4
Darwinism 2.0
‘Nothing in biology makes sense except in the light of evolution.’
– Theodosius Dobzhansky1
‘Nature, red in tooth and claw.’ Alfred Tennyson penned these evocative words at the dawn of the Darwinian age. Understandably, scientists and poets of the day were wont to dwell on the brutality of natural selection as manifested in the arms races of bodily adaptations, be they the razor-sharp teeth of the shark or the tough defensive shell of the tortoise. It is easy to understand how evolution may select for bigger wings, longer legs, keener vision, and so on in the relentless struggle for survival. But bodies – the hardware of life – are only half the story. Just as important – indeed, more important – are the shifting patterns of information, the command-and-control systems, which constitute the software of life. Evolution operates on biological software just as it does on hardware; we don’t readily notice it because information is invisible. Nor do we notice the minuscule demons that shunt and process all this information, but their near-thermodynamic perfection is a result of billions of years of evolutionary refinement.2
There is an analogy here with the computer industry. Thirty years ago personal computers were clunky and cumbersome. Innovations like the mouse, the colour screen and compact batteries have made computers far more efficient and convenient, as a result of which sales have soared. The capitalist version of natural selection has thus led to a vast growth in the population numbers of computers. But alongside the hardware innovations there has been an even more impressive advance in computer software. Early versions of Photoshop or PowerPoint, for example, are a pale shadow of those currently available. Above all, the speed of computers has increased vastly, while the cost has plummeted. And software improvements have contributed at least as much as hardware embellishments to the success of the product.
It took a century following the publication of Darwin’s theory for the informational story of life to enter the evolutionary narrative. The field of bioinformatics is now a vast and sprawling industry, accumulating staggering amounts of data and riding high on hyperbole. The publication in 2003 of the first complete human genome sequence, following a mammoth international effort, was hailed as a game-changer for biology in general and medical science in particular. Although the importance of this landmark achievement should not be diminished, it soon became clear that having complete details of a genome falls far short of ‘explaining life’.
When Darwin’s theory of evolution was combined with genetics and molecular biology in the mid-twentieth century, in what is termed the ‘modern synthesis’, the story seemed deceptively simple. DNA is a physical object; copying it is bound to incur random errors, providing a mechanism for genetic variation on which natural selection can act. Make a list of the genes and the function of the proteins they code for and the rest will be mere details.
About twenty years ago this simplistic view of evolution began to crumble. The road from an inventory of proteins to functional three-dimensional anatomy is a long one and the protein ‘parts list’ provided by the genome project is useless without the ‘assembly instructions’. Even today, in the absence of foreknowledge, nobody can predict from a genomic sequence what the actual organism might look like, let alone how random changes in the genome sequence would translate into changes in phenotype.
Genes make a difference only when they are expressed (that is, switched on), and it is here, in the field of gene control and management, that the real bioinformatics story begins. This emerging subject is known as epigenetics, and it is far richer and more subtle than genetics in isolation. More and more epigenetic factors which drive the organization of biological information patterns and flows are being discovered. The refinement and extension of Darwinism that is now emerging – what I am calling Darwinism 2.0 – is yielding an entirely new perspective on the power of information in biology, ushering in a major revision of the theory of evolution.
ELECTRIC MONSTERS
‘There is more to heredity than genes.’
– Eva Jablonka3
‘It came from space! Two-headed flatworm stuns scientists.’4 So proclaimed a British online publication in June 2017. The subject of the article, which inevitably involved ‘baffled boffins’, was the appearance of monsters in the International Space Station. The monsters didn’t invade the station; they emerged as part of an experiment to see how lowly flatworms got on in orbit if their heads and tails were chopped off in advance. It turns out they got on very well. One in fifteen came home with two heads in place of the one they had lost.5
The space worms are just one rather dramatic example of the exploding field of epigenetics. Loosely defined, epigenetics is the study of all those factors which determine the forms of organisms that lie beyond their genes (see Box 9). The two-headed worms are genetically identical to their more familiar cousins, but they look like a different species. Indeed, two-headed worms reproduce and beget more two-headed worms. No wonder the boffins were baffled. In this case the chief boffin was Michael Levin of Tufts University, who happens to be a collaborator with our research group at ASU.
To put the worm work in context, recall from the previous chapter how the development of the embryo (morphogenesis) provides a graphic example of the power of information to control and shape the form of the organism, although many of the actual mechanisms at work remain puzzling. I explained that the information needed to build and operate an organism lies to a great extent in the system’s ability to switch genes on or off and to modify proteins after the genetic instructions are translated. The regulation of information flow via chemical pathways – involving molecules such as the methyl group, histone tails and micro-RNA (see Box 9) – and the coupling of this gene-switching repertoire to a plethora of shifting chemical patterns is as yet dimly understood. Epigenetics thus opens up a vast universe of combinations and possibilities. I mentioned how the diffusion of specialized molecules called morphogens play an important role in controlling the unfolding dynamics of development, but that turns out to be only part of the story. In the last few years it has become clear that another physical mechanism could be even more significant in morphogenesis. Known as electro-transduction, it deals with changes to an organism’s form arising from electrical effects.
Box 9: Beyond the gene
Genes are switched on and off as needed through the life of an organism. There are many ways a gene can be silenced. A common method is methylation, in which a small molecule of the methyl group becomes attached to the letters C in a gene and physically blocks the gene read-out mechanism. Another is RNAi, a tiny RNA fragment just twenty-something letters long discovered serendipitously by botanists who were trying to make prettier flowers. In this mechanism, the gene is read out from DNA as usual, but RNAi (i stands for interference) mugs the messenger RNA while it is busy taking its data read-out to the ribosome and chops it in two, thus (somewhat brutally) junking the message. In complex organisms, genes can be smothered by being buried in highly compacted regions of chromatin.
In addition to gene switching, several other variables are at play. An expressed gene may produce a protein that subsequently becomes modified in some way. For example, proteins called histones assemble into little yo-yo structures known as nucleosomes around which DNA wraps itself. A single human chromosome may contain hundreds of thousands of nucleosomes. The yo-yos are not just structural elements but implicated in gene regulation. A variety of small molecules can become attached to the histones to make tails; there is some evidence that these molecular tags themselves form a code. Also, the spacing between nucleosomes along the DNA is neither regular nor random, and it seems that the positioning patterns contain important information in their own right. All these variables are very complex and the details are still not completely understood, but it is clear that modifications made to proteins after their manufacture are important regulatory elements in the cell’s information-management system. A further complication is that a ‘gene’ is not necessarily a continuous segment of DNA but may be made of several pieces. As a result, the mRNA read-out has to be cut and spliced to assemble the components correctly. In some cases, there is more than one splicing, meaning that a single stretch of DNA can code for several proteins at once; which protein is expressed depends on the specifics of the splicing operation, which is itself managed by other genes and proteins … and so on.
