If you look back at discoveries made in biology, many of them were thanks to a model organism. Gregor Mendel had his pea plants, Thomas Hunt Morgan had his flies, and Sydney Brenner has C. elegans. Indeed, the Notch, Wnt, and Hedgehog pathways were found in fruit flies, and the yeast genome was among the first to be sequenced.
While advances in sequencing make it possible for researchers to work on nearly any organism they choose, model organisms do offer some advantages. "I think people realize there are just some things you can't do on a large scale, or can't do effectively, and so you use model organisms," says Steven Henikoff at the Fred Hutchinson Cancer Research Center. "Then, of course, if you look back at discoveries, the major discoveries in science, in biology, you are going to find that somebody started out on some model organism."
Models are easier to work with than humans are. Models can be manipulated both environmentally and genetically in ways that humans just can't. Models tend to have smaller genomes, making genetic work easier to do. Further, models can be used to establish causality between a gene variant and a disease or phenotype.
"In model organisms, one can actually do experiments," says Trudy Mackay from North Carolina State University. "Having found an association … you can then follow it up by doing directed crosses. You can knock it out, you can make a mutation of it, you can actually in yeast — many model organism people have yeast envy — if you have an allele that has a particular single base pair change, you can actually put two alternative alleles in a completely common background and that's the gold standard of verification that that allele actually has an effect."
Fruit flies have long been used to study genetics, and C. elegans has been a mainstay of development research. These and other model organisms are being put to work in genomics, bringing their rich histories along for the ride.
A number of databases and tools exist for model organisms. There are FlyBase, WormBase, and The Arabidopsis Information Resource, among others, which keep organized all the data that's been collected on these organisms over the years, even adding newly acquired genomic information. Model organisms like flies, C. elegans, and Arabidopsis also have well-annotated genomes. "FlyBase does a really good job keeping the annotations of the sequences really up to date and rich because you can learn about what this gene does and everything pretty clearly by going to FlyBase," Henikoff says.
The ModEncode Projects to determine functional elements in C. elegans and Drosophila has also generated a lot of information. "I think the ModEncode Projects have been enormously successful. We've gotten so much data and we've been able to integrate it into the databases that house all of the rich information that has been accumulated about these organisms over the years," says Jason Lieb from the University of North Carolina, Chapel Hill.
[ pagebreak ]
While model organism researchers are benefitted by these resources and tools, they are also developing new and better ones. NC State's Mackay, along with Richard Gibbs and Stephen Richards at Baylor College of Medicine, developed the Drosophila Genetic Reference Panel, a population of 192 inbred fruit fly lines that vary for quantitative phenotypes, and that the researchers are sequencing.
Mackay likens it to a 1,000 Genomes Project for the fly, "the idea being that fly stocks are a living library of common genetic variation and quite representative of rare variants." She adds that "this resource with the natural variants is totally complementary to the other resources we have, which are mutations and RNAi constructs that disrupt the genes in the genome. We'll be able to go back and forth between natural variants and induced mutations."
Mackay's lab recently collaborated with Liesbeth Zwarts and Patrick Callaerts from the Flemish Institute for Biotechnology to study the genetic architecture of aggression in Drosophila using the mutation resources available for the fruit fly. Previously, Mackay's group had screened transposon-tagged mutations for ones linked to aggression and found a number of them. For this study, they homed in on six particular mutations and looked at how homozygous mutations affected brain morphology and how those mutations interacted with each other.
"What we found was quite interesting, that there was a lot of the enhancing and suppressing effects, which we call epistasis, and that the enhancing and suppressing effects occurred at all levels, aggressive behavior of the animals to brain morphology to gene expression," Mackay says. "With only just six mutations, there's a huge impact on the genome and we have complicated interactions among pairs of mutations. ... We think that if we can generate this with only six mutations, it means that the traits that are segregating many, many loci are going to be much more complicated than previously appreciated." This work was published in PNAS in October.
It's not just Drosophila that has newer resources. On the mouse side, there are two new populations for researchers to work with, the Collaborative Cross and the Diversity Outcross mice, developed by Jackson Laboratory's Gary Churchill and others. Both of these populations have the same 45 million SNPs and 400 million copy-number variants segregating within their population, though those populations have different underlying structures. The Diversity Outbred mice are meant to be used for high-resolution mapping while the Collaborative Cross mice are for validation and mechanistic studies.
