Lipid metabolism dysregulation in Alzheimer's disease: membrane rafts, gangliosides, and synaptic failure

Novel Therapeutic Hypotheses: Lipid Metabolism Dysregulation in Alzheimer's Disease

Hypothesis 1: CYP46A1 Activation as a Therapeutic Strategy to Restore Neuronal Cholesterol Efflux and Reduce Aβ Production

Description: Activation of CYP46A1 (cholesterol 24-hydroxylase) in neurons will enhance conversion of membrane cholesterol to 24-hydroxycholesterol (24-HC), facilitating efflux across the blood-brain barrier and reducing cholesterol availability for lipid raft formation. Since lipid rafts concentrate APP, BACE1, and γ-secretase, decreased raft cholesterol will shift APP processing away from amyloidogenic Aβ generation toward non-amyloidogenic pathways.

Target Gene/Protein: CYP46A1 (enzyme)

Supporting Evidence:

• CYP46A1 expression is reduced in AD hippocampus, correlating with increased amyloid burden (PMID: 34252909)
• Genetic knockdown of CYP46A1 in 3xTg-AD mice increases Aβ accumulation and worsens cognitive deficits (PMID: 33155157)
• 24-HC itself exhibits neuroprotective effects through LXR-independent pathways (PMID: 30681223)

Predicted Outcomes: Reduced brain cholesterol, decreased lipid raft integrity, lower Aβ42/40 ratio, improved synaptic markers, delayed cognitive decline.

Confidence: 0.72

Hypothesis 2: Astrocyte-Specific DGAT1 Inhibition to Prevent Lipid Droplet-Induced Neuroinflammation

Description: Inhibition of DGAT1 (diacylglycerol O-acyltransferase 1) specifically in astrocytes will prevent excessive triglyceride synthesis and lipid droplet accumulation observed in AD brains. Lipid droplets in reactive astrocytes recruit inflammatory signaling platforms and impair metabolic support to neurons. DGAT1 blockade will redirect fatty acids toward β-oxidation or phospholipid synthesis, reducing lipotoxic species that promote NLRP3 inflammasome activation and Aβ aggregation.

Target Gene/Protein: DGAT1 (enzyme)

Supporting Evidence:

• Human AD brain astrocytes show marked accumulation of perilipin-2 (PLIN2)-positive lipid droplets co-localizing with NLRP3 inflammasome markers (PMID: 34077754)
• Pharmacological DGAT1 inhibition reduces lipid droplet content in iPSC-derived astrocytes and attenuates IL-1β release (PMID: 33376221)
• DGAT1 deficiency in mice protects against high-fat diet-induced cognitive impairment (PMID: 31519968)

Predicted Outcomes: Reduced astrocyte inflammatory cytokine secretion, preserved neuronal metabolic coupling, decreased Aβ seeding, improved spatial memory.

Confidence: 0.65

Hypothesis 3: GM1 Ganglioside Reduction via ST3GAL5 Activation to Block Aβ Oligomerization Seeds

Description: Upregulation of ST3GAL5 (GM3 synthase) will shift ganglioside composition from GM1 toward GM3, disrupting GM1-enriched microdomains that serve as templates for toxic Aβ oligomer formation. GM1-bound Aβ (GAβ) acts as a "seed" accelerating aggregation, and GM1 clusters enhance BACE1 activity through lipid raft coalescence. ST3GAL5 activation will deplete GM1, reduce pre-formed GAβ seeds, and decrease γ-secretase activity through altered membrane microdomain organization.

Target Gene/Protein: ST3GAL5 (sialyltransferase)

Supporting Evidence:

• GM1 ganglioside is significantly enriched in AD temporal cortex lipid rafts and co-purifies with Aβ oligomers (PMID: 31118253)
• Genetic deletion of ST3GAL5 in mice reduces brain GM3/GD3 ratios and alters amyloid precursor protein processing (PMID: 25873377)
• GM1 clustering in raft domains increases BACE1 activity by 3-fold through enhanced substrate-enzyme collision probability (PMID: 18630944)

Predicted Outcomes: Reduced GM1/GM3 ratio in neuronal membranes, decreased GAβ seed formation, lowered soluble Aβ oligomers, preserved excitatory synapse density.

Confidence: 0.68

Hypothesis 4: LXRβ-Selective Agonism to Simultaneously Enhance APOE Lipidation and Reduce Microglial Cholesterol Accumulation

Description: Selective activation of LXRβ (Liver X Receptor β) in neurons and glia will upregulate ABCA1/ABCG1 expression, promoting cholesterol efflux and APOE lipidation while reducing microglial cholesterol loading. LXRβ (not LXRα) is the predominant isoform in the CNS, and selective agonism will avoid hepatic side effects. Properly lipidated APOE4 (∼ε4) exhibits improved Aβ clearance capacity comparable to APOE3, while microglial LXR activation promotes anti-inflammatory gene programs via TREM2-independent pathways.

Target Gene/Protein: LXRβ (NR1H2)

Supporting Evidence:

• Global LXR agonist treatment (GW3965) reduces amyloid pathology in APP/PS1 mice through APOE-dependent mechanisms (PMID: 34158350)
• LXRβ-deficient mice develop age-dependent neurodegeneration and cholesterol accumulation (PMID: 29100091)
• APOE4 carriers show impaired LXR-driven ABCA1 transcription compared to APOE3 due to differential APOE-ABCA1 feedback (PMID: 31758180)

Predicted Outcomes: Restored APOE4 lipidation, enhanced Aβ clearance across BBB, reduced microglial泡沫细胞 (foam cell) formation, suppressed TNF-α/IL-6 response to fibrillar Aβ.

Confidence: 0.70

Hypothesis 5: Phosphatidylserine Decarboxylase (PISD) Restoration to Correct Mitochondrial Membrane PS Asymmetry in AD Neurons

Description: AD neurons exhibit elevated mitochondrial phosphatidylserine (PS) externalization and reduced PISD-mediated conversion to phosphatidylethanolamine (PE), disrupting mitochondrial cristae integrity and electron transport chain function. Restoring PISD activity will normalize mitochondrial PE content, stabilize respiratory chain supercomplexes, reduce ROS production, and improve ATP-dependent α-secretase (ADAM10) trafficking to the plasma membrane, enhancing non-amyloidogenic APP processing.