Perhaps the biggest source of variety comes from the fact that, at least in complex organisms like animals and plants, the vast majority of DNA does not consist of genes coding for proteins anyway. The purpose of this ‘dark sector’ of DNA remains unclear. For a long while much of the non-coding DNA segments were dismissed as junk, as serving no useful biological function. But increasingly there is evidence that much of the ‘junk’ plays a crucial role in the manufacture of other types of molecules, such as short strands of RNA, which regulate a whole range of cellular functions. Cells are beginning to look like bottomless pits of complexity. The discovery of all these causal factors which are not located on the actual genes is part of the field known as epigenetics. It seems that epigenetics is at least as important as genetics as far as biological form and function are concerned.
In a pale echo of Frankenstein, it turns out that electricity is indeed a life force, but not quite in the way Mary Shelley (or rather Hollywood’s version of her story) imagined. Most cells are slightly electrically charged. They maintain this state by pumping positively charged ions (mostly protons and sodium) from inside to outside through the membrane that encloses the cell, creating a net negative charge. Typical potential differences across the membranes are between 40 and 80 millivolts. Although that doesn’t seem high, the membrane is so thin that this small voltage gradient represents an enormous local electric field – greater than those found near the Earth’s surface during a thunderstorm – and it’s actually possible to measure it. By using voltage-sensitive fluorescent dyes researchers can make pictures of the field patterns.
In a series of spectacular experiments at Tufts University, Michael Levin – he of the space worms – has demonstrated that electric patterning is important in sculpting the final morphology of an organism as it develops. Variations in voltage across large areas of the body serve as ‘pre-patterns’ – invisible geometrical scaffolds that drive downstream gene expression and thereby affect the path of development. By manipulating the electric potential differences across selected cells, Levin can disrupt the developmental process and create monsters to order – frogs with extra legs and eyes, worms with heads where tails should go, and so on.fn1
One series of experiments focuses on tadpoles of Xenopus, the African clawed frog. Normal frog embryos develop a characteristic pattern of pigmentation after a fraction of cells in the mid-region of the head and trunk begin to produce melanin. Levin treated the tadpoles with ivermectin, a common anti-parasite agent that electrically depolarizes cells by changing the flow of ions between the cell and its surroundings. Altering the electrical properties of so-called instructor cells had a dramatic effect, causing the pigmented cells to go crazy, spreading cancer-like into distant regions of the embryo. One perfectly normal tadpole developed a metastatic melanoma entirely from the electrical disruption, in the absence of any carcinogens or mutations. That tumours may be triggered purely epigenetically contradicts the prevailing view that cancer is a result of genetic damage, a story that I shall take up later in the chapter.
All this was remarkable enough. But an even bigger surprise lay in store. In a different experiment at Tufts University, devised by Dany Adams, a microscope was fitted with a time-lapse camera to produce a movie of the shifting electric patterns during the development of Xenopus embryos. What it showed was spectacular. The movie began with a wave of enhanced electrical polarization sweeping across the entire embryo in about fifteen minutes. Then various patches and spots of hyperpolarization and depolarization appeared and became enfolded as the embryo reorganized its structure. The hyperpolarized regions marked out the future mouth, nose, ears, eyes and pharynx. By altering the patterns of these electrical domains and tracing how the ensuing gene expression and face patterning changed, the researchers concluded that the electrical patterns pre-figure structures scheduled to emerge much later in development, most strikingly in the face of the frog-to-be. Electrical pre-patterning appears to guide morphogenesis by somehow storing information about the three-dimensional final form and enabling distant regions of the embryo to communicate and make decisions about large-scale growth and morphology.
Embryo development is a dramatic example of biological morphogenesis. Another is regeneration. Some animals can regrow their tails, even entire limbs, if they are lost for some reason. And sure enough, there’s an electrical story here too. Levin’s experimental creature of choice is a type of flatworm called a planarian (the ‘space worm’ species). These tiny animals have a head at one end with eyes and a brain to go with it, and a tail at the other end. Planaria are a favourite with teachers because if they are chopped in two they don’t die; instead, writes Levin,
the wound on the posterior half builds a new head, while the wound on the anterior half makes a tail. Two completely different structures are formed by cells that, until the cut occurred, were sharing all aspects of the local environment. Thus, still poorly understood long-range signals allow the wound cells to know where they are located, which direction the wound is facing and what other structures are still present in the fragment and do not need to be replaced.6
Levin discovered there is a distinctive electric pattern throughout the cut fragment, as there is indeed around wounds generally. Levin used drugs called heptanol and octanol, which sound like rocket fuel but serve to interfere with the ability of cells to communicate electrically with each other and hence modify the activity of the bioelectric circuit controlling how the tissues around the wounds decide their identity. By this means he was able to get a severed head to grow not a tail but another head, thus creating a two-headed, zero-tailed worm (see Fig. 12). Likewise, he can make two-tailed worms with no head. (He can even make worms with four heads or four tails.) The biggest surprise comes if the experimenters chop off the aberrant supernumerary head of a two-headed worm. You might expect this to rid the worm of any further two-headed aspirations, but it turns out that if the remainder of the worm is then cut in half two new two-headed worms are made! This is a dramatic example of epigenetic inheritance at work (see Box 10). The key point is that all these monster worms have identical DNA sequences yet dramatically different phenotypes. A visitor from Mars would surely classify them as different species based on their morphology. Yet, somehow, the physical properties of the organism (in this case, the stable states of the electric circuits) convey altered morphological information from one generation to the next.
Which brings up two important questions: where is the morphological information stored in these creatures, and how is it passed on between generations? Obviously, the information is not in the genes, which are identical. The DNA alone does not directly encode shape (anatomical layout) or the rules for repairing that shape if damage occurs. How do tissues know to keep rebuilding, say, the head of a planarian and stop when the right size has been completed? The standard reductionist explanation is to attribute the organism’s regenerative capabilities to a set of inherited genetic instructions, along the lines of ‘what to do if you are chopped in two: grow a new tail if you have lost your original’, and so on. But given that a two-headed planarian has the same set of genes as a normal planarian, how does a newly bisected two-headed worm tell its exposed stump ‘make a head’ in defiance of the normal protocol of ‘make a tail’? What epigenetic apparatus, exactly, is adapted by the momentary electrical tinkering yet remains locked in place for generation after generation of monsters even after Levin’s rocket fuel is removed?
