“Is Evolution Predictable?” asks a piece in Science. Here’s the first paragraph:
If one could rewind the history of life, would the same species appear with the same sets of traits? Many biologists have argued that evolution depends on too many chance events to be repeatable. But a new study investigating evolution in three groups of microscopic worms, including the strain that survived the 2003 Columbia space shuttle crash, indicates otherwise. When raised in a lab under crowded conditions, all three underwent the same shift in their development by losing basically the same gene. The work suggests that, to some degree, evolution is predictable.
The “some degree” part is the catch. I’m a big fan of general ideas, but the more I learn about evolution the more suspicious I become of broad truths. A given dynamic often has some degree of validity, but extending it too far leads to error or confusion in innumerable specific cases. Evolution may be the most robust and powerful theory for deductive inference in biology, but even here rationalism has its limits. For example, before the rise of molecular methods in exploring polymorphism the debates as to the nature of genetic variation in natural populations tended to focus on outcomes based on adaptive pressures. One school followed R. A. Fisher and argued that polymorphism was strongly constrained by negative selection, with periodic bouts of genetic diversity at a given locus as a positively selected allele was in transience between ~0% and ~100%. Sewall Wright on the other hand suggested that balancing selection (e.g., frequency dependence, heterozygote advantage, environmental heterogeneity) would maintain polymorphism within a population. The logic in both cases was clear, crisp, and plausible. But it turned out that in a deep way the argument was in the “not even wrong category.” Neutral theory and its heirs pointed out, correctly it seems, that at the molecular level most variation was driven by non-adaptive forces such as random genetic drift. Though some thinkers had conceptualized the model in its broad outlines prior to the empirical results, it was the latter which crystallized the need for a robust model and marginalized the older debate centered around adaptation and natural selection. But even here neutrality is not a model to explain it all. There are cases where adaptation and natural selection are relevant. In some instances you see classical dynamics with transients generated by positive selection sweeping through populations, and in other cases balancing dynamics may be operative. The overall point is that we must always be careful about bald assertions of the form “the latest research overturns….” in this area. Evolution is such a sprawling and cosmopolitan intellectual empire. Nature is subtle and richly textured, and our conceptual frameworks map onto the shape of reality only coarsely.
As for the paper itself, it’s nice and elegant. Patrick Phillips, who knows a thing or two about evolution and elegans is quoted in Science as saying that “”It’s an amazing study….” The letter to Nature is Parallel evolution of domesticated Caenorhabditis species targets pheromone receptor genes. Here’s the abstract:
Evolution can follow predictable genetic trajectories…indicating that discrete environmental shifts can select for reproducible genetic changes…Conspecific individuals are an important feature of an animal’s environment, and a potential source of selective pressures. Here we show that adaptation of two Caenorhabditis species to growth at high density, a feature common to domestic environments, occurs by reproducible genetic changes to pheromone receptor genes. Chemical communication through pheromones that accumulate during high-density growth causes young nematode larvae to enter the long-lived but non-reproductive dauer stage. Two strains of Caenorhabditis elegans grown at high density have independently acquired multigenic resistance to pheromone-induced dauer formation. In each strain, resistance to the pheromone ascaroside C3 results from a deletion that disrupts the adjacent chemoreceptor genes serpentine receptor class g (srg)-36 and -37. Through misexpression experiments, we show that these genes encode redundant G-protein-coupled receptors for ascaroside C3. Multigenic resistance to dauer formation has also arisen in high-density cultures of a different nematode species, Caenorhabditis briggsae, resulting in part from deletion of an srg gene paralogous to srg-36 and srg-37. These results demonstrate rapid remodelling of the chemoreceptor repertoire as an adaptation to specific environments, and indicate that parallel changes to a common genetic substrate can affect life-history traits across species.
This research had multiple stages and methods. I’ll focus on the “big picture” evolutionary ones. First, the proximate phenomenon which they’re spotlighting is the life history of nematode worms. Most of the worms used in research today are a specific strain of C. elegans. When I say specific strain, I mean that they derive from a specific line isolated ~50 years ago. This worm consists predominantly of hermaphroditic self-fertilizers, though a small minority are males. So there is some traditional sex (between males and hermaphrodites). Most of the work in this paper focused on three classes of elegans, a conventional population, and two independent lines adapted for the peculiarly nutrient rich environment of laboratories. In the normal course of events of elegans populations reach their Malthusian limit or carrying capacity, and the chemical signals in the environment trigger a shift of many into an alternative morph which serves to flip the individual into stasis, the dauer. In the context of the laboratory this life history adaptation is arguably maladaptive, insofar as high density does not correlate with a necessary exhaustion of nutrient supplies. Humans can continue to add more nutrients as necessary into the medium in an absolutely unnatural manner.