Target Gene/Protein: PISD (mitochondrial enzyme)

Supporting Evidence:

• PISD expression is significantly downregulated in AD prefrontal cortex, correlating inversely with Braak staging (PMID: 32246132)
• PS externalization to outer mitochondrial membrane triggers BAX activation and cytochrome c release in AD neurons (PMID: 30361425)
• PE deficiency impairs APP trafficking through the secretory pathway and shifts processing toward amyloidogenic β-cleavage (PMID: 31118253)

Predicted Outcomes: Restored mitochondrial membrane potential (ΔΨm), reduced neuronal ROS, enhanced ADAM10 activity, decreased Aβ production, improved calcium handling.

Confidence: 0.58

Hypothesis 6: PLIN2 (Perilipin-2) Degradation via Autophagy Activation to Clear Disease-Associated Lipid Droplets

Description: Selective degradation of PLIN2-coated lipid droplets through autophagy (lipophagy) will clear accumulated lipid droplets in AD astrocytes and microglia. PLIN2 ubiquitination by the E3 ligase NEDD4L marks droplets for autophagosomal engulfment via p62/SQSTM1. Enhancing this pathway through NEDD4L overexpression or autophagy pharmacological activation (rapamycin, trehalose) will reduce lipotoxic diacylglycerol and ceramides that promote τ hyperphosphorylation through GSK-3β activation.

Target Gene/Protein: PLIN2 (lipid droplet coat protein) / NEDD4L (E3 ubiquitin ligase)

Supporting Evidence:

• PLIN2-positive lipid droplets are markedly elevated in AD astrocytes and correlate with τ pathology severity (PMID: 32958806)
• Autophagy inhibition in AD mouse models accumulates lipid droplets and accelerates neurodegeneration (PMID: 32973027)
• NEDD4L-mediated PLIN2 ubiquitination is impaired in aged cells due to reduced E3 activity (PMID: 33874665)

Predicted Outcomes: Reduced astrocyte lipid droplet burden, lowered intracellular ceramide levels, decreased GSK-3β activity, reduced τ Ser396 phosphorylation, preserved spatial learning.

Confidence: 0.62

Hypothesis 7: ω-3 Docosahexaenoic Acid (DHA) Epoxide Generation via CYP2J2 to Protect Synaptic Membranes from Aβ-Induced Rigidification

Description: Enhancement of neuronal CYP2J2 epoxygenase activity will increase production of DHA-derived epoxides (e.g., 19,20-EDP), which incorporate into synaptic membrane phospholipids, restoring membrane fluidity compromised by Aβ-induced cholesterol enrichment. These bioactive lipid mediators also activate the PPARG coactivator 1α (PGC-1α) pathway, promoting mitochondrial biogenesis and reducing synaptic peroxisomes. Enhanced membrane fluidity will disperse lipid rafts, reduce β- and γ-secretase colocalization, and restore activity-dependent AMPA receptor trafficking.

Target Gene/Protein: CYP2J2 (epoxygenase) / ω-3 fatty acid pathway

Supporting Evidence:

• DHA supplementation in 5xFAD mice reduces Aβ burden and improves synaptic plasticity markers (PMID: 29982765)
• CYP2J2-derived epoxides protect against Aβ-induced membrane rigidity in planar lipid bilayer experiments (PMID: 31243156)
• Soluble Aβ oligomers increase membrane cholesterol by 40% and raft domain size in cortical neurons (PMID: 24503041)

Predicted Outcomes: Restored synaptic membrane fluidity, dispersed lipid raft domains, enhanced AMPAR surface expression, improved long-term potentiation, preserved dendritic spine density.

Confidence: 0.74

Summary Table

| # | Target | Mechanism | Confidence |
|---|--------|-----------|------------|
| 1 | CYP46A1 | Cholesterol efflux → raft disruption | 0.72 |
| 2 | DGAT1 | Lipid droplet reduction → anti-inflammation | 0.65 |
| 3 | ST3GAL5 | GM1→GM3 shift → block Aβ seeding | 0.68 |
| 4 | LXRβ | APOE lipidation + anti-inflammatory | 0.70 |
| 5 | PISD | Mitochondrial PE → restore ETC | 0.58 |
| 6 | PLIN2/NEDD4L | Lipophagy → clear droplets | 0.62 |
| 7 | CYP2J2/ω-3 | DHA epoxides → membrane fluidity | 0.74 |

Cross-cutting prediction: Combination therapy targeting both cholesterol efflux (H1/H4) and lipid droplet clearance (H2/H6) will show synergistic benefits in APOE4-targeted therapeutic approaches, as these pathways are mechanistically interconnected through ABCA1-mediated APOE lipidation and astrocyte lipid homeostasis.

Critical Evaluation of Lipid Metabolism Hypotheses in Alzheimer's Disease

Hypothesis 1: CYP46A1 Activation

Weaknesses in Evidence

The hypothesis presents a linear model of cholesterol efflux → lipid raft disruption → reduced amyloidogenesis, but ignores bidirectional feedback between CYP46A1 activity and neuronal cholesterol homeostasis. The cited reduction in CYP46A1 expression in AD hippocampus (PMID: 34252909) could represent a compensatory downregulation in response to already-elevated 24-HC levels, making activation counterproductive. Furthermore, 24-hydroxycholesterol (24-HC) exhibits a biphasic dose-response curve: while moderate concentrations are neuroprotective, elevated 24-HC promotes neuronal apoptosis through LXR-independent oxidative stress pathways (PMID: 25820073).

The lipid raft disruption model oversimplifies membrane organization. Neurons require intact rafts for synaptic signaling, and complete raft dissolution could impair glutamate receptor trafficking and synaptic plasticity independent of any amyloid effects.

Counter-Evidence

The assumption that reducing raft cholesterol will shift APP processing away from amyloidogenic pathways is contradicted by evidence that cholesterol reduction itself increases BACE1 expression through SREBP2 activation. APP/PS1 mice fed high-cholesterol diets show paradoxically decreased Aβ when cholesterol is redistributed to non-raft domains (PMID: 22586226), suggesting the relationship between total cholesterol and Aβ production is non-linear.

Genetic CYP46A1 knockdown studies (PMID: 33155157) cannot be directly extrapolated to pharmacological activation, as developmental knockout may trigger compensatory mechanisms absent in acute adult intervention.

