Jason Rosenhouse has dug into the details of the evo-devo chapter of Behe’s The Edge of Evolution and found some clear examples of dishonest quote-mining (so what else is new, you may be thinking—it’s what creationists do). I’ve warned you all before that when you see an ellipsis in a creationist quote, you ought to just assume that there’s been something cut out that completely contradicts the point the creationist is making; Rosenhouse finds that Behe gets around that little red-flag problem by simply leaving out the ellipses.
I just want to expand a little bit on one point Behe mangles and that Jason quotes. It turns out I actually give a lecture in my developmental biology courses on this very issue, the mathematical modeling antecedents to insect segmentation, so it’s simply weird to see Behe twisting a subject around that is so well understood in the evo-devo community, and that was actually well explained in Sean Carroll’s Endless Forms Most Beautiful.
First, here is what Behe says. There is a germ of truth to these particular quotes, but the conclusions and the details are all wrong.
The next unwitting evo-devo point is even more striking: Basic features of life were totally unpredicted by Darwin’s theory. In fact, reasoning straightforwardly in terms of Darwin’s theory led badly astray even the most eminent evolutionary biologists, who reached conclusions completely opposite to biological reality..
As you’ll see, there were unexpected details of developmental patterning that were not predicted; that part is true, and unsurprising. If there’s one thing we know about evolution, it is that outcomes are influenced by chance, which means we can’t very well sit down ahead of time and lay out the entire future of a lineage, anymore than we can sit down to a poker game and predict how every hand will be played.
But there are other problems with Behe’s claim. What he’s about to explain are not “basic features of life”, but the specifics of metazoan pattern formation. We know already that there are multiple ways you can generate patterns in an organism; the “mistake” we saw made was that developmental biologists sensibly proposed the simplest explanation first, as wise old Ockham would have instructed us, and discarded that as more complications were discovered.
Another error by Behe: these were not ideas derived from “Darwin’s theory”. Darwin’s theory was a general description of how organisms evolved. He did not know anything about genes, morphogens, reaction-diffusion models, computers, or any of the concepts used in this kind of work.
And finally, the early conclusions were not opposite biological reality. The early modeling is still useful and can be used to explain what’s going on in, for instance, vertebrates; it simply turned out that the animal model used in early investigations of the molecular basis of pattern formation, Drosophila, is highly derived and has acquired some very specific, hard-coded regulatory elements on top of a general principle.
Behe is all worked up about one particular example of pattern formation, segmentation. Many metazoans are built around repeating structures—for example, our vertebrae, the stripy organization of annelids, and the obvious cuticular segmentation of arthropods—and prior to the molecular genetics revolution in developmental biology that began in the 1980s, researchers sought out minimal models to explain how you could generate a pattern of repeating elements in an animal. This was some clever and elegant work, but Behe simply and incorrectly dismisses it as entirely wrong by quoting Sean Carroll out of context, and as if the results somehow invalidate evolutionary biology.
Mathematicians, too, were fooled, “Many theoreticians sought to explain how periodic patterns [such as fruit fly embryo segments] could be organized across large structures. While the maths and models are beautiful, none of this theory has been borne out by the discoveries of the last twenty years.” “The continuing mistake is being seduced into believing that simple rules that can generate patterns on a computer screen are the rules that generate patterns in biology.”
I’m fairly familiar with the older literature on pattern formation—I had the good fortune to have gotten my undergraduate training in developmental biology under the old paradigm, and gone on to grad school as the evo-devo influence was just beginning to grow. Mathematicians weren’t “fooled”. They were trying to model how, for instance, a chemical gradient could be transformed into a reiterating pattern of specific activation of other chemicals. It was good stuff; read some of Hans Meinhardt’s modeling work, for instance, which was very useful in explaining patterns of gene activation, once the genes had been identified. It was also not just free-floating speculation with a computer—Klaus Sander, as one example, was an excellent experimentalist who postulated the existence of gradients of morphogens from perturbations of embryos before the specific molecular agents were identified.
It was all interesting work, but as Jason points out and as Carroll clearly states, it wasn’t derived from Darwinian principles. It was from a very clever computer scientist and mathematician.
The revelation of how these stripe-making switches work clarified a long-standing question in the study of pattern formation in biological structures. For several decades, mathematicians and computer scientists were drawn to the periodic patterns of body segmentation, zebra stripes, and seashell markings. Heavily influenced by a 1952 paper by the genius Alan Turing (a founder of computer science who helped crack the German Engima code in World War II), “The Chemical Basis of Morphogenesis,” many theoreticians sought to explain how periodic patterns could be organized across entire large structures. While the math and models are beautiful, none of this theory has been borne out by the discoveries of the last twenty years. The mathematicians never envisioned that modular genetic switches held the key to pattern formation, or that the periodic patterns we see are actually the composite of numerous individual elements.
