A challenge in biology is to understand complex traits, which are influenced by many genetic variants. Studies in yeast provide the prospect of analysing such genetic variation in detail in other organisms, including humans.
Ever since the modern understanding of evolution and genetics in terms of natural selection and Mendelian inheritance was formulated, generations of scientists have struggled to explain the genetic bases and evolutionary significance of the remarkable variation among individuals, which is observed in many species. Despite considerable progress in finding the main genes that determine genetically simple traits, genetic variants of individually small effect that influence the so-called complex traits — which include height, weight and disorders such as neuropsychiatric diseases and cancers — have proved elusive. On page 1039 of this issue, Ehrenreich et al.1 report a method in yeast that offers great statistical power for identifying multiple genomic regions that contribute to complex traits. Their work affords significant hope that similar genomic studies will be possible in many other species.
Over the past 20 years, various analytical and empirical approaches have been developed to find gene variants that influence complex traits2,3. One experimental technique, called bulk segregant analysis4, examines progeny from crossing two different yeast strains and can potentially pinpoint multiple genes contributing to trait differences, especially when coupled with high-throughput analysis of the progeny's genotype5. More recently, these techniques have been merged with ways to select for progeny with extreme traits6, thus allowing greater mapping precision and power. Ehrenreich and colleagues' paper now hints that we are finally poised for what may be a step change in our understanding of the genetic basis of organismal diversity.
The authors1 report genetic variations in yeast that mediate 17 complex traits related to resisting chemicals. The main innovation here is to couple the generation of very large populations of progeny from an inter-strain cross with the selection of extremes of a trait from the population. Ehrenreich et al. then compare the frequencies of marker alleles (gene variants that are not thought to influence the trait but that flag up particular genomic regions) in the selected and unselected populations and test for significant deviation. When allele frequencies differ significantly between the two populations, the authors infer the presence in that region of a genetic variant influencing the trait. Because this method uses a large base population, and compares the extremes of a trait distribution, it can detect variants that make only modest contributions to the variation of the trait, which in principle allows comprehensive genetic dissection. As a key proof of concept, the authors evaluate the contribution of the variants they identified for one trait: resistance to the DNA-damaging agent 4-nitroquinoline. They find that differences in the genomic regions pinpointed by these variants can explain 54% of the variation among all the progeny of the same cross — an unusually high proportion in comparison with most traits studied in most species so far.
The genetic architectures of the 17 chemosensitivity traits studied by Ehrenreich et al. show remarkable diversity — ranging from ones influenced mostly by a single gene to others affected by up to 20 genes in different genomic regions. Thus, even in a single species, and for apparently similar traits, the underlying genetic architectures can be remarkably variable. Early indications from other studies suggest that such diversity of genetic architecture is to be broadly expected across many species. In humans, for example, comprehensive study of common gene variants has explained virtually none of the risk associated with most neuropsychiatric disorders7, but the yield for many autoimmune diseases has been much better: in Crohn's disease, for instance, up to 20% of the genetic component of risk has been explained2. One of the more interesting areas of study for the future will be the underlying evolutionary and developmental reasons behind such broad variation in the genetic architectures of different types of complex trait.
It is worth emphasizing that many of the key elements critical to the authors' success with yeast1 are applicable to other organisms (including humans), for example through selection of the extremes from a large base population and rapid assessment of gene-variant frequencies. For humans, selection of extremes happens routinely as part of clinical care. Individuals who are formally diagnosed with epilepsy or schizophrenia, for instance, represent just 1% or so of the general population. Other traits of immediate medical interest represent even greater extremes; these include drug-induced liver injuries and heart arrhythmias, two frequent reasons that potential therapeutic agents are halted in mid-development and prescription drugs are pulled from the market. Such extremes identified as 'cases' in the clinic can be compared with the general population or, in other settings, the extremes of a given distribution (for example, weight, height, birth weight and blood pressure) can be compared to identify the underlying genetic variations influencing the trait. Next-generation sequencing now makes comprehensive identification of virtually all gene variants present in large samples logistically possible.
For model organisms other than yeast, mapping with large sample sizes and selection of extreme traits have already been applied to the nematode Caenorhabditis, the fruitfly Drosophila and the plant Arabidopsis to analyse traits such as nutrient uptake, starvation resistance and heavy-metal tolerance. The advantage of yeast, however, is its very high meiotic recombination (crossover) rate — orders of magnitude higher than the average rates observed in these other species — which enables the effects of individual genes to be localized with greater precision. Nonetheless, Ehrenreich and colleagues' automated selection approach1 can be adapted for other model systems to localize modest-effect variants contributing to ecological and behavioural traits.
That said, we do not wish to trivialize the amount of work ahead, nor the complications inherent in studies of organisms other than yeast. Despite startling advances in sequencing technology, it remains difficult, for instance, to identify accurately many classes of genetic variation — in particular, structural variants and any variants in 'difficult-to-sequence' genome regions. Accurate measurement of trait features also remains a challenge, especially for behavioural traits and, in humans, for most mental illnesses. Even when causal variants for a given trait are identified, we face the still more difficult task of determining the functional effects associated with each of the variants and how to piece them together.
Nevertheless, a field of study sometimes suddenly enters a golden age when the necessary conceptual and technological ingredients combine to permit discovery at a rate that would have been difficult to imagine, let alone achieve, at any previous time. The study of the genetic basis of organismal diversity seems set at last to enter its own golden age, where even genetically complex traits will yield many of their long-held secrets.
Ehrenreich, I. M. et al. Nature 464, 1039–1042 (2010).
McCarthy, M. I. et al. Nature Rev. Genet. 9, 356–369 (2008).
Kao, C. H., Zeng, Z. B. & Teasdale, R. D. Genetics 152, 1203–1216 (1999).
Michelmore, R. W., Paran, I. & Kesseli, R. V. Proc. Natl Acad. Sci. USA 88, 9828–9832 (1991).
Borevitz, J. O. et al. Genome Res. 13, 513–523 (2003).
Nuzhdin, S. V., Harshman, L. G., Zhou, M. & Harmon, K. Heredity 99, 313–321 (2007).
Need, A. C. & Goldstein, D. B. Dialogues Clin. Neurosci. (in the press).
About this article