How to Choose the Right Alzheimer's Disease Research Model

Alzheimer's disease (AD) is a complex, progressive neurodegenerative disorder. Its core pathological hallmarks are the senile plaques formed by abnormal deposition of β-amyloid (Aβ) in the brain and the neurofibrillary tangles (NFTs) caused by hyperphosphorylation of Tau protein. Despite enormous global investment, AD pathogenesis remains incompletely understood and no curative therapy exists. Against this background, animal models that reproduce the pathological progression and clinical symptoms of human AD have become indispensable tools for understanding disease mechanisms, testing hypotheses, and screening candidate therapies.

The development of AD animal models mirrors the deepening of our understanding of the disease itself: from early simplified models built on a single pathological hypothesis (such as cholinergic damage), to transgenic models that reproduce core molecular pathology through genetic engineering, and now to a new generation of models designed to mimic human genetic variation with precision. Each stage has pushed the boundaries of AD research further.

Figure 1. Major pathological features of AD in the central nervous system ^[1].

The Past: The Dominance and Limits of the Aβ Hypothesis

Since the late 1980s, the "Aβ cascade hypothesis" dominated the field, asserting that abnormal Aβ deposition initiates AD pathology, with Tau hyperphosphorylation occurring as a downstream event. Within this framework, most therapeutic strategies targeted Aβ — clearing existing plaques, inhibiting Aβ production, or preventing its aggregation.

Non-Transgenic Models and Limitations

Early non-transgenic models used chemical, physical, or surgical methods to mimic isolated symptoms rather than full pathology:

• Cholinergic dysfunction models: Scopolamine or ibotenic acid lesions model cognitive/learning deficits, but fail to generate Aβ plaques or Tau tangles.
• Aβ injection models: Direct intracerebroventricular injection of aggregated Aβ rapidly induces memory deficits, offering a fast tool for toxicity studies.
• Other lesion models: D-galactose (oxidative stress/aging) and aluminum chloride cause neurotoxicity, but poorly mirror human AD etiology and progression.

Figure 2. Non-transgenic mouse models used in AD research ^[2].

The existence and limitations of non-transgenic models reflect an early, partial understanding of AD that emphasized single symptoms rather than the complex pathological cascade. As research advanced, scientists recognized that AD pathology extends well beyond this to include Aβ plaques and Tau tangles. This limitation drove the field toward transgenic models that better represent complete AD pathology.

Transgenic mouse models: amyloid (Aβ) models

To model AD pathology and advance research, the field turned to transgenic mice, introducing human Aβ-related genes — such as amyloid precursor protein (APP) and presenilin 1 (PSEN1) — into the mouse genome. Frequently used models include APP/PS1, Tg2576, and 5xFAD ^[3]. These models generally overexpress APP or carry additional PSEN1 mutations to promote Aβ deposition and AD-like amyloid plaque pathology.

Table 1: Comparison of commonly used mouse models

The failure of many early Aβ-targeting drugs in clinical trials suggests that simply lowering Aβ burden does not necessarily deliver cognitive benefit. That said, anti-Aβ monoclonal antibodies such as lecanemab have shown slowing of clinical progression in early symptomatic AD, providing clinical evidence for Aβ-targeted intervention. Current therapies still cannot reverse the disease and carry risks such as amyloid-related imaging abnormalities (ARIA). Together, these advances and limitations have pushed researchers to focus on treatment timing, Tau pathology, and other disease mechanisms ^[4].

The Present: Multi-Mechanism Therapies and Brain Delivery Technologies Shape Model Selection

AD drug development is now advancing interventions against Aβ, Tau, and other mechanisms in parallel. For animal model research, the requirement is to match model characteristics to the drug's mechanism, route of administration, and endpoints — with reduction of pathological protein, brain drug exposure, and cognitive improvement each verified separately. The advances below illustrate the implications for model selection.

Tau-targeted therapy deserves particular attention. Beyond blocking Tau aggregation or propagation, nucleic acid drugs that reduce Tau production have entered clinical validation. In July 2026, Biogen’s Phase II trial for the antisense oligonucleotide diranersen (BIIB080) showed the drug reduced CSF Tau by 50%–65%, lowered pathological burden on PET scans, and slowed cognitive decline by 26% over 18 months. Although the study missed its primary dose-response endpoint and requires Phase III confirmation, these results strongly validate ongoing research into Tau-lowering treatments and emphasize the need for humanized MAPT models to develop sequence-dependent drugs ^[5].