The biggest problem here is not untangling the story of what proteins are where but how the system as a whole processes information on size, shape and topology over a scale much larger than any single cell. What is needed is a top-down view that focuses on information flow and mechanisms for encoding the shapes of large and complex structures. So far, however, that code, or the nature of the signals conveying the building and repair instructions, remains shadowy. One way forward is to imagine that there is some sort of ‘information field’ permeating the organism, which, after Levin and his collaborators have adulterated it, somehow embeds details about the large-scale properties of the monster-in-waiting, including its three-dimensional form. Just how that might work is anybody’s guess. The way Levin expresses it, there is a pre-existing ‘target morphology’ that guides a variety of shape-modulating signals and is stored, interpreted and implemented by a combination of chemical, electrical and mechanical processes acting in concert:
A ‘target morphology’ is the stable pattern to which a system will develop or regenerate after perturbation. Although not yet understood mechanistically, regeneration ceases when precisely the right size structure has been rebuilt, indicating a coordination of local growth with the size and scale of the host.7
Understanding the growth of complex forms in biology has enormous medical implications, ranging from birth defects to cancer. If these forms are mediated at least in part by electric patterning, or indeed by any encoding that we can learn to rewrite (and let cells build to specification), there is scope for correcting and controlling pathologies. The holy grail of regenerative medicine is to be able to regrow entire organs. The human liver will in fact regrow to its normal size following surgical resectioning. Again, how it knows the final shape and size is puzzling. If similar regeneration could be extended to nerves, cranio-facial tissue and even limbs, the applications would be stupendous. But achieving these goals requires a much better understanding of living systems as cohesive, computational entities that store and process information about their shape and their environment. Above all, we need to discover how informational patterns – electrical, chemical and genetic – interact and translate into specific phenotypes.
Electro-transduction is just one example of how physical forces can affect gene expression. Mechanical pressure or sheer stress acting on the cell as a whole can sometimes produce changes in the cell’s physical properties or behaviour. A well-known example is contact inhibition. Cells in a Petri dish will happily divide if they are treated well and receive nutrients, but division will stop if the cells become claustrophobic, such as when the colony pushes up against a boundary and the population becomes overcrowded. Cancer cells turn off contact inhibition. They also undergo drastic changes in their shape and stiffness when they leave the primary tumour and spread around the body. Another example: when a stem cell is placed next to a hard surface, it will express different genes than when embedded in soft tissue, affecting the type of cell it differentiates into, a phenomenon of obvious importance to embryogenesis. There is a popular aphorism in the cancer research community: ‘What a cell touches determines what a cell does.’ The mechanism behind these sorts of phenomena is known as mechano-transduction, meaning that an external mechanical signal – a gross physical force – is transduced into altered gene expression in response.8
The two-headed space worms provide a striking illustration of mechano-transduction in zero-gravity conditions. Another space wonder comes from my own university and the experiments of Cheryl Nickerson. She has been working with NASA to study changes in the gene expression of microbes when they go into Earth orbit. Even the humble salmonella bacterium can somehow sense it is floating in space and changes its gene expression accordingly.9 The finding has obvious implications for astronauts’ well-being, because a nasty bug that might be held in check on Earth may make someone sick in space. Related to this is the fact that humans normally carry about a trillion microbes around inside them, many in the gut, forming what is called their microbiome; it plays an important role in human health. If there are changes in gene expression within the microbiome due to long periods in zero- or low-gravity conditions, it could become a serious obstacle to long-term spaceflight.10
Let me mention a couple more intriguing discoveries to finish up this section. Salamanders are well known for their ability to regenerate entire limbs. It turns out that if the sacrificed leg has cancer, and is severed in mid-tumour, the new leg is cancer-free. Evidently, the limb morphology, somehow encoded in the stump, is re-programmed to make a healthy leg. This runs counter to the conventional wisdom that rapid cell proliferation – a feature of limb regeneration – is a cancer risk (cancer is sometimes described as ‘the wound that never heals’). Indeed, many studies have also shown the ability of embryos to tame aggressive cancer cells. Another oddity concerns deer antlers, which drop off and regrow every year. In some species of deer, if you cut a notch in the antlers, next year’s regrown antlers come complete with an ectopic branch (tine) at that same location.fn2 Where, one wonders, is the ‘notch information’ stored in the deer? Obviously not in the antlers, which drop off. In the head? How does a deer’s head know its antler has a notch half a metre away from it, and how do cells at the scalp store a map of the branching structure so as to note exactly where the notch was? Weird! In the magic puzzle box of life, epigenetic inheritance is one of the more puzzling bits of magic.
Box 10: Epigenetic inheritance
Like the proverbial farmer’s wife, the German evolutionary biologist August Weismann cut off the tails of mice over many generations, but he never succeeded in producing a tailless mouse – a blow to Lamarck’s theory of evolution by the inheritance of acquired characteristics. However, the recent surge in the study of epigenetics is painting a more nuanced picture. Within the body, a cell’s type is conserved when the cell divides: if, say, a liver cell replicates, it makes two liver cells, not a liver cell and a skin cell. So the epigenetic markers (for example, methylation patterns) that determine gene expression (‘Thou shalt be a liver cell’) will be passed on to the daughter cells. But what about epigenetic changes passed from one generation of the whole organism to the next, for example, from mother to son? That is a very different matter; if it occurred, it would strike at the very basis of Darwinian evolution. There is not supposed to be any mechanism for changes to an organism’s body to get into its germ line (sperm and eggs) and affect its offspring. Nevertheless, evidence for intergenerational epigenetic inheritance has long been staring biologists in the face. When a male donkey is crossed with a female horse it produces a (sterile) mule. If a female donkey is crossed with a male horse, the result is a hinny. Mules and hinnies are genetically identical but very different-looking animals: they are epigenetically distinct, and so they must carry epigenetic determinants that depend on the sex of their parents. Other examples have been discovered where genes from the parents are imprinted with epigenetic molecular marks that manage to get into the germ cells and survive the reproductive process. Moreover, botanists know many cases where epigenetic changes accumulated during a plant’s life are passed on to the next and subsequent generations. Even in humans some studies have uncovered hints of something similar. One of these involved Dutch families who suffered near-starvation in the Second World War because they were bypassed by the Allied advance and unable to receive food. Children of the survivors were born with lower than average body weight and stayed below average height all their lives. More surprisingly, their own children seem to be smaller too.
So where, exactly, is the epigenome? Genes are physical objects with a definite location in the cell: a specific gene lies at a certain position on DNA. You can see them with a microscope. When it comes to epigenetics, however, there is no ‘epigenome’ in the same physical sense, no well-defined object at a specific location in the cell. Epigenetic information processing and control is distributed throughout the cell (and perhaps beyond the cell too). It is global, not local, as is the case with the cellular analogue of von Neumann’s supervisory unit I mentioned in the previous chapter.
FLIRTING WITH LAMARCKISM
‘Chance favors the prepared genome.’
– Lynn Caporale11
Decades before Darwin released On the Origin of Species, a French biologist published a very different theory of evolution. His name was Jean-Baptiste Lamarck. The centrepiece of Lamarckian evolution is that characteristics acquired by an organism during its lifetime can be inherited by its offspring. Thus, if an animal strives in this or that manner (tries to run faster, reach higher …) in its relentless struggle for survival, its progeny will come with an inherited slight improvement (a bit swifter, a bit taller). If this theory is correct, it would provide a mechanism for fast-paced purposeful change directed towards betterment. My mother often remarked that she could really use another pair of hands when doing housework. Imagine if, as a result of this need, her children had been born with four arms! By contrast, Darwin’s theory asserts that mutational changes are blind; they have no link to the circumstances or requirements of the organism that carries them. If a rare mutation confers an advantage, then that’s pure dumb luck. There is no directional progress, no systematic inbuilt mechanism for improvement.