"If you have inbred strains — which we do — you can replicate, you can put them in different environments, and you can do all the mechanistic studies, invasive procedures and so forth that you need," Churchill says. He adds that his team has a number of papers coming out soon that show that they only need a few hundred mice, to get single-gene mapping resolution in the Diversity Outbred population.
[ pagebreak ]
To the cause
The need for such resources is only increasing. The myriad sequencing and genome-wide association studies are linking regions to diseases or phenotypes but the cause or mechanism often remains to be found. Genome-wide association studies won't be useful unless there is a way to follow up on and verify those associations. "GWAS studies are not going to be very useful unless you have a mechanism to follow up, and model organisms provide that," Churchill says. Models, he adds, will give researchers clues about which GWAS hits to follow up on.
"You ultimately have to test the function of those associations to nail down what the mechanism is," UNC's Lieb adds. "That is what models always have been for. In a sense, models in genomics will do for genomics what they have always done, which is to provide a system in which to do controlled science."
Further, direct discoveries in model organisms continue to inform human studies. "The ability to make discoveries directly in model organisms didn't go away," Churchill says.
At the Hutch, Henikoff focuses on epigenomics, and studies DNA methylation and chromatin. "The components of chromatin — not just the nucleosomes and all, but also the chromatin remodelers, the modifying enzymes, the variants — they are all almost universal for eukaryotes. What we learn can apply to humans just as easily as to other model organisms," he says.
Many model organism studies do eventually feed into or inform human studies. "Model systems are going to provide you with clues as to which are more important," Jackson's Churchill says.
Henikoff and his team found that, in Arabidopsis, there is an antagonism between the histone variant H2A.Z and DNA methylation. Using Arabidopsis mutants, Henikoff says his team could "go well beyond correlation" to show that H2A.Z and DNA methylation are mutually antagonistic. "One is keeping the other out," he says. Then they moved into a mouse model of B-cell lymphoma where they found the same antagonism, though they could not prove causality in that system.
"I think that is a good example of how you can use model organisms where you can do genetic manipulation — and also it is a smaller genome so you could do it more thoroughly — and then we could hop to another model, one which is directly relevant to cancer and you can't really do that thing with humans," Henikoff says. "So, that has lots of advantages there. I like all model organisms, and each can have its place and we'll hop from one to another depending on what the question is."
[ pagebreak ]
Similarly, UNC's Lieb has a pipeline of sorts now set up in his lab. He developed a technique called FAIRE-seq that maps open chromatin, a methodology he developed in yeast. Jorge Ferrer from the Hospital Clínic de Barcelona did a mini-sabbatical in the Lieb lab and wanted to use FAIRE-seq to map open chromatin in pancreatic islet cells. The researchers then overlaid the information they uncovered about regulatory regions with SNPs linked to type 2 diabetes risk to narrow down the search for functional SNPs.
"We showed in a proof-of-principle way that you can do that using this method," Lieb says. "That's a great example of how you can move from basic discoveries in yeast, including the idea that nucleosomes are evicted from regulatory sites, and then a method developed to find those nucleosome eviction events, and then moving that to a mammalian system with relevance to type 2 diabetes. That's the model for what we'd like to continue doing." He and his lab are now applying this approach to other diseases, including breast cancer and progeria.
Model organisms are, however, just that — models. And research done in model organisms sometimes comes with a caveat.
Jackson Lab's Churchill notes that many cancer researchers bemoan that mouse models of cancer are not just like human cancers. But human cancers aren't good models of cancer either, he argues. "What they mean is that this particular knock-out mouse in a fixed background doesn't get a tumor that's just like every cancer patient on the face of the Earth. But the flip side to the argument is that no human is a good model for human cancer either because every human cancer is different and there are of course broad classes and similarities, but to expect to emulate a human condition in all of its diversity in a single, fixed genetic background is absurd." He adds that's why he wants to study diseases in a broad genetic background, like the two mice populations he helped develop.
Model systems also have different evolutionary and population histories from the organism they are being related to. Blake Meyers from the University of Delaware studies microRNA evolution and function. He and his team found a class of miRNAs in Medicago and soybean that they then also studied in potato and tomato. This class of miRNA, however, is not conserved in Arabidopsis. "The fun thing about that was that it showed that Arabidopsis, while it's a good model, good models aren't necessarily the whole story," Meyers says.