There is some beautiful work on reaction-diffusion models and chemical oscillators that generate nice striped patterns on computer screens and in petri dishes. It was very seductive, and you can’t blame the early investigators for thinking that maybe this was the answer to how embryos assigned cells to segments. It was nice math, and it was all so very elegant and simple. All you needed was a source and a sink for a couple of diffusible reactants with some specific properties, and voila, repeating stripes emerge.
However, the work on Drosophila segmentation (and Carroll was a prominent contributor to that) showed a different, more complex pattern, that instead of a few components that oscillated in their expression along the body axis, there were many components, and that each one had complex regulatory elements that responded in a discrete fashion to a gradient of a morphogen and to specific epistatic interactions with many other genes.
Neither result, either a simple system with few components or a complex system with many genes intricately regulated, contradicts evolutionary theory. For Behe to imply that modern evo-devo in any way justifies doubt about evolution is dishonest—the regulatory elements that Carroll is describing evolved. There’s a reason I reposted that article on pair-rule genes yesterday — these are the genes Carroll and Behe are discussing. Behe looks at that complicated hodge-podge of madly interacting genes, and thinks it must be designed; Carroll and I look at it and see bricolage, semi-random elements recruited ad hoc and incrementally into a functional assemblage.
What Carroll is talking about when he says that patterns are a “composite of numerous individual elements” rather than the elegant, simple mathematical rules of the early developmental modelers is this: when we look in Drosophila at a gene like even-skipped that has a beautifully periodic expression pattern — it’s turned on in segments 1, 3, 5, 7, etc. — we don’t find a general rule that controls gene expression in a clever way, we find brute force clumsiness. I know there are some people who’ve done some computer programming reading this. Imagine you’ve got some simple problem you’re supposed to solve, where when a certain variable is odd, you do X, and if it’s even, you do Y. You’d probably code something like this:
if ((segment_number mod 2)==1) then X else Y;
You’d try to devise some simple function that varies with the desired pattern, and use that to control conditional expressions. That is not how the fruit fly does it. The fruit fly does it in the way the idiot freshman in intro computer science would do it.
if segment_number==1 then X; if segment_number==2 then Y; if segment_number==3 then X; if segment_number==4 then Y; if segment_number==5 then X;
Expression in each segment is hard-coded in the regulatory elements of the gene. Yikes. There is nothing clean and simple here. And it gets even worse: the fly has nothing equivalent to a simple “segment_number” variable. What it has is a collection of other genes with variable levels of expression, and the conditional test is to read the state of those other genes in the cell and make regulatory decisions. Here, for instance is one even-skipped stripe, in parasegment 3, with it’s level of expression (the y axis) plotted on position along the anterior-posterior axis by parasegment number. You can see where the boundaries of its expression come from — it’s activated by the hunchback and bicoid gene products, but it’s repressed by giant and Krüppel. We could probably code this as:
if bicoid && hunchback && (not giant) && (not Krüppel) then even_skipped;
Clumsy as that is, it gets clumsier. Even-skipped also has to be active in parasegments 1 and 5 and 7 and so forth, and so we’re going to have a whole swarm of sorta-Boolean conditionals encoded in the regulatory region of even-skipped. It’s a major patchwork kludge, where tiny, easy modifiers have been swapped in by random processes over evolutionary time, each one nudging the pattern of expression closer or more robustly towards the regular pattern. This isn’t designed at all. This is what we expect from a process of random change refined by selection, and it isn’t what you get from planning.
I do have to disagree with Carroll on one thing, though. He claims that “none of [the earlier modeling work] has been borne out by the discoveries of the last twenty years.” It’s literally accurate, but a bit unfair in spirit. While the specific predictions flopped, I don’t think that’s to their discredit at all — they devised testable models and made predictions about the presence of morphogens, for instance — and their general principles are useful. I think Meinhardt, for instance, has made useful theoretical contributions to understanding how gradients contribute to morphogenesis, and his work was readily adaptable to the experimental results and the details of the molecular engines behind segmentation. Most importantly, though, don’t get trapped in Drosophila thinking. Drosophila is highly derived and intensely weird, and all those kludgy little hard-coded patches have been added over millions of years on top of what seems to be a much simpler, more primitive system … and that system seems to use a simpler network of molecular oscillators to set up periodic patterns. That simpler system seems to be what we vertebrates use. It’s nothing like what anyone predicted before, that’s true, but it is an elegant clock-and-wavefront pattern that works well.
I’m going to be very cruel and stop there. Generating periodic patterns is the subject of my next column for Seed, so you’ll have to wait for the next issue to come out. I did say a few words about the vertebrate mechanism before, though, so you can at least get a hint.
Meinhardt H (1977) A model of pattern formation in insect embryogenesis. J Cell Sci 23(1):177.
Meinhardt, H (1988). Models for maternally supplied positional information and the activation of segmentation genes in Drosophila embryogenesis. Development 104 (Suppl.), 95-110.
Sander, K (1988). Studies in insect segmentation: from teratology to phylogenetics. Development 104 (Suppl.), 111-121.