GENOMIC IMPRINTING
[DAVID HAIG: ]My work over the last decade or so has been principally concerned with conflicts within the individual organism. In a lot of evolutionary biology, the implicit metaphor is that the organism is a machine or, more specifically, a fitness-maximizing computer trying to solve some problem. Maximizing fitness is analogous to maximizing a utility function in economics. I'm interested in situations where there are conflicts within the individual, in which different agents within the self have different fitness functions, as well as the internal politics resulting from those conflicts of interest.
The area to which I've given the greatest attention is a new phenomenon in molecular biology called genomic imprinting, which is a situation in which a DNA sequence can have conditional behavior depending on whether it is maternally inherited—coming from an egg—or paternally inherited—coming through a sperm. The phenomenon is called imprinting because the basic idea is that there is some imprint that is put on the DNA in the mother's ovary or in the father's testes which marks that DNA as being maternal or paternal, and influences its pattern of expression—what the gene does in the next generation in both male and female offspring.
This is a complicated process because the imprint can be erased and reset. For example, the maternal genes in my body when I pass them on to my children are going to be paternal genes having paternal behavior. If my daughter passes on paternal genes to her children, even though she got the gene as a paternal gene from me it would be a maternal gene to her own offspring. Molecular biologists are particularly interested in understanding the nature of these imprints, and how it is possible to modify DNA in some way that is heritable but can then be reset. My own interest has been understanding why such odd behavior should evolve. I've been trying to find situations in which what is best for genes of maternal origin is different from what maximizes the fitness of genes of paternal origin.
The best way to understand the underlying theory is with a famous anecdote accredited to J.B.S. Haldane, the great British geneticist, who is said to have claimed that he would give his life to save more than two drowning brothers or more than eight drowning cousins. The logic is that if Haldane is only concerned with transmitting his genes to future generations, this is the right thing to do. On average, a gene in his body has one chance in two of being present in a brother. If he sacrificed the copy of a gene in his body to rescue three brothers, on average he'd be rescuing one and a half copies of the gene in his three brothers; placing him ahead in the genetic accounting. But when it comes to cousins, each only has one chance in eight of carrying a random gene in Haldane's body. To benefit from the sacrifice of one copy of a gene in himself, he needs to rescue nine or more cousins. This was formalized by Bill Hamilton in his theory of inclusive fitness.
My theory can be illustrated by rephrasing Haldane's question and asking: Would Haldane sacrifice his life for three half-brothers? For the sake of the story let's say that these are his maternal half-brothers—offspring of his mother but with different fathers. The traditional answer to that question is no, because if you pick a random gene in Haldane, it's got one chance in four of being present in a half-brother. Thus, a random gene would have an expectation of rescuing three quarters of a copy—three times one quarter—for the loss of one copy in Haldane. However, if imprinting is possible, genes may have information about their parental origin, and this can change the accounting.
From the point of view of a maternally derived gene in Haldane, the three half-brothers are all offspring of his mother, so his maternally derived genes have a probability of one-half being present in each half-brother. For the sacrifice of one copy of the gene in himself, Haldane would be rescuing one and a half copies, on average, of his maternally derived genes. Natural selection acting in that situation on genes of maternal origin would favor the sacrificial behavior.
However, things look very different from the point of view of Haldane's paternal genes. Those three half-brothers are the offspring of different fathers, making them complete non-relatives. If genetic accounting were all that was important, no sacrifice, no matter how small, would justify any benefit, no matter how great, to his paternal half-sibs. Therefore, in this case, selection on paternally derived genes would prevent Haldane performing this sacrificial action.
This illustrates that different selective forces can act on different genes within an individual, pulling him in different directions, resulting in internal genetic conflicts. I suspect that how these conflicts are resolved is a matter of history, genetic politics, and knowing the details of the system. To answer questions like these, a lot of insight is going to come from the social sciences. Political science in particular is all about dealing with conflicts of interest within society with the formations of parties and factions, and I believe that if there are conflicts within the individual, you'll have a similar sort of internal politics.
