Life Finds a Way

by Andreas Wagner

  To see why an animal would advertise its presence with a flourish, it helps to know some other organisms that do the same. They include the gaudy but venomous coral snake and the dazzling but toxic poison dart frog. Their message could not be clearer: back off.

  While passion-vine butterflies do not have dangerous fangs, they have a special trick to keep their enemies at bay: their larvae feed on passion vines, which produce dangerous self-defense chemicals, including cyanogenic glycosides. Butterfly larvae can tolerate these poisons, but once a butterfly larva has ingested them, the animal becomes toxic as well.26

  Like billboards on a highway, which are most effective when you see them more than once, warning colors—the technical word is aposematic colors—are best remembered if many animals carry them. In other words, poisonous animals display strength in numbers. If many toxic butterflies in a patch of forest share the same color pattern, they reduce any one animal’s risk of getting eaten. A naïve predator that survives biting a chunk out of one distasteful butterfly will remember that experience for life and avoid all others. But it will gladly take a bite out of a butterfly with a new color pattern, as experiments performed in 1972 by University of Washington zoologist Woodruff Benson prove. He painted over the red stripe on the wings of Heliconius butterflies with the wings’ black background color. Sure enough, when he released the altered butterflies, he found that more of them than the originals were killed over time, and more of the survivors were mutilated, showing bite marks from predatory birds, reptiles, or mammals.27

  With all of this in mind, imagine a fitness landscape whose two cardinal axes distinguish different color patterns on a butterfly wing. For example, one axis might quantify the amount of red, and the other the amount of yellow, relative to a black background. If many butterflies share a similar protective pattern, they create a peak in this landscape. Mutant butterflies whose colors lie off-peak are not protected and must run the gauntlet of hungry predators.

  In the fitness landscape of warning coloration, a peak pulls evolving butterflies toward it because it guarantees safety in numbers. This pull turns out to be so great that even different species of passion-vine butterflies—distinguishable by their antennae, genitals, and other features—have evolved the same warning coloration.28 They are all better off near the peak than anywhere else in the landscape. This is a remarkable example of convergent evolution, a process in which natural selection helps make different species more similar. It is also an example of Müllerian mimicry, the phenomenon where some toxic species mimic other toxic species, named after its discoverer, the nineteenth-century German naturalist Fritz Müller.

  In contrast to a peppered moth, whose wing color needs to match that of a tree’s bark, a butterfly’s warning color is arbitrary, as long as many other butterflies share it and predators can recognize it. Nothing would prevent Heliconius butterflies in different geographical areas from showing different color patterns. In one population, all individuals might share that black wing with a single red stripe, whereas in another they might sport the sunburst of orange and yellow.

  This is indeed the case, and in not just two but more than a dozen different areas, some covering thousands of square miles in the Amazon basin. What is more, different areas don’t just harbor butterflies with different warning flags. Two species that mimic each other in one area often also mimic each other in another area. That would be less remarkable if the protective color patterns in the two areas were the same, because it could be explained if the species migrated between areas. But the color patterns in different areas can be completely different from one another. In other words, species in different areas have converged independently on their area’s protective color pattern. Such multiple instances of convergent evolution underscore the power of the protection provided by a prevalent color pattern.29

  We may never know with certainty how the geographic diversity of these warning color preferences originated, but a hint comes from the much cooler climate that existed on our planet in the Pleistocene starting some 2.5 million years ago. During this time, when large regions of the planet were sheathed in ice, the Amazonian forest habitats of Heliconius may have retreated to smaller forest islands separated by vast areas of open grasslands that could not be traversed by butterflies.30 Imagine such isolated pockets of hospitable land as hothouses of evolution, where different butterfly populations could evolve different warning colorations. Once the globe warmed up again, these forest islands expanded into enormous and continuous swaths of rainforests. Butterfly populations expanded with them but were kept separate by natural barriers such as rivers and mountains.

