Natural History, September 2001
More knowledge doesn't necessarily translate into less confusion. In 1953, when James Watson and Francis Crick discovered the double-helix structure of DNA, biologists were poised for all the pieces of evolutionary theory to fall quickly into place. Darwin had shown how natural selection could transform anatomy and patterns of behavior, and now scientists looked forward to detecting the fingerprints of natural selection at the molecular level. But it turned out that natural selection was not the only force that could significantly alter DNA, and no one knew how to identify the source of any given change. It has taken decades to figure out a way.
This predicament took many biologists by surprise, because once the structure of DNA was understood, they thought they had a good handle on how natural selection works, Researchers determined that a gene consists of a stretch of DNA--a sequence of molecules called nucleotides that serves as a blueprint for the synthesis of a protein. If the nucleotide sequence somehow mutates, it may produce a different protein. The new protein may turn out to be so defective that the individuals who inherit the mutation die or become infertile; either way, such harmful mutations may end up being almost completely weeded out of a species. But some mutations may give their owners a reproductive edge, and before long those mutations will become widespread--even universal--in a species. In the course of thousands and millions of years, a series of such beneficial changes could thus rework the genome of a species.
By the 1960s, however, researchers had realized that this was far from the whole story of evolution. Richard Lewontin, of Harvard University, and his colleagues collected fruit fly proteins and tallied the many variant forms of each. (The technology for sequencing the actual nucleotides was still two decades away.) They studied a selection of proteins in a single population of flies and found that fully 30 percent of the molecules existed in more than one form. If natural selection was weeding out mutations, then it was an awfully careless gardener.
Lewontin's research, and similar results from other scientists, led the Japanese biologist Motoo Kimura to launch a full-blown assault on the standard picture of how nucleotides evolve. Yes, he argued, some mutations were harmful and some were advantageous. But between those two extremes was a huge range of mutations that had no effect at all. Over the generations, these neutral mutations, as they came to be known, might become more common solely as a result of random fluctuations (a process called neutral evolution). Kimura predicted that most variation in DNA would turn out to be produced by neutral evolution and not by natural selection.
The irony was inescapable: scientists finally had a chance to tune in to evolution on its most basic level, but the signal of natural selection seemed to be swamped by the static of neutral evolution. If your radio can't pick up a clear signal, however, maybe what you need to do is build a better radio. And biologists set out to do just that. By the 1970s, researchers had realized that there was a way--at least in theory--to isolate the signal of natural selection, thanks to the way cells build proteins.
A cell makes a copy of a gene's sequence and then ships this code out to a protein-building factory called a ribosome. There the genetic code acts as a guide for building a protein out of a set of molecules called amino acids. But a gene's nucleotides don't correspond to the sequence of amino acids in the protein in a one-to-one fashion. Before adding an amino acid to the end of a protein under construction, ribosomes have to read three nucleotides in a row. And in many cases, different trios of nucleotides can direct a ribosome to add the same amino acid. For example, if the nucleotide guanine appears three times in a row, a ribosome adds the amino acid glycine to the protein it is building. But it will add glycine even if the genetic sequence reads "guanine-guanine-thymine."
So imagine that a gene whose sequence includes three guanines mutates, with the third guanine turning into thymine. Even though the gene has changed, it still calls for the same old glycine. The new protein will not just be similar to the old one; it will be identical. Geneticists call these kinds of mutations "silent substitutions," as opposed to "replacement substitutions," which actually change the structure of a protein.
A comparison of the numbers of silent and replacement substitutions acquired by a gene should reveal whether it has experienced natural selection. Genes mutate randomly--both at sites where silent mutations occur and at replacement sites. If they are evolving neutrally, both kinds of mutations are equally likely to become universal. In other words, if 1 percent of the silent sites change in a million years, so will 1 percent of the replacement sites. But natural selection favors certain replacement mutations and makes them spread faster than neutral evolution can. Mutations that change a protein to some superior version will accumulate more quickly than will silent substitutions. If, say, 1 percent of the silent sites in a gene change in 10 million years, natural selection might in fact change 5 percent of the replacement sites.
