{ "id": "R-003", "version": "1.0", "date": "2026-10-11", "title": "TREM2: when does it help, do little, or cause harm?", "status": "AI-assisted targeted literature comparison; not peer reviewed", "decision": "Prioritize a state- and pathology-matched test of neuronal protection. Broad TREM2 activation has no established clinical benefit in the reviewed trial; no effective human subgroup or optimal dose is identified here.", "sources": [ { "id": "S1", "citation": "Ulland et al., Cell (2017)", "title": "TREM2 Maintains Microglial Metabolic Fitness in Alzheimer's Disease", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC5573224/", "access": "Abstract and relevant full-text results" }, { "id": "S2", "citation": "Piers et al., FASEB Journal (2020; online 2019)", "title": "A locked immunometabolic switch underlies TREM2 R47H loss of function in human iPSC-derived microglia", "url": "https://pubmed.ncbi.nlm.nih.gov/31907987/", "access": "Indexed abstract and publisher metadata" }, { "id": "S3", "citation": "Feiten et al., Nature Communications (2026)", "title": "TREM2 expression level is critical for microglial state, metabolic capacity and efficacy of TREM2 agonism", "url": "https://www.nature.com/articles/s41467-026-68706-8", "access": "Relevant publisher full-text results, Figure 7 and dosing methods" }, { "id": "S4", "citation": "Wang et al., Neural Plasticity (2020)", "title": "TREM2 Overexpression Attenuates Cognitive Deficits in Experimental Models of Vascular Dementia", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC7306072/", "access": "Abstract, methods and relevant results" }, { "id": "S5", "citation": "Loss of TREM2 / Tg-SwDI study, Alzheimer's & Dementia (2024)", "title": "Loss of TREM2 diminishes CAA despite an overall increase of amyloid load in Tg-SwDI mice", "url": "https://pubmed.ncbi.nlm.nih.gov/39308178/", "access": "Indexed abstract and figure descriptions" }, { "id": "S6", "citation": "Jain et al., Journal of Experimental Medicine (2023; online 2022)", "title": "Chronic TREM2 activation exacerbates Aβ-associated tau seeding and spreading", "url": "https://rupress.org/jem/article/220/1/e20220654/213537/Chronic-TREM2-activation-exacerbates-A-associated", "access": "Abstract and relevant experimental results" }, { "id": "S7", "citation": "Etxeberria et al., Journal of Neuroscience (2024)", "title": "Neutral or Detrimental Effects of TREM2 Agonist Antibodies in Preclinical Models of Alzheimer's Disease and Multiple Sclerosis", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11255434/", "access": "Abstract and relevant results/discussion" }, { "id": "S8", "citation": "Mummery et al., Nature Medicine (2026)", "title": "The TREM2 agonistic antibody AL002 in early Alzheimer's disease: a phase 2 randomized trial", "url": "https://www.nature.com/articles/s41591-026-04273-1", "access": "Publisher abstract, public figure descriptions and disclosures; subscription full text not reviewed" }, { "id": "S9", "citation": "Price et al., Journal of Neuroinflammation (2020)", "title": "Therapeutic Trem2 activation ameliorates amyloid-beta deposition and improves cognition in the 5XFAD model of amyloid deposition", "url": "https://link.springer.com/article/10.1186/s12974-020-01915-0", "access": "Publisher full text: systemic administration, results and behavioral methods" }, { "id": "S10", "citation": "Chen et al., Journal of Neuroinflammation (2026)", "title": "A ligand-mimetic anti-TREM2 agonist antibody elevates soluble TREM2 and ameliorates pathology in mouse models of Alzheimer’s disease and multiple sclerosis", "url": "https://link.springer.com/article/10.1186/s12974-026-03733-2", "access": "Publisher full text: 5xFAD methods and results; antibody signaling and sTREM2 results" } ], "rows": [ { "id": "C1", "source": "S1", "category": "benefit", "level": "Mouse / mechanism", "context": "TREM2-deficient 5XFAD; impaired cellular energy handling", "intervention": "Cyclocreatine metabolic support; not direct receptor agonism", "timing": "Rescue experiment in a deficient amyloid model; regimen not fully extracted", "comparator": "Deficient mice without metabolic rescue", "result": "Improved microglial clustering and reduced plaque-adjacent neuritic dystrophy.", "limit": "Rescuing a deficiency does not establish a benefit from increasing normal TREM2. No human cognitive outcome.", "location": "Metabolic rescue results", "sample": "Endpoint-specific biological replicate counts