Version history

{
  "session_id": "sess_SDA-2026-04-25-gapdebate-afba1a80bd_c747c608",
  "round_number": 2,
  "agent_persona": "persona-skeptic",
  "agent_backend": "codex_cli",
  "action": "critique",
  "content": "Overall skeptical read: the debate is probably mixing three separable variables that have not been cleanly orthogonalized experimentally: `location`, `ligand identity`, and `receiver-cell state`. The strongest evidence supports synaptic C1q/C3/CR3-mediated pruning in development and AD models, but that does not by itself prove a distinct microglial surface-signaling program for C1q, nor a binding-partner hierarchy that dominates location. Much of the translational logic is still mouse-heavy and disease-model dependent. Key anchors: synaptic pruning by C1q/C3 in development ([PMID: 18083105](https://pubmed.ncbi.nlm.nih.gov/18083105/)); early AD-model synapse loss via C1q/C3/CR3 ([PMID: 27033548](https://pmc.ncbi.nlm.nih.gov/articles/PMC5094372/)); direct `Aβ-C1q` binding and complement activation ([PMID: 8176223](https://pubmed.ncbi.nlm.nih.gov/8176223/), [PMID: 11714802](https://pubmed.ncbi.nlm.nih.gov/11714802/)); noncanonical C1q receptor signaling outside the CNS via `LAIR-1` ([PMID: 23093673](https://pmc.ncbi.nlm.nih.gov/articles/PMC3503216/)); `TREM2-C1q` interaction in neurodegeneration ([PMID: 37442133](https://pubmed.ncbi.nlm.nih.gov/37442133/)); cell-type/synapse-class selectivity in AD mice ([PMID: 37118504](https://pubmed.ncbi.nlm.nih.gov/37118504/)); neuronal pentraxins as C1q partners ([PMID: 33628204](https://pubmed.ncbi.nlm.nih.gov/33628204/)); ApoE-C1q binding in inflammation ([PMID: 30692699](https://pubmed.ncbi.nlm.nih.gov/30692699/)).\n\n1. `Synaptic C1q prunes; microglial C1q signals state`\nWeak evidence: the synaptic arm is well supported, but the microglial “surface-associated C1q reprograms microglia independent of cascade” arm is much less directly shown in CNS microglia. Most receptor-signaling evidence for C1q comes from peripheral myeloid cells, not resident brain microglia. Alternative mechanism: apparent “microglial C1q signaling” could just reflect autocrine/paracrine exposure to downstream complement fragments, Fc signals, or generic phagocytosis of opsonized cargo. Translational risk: a therapy built on this split may fail if microglia do not actually segregate into cascade-dependent and cascade-independent C1q programs in human brain. Falsify by forcing equal ligand identity and equal microglial state while varying only C1q location; if microglial-tethered C1q does not induce a distinct transcriptional/phagocytic program when downstream complement is genetically blocked, the claim weakens sharply.\n\n2. `Binding partner identity matters more than location`\nWeak evidence: this is plausible, but the hierarchy claim is mostly asserted rather than demonstrated. There is direct evidence for `Aβ` and neuronal pentraxin binding to C1q, but not a general proof that ligand identity dominates spatial context across CNS substrates. Alternative mechanism: microenvironmental geometry, membrane density, local complement regulators, and which cell sees the complex first may matter more than ligand chemistry alone. Translational risk: “ligand-selective” targeting is hard because C1q is promiscuous and disease tissue contains mixed complexes. Falsify with a factorial experiment that independently varies `ligand`, `location`, and `receiver cell`; if location explains as much variance as ligand identity in complement activation or engulfment, this hypothesis is overstated.\n\n3. `Astrocyte thrombospondin-rich ECM creates silent C1q tagging`\nWeak evidence: this is the weakest hypothesis in the set. The support is indirect and risks conflating canonical C1q with the broader C1q/TNF family or thrombospondin-related synaptogenic biology. I do not see strong primary evidence that `THBS1/2` are bona fide CNS C1q-binding scaffolds that impose a low-inflammatory state. Alternative mechanism: perisynaptic ECM effects may be mediated by altered synapse maturation, glial access, or complement regulator distribution rather than a specific `C1q-THBS` complex. Translational risk: high chance of chasing an interaction that is not central in vivo. Falsify by direct biochemical binding and in situ proximity assays first; if endogenous brain `C1q-THBS1/2` complexes are not robustly detectable, the model should be dropped before therapeutic reasoning.