Life Finds a Way

by Andreas Wagner

Figure 4.1.

  Fitness landscapes, of course, do not exist in only three, four, or five dimensions: they exist in hundreds and thousands. Each dimension corresponds to one feature of an organism that can vary, such as a location in the organism’s genome—be it a gene or an individual DNA letter—that can vary in its DNA text. For many purposes, thinking of such landscapes in three dimensions works well, but a paltry three dimensions fail us when we try to understand the ability to bypass fitness valleys. Fortunately, even though our brains cannot visualize high-dimensional landscapes, we can still explore and map them.

  My research group in Zurich and many others are doing just that, using state-of-the-art laboratory evolution experiments and computing technology. In our experiments, we evolve RNA enzymes and protein enzymes, like the beta-lactamase that neutralizes the antibiotic penicillin. To do this, we manufacture huge populations of identical, high-functioning molecules. We then sprinkle a population’s individuals with mutations in their molecular text, select for mutants that still perform well, and repeat. And while we do this, we sequence the molecular texts of the evolving individuals—thousands of them—and ask our computers to track their location on the landscape.

  What such populations teach us is just as important for evolution’s creative powers as it is surprising and bizarre.

  In these experiments, any one population starts out at a peak of its adaptive landscape. If this peak were isolated on the landscape’s high-dimensional plains, like Africa’s Mount Kilimanjaro, the population’s member molecules would remain huddled around the peak. All mutants would be inferior to the starting molecule, and natural selection would relentlessly purge them from the population. But that’s not what we see. The population does not stay put. It spreads through the landscape. Inexorably, steadily, and with every cycle of mutation and selection, its members take a few further mutational steps. What is more, individual members move away from the starting peak in many different directions. While they move, they get neither much better nor much worse at what they do, staying at about the same elevation on the landscape. The starting peak, it seems, is connected to other, nearby peaks. Dozens of “ridges” emanate from it and lead to these other peaks, from which dozens more ridges connect to even farther peaks, and so on.

  Experiments like these teach us something profound about the architecture of multidimensional adaptive landscapes: a peak is usually not a single location, like the top of Mount Kilimanjaro, but more like a network of high-altitude paths that form a sprawling spiderweb extending far through the landscape. The paths connecting different peaks need not be completely flat, but they also do not snake up and down by much. We know this because our experiments use huge populations of many billions of molecules, and in such populations genetic drift is too feeble to allow any one individual to descend far below a peak.

  Laboratory evolution experiments can only explore a tiny speck of an adaptive landscape, but in its four-billion-year experiment, nature has explored much greater swaths. It has also discovered how far molecules can travel on spiderwebs like these while preserving their unique skills. Take hemoglobin, the oxygen-binding transport protein in our blood. Thousands of other species—mice, reptiles, fish, insects, and even plants—harbor oxygen-binding proteins like it, each of them the endpoint of a long evolutionary journey that started from some ancestral oxygen-binding protein more than a billion years ago. And in this journey, oxygen-binding proteins have not remained unchanged. Their amino acid texts have changed slowly and inexorably, letter by letter by letter. Starting from their ancient common ancestor, these proteins have spread out on a vast spiderweb of useful oxygen transporters. By now, they have traveled so far that they share fewer than fifteen of their hundred-odd amino acid letters. They have become very different texts, different solutions to the same problem of binding oxygen.1

  And hemoglobin shares this pattern of evolution with myriad other evolving molecules, from the catalytic holdovers of a sunken RNA world, to innumerable proteins that catalyze biochemical reactions, communicate between cells, support our bodies, and help us move. Whatever each of them does, it does it very well. And the fitness peak each occupies is not merely a single peak, but rather a sprawling network of high-dimensional paths that extends far and wide through an adaptive landscape. Don’t try too hard to visualize this spiderweb, because its very existence would be impossible in three dimensions. It requires hundreds of dimensions and the many directions in which they can be explored.

