# Critical Evaluation of Neurodegeneration Therapeutic Hypotheses
---
## Hypothesis 1: TREM2 Microglial Activation
**Original Confidence: 0.78 → Revised: 0.62**
### Weak Links
- **Dose-dependency assumption unexamined.** TREM2 signaling has a documented biphasic character — agonistic antibodies at high concentrations can cause receptor internalization and desensitization (Painter et al., 2018). The therapeutic window for 4D9 agonism is not established in the primary literature.
- **Mouse model confounding.** The 5xFAD/Trem2−/− cross is problematic as a therapeutic-test platform: deleting TREM2 entirely removes the receptor target, so an agonistic antibody has nothing to act on. A heterozygous knockdown or conditional knockout would be more appropriate.
- **DAM model oversimplified.** The single-cell transcriptomics literature (Keren-Shaul et al., 2017; Mancuso et al., 2019) now identifies at least four microglial states beyond the binary DAM framework. Attributing pathology solely to failed DAM1→DAM2 transition is reductionist.
### Counter-Evidence
- Paradoxically, *loss* of TREM2 in some 5xFAD crosses reduces plaque burden (Wang et al., 2015), complicating the directional assumption that more TREM2 signaling = better clearance.
- TREM2 R47H effects on ligand binding affinity are contested; some biophysical studies suggest the variant retains near-wildtype binding to APOE and phospholipids, implying the risk mechanism may not be simply through ligand engagement.
- Human CSF sTREM2 shows a biphasic temporal pattern in AD ( Suárez-Calvet et al., 2016) — elevated early, suppressed late — suggesting that global TREM2 activation could be harmful at certain disease stages.
### Falsifying Experiments
1. Test whether 4D9 agonistic activity is confirmed in a cell-based assay (e.g., calcium flux in Trem2-expressing BaF3 cells) before animal testing — if it lacks agonist activity, the premise collapses.
2. Generate *conditional* Trem2 knockout mice crossed to 5xFAD, so microglial deletion can be timed post-plaque formation to test whether restoration of TREM2 after pathology onset is still beneficial.
3. Single-cell RNA-seq of microglia after treatment to confirm DAM gene induction (Cst7, Lpl, Trem2) rather than relying solely on morphological plaque metrics.
4. If the therapeutic produces cognitive benefit but no change in plaque load, the mechanism needs reconsideration — the hypothesis currently tightly couples plaque compaction to cognition.
---
## Hypothesis 2: LRRK2 G2019S Lysosomal Enhancement
**Original Confidence: 0.72 → Revised: 0.58**
### Weak Links
- **Mechanistic specificity gap.** The proposed link between LRRK2 kinase hyperactivation and *lysosomal* acidification impairment is inferred from RAB GTPase dysregulation but lacks a defined biochemical chain. LRRK2 G2019S predominantly phosphorylates RAB10 and RAB12; the connection to lysosomal v-ATPase assembly is indirect and contested.
- **Incomplete animal model fidelity.** G2019S knock-in mice show minimal spontaneous α-synuclein pathology without additional stressors (Herzing et al., 2018). This undermines the assumption that LRRK2 hyperactivation is sufficient to impair lysosomal clearance.
- **Species-specific toxicity barrier.** LRRK2 inhibitor toxicity in non-human primate lung (toxicology studies cited in BIIB122/PNI-149 development) is a major translational concern that the hypothesis does not address.
### Counter-Evidence
- The LRRK2 G2019S mutation has incomplete penetrance and variable expressivity — many carriers reach old age without PD, suggesting that G2019S alone is not a deterministic driver.
- In some studies, LRRK2 G2019S patient-derived neurons do not show marked lysosomal deficits compared to isogenic controls (Sandner et al., 2021).
- BIIB122 (dnakine) is being developed primarily for idiopathic PD, not specifically for G2019S carriers, indicating the field has de-prioritized the mechanistic hypothesis.
### Falsifying Experiments
1. Directly measure lysosomal pH (lysosensor imaging) and cathepsin activity in G2019S knock-in neurons before and after LRRK2 inhibitor treatment — if lysosomal function is not impaired at baseline, the mechanism is unsupported.
2. Test whether the G2019S PFF cross produces worse pathology than PFF alone — if it does not, LRRK2 G2019S is not a meaningful potentiator.
3. Include a behavioral readout that is sensitive to enteric nervous system pathology (e.g., gastrointestinal transit time) to determine whether peripheral pathology drives central spread.
4. Measure LRRK2 kinase activity in patient-derived neurons vs. isogenic controls to confirm hyperactivation at the relevant biochemical node.
---
## Hypothesis 3: FUS Nuclear Import Restoration
**Original Confidence: 0.68 → Revised: 0.55**
### Weak Links
- **Nuclear/cytoplasmic ratio as surrogate endpoint.** High-content imaging of the N/C ratio is a morphological readout that does not capture functional restoration of FUS-dependent splicing. Splicing may be disrupted by cytoplasmic FUS aggregation through mechanisms other than nuclear loss (e.g., RNA granule sequestration of splicing factors).
