IT STRUCK me in an examination room. As my colleagues questioned a graduate student about genetics, my mind began to wander between the genetic code and information theory. And that’s when it hit, like a strong wind slamming open an unseen door. A single amino acid can be encoded in as many as six different ways. I suddenly realised this means that DNA has the flexibility to carry multiple overlapping messages. Sure, genomes contain more than just genes: they also hold instructions about where in our body and when in our lives to make each protein. But what if our DNA also contained information that made mutation more likely in some parts of the genome and less likely in others? Such a genome would have the potential to influence its own evolution, protecting essential DNA sequences in some places while elsewhere unleashing genetic variations that could explore evolutionary possibilities.
There was just one problem. If I was correct, it would mean that not all genetic mutations are random. Such a notion was as heretical two decades ago when I first had this idea as it is today. Nevertheless, in the intervening years, breathtaking advances in genetics have revealed that less than 5 per cent of our genome codes for the amino acids in our proteins, providing even more room for extra information than I imagined. What’s more, there is clear evidence from organisms as diverse as humans and bacteria that genomes do indeed contain information that can focus mutations in certain areas and direct it away from others.
Yet students are still taught that evolution works through completely random genetic variation acted upon by natural selection, leading to the survival of those organisms with genes best suited to their environment. Surely it is time to rethink the idea that evolution is purely a game of chance: to accept that genomes could have evolved information that allows them to influence genetic change and affect their own chances of survival?
Advertisement
Although this may sound like heresy, it does not challenge Darwin’s evolutionary theory. Instead it updates it, by recognising that natural selection acts not only on fins and wings, but also on the very mechanisms that copy and repair DNA. It makes perfect sense that these mechanisms would be subject to selection pressures: genomes that managed to steer their mutations, even a little, would have pulled way ahead of others that were throwing darts in the dark. Such genomes would have a better chance of producing fit offspring, so they would tend to become more widespread than genomes that relied completely on random change.
Accepting all this has radical implications. It should make us wary of thinking that just because we can translate sequences of DNA bases into amino acids, we understand all the information our genomes contain. The realisation that genomes may contain more information than we imagined should inform our decisions about genetic engineering and medicine. And at a philosophical level, if genomes can learn about the world through natural selection, perhaps we should even consider them to be intelligent.
As our ability to examine entire genomes allows us to peer under the hood of evolution, it is increasingly obvious that the long evolutionary journey has not been just a series of random stumbles. Certain types of DNA sequence seem to invite mutation and others seem to repel it. One type of mutation-attracting sequence, found in bacteria as well as us, is “slippery” DNA. Slippery sequences comprise a series of DNA’s chemical bases or “letters” repeated over and over again. The repeating unit may be simply a single DNA base – adenine, guanine, cytosine or thymine – or a short string of bases such as CAG. The machinery that makes copies of DNA for the organism’s offspring can lose register in such repetitive sequences, copying a unit of the repeat more than once, or skipping over it. TTTT may become TTTTT or CAGCAGCAG become CAGCAG. These deletions and additions generate a great deal of genetic diversity.
The mutation rate of slippery DNA can be 1000 times higher than elsewhere in the genome, so in that sense it is not random. But the important question as far as evolution is concerned is whether mutation is random with respect to its effect on survival. Are slippery sequences just wet stones scattered randomly across the landscape, or are they some kind of guidepost?
If a slippery sequence occurs in a place in the genome where it makes no difference to the organism’s fitness, natural selection will neither favour nor remove it. If it emerges in a place where any mistake is deadly, organisms with that poorly placed sequence will leave fewer descendants. But if a slippery sequence appears in a spot where it actually provides an advantage, scattering progeny this way and that around dangerous barriers to survival, natural selection will favour it. Generations into the future, a greater proportion of genomes will have slippery DNA at that spot. In a sense, the DNA would have “learned” through natural selection that there is an advantage to having a high mutation rate at that spot.
And that is indeed what happens: certain slippery sequences do boost survival. Richard Moxon and his colleagues at the University of Oxford point to the important role of such sequences in DNA that determines bacterial surface proteins (The Journal of Clinical Investigation, vol 107, p 657). Pathogens with “coats” that frequently change thanks to genetic slips have several advantages. Each generation’s diverse proteins mean the bacteria can stick to a variety of different tissues within their host, exploring new places to grow. What’s more, bacteria that keep changing their coat remain one step ahead of the host’s immune system, which is constantly on the alert for the bacteria in its former guise. Many bacteria owe these survival skills to the high mutation rate of slippery DNA.
We cannot yet be sure whether the slippery sequences in our own genome are similarly well placed to give us an evolutionary advantage. But we do know that the human genome has several other ways of directing mutation to occur in spots where variation is crucial for survival. Just as the bacteria change their coats to survive, our immune systems need to keep pace with the change to stay on top of such invaders. In the evolutionary arms race between host organisms and their pathogens both parties have to run just to stand still, so human genomes that evolved the ability to increase the mutation rates of antibody genes would have a survival advantage. But there is a catch.
Our antibody genes must satisfy two seemingly conflicting requirements. On the one hand, the body’s array of antibodies must be as diverse as possible, so that at least one of them will be able to latch onto any given pathogen, even a pathogen the immune system hasn’t seen before. On the other, the antibodies must have a reliable, stable, pathogen-disposal plan. If antibody genes achieved diversity through random mutation, mutations would be as likely to destroy the ability to dispose of pathogens as to gain a grip on new ones. Evolution’s ingenious solution is that we inherit our antibody genes in pieces.
One stretch of our DNA encodes variable regions: a large palette of possible pathogen-binding segments. Another region encodes a few well-conserved pathogen-disposal mechanisms. In each cell destined to become an antibody factory, a piece of DNA encoding one of the wide selection of variable regions is moved into place next to the conserved DNA that tells the antibody how to dispose of pathogens.
