WHEN YOU THINK OF DOMESTICATED ANIMALS, THE GUINEA PIG is not one of the first that comes to mind. Yet the guinea pig was domesticated a thousand years before the horse—not as a pet but for food. In fact, to this day, guinea pigs remain a dietary staple in the Andes of Peru and Bolivia, where they were originally domesticated. It was only thousands of years later, after they were transported to Europe during the seventeenth century, that they were made into pets, and later the paradigmatic subjects for scientific experiments.
It was in part because of their circuitous route—by ship—to European ports that guinea pigs acquired their common moniker. And a curious name it is. The guinea pig is neither from Guinea, nor a pig. The “guinea” in “guinea pig” refers to the Guinea coast of West Africa, a stopover and resupply point for European ships heading to and from South America. Though the Guinea pigs were originally loaded on the ships in South America, many Europeans mistakenly assumed that they had been supplied in Guinea. European sailors, like the Andean people who domesticated them, loaded their ships with guinea pigs because they were a good source of protein for the long journey. Some of those that didn’t get eaten became the first guinea pig pets.
The “pig” in “guinea pig” probably derives in part from the fact that they didn’t look anything like European rodents—mice, rats, and such—though their resemblance to pigs is not all that evident either. But so they seemed at least to the father of scientific classification, Linnaeus, who baptized them with the species name porcellus (the Latin word from which “pork” is derived), perhaps because their bodies are fairly squat and their tails tiny.
The guinea pig is actually a type of cavy, a group of South American rodent species that generally inhabit high-altitude grasslands. The bulk of a wild cavy’s diet is grass; indeed, they are considered by some the ecological equivalent of cows. But guinea pigs are not as grass bound as cows; they can thrive on a diverse diet, which is one reason they are easy to maintain in captivity. In the Andes they often have the run of the house until the dinner bell tolls.
In the eyes of well-fed Europeans, guinea pigs seemed irresistibly cute and cuddly. Hence their conversion in Europe from food to pets. During this second round of domestication, guinea pig fanciers selectively bred for a variety of coat colors as well as for long and curly hair, which diverged quite dramatically from the wild type. It was primarily because of this variation in pelage that guinea pigs became the first mammalian model for genetic studies.
William Castle and Sewall Wright
At about the same time that Morgan was setting up his Fly Room at Columbia University (see Chapter 2), William Castle set out to do comparable genetic research at Harvard. Somewhat ironically, it was Castle who first saw in fruit flies their potential value for genetics,1 but unlike Morgan he chose to stick with mammals. Castle had labs devoted to rabbits, mice, and rats, but he had a special fondness for guinea pigs. He became so enamored that he traveled to South America to collect their wild relatives for his breeding experiments. Guinea pigs are far from ideal subjects for genetic studies, though, far less suitable certainly than fruit flies. Fruit flies can cycle through fifty generations in a year, guinea pigs two, maybe three, if you are lucky. Morgan’s group had identified over one thousand mutations before Castle’s group found ten. Yet guinea pigs have some advantages over fruit flies, if only that any mutation is easier to spot. Mutations in coat characteristics are especially easy to see.
Castle had a hands-off managerial style similar to Morgan’s. He gave his students great leeway in selecting their projects. And though Castle’s lab was not as teeming with future luminaries, it did produce several who became members of the National Academy of Sciences, including one very special student named Sewall Wright, whose contributions to genetics were unsurpassed in their breadth.2 Wright not only greatly enlarged the scope of classical genetics as it relates to heredity; he also thought carefully about the gene as a physiological element and how a gene could developmentally influence a trait like coat color. For this reason, he can be considered one of the fathers of developmental genetics. But Wright is best known as one of the co-founders of the field of population genetics, through which he had a huge influence on evolutionary biology. Wright had a unique set of aptitudes and interests, which made all this possible.
