The trouble with mice as behavioral models for Alzheimer’s and other neurologic diseases

There’s been a lot of talk about overinvestment in interventions aimed at amyloid in the weeks since Biogen discontinued a late-stage study of aducanumab, an experimental therapy for Alzheimer’s disease.

Although much of the focus has been on the amyloid hypothesis at the heart of that work — and other failed treatments — I believe we are overlooking another key driver for numerous translational failures: the overreliance on behavioral readouts from contrived transgenic rodent models to guide drug development for Alzheimer’s, Parkinson’s, and other neurologic diseases. We need to find ways to move beyond this flawed paradigm.

The Food and Drug Administration has put in place well-established guidelines for preclinical safety studies. But the FDA does not typically weigh in on preclinical effectiveness studies. It’s generally up to individual companies to decide when an investigational compound has shown enough potential to merit advancing it into high-risk and high-cost clinical trials.

In many areas of drug development, the most important preclinical metric is whether the compound showed signs of efficacy in an animal model, most often mice.

This makes sense for some diseases, and for certain forms of data. In movement disorders, for example, motor dysfunction is arguably conserved between mouse and human. Similarly, if an agent heals a wound in a mouse, it’s reasonable to assume it may do the same in people. But this breaks down when it comes to neurologic disorders that manifest themselves in distinctly human cognitive and functional behaviors. It is hard to argue with a straight face that assessing how speedily mice find their way around a maze appropriately models the complex interplay of cognitive, behavioral, and functional symptoms in human diseases such as Alzheimer’s or autism spectrum disorder.

Disclaimer: I’m not a neurologist. I do, however, run a neuroscience-focused biotech company focused on enhancing synaptic integrity to treat neurologic disorders, and I’ve worked with incredibly talented neuroscientists for the better part of 15 years, so my views are informed by their collective wisdom. And sometimes it takes an outsider to point out the obvious.

What seems obvious to me is that the animal model paradigm is broken when it comes to complex neurologic diseases.

Let’s start with the highly inbred mouse models largely used for this kind of work. They lack the genetic diversity that may contribute to the severity and complexity of Alzheimer’s and other complex diseases in people. This is arguably one reason why many preclinical studies in rodent models have failed to translate to humans.

A recent paper in the journal Neuron illustrates this point. Researchers started with a particular strain of mice created with a set of genes known as 5XFAD, which contains multiple mutations linked to Alzheimer’s in humans. These mice were then cross-bred with 28 different mouse strains. The genetic makeup and background of the offspring varied considerably, although all had the 5XFAD genes. As it turned out, the gene set had vastly different effects on behavior in different mouse strains.

Most of the 5XFAD mice showed impaired memory with aging and, taken as a group, there was an average acceleration of impairment compared to their parent strains. Yet the effect size varied considerably. In some strains, 5XFAD had a strong effect, in others it had little or no effect. In a few, the gene set was remarkably associated with improved memory relative to the parent strain! Clearly, the effects of an experimental compound would likely vary a great deal depending on which strain of mice was used for preclinical testing.

If mice genetics are complicated, human genetics is infinitely more so. Furthermore, it is impossible to model in mice all of the complex environmental inputs — everything from nutrition to education to exercise levels — that affect cognition, function, and behavior in people with neurologic disorders.

Other variables such as transgenic drift, technician variability, and confirmation bias also play roles in the lack of translational power with behavioral phenotypes from transgenic animal models.

I don’t mean to suggest that transgenic animal models have no utility in translational neuroscience. They are clearly helpful for basic research and to study drug effects at the molecular, structural, and functional levels. But just as clearly it’s problematic across multiple neurologic diseases to rely on rodent behavioral studies as a critical go/no go signal for advancing therapies into clinical trials.

So how do we move forward?

To begin, it helps to have this discussion frankly — and to have it with investors, board members, and executive teams. We must all be hesitant to overweight behavioral phenotypes when making key decisions in late-preclinical neuroscience drug discovery.

We also must continue to advance novel preclinical approaches to evaluate the effect of a therapy before deciding whether it merits a clinical trial. Across the industry, we’ve seen advances in tools to measure molecular, structural, functional, and circuitry changes in the brain. Now it’s time to give those experimental results more weight as we evaluate a compound’s chances for success.

Affirmative data from these tools do not, of course, constitute proof that a drug will do the same thing in the human brain that was observed in the rodent brain. But they do suggest that the therapy has quantitative molecular and structural effects on neurons, brain circuitry, and the like — effects on systems that we have solid reasons to believe more closely mirror the human brain than rodent behavior mirrors human behavior. Still, many in the field are unwilling to abandon requiring rodent behavioral activity as a must-have before moving a compound into the clinic.

We’ll never know how many compounds were moved into the clinic based on questionable behavioral data. We’ll also never know how many otherwise promising compounds were shelved for failure to show “efficacy” in improving cognition in a flawed mouse model.

And none of this touches on the ethical considerations of using animals for research purposes, which should always be conducted in the most humane and thoughtful ways.

For all of our collective disappointment over yet another failed amyloid therapy, I am confident that the next decade of neuroscience research and development will be better than the last — as long as we don’t continue overinvesting in mechanisms that have proved fruitless and as long as we don’t continue to use behavioral data from contrived mouse models as a critical gatekeeper before clinical trials.

It’s time for the industry to evolve. We should start by taking a hard look at old paradigms.

Adam Rosenberg is the president and chief executive officer of Rodin Therapeutics, a company based in Cambridge, Mass., aimed at discovering and developing therapies to enhance synaptic integrity to treat neurodegenerative, neurodevelopmental, and neuropsychiatric diseases.