Mouse Models: Handle with Care
Researchers find many “ALS mouse” results don't translate to humans


by Jim Schnabel

May 19, 2008

For more than a decade, one of the most important tools for researching amyotrophic lateral sclerosis (ALS) has been the so-called SOD1 mouse. Designed to express multiple copies of a defective gene found in some people whose families have a history of developing ALS, this transgenic mouse develops an accelerated ALS-like brain disease at about four months of age. Researchers routinely have identified ALS drugs for human testing by first seeing whether they extend lifespan in these mice.

“We all assumed that we could find a better drug by using the mice,” says Jeff Rothstein, professor of neurology at Johns Hopkins and the director of the Robert Packard Center for ALS Research. “So many drugs have gone forward essentially [because of] one mouse trial.”

But faith in the SOD1 mouse has been seriously weakened. In a large-scale, multicenter clinical trial published in The Lancet late last year, the drug minocycline, which had been reported to slow the disease in SOD1 mice, was found to accelerate the disease in humans with ALS. In January 2008, another group of researchers published a study questioning the design of previous SOD1 mouse trials. The researchers reported that when they tested medications in SOD1 mice under tighter protocols, eight other drugs that had once seemed to work in the mice no longer showed significant effects.

“You’ll still see papers published, even in high-profile journals, describing the ‘largest effect yet’ in the mouse,” says Rothstein. “But it’s now clear that there’s no correlation between having an effect in a mouse and any result in humans.”

Cause of results less than clear

The analysis of previous mouse studies, published in the journal Amyotrophic Lateral Sclerosis, was performed by the ALS Therapy Development Institute (ALS-TDI), in Cambridge, Mass., and at the University of California at San Francisco. Pointing to confounding factors that had not been adequately controlled in previous drug trials in SOD1 mice, the researchers concluded that “the majority of published effects are most likely measurements of [confounder-related] noise ... as opposed to actual drug effect.”

One of the strongest confounding factors, according to the researchers, was a genetic variation among the mice. Designed to express 23 copies of a deleterious SOD1 enzyme mutant, the mice sometimes expressed markedly fewer copies, and apparently as a result lived significantly longer on average.

Other factors included the gender of the mice, since females tend to live longer, and their litter membership, since littermates tend to live to the same age.

Sean Scott, president of ALS-TDI, notes too that colony-wide infections may have played a significant role.

Typically such studies are run in large, university-based animal colonies, he says. “And in these colonies the animals are just rife with viruses and other pathogens. And the protocols there really cause the vets to treat the animals that are in the study as opposed to ‘censoring’ them, the way we would. So if an animal develops an eye infection, for example, it stays in the study. Our concern is that drugs like minocycline are treating some infection that’s endemic to the colony, and all you’re seeing is relief from the colony infection. And I believe that kind of thing is more common than everybody thinks.”

Paul Gordon, the Columbia University clinician who was lead author on the Lancet minocycline study, also comments that ALS drug trials have been done “a little bit differently in the animals than in humans.” Gordon cites two differences—frequently larger drug doses for ALS mice and earlier treatment in the disease course—as possible additional explanations for promising mouse trials that have almost entirely failed to translate into human results.

“There are many aspects of a human trial that actually should be matched in a mouse trial,” says Rothstein.

Scott and his colleagues have concluded, however, that this is easier said than done. “I think the answer everybody would love to hear is that if you adhere to these guidelines we put out, then everything will be all right,” Scott says. “In reality it’s a complicated model to run and it almost needs to be run in an industrial system, where the people handling it are doing it over and over again. All too often in the academic setting, the folks that are handling the animals are grad students who have never seen them before. I think the unfortunate reality is that preclinical drug development can’t be a casual or one-off endeavor.”

Mouse ALS versus human ALS

One of the results of the ALS-TDI study concerned the drug riluzole, which has been approved by the FDA for its demonstrable but weak effect in extending the lifespan of people with ALS. Riluzole had been reported to extend the lives of SOD1 mice too, but when Scott and his colleagues re-tested SOD1 mice under more rigorous protocols, they found the drug had no significant effect.

The later study might simply have included too few mice to enable the statistical detection of riluzole’s weak effect. But an alternate possibility is that riluzole really has no effect in SOD1 mice, though it works to some degree in humans. If that were so, it would eliminate the one significant drug for which mouse trial and human trial results had seemed to agree, and would in turn beg the question: Do the SOD1 mice have any relevance at all to human ALS?

