by microglia/astrocytes in AD tissue: https://pubmed.ncbi.nlm.nih.gov/37812432/
• Gerson and Kayed, 2013, tau oligomer seed formation and propagation: https://pubmed.ncbi.nlm.nih.gov/23882255/
• Vasconcelos et al., 2016, Aβ can heterotypically seed tau propagation: https://pubmed.ncbi.nlm.nih.gov/26739002/
• De Rossi et al., 2017, BIN1 and tau pathology relationship is not straightforward: https://pubmed.ncbi.nlm.nih.gov/29479533/

If you want, I can convert this into a SciDEX-ready table with fields for `support_score`, `novelty_score`, `killer_experiment`, and `skeptic_priority_rank`.

Most investable survivors are `6 > 4 > 1 > 2`. I would not spend serious translational budget yet on `7`, and I would treat `3` and `5` as modifier mechanisms rather than lead programs.

| Rank | Hypothesis | Druggability | Biomarkers | Best model systems | Safety / translational risk | Realistic path |
|---|---|---|---|---|---|---|
| 1 | `6` Tau missorting transitions into autonomous tau seeding | High, relative to others. Clear intervention classes: anti-tau antibodies, seed-blocking biologics, ASOs, uptake blockers. | CSF/plasma p-tau217, p-tau181, MTBR-tau, tau seeding assays, tau PET, synaptic markers like NfL/neurogranin. | Human iPSC excitatory neuron networks, microfluidic compartments, seeded organoids, APP+tTAU mouse combinations with amyloid withdrawal. | Main issue is timing: likely only works early. Anti-tau antibodies may show modest efficacy if seeds are mostly intracellular. | Strong preclinical package possible in 18-30 months; IND-grade program 3-5 years; roughly `$15M-$40M` for serious preclinical-to-IND effort depending on modality. |
| 2 | `4` Microglia/complement sustain post-Aβ degeneration | Moderate. Targets exist: C1q, C3, CR3, TREM2-state modulators. Biology is druggable, but CNS immunology is tricky. | CSF/plasma YKL-40, sTREM2, C1q/C3 fragments, SV2A PET for synapses, tau PET, NfL. | Human tri-cultures, xenografted human microglia mice, amyloid-plus-tau models with delayed amyloid clearance, spatial transcriptomics. | Biggest risk is on-target immunologic liability and blocking beneficial pruning/repair. Likely better for slowing synapse loss than reversing tau polarity. | Mechanistically testable in 12-24 months; translational de-risking 2-4 years; `$10M-$25M` for platform-quality preclinical program. |
| 3 | `1` Fyn-anchored dendritic tau/NMDAR complex persists after Aβ | Moderate. Fyn is druggable, but prior CNS kinase efforts have struggled. Tau-lowering may be better than direct Fyn inhibition. | Phospho-NMDAR/SRC signatures in CSF are weak clinically; better to pair CSF tau markers with EEG/network hyperexcitability, SV2A PET, synaptic fluid biomarkers. | Primary neurons are useful, but human iPSC neuron-astrocyte co-cultures and MEA readouts are more relevant; inducible Aβ pulse-washout paradigms are essential. | Narrow therapeutic window, CNS kinase tolerability, and risk that rescue only reflects generic anti-excitotoxic effects. | Good mechanistic program in 12-18 months; hard to make into differentiated clinical asset unless paired with tau-lowering; `$8M-$20M` preclinically. |
| 4 | `2` GSK3β/CDK5 feedback maintains missorting | Low-moderate. Targets are druggable in principle, but specificity and therapeutic index are poor. Best use is mechanistic validation, not near-term drug program. | p-tau epitope panels, kinase activity proxies, phosphoproteomics, NfL. | Compartmentalized neurons, live-cell phospho-tau imaging, inducible kinase perturbation after washout. | Pleiotropy is the problem. GSK3β/CDK5 inhibition affects many essential functions, so safety margin is likely weak for chronic AD use. | Fast falsification in 9-15 months; not attractive as a standalone development thesis unless it points to a more specific downstream node; `$3M-$8M` for decisive mechanistic work. |

My call on the others:

• `3` Proteostasis failure is biologically credible but too nonspecific for first-pass translation. Good as a downstream readout, weak as a lead therapeutic thesis.
• `5` Astrocytic EAAT2 failure is plausible and may matter in subgroups with hyperexcitability, but it is unlikely to be the core persistence mechanism.
• `7` BIN1 should be treated as a genetic modifier study, not a drug program.

Best biomarker package across the whole question:

• Entry/target engagement: amyloid PET or soluble Aβ measures proving true clearance.
• Persistence biology: CSF/plasma p-tau217, MTBR-tau, tau seeding assay.
• Damage state: NfL, neurogranin, SV2A PET.
• Mechanism-specific add-ons:
• `4`: sTREM2, C1q/C3, microglial imaging/transcriptomics.
• `1`: EEG/MEA hyperexcitability, calcium imaging.
• `2`: phosphoproteomics.
• `6`: biosensor seeding assays from conditioned media and cell fractions.

Best experimental strategy:

Build one rigorous `pulse Aβ -> verified complete removal -> delayed intervention` assay.

Run it first in human iPSC neuron monoculture and tri-culture.

Prioritize endpoints that separate `tau polarity`, `tau seeding`, and `ongoing degeneration`.

Advance only mechanisms that remain abnormal after orthogonally proving near-zero residual Aβ.

If you want, I can turn this into a SciDEX-ready table with `tractability_score`, `trial_readiness_score`, `killer_experiment`, `timeline_years`, and `budget_range_usd`.