CONTACT: Dawn Levy, News Service (650) 725-1944; e-mail: email@example.com
A photo of Aviv Bergman and Mark Siegal is available
on the web at http://newsphotos.stanford.edu. Photo
credit: L.A. Cicero
Computational geneticists revisit a mystery in evolution
You and I are both human, with hearts that beat at roughly the same rates, nervous systems that churn out just about the same chemicals, bodies that are similar enough to peg us as people and not chimpanzees. But despite the fact that we both belong to the same species, our genes are pretty different. Only half the genes are identical among siblings who aren't twins, for example, and for most of us, the degree of genetic relatedness is much smaller.
Why, biologists first asked 60 years ago, do members of the same species have such similar traits, or phenotypes, despite the fact that they have such diverse genes, or genotypes? They couldn't fully explore that question until now -- when, aided by computers, they can sift through mountains of experimental data. In the June 24 issue of the Proceedings of the National Academy of Sciences, senior research scientist Aviv Bergman of Stanford's Center for Computational Genetics and Biological Modeling (CCGBM) and postdoctoral scholar Mark Siegal of the Department of Biological Sciences provide a surprisingly simple answer.
Invariant traits -- such as having five fingers to a hand instead of four or six -- don't become universal because Nature has somehow selected special genotypes that faithfully direct development of the trait under a wide variety of conditions, the researchers argue. Instead, they show, it is the complexity of our genotypes -- the many genes that interact in networks during development, inhibiting and activating each other and even regulating themselves -- that provides fidelity. Indeed, Bergman and Siegal show that any functional genetic network that is complex enough has this built-in property of fidelity. This is true whether natural selection on the phenotype produced by the network during development is strong, weak or absent. Natural selection may be important in shaping traits that aid in reproduction and survival, but Bergman and Siegal show that it doesn't matter much during development, when, biologically speaking, all roads lead to Rome.
"We're taking a more sophisticated view of evolution as a process," says Bergman, who co-directs the CCGBM with Marcus Feldman, the Burnet C. and Mildred Finley Wohlford Professor in the School of Humanities and Sciences. "We need to take into account not only the genetic system by which the hereditary information is passed on from one generation to the next, but also the developmental system by which the information contained in the fertilized egg is expanded into the functioning structure of the reproducing individuals." Researchers at CCGBM, established in 1997 with a grant from the Paul G. Allen Foundation, conduct interdisciplinary research into quantitative problems of biology.
"Evolutionary biologists tend to think of natural selection as the first possibility of a mechanism for explaining most things that they observe," says Siegal. "So it's natural that the first attempts to explain this disconnect between great genotypic variation and little phenotypic variation was through natural selection."
In 1942, even before it was known that genes were made of DNA, British biologist Conrad H. Waddington coined the term canalization to describe the "funneling" that occurs during development to produce just a few end products, or traits, such as the beautifully patterned wings of a butterfly. He envisioned development proceeding the way a ball rolls down a mountain, traveling mainly along well-worn grooves and having the option of rolling one way or the other at only a few forks in the road. The ball rolls to the right, and the result is, say, development of an elaborately patterned forewing. It rolls left, and a different-looking hindwing forms. The puzzle that attracted Bergman and Siegal was not so much the nature of the genetic switches that operate at the "forks" but instead what causes the "grooves" that keep development faithfully rolling along when both environmental disturbance and genetic mutation could potentially set it off course.
"You can throw a lot of insults at an organism -- either genetic ones by mutation or environmental ones by changing the temperature or changing the chemical composition of the food -- and in spite of all of those insults, development is pretty robust," explains Siegal, who also conducts evolutionary research on fruit flies in the laboratory of Bruce Baker, the Dr. Morris Herzstein Professor in Biology. From larvae incubated in a lab at 18 or 28 degrees Celsius, for instance, similar-looking flies will develop, even though chemical reactions are twice as fast at hotter temperatures than colder ones. The developmental pathway and end products (traits) seem immune to such insults.
