The idea that gene variants alone control an organism's traits is overly simple. A study of the effects of gene interactions on the outcomes of random variation in gene expression reveals the complex reality. See Letter p.250
Ever since Gregor Mendel and Thomas Morgan crystallized the idea of the gene as the heritable unit, biologists have been trying to map the relationship between genotype (the complement of gene variants present in an individual) and phenotype (an individual's physical characteristics). On page 250 of this issue, Lehner and colleagues1 provide insight into the complexity of this relationship by showing how random variation in the expression of members of pairs of interacting genes can act synergistically to affect the development of a complex organism — the roundworm Caenorhabditis elegans.
Notable advances in DNA sequencing during the past few years have made it possible to define the genotype of any individual rapidly and cheaply. As a result, scientists glibly talk about the possibility of pinpointing individual genes that influence every aspect of our being, from our propensity to get cancer to our chances of living to be 100 years old. However, this appealing vision oversimplifies a much more complex reality. There is no one-to-one relationship between genotype and phenotype — even identical twins can have radically different personalities, disease susceptibilities and life trajectories.
Although many factors contribute to the mapping of genotype to phenotype, the influence of two of these is especially challenging to account for. The first is 'noise' — the molecular fluctuations inherent in biological systems that cause random switching of genes on and off. The second is that genes do not act alone, but work together to form functional cellular networks. That results in complex genetic interactions in which the phenotypic impact of a particular variant (or allele) of a gene is dependent on an individual's full genetic make-up. For example, it is not uncommon for individual alleles to be beneficial on their own, but lethal when two copies are present.
Alone, either of these factors could generate an elaborate relationship between genotype and phenotype, but there is inevitably a complex interplay between the two. Lehner and colleagues' study1 examines this interplay. Previous work2 has investigated how noise in gene expression influences the effects of a defective allele of the C. elegans skn-1 gene. Notably, the skn-1 mutant allele leads to defects in intestinal development in some, but not all, individuals; such incomplete development of a phenotypic trait is known as partial penetrance.
Using a clever approach (known as fluorescence in situ hybridization) that enables the exact number of RNA molecules in a cell to be counted, this study2 established that random fluctuations in certain factors lead to a variable 'all-or-none' expression of the transcription factor gene elt-2, which is dependent on skn-1. In wild-type worms, the system is robustly buffered against such fluctuations. But in the skn-1 mutants, the variability is unleashed, so that elt-2 is not activated in some cells, hence preventing normal development of the gut. These results illustrate how quantitative fluctuations in upstream components of a signal-transduction system can lead to qualitative differences in a downstream developmental event.
Lehner and colleagues1 further explore the impact of stochastic fluctuations on the partial penetrance of a mutation, while also examining their interplay with the gene network in which the mutant allele is involved. In particular, the authors address the question of how fluctuations in factors expressed early in development affect subsequent developmental switches. This necessitated the development of a new experimental strategy, because in the previous work2 the worms were killed to measure messenger RNA levels, thereby precluding the observation of later developmental events.
The authors1 used transcriptional reporters — in which a second copy of the gene of interest drives the expression of a fluorescent protein — to follow variations in gene expression in live worms. A key requirement of this approach (verified by the authors) is that fluctuations in expression of the reporter accurately reflect fluctuations in expression of the endogenous genes. This means that the reporter is tracking expression noise resulting from differences in cellular environment — extrinsic noise3 — rather than the random turning on and off of individual promoters.
Lehner and co-workers first investigated why mutations in the T-box transcription factor gene tbx-9 that completely abolish the gene's function cause an incompletely penetrant defect in C. elegans larval development: roughly half of the animals lacking tbx-9 develop normally; the other half have muscle and epidermal defects. The authors observed that overexpression of tbx-8 — a gene closely related to tbx-9 — eliminates the defects caused by loss of tbx-9. What's more, loss of tbx-9 in their mutants caused upregulation of tbx-8.
More strikingly, the authors observed that differences in tbx-8 expression were a strong predictor of the phenotypic effects of tbx-9 loss — tbx-9-mutant worms that expressed tbx-8 at a low level were considerably more likely to develop abnormally than were genotypically identical individuals that expressed tbx-8 more strongly. Thus, stochastic differences in the feedback loop that compensates for tbx-9 loss by upregulating tbx-8 contribute to the variable penetrance of the tbx-9 mutation. Lehner and colleagues observed a similar set of effects with the transcription factors flh-1 and flh-2, which, like tbx-8 and tbx-9, are a pair of related genes that resulted from ancient gene duplication. This type of noisy feedback between related genes may therefore be quite common.
Although these findings represent a step forwards, the variability of tbx-8 expression could not fully account for the variable penetrance of the tbx-9 mutation. What, then, is the source of the remaining variability? Molecular chaperones — proteins that assist other proteins in folding — can buffer a wide range of cellular defects, especially the Hsp90 chaperone4 (also known as DAF-21 in C. elegans). Lehner and co-workers therefore investigated random fluctuations in chaperone levels in individual worms as a possible explanation of the missing variability. They found that fluctuations in expression of daf-21 were a strong predictor of the effects of tbx-9 loss. Importantly, the effects of the differences in tbx-8 and daf-21 expression were independent of each other but synergistic: more than 90% of worms with high levels of daf-21 and tbx-8 expression developed normally, whereas roughly two-thirds of the animals in which expression of both genes was low did not.
Lehner and colleagues found that the observed variability in penetrance of tbx-9 loss in their study can, to a remarkable degree, be accounted for by variations in expression of tbx-8 and daf-21. This, in turn, raises the question of what underlying mechanisms in the cell cause the observed variability and how many other genes are affected by these fluctuations. In the case of daf-21, the authors show that the fluctuations are probably part of coordinated changes in the expression of a broad range of chaperone genes, including one known as hsp-4. Such 'noise regulons' — synchronized stochastic changes in functionally related genes — are particularly suitable for causing coordinated effects on cell physiology, because the related genes act together to carry out common functions5 (Fig. 1). However, the authors did not find evidence that fluctuations in tbx-8 expression correlate with variations in the expression of chaperones, and it may be that there are no other genes whose noise co-varies with that of tbx-8. A more tantalizing possibility is that variations in the expression of tbx-8 also reflect a coherent cellular state in which other members of the tbx-8/tbx-9 pathway fluctuate in unison. This raises the further question of how many such noise regulons exist, and what controls their activity.
More broadly, it is clear from recent studies1,2 that stochastic fluctuations can have a big impact on the penetrance of gene alleles. A better understanding of the structure of noise — which genes tend to fluctuate together, and how these fluctuations are controlled — should provide crucial insight into the nature of the genotype–phenotype relationship.