In theory, then, it should be simple to check a gene for natural selection. All you'd have to do is tally the fraction of replacement sites that have mutated and compare that with the fraction of silent sites. If the ratio is one to one, the gene has evolved neutrally. If the ratio is higher, natural selection has been at work. And the higher the ratio, the more intense the natural selection.
While simple in concept, this evolutionary "radio" was actually staggeringly hard to build. For one thing, researchers can't go back millions of years to read a gene's ancestral sequence, nor can they know the precise history of mutations that led up to its current form. But biologists can make some inferences by comparing the genes of closely related animals. Imagine, for example, that a certain gene for red blood cells is identical in every ape but that the human version of the gene differs at the 123rd position. It would be a safe bet that the human gene started off like the others and later mutated at position 123. But the evidence from real genes is rarely so clean, and thus some uncertainty inevitably creeps in.
Making matters fuzzier, until only a few years ago biologists had relatively few genetic sequences available to analyze. Like a political poll of your four best friends, studies of these genes were usually not statistically significant. Another problem was that the models of evolution used by biologists for detecting natural selection assumed that mutations were random, which is not entirely true. For example, the chemical nature of DNA makes certain nucleotides more likely to mutate into certain other ones. With so much uncertainty in their methods, biologists attempting to tune in to natural selection couldn't be sure they weren't getting a false signal.
Not until the mid-1990s were researchers able to sidestep these uncertainties. By then, genetic data banks had begun to swell with sequences, and computers were becoming outrageously fist. Taking advantage of both trends, researchers began analyzing DNA using powerful statistical equations. One of the most widely used methods is called maximum likelihood. Although there's some head-spinning math behind the procedure, it basically works this way: First, pick a group of related species you want to study. Then compare various chunks of their DNA to determine their relationships and build their evolutionary tree. Now choose a gene whose history you want to track along the branches of the tree.
Ziheng Yang, of University College London, and his colleagues have led the way in this approach. They program computers to look at every possible series of mutations that could have yielded the existing variants of a gene from a common ancestor. By sorting through the billions of possible scenarios and calculating the effects of different degrees of natural selection, the computer finds the level of natural selection that offers the maximum likelihood of having produced the actual gene sequences.
Other researchers are using methods such as Yang's to analyze hundreds of genes from numerous species, including humans. Many of these genes indeed turn out to be evolving neutrally, just as Kimura predicted they would. But some glow with natural selection's fire. The genes for some of the immune-system proteins that recognize invading pathogens have been adapting particularly quickly. Known as the major histocompatibility complex (MHC), these proteins are found within each cell, and their job is to lock on to fragments of viruses and other parasites within the cell and bring them to the cell's surface. Special immune-system cells passing by can recognize the fragments so displayed and force the infected cell to self-destruct. It turns out that some genes that build MHC proteins have acquired four times more replacement mutations than silent ones.
Scientists can tune in even more precisely to natural selection--determining not just whether a gene has been affected by it but which of its individual nucleotides has. Using the maximum likelihood method, for instance, Yang examines each site in a gene and tries to predict its nucleotide on the basis of the rest of the gene's sequence as well as the sequence of the gene in related species. The computer makes a series of predictions based on different assumptions about the level of natural selection acting on the site and then picks the level that works best.
Apparently, only a few nucleotides in MHC genes have been strongly affected by natural selection. If you look at their positions, they seem to be scattered randomly here and there within the gene. But the hidden order becomes apparent once you take account of the actual shape of MHC proteins. All the amino acids altered by natural selection are found at a cleft in the proteins--the place where they actually make contact with the pathogens.
Researchers had anticipated exactly these results, basing their prediction on the struggle between hosts and parasites. If a pathogen evolves a new structure, an old MHC protein may be less capable of grabbing onto it, making it harder for the host to fight off the disease. As a result, natural selection should favor animals with new MHC proteins that offer a better grip. Not surprisingly, researchers have also found signs of intense natural selection in the genes that construct the surface proteins of bacteria and viruses. The changes in their structure enable the pathogens to evade recognition by the immune system. Through studying these hot spots of adaptation in disease genes, scientists may be able to make new vaccines to keep up with their evolution.
