Oh, another "I should have been an evolutionary microbiologist" post:
Over at New Scientist there's an article about the genetic differences between humans and chimps. They talk a bit about how a superficially small percentage of differences among base pairs (around 1.5%) can lead to modifications in a large number of genes (~80%) since genes are far fewer in number because they typically consist of a very large number of base pairs. Okay, as they say, simple math.
But tucked in the end is surprise that tiny changes in base pairs can radically alter the shape of the proteins coded for by genes. "[W]e thought that protein structures would be highly conserved."
I shouldn't really rely on New Scientist to give me a sense of state-of-the-art consensus in a particular scientific community, but to me it seems obvious that even as evolution selects for particular traits, it's also going to select for adaptability. And not just adaptability in the sense of an organism's being able to react to changes in its environment, but also adaptability that allows ontogeny to proceed flexibly in different directions depending on the prevailing environmental conditions, and, as in this case, genetic heritage that allows minor genetic differences to result in big phenotypical differences that are at the same time useful. Genotypes selected for this quality would contain genes highly adapted to a particular environment, but they would also allude to many, many other genes, not actually contained in the genome but only one or two hops of mutation away, or maybe sequences that would appear if the genes in the genotype were reshuffled, or genes that code for traits that appear emergenetically with other genes in a population. Some of them would be atavisms, but they need not ever have appeared in the population before. All of these ghost genes could be selected for in the abstract as a way of coping with changing environments. It seems to me that the selection pressure here would be quite high, given the disparity between the average life of an organism and the average time between appearances of new genes in a population.
I feel like too many people (and I don't know if this is scientists or just science writers) have one of three conceptions of the evolutionary process. One is an engineer's point of view, which says that if you are going to design a bird, then you need genes for wings and genes for beaks and genes for feathers, and then you put them all together. And yes, I bet if you parsed out an entire bird genome, you would discover genes that mapped roughly to those phenotypical traits. But you run into two problems. The first is a familiar one, which is the gene shortage. There just aren't enough genes in a bird genome to even begin to explain all the phenotypical details of a bird. Even the rather gross and naive mappings seem to run out before you even get started. The other problem is the one that creationists always point to: how do you develop genes for wings without genes for feathers? And even if you develop them simultaneously, what have you got in the meantime, and what purpose do they serve?
Another view of evolutionary development I call the "Lego" theory. In this view, evolution is just a process of rearranging building blocks that are really, really modular. Some are more specialized than others, but they are generally modular and numerous enough and small enough that you can build any arbitrary shape out of them. The blocks themselves are generall held to be discrete and/or indivisible. Computer scientists seem to really like this idea, because they're all about modularity, and such building blocks look to them like the "it's all ones and zeroes" doctrine. And I guess if you look at a genome, it's all A, C, T, and G (or strings of codons, or proteins, if you want to get fancy). But the Lego paradigm has problems of its own. First of all, as my big brother would remind me when his Godzilla impression visited Melissatown, Lego structures are really brittle. Take out pieces and they fall apart. Get the pieces in the wrong place, and they don't fit together at all. And because the pieces are so tiny you need to have a very big, very specific plan for assembling them. As you may remember from childhood, they don't assemble themselves. Aside from the question of where you're going to store this plan in all its specificity, the space of all possible plans is going to be really quite large. The Lego model solves one problem of the engineer view, allowing you to imagine intermediate steps along the way to evolving complex traits, but it fails to explain what the engineer view does best: design coherence. If you deviate from a particular plan, it's possible to derive highly specialized, coherent designs, but you're more likely to end up with what I always ended up with, hodge-podge designs that didn't hold together very well, multicolored cinderblock houses on wheels with flowers growing on the roof.
The you have the rather thudding tree view. This is the idea that divergences follow divergences, branchings follow branchings, and eventually enough of these discrepancies accumulate exponentially to explain the diversity of life. This view is appealing because it probably does reflect the historical progress of evolution on both a macro (divergence of species) and micro (variation within a population) scale. That's about the limit of its explanatory power, unfortunately. It doesn't really tell you very much about ontogeny, or phylogeny, or what kinds of genes you might find in a genome. Just that the road not taken has made all the difference. It's true in a sort of trivial sense.
I like what I call the origami view. In origami, every practitioner learns what are called "bases," simple forms that serve as starting points for more complicated models. All models derived from a particular base share certain features, like the number of points that can be turned into limbs or wings or whatever. They are fertile ground for elaboration. But more importantly, bases don't differ all that much from each other, and techniques for elaborating on points can be transferred from one base to another. Unlike the engineered design, it is easy to see small differences in elaboration can remain useful. Unlike the modular Lego design, the final destiny of every piece does not have to be micromanaged from a global master plan, and small variations can be quite productive. And it fills out the evolutionary tree by helping to give shape to the kinds of variations and branchings we would expect to see.
But in this model, I'm not just talking about the way details accrete to gross structures. Obviously that is a familiar process in both the development of an organism and the evolution of a species. Instead, consider the problem you would face if you wanted to design a new origami model, say, something complicated like an origami giant squid. You could, if you wanted, just folding a piece of paper in random ways, fumbling for inspiration. But if you were an experienced paper folder, you would know about these already existing bases, and you might be able to repurpose one of them. And if that didn't work, you might be able to take the generalized procedures for folding bases to develop a new, more useful base. Not only does a particular origami form contain adaptability because it is derived from a base that easily admits elaboration, but origami as a discipline is very adaptable because its techniques are transferrable to different contexts among many different forms. The space of possible forms is reduced, but the distance between significant forms is much smaller.
It's a truism, of course, that one expects to find more variation in traits if that variation leads to less variation in reproductive fitness. On the other hand, traits that cause drastic changes in reproductive fitness will hardly vary at all, because of the obvious evolutionary pressure not to vary. One obvious response to such evolutionary pressure is to try to conserve such traits (and their underlying genetics). But conserving traits is maladaptive, in a certain way, because it makes it harder for a species to adapt if the environment changes in the future. Now, of course, the species doesn't care about theoretical future environments, only about surviving in the current one. But over the course of billions of years of evolution, it retains the memory, if you will, of many past changes in environment and the need to adapt. Therefore, genes that have proven themselves not just valuable in the current environment, but also a fertile substrate for novel adaptations, will have been selected for. Or to put it a different way, these genes respond to a situation where small changes in traits result in large changes in reproductive fitness by reducing the latter. The pressure to anticipate future changes in environments exists in the immediate present after all!
Adaptability is a trait that is selected for, or maybe I should call it a meta-trait, because it is not associated with a given gene, but rather with the entirety of the genome. It's easy to imagine that big changes in an organism's form can be associated with tiny changes in embryonic development. I would take this a step further and say that this fact's enhancement of adaptability is part of what selected for the embryonic style of development in the first place.
Redundancy, the repetition or backing up of genes, might be another such meta-trait. Robustness, a measure of the number of changes can be made without affecting the phenotype, might be another one. Punniness might be another. Gene density? There have to be more. One nice thing about metatraits is that they might be things that would select for a particular way of encoding genomes, rather than their actual content. Things like codon usage, for instance.
Well, this is getting pretty far afield. I'm sure someone's thought of all this already. But this is a long-winded way of saying that the result that protein shapes are not conserved shouldn't be in the least surprising. Especially, if we're talking about non-polymorphous genes that differ across species. That's where the action is going to be, for crying out loud.
--Melissa O, at 01:04