I'm particularly interested in looking at situations in the real world where the Haldane story I just gave would apply—where there are potential conflicting selective forces acting within the individual. So far I've talked about conflicts between genes of maternal and paternal origin, but there are also possible conflicts between genes sitting on the sex chromosomes and genes sitting on the other chromosomes, or between genes sitting in the nucleus and genes sitting in mitochondria, or between our genetic inheritance and cultural transmission. I'm trying to develop a set of theories and tools for dealing with such situations.
Genomic imprinting is a fascinating phenomenon, and raises an interesting question: If information about the sex of the parent in the previous generation can be transmitted by such mechanisms, is there other historical information input from the environment that can be transmitted to the current generation and influence genetic expression? Would it be possible that if my great-grandmother experienced a famine or lived in a time of war, that this has put an imprint on the genome which is influencing gene expression in my own body?
My interest in genetic imprinting began while I was completing my doctorate at Macquarie University in Sydney. I began studying plant ecology and, in particular, how regeneration after fire takes place. I wandered around the bush a bit looking at plants, but my heart really wasn't in that. Through good fortune I got an opportunity to do a theoretical study on the evolution of the life cycles of plants, applying kin selection theory—the theory of parent-offspring conflict developed by Robert Trivers—to plants. By thinking about what's happening within seeds, I essentially had a theory of genomic imprinting ready to go the moment I heard of the phenomenon.
In a 1974 paper on parent-offspring conflict Trivers pointed out that there was often an implicit assumption that what was good for a parent was also good for the offspring. In terms of genetic transmission, it would seem that offspring are parents' stake in the future, so parents should be doing their best for them. What Trivers argued, however, was that parents would be selected to maximize their total number of surviving offspring—which may be quite different from maximizing the survival of any particular individual offspring. He suggested that there is a tradeoff between producing lots of offspring and investing relatively little in them versus producing a small number of offspring and investing a lot in each. He thought that over evolutionary time offspring would begin to compete with their siblings for available resources. And in turn, sibling rivalry would result in conflict between offspring and parents, since over time offspring would be selected to try to get more than their fair share of resources from their parents—more than the parents were selected to supply—whereas parents would be selected to spread their resources more evenly over a larger number of offspring. Trivers's theory was that this could lead to evolutionary conflicts.
I was asked to talk at the National Institute of Health in a workshop on imprinting and human disease. My goal was to suggest how evolutionary theory would provide new insights into human disease. An obvious case was in human pregnancy, where Trivers's theory of parent-offspring conflict could help to understand why pregnancy is so often associated with medical complications. Since then, looking at maternal fetal interactions has been another area in my research.
Trivers's theory has a lot to say about why pregnancy doesn't work particularly well. If we look at most of the products of natural selection, like the hand, the liver, the heart, or the kidney, these are wonderful bits of engineering that function very well for 60 or 70 years. But why are there so many problems in pregnancy? Pregnancy is absolutely essential to reproduction, so you might expect that this would be one part of our human physiology that had been perfected by natural selection. But there is an important evolutionary difference between the function of the heart and what's going on in pregnancy. When we look at the selective forces acting on the function of the heart, there's no evolutionary conflict. All of the genes involved in the development and function of the heart belong to the same genetic individual and, in a sense, have the same genetic interest: the maximization of the number of offspring of that individual. In the absence of conflict we've got a simple optimization problem, and you get an optimal solution.
But in the relationship between mother and fetus—because of the parent-offspring conflict that Trivers pointed out—we've now got conflicting forces. The offspring is being selected to take a little bit extra from the mother, and the mother is selected to resist some of the offspring's demands. Those selective forces tend to act at cross purposes and cancel each other out.
One very important problem during pregnancy is the communication of information between mother and offspring. In communication within the body there's no conflict, since selection causes cells to send messages as cheaply and as efficiently as possible. But when you're looking at the exchange of messages between mother and fetus, there's a problem of credibility, since their interests are not identical. In some situations, there's an evolutionary incentive to send misleading messages, and corresponding selection for receivers to distrust messages being received.