  What is more, all this applies not only to individual molecules, but also to complex assemblages of these molecules—the biochemical machineries that build and maintain our bodies. Two of these machineries are especially important. One is run by the regulator proteins that control the transcription of many genes. These regulators do not act alone, but instead in complex regulatory circuits whose members regulate both each other and the transcription of hundreds of other genes. The second machinery is metabolism, a complex network of thousands of chemical reactions, each catalyzed by a dedicated enzyme encoded in some gene. Our metabolism and those of all other organisms procure energy and building materials—nutrients—to manufacture the numerous molecules life needs to persist.

  These regulatory and metabolic machineries are encoded in the genome, and when the genome varies through DNA mutations, so do they. This is why a population’s individuals differ in their metabolism, some harvesting energy more efficiently than others, some storing more fat, some tolerating certain foods better, and so on. And this is why some regulation circuits build better bodies than others, constructing larger wings, stronger hearts, or faster neurons. In other words, regulation and metabolism can evolve and improve by DNA mutation and natural selection. But just as important is this: even the circuits and metabolisms achieving peak performance are not all the same. The peak they occupy is not a single Kilimanjaro-like hump in the landscape. Rather, it is a sprawling network of ridges. Along this network, diverse forms of regulation and metabolism can co-exist. They are each able to build and maintain an optimally functioning body but do so in different ways.

  Entire books could be written about this unexpected high-dimensional world. In fact, one has been written. In Arrival of the Fittest, I tell the fascinating story of where these spiderwebs come from, why they are near-universal, and why they are crucial for evolution.2 But what matters here is that they help nature in the act of creation and that we have the experiments to prove it.

  In one of these experiments, Eric Hayden, then a postdoctoral researcher in my laboratory, started with a ribozyme (already mentioned in Chapter 2) that can link an RNA molecule with a specific letter sequence to itself. From this starting point, Eric sought to evolve a more flexible ribozyme that could link a different RNA molecule to itself. Actually, he performed two experiments, each starting with a different population of molecules. The first population was concentrated on one peak of the starting ribozyme’s adaptive landscape—actually, in one location of a high-altitude network of ridges—whereas the second was sprawled out along this network. He asked which of the populations would be the better innovator—which would discover the new molecule faster.

  After a mere eight cycles of mutation and selection we had the answer. The population spread far and wide discovered the more flexible ribozyme six times faster. And the reason is not hard to understand: this new ribozyme occupied a higher peak that existed some distance away from the starting ribozyme, and in the population spread out across the ridges below this new peak, some individual ribozymes just happened to be close to the new peak. They had a head start and could get to the top faster.3

  Much the same story is told by the antibiotic resistance protein beta-lactamase. An experiment in the laboratory of Israeli biochemist Dan Tawfik showed that evolving beta-lactamase genes that are spread out along ridges near their adaptive peak have better odds of developing the novel ability to destroy both cefotaxime and penicillin. Different molecule, same reason: the sprawling network can help them get
close to a new peak without their dipping into a deep valley.4

  That very fact can help explain why resistance against ever-new antibiotics evolves so rapidly. Not only do bacteria divide fast and live in huge populations, but their antibiotic resistance genes are also highly diverse and scattered all over the adaptive landscape, thanks in part to the sprawling spiderwebs of ridges permeating this landscape.5 This diversity increases the odds that one of these genes will find itself near a peak of resistance against a new antibiotic.

  In sum, the peculiar architecture of adaptive landscapes, where each peak really is a network of multi-dimensional ridges, helps evolution solve difficult problems. It helps create diverse organisms and molecules, some of which may be close to even higher peaks—better solutions to old problems or inventive solutions to new problems.

  “Beam me up, Scottie” is an indelible line etched into the collective memory of those who are familiar with the original Star Trek television series.6 James Kirk, the captain of the starship Enterprise, spoke it whenever he needed to get himself out of a tight spot, usually among hostile creatures on an alien planet, whereupon the starship’s engineer, Montgomery Scott, would magically teleport Kirk back onto the mothership. Sadly, like the Enterprise, this kind of teleportation remains a matter of science fiction.