When scientists began to tune in to natural selection, they also expected to hear a loud signal coming from the genes involved in sex. Biologists knew from studies on a wide range of animals that males compete with other males to fertilize eggs and that females choose potential mates based on the males' appearance and, perhaps, even on their sperm. Last year Gerald Wyckoff, Wen Wang, and Chung-I Wu, of the University of Chicago, discovered that a few genes that help build sperm in humans were undergoing rapid selection. And last February, Willie Swanson and his colleagues at Cornell University teamed up with Ziheng Yang; together they detected equally strong selection at work on genes that build the proteins on the surface of human egg cells. As was the case with the MHC, they found that natural selection was at work on only a few nucleotides--specifically, the ones that build part of the protein that makes contact with the head of a fertilizing sperm.
These biologists have only pointed to the regions where natural selection is taking place; they don't yet know exactly what adaptive pressures are driving it. One possibility is that receptors on the eggs enable them to choose certain sperm for fertilization. As male genes for the surface of sperm adapt to these receptors, the female genes for eggs evolve into new forms.
When we think of what changes in our ancestors gave rise to Homo sapiens, the genetics of fertilization and head colds may not pop to the top of the list. The only reason that genes involved in sex and sickness were the first to reveal evidence of natural selection is simply that biologists turned to them first. Researchers are now looking for evidence of natural selection in other kinds of genes. One of the most surprising results of this search came in August 2000 from Gavin Huttley, a geneticist at Australian National University. Huttley has discovered that a gene called BRCA1 has experienced strong natural selection over the past few million years.
BRCA1 has been in the news in recent years because it sometimes mutates into a form that has been linked to several types of cancer, such as breast and prostate cancers. Normally, the gene plays vital roles in the development of embryos and the repair of damaged DNA. Huttley compared the BRCA1 gene in humans with those in other apes and discovered that it experienced intense natural selection in our ancestors. It's possible, he concludes, that the cancers BRCA1 can cause when it mutates are a price we pay for having become human.
Other genes may have evolved in important ways, too, but natural selection may have affected only a few key mutations, making them hard to detect. Wyckoff, Wu, and their University of Chicago colleague Justin Fay predict that many such genes will be discovered. In the July 2001 issue of Genetics, they reported on a survey they had made of several hundred genes sequenced during the Human Genome Project. Rather than analyzing individual genes, they pooled all of them together and compared them with the same sequences in other primates. By analyzing so much data at once, they hoped to pick up faint signals of natural selection that they might otherwise have missed in any single gene. The scientists were not disappointed. In the 30 million years since our ancestors split from those of the Old World monkeys, more than a third of the changes in our amino acids have been the result of natural selection. Wyckoff, Wu, and Fay calculate that, on average, one of these adaptive changes happened every eighty years or so.
Natural selection is widespread in the human genome; now biologists just have to figure out exactly which genes have experienced it. Not surprisingly, researchers such as Wyckoff and his coworkers (notably graduate student Steve Dorns, who works at the laboratory of Bruce Lahn, of the University of Chicago and Howard Hughes Medical Institute) are examining the genes that help build and operate the human brain. After all, the human brain is in some ways very different from those of apes. The most obvious difference is size: relative to body weight, a human brain is twice as big as a chimp brain.
The initial results are telling--but not what you might expect. In the few brain-related genes that scientists have closely studied in primates, silent substitutions far outnumber replacement substitutions. This sort of ratio is the sign of a third kind of evolution, known as purifying selection. Some particularly intricate parts of the body cannot tolerate even a little tinkering with their structure, so any change to the proteins involved gets ruthlessly weeded out. The primate brain may be one such delicately constructed organ.
Ancestral humans somehow overcame this resistance to change and acquired a few precious mutations that made them (and us) particularly clever. But finding a way to detect those crucial changes may prove to be one of our greatest mental challenges.
Copyright 2001 Carl Zimmer