In fact Kirschner and Gerhart do not so much resolve Darwin’s dilemma as invent a new and, in their view, better theory. Reviewers have not uniformly agreed.17
Kirschner and Gerhart, and indeed the entire evo-devo field, inadvertently do more to undermine Darwin than to save his theory. The first and most obvious concession made by evo-devo (tacitly or otherwise) is that profound, fundamental evolutionary questions had heretofore been utterly unexplained. The rise of multicellular animals, the appearance of novel processes and structures, not to mention novel cell types—none of those had been explained by Darwin’s basic theory, even as elaborated by the “neo-Darwinian synthesis” of the mid twentieth century.
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. Consider the following examples:
François Jacob wrote, “When I started in biology in the 1950s, the idea was that the molecules from one organism were very different from the molecules from another organism. For instance, cows had cow molecules and goats had goat molecules and snakes had snake molecules, and it was because they were made of cow molecules that a cow was a cow.”18 As Jacob ultimately learned, however, that was completely wrong.
In the 1960s Ernst Mayr, an architect of the neo-Darwinian synthesis, confidently predicted on Darwinian grounds that “the search for homologous genes is quite futile,” of which Sean Carroll notes, “The view was entirely incorrect.”19 In retrospect, it is astounding to realize that the strong molecular similarity of life, which Darwinists now routinely (and incorrectly) appropriate as support for their entire theory, was not anticipated by them. They expected the opposite.
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.”20
Writes Carroll, “The most stunning discovery of Evo Devo [that similar genes shape dissimilar animals]…was entirely unanticipated.”21 And “biologists were long misled” to think that simple legs were quite different from complex legs. “But it is wrong.”22
Kirschner and Gerhart are repeatedly surprised: They have a section entitled “The Surprising Conservation of Compartments.”23 And “It came as a surprise (if not a shock)” to find the same regulatory genes expressed in the heads of Drosophila and mammals. “Until that time, it was widely thought that the vertebrate head is entirely novel, the invention of our phylum.”24 According to Walter Gehring, the same goes for eyes. “This is an unexpected finding since the single lens eye of vertebrates was generally considered to have evolved independently of the compound eye of insects because these two eye types are morphologically completely different.”25
Time and again, by intentionally reasoning about animal life on Darwinian principles, the best minds in science have been misled. They justifiably expected randomness and simplicity, but discovered depths of elegance, order, and complexity. As National Academy of Sciences president Bruce Alberts exclaimed, “We can walk and we can talk because the chemistry that makes life possible is much more elaborate and sophisticated than anything we students had ever considered.”26
A third point is that, although it is polite and deferential, discontent with traditional Darwinism rumbles among many scientists who think most intently about evolutionary issues. As Smithsonian paleontologist Douglas Erwin wrote in his review of The Plausibility of Life, “Kirschner and Gerhart’s book must be placed in the context of a number of other recent contributions to evolutionary thought, all of which argue that the current model of evolution is incomplete [emphasis added].”27 Well, is the current model incomplete? “Is there reason to think that our view of evolution needs to change? The answer is almost certainly yes” [emphasis added], avows Erwin, quickly adding “although not, as the purveyors of creationism /intelligent design would have it, because the reality of evolution is under question.” Apparently, by “the reality of evolution,” Erwin means common descent, although he does not use this term.
In fact, some recent authors promoting modularity strongly insinuate that Darwin’s theory as it has been understood by most biologists of the past century and a half could not account for major features of life. Only now is it credible. If the most recent findings were not correct, they say, Darwinism would be forlorn. Toward the end of their book Kirschner and Gerhart coyly ask;
Can evolution be imagined without facilitated variation? What capacity to evolve would a hypothetical organism have if it did not have facilitated variation? If animals did not use and reuse conserved processes, they would, we think, have to evolve by way of total novelty—completely new components, processes, development, and functions for each new trait. Under these circumstances the demands for “creative mutation” would be extremely high, and the generation of variation might draw on everything in the phenotype and genotype.28
Their clear implication is that without facilitated variation—their own brand-new proposal—Darwinism would fail.
As a computer scientist interested in evolutionary algorithms, University of Southampton lecturer Richard Watson comes at the topic from a different angle, but he arrives at the same conclusion as Kirschner and Gerhart. In Compositional Evolution, Watson lays it on the line:
In computer science we recognize the algorithmic principle described by Darwin—the linear accumulation of small changes through random variation and selection—as hill climbing, more specifically random mutation hill climbing. However, we also recognize that hill climbing is the simplest possible form of optimization and is known to work well only on a limited class of problems [emphasis added to the last sentence].29
Those problems include very simple ones that can be solved by changing just one or a few variables—as in the evolution of drug resistance or the resistance of humans to malaria. “Darwin’s masterful contribution was to show that there was some principle of optimization that could be implemented in biological systems,” allows Watson—just not the right one for complex systems. Watson proposes his new idea of “compositional evolution,” which boils down to more modularity. Without compositional evolution, implies Watson, evolution by unintelligent processes would be a no-go.