not extracted" }, { "id": "C2", "source": "S2", "category": "benefit", "level": "Human-derived cells", "context": "iPSC-derived microglia carrying the R47H risk variant", "intervention": "Manipulation of PPARγ / p38MAPK-related signaling", "timing": "Cell-culture intervention; no clinical disease stage", "comparator": "Variant and comparator cell conditions", "result": "Metabolic switching and phagocytic function could be rescued.", "limit": "A cellular functional result, not neuronal protection or patient benefit. Donor replication needs full-text verification.", "location": "Indexed abstract", "sample": "Donor and experiment counts not extracted" }, { "id": "C3", "source": "S9", "category": "benefit", "level": "Mouse / behavior", "context": "Male 5XFAD amyloid model without added human tau seeds", "intervention": "AL002a, 50 mg/kg intraperitoneally each week; MOPC control", "timing": "Start at 4 months; 14 weeks of treatment; behavior during the final 2 weeks", "comparator": "Isotype-matched MOPC; additional wild-type groups", "result": "Less amyloid deposition and improved spatial learning and novel-object recognition.", "limit": "One aggressive amyloid model; comparison with C6 also changes age, dose, sex composition and endpoints.", "location": "Systemic administration; behavioral and chronic-treatment results", "sample": "5XFAD: 12 per antibody group; wild type: 8 AL002a, 7 control" }, { "id": "C4", "source": "S10", "category": "benefit", "level": "Mouse / pathology", "context": "5xFAD amyloid model; ligand-mimetic antibody 03O05", "intervention": "30 mg/kg intraperitoneally weekly for 12 weeks", "timing": "Start at 7 months; tissue 2 days after the final dose", "comparator": "Vehicle-treated 5xFAD; age-matched wild type", "result": "Higher soluble TREM2, fewer filamentous plaques and reduced neuronal dystrophy.", "limit": "The disease experiment used vehicle rather than an isotype antibody control. No matched comparison proves superiority over other agonists or benefit in people.", "location": "In vivo efficacy methods; Figures 3–6", "sample": "14–15 mice per group in the chronic study" }, { "id": "C5", "source": "S3", "category": "mixed", "level": "Mouse / cell state", "context": "APP SAA amyloid reporter mice; low, mid and high TREM2-expressing microglia", "intervention": "Brain-penetrant ATV:4D9 versus isotype; 1 mg/kg monthly", "timing": "Four doses at ages 8–11 months; collection at 12 months", "comparator": "Isotype within expression strata", "result": "Mid-expression cells showed increased metabolic indicators; low cells changed little; high cells showed attenuated or opposite responses.", "limit": "These are metabolic profiles, not established benefit or injury. A high-expression response must not be labeled neuronal harm. Delayed collection limits acute-response interpretation.", "location": "Figure 7 and chronic antibody-treatment methods", "sample": "Assay-specific biological replicate counts not fully extracted" }, { "id": "C6", "source": "S6", "category": "harm", "level": "Mouse / amyloid + tau", "context": "5XFAD with human Alzheimer-derived tau injected at 6 months", "intervention": "AL002a 80 mg/kg intraperitoneally weekly versus IgG", "timing": "Start 1 week BEFORE tau injection; continue to 9 months", "comparator": "Control IgG in the same seeded model", "result": "More tau seeding/spreading and neuritic dystrophy; total amyloid burden did not fall.", "limit": "Harm was not restricted to treatment after established tau pathology. Artificial seeding and other design differences prevent attribution to tau alone.", "location": "Figure 1; chronic treatment and tau pathology results", "sample": "Chronic groups: 27 antibody (13 female/14 male), 26 control (14 female/12 male); endpoint subsets vary" }, { "id": "C7", "source": "S7", "category": "null", "level": "Mouse / AD models", "context": "PS2APP and TauPS2APP; multiple agonist antibodies", "intervention": "Para.09, mAb5 and mAb7 across separate designs", "timing": "Early, middle and later treatment designs; 12- or 16-week regimens", "comparator": "Matched experimental controls in each design", "result": "Limited or no improvement in the tested Alzheimer-related pathological and neuronal endpoints.", "limit": "A null finding is not proof of equivalence. CNS exposure, antibody format and timing remain possible explanations, not resolved causes.", "location": "AD-model results and discussion", "sample": "No