  At least in our daily lives. Because it turns out that nature uses something like teleportation to reach faraway places—not quite a starship, but distant locations on the landscapes where nature’s creativity unfolds. Everybody knows about it, most people like it, but many fewer truly appreciate why it is important: sex.

  Each of the twenty-three chromosomes that host our genes comes in a pair—that’s why we have two copies of each gene. During a special kind of cell division that creates sperm and egg cells, a chromosome pair’s two members line up to exchange some of their DNA text. Imagine each pair as two equally long shoelaces in different colors, say, black and white, lined up in parallel on a flat surface, end to end, one on the left, and the other on the right. While keeping these shoelaces aligned, cut them in an arbitrary place—actually, cut them in a few places—and then swap the resulting fragments between the left and right. When you are done, glue the pieces back together, and you are left with a pair of shoelaces whose members change color—black to white or white to black—at least once along their length. Wherever the left shoelace changes from black to white, the right one changes from white to black.

  When our bodies manufacture sperm and egg cells, this aligning, random cutting, swapping, and regluing—the scientific term is recombination—happens to each chromosome pair, which becomes a mosaic of DNA strings, just like those rearranged shoelaces become a mosaic of black and white strings. And one member of each pair gets stuffed into each sperm or egg cell.

  When two parents conceive a child, the sperm spills the father’s rearranged DNA into the female egg, which contains the mother’s rearranged DNA. The end result is a fertilized cell that has two copies of each of the twenty-three chromosomes that were recombined inside the mother and father.

  In reality, the two DNA strings differ not in their color but in the sequence of their DNA letters, and only ever so slightly, at about one in every one thousand letters for a typical human being.7 That is, if you were to walk along any one of your chromosome pairs, you would find that once every one thousand letters, one of the pairs has one letter, say, A, whereas the other has another letter, C, G, or T. In all the remaining letters, 99.9 percent of them, the two chromosomes are identical.

  In other words, the two members of each of your twenty-three chromosome pairs barely differ. But because there is so much DNA in them—some three billion letters, packed into all those chromosomes—these differences add up, such that, overall, three million DNA letters differ between the two copies of all of your chromosomes.8

  Knowing this, we can also find out how many letters differ between the newly assembled genome of a child and the genome of either parent. One member of each chromosome pair comes from the father, so it does not differ from that father’s genome.9 The other one comes from the mother, which differs from the father by about three million letters. Overall, the child’s genome thus differs from the father’s genome by some 1.5 million letters—the average of zero and three million. By the same calculation, the child’s genome differs from that of its mother by the same amount, some 1.5 million letters, or 0.05 percent of its genome.

  This percentage may not sound like much, but adaptive landscapes can help us grasp its true magnitude. If a single step on an adaptive landscape—a single letter change in a genome—covered as much distance as an average human covers in a single step, then this kind of genome swapping would teleport a child about seven hundred miles, traversed in a single leap. And genomes take such leaps every time two parents have a child.10 If you were to travel this far from the rolling plains around Wichita, Kansas, you could find yourself in the middle of the Rockies in Colorado or Utah, with plenty of new peaks to explore.

  The DNA of human parents—or any two organisms of the same species—usually differs much less than does the DNA of different species. But the more two parents differ in their DNA, the further recombination can leap, and the greater its creative powers can become. Recombination leaps especially far when parents from different species mate and produce offspring—the kind that is called a species hybrid. To be sure, some hybrids are dead ends of evolution—their parents are very different or genetically incompatible, so a fetus cannot develop, or it develops but is sterile. Examples include the mule, a horse–donkey hybrid; the zorse, a zebra–horse hybrid; and the liger, a lion–tiger hybrid. But hybridization can also be very successful. It can even launch entirely new species—instantly.11

