Masters of Complexity: Edison Liu On The Future Of The Mouse Model

By Allison Proffitt 

June 8, 2017 | Believe it or not, when the Jackson Lab was born in 1929, many researchers didn’t believe it was possible to study genetics in mammals, assuming inbreeding in mammals was impossible. A lot has changed in biology and genetics since then, and there are now more mice living at the Jackson Lab than people living in the state of Maine.

Edison Liu, president and CEO of The Jackson Laboratory, is harnessing the JAX spirit to push the organization—and hopefully the field—toward a more mature understanding of complexity.

“The paradox is that with our technologies in genomics, we can know every single genetic detail, every single nucleotide and measure every phenotypic output,” Liu said during his plenary talk last month at the 2017 Bio-IT World Conference & Expo. “But the challenge now is how to measure complexity, to understand the complexity,… to create complexity so we can prove a principle, and how to master that complexity,” he said.

It’s been the focus of his energy for some time now. The field is moving from classical genetics to complex, or functional genetics, he told me during a conversation last summer. “We [JAX] are going to be the masters of complexity, perhaps one of the only institutions completely focused on complex genetics and functional genomics with the scale of operations that we have.”

The old version of mouse models—inbred lines with a single mutation meant to represent an entire human disease—is disastrously simplistic, Liu believes. “It’s akin to taking one genetically ill individual with a private mutation that causes cystic fibrosis, and then trying to generate an entire drug panel for diabetes using that one patient. It makes no sense whatsoever!”

The need, he told the packed auditorium at Bio-IT World, is for a new kind of mouse model, one capable of complexity to accurately model the complex diseases we want to tackle. For instance, we now think there are 75-90 genes involved in Type 2 diabetes. Those types of problems can’t be tackled with single-gene knockouts.

But likewise, he insisted, they aren’t only the purview of algorithms and computer modeling either. Liu didn’t argue against the power of computer modeling and simulation, just that live mouse models let us explore hypotheses in different and important ways. Both, Liu said, are important to discovery.

“We can no longer rely on in silico predictions, and we have to be able to reconstruct [genomics] in whole organisms,” Liu told me last summer. We believe that computational science is going to be the bedrock [of genomics], like molecular biology and cellular biology in biology. You have to have it! But to think you can rely on interpreting the genome to give you all answers, I think, is a fallacy. It needs the validation and biological experimentation.”

Making the Model Work

As an example, Liu highlighted the work of Greg Cox at JAX, who identified the murine gene responsible for a muscle wasting phenotype. It turned out to be the same mutation a physician had seen in his human patient with Spinal Muscular Atrophy with Respiratory Distress, SMARD. By crossing two inbred mouse lines, Cox found that he could induce a rescue mutation that ameliorated symptoms in some mice. The disease-causing mutation is on chromosome 11. The rescue mutation is on chromosome 13, and Cox and his team have narrowed down the options to three candidate genes.

This type of crossbreeding would never be possible in the human population, Liu points out. In fact, he argues that the work would never have been accomplished with sequencing alone.

How do you apply that to the human condition? Liu referenced Nadia Rosenthal’s work on heart attacks. We see clear differences in heart attack outcomes based on stress level, gender, and life style in addition to genetic variants, Liu said. Rosenthal asked: “Are there variances to the outcome of the myocardial infarction, which at the endpoint is dead tissue and heart failure, that are based on genetics?” She looked at the outcomes in different strains of mice by purposefully “outbreeding” a diverse population of mice. Using only eight founder mice and allowing them to breed at random, Rosenthal was able to model more variances than are found in the human population, exposing a wide range of variability in cardiac scarring, vascularity, and immune system.

“Those combinations gave a gradation of severity in a reproducible manner,” Liu told the Bio-IT World audience. Rosenthal mapped her findings and has a candidate gene: MYO18B.

Diversity outbreeding and other techniques have moved mouse models from version 1.0—single-gene knockouts—to version 2.0, Liu said. Clinical sequencing is revealing hundreds of thousands of candidate mutations in any one disease, he said. How do you then by statistical association alone make any claims on causality? You must move into a model system.

Liu’s team is inducing mutations in different genetic backgrounds of mouse—he mentioned using CRISPR for highly-targeted changes in the genome—phenotyping the mice in a highly rigorous manner to gain an understanding of the cellular mechanisms, then extracting cells to do drug screening. “This brings us back to the clinical sector,” Liu said, “and in this iterative manner find some truths about a cure.”

Liu said the goal may not be to attack a specific mutation, but instead to activate a modifier gene, much like Cox’s work demonstrated. “It’s an orthogonal therapeutic.”

Complex Futures

JAX is changing the way mouse research is done to facilitate complexity. We can now electroporate embryos with guide RNAs, Liu said. “It frees us from the artisanal approach of single injections to ones we can do in a 96-well format.” Taking it a step further, researchers added a series of signature motifs to the tail of the RNA that acts as a “digital code,” to enable multiplexing. We can do this to activate, repress, or delete marked genes, Liu said.

The theoretical limit to the approach is 60,000, though researchers haven’t approached that number yet. Still Liu is hopeful. “We could change the genome in subtle or dramatic ways 60,000 loci at time,” he said. “The idea here is, instead of having a mouse for every mutation, we can actually do a mouse for every abnormal network. That would raise the complexity,” Liu said, “and give us a higher resolution look at diseases.”