One thing that's happening during pregnancy is that there's a lack of the usual feedback controls, checks and balances. I read grant applications for scientists proposing to study maternal-fetal relations, and they tend to portray it in very rosy terms, as an almost loving exchange of messages between mother and fetus. But in pregnancy an embryo implants itself in the abdominal cavity or in the fallopian tube—in a completely inappropriate position in the body—and develops autonomously in the absence of any appropriate maternal messages. I believe there's actually very little communication going on between the mother and the fetus during the pregnancy. Rather, you're looking at various fetal attempts to manipulate maternal physiology and metabolism for fetal benefits.
During pregnancy the mother's hormonal communication systems are coming under joint control of both the mother and the fetus. The fetus secretes a number of hormones into the mother's body to achieve various effects, particularly increasing the nutrient levels of the maternal blood. In the early stages of human pregnancy, the embryo embeds itself in the uterine wall and taps into the maternal blood system, releasing hormones into maternal blood that can influence the mother's physiology, blood sugar levels, and blood pressure. The higher the levels of sugar and fats in maternal blood, the more nutrients the fetus can obtain. Typically, hormones are molecules produced in tiny amounts that have big effects, at least when communication occurs within a single body and there is no conflict between sender and receiver. However, in pregnancy, one individual (the fetus) signals to another (the mother) and there is potential for conflict. Natural selection favors increased production of the hormones by offspring to get a bigger effect, while at the same time it favors maternal receiving systems that become more and more resistant to manipulation. There is thus potential for an evolutionary escalation that sometimes results in placental hormones being produced in absolutely massive amounts. It's estimated that about a gram a day of human placental lactogen is secreted into the maternal blood stream, and yet it has relatively minor effects.
I think this observation, that placental hormones tend to be produced in very large amounts, is the best evidence for the existence of maternal-fetal conflict. The fetus secretes these hormones into the mother's body in an attempt to persuade the mother to do something that she might not necessarily want to do. Think of placental hormones as the equivalent of the junk mail that you get in your mail box. These messages are trying to persuade you to do something. They're relatively cheap to produce so they're distributed in vast quantities but have relatively minor effects. They must work sometimes, but it's very different from the sort of intimate whisper you might get between two individuals who have common interests.
The most successful application of my ideas on imprinting has been to the study of growth during pregnancy, and the prediction that paternally derived genes are selected to produce larger placentas that extract more resources from mothers. But the basic idea of the theory applies to any interactions among relatives that are what I call asymmetric kin; that is, relatives on the maternal side of the family but not on the paternal side, or vice versa. I suspect that genomic imprinting is going to be relevant to understanding the evolution of social interactions. There's also evidence now that imprinting is implicated in some forms of autism. There are a number of imprinted genes that are known to be imprinted in the brain, and I'm interested in exploring those ideas.
The most exciting empirical work that's been done to test my ideas came out of Shirley Tilghman's lab before she became President of Princeton. Hers was one of the first labs to describe an imprinted gene. Paul Vrana, a postdoc of Tilghman's, looked at crosses between two species of mice, one of which had a very high rate of partner change—multiple fathers within a litter—whereas the other was a so-called monogamous mouse, where a single father fathered all the offspring in a litter and the female had about an 80 percent chance of staying with the father to produce the next litter. The researcher predicted that the conflict between maternal and paternal genomes would be more intense in the mouse with multiple paternity than in the monogamous mouse, and in fact, when you cross them you get a dramatic difference in birth weight.
If the father came from the species with multiple paternity, there had been intense selection on paternal genomes to extract more resources from mothers. This paternal genome would be matched against a maternal genome that had not been strongly selected to resist paternal demands. In this direction of the cross, offspring were larger than normal, whereas in the reciprocal cross where the paternal genome came from the monogamous species and the maternal genome from the polyandrous species, offspring were smaller than normal. Paul Vrana was able to show that this difference was largely due to imprinted genes in these two species. This suggests that divergence of imprinted genes may contribute to the speciation process, and in particular that changes in social systems and mating systems can cause changes in the expression of imprint. These can then contribute to reproductive isolation between sister species.