Evolution would certainly work much faster and more efficiently if nature engineered just the right mutations to help out, in the way Lamarck envisaged. Nevertheless, biologists long ago dismissed the idea as too much like the guiding hand of God, preferring instead to appeal to random chance as the sole explanation for variation. And that’s the way it was for many decades. Now, however, doubts have crept in. Levin’s monster worms are surely a clear example of the inheritance of acquired characteristics – acquired in this case by laboratory butchery. Many other examples are known. So has the time come to abandon Darwinism and embrace Lamarckism?
Nobody can deny that natural selection encourages the survival of the fittest. Organisms show variations, and nature selects the fitter. But there have always been niggling worries. Nature can only work with the variants it has, and a fundamental question is how those variants arise. Survival of the fittest maybe, but what about the arrival of the fittest, as the Dutch botanist Hugo de Vries dubbed it a century ago? In biology, remarkable innovations with far-reaching consequences abound: photosynthesis, the bony skeleton of vertebrates, avian flight, insect pollination, neural signalling, to name but a few. The problem of how life generates so many ingenious solutions to survival problems is today the subject of lively investigation.12 If something works well, random changes are likely to make it worse, not better. Even with a duration of 3 or 4 billion years, is it possible that so much organized complexity – eyes, brains, photosynthesis – has arisen just from random variation and natural selection?13
Over the years many scientists have expressed scepticism. ‘A simple probabilistic model would not be sufficient to generate the fantastic diversity we see,’14 wrote Wolfgang Pauli, a quantum physicist and contemporary of Schrödinger. Even distinguished biologists have expressed their doubts. Theodor Dobzhansky wrote: ‘The most serious objection to the modern theory of evolution is that since mutations occur by “chance” and are undirected, it is difficult to see how mutation and selection can add up to the formation of such beautifully balanced organs as, for example, the human eye.’15 Many of these problems would evaporate if some vestige of Lamarckian evolution were at work.
In 1988 a group of Harvard biologists claimed they had witnessed a clear-cut example of the propitious arrival of the fittest. A team led by John Cairns made the provocative claim ‘Cells may have mechanisms for choosing which mutations will occur.’16 Choosing? Published in Nature, and coming from a highly respectable laboratory, the announcement was received with consternation. To understand their experiments, recall from here that the bacteria E. coli like to eat glucose but can flip a switch to enable them to metabolize less-tasty lactose if pressed. Cairns’ group worked with a mutant strain of E. coli that was unable to process lactose, and then challenged them with a lactose-only diet. They observed that some of these starving bacteria spontaneously mutated to the lactose-utilizing form. In itself, this is no threat to orthodox Darwinism, so long as the said mutations arise as lucky flukes. But when the Harvard team worked out how likely that was, they concluded that their bacteria had uncanny rates of success, beating the raw odds by a huge margin. The researchers wondered, ‘Can the genome of an individual cell profit by experience?’17 just as Lamarck had proposed. They hinted that the answer might be yes, and that they were dealing with a case of mutations ‘ “directed” toward a useful goal’.
Reacting to the furore, Cairns did some follow-up experiments and backtracked on the more sensational aspects of the claim. But the genie was out of the bottle, and there ensued a surge of experiments by his and other groups. A lot of E. coli suffered glucose deprivation. When the dust settled, this is what emerged. Mutations are not random: that part is correct. Bacteria have mutational hotspots – specific genes that mutate up to hundreds of thousands of times faster than average. This is handy if it is advantageous for the bacteria to generate diversity. A case in point is when they invade a mammal and have to do battle with the host’s immune system. Bacteria have identifying surface features that act a bit like a soldier’s uniform. The host’s immune system recognizes the pathogen from the details of its coat. A bacterium that can keep changing uniform will obviously have a survival advantage, so it makes good Darwinian sense for the ‘uniform genes’ to be highly mutable. For circumstances such as this, bacteria have evolved certain ‘contingency genes’ that are more mutable than others, implying a greater likelihood of mutations arising in these genes. Within that contingency, however, it’s still a hit-and-miss affair. There is no evidence for bacteria ‘choosing’ specific mutations to order, as Cairns originally hinted.
A more arresting example concerns bacteria that can selectively switch on elevated mutation rates in just the right genes to get them out of trouble. Barbara Wright of the University of Montana looked at mutants of poor old E. coli that possessed a defective gene, one that codes for making a particular amino acid.18 You and I usually get our amino acids from food, but if we’re starving our cells can make their own. Same with bacteria. What Wright wondered was how a starving bacterium with a faulty amino acid gene would respond. The bacterium gets the signal ‘need amino acids now!’ but the defective gene cranks out a flawed version. Somehow, the cell senses this danger and turns up the mutation rate of that specific gene. Most mutations make things worse. But among a colony of starving bacteria there is now a good chance that one of them will get lucky and suffer just the right mutation to repair the defect, and that cell saves the day. It gets the bacterial equivalent of another pair of hands. The term used for this biased mutation is ‘adaptive’, because it makes the organism better adapted to its environment.
A long-time pioneer of adaptive mutations is Susan Rosenberg, now at Baylor College of Medicine in Houston, Texas. Rosenberg and her colleagues also set out to get to the bottom of how starving bacteria manage to mutate their way to a culinary lifeline with such uncanny panache. They focused on the repair of double-strand breaks in DNA – a never-ending chore required so that cells can carry on their business as usual.19 There are various methods used to patch such a gap, some high quality, others less so. Starving bacteria, Rosenberg discovered, can switch from a high-fidelity repair process to a sloppy one. Doing so creates a trail of damage either side of the break, out as far as 60,000 bases or more: an island of self-inflicted vandalism. Rosenberg then identified the genes for organizing and controlling this process. It turns out they are very ancient; evidently, deliberately botching DNA repair is a basic survival mechanism stretching back into the mists of biological history. By generating cohorts of mutants in this manner, the colony of bacteria improves its chances that at least one daughter cell will accidentally hit on the right solution. Natural selection does the rest. In effect, the stressed bacteria engineer their own high-speed evolution by generating genomic diversity on the fly.
Is there any hint from Rosenberg’s experiments that these wily bacteria can also generate the ‘right’ mutations with better than chance, as Cairns originally implied? Do the fittest ‘arrive’ with prescient efficiency? This is not a simple yes-or-no issue. It’s true that the cells don’t adopt a scattergun approach in their mutational rampage: the elevated mutations are not distributed uniformly across the genome. Rosenberg confirmed, however, the existence of certain favoured hotspots which are more likely than chance to house the genes needed to evolve out of trouble. But unlike the highly focused mechanism that Barbara Wright discovered, which targets specific faulty genes expressing themselves badly, Rosenberg’s mutations affect all genes in the hotspot regions indiscriminately, regardless of whether they are struggling to churn out proteins or just sitting idle. In that sense it is a more basic yet also more versatile mechanism.