- **The "Transportin-1 binding deficit" mechanism is not the primary pathogenic event in most ALS-FUS cases.** The strongest evidence is for FUS P525L, which disrupts the NLS directly. For other ALS-linked FUS mutations (R521C, R522G), the Transportin-1 hypothesis is less supported.
- **Organoid/spinal organoid modeling limitations.** Spinal organoids lack mature motor neuron electrophysiology and do not replicate the chronicity of ALS progression.
### Counter-Evidence
- ALS-FUS patient neurons show stress granule formation and liquid-liquid phase separation abnormalities as early or earlier events than nuclear import disruption.
- The proposed "synaptic and mitochondrial gene splicing" disruption has not been systematically validated by RNA-seq in purified motor neurons from FUS-ALS patients.
- ASO approaches for FUS mutations have so far focused on knockdown (which may be counterproductive given FUS's essential cellular roles), not import restoration.
### Falsifying Experiments
1. Perform RNA-seq (not just qPCR for candidate genes) on motor neurons from FUS P525L iPSCs after hit compounds to identify which splicing events are actually restored.
2. Test whether compounds that increase nuclear FUS ratio also reduce stress granule pathology — if they don't, the N/C ratio is an insufficient proxy.
3. Include a survival assay in spinal organoids at a timepoint where mutant FUS motor neurons show >30% reduced viability vs. isogenic controls. Compounds that merely shift the N/C ratio without improving survival are mechanistically insufficient.
4. Validate hits in a non-iPSC system (e.g., primary rodent motor neurons) to exclude artifacts from reprogramming.
---
## Hypothesis 4: PINK1/Parkin Mitophagy Activation
**Original Confidence: 0.65 → Revised: 0.52**
### Weak Links
- **Mechanism relevance to sporadic PD is unproven.** PINK1 and Parkin mutations cause autosomal recessive early-onset PD — this is a minority of PD cases (~2-3% combined). Whether pharmacological mitophagy enhancement applies to idiopathic PD is a major extrapolation.
- **Urolithin A specificity.** Urolithin A is a gut microbiome-derived metabolite that induces general autophagy and mitochondrial biogenesis via PGC-1α and AMPK activation. It does not directly activate the PINK1/Parkin pathway. Calling it a "mitochondrial-targeted small-molecule activator of the PINK1/Parkin pathway" overstates its mechanistic specificity.
- **Mouse model does not replicate human PD.** PINK1 knockout mice do not show the robust dopaminergic neuron loss seen in PINK1-deficient humans. The behavioral and histological phenotype is subtle, making therapeutic rescue difficult to interpret.
### Counter-Evidence
- Urolithin A has been studied primarily in C. elegans and aged rats; the magnitude of effect on dopamine neuron survival in a specific PD model is not established.
- PINK1-deficient flies show dramatic mitochondrial pathology, but mammalian models do not phenocopy this, suggesting evolutionary divergence in pathway necessity.
- PINK1 and Parkin are activated by mitochondrial membrane potential loss — this is inherently a damage-response mechanism, not a homeostatic one. Forcing activation in healthy neurons may produce mitochondrial remodeling that is not benign.
### Falsifying Experiments
1. Treat *wild-type* mice with urolithin A to establish whether it causes any mitochondrial abnormalities or "excessive" mitophagy in non-damaged neurons.
2. Compare urolithin A with a direct PINK1 activator (if such a compound exists) to determine whether the PINK1/Parkin pathway is specifically engaged vs. general autophagy.
3. Include mitochondrial complex I activity assays in the ventral midbrain — if urolithin A improves behavior via PINK1-independent pathways, the mechanistic claim must be revised.
4. Test whether the CatWalk motor improvement correlates with TH+ neuron survival or only with functional mitochondrial metrics. If behavior improves without neuroprotection, the therapeutic mechanism is different from the hypothesis.
---
## Hypothesis 5: C9orf72 ASO Treatment
**Original Confidence: 0.82 → Revised: 0.78**
### Weak Links (relatively few — this is the strongest hypothesis)
- **C9 haploinsufficiency vs. toxic gain-of-function unresolved.** The hypothesis invokes both DPR toxicity and C9 protein haploinsufficiency, but ASOs primarily reduce repeat transcripts. The relative contribution of each to neurodegeneration is unknown, and ASO effects on DPR may not address any haploinsufficiency component.
- **TDP-43 pathology is downstream, not necessarily reversible.** Even if ASOs reduce repeat transcripts, TDP-43 inclusions may represent a point-of-no-return beyond which nuclear TDP-43 localization is insufficient to restore splicing. The window of reversibility has not been mapped.
- **Clinical trial biomarker limitations.** NCT04165729 demonstrated safety and CSF poly(GP) reduction, but poly(GP) is a proxy for transcription — whether it tracks neurodegeneration or clinical progression is unproven.
### Counter-Ev