APOE4-related lipid metabolism, microglial function, and neuroinflammation offer further entry points. A 2026 preclinical study found APOE4 influences microglial cholesterol efflux and Aβ phagocytosis via Asxl1–LXRα-related epigenetic regulation, suggesting genetic background and cell state belong in evaluation. But mechanism does not equal benefit: the TREM2 agonist antibody AL002 showed target engagement and microglial activation, yet failed to improve clinical outcomes in Phase II. Such models should therefore assess immune response, pathological burden, and neurological function together, not inflammatory markers alone ^[6-7].

Blood-brain barrier (BBB) delivery offers another route for large molecules. Trontinemab pairs Aβ targeting with a transferrin receptor 1 (TfR1)-mediated brain shuttle to enhance antibody transport across the BBB.

Figure 3. Comparison of anti-amyloid antibody delivery with and without brain shuttle technology ^[8].

At AAIC 2026, Roche presented updated early clinical data and the Phase III PrevenTRON study design in cognitively unimpaired individuals, though clinical benefit requires confirmation. Developing such drugs necessitates receptor-humanized models to address human-mouse binding differences, differentiate vascular from parenchymal exposure, and evaluate safety within an AD pathological context.

These developments also refine Tau model usage: sequence-targeting drugs (ASOs or siRNAs) require models retaining human MAPT target sequences, while anti-aggregation therapies demand reproducible Tau pathology. Because reducing MAPT mRNA does not guarantee decreased Tau aggregation or cognitive recovery, researchers must select protein, pathological, and functional endpoints aligned with their specific research objectives.

Figure 4. Timeline of transgenic mouse model development for AD research, compiled from the literature ^[10].

For studies that must address Aβ and Tau pathology together, the 3xTg-AD triple-transgenic mouse is a commonly used tool. This model combines the APP Swedish, Psen1 M146V, and MAPT P301L mutations to study interacting Aβ and Tau pathology and evaluate combination therapies. Because pathological timing varies by genetic background, sex, and breeding colony, experiments require strain-validated schedules rather than fixed age timetables. Furthermore, since the MAPT P301L mutation originates from frontotemporal lobar degeneration rather than typical AD, 3xTg-AD effectively addresses specific dual-pathology questions but cannot fully replicate sporadic AD ^[11].

Figure 5. Timeline of AD pathogenesis in the 3xTg-AD model ^[12].

The Future: New Model Development and Precision Drug Evaluation

To support Tau-targeted drug development, Cyagen created Tau-humanized mouse models, led by the foundational hutau(MAPT) model, which replaces mouse Mapt with human MAPT to express human 3R and 4R Tau isoforms for splicing and distribution studies. Expanding on this platform, Cyagen has developed hutau-P301L (Cat. No. C001835) and hutau-P301S (Cat. No. C001836) humanized models for mutant Tau research and drug evaluation. These mutations are primarily associated with hereditary tauopathies and do not directly reproduce AD-associated Tau pathology. Their suitability for AD research should therefore be assessed using validated expression, pathology, and behavioral data for each strain.

Cyagen offers humanized TFRC and CD98HC models for blood-brain barrier delivery, APOE4 and other novel AD-targeted gene humanized models, multi-target cross models, and gene-mutated iPSC-derived neurons to evaluate brain delivery, immune responses, pathology, and neurological function.

Validation data show that intracerebroventricular siRNA administration reduced human MAPT mRNA across multiple brain regions in hutau(MAPT) mice (Figure 5). These results support the model’s use in evaluating therapies targeting the human MAPT sequence. Since intracerebroventricular administration bypasses the BBB, brain penetration following peripheral dosing requires separate evaluation.

Figure 6. hutau(MAPT) mice used for preclinical efficacy assessment of different human Tau-targeting siRNAs (data provided by a customer).

Humanized models provide Tau-targeted drugs with evaluation tools matched to human sequences. In practice, target-expression models should be distinguished from disease-phenotype models: the former verify whether a drug engages the human target, the latter test whether the intervention changes pathology or function. If the study involves peripheral administration, brain delivery and tissue distribution assessments must also be included.

Conclusion: Match the Model to the Research Question

From non-transgenic models to classic transgenic models and now humanized models, the toolkit for AD research keeps expanding. The key to model selection is clarity about the question being asked: verifying target engagement, observing pathological change, or evaluating brain penetration and functional improvement. Aβ, Tau, immunometabolism, and brain delivery research each have different requirements, and no single model covers every element. Combining validated animal models with human cells or brain organoids, and using clinically relevant evaluation metrics, helps produce a more accurate assessment of a candidate therapy's potential and limitations.

👉 [Find the Right Preclinical Model for Your Research]

👉 [Explore Our Alzheimer's Disease Preclinical CRO Solutions]

👉 [Discuss Your Next Preclinical Study with Our Experts]

Reference

[1] Sonsalla MM, Lamming DW. Geroprotective interventions in the 3xTg mouse model of Alzheimer's disease. Geroscience. 2023 Jun;45(3):1343-1381. doi: 10.1007/s11357-023-00782-w. Epub 2023 Apr 6. PMID: 37022634; PMCID: PMC10400530.

