Version history

{
  "session_id": "sess_SDA-2026-04-04-gap-tau-prop-20260402003221",
  "round_number": 3,
  "agent_persona": "persona-domain_expert",
  "agent_backend": "minimax-M2.7",
  "action": "debate",
  "content": "# Domain Expert Position: Tau Propagation Mechanisms and Therapeutic Interception Points\n\n## The Interception Point Problem: An Industry-Wide Target Validation Failure\n\nThe anti-tau immunotherapy field has suffered a series of high-profile clinical failures that demand explanation beyond \"wrong patient population\" or \"too late intervention.\" Gosuranemab (Biogen/AbbVie), semorinemab (Genentech/AC Immune), and tilavonemab (AbbVie) all failed Phase 2 or Phase 3 trials in Alzheimer's disease despite compelling preclinical rationale (Cummings et al., 2019, PMID: 31062825). As a domain expert, I argue these failures reflect a fundamental misunderstanding of **which tau species** should be targeted and **when** during the propagation cascade.\n\nThe evidence is now clear that tau propagates via multiple distinct mechanisms operating in parallel: vesicular pathways involving exosomes and extracellular vesicles carrying LAMP2A/HSP70-dependent tau cargo (Xu et al., 2025, PMID: 40187566), non-vesicular free tau aggregates, and potentially tunneling nanotubes. Polanco and colleagues (2023, PMID: 36316035) used CRISPRi screening to identify regulators shared between exosomal and vesicle-free tau propagation pathways, revealing that these mechanisms are mechanistically intertwined but genetically separable. The heterogeneity of tau species in human tauopathies is now well-documented (Leroux et al., 2022, PMID: 34563677), with distinct EV populations showing different cargo compositions depending on the specific tauopathy subtype.\n\n## The Skeptic's Challenge Is Partially Valid\n\nThe skeptic correctly identifies that the histopathological \"spread\" pattern is correlative rather than definitively causative. Sequential neurofibrillary tangle involvement from entorhinal cortex to downstream regions is *consistent with* transcellular propagation but does not *prove* it—prion-like spread is one of several hypotheses. Alternative explanations include \"common vulnerability\" (regions with shared transcriptional profiles become affected independently) or \"transneuronal spread\" (a related but mechanistically distinct process). The field's conflation of these possibilities has led to therapeutic strategies targeting extracellular tau when the causal pathology may originate intracellularly.\n\nHowever, the evidence from Wang et al. (2017, PMID: 28086931) demonstrating that neuronal exosome-derived human tau is toxic to recipient mouse neurons in vivo provides more direct evidence for a functional consequence of transcellular transfer than the skeptic acknowledges. The recent identification of specific KFERQ-like motifs mediating tau incorporation into exosomes via LAMP2A and HSP70 (Xu et al., 2025) represents the kind of mechanistic precision the skeptic demands.\n\n## A Path Forward: Polypharmacology at Defined Propagation Nodes\n\nI propose that the field's failure is not evidence against propagation as a therapeutic target, but against the **single-node intervention** strategy. The MAPT protein (UniProt: P10636) is a 758-amino acid microtubule-associated protein with at least 15 documented post-translational modification sites, suggesting that therapeutic modulation must account for context-dependent conformational states.\n\nThe most promising interception points are now becoming clear:\n1. **Intracellular seeding inhibition**: Small molecules targeting the tau repeat domain (R1-R4) that blocks template-directed misfolding\n2. **Secretion suppression via EV pathway manipulation**: Targeting the LAMP2A/HSP70 axis or syntenin-dependent exosome biogenesis\n3. **Uptake receptor blockade**: Heparan sulfate proteoglycans and LDLR family members have been implicated in tau internalization\n\nTau aggregation inhibitors like LMTX (methylene blue derivative from TauRx) have proceeded furthest clinically, with mixed Phase 3 results suggesting narrow therapeutic windows. The lesson is that aggregation inhibitors must be administered before significant seed accumulation occurs—a window that is essentially closed at clinical presentation.\n\n**Confidence: 0.75** — The heterogeneity of propagation mechanisms is well-established, but their relative contributions in human disease versus preclinical models remain contested. My confidence is higher for the mechanisticbiology than for therapeutic translation feasibility.",
  "tokens_used": "67091",
  "persona_id": "persona-skeptic",
  "confidence": 0.75,
  "argument": "# Domain Expert Position: Tau Propagation Mechanisms and Therapeutic Interception Points\n\n## The Interception Point Problem: An Industry-Wide Target Validation Failure\n\nThe anti-tau immunotherapy field has suffered a series of high-profile clinical failures that demand explanation beyond \"wrong patient population\" or \"too late intervention.\" Gosuranemab (Biogen/AbbVie), semorinemab (Genentech/AC Immune), and tilavonemab (AbbVie) all failed Phase 2 or Phase 3 trials in Alzheimer's disease despite",
  "evidence": " compelling preclinical rationale (Cummings et al., 2019, PMID: 31062825). As a domain expert, I argue these failures reflect a fundamental misunderstanding of **which tau species** should be targeted and **when** during the propagation cascade.\n\nThe evidence is now clear that tau propagates via multiple distinct mechanisms operating in parallel: vesicular pathways involving exosomes and extracellular vesicles carrying LAMP2A/HSP70-dependent tau cargo (Xu et al., 2025, PMID: 40187566), non-vesicular free tau aggregates, and potentially tunneling nanotubes. Polanco and colleagues (2023, PMID: 36316035) used CRISPRi screening to identify regulators shared between exosomal and vesicle-free tau propagation pathways, revealing that these mechanisms are mechanistically intertwined but genetically separable. The heterogeneity of tau species in human tauopathies is now well-documented (Leroux et al., 2022, PMID: 34563677), with distinct EV populations showing different cargo compositions depen",
  "data_evidence": "{\"tool_call_count\": 8, \"tools_used\": [\"open_targets_associations\", \"search_trials\", \"uniprot_protein_info\", \"paper_corpus_search\", \"search_trials\", \"gwas_genetic_associations\", \"search_trials\", \"paper_corpus_search\"]}"
}