Wright, who was something of an autodidact, also came from a different biological tradition than Morgan and his students: physiology. One consequence of his background was his concern with what genes actually do physiologically. Of course, in this premolecular age, in which the physical gene had not yet been characterized, such inferences were extremely indirect. Wright nonetheless was remarkably prescient as to the physiological actions of the genes he studied. Moreover, because of his concerns about the concrete physiological gene, as opposed to the Mendel-Morgan notion of genes as abstract particles of inheritance, Wright was less inclined to try to shoehorn seemingly anomalous results into the Mendelian framework.
In many breeding experiments conducted during this period, there would be slight to sizeable departures from what would be expected from Mendel’s laws. The sizeable departures were taken seriously, but lesser departures were generally considered within the margin for error. Wright, though, chose to emphasize this residual variation rather than ignore it. Ultimately, this tactic caused Wright’s view of genes and gene actions to diverge substantially from the mainstream of classical genetics. In particular, he tended to a more complex view of genetic inheritance than was the norm of the time. First, Wright came to believe that the inheritance of traits such as coat color involved more loci and alleles than most geneticists. Second, he emphasized the fact that alleles at different loci can interact in complex ways that cannot be calculated from their separate effects, a phenomenon known as epistasis.
Though now widely accepted, this view was not at all well received by his contemporaries. Wright had other ahead-of-his-time ideas as well. For example, Wright was much more open to the possibility that the environment contributed to coat color and other “genetically determined” traits, and was a pioneer in the study of gene-environment interaction. This, too, ultimately stemmed from his being steeped in the physiological or developmental view of genes and gene actions. Because Wright viewed genes as physiological entities, it was much easier for him to envision that the effect of a given gene can depend on environmental factors that also influence the same physiological processes.
But perhaps most heretical was the importance that Wright ascribed to random events in modulating gene effects and development. Wright saw randomness at every level of biological processes, from the biochemical to that of the whole guinea pig, and beyond that, to whole populations of guinea pigs.
Wright’s somewhat iconoclastic brand of (physiological) genetics was always overshadowed by the classical school, which led from Morgan to Watson and Crick, but it was Wright’s approach to genetics, not Morgan’s, that eventually laid the foundation for epigenetics. The study of epigenetic inheritance, the subject of this chapter, certainly owes much more to Wright than it does to Morgan.
The Agouti Gene
By the time Wright commenced his research on guinea pigs, several genes were known to contribute to coat color. Wright devoted much of his early career to figuring out how they interact to produce the various color patterns guinea pig fanciers had created, plus some additional ones that Castle had created by hybridizing domesticated guinea pigs with their wild progenitors. Wright began with the Mendelian assumption that each color-related gene (locus) acted independently of the others and that there were two variant alleles at each locus: the wild-type allele, and a mutant allele made more common through the selective breeding of guinea pig fanciers. He further assumed, like Morgan, that the wild-type alleles were dominant and the new mutant alleles recessive. These Mendelian assumptions served Wright well for the most part, but there was always a substantial amount of residual variation that could not be explained within the standard Mendelian framework.
Wright’s research on the agouti locus is particularly germane for our purposes here.3 The agouti locus is named for the color pattern displayed by agoutis, which are basically long-legged versions of guinea pigs. The wild-type agouti allele, call it A, is usually associated with a distinctive color pattern. Each hair starts out black; that is, it has a black tip. As the hair continues to grow, however, it becomes yellowish to red, then black again at the base. These banded hairs, it turns out, are typical of most wild mammals, not just agoutis, including the wild progenitor of guinea pigs. And the agouti gene is found in all mammals including humans.
Breeders had subsequently selected for a mutant allele, a, that caused the width of the yellow band to increase at the expense of the black, resulting in various reddish-yellow color patterns. But these two alleles could not account for all of the variation in hair banding. Wright demonstrated that there was at least one more genetic factor involved, a second mutant allele at the agouti locus. But even given this third allele, there was still unexplained variation. Wright was certainly willing to acknowledge that environmental factors might be complicating things, but given the state of the art of that time, he could not have envisioned how. He would not, however, have been surprised at the answer. It falls quite neatly into his worldview.