Here is an analogy. Imagine being trapped inside a burning building. You guess there might be a window somewhere that will open and let you escape, but which one? Maybe there are dozens of windows. A really smart person would have figured out a fire escape procedure in advance, just in case. But you didn’t. So what is the next smartest thing to do? It is of course to try each of the windows one by one. In the absence of any other information, a random sampling procedure is as good as any. A really dumb thing to do would be to scurry around totally at random, going into cupboards or ducking under beds. Targeted randomness is more efficient than complete randomness. Well, bacteria are not super-smart but nor are they really dumb: they concentrate their chances where it is most likely to do some good.
How did all this mutational magic come to exist? In retrospect, it’s not too surprising. Clearly, evolution will work much better if the mechanisms involved are flexible and can themselves evolve – what is often called the evolution of evolvability. Long, long ago, cells that retained an ability to evolve their way out of trouble would have been at an advantage. An evolution-boosting mechanism that is switched on when conditions demand and dialled back when times are great is a boon. The adaptive response to stressfn3 is almost certainly an ancient mechanism (really a set of mechanisms involving a spectrum of processes, from haphazard to focused and directed) that evolved for good biological purposes. The biologist Eva Jablonka describes adaptive mutations as an ‘informed search’. She concludes, ‘The cell’s chances of finding a mutational solution are enhanced because its evolutionary past has constructed a system that supplies intelligent hints about where and when to generate mutations.’20 It is important to understand that this is not a falsification of Darwinism but an elaboration of it. This is Darwinism 2.0. Biochemist Lynn Caporale writes: ‘Rejecting entirely random genetic variation as the substrate of genome evolution is not a refutation, but rather provides a deeper understanding, of the theory of natural selection of Darwin and Wallace.’21 The refinement that emerges from these recent experiments displaying a Lamarckian flavour is that nature selects not just the fittest organisms but the fittest survival strategies too.
The foregoing ideas illustrate how organisms use information from the past to chart the future. This information is inherited both over deep time (for example, the contingency genes I discussed here) and epigenetically, from the previous generation. Life may therefore be described as an informational learning curve that swoops upwards. Organisms do not have to proceed by trial and error and ‘reinvent the wheel’ at each generation. They can profit from life’s past experience. This progressive trend stands in stark contrast to the second law of thermodynamics, which tells a story of degeneration and decay.
DEMONS IN THE GENES
Surprising though adaptive mutations may be, they still imply that genomes are the passive victims of randomly inflicted externally generated blows or blunders, albeit with rigged odds. It’s still a chancey business. But suppose cells in trouble didn’t have to rely on external forces at all to cause mutations? What if they could actively manipulate their own genomes?
Actually, it is clear that they do. Sexual reproduction involves several slicing-and-dicing genomic reconfigurations, some random, some supervised. Many intermingling methods are out there, each of which involves cells shuffling their own DNA in a carefully arranged manner. And sex isn’t the only example. Correcting the errors that occur during DNA replication requires another set of genomic management operations. Most of the primary damage to DNA, for example by radiation or thermal disruption, never makes it to the daughter cells because it is repaired first. Human DNA would suffer devastating mutational damage, estimated at an overall 1 per cent copying error rate per generation, without all the in-house, high-tech proofreading, editing and error correcting which reduces the net mutation rate to an incredible one in 10 billion. So cells are able to monitor and actively edit their own genomes to a high degree of fidelity in an attempt to maintain the status quo.
But now we encounter a fascinating question: can cells actively edit their genomes to change the status quo? Decades before the work of Cairns and Rosenberg this question was investigated in a series of remarkable experiments by the distinguished botanist and cell biologist Barbara McClintock. Starting in her student days in the 1920s, she experimented with maize plants and established many of the basic properties of chromosome structure and organization we know today, for which she subsequently received a Nobel Prize in Physiology or Medicine – the first woman to win the prize unshared in that category. With the help of a basic microscope, McClintock looked to see what happened to the chromosomes of the maize plants when they were exposed to X-rays. What she reported caused such a ballyhoo and attracted so much scepticism that in 1953 she felt moved to stop publishing her data. What was uncontentious were her observations that chromosomes break into fragments when irradiated. But the big surprise was the fact that the pieces could rejoin again, often in novel arrangements. Humpty-dumpty could be reassembled in a baroque sort of way. Chromosome reorganization writ large might seem lethal, and it often was. But it was not always so: in some cases, the mutant plants went on to replicate their grossly modified genomes. Crucially, McClintock found that the large-scale mutations were far from random; it appeared that the maize cells had a contingency plan for the day their genomes were smashed. Even more amazingly, if the plants were stressed, for example by infection or mechanical damage, spontaneous chromosome breakage could occur without the benefit of X-ray disruption; the broken ends were rejoined after the chromosome replicated. In 1948 McClintock made her most startling discovery of all. Segments of chromosomes could be transposed – switch places on the genome – a phenomenon popularly known as ‘jumping genes’. In the maize plants this produced a mosaic coloured pattern.
Today, genomic transpositions are recognized as widespread in evolution. It has been estimated that up to half the human genome has undergone such genetic gymnastics. Cancer researchers are also very familiar with transpositions. A much-studied example is the Philadelphia chromosome (after where it was discovered) which can trigger leukaemia in humans; it involves a chunk of chromosome 9 being transposed for a chunk of chromosome 22. In certain late-stage cancers, chromosomes can become so deranged as to be almost unrecognizable, with wholesale rearrangements, including entire chromosome duplications and stand-alone fragments replacing the orderly arrangement found in healthy cells. An extreme case is called chromothripsis, in which chromosomes disintegrate into thousands of pieces and rearrange themselves into scrambled monsters.
In spite of the grudging acknowledgement that McClintock was right, her results remain disturbing because they imply that the cell can be an active agent in its own genomic change. She herself evidently thought as much. On the occasion of her Nobel Prize, awarded for ‘the discovery of mobile genetic elements’, she had this to say:
The conclusion seems inescapable that cells are able to sense the presence in their nuclei of ruptured ends of chromosomes and then to activate a mechanism that will bring together and then unite these ends, one with another … The ability of a cell to sense these broken ends, to direct them toward each other, and then to unite them so that the union of the two DNA strands is correctly oriented, is a particularly revealing example of the sensitivity of cells to all that is going on within them … A goal for the future would be to determine the extent of knowledge the cell has of itself, and how it utilizes this knowledge in a ‘thoughtful’ manner when challenged … monitoring genomic activities and correcting common errors, sensing the unusual and unexpected events, and responding to them, often by restructuring the genome. We know about the components of genomes that could be made available for such restructuring. We know nothing, however, about how the cell senses danger and instigates responses to it that often are truly remarkable.22
It turned out that transposition and mobile genetic elements were only the tip of the iceberg. When facing challenges, cells have many ways to ‘rewrite’ their genomes, just as computer programs have bugs removed or are upgraded to perform new tasks. James Shapiro, a collaborator with McClintock as a young man, has made a comprehensive study of the mechanisms involved. One of these is called reverse transcription, whereby RNA, which normally transcribes sequences from DNA, is sometimes able to write its own sequence back into DNA. Because there are many mechanisms for RNA sequences to be modified after they have transcribed the information from DNA, reverse transcription opens the way for cells to alter their own DNA via RNA modification. A specific reverse transcription gene that has been studied in detail is BC1 RNA, which is important in the neural systems of rodents.23
It is now recognized that diverse processes of reverse transcription have played a major role in evolution and may, for example, account for a large fraction of the genetic differences between humans and chimpanzees.
