The Multifaceted Challenge of Modeling Alzheimer's Disease

Field Application Scientist Taconic Biosciences

The mechanisms behind Alzheimer’s disease continue to elude investigators, with each new study lending credence to different pathological hypotheses. Recently, therapies like Biogen’s Aduhelm (which has secured FDA approval) and Eli Lilly’s donanemab (which has received breakthrough therapy status) have demonstrated a reduction in amyloid plaques, fueling hope for a cure. Yet the jury is still out on whether the presence of amyloid plaques fully correlates with the development of, and cognitive deficits associated with, Alzheimer’s disease. In this complicated, ever-changing research landscape, it’s vital that preclinical mouse models evolve to support the study of the varied mechanisms believed to be at work.

Mechanisms underlying Alzheimer’s disease

Alzheimer’s disease is a multifaceted progressive neurodegenerative disease, of which two forms exist. Familial or early-onset Alzheimer’s disease, which is caused by dominant mutations in a few key genes, accounts for <5% of patients. Sporadic or late-onset Alzheimer’s disease, which is associated with both genetic and environmental factors, accounts for >95% of patients.

There are many hypotheses of Alzheimer’s disease, yet two have dominated the field: the amyloid cascade hypothesis and the tau hypothesis. According to the amyloid cascade hypothesis, mutations in the amyloid precursor protein (APP) lead to amyloid accumulation or the formation of misfolded amyloid β (Aβ) around neurons, forming amyloid plaques. According to the tau hypothesis, neurofibrillary tangles are composed of hyperphosphorylated microtubule-associated protein tau (MAPT). Despite research focused on these hypotheses, the molecular links between Aβ and tau remain unclear.

With even an approved, much-touted therapy like Aduhelm generating significant controversy and debate, including concerns regarding its effectiveness in mitigating cognitive impairment, it’s apparent that more research is necessary—requiring preclinical models that enable investigators to study hypotheses beyond Aβ to advance our understanding of this disease and expedite our development of effective therapeutics.

The state of Alzheimer’s disease mouse models

There are over 200 Alzheimer’s disease preclinical models, many of which are transgenic (Tg) mice. They are summarized in a recent review (Cacabelos et al. Expert Opin. Drug Discov. 2021 Aug 25; 1–26). The most widely used models focus on amyloid plaques and overwhelmingly consist of genetic mutations representative of familial forms of Alzheimer’s disease, including the insertion of human genes encoding APP, presenilin 1 (PSEN1), and/or presenilin 2 (PSEN2), leading to the overproduction of Aβ. Insertion of the MAPT gene, which leads to neurofibrillary tangle accumulation, is also widely investigated.

To study the interaction between Aβ and tau, individual models can be crossed. For example, the TAPP model crosses Tg2576 (mice expressing the APP Swedish mutation) with JNPL3 (mice expressing the P301L mutation in tau). The 3xTg model consists of APP (Swedish mutation), PSEN1, and MAPT for investigation of both plaques and tangles, while the 5xFAD model, considered more progressive, consists of APP (Swedish, Florida, and London mutations) and PSEN1 (M146L and L286V mutations) for plaques but lacks tangles. Recently, the 5xFAD was crossed with P301L to produce the 6xTg model.

These models share the premise that rapid disease onset follows Aβ overexpression and/or tau overexpression. This premise, however, poses three difficulties: there are aging-related discrepancies due to early and aggressive onset of the disease, sometimes as early as in juveniles; transgene overexpression leads to plaque and tau overaccumulation that is not physiologically relevant to what is observed in human disease; and additional mechanisms beyond Aβ and tau are ignored.

Where Alzheimer’s models are headed

The ability to effectively model sporadic disease will be critical to advancing Alzheimer’s research. Such models more closely represent the clinical population, but they have not been widely adopted, in part because they are complex to make and require long lead times to detect phenotypical alterations. The Model Organism Development and Evaluation for Late-onset Alzheimer’s Disease (MODEL-AD) consortium was recently established with the goal of developing at least 50 sporadic models (Oblak et al. Alzheimer’s Dement. 2020; 6: e12110), which is an important step.

There are many validated risk factors in sporadic Alzheimer’s disease, with the two most promising targets being apolipoprotein E4 (APOE4) and triggering receptor expressed on myeloid cells 2 (TREM2). APOE4 is the most common risk factor and is involved with lipid metabolism and inflammation, making it ideal for comorbidity assessment. However, it is absent in mice, so models without humanized APOE4 miss critical mechanisms associated with the human disease. TREM2 is expressed by microglia, which are innate brain immune cells that mediate inflammation and may play a protective role by promoting plaque and tau clearance.

One key approach for advancing sporadic Alzheimer’s disease models and improving their translatability will be the enhancement of existing models by crossing them with humanized or knock-in APOE4 and/or TREM2. Additionally, designing models on diverse genetic backgrounds rather than on inbred mice, or designing models with the microbiota of wild mice (wildR), will increase translatability to human heterogeneity. Regardless, there is a need to standardize the strain background because it has been shown that the strain background can affect disease pathology even within the same model.

The immune and inflammatory systems differ between mice and humans. With a surge of interest in microglia, it has been observed that human microglia respond differently to plaques than mouse microglia, and so models with humanized microglia may be critical moving forward. Indeed, researchers are going as far as transplanting human pluripotent stem cells or selected human microglia directly in Alzheimer’s models that are also immunodeficient. Regardless of microglial differences, exciting research is fueling new mechanistic insights underlying inflammation and the immune response observed in both humans and mice.

As reported recently (McAlpine et al. Nature 2021; 595: 701–706), communication between astrocytes and microglia may be mediated by interleukin-3 (IL-3) in both humans and mice. Specifically in diseased patients and mice, astrocytes will release IL-3, which initiates microglia to clear Aβ and tau, ultimately alleviating pathology and cognitive deficits. Mice lacking IL-3 had significantly more Aβ and cognitive impairment than mice with IL-3.

It’s important to continue improving the original Aβ and tau models as well. New models need to feature more physiologically relevant levels of human amyloid proteins, for example. The APPNL-F model avoids artifacts from APP overexpression using a knock-in approach to express humanized APP.

Interest in tau is increasing, and data from a recent human positron emission tomography (PET) study illustrated that tau may be a superior diagnostic tool for disease progression compared to amyloid (Ossenkoppele et al. JAMA Neurology 2021; 78: 961–971). However, tau mouse models need to evolve as there are actually no MAPT mutations associated with human Alzheimer’s disease. Research should also focus on the mechanisms underlying why plaques and tangles form. As reported in a recent article (Lam et al. PLOS Biology 2021; 19(9): e3001358), Aβ may actually originate from outside the brain. Amyloid protein made exclusively in the liver was transported into the brain by lipoproteins and contributed to disease progression in mice.

Since no single model of Alzheimer’s disease is able to fully recapitulate the many facets of this condition, the most effective approach to improving translatability from preclinical research to the human patient is likely to be the incorporation of multiple models, each contributing different insights. Both genetically engineered models of the disease and aged models that allow the study of age-related phenotype progression are likely to play a role.

Notwithstanding the multifaceted challenges and myriad hypotheses about Alzheimer’s disease, mouse models continue to advance our understanding of this complex condition. Recent developments demonstrate hope for viable treatments, yet there is much work to be done. As preclinical models evolve, investigators will continue to leverage these tools as they work to explore and develop therapies that may alter life for Alzheimer’s patients.

Moriah L. Jacobson, PhD ([email protected]), is a field application scientist at Taconic Biosciences.