Once the variable pathogen-binding piece is in place, it is subjected to biochemical mechanisms that focus additional variation in precisely those places in its structure that determine which pathogen it can grab. How this happens is not entirely clear, but Michael Neuberger and his team at the University of Cambridge have suggested that an enzyme attracted to this location cuts off part of the base cytosine. Such damage attracts repair enzymes but, in this case, they don’t reverse the damage, and instead leave mutations.
However the biochemical mechanisms work, the upshot is that our immune system has evolved the ability to direct mutation to exactly where it is needed, while protecting stretches of DNA that are best conserved. But antibody genes are not the only genes that face the challenge of conflicting requirements of variation and conservation. Might genomes have evolved the ability to focus mutation and cut and paste specific sections of DNA in other genes too? Colour vision is a case in point.
To come up with a protein that detects red light, evolution doesn’t need to start from scratch. It can make a copy of the pre-existing genes for the protein that detects green, and then tinker. Indeed, many new genes evolve through mutations that modify copies of working genes. Geneticists have assumed that these mutations are completely random, but successful adaptation repeatedly has the same requirement: to vary certain patches of DNA while conserving others. If this kind of tinkering works best, then surely natural selection will favour genomes that tend to steer mutation towards those parts of the genome where it is likely to increase fitness and away from those where genetic change would be damaging.
Now that we can compare entire genomes, scientists can study the exploratory cutting and pasting that occurs during the formation of sperm and egg cells, and they have found that it certainly does not look completely random. Last year, Pavel Pevzner and Glenn Tesler from the University of California, San Diego, used a computational comparison of the human and mouse genomes to estimate that in humans there are about 400 of these cut-and-paste “fault zones”, representing 5 per cent of the genome. And just a few months later, a team led by Jim Kent and David Haussler from the University of California, Santa Cruz, confirmed that DNA is indeed more likely to be cut and pasted at some spots than at others, and they identified some of these hotspots. It is not yet clear why they are prone to genetic upheaval, but it’s possible they are particularly accessible or susceptible to the enzyme SPO11, which initiates recombination of chromosomes when they are being repackaged for the next generation.
While slippery genes and genetic cutting and pasting can lead to novelty, in contrast, the human Y chromosome has revealed a way of undoing mutations. Until it was sequenced last year, most researchers had assumed that this partnerless little chromosome was doomed to decline, gradually eroded by the damaging effects of random mutation. With only one copy of the Y in a typical male genome, if repairs need to be made there is less back-up of information than in the other paired chromosomes. But some researchers believe the Y may carry its own back-up copy in the form of information provided by “palindromic” DNA sequences, which are analogous to words that read the same backwards or forwards (Nature, vol 423, p 873).
The two long strands that make up each double helix of DNA are held together by pairing between opposing bases on each strand: A pairs with T and G pairs with C. So a strand that begins with, say, AGTTCC and ends in GGAACT can fold back onto itself, creating a little hairpin in which each base is paired with its complement. If a mutation changed the first G to an A, for example, the cell’s molecular machinery would notice that the bases no longer matched up in the hairpin, and could restore the A to a G.
This clever trick is not confined to humans. The tendency of certain DNA sequences to form self-correcting palindromic hairpins was first identified in the 1980s by Lynn Ripley, now at New Jersey Medical School, in her work with a lunar-lander-shaped virus called T-4 that infects bacteria: a long evolutionary distance from humans.
From strategically placed slippery sequences and focused variation in antibody genes to palindromic DNA, the genome does seem to have evolved ways to affect its own fate. Sitting in that examination room, two decades ago, I realised that because almost all amino acids are encoded by more than one sequence of DNA bases, information that focuses mutations could be embedded in the choice of how a protein’s amino acid sequence is encoded.
One way of encoding a string of amino acids may have a high intrinsic rate of change because it generates slippery sequences, while another way of encoding the same amino acids may form part of a self-correcting palindrome (see Table). For example, the amino acid glycine can be encoded as one of four different triplets, including GGG. And the amino acid pair glycine-glycine can be encoded in 16 different ways, one of which – GGGGGG – looks more slippery than the others. Thus a more stable or more mutation-prone sequence may arise in a certain spot purely by chance. But then natural selection will act, retaining the sequence if it provides a repeated advantage, or culling those genomes where the rate of variation at that spot works against it.
Genomes face certain kinds of threats to their survival every time they pass from generation to generation. Through natural selection, they “learn” about the world by exploring and surviving. If natural selection is a teacher, and genomes can learn, should we think of them as intelligent? This question has important implications not only for evolutionary theory, but also for genetic medicine. Until we understand the genome’s multiple, interlinked functions, we cannot fix “errors” and assume that we know what effect that will have. At the very least we should consider the possibility that a supposedly faulty gene may in fact have become widespread because it protects people, either by blocking a renegade gene somewhere else in the genome, or by providing its own advantage under certain circumstances – just as sickle cell trait does against malaria.
We also know that some genes that look very broken to us have saved people’s lives. For example, there is a gene that encodes a protein used by HIV as a doorway to get inside our cells. In some people, this gene is “broken”, making them resistant to AIDS. If we had “fixed” those doorways before HIV took hold, our genome would never have shown us the life-saving potential of drugs that could block the opening of this door. Luckily we did not have the ability to purge this gene from our collective genome before HIV struck. The first door-blocking drug has just been approved for treating advanced cases of AIDS.
This should make us pause for reflection. We should approach our genome with humility – indeed, with awe – not as engineers and doctors but rather as students, for after billions of years of evolution it surely has much to teach us about survival. And as we look closely, will we come to consider genomes to be in some way intelligent? Perhaps so. After all, what makes us so smart?