The second bit of work is being done in, of all places, a liver oncology lab at the Duke University Medical Center that is studying genomic imprinting. Out of curiosity, Randy Jirtle and Keith Killian looked at marsupials and then at the platypus—an egg-laying mammal—to see where imprinting arose. They found that imprinting is absent in the platypus, at least for the genes they looked at, but was present in marsupials. Thus, imprinting appears to have arisen more or less coincident with the origin of live birth, before the common ancestor of marsupials and placental mammals. There are some exciting areas of research of that kind.
There are also some other recent intriguing observations out there that beg for a theoretical explanation. There's evidence in the mouse, for example, that the paternal genome particularly favors development of the hypothalamus, whereas the maternal genome favors development of the neocortex. I've suggested that some maternal-paternal conflicts can be seen within the individual between different parts of the brain favoring different sorts of actions. I don't have a good explanation of why that's occurring in the mouse, but I would love to know. At a broader level, perhaps these theories have something to say about the subjective experience of internal conflicts—why we sometimes have great difficulty making up our minds. If the mind were purely a fitness-maximizing computer with a single fitness function, then this paralyzing sense of indecision we often feel would make no sense. When we are forced to make a difficult decision it can sometimes consume all our energies for a day, even though we'd be better off making a decision one way or the other. Perhaps that can be explained as a political argument going on within the mind between different agents with different agendas. That's getting very speculative now, though.
In the future I'd also like to get back to plants. I've put a lot of work into thinking about plant life cycles, and the work that I did in my Ph.D. has had relatively little impact, so I'd like to go back and rethink some of those ideas. I've thought of writing a book called Sociobotany that would do for plants what Trivers, Wilson, and Dawkins did for animal behavior. Botany tends to look at the different stages in the life cycles of a plant as cooperating one with the other. But Trivers's theories of parent-offspring conflict are very relevant to understanding some odd features of seed development and the embryology of plants. One of my favorite examples of this phenomenon can be seen in the seeds of pine trees and their relatives. The seed contains multiple eggs that can be fertilized by multiple pollen tubes, which are the functional equivalent of sperm. Within the seed, multiple embryos are produced that then compete to be the only one that survives in that seed. As this happens there's very intense sibling rivalry and even siblicide going on in the seed. Because of oddities of plant reproduction, the eggs that produce those embryos are all genetically identical one to the other, so all the competition among the embryos is between the genes that they get from their fathers through the pollen tube. Because of this, I expect there to be imprinting in the embryos of pine trees.
Another interesting case is found in Welwitschia, a very odd plant that grows in the Namibian desert. Here, once again because of oddities of the plant's genetics, the egg cells are no longer genetically identical one to the other, and they compete with each other to produce the embryo that survives in that seed. Rather than waiting for the pollen tube to reach the eggs, the eggs grow in tubes up to meet the pollen tubes. There's actually a race to meet the pollen tubes growing down to meet the eggs. Fertilization occurs and then the embryos race back down into the seed to gain first access to the food reserves stored in the seed. This odd behavior was just a strange observation of plant embryologists, but I think the application of ideas of conflict between different genetic individuals gives a very pleasing explanation of why you observe this behavior in Welwitschia but not in other groups where the eggs are genetically identical to each other.
Some of these ideas also intersect with the work of evolutionary psychologists. Although I don't interact with them on a daily basis, they're very keen on my work, and I follow theirs. A true psychology has got to be an evolutionary psychology. Whether every theory that goes under the name of evolutionary psychology is evolutionarily justified is a different question, but in terms of the question whether Darwin is relevant to understanding the mind and human behavior, evolutionary psychologists have got it right. We are evolved beings and therefore our psychology will have to be understood in terms of natural selection, among other factors.