Though Wright continued to use guinea pigs for his genetic studies, most subsequent research at the agouti locus was conducted on mice. Wright’s original research on the agouti locus in the guinea pig laid the groundwork for the research on this same gene in mice, including the recent epigenetic research that we will explore here.
In the nineteenth century, the pet trade in mice rivaled that of guinea pigs. Mouse fanciers had also uncovered recessive mutations in the agouti locus that caused the yellow band to widen, resulting in a yellowish coloration. But the number of mutations at the agouti locus increased considerably when scientists took over the breeding. Moreover, some of these mutations were dominant to the wild-type allele, A. One such mutation, lethal yellow (AL), was generally lethal. Other dominant mutations were not lethal. These included the mutation viable yellow (Ayr), so called because, unlike mice with the lethal mutation, mice with this mutation survived, albeit with significant physiological defects. Pleiotropic is the term used to describe alleles such as Avy that have multiple physiological effects.
Pleiotropy simply reflects the fact that the protein products of most genes are expressed in more than one cell type. As such, a gene can participate in more than one physiological or developmental process. In this case, the most obvious developmental process—to the human eye—that the agouti gene participates in is the one that determines hair color. The agouti protein affects hair color by interfering with the binding of the hormone that promotes melanin (associated with black pigment) production to its receptor.4 But melanin is produced in many cell types other than hair follicles; the agouti protein interferes with melanin production in all of them, including those found in the liver, kidney, gonads, and fat.5 The result of all this interference is lethal in AL (lethal yellow) mice and grossly compromises health in Avy(viable yellow) mice. Among the adverse health consequences of this mutation are obesity, diabetes, and various kinds of cancer.6
Unlike AL mice, which are always yellow, the coat color of Avy mice is quite variable, ranging from almost pure yellow to a wild-type coloration, called pseudoagouti. You can predict the health of a viable yellow mouse by its coat color. Those viable yellow mice with yellow coloration are obese, diabetic, and cancerous; those with the pseudoagouti coloration, however, have none of these defects.7
How do we explain this variable coloration and associated health defects of viable yellow mice? One explanation, consistent with Wright’s approach, concerns what has come to be called genetic background. For Wright, in contrast to most of his contemporaries, the effect of a gene (allele) like viable yellow on a trait such as coat color depends on a lot of other factors. Among these other factors are other genes. That is, the effect of the viable yellow allele on coloration depends, in part, on what other alleles are present at other genetic loci. Not all other loci, of course, but a lot more loci and alleles than most of his contemporaries were willing to acknowledge.
But the viable yellow allele has variable effects on coloration and health even when the genetic background is kept constant. That is, even genetically identical mice with this mutant allele are quite variable in coat coloration and health. Within a single litter of genetically identical viable yellow mice, you can find the full range of coat patterns from yellow, to mottled, to pseudoagouti, and with them the associated variation in health.8
Epigenetics at the Agouti Locus
The color differences in these mice, it turns out, are associated with differences in the epigenetic state of the viable yellow allele. It is unmethylated in the yellow mice but highly methylated in the pseudoagouti mice. Mottled mice are intermediate methylation-wise.9
So why do some of these genetically identical viable yellow mice have methylated viable yellow alleles, while others don’t? In part, it depends on the coat color, and hence epigenetic state, of the mother. Female mice with the yellow coat color tend to produce yellow offspring, never offspring with the pseudoagouti phenotype. Mothers with the pseudoagouti coloration produce few yellow offspring and more pseudoagoutis.10 Moreover, a grandmother’s coloration also influences the coat coloration of her grand-offspring.11 There is no relationship between the coat color of a father and that of his offspring.
This may sound familiar, like the transgenerational maternal effects on the stress response that we discussed in the previous chapter. But though these curious patterns of coat color inheritance certainly constitute a maternal effect, this one occurs much earlier in development. When the fertilized eggs of yellow mothers were transferred to black mothers, they still tended to be yellow at birth.12 So there is no effect of the intrauterine environment here. Instead, an epigenetic attachment to the Avy allele, one that alters coat color in otherwise genetically identical mice, has been directly transmitted from mother to pup. This is true epigenetic inheritance.