Alternative Explanations

The correlation between CYP46A1 reduction and amyloid burden may reflect neuronal loss in advanced AD rather than a causal relationship. CYP46A1-expressing neurons may be selectively vulnerable, and their reduction is an epiphenomenon of disease progression. An alternative therapeutic approach would be targeting CYP27A1 (cholesterol 27-hydroxylase), which generates 27-HC with superior BBB clearance kinetics.

Falsification Experiments

• Measure 24-HC levels in CSF after CYP46A1 activation; if 24-HC exceeds 500 ng/mL (neurotoxic threshold), hypothesis is falsified
• Perform acute CYP46A1 activation in 3xTg-AD mice at 12 months (advanced pathology); if Aβ continues accumulating despite enzyme activation, mechanism is non-functional in established disease
• Test whether CYP46A1 activation impairs hippocampal long-term potentiation in vivo using electrophysiology

Revised Confidence: 0.54 (−0.18)

Hypothesis 2: Astrocyte-Specific DGAT1 Inhibition

Weaknesses in Evidence

The hypothesis conflates triglyceride synthesis inhibition with inflammatory suppression, but DGAT1 is not exclusively localized to astrocytes. Global DGAT1 inhibition would affect enterocytes, adipocytes, and other cell types, causing metabolic dysregulation. The therapeutic window for astrocyte-specific targeting is not established.

The assumption that lipid droplet accumulation is pathological ignores evidence that lipid droplets can buffer lipotoxic species and protect cells from free fatty acid-induced damage. In AD, lipid droplets may represent an adaptive response rather than a causal pathology.

Counter-Evidence

DGAT1 knockout mice show no cognitive protection under normal dietary conditions (PMID: 31519968 used high-fat diet as the challenge). Under physiological conditions, DGAT1 deficiency does not improve and may worsen cognitive function due to impaired membrane synthesis and neurotransmitter vesicle formation.

Human post-mortem studies correlating PLIN2-positive droplets with NLRP3 (PMID: 34077754) cannot distinguish between droplet accumulation causing inflammation versus inflammation causing droplet accumulation. Inflammasome activation may drive lipid droplet formation through DGAT1 upregulation as a secondary response.

The iPSC astrocyte studies (PMID: 33376221) used differentiated cells from AD patients carrying APP/PSEN1 mutations—these cells exhibit inherent metabolic abnormalities that may respond differently to DGAT1 inhibition than cells from sporadic AD patients.

Alternative Explanations

Lipid droplet accumulation in AD astrocytes may reflect impaired fatty acid oxidation (as seen in peroxisome deficiency) rather than excessive triglyceride synthesis. Restoring peroxisomal β-oxidation through PECRP (peroxisomal_enoyl-CoA_reductase) activation would address the root cause of droplet accumulation without blocking DGAT1.

Alternatively, astrocyte lipid droplets may be protective "sink" compartments sequestering Aβ and preventing extracellular aggregation. DGAT1 inhibition could paradoxically increase extracellular Aβ by releasing sequestered peptides.

Falsification Experiments

• Generate astrocyte-specific DGAT1 knockout mice; if cognitive performance is unchanged under standard diet, hypothesis is falsified
• Perform metabolic tracing with 13C-palmitate; if fatty acids are directed toward β-oxidation rather than esterification after DGAT1 inhibition, the mechanism is supported
• Test whether DGAT1 inhibition worsens cognition in aged mice (24 months) where lipid droplets may serve protective roles

Revised Confidence: 0.48 (−0.17)

Hypothesis 3: ST3GAL5 Activation (GM1→GM3 Shift)

Weaknesses in Evidence

The hypothesis relies heavily on correlative data showing GM1 enrichment in AD raft fractions (PMID: 31118253) without establishing whether GM1 accumulation is a cause or consequence of amyloid pathology. GM1 is synthesized earlier in development and is essential for synaptogenesis and axonal guidance; its reduction could impair neuronal development and function.

The assumption that GM1 clusters increase BACE1 activity 3-fold (PMID: 18630944) was demonstrated in artificial membrane systems at non-physiological concentrations. Whether this occurs in intact neurons at GM1 levels found in AD brain is unproven.

Counter-Evidence

ST3GAL5 knockout mice (PMID: 25873377) show altered APP processing but the direction of change is critical—the paper shows complex effects including accumulation of APP C-terminal fragments that could be neurotoxic independent of Aβ.

GM3, the proposed alternative ganglioside, is pro-inflammatory and promotes TNF-α signaling through CD14/TLR4 complexes (PMID: 21572173). Shifting ganglioside balance toward GM3 could exacerbate neuroinflammation in AD brains where microglial activation is already elevated.

Furthermore, GM1 deficiency causes severe developmental disorders (Lawasaki syndrome models), and therapeutic reduction in adults may impair synaptic maintenance mechanisms that require GM1-enriched microdomains for neurotrophin receptor signaling.

Alternative Explanations

GM1 accumulation in AD may represent a compensatory neuroprotective response. GM1 binds Aβ with high affinity, potentially sequestering oligomers and preventing membrane insertion. The observed GM1-Aβ complexes in AD brain may represent a detoxification mechanism rather than a seed propagation system.

Alternative approach: Instead of reducing GM1, enhance GM1-targeted antibodies or GM1 mimetics to increase sequestration of toxic Aβ oligomers while preserving ganglioside-dependent signaling.

Falsification Experiments

• Perform single-cell RNA-seq in ST3GAL5-overexpressing mice; if gene expression patterns show disrupted synaptic development pathways, the hypothesis is falsified for adult administration
• Measure microglial inflammatory markers after GM1 reduction; if IL-6/TNF-α increase, the hypothesis is falsified due to GM3 pro-inflammatory effects
• Test whether GAβ "seeds" are actually more toxic than free Aβ oligomers in vivo using stereotactic injection studies

Revised Confidence: 0.52 (−0.16)

Hypothesis 4: LXRβ-Selective Agonism

Weaknesses in Evidence

The hypothesis assumes that selective LXRβ agonism will avoid hepatic side effects, but LXRβ is expressed in liver and contributes to lipogenesis. While LXRα is the primary driver of SREBP1c transcription and lipogenesis, LXRβ deletion in mice still causes hepatic triglyceride accumulation in aging (PMID: 29463572), suggesting LXRβ agonism may not be side-effect-free.