single pooled n; full endpoint-level extraction pending" }, { "id": "C8", "source": "S7", "category": "harm", "level": "Mouse / demyelination", "context": "Toxin-induced demyelination and repair models", "intervention": "TREM2 agonist antibodies tested during injury/recovery", "timing": "Repair-stage experiments; not a dementia clinical stage", "comparator": "Experimental control treatment", "result": "Recovery from demyelination was impaired in the tested settings.", "limit": "A safety-context finding, not direct evidence of harm in human dementia. C7 and C8 are one publication, not independent replications.", "location": "Demyelination results", "sample": "Endpoint-specific n and regimens not fully extracted" }, { "id": "C9", "source": "S4", "category": "benefit", "level": "Mouse / ischemic injury", "context": "Young adult male mice after transient bilateral carotid occlusion", "intervention": "Lentiviral TREM2 overexpression", "timing": "After 24-hour recovery; spatial learning assessed about 4 weeks later", "comparator": "Injury/control-vector and sham conditions", "result": "Better spatial-learning performance and hippocampal neuronal preservation.", "limit": "Acute ischemia differs from aged vascular amyloid disease. Gene overexpression is not an antibody treatment.", "location": "Animal groups, treatment and behavioral results", "sample": "Three groups of 10; individual endpoint counts require verification" }, { "id": "C10", "source": "S5", "category": "mixed", "level": "Mouse / vascular amyloid", "context": "CAA-prone Tg-SwDI mice; Trem2 wild type, heterozygous and knockout", "intervention": "Lifelong genetic loss, not a timed treatment", "timing": "Assessed at 16 months", "comparator": "Different Trem2 genotypes in the same model", "result": "Knockout reduced vascular amyloid but increased overall brain amyloid.", "limit": "Opposing compartment outcomes are a trade-off. Knockout cannot be read backwards as proof that an agonist causes human vascular harm.", "location": "Indexed abstract and Figure 1 descriptions", "sample": "Figure 1 total amyloid: WT 7, heterozygous 7, knockout 6; not all endpoints" }, { "id": "C11", "source": "S8", "category": "null", "level": "Human / randomized trial", "context": "381 participants with early Alzheimer’s disease", "intervention": "AL002 15, 40 or 60 mg/kg IV every 4 weeks versus placebo", "timing": "48–96 weeks; primary estimate at week 96", "comparator": "Randomized, double-blind placebo control", "result": "Target engagement without a significant primary CDR-SB benefit. MRI changes resembling ARIA were the most frequent reported treatment-emergent adverse events.", "limit": "No established clinical efficacy for this regimen. Confidence intervals allow benefit and harm; adverse-event rates were not extracted from the subscription full text.", "location": "Public abstract; NCT04592874", "sample": "381 randomized across four groups" } ], "comparisons": [ { "title": "Amyloid alone versus added tau", "sources": [ "S9", "S6" ], "observation": "The same murine agonist appears in a beneficial amyloid experiment and a harmful tau-seeding experiment.", "inference": "Tau context is a priority moderator to test, but age, dose and other design differences prevent a causal explanation from this pair alone." }, { "title": "Timing is not a universal early-good / late-bad rule", "sources": [ "S6", "S10" ], "observation": "The harmful seeded experiment began before tau injection; a different antibody had favorable pathology findings when started in older amyloid mice.", "inference": "Stage must be defined by pathology and cell state, not age alone. The papers do not locate a human treatment window." }, { "title": "Activation mode deserves a controlled comparison", "sources": [ "S10", "S6", "S8" ], "observation": "03O05 increased sTREM2. The human AL002 trial lowered CSF sTREM2; acute AL002a experiments also reported increases.", "inference": "Do not collapse antibodies or soluble and membrane TREM2 into a simple good/bad rule. Shedding, signaling duration and brain exposure need matched measurements; higher sTREM2 is not itself proof of protection." }, { "title": "A cellular response is not a clinical success", "sources": [ "S1", "S2", "S3", "S8" ], "observation": "Functional rescue and expression-dependent metabolic responses coexist with a negative human efficacy trial.", "inference": "Select interventions by neuronal outcomes and safety, alongside target