Nor is the backflow of information limited to RNA → DNA. Because genome repair is controlled by complex interactions in the cell, the decision ‘to repair or not to repair’ or ‘how to repair’ can depend on a variety of proteins that have been modified after their formation. The upshot is that proteins, and modifications they have acquired during the life cycle of the cell, can influence genomic content: the epigenetic tail wagging the genetic dog. In all, Shapiro has identified about a dozen different mechanisms whereby a cell, operating at a systems level, can affect the information content of its own DNA, a process he calls natural genetic engineering. To summarize the central dogma of neo-Darwinian biology, information flows from inert DNA to mobile RNA to functional proteins in a one-way traffic. To use a computer analogy, the Darwinian genome is a read-only data file. But the work of McClintock, Shapiro and others explodes this myth and shows it is more accurate to think of the genome as a read–write storage system.
The refinements of Darwinism I have described in this chapter go some way to explaining the puzzle of the arrival of the fittest. At present, there are collections of case studies hinting at different mechanisms, many with Lamarckian overtones, but as yet no systematic information-management laws or principles have been elucidated that govern these phenomena. However, it’s tempting to imagine that biologists are glimpsing an entire shadow information-processing system at work at the epigenetic level. ‘Nature’s many innovations – some uncannily perfect – call for natural principles that accelerate life’s ability to innovate …’24 writes Andreas Wagner, an evolutionary biologist from Switzerland. ‘There is much more to evolution than meets the eye … Adaptations are not just driven by chance, but by a set of laws that allow nature to discover new molecules and mechanisms in a fraction of the time that random variation would take.’25 Kevin Laland, an evolutionary biologist at the University of St Andrews, is co-founder of what has been dubbed the ‘Extended Evolutionary Synthesis’. ‘It is time to let go of the idea that the genes we inherit are a blueprint to build our bodies,’ he writes. ‘Genetic information is only one factor influencing how an individual turns out. Organisms play active, constructive roles in their own development and that of their descendants, so that they impose direction on evolution.’26
Orthodox biologists are not taking this assault lying down. The heresy of Lamarckism is always guaranteed to inflame passions and the Extended Evolutionary Synthesis remains a contentious challenge, as is the claim that epigenetic changes can be passed down the generations. Just how much the ‘purist’ version of Darwinism needs to be adapted is controversial.27 It’s fair to say that the battle is far from over.
CANCER: THE HARSH PRICE OF MULTICELLULARITY
Genomes can undergo profound changes not just on evolutionary timescales over millions of years but during the lifetime of an organism. The most dramatic example of the latter is provided by cancer, the world’s number-two killer. Dreadful though this disease may be, it provides a fascinating window on our evolutionary past.
There is no hard-and-fast definition of cancer; instead it is characterized by about a dozen ‘hallmarks’.28 Advanced cancers in humans may display all or only some of the hallmarks. They include a surging mutation rate, unrestrained cell proliferation, disabling of apoptosis (programmed cell death), evasion of the immune system, angiogenesis (organization of a new blood supply), changes in metabolism and – the most well-known and medically problematic – a penchant for spreading around the body and colonizing organs remote from the site of the primary tumour, a process called metastasis.
Cancer is the most studied subject in biology, with over a million published papers in the last fifty years. It may therefore come as a surprise to the reader to learn that there is no agreement on what cancer is, why it exists and how it fits into the great story of life on Earth. Very little attention has been given to understanding cancer as a biological phenomenon, as opposed to a disease to be annihilated by any means at hand. Most of the gargantuan research effort across the world has been devoted to destroying cancer. Nevertheless, standard cancer therapy – a mix of surgery, radiation and chemical toxins – has changed little in decades.fn4 Survival rates for all but a handful of cancer types have improved only modestly or not at all: life extension through chemotherapy is mostly a rearguard action against the inevitable, measured in weeks or months rather than years. This dismal state of affairs can’t be blamed on lack of funding. The US government alone has spent $100 billion on cancer research since President Nixon supposedly declared war on it in 1971, while charities and drug companies have poured in billions more.
Perhaps the lack of progress is because scientists are looking at the problem in the wrong way? Two common misconceptions are that cancer is a ‘modern disease’ and that it primarily afflicts humans. Nothing could be farther from the truth. Cancer or cancer-like phenomena are found in almost all mammals, birds, reptiles, insects and even plants. Work by Athena Aktipis and her collaborators shows the existence of cancer, or cancer analogues, across all metazoan categories, including fungi and corals.29 (See Fig. 13.) Instances of cancer have even been found in the simple organism hydra.30
The fact that cancer is so widespread among species points to an ancient evolutionary origin. The common ancestor of, say, humans and flies dates back 600 million years, while the broader categories of cancer-susceptible organisms have points of convergence over 1 billion years ago. The implication is that cancer has been around for as long as there have been multicelled organisms (metazoa). This is reasonable. It goes without saying that cancer is a disease of bodies; it makes little sense to say that an isolated bacterium has cancer. But bodies did not always exist. For 2 billion years life on Earth consisted of single-celled organisms only. About 1.5 billion years ago the first multicellular forms appeared, during the geological epoch known as the Proterozoic (‘earlier life’ in Greek).fn5
The transition to multicellularity entailed a fundamental change in the logic of life. In the world of single cells, there is but one imperative: replicate, replicate, replicate! In that sense, single cells are immortal. Multicelled creatures, however, do things very differently. Immortality is outsourced to specialized germ cells (for example, eggs and sperm) whose job is to carry the organism’s genes forward into future generations. Meanwhile, bodies, which are a vehicle for these germ cells, behave very differently. They are mortal. The cells of the body (somatic cells) retain a faint echo of their past immortality in a limited ability to replicate. A typical skin cell, for example, can divide – between fifty and seventy times. When a given somatic cell reaches its use-by datefn6 it either goes dormant (a state called senescence) or commits suicide (apoptosis). That does not spell the end of the organ, because replacement cells of the same type are made by stem cells. But eventually the replacement process also wears out and the whole body dies, leaving the germ-line progeny, if any, to carry the genetic heritage into the future.