In sum, the new evolutionary writings have unintentionally done much to damage Darwin, but have not offered convincing alternatives to replace him.
IT ONLY GETS WORSE
Let’s acknowledge that genetics has yielded yet more terrific (and totally unanticipated) evidence of common descent. Has evo-devo produced a new way for random mutation to explain basic features of animal life? No, exactly the opposite. It’s not hard to see why, more than twenty years after the first animal control proteins were sequenced, evolutionary biologists are still utterly unable to give a concrete account of how to explain the unintelligent evolution of animal forms.
In Chapter 7 we encountered unanticipated bottom up–top down construction in systems that build cellular machinery such as the cilium. In retrospect, we realized that the need for specific systems to construct a cilium, as well as for intricate genetic control programs to coordinate the construction, greatly complicates the task of explaining them. The control systems are a further layer of complexity—on top of the complexity of the finished systems themselves—which we in our innocence had not fathomed would be required. The need for control systems does not make the task of Darwinian explanation easier; it makes it far worse.
In the same way as for molecular machinery, in the past several decades developmental biology has unexpectedly discovered the need for c
areful, bottom up–top down planning in the construction of the entire animal. As Eric Davidson trenchantly noted, “Development imposes extreme regulatory demands…. All of the structural characters of the edifice, from its overall form to minute aspects that determine its local functionalities such as placement of wiring and windows, must be specified in the architect’s blueprints” (my emphasis). As with molecular machinery, the elaborate assembly control instructions for whole animals are a further layer of complexity, beyond the complexity of the animal’s anatomy itself. The inadequacy of Darwinism to account for the intricacies of animal development has not been lessened by recent discoveries; it has been greatly exacerbated.
ALL THINGS CONSIDERED
Even though the castaway of Chapter 8 didn’t have hard estimates of probabilities, in light of his experience of nature and his sure knowledge of the design of the wrecked ship, he confidently judged that the neatly piled square of stones and other island features were purposefully designed, rather than the result of some bizarre accident such as a lightning strike. Similarly, although hard numbers are difficult for us to come by, in light of our knowledge of the design of spectacular molecular systems such as the cilium and our experience of nature (particularly our experience with the havoc wreaked by random mutation—even when it “helps”), we can confidently judge that the kind of coherent, multistep control system that Davidson’s observation indicated was demanded to build an animal body was purposely designed.
But how deep does that design extend? There are many distinct animal forms, which biologists have long placed into hierarchical categories such as phyla, classes, and orders. We must remember that randomness does occur and can explain some aspects of all areas of life. So, based on developmental biology and our new knowledge of life’s molecules, can we draw a reasonable, tentative line between Darwin and design in animal evolution? Does design stop at, say, the level of phyla? Or classes? For example, given a generic animal in the distant past with twofold, bilateral symmetry, is it biologically reasonable to think that at that point the rest of the animal world could evolve by random mutation? Or not? Again, we have to keep in mind that few pertinent, quantitative experiments directly applicable to that question have been done. What’s more, further lab work will almost certainly uncover much greater complexity in animal development and other relevant facts, so our appraisal will have to be revised as more information comes in. Nonetheless, there are enough data already in hand to form a reasoned estimate.
To prepare to locate a provisional edge of animal evolution, let’s consider several important factors. When we pass from considering single-celled creatures to multicelled animals, two big things change, in opposite directions. First, we find that animals already are endowed with a passel of toolbox components such as Hox genes to play with (which is where evo-devo musings generally start). That just might open up random-evolutionary possibilities.
(As an aside, it is fascinating to note that the appearance of Hox toolbox components seems to have significantly predated the appearance of new animal forms. As Sean Carroll remarks:
The surprising message from Evo Devo is that all of the genes for building large, complex animal bodies long predated the appearance of those bodies in the Cambrian Explosion. The genetic potential was in place for at least 50 million years, and probably a fair bit longer, before large, complex forms emerged.30
Another surprise to Darwinists! To an intelligent design proponent such as myself, this is a tantalizing hint that parts were moving into place over geological time for the subsequent, purposeful, planned emergence of intelligent life.)