\n\n4. `APOE isoform reshapes the C1q binding landscape`\nWeak evidence: ApoE and C1q can interact, but the leap from that to isoform-specific synaptic versus plaque C1q partitioning is still speculative. The causal chain `APOE4 -> altered C1q partner landscape -> more pruning/inflammation` is not yet cleanly established. Alternative mechanism: ApoE isoforms could alter plaque compaction, lipid trafficking, or microglial activation upstream of C1q rather than through direct C1q-complex remodeling. Translational risk: ApoE effects are pleiotropic, so a C1q-centered intervention may miss the dominant ApoE biology. Falsify by swapping ApoE isoforms while holding `Aβ` burden, lipidation state, and microglial genotype constant, then testing whether C1q interactomes and functional outputs still diverge.\n\n5. `TREM2 state decides whether C1q tagging is adaptive or toxic`\nWeak evidence: this one is stronger than most because there is direct evidence that TREM2 binds C1q and restrains complement-mediated synapse loss. But the “same tagged substrate, divergent outcome solely by TREM2 state” framing is still too simple; astrocytes and other receptors also shift the outcome. Alternative mechanism: TREM2 effects may mainly alter microglial metabolism, survival, clustering, or plaque handling, with C1q being one branch rather than the master switch. Translational risk: TREM2 modulation can have stage-specific and pathology-specific effects, so combining it with C1q targeting may not generalize across AD, tauopathy, and human aging. Falsify by presenting identical C1q-opsonized substrates to `TREM2 WT` and `KO` microglia in a complement-defined system while measuring uptake efficiency versus inflammatory amplification; if TREM2 loss changes only uptake kinetics but not injury-state transition, the claim is overstated.\n\n6. `Block C1r/C1s, spare C1q recognition`\nWeak evidence: conceptually attractive, but it assumes C1q recognition is mostly beneficial and downstream protease activation is mostly harmful in the CNS. That partition is not settled. Some harmful effects may arise from C1q binding itself, while some downstream activation may aid debris handling. Alternative mechanism: pathology may depend on amount, chronicity, and local regulator failure, not simply the recognition-versus-activation step. Translational risk: selective `C1r/C1s` blockade might preserve some unwanted C1q receptor signaling, or impair host defense enough to erase CNS benefit. Falsify by head-to-head comparison of `anti-C1q` versus `C1r/C1s` inhibition in the same model with matched target engagement, including infection susceptibility and plaque/debris handling, not just synapse counts.\n\n7. `C1q has distinct inhibitory-versus-excitatory synapse roles`\nWeak evidence: there is evidence of cell-type/synapse-class selective engulfment in AD mouse models, but that does not prove C1q itself has fundamentally different biochemical programs at inhibitory versus excitatory synapses. The selectivity could arise downstream from glial preference, anatomy, or activity state. Alternative mechanism: synapse-class vulnerability may be driven by network activity, local ECM, or astrocyte-versus-microglia division of labor rather than synapse-intrinsic C1q coding. Translational risk: circuit findings from one model or brain region may not generalize to human disease. Falsify with purified inhibitory and excitatory synaptosomes matched for damage state and exposed to identical C1q/complement conditions; if class selectivity disappears outside the intact circuit, the effect is probably contextual rather than intrinsic.\n\nBottom line: hypotheses `1`, `5`, and `6` are the most testable and biologically grounded; `2` is plausible but underspecified; `3` is currently too conjectural; `4` and `7` may be partly true but are likely over-attributing causality to C1q instead of broader glial/pathology context. The cleanest decisive experiment is a full-factorial design that independently manipulates `C1q location`, `binding partner`, `complement activation competence`, and `microglial state` in one humanized system.",
  "tokens_used": "2084",
  "persona_id": "persona-skeptic"
}