The Edge of Evolution

by Michael J Behe

  Unlike malaria and HIV, which both have to fend for themselves in the wild and fight tooth and claw with the human immune system, the E. coli in Lenski’s lab were coddled. They had a stable environment, daily food, and no predators. But doesn’t evolution need a change in the environment to spur it on? Shouldn’t we expect little evolution of E. coli in the lab, where its environment is tightly controlled? No and no. One of the most important factors in an organism’s environment is the presence of other organisms. Even in a controlled lab culture where bacteria are warm and well fed, the bug that reproduces fastest or outcompetes others will dominate the population. Like gravity, Darwinian evolution never stops.

  But what does it yield? In the early 1990s Lenski and coworkers began to grow E. coli in flasks; the flasks reached their capacity of bacteria after about six or seven doublings. Every day he transferred a portion of the bugs to a fresh flask. By now over thirty thousand generations of E. coli, roughly the equivalent of a million years in the history of humans, have been born and died in Lenski’s lab. In each flask the bacteria would grow to a population size of about five hundred million. Over the whole course of the experiment, perhaps ten trillion, 1013, E. coli have been produced. Although ten trillion sounds like a lot (it’s probably more than the number of primates on the line from chimp to human), it’s virtually nothing compared to the number of malaria cells that have infested the earth. In the past fifty years there have been about a billion times as many of those as E. coli in the Michigan lab, which makes the study less valuable than our data on malaria.

  Nonetheless, the E. coli work has pointed in the same general direction. The lab bacteria performed much like the wild pathogens: A host of incoherent changes have slightly altered pre-existing systems. Nothing fundamentally new has been produced.25 No new protein-protein interactions, no new molecular machines. As with thalassemia in humans, some large evolutionary advantages have been conferred by breaking things. Several populations of bacteria lost their ability to repair DNA. One of the most beneficial mutations, seen repeatedly in separate cultures, was the bacterium’s loss of the ability to make a sugar called ribose, which is a component of RNA. Another was a change in a regulatory gene called spoT, which affected en masse how fifty-nine other genes work, either increasing or decreasing their activity. One likely explanation for the net good effect of this very blunt mutation is that it turned off the energetically costly genes that make the bacterial flagellum, saving the cell some energy. Breaking some genes and turning others off, however, won’t make much of anything. After a while, beneficial changes from the experiment petered out.26 The fact that malaria, with a billion fold more chances, gave a pattern very similar to the more modest studies on E. coli strongly suggests that that’s all Darwinism can do.

  THE PROTEIN EDGE

  To put the difficulty of developing one or two protein-protein binding sites in perspective, Table 7.1 lists some approximate population sizes and likelihoods for some selected events. Figure 7.4 graphs results from the four dissimilar species: human, E. coli, HIV, and malaria. The number of cellular protein-protein binding sites developed by random mutation and natural selection for each (one for human—due to sickle cell hemoglobin—and zero for the other species) is plotted against the species’ population size. The bottom axis of the graph extends from 100 to 1040; since there likely have been fewer than 1040 organisms during the entire history of the earth, the bottom axis represents all of life.

  FIGURE 7.4

  Graph of the number of protein-binding sites produced by random mutation and natural selection versus the population size for human (circle), E. coli(triangle), HIV (square), and malaria (diamond). (The values used for the population sizes are, respectively: 10 8,10 13,10 20, and 5 x 10 21—10 8 is the approximate number of humans needed to produce a sickle hemoglobin mutation; 10 13 is the total number of E. coliin the experiments of Richard Lenski; 10 20 is the estimated number of HIV in the past several decades, worldwide; the value of 5 x 10 21 is calculated from 10 20 malaria cells each year for the past fifty years, the approximate time since chloroquine was introduced.) Humans developed one binding site (for sickle hemoglobin);the other species developed none. The top of the area shaded gray marks the molecular edge of evolution. Notice that the vertical axis is discontinuous. The star in the upper right marks the approximate number of different kinds of protein-protein binding sites in a typical cell. Extrapolating from the observational data shows random mutation accounts for very few of those sites.

  Most people of course are familiar with the ordinary concept of an average; for example, the average of 0 and 40 is 20. In mathematics there is a concept called the geometric average. It is the number whose exponent is the average of the exponents of other numbers. For example, the geometric average of 100 and 1040 is 1020. In a geometric sense, 1020 is midway to 1040. So in a geometric sense, the observational data plotted in Figure 7.4 cover half of all life that has ever existed on earth, since the data extends past 1020 on the bottom axis. That allows us to be very confident in extrapolating from the data. The arrow in the upper-right-hand corner of the figure is directed toward a large point near 10,000, which represents the rough number of protein-protein binding sites in a typical cell. Somehow all those binding sites developed during the history of life. Straightforward extrapolation from the observational data plotted at the bottom of Figure 7.4 strongly indicates that random mutation accounts for very few of them.

