Progress towards improving mouse models for preclinical testing of treatments for Alzheimer’s disease
Abstract
Alzheimer’s disease (AD) research has accomplished significant strides in recent years, particularly with the first regulatory approvals of disease-modifying therapies and a comprehensive revision of disease diagnostic criteria. Both developments place new demands on preclinical mouse models used to test potential treatments. This review summarizes recent progress in optimizing mouse models for Alzheimer’s research, using the NIA-funded Model Organism Development and Evaluation for Late-onset Alzheimer’s Disease (MODEL-AD) program as a case study, with specific focus on their relevance for preclinical therapeutic testing, their genetic and molecular fidelity to the human disease, and the contribution of the MODEL-AD program. Key advancements discussed include the shift from transgenic overexpression models to knock-in and genome-replacement strategies; the development of late-onset Alzheimer’s models (LOAD), LOAD1, LOAD2, LOAD3 and related LOAD-platform mice expressing humanized APOE, APP, and MAPT; the integration of environmental risk factors such as high-fat diet and toxic metal exposure into gene x environment interaction models; detection of prodromal stages of disease using multi-omics signatures; and the use of clinically relevant biomarkers, including plasma NfL, GFAP, and p-tau217, as cross-species translational endpoints. We discuss the importance of diverse genetic backgrounds, the contribution of the AD Knowledge Portal as an open-science data-sharing resource, and the principles of rigorous preclinical study design required to maximize translational success. We also discuss the limitations of these models and the evidence still needed to establish their predictive validity for therapeutics. These advances do not yet amount to validated models of late-onset AD, but they give the field better tools and frameworks for selecting and validating mouse models for specific human mechanisms and therapeutic questions. Animal models based on the genetics of late-onset AD (LOAD), rather than familial AD (fAD), are needed to model mechanisms relevant to the majority of AD patients, who do not carry autosomal dominant mutations; fAD-based models remain valuable for amyloid, CAA, and ARIA biology. Knock-in and genome-replacement mouse models expressing humanized variants of APOE, APP, and MAPT at physiological levels avoid the non-physiological expression levels and insertional mutagenesis of transgenic overexpression lines, making them better-controlled platforms for many questions. The MODEL-AD program has created a series of progressively refined LOAD platform mice, LOAD1 (APOE4/Trem2*R47H), LOAD2 (APOE4/Trem2*R47H/hAβ), and LOAD3 (APOE4/hAβ/hMAPT), progressively adding major LOAD risk alleles to a common platform. Improved genetic (construct) validity does not by itself establish predictive validity: no knock-in LOAD model has yet been benchmarked against clinical trial outcomes, and doing so requires testing drugs with known clinical results. Environmental risk factors, including high-fat diet, vascular risk, and toxic metal exposures, can be combined with genetic risk factors and aging to produce clinically relevant disease phenotypes suited for testing therapeutics independent of amyloid and tau pathology. Clinically aligned phenotyping modalities such as plasma NfL, GFAP, and p-tau217, in vivo PET/MRI, and touchscreen cognitive assessment, aligned with the 2024 revised Alzheimer’s Association criteria, should be incorporated into mouse model evaluation pipelines to maximize translational utility. Genetic diversity across inbred backgrounds, wild-derived strains, and outbred stocks substantially influences the expression of AD phenotypes, including neuroinflammation and cerebral amyloid angiopathy; this diversity must be harnessed rather than minimized. Each therapeutic compound should be matched to the right target, in the right model, at the right age, using the right assay, with study designs that mirror clinical trial standards including pre-specified endpoints, randomization, blinding, and inclusion of equal numbers of both sexes.