  Whatever the true origin of their different color patterns, the main take-home message is that the fitness landscape of passion vine butterfly coloration is not simple. It has multiple peaks, each corresponding to the warning color that dominates in a different region of the Amazon basin.31

  When Sewall Wright conceived the fitness landscape, he did not have ammonites or butterflies in mind; he was thinking about his breeding experiments and the complex gene interactions they revealed. Wright’s math showed that such interactions could bring forth fitness landscapes with many more than two or even a dozen peaks. More than that, he realized that their topography could be so complex that it defies imagination.

  To see where Wright was coming from, let’s revisit the peppered moth. While its wings can display many shades of gray, it turns out that a population of moths would mostly consist of two types: a light one referred to as typica and a dark one referred to as carbonaria.32 In genetics jargon, these moths have two different phenotypes—a term that refers to any observable feature of an organism—and these phenotypes are encoded by two different genotypes, the DNA that is responsible for their appearance. The two genotypes are two different alleles of the same gene, which can be inherited in the indivisible, atom-like fashion that Gregor Mendel first discovered when he crossed pea plants in his monastery garden.33 And because a moth’s wing is basically either light or dark, one can replace the continuous light–dark axis of the one-dimensional landscape from Figure 1.2 with something simpler—the two points at the ends of the light–dark continuum, as shown in Figure 1.5a. Each of these points has a different value of fitness that describes how well a moth of that color can survive and reproduce. (The figure does not show that value.)

  If it were only wing color that mattered to a moth’s survival, the story would end here. But other traits also contribute, and that’s where the complications begin. Wing size is one of these traits, and we know that mutations in some genes can alter it.

  Figure 1.5.

  Moths with one allele of such a gene would have normal large wings, whereas moths with the other, mutant allele would have smaller wings. Smaller wings reduce lift and impair flight and thus decrease fitness. Together with the two wing-color alleles, the two wing-size alleles can form four possible genotypes, which can be visualized as the four corners of the square shown in Figure 1.5b.

  It gets still more complicated. Now consider a third gene, this one influencing the size of the moth’s antennae. These marvelous sensory organs allow a male to home in on a female that is miles away and to follow a faint scent trail of just a handful of female pheromone molecules per cubic meter. Moths with one variant of this antenna gene have normal antennae, whereas moths with another have smaller antennae that are less sensitive, so they might get lost when tracking a female. Needless to say, getting lost while searching for your mate is not great for your ability to reproduce, another important aspect of fitness. With the addition of these alleles that encode antenna size we have eight possible genotypes: two for antenna size, two for wing color, and two for wing size. They can be placed on the corners of the cube in Figure 1.5c, where the paired, leaf-like objects stand for the antenna. (Just like Figures 1.5a and 1.5b, this figure does not show the genotypes’ fitness values.)

  Other genes affect further traits, such as the acuity of vision and the ability to endure starvation,
evade attackers, or extract energy from nectar. With each new trait and allele pair that we add, the number of genotypes doubles. For one, two, and three traits, we were able to write the possible genotypes as the end points of a line, the corners of a square, and the vertices of a cube—objects in one, two, and three dimensions. But for four traits and their sixteen possible genotypes, we would need an object like a cube—but in four dimensions. Mathematicians call such high-dimensional cubes hypercubes. We can’t visualize them well, but mathematics can describe them because their geometry follows straightforward laws. For example, the number of a hypercube’s vertices doubles with every added dimension. A four-dimensional hypercube has sixteen vertices, a five-dimensional hypercube has thirty-two, a six-dimensional has sixty-four, and so on.

  Despite the leading role that moths played in early evolutionary biology, they were soon surpassed by the tiny fruit fly Drosophila melanogaster. Geneticists cherish the fruit fly for several reasons: It is small, so one can easily keep thousands of flies in the lab. It’s not a picky eater—a bit of yeast, cornmeal, or sugar, and happiness ensues. It reproduces very quickly. And despite its small size it has many traits that one can study with just a low-powered microscope, such as the shape of its wings, the color of its eyes, or the size of its antennae.