Indeed, some biologists argue that canalization may have evolved as a response to environmental change. Under this scenario, Bergman explains, "When mechanisms evolved to dampen the effect of environmental variation on the phenotype, as a side effect they also happened to buffer genetic variation." But the results of Bergman and Siegal suggest that environmental perturbation is not necessary for canalization to evolve. "We don't know all the details of what makes that funneling process work," Siegal admits. "But our contribution to it is giving one possible reason that hasn't in our view been considered enough."
Scientists used to think that developmental fidelity evolved via natural selection, principally through survival and reproduction of organisms with redundant genetic systems -- that is, ones with copies of important gene sequences. But Siegal and Bergman's results indicate that redundancy may only be one small manifestation of a bigger theme: the complexity of gene networks. In short, more complex systems are more resistant to change in their outputs.
"It is typically assumed that important properties of organisms are crafted by natural selection," says Dmitri Petrov, assistant professor of biological sciences. "What Siegal and Bergman show is that robustness in the face of mutation, or canalization, may be a byproduct of complexity itself and therefore that robustness may be only very indirectly a product of natural selection."
Says Siegal: "It might be that the complex nature of the genetic system itself is going to give you canalization independent of natural selection. This complexity goes beyond mere redundancy, incorporating all kinds of elaborate connections in the gene network."
That doesn't mean natural selection doesn't play an important role. Continues Petrov: "Natural selection has shaped the genetic networks of complex organisms so that they produce appropriate phenotypes -- the more highly interconnected these networks are, the more robust the corresponding phenotypes are. The importance of this result is that it shifts the focus of the field away from abstract models of natural selection and toward actual genetic networks. In so doing, it will provide a new perspective for analyzing and understanding the current outpouring of genetic data in model organisms."
A new perspective could prove useful -- because invoking natural selection to explain the disparity between genotypic and phenotypic variation has several problems. First, a prerequisite for canalization is genetic variation -- but if selection for a trait is too strong, it shrinks the gene pool. "Once that limits the genetic variation, it removes the pressure to have canalization," Bergman says.
Second, modeling has shown that if nature "selects" a trait, canalization evolves -- but very, very slowly, over millions and millions of generations. "When you start thinking about time scales like that," Siegal says, "you have to wonder whether any evolutionary force can be consistent over that amount of time to actually cause the outcome that you see."
And third, what's "optimal" today may not be optimal tomorrow. Says Bergman: "As [scientist Stephen Jay] Gould said, as the environment changes what was once fit may not be fit today, and with further change in the environment could become fit again."
For their project, Siegal and Bergman chose to model an abstract system that is important in the development of most organisms -- transcription factors, or proteins that regulate the expression of genes. In the model they developed, 10 genes each encode a protein that in principle is capable of regulating the expression of each of the other nine genes, as well as itself. To compare the complexity of the abstract system with that of an actual system, consider that yeast, for example, has about 6,000 genes, around 500 of which regulate each other.
Bergman and Siegal's collaboration comes at a time when -- thanks to the use of microarray technology in a new field known as functional genomics -- scientists have greater knowledge about sophisticated gene interactions during development. This technology helps scientists analyze the functions of genes in an organism's genome all the genes that make up its genetic blueprint and allows them to look closer than ever before at the intricacies of heredity. So, although the song remains the same as that sung by previous giants of biology, such as Darwin and Gould, Bergman and Siegal are studying the individual musical notes to better understand how evolution's song plays out.
"The evolution of genetic robustness is a whole new game now that we have the results from Drs. Bergman and Siegal," says Gunter Wagner, professor of ecology and evolutionary biology at Yale University. "[It] shows that selection against lethal mutations [those that make the network incapable of producing any phenotype] can lead to the evolution of mutational robustness of any character state, even in the absence of stabilizing selection for that character state itself."
Says Siegal: "In many ways canalization was sort of a smokescreen that was dividing evolutionary biologists and developmental biologists. The developmental biologists were studying their genetic networks and the evolutionary biologists were in the abstract saying, 'Well, these networks must have evolved to produce certain properties, like robustness in the face of mutational insult.' But since we have shown in our model that it's actually the nature of the developmental system that can give you this property, they're really not two separate things to study. They're the same thing to study. I think a lot will come out of looking at actual genetic networks and how the structure of those networks gives them the property of being robust."
By Dawn Levy