  Successful hybridization is especially common in plants, where it creates up to 10 percent of new plant species.12 It also often enables the new species to boldly go where no parents have gone before. Case in point: two hybrid species of US sunflowers from the genus Helianthus. Their parents dwell on the Great Plains, but one of the hybrid newcomers can survive in the deserts of Nevada—it is aptly named Helianthus deserticola. The other thrives in Texan salt marshes. Both new habitats would be deadly for either parent.13

  Animals, too, can hybridize successfully. In 1981, for example, Princeton researchers Peter and Rosemary Grant discovered a new hybrid Galápagos finch when they encountered an especially unusual male specimen on the island of Daphne Major. Not only was this bird 50 percent heavier than other finches, it also sang a new song, had an unusually large head, and had a beak that allowed it to crack seeds that were inaccessible to other finches. By observing “Big Bird’s” descendants through seven generations during the next twenty-eight years—talk about tenacity—the Grants found that its new features were indeed helpful. When a drought wiped out 90 percent of finches between 2003 and 2005, its descendants were among the survivors. DNA analysis showed that Big Bird was a hybrid of two other Galápagos finch species. More than that, it proved that multiple other species of Darwin finches are hybrids.14

  Bacteria cannot hybridize like plants and animals do. However, teleportation across adaptive landscapes is so important that nature empowered even bacteria to do it, though it equipped them with a mechanism very different from ours. Bacterial genomes can include genes that enable one bacterium to donate DNA to another, recipient bacterium. These genes enable the donor to build a long, hollow protein tube—the technical term is a sex pilus—that latches onto a nearby recipient cell, reels it in, and helps transfer DNA to the recipient. The process is also called horizontal gene transfer, and the transferred DNA can help the recipient survive in new environments. Horizontal gene transfer may seem vaguely similar to our version of sex, but it differs in important respects. Bacteria do not have sex every generation, nor do they reproduce with sex like we do. They merely transfer a copy of DNA—up to hundreds of genes—to another cell. Sometimes the process even transfers the very genes needed to build the sex pilus. The result is somet
hing like a sex change that converts a “female” into a “male.”15

  But the most important difference from our sex is this: bacterial sex is available to organisms a hundred-fold more diverse than two humans or even two sunflowers.16 If we had a bacterium’s recombination powers, we would not just routinely blend genes from other humans into our genome, but also those from chimpanzees, mice, birds, or even reptiles and fish. Bacteria can even exchange DNA with animals and plants.17 Just imagine the consequences—on our lifestyle, on world hunger, and on the global economy—if we could acquire a plant’s ability to harvest energy from light and to build our bodies from the air’s carbon dioxide.18

  In sum, bacteria can catapult themselves not just hundreds of miles, but thousands of miles, through a vast genetic landscape, all courtesy of gene transfer.

  It is not surprising then that bacterial teleportation has led to innovations almost as radical as humans capable of photosynthesis, because bacteria all around us exploit teleportation to experiment with ever-new combinations of genes. Among their creative discoveries are gene combinations that help them survive—and even thrive—on nasty man-made molecules, such as pesticides like DDT and pentachlorophenol, or the highly toxic industrial waste product dioxin.19

  Molecules like these were invented by chemists only within the last century, which gives you an idea of how quickly bacteria created the innovations that turned such molecules into food: a mere instant of evolutionary time. And once gene transfer has helped create any one such innovation, it helps spread the innovation from one bacterial species to another and beyond. That’s why the skill of surviving an antibiotic can spread rapidly among different species, so rapidly that human innovators have trouble creating new antibiotics fast enough to catch up.20

  Fortunately, nature’s lessons about distant jumps through an adaptive landscape are not lost on human innovators. They are imagining new mechanisms for genetic teleportation that are more powerful than even bacterial sex. These strategies create new DNA in test tubes through an orgy of molecular recombination that even nature would be hard-pressed to match.21