  These advantages allowed geneticists like Thomas Hunt Morgan to scour thousands of fruit flies for mutant genes. Starting in 1908, Morgan toiled for two years before his first big break, when he discovered an allele he christened white because it turns the usually bright-red eyes of flies white.34 After this first mutant allele was discovered, others quickly followed over the next years. They modified not only eye color, but all sorts of traits, including the size and shape of wings and body, the structure of important sensory organs like eyes, antennae, and bristles, as well as key traits like fertility and life expectancy.

  By the time Wright proposed the fitness landscape concept in 1932, fruit fly experiments had already identified mutations in four hundred different fruit fly genes.35 Even if each of these four hundred genes had only two alleles, there would be 2400—or 10120—possible genotypes, each with a fitness value potentially different from every other genotype. That’s a very large number, much larger than the comparatively puny number of 1090 hydrogen atoms in the universe. As with the moths in Figure 1.5, each of these genotypes can still be placed on one vertex of a cube, albeit a four-hundred-dimensional one, with as many vertices as there are genotypes. The resulting landscape looks nothing like a familiar three-dimensional mountain range. Instead, each vertex of the cube corresponds to one “location” in the landscape—a fly with a specific allele combination—and its fitness is the “altitude” at this location.

  This abstraction, far removed from our daily experience of a landscape, is what Wright had in mind when he introduced the landscape idea. But, like the rest of us, he could not visualize it with limited three-dimensional geometry. So he did what most of us do when faced with complexity far beyond our mind’s grasp: he ignored it. He continued to talk about fitness landscapes as if they were three-dimensional and showed the kinds of peaks and valleys we are familiar with. And who can blame him? All our intuition about geometry comes from the three-dimensional world we live in. It may not apply to higher dimensions, but it is all we have.

  Despite their limitations, even highly simplified landscapes and their peaks can be enormously valuable. Their topography can hold clues to how innovations emerge in biological evolution and how its creative process produced well-camouflaged moths, efficiently swimming ammonites, and gaudy poisonous butterflies. What’s more, we shall see later that these landscapes are just as useful for understanding other forms of creativity. And even where the three-dimensional landscape metaphor fails—especially where it fails—it can teach us important lessons about creativity.

  The complexity of evolution’s landscapes also makes another point: when geneticists like Morgan and Wright first glimpsed life’s genetic complexity, they got more than they had bargained for.

  But, as it turns out, they had seen nothing yet.

  Chapter 2

  The Molecular Revolution

  Morgan and his research associates—also known as the fly boys—discovered much more than the white gene. They also discovered that genes are located on chromosomes, a discovery that earned Morgan the 1933 Nobel Prize. And Morgan invented genetic mapping, which allowed scientists to locate genes like white on each of the five fruit fly chromosomes. Morgan’s work still resonated half a century later, when his ideas helped locate in the human genome those genes involved in diseases like breast cancer. But an even bigger prize eluded him: to understand how different alleles of a gene cause different phenotypes to appear. That discovery had to wait for a molecular revolution in biology, which Morgan’s work had prepared, but which would not get going until decades later.

  It began when Oswald Avery showed in 1944 that DNA extracted from the corpse of a pneumonia bacterium can transform other, harmless bacteria into dangerous killers, as dangerous as live pneumonia bacteria. And it continued when, in 1953, James Watson and Francis Crick first elucidated the chemical structure of genotypes when they discovered DNA’s double-stranded spiral staircase.1 Each strand of this celebrated DNA double helix is built from four different nucleotide building blocks, distinguished by the four bases adenine, guanine, cytosine, and thymine and abbreviated by the letters A, C, G, and T that together form DNA’s molecular alphabet. A molecule with this structure is an ideal information carrier because different sequences of the four letters, like different texts in the English language, can encode different information—the kind that parents pass on to their offspring.

