In the years since they were first developed, microarrays have been applied to an extraordinary range of situations. But in 2001, some researchers began to voice concerns that the increasingly ubiquitous technology was ignoring the power of classical genetics. Microarrays could be combined, they said, with genome-wide linkage analysis to give the geneticist a new way to examine complex traits.1
The technique, which some call genetical genomics, essentially treats RNA transcript levels as quantitative traits that can be mapped to quantitative trait loci (QTLs) that influence their expression. Such studies have shown that transcript levels are heritable and can be analyzed in a high-throughput manner, generating excitement about that possibility that virtually any genetic variation can be teased apart. This is a "new stage of genetics," says Robert Williams, codirector of the Center of Genomics and Bioinformatics at the University of Tennessee. "One where we can now study tens...
Although yeast gave genetical genomics its first solid example, it didn't take long for its appeal to grow. "Many groups from all over the world are now starting up this type of experiment," says Ritsert Jansen of the Groningen Bioinformatics Center, the Netherlands. Jansen's lab is actively studying
While such integration is still far from a reality, this issue's Hot Papers were some of the first to open up this possibility. For this reason, " [they] attracted a lot of attention," says Jansen. They also began to explore the question of what type of variation contributes to complex traits in a number of systems.
Cheung and colleagues went straight to humans. They originally analyzed 35 individuals from Utah pedigrees, finding that expression phenotypes aggregated in families.3 "This put forward the idea that gene expression can also be considered a human phenotype," says Cheung. In subsequent work, they looked at 3,554 genes from 14 families and found several hotspots in the human genome that regulate the expression of numerous genes.5 Cheung's group is now hunting for these regulators, which potentially control complex networks of genes.
Schadt and colleagues at Rosetta (a wholly owned subsidiary of Merck) are investigating the same question. Their original work focused heavily on mice. They analyzed a population of 111 F2 mice and linked expression patterns to obesity.4 "We are really into [genetical genomics] at the moment," says Schadt, "the reason being that it looks like it works." The group's research has already linked four genes – including complement component 3a receptor 1 (
Additional studies continue to surface. Timothy Aitman and colleagues at Imperial College London profiled the BXH/HXB panel of rat recombinant inbred strains and found 73 candidate genes for hypertension.6 In another study, Williams and colleagues profiled the well phenotyped BXD panel of mouse recombinant inbred strains, finding, for example, a locus (near marker
DEEPER GENETIC QUESTIONS
Genetical genomics provides the opportunity to ask broader questions about gene control. In their original study, Kruglyak's lab found that the expression levels of 570 yeast genes were linked to one or more loci.2 These loci were considered to act in
© 2004 Elsevier
This example shows how expression phenotypes are mapped to a locus in yeast. (A) Expression is measured in quadruplicate for each parent and once for each of eight segregating offspring. (B) Genotype is characterized with the use of markers. Here, three are used, and green and blue represent the parental origin of the genomic segment. (C) The expression phenotypes for segregants are compared for each marker. The horizontal lines represent mean expression level, and a regulator locus can be pegged to the marker two where the difference is most pronounced. If the actual gene is near the regulator, the expression phenotype show
"People had been using microarrays for years before us, but we combined genetic linkage to examine segregating polymorphisms," says Kruglyak. What was found "added the wrinkle that just because you have a segregating expression level, it doesn't mean it's caused by a polymorphism in the gene."
In a follow-up study, Kruglyak's group found that most regulatory variation is not caused by genetic changes in transcription factors.8 This suggests that the elements regulating variation in gene expression are spread across different gene classes.
The work in Kruglyak's group was "one of the first steps to seeing if expression differences can be correlated with genetic differences," says Cheung, who was impressed with the technique right away because it allows gene regulatory regions, no matter where they are, to be uncovered. "We really don't know the transcriptional control for most of the genes in our genome," says Cheung, "so why not use a blind genetic approach and look at everything?"
POTENTIAL AND LIMITATIONS
The excitement over the potential of genetical genomics is partly coming from the similar data it's generating, which in each system is pointing to
Jansen points out that the degree of complexity in a complex trait is still unknown. "But there is enough evidence that we can make progress," says Jansen. "In fact, the genetics of an expression phenotype may seem to be a bit more informative than that of classical phenotype."
Karl Broman, assistant professor of biostatistics at Johns Hopkins University, says the studies using genetical genomics are quite interesting, but cautions that it remains to be seen what can be learned about human disease from them. "The difficult thing is to turn observed associations between genes into some causal pathway," says Broman, where one gene causes the expression in another gene.
Jansen adds that traits encoded by many genes of small effect will require large population studies, which is not an inexpensive endeavor when each offspring must undergo microarray analysis. "Industry can pay for this," says Jansen. Also, large collaborations, such as the Complex Trait Consortium
A form of "systems genetics" is where this is all headed, says Williams. This will require a great deal of computation, the techniques for which are still under development. "It won't be easy, but I don't think you can understand the function of genes in isolation, since they are inherently modular," Williams says. "This will put the function back into functional genomics."