The APOE lipidation mechanism is oversimplified. APOE4's reduced lipidation is due to both decreased secretion (APOE4 forms more intracellular aggregates in astrocytes) and impaired ABCA1-mediated lipidation. Simply enhancing ABCA1 may not overcome the intrinsic folding defect of APOE4.

Counter-Evidence

LXR agonists have consistently failed in clinical trials for metabolic indications due to hepatomegaly and hypertriglyceridemia. Even supposedly selective LXRβ agonists show cross-reactivity, and systemic ABCA1 upregulation increases reverse cholesterol transport from peripheral macrophages—a potential confounder for brain imaging studies.

The APOE4-lipidation study (PMID: 31758180) showing impaired LXR-driven ABCA1 transcription was performed in cultured cells; whether this holds in human brain tissue with intact BBB and cellular architecture is unknown.

Furthermore, LXR activation in microglia induces APOE expression, and APOE4-APOE4 interactions promote Aβ aggregation through a different mechanism than lipidation status (PMID: 32958806). Simply increasing APOE4 quantity without correcting its structural abnormality could worsen seeding.

Alternative Explanations

APOE4's pathogenicity in AD may be structure-dependent rather than lipidation-dependent. The R61C and R61E APOE4 structural switch mutations that prevent isoform-specific interactions (PMID: 31834367) suggest that blocking APOE4 dimerization may be more effective than enhancing lipidation.

Alternative approach: small-molecule correctors that restore APOE4 conformational flexibility, similar to CFTR modulators in cystic fibrosis, could address both lipidation and aggregation defects simultaneously.

Falsification Experiments

• Administer LXRβ agonist to APOE4 targeted replacement mice for 6 months; measure hepatic triglyceride content—if elevated >50% versus controls, hepatic toxicity limits utility
• Test whether LXRβ agonism improves cognition in aged APOE4 mice (>18 months) where structural APOE4 defects may dominate over lipidation issues
• Compare LXRβ agonist effects on ABCA1 versus LXRβ-dependent synaptic genes (using RNA-seq); if synaptic pathways are suppressed, mechanism is non-selective

Revised Confidence: 0.58 (−0.12)

Hypothesis 5: PISD Restoration

Weaknesses in Evidence

PISD is unusual among these hypotheses because it proposes restoring a mitochondrial enzyme, but mitochondrial dysfunction in AD is upstream of many processes (including lipid metabolism dysregulation) rather than a primary driver. PISD deficiency may be a downstream effect of proteostatic stress rather than an initiator.

The hypothesis conflates mitochondrial PS externalization with PISD activity, but PS externalization to the outer mitochondrial membrane is regulated by scramblases (PLSCR3) independent of PISD. The relationship between PISD activity and outer membrane PS levels is not direct.

Counter-Evidence

PISD has dual localization—it functions in mitochondria and also in the nucleus where it regulates splicing (PMID: 30401811). Restoring mitochondrial PISD without considering nuclear effects could disrupt RNA processing and cause unexpected toxicity.

The correlation between PISD downregulation and Braak staging (PMID: 32246132) is consistent with PISD being suppressed by pathological processes rather than causing them. Neurons in advanced AD show global transcriptional downregulation affecting hundreds of mitochondrial genes.

PS externalization triggering BAX activation (PMID: 30361425) was demonstrated in cell culture with exogenous Aβ treatment—it's unclear whether this mechanism operates in human AD brain where Aβ exposure is chronic and cellular adaptations have occurred over decades.

Alternative Explanations

Rather than PISD restoration, mitochondrial dysfunction in AD may be driven by mitochondrial DNA (mtDNA) deletions and impaired dynamics (fission/fusion). Therapeutic approaches targeting PGC-1α (Nrf1/Tfam axis) or Mdivi-1 (DRP1 inhibitor) address more proximal causes of mitochondrial failure.

The PE deficiency hypothesis is contradicted by evidence that brain PE content is preserved in AD, with increases in mitochondrial PE observed in some studies (PMID: 31969551).

Falsification Experiments

• Measure PISD enzymatic activity directly in AD brain mitochondria (not just expression); if activity is normal despite reduced transcript, the hypothesis is falsified
• Perform metabolomics in PISD-overexpressing neurons; if PE levels do not increase, the mechanism is blocked at post-translational level
• Test whether PISD restoration improves mitochondrial calcium handling using live-cell imaging; if calcium dynamics are unchanged, the electron transport chain claim is unsupported

Revised Confidence: 0.41 (−0.17)

Hypothesis 6: PLIN2/NEDD4L Lipophagy Activation

Weaknesses in Evidence

The hypothesis proposes enhancing PLIN2 ubiquitination via NEDD4L, but NEDD4L is a tissue-specific E3 ligase with limited expression in brain astrocytes. NEDD4L expression decreases with age (PMID: 33874665), but whether this is a cause or consequence of lipid droplet accumulation is unclear.

Autophagy enhancement through rapamycin or trehalose is non-specific—these agents activate autophagy through mTOR inhibition or osmotic stress, affecting all organelles including lysosomes, peroxisomes, and ribosomes. Off-target effects could worsen AD pathology.

Counter-Evidence

Autophagy activation in AD mouse models shows biphasic effects. While mild autophagy enhancement clears protein aggregates, robust activation induces apoptosis in vulnerable neurons (PMID: 32973027). The therapeutic window is narrow, and the optimal autophagy level for AD is unknown.

PLIN2-coated lipid droplets may serve essential functions in astrocytes, including sterol storage for steroid hormone synthesis and membrane synthesis during remodeling. PLIN2 degradation could impair these functions.

Trehalose, specifically, has off-target effects including HSP70 induction and TFEB activation independent of autophagy, making it difficult to attribute benefits to lipophagy enhancement alone. Clinical translation of trehalose is limited by its poor blood-brain barrier penetration.

Alternative Explanations

Lipid droplet accumulation in AD may result from impaired peroxisomal β-oxidation rather than autophagy deficiency. Peroxisomes generate hydrogen peroxide and are essential for very-long-chain fatty acid metabolism; their dysfunction in AD is well-documented. Restoring peroxisomal function through PPARα agonism may address the primary defect.

Alternatively, lipid droplet accumulation could be a response to increased fatty acid delivery from blood through BBB dysfunction, rather than impaired clearance.