[2] Hernández-Rodríguez M, Vega López JM, Martínez-Rosas M, Nicolás-Vázquez MI, Mera Jiménez E. Murine Non-Transgenic Models of Alzheimer's Disease Pathology: Focus on Risk Factors. Brain Sci. 2025 Mar 19;15(3):322. doi: 10.3390/brainsci15030322. PMID: 40149843; PMCID: PMC11940003.

[3] Kosel F, Pelley JMS, Franklin TB. Behavioural and psychological symptoms of dementia in mouse models of Alzheimer's disease-related pathology. Neurosci Biobehav Rev. 2020 May;112:634-647. doi: 10.1016/j.neubiorev.2020.02.012. Epub 2020 Feb 15. PMID: 32070692.

[4] Pádua MS, Guil-Guerrero JL, Prates JAM, Lopes PA. Insights on the Use of Transgenic Mice Models in Alzheimer's Disease Research. Int J Mol Sci. 2024 Feb 28;25(5):2805. doi: 10.3390/ijms25052805. PMID: 38474051; PMCID: PMC10931675.

[5] Biogen. Biogen presents phase 2 CELIA data at AAIC demonstrating meaningful clinical outcomes and robust tau reduction with diranersen in early Alzheimer’s disease [Internet]. Cambridge (MA): Biogen; 2026 Jul 14 [cited 2026 Sep 17]. Available from: https://investors.biogen.com/news-releases/news-release-details/biogen-presents-phase-2-celia-data-aaic-demonstrating-meaningful.

[6] Lin L, Pan Z, Wei Z, Jiang X, Lai Y, Chen M, Lin F. ApoE4 Drives Microglial Lipid Dysregulation in Alzheimer's Disease via Epigenetic Reprogramming of the Asxl1/LXRα-H3K4me3 Axis. J Neuroinflammation. 2026 Mar 11;23(1):124. doi: 10.1186/s12974-026-03740-3. PMID: 41808104; PMCID: PMC13088513.

[7] Alector, Inc. Alector announces results from AL002 INVOKE-2 phase 2 trial in individuals with early Alzheimer’s disease and provides business update [Internet]. South San Francisco (CA): Alector, Inc.; 2024 Nov 25 [cited 2026 Sep 17]. Available from: https://investors.alector.com/news-releases/news-release-details/alector-announces-results-al002-invoke-2-phase-2-trial.

[8] Taylor E. Potential Alzheimer’s treatment, trontinemab, hits the news – how does it work and is it available? [Internet]. Registro Brasileiro de Ensaios Clínicos (ReBEC); 2025 Oct 3 [cited 2026 Sep 17]. Available from: https://ensaiosclinicos.gov.br/news/381. Originally published by Alzheimer’s Research UK.

[9] Roche. Roche presents new data in Alzheimer’s disease from across its integrated pharmaceutical and diagnostics portfolio at AAIC [Internet]. Basel (Switzerland): Roche; 2026 Jul 7 [cited 2026 Sep 17]. Available from: https://www.roche.com/media/releases/med-cor-2026-07-07

[10] Javonillo DI, Tran KM, Phan J, Hingco E, Kramár EA, da Cunha C, Forner S, Kawauchi S, Milinkeviciute G, Gomez-Arboledas A, Neumann J, Banh CE, Huynh M, Matheos DP, Rezaie N, Alcantara JA, Mortazavi A, Wood MA, Tenner AJ, MacGregor GR, Green KN, LaFerla FM. Systematic Phenotyping and Characterization of the 3xTg-AD Mouse Model of Alzheimer's Disease. Front Neurosci. 2022 Jan 24;15:785276. doi: 10.3389/fnins.2021.785276. PMID: 35140584; PMCID: PMC8818877.

[11] Kametani F, Hasegawa M. Reconsideration of Amyloid Hypothesis and Tau Hypothesis in Alzheimer's Disease. Front Neurosci. 2018 Jan 30;12:25. doi: 10.3389/fnins.2018.00025. PMID: 29440986; PMCID: PMC5797629.

[12] Reyna NC, Clark BJ, Hamilton DA, Pentkowski NS. Anxiety and Alzheimer's disease pathogenesis: focus on 5-HT and CRF systems in 3xTg-AD and TgF344-AD animal models. Front Aging Neurosci. 2023 Nov 10;15:1251075. doi: 10.3389/fnagi.2023.1251075. PMID: 38076543; PMCID: PMC10699143.