It's with thanks to noted chemist-blogger Derek Lowe that I discovered a recent paper that (once again) has me questioning one of the cornerstones of 21st century biology. I am talking about our over-reliance on the inbred mouse as a proxy for all things human health and disease. The paper, "Passenger Mutations Confound Interpretation of All Genetically Modified Congenic Mice," was published last month in Immunity.
The authors combed through a database of mouse genetic variation, looking to see whether mice that had genes inactivated (knockouts) or inserted (transgenic) really were identical to so-called wild type mice. What they found was that the process which we use to genetically modify mice doesn't just affect the genes we add or subtract. Instead, genes near the one you’re targeting may have naturally occurring variations or get damaged during the process. These variations will stay with your targeted gene as you breed it into other mouse strains in order to do experiments—the titular 'passenger mutations.'
That's something that researchers rarely account for, and those passenger mutations can confound our studies. So, it may be that a mouse missing gene A might be widely used as a model for disease X. But all the while, an unnoticed mutation in gene B is actually responsible for the phenotype or drug effect being investigated.
The finding comes at a time when science is having a little crisis of confidence. Exhibit A: a high profile attempt to replicate 100 psychology experiments was able to do so for fewer than half. In and of itself this isn't necessarily a bad thing, and there are lots of reasons why so many of those studies wouldn't duplicate. But it isn't the first time we've seen something like this—other attempts to reproduce published biomedical studies have also failed. Amgen, a biotech company, tried to replicate 53 'landmark' cancer studies and failed all but six times.
Since the point of biomedical research is to discover treatments (or cures) for human diseases, you can understand why it's important that the scientific foundations we're building upon to find those treatments are solid. Most of the potential drug compounds the pharmaceutical industry investigates fail during the development process. Even back in the mid-90s (when I was a pharmacology undergraduate), we were taught that maybe only one in ten thousand 'new chemical entities' would make it through. For the past three decades, industry has tried to improve those odds, latching onto a series of trends—combinatorial chemistry, high-throughput screening, genomics—meant to make everything more analytic and quantitative.
Back in the olden days, it wasn't that important to know how a drug worked, just that it did. Even now, there's still disagreement as to just how acetaminophen (or paracetamol, if you prefer) actually functions. But the olden days of drug discovery (and biomedical research) came with a much higher body count. Rats, hamsters, ferrets, rabbits, cats, dogs, pigs, sheep, cows, primates, and more were sacrificed to the altar of human understanding. This was expensive and perhaps a little harrowing or soul destroying for those carrying out the work.
The move to reduce the number of animals used in research was both well intentioned and desirable. Instead of using different species to stand in for humans depending on the disease, why not use cells grown in a lab or mice genetically altered to fit the disease?
The ubiquity of the mouse model in research is understandable. Mice are small, which means they're cheap to house and cheap to feed—if you want anything smaller you have to leave the mammals behind and switch to zebrafish or fruit flies. There are lots of inbred strains, which means you can use lots of them and they will be (supposedly) genetically identical to each other, thus controlling for a whole genome full of variables.
But those models may not be good proxies for human diseases. Before I left the bench, my research focused on atherosclerosis—heart disease. The problem is that mice don't really suffer from heart disease in the wild. So a number of different genetically modified mouse models have been generated to study atherosclerosis. One involves knocking out a gene called LDLR and then feeding the mice a high fat diet; another knocks out a gene called APOE.
But even these models aren't great: you can cause the build up of plaques in the walls of their blood vessels, but even extensive plaque deposits don't cause the mice to have strokes or heart attacks—the things that actually kill humans with the disease. So finding out that a drug reduces the plaque buildup in these mouse models may have little relevance for its use in people.
This issue can affect lots of other mouse models as well, particularly when it's a disease of the brain. During my years at NIH, we would often joke that science had now found plenty of cures for cancer, as long as one was talking about mice that we had to give cancer to in the first place.
Another thing that became ever clearer during those years was how much of the genetic differences between humans involves rare variants that might be found in a few families, or maybe even a single one (we call these private mutations). A lot of that rare genetic diversity is thought to be important for how the same environmental exposures can cause disease in one person while not in another, so studying it is rather relevant.
Yet we pride ourselves on using mice that are (allegedly) all genetically identical to each other. Would a cohort of more genetically diverse mus musculus be a better idea?
To me, all of this is both highly inefficient—drug companies and research funding agencies end up spending lots of money on research that will ultimately fail. And it's a solid argument in favor of more funding for basic (as opposed to applied or translational) science. The phrase "more research is needed" has long been banished from these pages as a trite conclusion for a writer looking to wrap up their science post, but here, more research really is needed. I don't think the use of research mice is going anywhere, so oughtn't we get a much better grip on what it is that we're trying to ask of them?
Finally, if you're a researcher working with knockout or transgenic mice, the authors of that Immunity paper have created a helpful Web tool that you can use to see if you might have some passenger mutations you need to start worrying about.
reader comments
38