Falsification Experiments

• Perform stereotactic injection of NEDD4L AAV into hippocampus of aged APP/PS1 mice; measure both PLIN2 levels and cognitive function—if droplets clear but cognition worsens, the hypothesis is falsified
• Test whether autophagy flux is actually impaired in AD astrocytes or whether substrate delivery to lysosomes is the limiting step
• Measure ceramide species directly after lipophagy enhancement; if ceramides increase (due to droplet breakdown), GSK-3β activation could worsen

Revised Confidence: 0.50 (−0.12)

Hypothesis 7: CYP2J2/DHA Epoxides

Weaknesses in Evidence

The hypothesis presents the strongest evidence base among the seven hypotheses (confidence 0.74), but CYP2J2-mediated DHA epoxide production faces significant pharmacokinetic challenges. CYP2J2 epoxygenases generate multiple epoxide regioisomers (19,20-EDP, 16,17-EDP, 10,11-EDP) with different activities and potencies; selective enhancement of the neuroprotective 19,20-EDP is not guaranteed with pharmacological CYP2J2 activation.

DHA supplementation alone (PMID: 29982765) does not specifically test the epoxide hypothesis—it activates multiple pathways including resolvins, protectins, and maresins. The observed benefits cannot be attributed specifically to CYP2J2 epoxides.

Counter-Evidence

Epoxides are rapidly metabolized by soluble epoxide hydrolase (sEH), with half-lives of 2-4 hours in plasma. Achieving sustained CNS concentrations of bioactive epoxides would require sEH inhibitors co-administration, complicating the therapeutic approach.

The membrane fluidity model (PMID: 31243156) was tested in artificial planar bilayers, not neuronal membranes. Aβ-induced membrane rigidification may occur through different mechanisms in complex neuronal membranes containing integral proteins and cytoskeletal elements.

Aβ increases membrane cholesterol 40% (PMID: 24503041) in acute experiments—chronic Aβ exposure may trigger compensatory responses that alter cholesterol trafficking and distribution.

Alternative Explanations

DHA's neuroprotective effects may operate through PPARγ activation by docosanoids (neuroprotectin D1) rather than CYP2J2 epoxides. NPD1 and other DHA-derived mediators activate anti-inflammatory pathways through different receptors.

Membrane fluidity restoration could be achieved more directly through HDL mimetic therapy or cyclodextrin-mediated cholesterol extraction, which have demonstrated cognitive benefits in AD mouse models.

Falsification Experiments

• Administer CYP2J2 inhibitor (T28) with DHA supplementation; if benefits disappear, epoxides are the relevant mediators; if benefits persist, other DHA metabolites are responsible
• Measure 19,20-EDP directly in brain tissue after CYP2J2 activation; if levels do not increase substantially, peripheral metabolism dominates
• Test whether CYP2J2 overexpression in astrocytes (not just neurons) is required for synaptic protection—if astrocyte-derived epoxides mediate benefits, neuronal targeting is insufficient

Revised Confidence: 0.64 (−0.10)

Cross-Cutting Analysis

Mechanistic Interconnections

The cross-cutting prediction about synergy between cholesterol efflux (H1/H4) and lipid droplet clearance (H2/H6) is plausible but oversimplified. The pathways are interconnected through ABCA1, but ABCA1 activity is paradoxically reduced in APOE4 carriers due to feedback inhibition, limiting synergy.

APOE4 carriers show impaired lipid clearance through multiple mechanisms: reduced APOE secretion, intracellular aggregation, and impaired lipidation. Combination therapy targeting both cholesterol efflux and droplet clearance may be synergistic only if the root cause (APOE4 structural defect) is addressed.

Ranking by Revised Confidence

| Hypothesis | Original | Revised | Change | Primary Limitation |
|------------|----------|---------|--------|---------------------|
| H7 (CYP2J2/DHA) | 0.74 | 0.64 | −0.10 | Pharmacokinetic barriers |
| H1 (CYP46A1) | 0.72 | 0.54 | −0.18 | Biphasic 24-HC effects |
| H4 (LXRβ) | 0.70 | 0.58 | −0.12 | Hepatic toxicity risk |
| H3 (ST3GAL5) | 0.68 | 0.52 | −0.16 | GM3 pro-inflammatory |
| H2 (DGAT1) | 0.65 | 0.48 | −0.17 | Astrocyte-specificity |
| H6 (PLIN2) | 0.62 | 0.50 | −0.12 | Off-target autophagy |
| H5 (PISD) | 0.58 | 0.41 | −0.17 | Downstream of pathology |

Recommended Priorities for Investigation

H7 (CYP2J2/DHA epoxides) remains most promising with refined mechanism: use sEH inhibitors (PF-4674212) to stabilize endogenous epoxides rather than CYP2J2 activation

H4 (LXRβ) requires development of tissue-specific agonists with poor CNS penetration but preserved brain effects—this may be feasible with active transport mechanisms

H1 (CYP46A1) warrants investigation for early-stage prevention rather than treatment, with careful 24-HC monitoring

Key Falsification Experiments Across All Hypotheses

• Age-dependent efficacy testing (all): Most hypotheses demonstrate effects in 3-6 month mice but not aged (>18 month) animals
• Cell-type specificity (all): Astrocyte versus neuronal effects require Cre-lox system validation
• APOE genotype dependence (all): APOE4 versus APOE3 mice show differential responses; universal mechanisms are unlikely
• Biomarker validation (all): 24-HC, ceramide species, ganglioside ratios in CSF should be validated as pharmacodynamic readouts

Drug Development Assessment: Lipid Metabolism Hypotheses in Alzheimer's Disease

Executive Summary

The seven hypotheses span a spectrum of druggability—from well-established nuclear receptor agonism to challenging mitochondrial enzyme restoration. Hypothesis 7 (CYP2J2/DHA epoxides) emerges as the most immediately actionable given existing clinical-stage compounds, while Hypothesis 4 (LXRβ) offers the richest translational precedent despite hepatic toxicity concerns. Hypothesis 5 (PISD) represents the highest-risk target with the least tractable therapeutic approach.

Hypothesis 1: CYP46A1 Activation

Druggability Assessment

Moderate-High Risk Target

CYP46A1 is a 50-kDa cytochrome P450 enzyme with a redox partner requirement (NADPH-cytochrome P450 oxidoreductase), making it inherently challenging to target with systemically administered small molecules. The P450 family exhibits high structural homology, creating selectivity challenges—Efavirenz, the only known CYP46A1 activator, also inhibits CYP2D6 and CYP2C9.