engagement, rather than treating a biomarker change as a therapeutic result." } ], "priorities": [ { "rank": 1, "title": "State-specific protection with a tau safety check", "reason": "Best next mechanistic question: when does functional microglial rescue coincide with less neuronal injury?", "test": "Compare matched amyloid-only and amyloid-plus-tau conditions, stratify baseline TREM2 expression and functional impairment, and measure actual CNS exposure. Use neuronal injury as the prespecified primary endpoint; metabolism, tau spread and vascular injury are explanatory or safety endpoints.", "gate": "Advance only if neuronal protection replicates independently without a prespecified unacceptable increase in tau or vascular injury. A metabolic improvement alone does not pass.", "status": "Priority hypothesis and proposed design; no experiment executed" }, { "rank": 2, "title": "Compare agonist mechanisms under matched conditions", "reason": "Binding epitope, shedding and signal duration could help explain conflicting outcomes; current comparisons are confounded.", "test": "Compare ligand-mimetic and other agonist approaches with matched exposure, appropriate isotype controls, Fc characterization, membrane/soluble TREM2 measures and the same injury endpoints.", "gate": "Deprioritize the proposed mechanism if its predicted advantage disappears with matched controls or if soluble TREM2 shifts without neuronal protection.", "status": "Exploratory mechanism; no validated superiority claim" }, { "rank": 3, "title": "Keep vascular and myelin outcomes as decision gates", "reason": "A favorable result in one tissue compartment may hide a worse result in another.", "test": "Extract vessel-wall versus parenchymal amyloid, barrier injury and repair outcomes separately. First complete endpoint-level sample sizes, blinding, exclusions and uncertainty estimates.", "gate": "Do not interpret lower total amyloid or a repair biomarker as sufficient evidence of overall benefit.", "status": "Required evidence work before selecting a translational candidate" } ], "clinicalEstimates": { "source": "S8", "endpoint": "Week-96 change in CDR-SB: least-squares mean difference versus placebo (95% CI)", "rows": [ { "dose": "15 mg/kg", "difference": -0.31, "low": -1.61, "high": 0.98 }, { "dose": "40 mg/kg", "difference": 0.13, "low": -1.18, "high": 1.43 }, { "dose": "60 mg/kg", "difference": -0.17, "low": -1.49, "high": 1.15 } ], "interpretation": "All intervals include zero; the primary endpoint was not met. These estimates concern this trial, not a pooled estimate for TREM2 biology." }, "method": { "scope": "Ten purposively selected primary publications; eleven context rows. Eight publications carried forward from R-002 plus two direct agonist comparators. This is a targeted narrative comparison, not a systematic review or meta-analysis.", "searchDate": "2026-10-11", "retrieval": "Follow-up of R-002 references and primary publisher/PubMed pages. Targeted searches concerned AL002a chronic activation, neutral/detrimental agonists, TREM2 expression-dependent metabolism and ligand-mimetic agonism. No exhaustive retrieval or screening denominator was recorded.", "classification": "Benefit, no demonstrated benefit and harm refer only to the stated endpoints and experimental context. Mixed includes trade-offs and cellular responses with uncertain therapeutic meaning. Categories are not certainty grades. Papers and rows are not votes.", "limitations": "Incomplete endpoint-level extraction, no formal risk-of-bias assessment, no pooled effects, no independent expert review and no reanalysis of underlying datasets. Missing sample sizes and regimens are explicitly marked. The trial subscription full text was not reviewed; several inherited records remain abstract-level. Funding and conflicts were not systematically compared. Access depth is listed per source. No exhaustive correction/retraction check is claimed.", "maintenance": "Dated editorial report, separate from automatic publication scans. Future evidence changes require a versioned review. The website did not run new biological experiments or discover a validated therapeutic connection.", "nextDeliverable": "A matched-comparison extraction sheet with all biological replicate counts, sex, dose/exposure, controls, blinding, exclusions and outcome uncertainties; followed by an independently reviewed protocol. No immediate human treatment experiment is proposed." } }