Why would any cell in its right mind sign up for a multicelled existence that involves a short burst of replication followed by suicide? What possible advantage can it gain in the great evolutionary survival game? As always in biology, there are trade-offs. By joining a collective of genetically similar cells, a given cell will still contribute to the propagation of most of its genes via the germ cells. If the collective as a whole possesses survival functions unavailable to single cells, then the arithmetic of genetic legacy may tip the balance in favour of the community over the go-it-alone approach. When the mathematics looks right, a deal is struck between individual cells and the organism. The cells join the collective project and die, and in return the organism takes on the responsibility to propagate the cells’ genes. Multicellularity therefore involves an implicit contract between the organism as a whole and its cellular members. It was a contract first signed in the Proterozoic era, over a billion years ago.
Multicellularity can be a good idea – it works for us! – but it does have a downside. When individuals join a communal effort there is always vulnerability to cheating. This is familiar from human society, where people receive a survival benefit from organized government through such things as defence, welfare and infrastructure but are expected to pay for it in taxes. As is well known, there is a strong temptation to cheat – to take the benefits on offer but dodge the taxes. It happens all over the world. To counter it, governments have invented layers of rules. (Tax law in Australia, for example, runs to a million words. In the US the tax code is almost infinitely complex.) The rules are then policed by government and law-enforcement agencies. In spite of this elaborate set-up, the system is imperfect; there is an arms race between cheats and enforcers: internet fraud and identity theft are prime current examples. A similar arms race is played out in multicellularity. To get individual cells to stick to the contract there have to be layers of regulatory control, policed by the organism as a whole, to deter cheats. Thus, a given somatic cell (skin cell, liver cell, lung cell …) will normally divide only when the regulations permit. When more cells of that variety are needed, the cell’s own internal ‘replication program’ will take care of it. But if division is inappropriate, then the regulatory mechanisms will intercede to either prevent it or, if the cell is persistently recalcitrant, order the death sentence: apoptosis. A graphic example of this strict policing occurs if a cell finds itself in the wrong tissue environment. For example, if a liver cell is accidentally transported to, or deliberately transplanted in, the lung, it won’t fare well. Chemical signals from the lung tissue recognize that they have an interloper (‘not one of us!’) and may order apoptosis.
What does cheating mean for a cell in a multicelled organism? It means a default to the selfish every-cell-for-itself strategy of unicellular life: replicate, replicate, replicate. In other words, uncontrolled proliferation. Cancer. Put simply, cancer is a breakdown of the ancient contract between somatic cells and organisms, followed by a reversion to a more primitive, selfish agenda.
Why does the policing fail? There can be many reasons. An obvious one is damage, say from radiation or a carcinogenic chemical, to one of the ‘police genes’. There is a class of genes, of which p53 is the best known, that serve as tumour suppressors. Damage p53 and a tumour may not be suppressed. Another trigger is immunosuppression. The adaptive immune system includes cancer surveillance as part of its remit. If the system is working properly, incipient cancer cells are spotted and zapped (or incarcerated and contained) before they can cause trouble. But cancer cells can cloak themselves chemically to hide from the immune surveillance police. They can also subvert the immune system by recruiting its scouts (macrophages) and ‘turning’ them, like captured spies, to work for them. Tumour-associated macrophages will screen a tumour and stymie the immune attack.
For cancer to take hold, two things have to happen. A normal cell has to embark on a cheating strategy and the organism’s police have to slip up somewhere. The conventional explanation is the somatic mutation theory, according to which genetic damage accumulates in somatic cells as a result of ageing, radiation or carcinogenic chemicals causing the cells to misbehave and go rogue, that is, embark on their own agenda. The resulting ‘neoplasm’, or population of new cells, rapidly develops (says the orthodox theory) the distinctive hallmarks I mentioned, including uncontrolled proliferation, plus a tendency to spread around the body and colonize remote organs. The somatic mutation theory assumes that the same hallmarks of cancer are reinvented de novo in each host solely by a sort of fast-paced Darwinian process of natural selection, in which the fittest (that is, nastiest) cancer cells outbreed their competitors via runaway replication, eventually killing the host (and themselves). Though entrenched, the somatic mutation theory has poor predictive power, its explanations amounting to little more than Just So stories on a case-by-case basis. Most seriously, it fails to explain how mutations confer so many fitness-improving gains of function in a single neoplasm in such a short time (yes, that again). It also seems paradoxical that increasingly damaged and defective genomes should enable a neoplasm to acquire such powerful new functionality and so many predictable hallmarks.
TRACING THE DEEP EVOLUTIONARY ROOTS OF CANCER
Over the past few years my colleagues and I have developed a somewhat different explanation of cancer that seeks its origins in the far past.31 We were struck by the fact that cancer almost never invents anything new. Instead, it merely appropriates already existing functions of the host organism, many of them very basic and ancient. Limitless proliferation, for example, has been a fundamental feature of unicellular life for aeons. After all, life is in the business of replication and cells have had billions of years to learn how to keep going in the face of all manner of threats and insults. Metastasis – the process whereby normally sedentary cells become mobile, quitting a tumour to spread around the body – mimics what happens during early-stage embryogenesis, when immature cells are often not anchored in place but surge in organized patterns to designated locations. And the propensity of circulating cancer cells to invade other organs closely parallels what the immune system does to heal wounds. These facts, which oncologists know well, combined with the predictable and efficient way that cancer progresses through its various stages of malignancy, convinced us that cancer is not a case of damaged cells randomly running amok but an ancient, well-organized and efficient survival response to stress.fn7 Crucially, we believe that the various distinctive hallmarks of cancer do not independently evolve as the neoplasm goes along – that is, are stumbled across by accident – but are deliberately switched on and deployed systematically as part of the neoplasm’s organized response strategy.
In summary, our view of cancer is that it is not a product of damage but a systematic response to a damaging environment – a primitive cellular defence mechanism. Cancer is a cell’s way of coping with a bad place. It may be triggered by mutations, but its root cause is the self-activation of a very old and deeply embedded toolkit of emergency survival procedures.fn8 The key distinction between the two theories can be illustrated with an analogy. Consider a playground victim of bullying who runs away as a survival strategy. The victim’s exit is self-propelled; the pushes and punches of his attackers may trigger his flight, but they are not themselves the ultimate cause of his motion – he is not pushed away, he runs away. Here is another analogy. If a computer suffers an insult – corrupted software or a mechanical blow – it may start up in safe mode (see Fig. 14). This is a default program enabling the computer to run on its core functionality even with the damage. In the same way, cancer is a default state in which a cell under threat runs on its ancient core functionality, thereby preserving its vital functions, of which proliferation is the most ancient, most vital and most protected. To trigger cancer, the threats don’t have to be radiation or chemicals; they could be ageing tissues, low-oxygen tension or mechanical stresses of various sorts, including wounding. (Or even electrical disruption – see here.) Many factors, individually or collectively, can cause the cell to adopt its inbuilt ‘cancer safe mode’.