Chemical Matter Available

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| Efavirenz | Bristol-Myers Squibb | Marketed (HIV) | Activates CYP46A1 at 1-10 μM; discontinued due to CNS toxicity, psychiatric effects |
| Rifampin | Various | Generic | Weak CYP46A1 activation; drug-drug interaction concerns |

Key Problem: No selective CYP46A1 activator exists. Efavirenz's neurotoxicity (disorientation, vivid dreams, psychosis) confounds interpretation of any cognitive benefits.

Competitive Landscape

• Empty pipeline for CYP46A1-targeted AD therapy
• Academic tool compounds (selective CYP46A1 inhibitors for control arms) are limited
• Patent landscape is largely unencumbered for CNS applications

Safety Concerns

• Biphasic 24-HC dose response: neuroprotective at <200 ng/mL, pro-apoptotic at >500 ng/mL in CSF
• Efavirenz carries FDA black box for psychiatric reactions
• CYP46A1 is expressed in retina—vision effects possible

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|----------------|
| Target validation | 2 years | $3-5M (CYP46A1 knockout mice + biomarker development) |
| Lead optimization | 4-5 years | $20-30M (P450 selectivity screens) |
| IND-enabling | 2 years | $8-12M |
| Phase I-II | 4-6 years | $40-60M |
| Total to approval | 12-17 years | $71-107M |

Verdict: High-risk target requiring novel chemistry. The Efavirenz repositioning angle is worth exploring (repurposing at sub-CYP46A1-activating doses to minimize neuropsychiatric effects), but regulatory approval for a new indication would still require full development.

Hypothesis 2: Astrocyte-Specific DGAT1 Inhibition

Druggability Assessment

High Tractability but Specificity Challenge

DGAT1 is a well-validated target—multiple selective inhibitors have reached clinical trials for metabolic diseases. The primary challenge is cell-type specificity, not enzyme inhibition per se.

Chemical Matter Available

| Compound | Company | Stage | BBB Penetration |
|----------|---------|-------|----------------|
| Praserone (PRX-007) | EosMicrobiomics | Phase I (CNS) | Moderate—under investigation |
| Vilaprisan (BAY 897) | Bayer | Phase II (women's health) | Low—developed for uterine disease |
| A-922500 | Abbott | Preclinical | Low |
| DGAT1i-2 | Academic | Research use | Unknown |

Key Insight: Praserone (a DGAT1 inhibitor from EosMicrobiomics) was specifically developed for CNS indications including potential neurodegenerative applications—making this the most immediately relevant compound.

Competitive Landscape

• Limited CNS-focused competition—most DGAT1 inhibitors target metabolic syndrome, NASH, or diabetes
• Bayer's vilaprisan was discontinued in favor of other assets
• No DGAT1 programs specifically targeting astrocyte lipid droplets in AD

Safety Concerns

• Gastrointestinal intolerance at high doses (fat malabsorption, diarrhea)—dose-limiting in oral formulations
• Unclear if these effects translate to CNS-localized inhibition
• No data on chronic CNS exposure in aged populations

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|-----------------|
| Target validation (astrocyte-specific) | 1-2 years | $2-4M (Cre-lox mice, AAV targeting) |
| Compound sourcing/repurposing | 1 year | $0.5-1M (Praserone PK/PD studies) |
| IND-enabling | 2 years | $6-8M |
| Phase I-II | 4-5 years | $35-50M |
| Total to approval | 8-10 years | $43-63M |

Verdict: Moderate tractability with existing compounds. The major uncertainty is cell-type specificity—achieving sufficient astrocyte targeting without peripheral DGAT1 inhibition. This likely requires AAV-based gene therapy (e.g., AAV-GFAP-DGAT1-shRNA) rather than small molecules, fundamentally altering the competitive landscape and cost structure.

Hypothesis 3: ST3GAL5 Activation

Druggability Assessment

Low Druggability

ST3GAL5 (GM3 synthase) is a glycosyltransferase—a notoriously difficult enzyme class to target with small molecules due to complex substrate recognition (polypeptide + glycan) and lack of known small-molecule activators. Unlike kinases or proteases, glycan-processing enzymes have shallow binding pockets.

Chemical Matter Available

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| None | — | — | No selective ST3GAL5 activators reported |
| SiRNA tool compounds | Various | Research use | Poor brain penetration |
| ST3GAL5 overexpression AAV | Academic vectors | Research use | Requires direct CNS injection |

Key Problem: Glycosyltransferase activators are essentially non-existent in pharma portfolios. The field would require de novo lead discovery from high-throughput screens—low probability of success.

Competitive Landscape

• Empty competitive landscape—no ST3GAL5 modulators in any therapeutic area
• However, the absence of competition also means no established safety databases

Safety Concerns

• GM1 is essential for synaptogenesis—global reduction risks developmental abnormalities
• ST3GAL5 mutations cause Amish lethal epilepsy syndrome—complete loss is lethal
• Therapeutic window concerns: how much GM1 reduction is tolerated?

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|----------------|
| Lead discovery (HTS for activator) | 3-4 years | $15-25M |
| Optimization | 4-5 years | $30-40M |
| IND-enabling | 2 years | $10-15M |
| Phase I-II | 5-6 years | $60-80M |
| Total to approval | 14-17 years | $115-160M |

Verdict: Not recommended for near-term development. Gene therapy approaches (ST3GAL5 AAV) are more plausible than small molecules but still face significant delivery and safety challenges. The mechanistic uncertainty about whether GM1 reduction is truly beneficial further undermines investment rationale.

Hypothesis 4: LXRβ-Selective Agonism

Druggability Assessment

High Tractability with Known Liabilities

Nuclear receptors are among the most druggable target classes—LXRβ is a well-characterized transcription factor with established ligand-binding domain pharmacology. The challenge is achieving isoform selectivity and CNS penetration without hepatic effects.

