Uncovering the causal basis of complex diseases or traits is challenging. As suggested by their name, a myriad of sources may be potentially contributing to variation. Phenotypes result from interactions between the genetic background (which itself may be compounded by multiple additive or interacting and multiplicative alleles), control of the expression within the context of that background, and the environment in which the expression occurs.
Using humans to sort out this complexity in an experimental setting is therefore difficult, if not pretty much out of the question. The advantages of using non-human model organisms to study complex disease and traits are numerous, including the ability to define the genotypic background (via strains), to perform experiments with replicate genotypes from those strains, to control the physical and social environment of the individuals under study, to assess tissue-specific expression, and to obtain more detailed phenotypic information than is typically enabled with human studies. Additionally, one can also perform mating experiments and perform experiments in replicate. For mammal models, there are also many homologous genes and phenotypes that are representative of their counterparts in humans.
For all of the advantages of non-human model organisms, one historical limitation of mouse models in particular was the lack of diversity inherent in strains, and whether the homozygosity and lack of variation biased study results. We recently attended the 13th annual Complex Trait Community meeting held this year in Berlin, which featured a wide range of studies using the Collaborative Cross mouse lines. These lines were created through a breeding design to mix and randomize genomes from 8 laboratory and wild-derived strains into an experimental population where each recombination site was unique and ideally whose size was sufficiently powered to support quantitative trait loci (QTL) analysis at near single-gene resolution1. The Collaborative Cross lines not only provide the advantages of typical mouse models, but also provides a diverse genomic background in which to test a series of hypotheses about genes implicated in human disease. The hypotheses may be motivated by QTL discoveries in large human genome-wide association studies (GWAS), or directly when phenotypes arise via recombination in mouse strains under study.
One study in particular that nicely set the stage was presented by Grant Morahan from the University of Western Australia. He used the set of 130 Collaborative Cross lines he produced, and took advantage of traits with known genes (coat color) to evaluate as a proof-of-concept whether genes can reliably be mapped and discovered with as few as 50 mouse strains. Genotyping animals with the Mega Mouse Universal Genotyping Array (aka “megaMUGA”, a custom Illumina mouse genotyping array with up to 77,800 SNPs designed by GeneSeek), allowed for haplotype reconstruction to identify segregating regions of the genome correlated with phenotype (coat color), effectively shrinking the genomic search space to map traits. Applying a quantitative score to the phenotypes and using a multinomial approach to mapping the quantitative trait loci, he could successfully map and identify the causative variants in all three loci. Performing 2,000 simulations of 50 randomly chosen strains, he found that with 50 strains, 77% of trials resulted in at least one locus being significantly identified. Only 2% of trials failed to produce suggestive mapping results, while less than 1% of trials yielded a false positive result. These results demonstrate that researchers can be confident the CC approach could be used for mapping genes and identifying causative SNPs that underlie QTLs.
A friendly yet exciting post-presentation discussion about a diet treatment (“high fat” vs. “ingredient composition”) typified some of the challenges of controlling the environment. If designing experiments to control environment is challenging in mouse studies, what hope is there of directly teasing out and quantifying these effects in human studies, where controlling and/or monitoring and accurately reporting one’s environment are fraught with challenges—imagine daily meticulous records of all of our social interactions, and what we consume, for starters…
There are clearly many genomic and experimental approaches that we can take to ultimately answer questions and get to the bottom of complex disorders. While the physiology and biology of mice do not perfectly represent those of humans, there are many homologous traits and genes that can be uncovered perhaps more easily in the model system. In the classic allegory where several people assess one part of an elephant and each arrive at different conclusion, perhaps a smaller animal with fewer parts may allow us to assess more at once.
1 Threadgill, DW and Churchill, GA. “Ten Years of the Collaborative Cross”, Genetics, February 1, 2012 vol. 190no. 2 291-294.