  Earlier in this chapter, using considerations from shape space, I conservatively estimated that the probability of developing a new protein-protein binding site by random mutation and natural selection would probably be on the order of a CCC—roughly the same difficulty as the development of chloroquine resistance in malaria, about one in 1020. After looking at the results from work on P. falciparum and HIV, that estimate may now seem much too generous. CCCs do happen, if the population numbers supply them. In recent years chloroquine resistance has popped up a number of times independently. Yet HIV, despite having undergone more of at least some kinds of mutations than cells have experienced since the beginning of the world, has produced no new interactions between viral proteins. Nor has malaria developed new cellular protein-protein interactions. It seems the likelihood of developing a useful new protein-protein binding site is actually worse than a CCC.

  Could the edge of evolution be as close as a single cellular protein-protein binding site, rather than two? After all, no new such interactions have been uncovered in malaria and HIV. Could it be that shape-space reasoning has significantly underestimated the difficulty of developing a single new binding site in the crowded, tightly regulated interior of a cell? That’s possible, and we always have to keep in mind that these estimates are rough and will be revised as more information becomes available. Still, I think it’s better to err on the side of caution, allow room for the odd exception like sickle hemoglobin, and draw the line at complexes of three kinds of proteins (that is, two binding sites), as I do in Figure 7.4.

  So let’s accept my earlier conservative estimation, and spell out some implications. The immediate, most important implication is that complexes with more than two different binding sites—ones that require three or more different kinds of proteins—are beyond the edge of evolution, past what is biologically reasonable to expect Darwinian evolution to have accomplished in all of life in all of the billion-year history of the world. The reasoning is straightforward. The odds of getting two independent things right are the multiple of the odds of getting each right by itself. So, other things being equal, the likelihood of developing two binding sites in a protein complex would be the square of the probability for getting one: a double CCC, 1020 times 1020, which is 1040. There have likely been fewer than 1040 cells in the world in the past four billion years, so the odds are against a single event of this variety in the history of life. It is biologically unreasonable.

  With the criterion of two protein-protein binding sites, we can quickly see why stupendously complex str
uctures such as the cilium, the flagellum, and the machinery that builds them are beyond Darwinian evolution. The flagellum has dozens of protein parts that specifically bind to each other; the cilium has hundreds. The IFT particle itself has sixteen proteins; even complex A, the smaller subset of IFT, has half a dozen protein parts, enormously beyond the reach of Darwinian processes. In fact, drawing the edge of evolution at complexes of three different kinds of cellular proteins means that the great majority of functional cellular features are across that line, not just the most intricate ones that command our attention such as the cilium and flagellum. Most proteins in the cell work as teams of a half dozen or more.

  If the great majority of cellular protein-protein interactions are beyond the edge of evolution, it is reasonable to view the entire cell itself as a nonrandom, integrated whole—like a well-planned factory, as National Academy of Sciences president Bruce Alberts suggested. This conclusion isn’t a “God of the gaps” argument. Nonrandomness isn’t a rare property of just a handful of extra-complex features of the cell. Rather, it encompasses the cellular foundation of life as a whole.

  8

  OBJECTIONS TO THE EDGE

  This chapter makes some important distinctions and addresses potential objections. It considers counterarguments to my attempt to define the edge of evolution—not philosophical ones, about the “other side” of that boundary, but technical and logical ones about the line itself. After that, at the end of the chapter, I cross the line.

  In order to be as confident as possible about where to draw the line marking the edge of evolution, we have to take into account all the relevant data. Not all protein interactions can be lumped into the same category; we have to make careful distinctions and then check them against the relevant facts. One small point to note, for example, is that it’s three or more different proteins binding specifically to each other that I assert is beyond Darwinian processes, not just three or more copies of the same protein. A number of proteins, like sickle hemoglobin, bind repeatedly to copies of themselves using the same binding site, like many copies of a single simple Lego part that can be stacked on each other. A “stack” of thousands of such proteins, all of a single type, is not beyond Darwinian possibility.

  Another, more important point to note is that I’m considering just cellular proteins binding to other cellular proteins, not to foreign proteins. Foreign proteins injected into a cell by an invading virus or bacterium make up a different category.1 The foreign proteins of pathogens almost always are intended to cripple a cell in any way possible. Since there are so many more ways to break a machine than to improve it, this is the kind of task at which Darwinism excels. Like throwing a wad of chewing gum into a finely tuned machine, it’s relatively easy to clog a system—much easier than making the system in the first place.2 Destructive protein-protein binding is much easier to achieve by chance.

  More interesting than proteins that just gum up cellular defenses are those that allow a pathogen to take advantage of a host cell system. For example, cells have several intricate systems that control their shape, one of which is based on a protein called actin. Actin can form long fibers by assembling many copies of itself, another example of a Lego stack. However, the assembly of actin fibers is tightly regulated by other proteins in the cell, so that it only takes place at the proper time and place. Several kinds of bacteria and viruses subvert the Lego-stacking process for their own benefit by attaching to one of the control proteins, tricking it into thinking actin should be assembled on the pathogen. In effect, the invading pathogen hijacks a cell process, which helps it to spread.