Although elements of the cancer default program are very ancient, dating back to the origin of life itself, some of the more sophisticated features recapitulate later stages in evolution, especially in the period between 1.5 billion and 600 million years ago, when primitive metazoans emerged. In our view, cancer is a sort of throwback or default to an ancestral form; in technical jargon, it is an atavistic phenotype. Because cancer is deeply integrated into the logic of multicellular life, its ancient mechanisms highly conserved and fiercely protected, combating it proves a formidable challenge.
Our theory makes many specific predictions. For example, we expect the genes that are causally implicated in cancer (usually called oncogenes) to cluster in age around the onset of multicellularity. Is there any evidence for this? Yes, there is. It is possible to estimate the ages of genes from their sequence data by comparing the number of differences across many species. This well-tried technique, known as phylostratigraphy, enables scientists to reconstruct the tree of life, working backwards from common features today to deduce the convergence point in the past (see Fig. 15).
A study in Germany using four different cancer gene datasets demonstrated the presence of a marked peak in genes originating at around the time that metazoa evolved.32 A recent analysis of seven tumour types led by David Goode and Anna Trigos in Melbourne, Australia, focused on gene expression.33 They sorted genes into sixteen groups by age and then compared expression levels in cancer and normal tissue for each group. The results were striking. Cancer over-expresses genes belonging to the two older groups and under-expresses younger genes, exactly as we predicted. Furthermore, they found that as cancer progresses to a more aggressive, dangerous stage the older genes are expressed at higher levels, confirming our view that cancer reverses the evolutionary arrow at high speed as it develops in the host organism, with the cells reverting to their primitive ancestral forms in a space of weeks or months. More generally, the Australian group found that genes associated with unicellularity are more active in cancer than those that evolved later, during the era of multicellularity.
In our own work at Arizona State University we looked at mutation rates.34 The atavistic theory predicts that older genes should be less mutated in cancer (after all, they are responsible for running the ‘safe mode’ program), while younger genes should be mutated more. My colleagues Kimberly Bussey and Luis Cisneros considered a total of 19,756 human genes and used an inventory of cancer genes called COSMIC compiled by the UK Sanger Institute. This data was combined with a database of genetic sequence data from about 18,000 species across all taxonomic groups that includes an analysis of evolutionary ages. This allowed my colleagues to estimate the evolutionary ages of the genes in the human genome. They found that genes younger than about 500 million years were indeed more likely to be mutated – in normal, but especially in cancer, tissue – while genes older than a billion years tend to suffer less mutations than average, as we expected. They also confirmed the German study that the ages of cancer genesfn9 display a cluster around the time of the onset of multicellularity, supporting our contention that cancer is driven by disruption of functions that evolved to achieve multicellular organization. COSMIC classifies genes into dominant and recessive, and my colleagues found that cancer genes with recessive mutations were significantly older than most human genes.
The most telling result came from addressing a rather different question: what are cancer genes good for? A database called DAVID organizes genes around their functionality. When Cisneros and Bussey fed the COSMIC data into DAVID what leapt out was that recessive cancer genes older than 950 million years were strongly enriched for two core functions: cell cycle control and DNA damage repair involving double-strand breaks (the worst kind of damage DNA can suffer). Looking at the evolutionary history of the genes involved, the researchers spotted something significant. The non-mutated genes in the same DNA repair pathways correspond tofn10 those genes in bacteria that drive the adaptive mutation response to stress – the very phenomenon I discussed earlier in this chapter (see here). And as in bacteria, these genes serve to turn up the cell’s mutation rate in a desperate effort to survive by evolving a pathway out of trouble. I explained Susan Rosenberg’s discovery that when a bacterium senses a double-strand break it can switch to a sloppy repair mechanism, creating a trail of errors (mutations) either side of the break. We were keen to know whether cancer cells also display a pattern of damage around double-strand repairs. It turns out they do. Out of 764 tumour samples from seven different sites (pancreas, prostate, bone, ovary, skin, blood and brain) my colleagues looked at, 668 had evidence of mutational clustering around the break. This all fitted in with our theory that cells under stress turn cancerous by reawakening ancestral gene networks which, among other things, create a high rate of mutations. Thus, one of the best-known hallmarks of cancer (and the main reason why it so often evades chemotherapy by evolving drug-resistant variants) turns out to be self-inflicted. This finding fits in well with the atavism theory: cancer is merely appropriating an ancient stress response, still used by bacteria today, which evolved way back in the era of unicellular life.
As in stressed bacteria, mutations in cancer cells are far from random: there are definite mutational ‘hotspots’ and ‘cold-spots’ (regions of low mutation). This makes perfect sense. Multicelled organisms should work hard to protect key parts of their genomes, such as those responsible for running the core functions of the cell, and devote less resources to the ‘bells and whistles’ associated with more recently evolved and less critical traits. A project led by Princeton’s Robert Austin and Amy Wu subjected cancer cells to a therapeutic toxin (doxorubicin) and studied the evolution of their resistance to this drug. Austin and Wu found that cold-spot genes were significantly older than average.35 These new results help explain why natural selection hasn’t eliminated the scourge of cancer. If tumours really are a reversion to an ancestral form, then we might expect that the ancient pathways and mechanisms that drive cancer would be among the most deeply protected and conserved, as they fulfil the most basic functions of life. They can’t be got rid of without disaster befalling the cells concerned. The mutator genes we investigated are just one example.
Another reason that evolution hasn’t eliminated cancer is because of the link with embryogenesis. It has been known for thirty years that some oncogenes play a crucial role in development; eliminating them would be catastrophic. Normally, these developmental genes are silenced in the adult form, but if something reawakens them cancer results – an embryo gone wrong developing in adult tissue. The writer George Johnson summarizes this well by referring to tumours as the ‘embryo’s evil twin’.36 Significantly, the early stages of an embryo are when the organism’s basic body plan is laid down, representing the earliest phase of multicelled life. When the cancer switch is flipped, there will be systematic disruption in both the genetic and epigenetic regulators of information flow, as the cells recapitulate the very different circumstances of early embryo development. This will involve both changes to the way regulatory genes are wired together and changes in patterns of gene expression. Our research group is trying to find information signatures of these changes. We hope it will prove possible to identify distinct ‘informational hallmarks’ of cancer to go alongside the physical hallmarks I mentioned – a software indicator of cancer initiation that may precede the clinically noticeable changes in cell and tissue morphology, thus providing an early warning of trouble ahead.
The atavistic theory of cancer has important implications not just for diagnosis but for therapy. We think the search for a general-purpose ‘cure’ for cancer is an expensive diversion, and that cancer, being so deeply entrenched in the nature of multicellular life itself, is best managed and controlled (not eliminated) by challenging the cancer with physical conditions inimical to its ancient atavistic lifestyle. Only by fully understanding the place of cancer in the overall context of evolutionary history will a serious impact be made on human life expectancy in the face of this killer disease.