But how did this epigenetic variation arise in the first place? From previous chapters, we might expect some sort of environmental effect. In this case, diet appears to have a role. When pregnant viable yellowmothers are fed a diet high in methyl donors, such as folic acid, the spectrum of coat colors in their offspring shifts toward the pseudoagouti end.13 Moreover, when the offspring, which experienced the methyl supplementation in utero, themselves became mothers, the color spectrum shift is sustained in their offspring.14 The transmission of the diet-induced change to the grand-offspring occurred even though these second-generation mothers received no further methyl supplementation.
The shift in the color spectrum caused by a high methyl diet was quite modest. Most of the epigenetic and hence color differences in genetically identical viable yellow mice must result from other factors. One source for this variation, which is receiving increasing attention of late, is chance. Much of what causes one individual with this allele to be yellow or pseudoagouti—with all of the associated health consequences—may be essentially random processes at the molecular level that affect methylation of the allele.15 So in essence, we have a case of a partly random epigenetic event that can be inherited. That sounds very much like a mutation.
Why Epigenetic Inheritance Was Not Supposed to Occur
For years, it was thought that true epigenetic inheritance was impossible. During the process of making sperm and eggs, all epigenetic marks were thought to be removed during a process called epigenetic reprogramming.16 Any epigenetic attachments that survived this process are removed during a new round of reprogramming soon after fertilization. Hence each new generation starts with a clean epigenetic slate. Recently, however, it has been demonstrated that epigenetic reprogramming does not wipe out every epigenetic mark. Some epigenetic changes, including those induced by environmental factors, are not erased; they are transmitted to the next generation.
The agouti locus is one of the best-documented cases of epigenetic inheritance among mice, but there are a number of other known cases in mice as well. One such case involves the Axin gene, which when methylated results in a kinked tail.17 The methylation pattern and hence kinked tail can be inherited from both mother and father. A number of genes related to olfaction, especially the detection of pheromones, also display putative epigenetic inheritance.18 In humans, there may be epigenetic inheritance at a locus that promotes a particular form of colon cancer.19 Since it is only recently that cases of anomalous (by Mendelian standards) inheritance have been viewed through an epigenetic lens, we might expect more cases of epigenetic inheritance in humans and other mammals will be discovered in the near future.
But there are reasons to suspect that epigenetic inheritance is less common in mammals than in other life-forms.20 Good examples of epigenetic inheritance have been identified in creatures as diverse as fruit flies and yeast.21 But some of the most dramatic examples of epigenetic inheritance occur in plants.22
As an experimental subject, the plant equivalent of a mouse is an unprepossessing member of the mustard family known only by its scientific name, Arabidopsis thalinia. In the wild, Arabidopsis thrives in diverse habitats throughout Eurasia; it also thrives in laboratory environments. It is a highly variable plant with respect to its size and flowering time, among other traits. Both the size of the plant and its flowering time are inherited epigenetically. Let’s consider first an epigenetic factor that affects the size of the plant.
In many organisms, but especially plants, there is a tradeoff between growth and defense against infection by pathogens. The more resources devoted to pathogen defense, the slower the growth rate. So plants that find themselves in a particularly pathogen-rich environment tend to be dwarfs. Pathogen resistance in A. thalinia is mediated by a number of resistance (R) genes. There is one particular cluster, located on chromosome 4, that is subject to epigenetic regulation. An epigenetic variant called bal causes one gene in this cluster to be chronically active. The gene behaves like the plant is under attack even when it isn’t. Plants with this epigenetic variant are scrawny and bedraggled looking, their leaves withered, and their roots underdeveloped. But genetically identical plants that lack this epigenetic factor are robust, even when they are grown in identical environments.23 Before the advent of epigenetics, it would have been assumed that the dwarf was a mutant, that there was some change in the sequence of the R gene. Now we know that even such stark differences within a species can be caused by differences in the epigenetic regulation of gene expression.