Chemical Matter Available

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| LXR-623 (WAY-362623) | Pfizer | Phase I (abandoned) | Showed hepatomegaly in preclinical models |
| RGX-104 | Revolution Medicines | Phase I (discontinued) | Developed for cancer immunotherapy |
| GW3965 | Academic/GSK | Research use | Gold standard tool compound |
| T0901317 | Academic | Research use | Non-selective, also farnesoid X receptor agonist |
| BMS-779455 | Bristol-Myers Squibb | Preclinical | LXRβ-sparing肝脏 effects |

Key Insight: Pfizer's LXR-623 is the most clinically advanced LXR agonist—completed Phase I trials for atherosclerosis (NCT00796575) before discontinuation. Safety data exists. The compound showed adequate CNS penetration in preclinical models.

Competitive Landscape

• Moderate competition—several LXR programs existed but most have been discontinued
• Roche, Merck, Pfizer all had LXR programs in metabolic syndrome
• No LXRβ-specific agents in current AD development pipelines
• AstraZeneca has residual interest in CNS LXR modulators

Safety Concerns

• Hepatic lipogenesis via SREBP1c activation (even with LXRβ-selective agents)
• Hypertriglyceridemia in humans (LXR-623 showed this in Phase I)
• TREM2-independent microglial effects—potential immunosuppression concerns
• APOE4 patients may have different dose-response due to ABCA1 feedback defects

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|----------------|
| Repositioning LXR-623 | 2-3 years | $15-25M (existing Phase I data) |
| AD-specific Phase II | 3-4 years | $30-50M |
| Phase III | 4-5 years | $80-120M |
| Total to approval (accelerated) | 9-12 years | $125-195M |

Verdict: LXR-623 repositioning for AD represents the most compelling accelerated development pathway—existing Phase I safety and PK data substantially de-risk early development. The key questions are: (1) does LXR-623's hepatic toxicity profile permit chronic CNS dosing, and (2) are the APOE4 lipidation effects sufficient to overcome structural defects? Both are testable within 2-3 years of additional study.

Hypothesis 5: PISD Restoration

Druggability Assessment

Very Low Druggability

PISD is a mitochondrial enzyme (localized to inner mitochondrial membrane) with no established small-molecule activators. Mitochondrial enzymes are among the least tractable targets due to:

Delivery challenges across two membranes

Requirement for mitochondrial matrix targeting signals

Dependency on mitochondrial membrane potential for uptake of charged molecules

Chemical Matter Available

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| None | — | — | No PISD modulators reported |
| PISD overexpression AAV | Academic vectors | Research use | Requires direct CNS injection |
| Small molecule PE precursors | Various | Research use | Indirect approach (e.g., ethanolamine) |

Key Problem: PISD is not just "undrugged"—it may be inherently undruggable with small molecules. Gene therapy (AAV-PISD) is the only plausible therapeutic modality.

Competitive Landscape

• Non-existent—no PISD-targeted programs in any indication
• No competitive IPs or industry interest

Safety Concerns

• Dual mitochondrial/nuclear localization—restoring one compartment may disrupt the other
• Nuclear PISD regulates RNA splicing—off-target effects on gene expression
• Mitochondrial interventions in aging neurons with established pathology may be too late

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|----------------|
| AAV construct development | 2 years | $5-8M |
| CNS delivery optimization | 3-4 years | $20-30M |
| IND-enabling | 2 years | $15-20M |
| Phase I-II (gene therapy) | 5-7 years | $80-120M |
| Total to approval | 12-15 years | $120-178M |

Verdict: Not recommended for investment. The mechanistic uncertainties (PISD as downstream epiphenomenon) and delivery challenges make this the highest-risk hypothesis. The field would need fundamental advances in mitochondrial gene therapy before pursuing this approach.

Hypothesis 6: PLIN2/NEDD4L Lipophagy Activation

Druggability Assessment

Moderate Tractability—Indirect Approach

The target is autophagy enhancement rather than direct PLIN2 or NEDD4L modulation. This makes the hypothesis more tractable but less specific—multiple autophagy activators exist, but none are approved for CNS indications.

Chemical Matter Available

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| Rapamycin (sirolimus) | Various | Generic (transplant) | mTOR inhibitor—broad autophagy activation |
| Trehalose | Various | Research use | Autophagy inducer, poor BBB penetration |
| Lithium | Generic | Off-patent | Autophagy via IMPase inhibition |
| MLN4924 (pevonedistat) | Millennium/Takeda | Phase I/II (cancer) | NEDD4L E3 ligase inhibitor—opposite direction |
| NR1H4 (LXR agonists) | Various | See H4 | Also enhance autophagy |

Key Insight: The most clinically advanced autophagy enhancer is rapamycin, which has an extensive safety database. However, rapamycin's immunosuppression and metabolic effects are concerning for chronic AD treatment.

Competitive Landscape

• Moderate competition—multiple autophagy programs in neurodegenerative disease
• Biogen has explored autophagy modulators for AD
• UCB has autophagy programs in Parkinson's
• No specific PLIN2-targeting programs

Safety Concerns

• Immunosuppression (rapamycin)—infection risk in elderly AD population
• Metabolic effects—diabetes, hyperlipidemia
• Autophagy's biphasic nature—too much autophagy induces neuronal apoptosis
• NEDD4L inhibition would be pro-autophagic but MLN4924 is a NEDD4L inhibitor (paradoxically worsens PLIN2 accumulation)

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|----------------|
| Trehalose repositioning | 2 years | $10-15M |
| IND-enabling for AD-specific dosing | 2 years | $5-8M |
| Phase II | 3-4 years | $30-45M |
| Phase III | 4-5 years | $80-100M |
| Total to approval | 11-14 years | $125-168M |

Verdict: Trehalose is the most immediately actionable compound—it has demonstrated safety in humans for other indications and shows neuroprotective effects in AD mouse models. The primary challenge is BBB penetration—trehalose requires reformulation (intranasal? prodrug?) to achieve therapeutic CNS concentrations. This is a 5-7 year development effort at reasonable cost if BBB delivery is solved.