  While that’s a fascinating and medically important process, the pathogen protein just triggers a pre-existing cellular mechanism. Like a tree limb that falls in the wind and hits the switch of a complex machine, turning it on, the pathogen protein does very little on its own. Darwinism can explain that aspect of the pathogen, but not the hijacked process it triggers. Like the development of antifreeze protein in Antarctic fish, such minimally coherent phenomena probably mark the far boundaries of what Darwinian processes can do in microbes.

  AN EXCEPTION?

  One apparently large exception to the difficulty of forming new cellular protein-protein interactions is sickle cell hemoglobin itself. Instead of needing several changes to make a new binding site, sickle hemoglobin needed just one. With just one change in its amino acid sequence, sickle hemoglobin developed a new binding site that allowed it to stick weakly to itself, and thus conferred resistance to malaria on Sickle Eve. Instead of needing a hundred billion billion people, the change required maybe just a hundred million. Why?

  The reason is that the red blood cell is very unusual. Most other types of cells contain many different kinds of proteins, no one of which overwhelms the cell. Because its job is to carry as much oxygen as it can from the lungs to the tissues, by contrast, the red blood cell is stuffed with one protein, hemoglobin, the oxygen-transporting protein—hundreds of millions of copies of it. Although it contains a number of other kinds of proteins as well, about 90 percent of red blood cell protein is hemoglobin. The very high concentration of hemoglobin makes it a lot easier for interactions between hemoglobin molecules to have a noticeable effect. To understand why, let’s go back to the swimming pool analogy and think about objects that fit each other, but poorly. On average they would perhaps spend about 1 percent of their time together. They are easily knocked apart, and then have to drift around for a long while until they accidentally came together again. Well, suppose in the pool we had not just one copy of those poorly fitting pieces, but millions. Now when the pieces stuck and then got knocked apart, they would have to drift for a lot less time to bump into another copy of their partner. Instead of searching for that one mate in the pool, the proteins would have millions to stick to. Because they would spend much less time searching for a copy of their partner, they’d spend a much larger fraction of time stuck together, even though their attachment was weak. If they were symmetrical like hemoglobin, with two identical sides, they could stick to a partner using each face, which could stick to another, and so on, until many of the copies congealed and gummed up the swimming pool. The bottom line is that if a protein is highly concentrated in a cell, as hemoglobin is in the red blood cell, a single, shape-changing mutation has a much better chance of making the protein stick to itself. Conversely, if hemoglobin were present at more typical protein levels, the sickle mutation wouldn’t work.3 At normal levels multiple amino acid changes would likely be needed to make hemoglobin stick to itself.

  A more interesting example than sickle hemoglobin is the case of a protein abbreviated FKBP. The change of one particular amino acid (at position 36) in this protein causes the protein to bind to itself with moderate strength (about a hundred times more strongly than sickle hemoglobin). Using a technique called X-ray crystallography, which allows scientists to visualize almost every atom in a protein, this mutant proved very unusual:

  The interface between the two proteins is characterized by a remarkably extensive and complementary set of contacts suggestive of a bona fide protein-protein interaction rather than an artificial pairing…. Thus the interaction strikingly resembles natural high-affinity protein-protein interfaces…. This result suggests that the…substitution may…relieve an inherent steric hindrance to intermolecular association…. The discrete…change elicited by the F36M mutation is remarkable and, to our knowledge, unprecedented.4

  In other words, it looked like the protein was pre-engineered to be complementary to itself, but was kept apart in the premutated version.5 Switching amino acids in the mutation removed a blockage. In other words, the behavior of the protein FKBP was unlike anything encountered before. The close fit of the protein may mean that it is actually built to self-associate in nature under some circumstances that had previously escaped attention. It might be an example of Darwinian destruction (the scientists unwittingly undid a previous mutation). In any case, FKBP shows the need to be very cautious in inte
rpreting a single experimental result. The subtle tasks of some proteins in the cell might require that they be poised to bind to each other. Mutating proteins as these scientists did could give us a false reading of the difficulty of the task facing evolution. To get a better understanding we should look beyond isolated results to the best general information on evolution we have.

  ACCIDENTAL JIGSAW PUZZLES

  In the last chapter I argued that design could be detected in the very fit of complex parts. But is that always true? Just the other day my six-year-old daughter knocked a vase off a shelf in our home, and it broke into several big chunks. The ragged breaks were complex. No other objects in our home or out of it matched them. Of course, the chunks fit perfectly together, yet they weren’t individually designed. Here’s another example. Suppose a rock fell into a puddle of water. During the night the water froze; a person who carefully removed the rock from the ice would see that the rock and the hole in the ice were exactly complementary to each other. They weren’t designed to match each other by an intelligent agent, as automobile parts are, nor did we have to search through a huge shape-space library to find them.