Flowering in A. thalinia is also epigenetically regulated. In 1990, a mutation was identified in some wild populations of Arabidopsis that caused a delay in flowering.24 The mutant, called fwa, caused plants that would normally flower in the spring to flower instead during the summer or fall. Multiple genetic tests indicated that fwa was a classical dominant Mendelian trait, like brown eyes in humans. The mutation was subsequently mapped to a single gene encoding a transcription factor. But scientists were puzzled when they could not find any difference in the sequence of the fwa mutant and the normal FWAallele. The mutation, it turned out, was not a mutation but an epimutation, an altered methylation pattern. And this epimutation has been stable, inherited in a quasi-Mendelian manner, for many years.25
Transgenerational Epigenetic Effects
Epigenetic inheritance, such as occurs at the agouti locus and in A. thalinia, is but one form of what I will call a “transgenerational epigenetic effect,” by which I mean an epigenetic effect transmitted from parent to offspring and beyond.26 This broader category includes the social inheritance of the stress response in mice and other forms of nongenetic inheritance that have an epigenetic component. To qualify as epigenetic inheritance in the strict sense, though, the epigenetic attachment, or mark, must pass through the epigenetic reprogramming process intact. In the case of the lick-deprived rats’ stress response, the epigenetic attachments are reconstructed anew each generation; the original epigenetic alterations do not survive epigenetic reprogramming. This is probably true of most transgenerational epigenetic effects that result from the maternal or social environment. That would include the effects of the Dutch famine discussed in Chapter 1. There is no compelling evidence that the grandmother effect described there is true epigenetic inheritance. But in another study of dietary effects in humans, the case for true epigenetic inheritance is much stronger.
There is an isolated Swedish population for which very accurate records of crop harvests have been kept for hundreds of years, so scientists can calculate the average amount of calories consumed in a given year. One notable result of this study is an association between the calories consumed by men during their adolescent years and the health of their grandchildren. Paternal (but not maternal) grandsons of males exposed to famine before adolescence were less susceptible to cardiovascular disease than the grandsons of those who did not experience famine.27 In contrast to the epigenetic effects on birth weight resulting from the Dutch famine, this association cannot be attributed to the maternal environment. The only biological thing a male contributes to his offspring and grand-offspring is his sperm. So this appears to be a case of environmentally induced epigenetic changes that qualify as epigenetically inherited.
It should be noted that epigenetic inheritance at the agouti locus is not very precise or efficient. The correlation between parent and offspring, while significant, is not high. This parent-offspring correlation is in fact much lower than that for the stress response in rats, which does not result from true epigenetic inheritance.
It is different in plants. True epigenetic inheritance is much more common in plants than in animals and can be stable over hundreds of generations—as stable, in some cases, as genetic inheritance. The reason for the higher incidence of epigenetic inheritance in plants is that their epigenetic reprogramming is much less pervasive and thorough. Many more epigenetic marks make it through this process unscathed.
Wright’s Legacy
Sewall Wright’s work on the inheritance of coat color in guinea pigs, which continued throughout his long life, is a testament to both his close observation and a theoretical insight that was informed, but not encumbered, by established doctrine. Wright’s approach to genetics differed markedly from that of most of his contemporaries in its emphasis on the gene as a physiological and developmental factor. Most geneticists of his time, including Morgan and his group, preferred to view the gene as an abstract hereditary factor. The payoffs for Morgan’s approach were immediate, whereas most of the dividends resulting from Wright’s approach would become apparent only decades later. It was Wright’s approach, however, not Morgan’s, which laid the groundwork for developmental genetics and its offshoot, epigenetics.
More specifically, Wright’s research on the agouti locus became the foundation for subsequent research on its developmental role, from hair color to obesity. Ultimately the ongoing research on the agouti locus resulted in the first well-documented case of true epigenetic inheritance in a mammal.
As mentioned earlier, epigenetic inheritance in the strict sense is but one form of the broader phenomenon of transgenerational epigenetic effects, other forms of which—such as the social transmission of the stress response in rats—we explored earlier. We will explore another kind of transgenerational epigenetic effect in Chapter 9, a rather odd one; but first some background information will be helpful, which comes to us by way of the mysterious X chromosome.