  When a cell decodes the information in a gene’s DNA, it first transcribes the DNA’s letter sequence into an RNA, or ribonucleic acid, copy. This RNA molecule is usually a mere intermediary. Its role is to be translated into a protein’s letter sequence of amino acids. Once created, this amino acid string is incessantly jostled by the endless vibrations—also known as heat—of nearby molecules that collide with it. The energy in these collisions helps a protein fold into an intricate three-dimensional shape that biochemists call a conformation or a fold. The folded protein also vibrates with heat, and these vibrations allow proteins to perform myriad useful jobs. Protein enzymes catalyze thousands of different chemical reactions that take place in organisms on this planet, each enabled by an enzyme’s unique three-dimensional shape. Proteins import hundreds of different nutrients into cells and help excrete just as many kinds of waste molecules. Proteins stiffen the molecular skeleton that prevents our cells from collapsing into amorphous blobs, and that makes brain cells visibly different from liver cells. Protein hormones keep our body working, like the insulin that controls our blood sugar, the prolactin that enables milk production, or various pain-reducing endorphins.2 And proteins keep life on the move by spinning bacterial flagellae—themselves made of protein—and by contracting mammalian muscles. Without these workhorse molecules, life might never have crawled out of the primordial soup. All these proteins are encoded in the genes of the organisms that produce them. We humans have more than twenty thousand genes, an organism like a fruit fly has some fifteen thousand genes, and simpler organisms like the bacterium Escherichia coli still have a few thousand.3

  Each of these genes can mutate anywhere along its DNA letter sequence. Such mutations occur when high-energy particles or atoms smash into DNA, when destructive by-products of metabolism react with DNA, or when DNA replication enzymes—another crucial kind of protein—commit errors while copying DNA. These processes create different kinds of mutations. One of them—also called a point mutation—is especially frequent and alters only a single letter in a gene. It is easy to calculate the total number of alleles that such molecular typos can create. For a gene with some one thousand nucleotide letters—not unusually long—the first letter must be one of the four possibilities: A, C, G, or T. Whichever it is, let’s say a C, the letter can change into
any of the three other letters (A, G, or T), so there are three mutant texts that can be created by altering the first letter. The same argument holds for the second of the thousand letters—it can change into three others—for the third letter, and so on, all the way through the one-thousandth letter. All of these possibilities add up to 3,000 new alleles that a single DNA typo can create. This number is even greater for longer genes, and for those mutations that alter more than one letter at a time.

  All of this means that biologists today must reckon with landscapes of staggering complexity, much greater than Wright’s landscapes with a few genes and a handful of alleles. If we just consider all the variants of a Drosophila genome that can be produced by a single typo in one of Drosophila’s fifteen thousand genes, the resulting landscape has a whopping 300015000 possible genotypes.4 That’s a one with more than twelve thousand trailing zeroes, long enough to fill several pages of this book. All these genotypes can still be arranged on the corners of a high-dimensional cube, but the number of vertices on this cube is much greater than the number on the cube that Wright envisioned. Wright’s hypercube already had more vertices than there are atoms in the universe. If each atom in our universe harbored another universe, and if each of these universes harbored as many atoms again as exist in our universe, the total number of atoms in all of these universes would still be dwarfed by the number of possible fruit fly genotypes.

  The molecular revolution also clarified how exactly mutations change genotypes and phenotypes. A mutation that changes a single letter in a gene displaces a genotype on the vast hypercube of all DNA sequences, from one vertex to an adjacent vertex, and such changes often alter the encoded protein, which transforms the phenotype. The white gene, for example, affects the eye color of fruit flies not because it encodes an eye color pigment, but because it encodes a transporter protein that delivers the molecular building blocks of such eye pigments. Mutations in this gene cause white eyes because they cripple the transporter, such that building blocks never reach their destination in the eye.5