Lessons learned from reading human genomes

Edison Liu, The Jackson Laboratory



Human genome sequencing has uncovered thousands of possible effectors of disease. Many of them are around regulatory regions, which makes it difficult to test in mice. There are causal variations- you are talking about combinatorics in a grand scale. Expand that to the entire genome of different strands and species, you have a large problem of multiple variations impacting phenotype outcomes. Most human diseases need to be modeled in a whole organism, cells are not enough. You can't model Alzheimers in a cell culture, need animal models.


Modeling genomic complexity is very hard. We are excited about the applications of CRISPR. At Jackson Labs, we believe that the mouse genetics is the over half of human biology. There's inappropriate extrapolation of mouse models to human disease; one has to know the limitations of model systems. The modeling has often been done on the wrong mutation, like by knock-out rather than using the humanized human variation. Insufficient genetic diversity in some of these mouse models, due to cost issues, really. If one was to transplant this concept of precise model systems, we have to have diversity of both the models as well as the mutations.


We generate new model systems and new transgenic mice at Jackson Labs. One of them is using diversity strains where we use 8 founder mice and using a controlled mating procedures, we develop in one case collaborative crosses where their jumbled DNA becomes inbred species with specific rearrangements. The entire diversity can be about 45 million SNPs, and 2M+ indels. The differences can be greater than between humans and neanderthals. The mouse model 2.0 is precise mouse models, which follows this paradigm.

We identify all mutations in patients, we construct these mutations in mice... in a stable genetic background of a pure inbred strain, or in highly variable genetic backgrounds, to then test the effects of the genetic background which we can experimentally extract the complements after. In these, we do extensive phenotyping. We just completed the construction of a $21M phenotyping facility.

Extracting cogenitor cells in the brain or heart can actually help with drug screening in cells, so we go back to the models and back to clinical trials. In genetic backgrounds, we can optimize the genetic models because of these modifiers that clearly change the phentoype and improve the stable genetic models we have and use these for drug screening.

You have seen many modifiers having dramatic impact, but this is one that was done by Greg Cox, for immunoglobulin u-binding protein 2 mutation. SMARD. Spinal muscular atrophy with respiratory disress. This gives motor neuron atrophy; in a specific mouse cross, using this, it completely reversed the phenotype. At JAX, we have several of these approaches locked in at the Center for Precision genetics (Burgess) where we look for patient-specific models tuned for the exact mutation for the disease, the same mutation, different genetic backgrounds.

At Alzheimer's disease precision models center (Howell and Carter), we have candidate mutations from human sequencing.

At Center for Genome Dynamics (Churchill), we have diversity outbred DO mice.

Cas9 stuff

Diversity of cas9 protein into mouse zygotes through a series of electroporation dramatically increases the efficiency of model creatio 2016

Casilio: a versatile CRISPR-cas9 pumilio byrid gene for regulation andrgenomic labeling

We are using electroporation of mouse embryos to deliver CRISPR-cas9. With the casilio system, it's a combination of cas9 and pumilio which is an RNA binding protein which can be tuned becaue of the combinations of repetitive sequences. By engineering that, we can produce a trinary system of the cas9, a guide RNA and then an effector that is linked to the PUFF motifs that are there, that can recognize very specific binding sites and binding sequences in the guide RNA. We can do multiplexing in a single cell of many different effectors, we can have multimerization to expand, and we can develop complexx formation with multiple protein factors.

In this manner, we have been able to create in a few months multiple effectors to activate, repress and change the epigenetic code as well as to actually label cells in multiple loci. As a nexampe, this is Oct4 and Sox2 simultaneously in a cell using two guide RNA and the recognition sites- you can see that activation of Oct4 can only occur when the Oct4 guide RNA is there, at the same time, we can actually downregulate it with a repressor in the same cell using a Sox2 guide RNA locked on to a repressor, and you can put those two together and it's highly dependent on the ... compound.

With these kinds of tools, we want to develop highly complex mouse models of the human condition, even converting many of the loci to human-like features.