Relevance of transgenic mouse models for Alzheimer's disease

Highlights

• Alzheimer's disease (AD) is a uniquely human condition. However, transgenic mouse models that mimic specific aspects of AD pathology are necessary to evaluate hypotheses and molecular mechanism not accessible in humans.

• Recognizing the complexity of AD is important if expectations from any form of research, whether it involves human subjects or mouse models, are to be managed.

• A number of models of AD-relevant pathology have been developed, each with strengths and limitations that should be fully understood before use.

• Transgenic mice are available that incorporate universal biological variables that drive AD risk and pathology (age, sex, APOE), and AD-related gene mutations (TREM2, ABCA7, PICALM) that affect a limited cohort of AD patients.

• The frequently expressed sentiment “We have cured AD mice hundreds of times, but the cures never work in humans, so we need better models” is false. No mouse model fully replicates AD and no treatment has fully “cured” even the limited pathology of any mouse model.

• The most relevant question for the use of a model should be: Is the choice of this rodent model appropriate for the question being addressed? While transgenic mice do not in fact have AD, they can tell us a great deal about specific aspects of AD disease mechanisms if the question is properly framed.

• After selecting the model, studies must be planned and executed with an understanding of what information is provided by the readouts.

• Current funding mechanisms and career development requirements are impediments to full hypothesis testing, either mechanistic or therapeutic.

• Sensationalization of data, promulgation of dogma, and copycat science are hindering the optimal use of transgenic models of AD. Developing new models will not solve this problem.

• Answers to fundamental questions about AD that remain elusive may be addressed using current models, or through careful consideration of what a new model should replicate.

• As always, the best science comes from asking the right questions, not from following the latest trends in either technology or posited causality.

Abstract

Over the last several decades, a number of mouse models have been generated for mechanistic and preclinical therapeutic research on Alzheimer's disease (AD)-like behavioral impairments and pathology. Acceptance or rejection of these models by the scientific community is playing a prominent role in how research findings are viewed and whether grants get funded and manuscripts published. The question of whether models are useful has become an exceptionally contentious issue. Much time and effort have gone into investigators debating comments such as “there are no mouse models of AD,” “…nice work but needs to be tested in another mouse model,” or “only data from humans is valid.” This leads to extensive written justifications for the choice of a model in grant applications, to the point of almost apologizing for the use of models. These debates also lead to initiatives to create new, better models of AD without consideration of what “better” may mean in this context. On the “other side,” an argument supporting the use of mouse models is one cannot dissect a biological mechanism in postmortem human tissue. In this chapter, we examine issues that we believe must be addressed if in vivo AD research is to progress. We opine that it is not the models that are the issue, but rather a lack of understanding the aspects of AD-like pathology the models were designed to mimic. The goal here is to improve the utilization of models to address critical issues, not to offer a critique of existing models or make endorsements.

Introduction

There are diseases that are uniquely human, since no orthologous disease has been identified in other species. Many adult-onset neurodegenerative diseases fall into this category, including Alzheimer's disease (AD). A corollary of this absence of equivalent diseases is that there are few, if any, effective treatments for this category of disease. Recent history suggests that we need to understand the molecular basis of pathogenesis in these diseases if we are to develop effective therapies. However, many kinds of analysis cannot be done in human patients. Therefore, we are often left with descriptions of postmortem patient tissues that retain few indications of early stage disease and often are confounded by generic neuronal damage (e.g., mitochondrial and other organelle dysfunction, markers of reactive oxygen species stress, lower neuronal marker levels, changes in neuron structure) that is not disease specific and has not been useful in developing treatments.

Sequencing of the human genome and identification of mutations that produce AD in humans led to hope that a genetic solution might be found, but this approach has failed to produce clinically useful therapies. This may reflect (at least in part) the delayed onset of these diseases, which in turn indicates a phenotypic complexity that is not simply a result of gene expression but instead reflects interactions with the environment and posttranslational effects. However, identification of gene mutations that result in AD has led to the generation of animal models expressing mutant human proteins as models. Typically, these models produce pathological changes that partially replicate changes seen in human patients but fail to capture the full characteristics of the disease. Undeterred, a wide range of therapeutic approaches have been considered and many that showed limited success in animal models have been taken to clinical trials and failed. Investigators have reacted by blaming the animal models for the failure to develop therapies, leading to ever more complex models or complete dismissal of animal models as a means to better understand the disease process. This creates a dilemma—if animal models are dismissed then how do we model AD? The time is right to consider the use of models in the study of AD and other adult-onset neurodegenerative diseases. We propose that animal models should not be dismissed but recognized for their strengths in modeling specific aspects of AD-relevant pathology.