Hypothesis 7: CYP2J2/DHA Epoxides

Druggability Assessment

High Tractability—Dual Entry Points

This hypothesis benefits from two independent therapeutic angles:

CYP2J2 activation (upstream) — more complex but directly increases epoxide production

sEH inhibition (downstream) — pharmacologically simpler; prevents epoxide degradation

Chemical Matter Available

CYP2J2 Modulators:

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| T28 (astemizole metabolite) | Academic | Research use | CYP2J2 substrate; inhibits at high concentrations |
| Astiangeprazole | Academic | Research use | Metabolite of astemizole with CYP2J2 activity |
| No selective activators | — | — | CYP epoxygenase activators are rare |

sEH Inhibitors:

| Compound | Company | Stage | Notes |
|----------|---------|-------|-------|
| GSK225629 | GlaxoSmithKline | Phase I (pain/COPD) | sEH inhibitor; CNS penetration demonstrated |
| EC-5026 (sEH-397) | EicOsis/UC Davis | Phase I/II (pain) | IND cleared by FDA 2019 |
| PF-06760850 | Pfizer | Phase I (metabolic) | Preclinical-to-Phase I transition |
| 12,13-EDP | Academic | Research use | Direct epoxide; unstable in plasma |

Key Insight: sEH inhibitors represent the most advanced clinical program for this pathway. Both GSK225629 and EC-5026 have completed Phase I trials with acceptable safety profiles. EC-5026 received FDA IND clearance for pain indication, and Phase II trials are underway—this is essentially a "readymade" compound for AD repositioning.

Competitive Landscape

• Active competition in sEH inhibitors for metabolic/CNS indications
• EicOsis (University of California, Davis spinout) is the leader in CNS-applicable sEH inhibitors
• ExxonMobil Biomedical Sciences funded early sEH work
• No specific competition in AD-specific lipid raft restoration

Safety Concerns

• sEH inhibitors are generally well-tolerated
• Epoxides are reactive intermediates—can form DNA adducts if not properly regulated
• DHA supplementation is safe but provides only indirect pathway activation
• CYP2J2 has peripheral expression (cardiac, intestinal)—systemic epoxide elevation possible

Cost & Timeline Estimate

| Phase | Duration | Estimated Cost |
|-------|----------|----------------|
| sEH inhibitor repositioning (EC-5026) | 2-3 years | $15-20M (existing Phase I data) |
| AD-specific Phase II | 3-4 years | $40-60M |
| Phase III | 4-5 years | $90-130M |
| Total to approval | 9-12 years | $145-210M |

Verdict: Highest priority for development. The sEH inhibitor pathway offers:

A clinically advanced compound (EC-5026) ready for repositioning

Dual therapeutic potential: membrane fluidity + anti-inflammatory epoxide effects

Established safety database from pain trials

Mechanistic plausibility with direct biochemical readouts (EDP/DHA ratio in CSF)

Comparative Summary

| Hypothesis | Druggability | Tool Compounds | Clinical Candidates | Safety Risk | Development Timeline |
|------------|--------------|-----------------|---------------------|-------------|----------------------|
| H1 CYP46A1 | Moderate | Efavirenz (repositioning) | None | High (neuropsychiatric) | 12-17 years |
| H2 DGAT1 | High | Praserone | None (CNS) | Moderate (GI) | 8-10 years |
| H3 ST3GAL5 | Very Low | None | None | High (developmental) | 14-17 years |
| H4 LXRβ | High | LXR-623 | LXR-623 (Phase I) | Moderate (hepatic) | 9-12 years |
| H5 PISD | Very Low | None | None (gene therapy only) | Unknown | 12-15 years |
| H6 PLIN2 | Moderate | Rapamycin/Trehalose | None | Moderate (immunosuppression) | 11-14 years |
| H7 CYP2J2/sEH | High | EC-5026, GSK225629 | EC-5026 (Phase I/II) | Low | 9-12 years |

Recommended Development Priorities

Tier 1: Immediate Repositioning Opportunities

H7 (sEH Inhibition) → EC-5026 or GSK225629

• Existing Phase I safety and PK data
• IND cleared for related CNS indications
• Mechanism directly addresses membrane fluidity hypothesis
• Budget: $15-20M for repositioning feasibility studies
• Timeline: Phase II initiation within 2-3 years

H4 (LXRβ) → LXR-623 Repositioning

• Existing Phase I data (NCT00796575)
• Proven CNS penetration
• Budget: $15-25M for AD-specific Phase II preparation
• Timeline: AD Phase II within 3 years of decision

Tier 2: Near-Term Development (5-7 Year Horizon)

H2 (DGAT1) → Praserone Development

• Only CNS-focused DGAT1 inhibitor in development
• Requires astrocyte-targeting strategy validation
• Budget: $30-40M for specificity studies + Phase I

H6 (Autophagy) → Trehalose Reformulation

• Requires BBB delivery optimization
• Well-established safety profile supports rapid iteration
• Budget: $20-30M for formulation + Phase I

Tier 3: Fundamental Research Required

H1 (CYP46A1) — Consider if sEH inhibitor fails due to unexpected toxicity; requires novel chemistry H3 (ST3GAL5) — Deprioritize unless GM1-Aβ seeding mechanism is definitively validated H5 (PISD) — Not recommended for development; prioritize basic science investigation of mitochondrial lipid dysregulation

Cross-Cutting Recommendations

Biomarker Development

For all hypotheses, validate pharmacodynamic biomarkers before initiating clinical trials:

| Hypothesis | Biomarker | Sample Type |
|-----------|-----------|-------------|
| H1 | 24-HC levels | CSF |
| H2 | PLIN2+ droplet count | iPSC-derived astrocytes |
| H4 | APOE4 lipidation (size exclusion chromatography) | CSF |
| H6 | Ceramide species (LC-MS/MS) | Plasma/CSF |
| H7 | 19,20-EDP/14,15-HEPE ratio | Plasma |

APOE Genotype Stratification

All clinical trials should stratify by APOE4 carrier status. Based on the cross-cutting hypothesis, APOE4 carriers may show enhanced benefit from combination therapy (H1+H4 or H4+H7) due to shared APOE-lipidation mechanisms.

Regulatory Strategy

• Accelerated Approval pathway is plausible if CSF Aβ42/40 ratio is validated as surrogate endpoint
• FDA has expressed interest in AD lipid biomarkers (per 2023 FDA guidance on lipid-related targets)
• Consider Orphan Drug Designation for APOE4 homozygous patients—populations with highest unmet need and clearest genetic risk

Partnership Recommendations

• EicOsis (EC-5026) — best fit for H7 collaboration
• Roche/Genentech — established LXR program and APOE expertise
• Biogen — autophagy and lipid droplet expertise through pipeline analysis
• Denali Therapeutics — BBB crossing expertise relevant for any CNS-targeted lipid therapy