But second, population sizes plummet, which greatly restricts Darwinian possibilities. No multicelled species can match the sheer population numbers that bacteria reach. When we consider animals, we now have many, many fewer than 1040 organisms—the number of bacterial cells that have likely existed on the earth since it formed. As pointed out earlier, the number of malarial parasites produced in a single year is likely a hundred times greater than the number of all the mammals that have ever lived on earth in the past two hundred million years.31
As the population sizes associated with multicellular organisms drop, they begin to fall out of the “couch potato” evolutionary class into the “frail old man” class. In other words, unlike single-celled organisms, larger multicelled animals can no longer be expected to jump more than one missing mutational step, simply because they have fewer chances to generate beneficial mutations. As a rule, each and every mutation—each nucleotide or amino acid change—along the path to a new feature would have to be either beneficial or at the very least not harmful. To reiterate Allen Orr’s conclusion, “Given realistically low mutation rates, double mutants will be so rare that adaptation is essentially constrained to surveying—and substituting—one-mutational step neighbors.”32
How does that affect our estimation? A reasonable, informed person would find it hard to disagree with Stone and Wray’s expectation, “as with amino acid substitutions within coding regions of genes, we predict that in many cases the consequences of a new binding site appearing within a promoter will be either detrimental or neutral; only in rare cases will it be beneficial.”33 So here is the key judgment: It seems a reasonable approximation to treat changes in switch regions, regulatory proteins, and so on, roughly the same as changes in protein-binding sites. That is, if some new control mechanism requires several coherent steps to set it up, for example two or three control proteins acting in concert, then it is reasonable to consider that as roughly equivalent to several proteins binding to each other in a useful multiprotein complex, and to rule out random mutation as an explanation for it.
Why does the fact of multiple coherent steps matter? In Chapter 3 we saw that resistance of malaria to chloroquine was found in only one in a hundred billion billion organisms ( 1020—a CCC) because it required skipping an evolutionary step. In Chapter 7 I argued that three different proteins (two new binding sites) forming a specific complex was beyond the molecular edge of evolution, because it was a double CCC, 1040. The likelihood of the event was so low it would not be expected to occur in the history of the earth, because an organism would have to jump a number of evolutionary steps. Here, with many fewer organisms available, the argument is that forming a new control mechanism for some feature of animal development involving about three or more different kinds of proteins or switches is also a reasonable place to draw the edge of evolution for animal form, because, again, evolutionary steps would probably have to be skipped.
Admittedly, this is a fuzzy estimate—necessarily so, because our current data are limited. Nonetheless, the uncertainty shouldn’t deter us from reaching at least some reasonably firm judgments, because some major control mechanisms uncovered so far are well beyond this measure.
DEEPER AND DEEPER
First, let’s consider a control mechanism that is known to be very complex—one that showcases the sense of Eric Davidson’s exclamation that “development imposes extreme regulatory demands.” A recent special issue of the Proceedings of the National Academy of Sciences explored “genetic regulatory networks”; that is, the control machinery that is necessary to build animal bodies. As the editors Michael Levine and Eric Davidson explain:
FIGURE 9.3 Schematic drawing of a developmental gene regulatory network for sea urchin endomesoderm. The network is strongly evocative of a complex electrical or computer-logic circuit. The figure is reproduced from http://sugp.caltech.edu/endomes/, courtesy of Eric Davidson, who wished to have noted that “Permission for use of this figure is not to be construed to indicate the agreement of its authors with the overall thesis” of this book.
Gene regulatory networks (GRNs) are logic maps [emphasis added] that state in detail the inputs into each cis-regulatory module, so that one can see how a given gene is fired off at a given time and place…. The architecture reveals features that can never be appreciated at any other level of analysis but that turn out to embody distinguishing and deep
ly significant properties of each control system. These properties are composed of linkages of multiple genes that together perform specific operations, such as positive feedback loops, which drive stable circuits of cell differentiation.34
Figure 9.3 is an illustration of the genetic regulatory system that turns on the genes that control the construction of a tissue called the endomesoderm in sea urchins. Notice the obvious, impressive coherence of the drawing. The figure is intended to be strikingly reminiscent of a complex electronic or computer-logic circuit, because in essence that is what genetic circuits are. The system contains a core of six genes that code for master regulatory proteins that eventually switch on scores of proteins that boast many more DNA switches, very far beyond the criterion of three proteins or switches. We can thus conclude this system is well beyond the edge of evolution. It was very likely purposely designed.
Eric Davidson and Douglas Erwin describe the core of the control system for sea urchin endomesoderm as a genetic regulatory network “kernel,” the most basic type of regulatory network now known.35 Kernels, they say, have a number of properties, including:1) they “specify the spatial domain of an embryo in which a given body part will form”; and 2) “interference with expression of any one kernel gene will destroy kernel function altogether”—in other words, they are irreducibly complex. If all the genes are necessary for kernel function, it would have required many coherent evolutionary steps to set up. Kernels in general can be expected to have a degree of complexity similar to that for sea urchin endomesoderm, so we can infer that other kernels also were designed.
Animals are divided into a number of groups according to their general “body plan.” For example, one group of animals, chordates (which includes vertebrates like us), have a nerve cord arranged in the back of their bodies, whereas arthropods, the group that includes insects and crustaceans, have a nerve cord in the front. Biologists count dozens of fundamentally different body plans. Types of animals that have the same body plan are generally grouped together in the same phylum, which is the biological classification right under kingdom (kingdom divides organisms into bacteria, plants, animals, and a few other categories).
The Edge of Evolution Page 21