Over time evolution adapts to this situation in some cases. Mutant forms of elegans have arisen which do not or can not enter into the dauer phase. This would be highly deleterious in the wild, a disastrous loss of functionality. Resource scarcity is a common feature of the natural world. But in the artificial environment of the laboratory the constraint is released. On the contrary, loss of function mutations which in the wild would be highly deleterious become beneficial, illustrating the context-sensitive nature of natural selection. The two mutant lineages under consideration here don’t enter into a static phase when natural maximal population density is reached, they continue to develop and reproduce as they would at lower densities. In an infinite resource environment the mutants can marginalize the “wild type,” because the latter are optimized for conditions which no longer hold.
By crossing inbred lineages of the distinct elegans populations they confirmed the likelihood that the loss of function was due to a mutation of large effect (Note: this sentence may be misleading, see this comment). When you have a quantitative trait whose genetic architecture is due to diffusion across many genes of small effect a cross between individuals with divergent phenotypes often spans the gamut of the range of the trait value. In this case the crosses exhibited a bimodal distribution in response to chemical signals; a tendency toward one morph or the other (wild type or loss of dauer mutant). This implies that a gene or a few genes are segregating within the population in two functionally distinct variants. By comparing one of the mutant lines with the wild type they localized a major genetic difference at the srg-36 and srg-37 loci on the X chromosome. Reintroducing wild type variants of the gene into the sequences of mutants resulted in the reversion back to wild type phenotype. An interesting point is that there were differences in the mutation across the two non-wild type lineages. This confirms the likelihood that what you have here are two independent mutations which converged upon a common phenotype. Since it isn’t necessarily hard to break a function, this shouldn’t be too surprising.
So with the elegans they reported results which confirmed that three lines derived from a common inbred strain ~50 years ago varied in phenotype in such a fashion than the two mutants exhibited parallel adaptations to the peculiar environment of the laboratory medium. The genetic changes were broadly similar, though differed in important details which confirm the likelihood of independence of mutation. But the authors also looked at another worm species, briggsae, which diverged from elegans ~20 million years ago. Because of the date of divergence they couldn’t simply compare briggsae to elegans in a straightforward fashion on the genomic level (the genes weren’t perfectly analogous across species). But they did report that a similar loss of function mutation on an srg gene which inhibits dauer formation can be found in briggsae.
The authors conclude:
These observations indicate that genetic trajectories during evolution are constrained and that adaptation can, at least to some extent, be predictable. One class of known adaptive genes are input–output genes, developmental regulators with complex cis-regulatory motifs that provide a molecular substrate that allows sculpting of developmental patterns…The genes srg-36 and srg-37 seem to fall into a second class of adaptive genes, including opsin genes and taste receptors…sensory receptors whose diversity allows circumscribed adaptation to environmental changes without pleiotropic effects.
Notice words like “to some extent.” These specific results are great, but what are their general insights? How general are they? I don’t know. A few years ago a nice paper came out which ended with: “Our findings indicate that even if selection for the ancestral function were imposed, direct reversal would be extremely unlikely, suggesting an important role for historical contingency in protein evolution.” Researchers can find both contingency and inevitability, depending upon circumstances. The Science piece notes the peculiarity of these genes (also alluded to above in the letter itself):
Fifty to 100 genes affect whether a worm enters the dauer state. In theory, deletions on any of them could keep worms from becoming dauer larvae. But many of these genes affect several aspects of the animal’s development and physiology, whereas the pheromone receptors simply sense the environment and thus can be lost harmlessly, Bargmann suggests. The study may point to “a general rule,” adds Phillips: that evolution tends to delete genes whose loss will not have widespread effects, an idea that is very slowly gaining ground.
There are lots of ways to break this particular function, but only this particular way seems to be likely and not have other serious side effects. In other words the release of constraint on this locus has to be framed in the context of a lot of restriction on the rest of the genome. We need to think of evolution as more than a flux of populations, rather, evolution operates in a genomic landscape, which is characterized by its own internal logic. There’s an underlying unity and consistency to all of this. The disjunction between the coarse generalities possible in human language and the true generalities has more to do with our cognitive limits in conceptualization than the constraints on the science. Many of the generalities about evolutionary process are entirely robust, when accounting for background conditions. I suspect one of the problems is that many researchers assume particular background conditions as implicit in a specific context, but that does not carry through when it comes to more public discourse because naturally the target audience does not have a thick network of background facts into which new results can be snugly nested. The fundamental core logic of evolutionary theory is rather elegant and simple. How it plays out though is very complex.