The mutant genes expressed by SOD1 mice are found in only a few people who have ALS; the vast majority of cases are termed “sporadic,” of so-far unknown origin. Because the nervous system damage seen in these mice is so similar to that found in humans with ALS, it has been assumed that the disease process is largely the same and is merely initiated in a different, more artificial way in the mice. Gordon emphasizes that the jury is still out, but he admits the possibility that SOD1 mice are “just not a good model for human ALS.”

Scott says that he and his colleagues worry especially about the artificially large number of copies of the SOD1 gene designed into the genomes of the mice. Even the tiny minority of people with ALS who do have the SOD1 mutation generally have only one copy of the gene in their cells. “It’s something that we ponder in our lab. Is that thing driven so hard by 23 copies of the gene that if somebody were to come back from the future and hand us the cure, it wouldn’t affect the mouse? We have no idea.”

There may be further complications to the use of SOD1 mice. In a study published online March 16 in Nature Neuroscience, researchers at the University of Rochester and UC-San Diego described changes to the linings of blood vessels in SOD1 mice that caused the normal barrier between bloodstream and central nervous system to leak. The resulting “microhemorrhages” into the brain, according to the researchers, may have triggered the ALS-like brain damage later seen in the mice as they aged.

“Yet you don’t see any of that in human ALS,” notes Rothstein. “It’s irrelevant to man.”

Biomarkers

Rothstein points out that several other transgenic mouse models have been developed for ALS, and although none has been through extensive drug testing, some or all of them may ultimately prove more useful than SOD1 mice. The one Rothstein regards as the best, however, develops ALS-like symptoms only after it has lived for about a year. “That’s a long preclinical study,” Rothstein says. “None of these mice are as easy to work with as the [SOD1] mouse.”

“At the moment there literally is no other drug model that’s appropriate for pharmaceutical screening,” says Scott.

However, to the extent that SOD1 mice continue to be used to screen new ALS drugs, they are likely to be used differently, with less emphasis on broad, “noisy” targets such as prolonging the lifespan and more emphasis on improving the measures of narrowly-specified “biomarkers” such as proteins in the blood or spinal fluid that are known to rise or fall with the course of the disease.

The problem here, as Gordon puts it, is that “we don’t have biomarkers for ALS. You know, in disorders like HIV it’s possible to measure viral load, and in leukemias you can measure white blood cell count and those types of things. We have nothing like that in ALS and so we’re left with these clinical measures [such as lifespan extension] for which it takes months and months to measure change.” Researchers are urgently seeking such biomarkers of ALS.

“So we go back to the human and ask, what are the defects in the disease, what are the pathways we’re trying to alter?” says Rothstein. “You should have a body of science, from looking at patients, that says, this is a good pathway, and oh by the way my drug works on that pathway in the mouse.”

Scott says his lab has undertaken a large study along such lines, “to answer what’s similar and what’s different at the molecular level between human ALS and mouse ALS, so we’re only chasing things we know are altered in both.” By the end of this year, he hopes his group will have enough data to ensure that “we’ll never take a shot at a target that we don’t already know is altered [by ALS] in both humans and mice.”

Is there a problem with neurodegenerative mouse models generally?

Mouse models have been developed for other neurodegenerative diseases, including Alzheimer’s, Parkinson’s and Huntington’s diseases. The only mice to have been tested as much as the SOD1 mouse are the “APP” mice, which overexpress amyloid precursor protein and develop Alzheimer’s-like brain damage. Like the SOD1 mice, APP mice are also beginning to be viewed with skepticism.

Samuel Gandy, a cell biologist and neurologist at Mount Sinai Medical Center who also serves as chairman of the National Medical and Scientific Advisory Council of the Alzheimer’s Association, notes that several classes of drugs that had seemed to work in the APP mice recently failed in human clinical trials.

The fact that something similar has happened with SOD1 mice in ALS suggests to Gandy that there may be something fundamentally wrong with these models or the way they have been used.

“I don’t think that the neurodegenerative disease mouse-model people outside ALS are yet as chastened as they might be, by the ALS mouse experience,” he says.

Scott, of the ALS Therapy Development Institute, doesn’t have experience with models outside ALS but says that, given his group’s findings about confounding factors in ALS mouse studies, “I would be very surprised if any of the Alzheimer’s, Huntington’s disease or Parkinson’s disease models escaped this type of problem.”

For his part, Rothstein hopes that more failures of neurodegenerative disease mouse models aren’t on the way. But, he says, “I certainly worry about that.”