Evolution of genomic imprinting

Diploid sexual organisms inherit one copy of each gene maternally, and one paternally. At some loci, only one of the two copies is expressed; which copy is expressed is determined by whether it is inherited maternally or paternally. Evidence from mice and Arabidopsis suggests that this imprinting process is mediated by cytosine methylation 15 17 21. Additionally, the parent-specific pattern of expression appears to be established during gametogenesis, through a mark on the region to be methylated 14. Although the mechanisms of imprinting are well-studied, the forces promoting the evolution of imprinting, and controlling the evolution of imprinted genes are not entirely clear 19 13.

A number of imprinted genes play a role in the fetal growth of mammals, and are imprinted primarily during the early stages of development (e.g., Igf2 4, Igf2r 2, p57^KIP2 11). Similarly in plants, imprinting appears to be crucial to endosperm development (a nutritive structure in the seed), but has little influence on the viability of the adult plant. Further, all known imprinted plant genes are imprinted in the endosperm 1. The Arabidopsis gene, MEDEA (MEA), which suppresses endosperm growth 5, is one example of a gene that is maternally expressed in the endosperm, but is biparentally expressed in the embryo and adult tissues10.

The most widely discussed theories for the evolution of imprinting involve a resource allocation conflict between a mother and her offspring 20. Because larger offspring are more viable (at least in seed plants 9), there will be selection for offspring to acquire as many maternal resources as possible. Conversely, there may be selection for a mother to distribute resources among all of her offspring. Additionally, if there is multiple paternity within or between broods, there may be a conflict of interest between maternal and paternal genomes, with selection for strong resource acquisition by the paternal genome. Mathematical models have shown that a conflict over resources can permit the evolution of imprinting. Theories predict that an allele which enhances offspring growth when paternally inherited, and/or represses offspring growth when maternally inherited can invade a population through selection 20.

There are several reasons to suspect that the sequence evolution of imprinted genes may differ from that of other non-imprinted genes. Spencer 19 notes that although imprinted genes are expressed in a non-Mendelian fashion, their inheritance is standard. Thus, their effective population size is no different than that of non-imprinted autosomal genes, and the strength of genetic drift should not be influenced by imprinting. However, there should be differences in the effect of selection and mutation. For instance, deleterious mutations in an imprintable gene will be hemizygous whenever they are expressed and are therefore similar to dominant or partially dominant mutations in non-imprintable genes 18. Spencer 18 predicts that if mutation rates are higher in one sex than the other, the genetic load of genes silenced when maternally inherited should differ from the genetic load of genes silenced when paternally inherited. Because a mutation occurring in an imprinted gene copy will not be expressed immediately, and because of the hypothesized higher mutation rate in the male germ line of animals, genes that are paternally inactivated are expected to carry more deleterious mutations than genes that are maternally inactivated. This is a point in which plants and animals would be expected to differ, because in plants, the soma and germplasm are not separate. Many somatic cell divisions precede gamete production, with no systematic differences in the number of cell divisions leading to male and female gametes 23.

Only a few papers have studied the sequence evolution of imprinted genes, relative to non-imprinted genes (below), and none of those studies consider species-level diversity. Empirical research has been primarily concerned with testing the hypothesis that conflict for resources created antagonistic coevolution between growth enhancers and suppressors, that this conflict provided the conditions for the evolution of imprinting, and that conflict is still driving the evolution of imprinted genes.

Hurst and colleagues 12 16 expected to see rapid rates of evolution in imprinted genes if antagonistic coevolution was driving their evolution. Some evidence from sequence data in mammals is consistent with this prediction; a number of genes involved in maternal-fetal interactions, but of unknown imprinting status, show high rates of sequence evolution (placental lactogens 1 and 2, growth hormone, growth hormone releasing factor, and prolactin; reviewed in 6). However the 22 imprinted genes analyzed 12 16 in a comparison of mouse and rat do not show rapid sequence evolution, relative to non-imprinted genes. A more detailed analyses of one imprinted gene, the insulin-like growth factor II receptor (Igf2r), revealed that the signal sequence, which controls cellular localization, is evolving more quickly than the signal sequences of non-imprinted genes, and at a rate comparable to that of rapidly evolving immune-system genes 16 22. Thus, it was suggested that the signal sequence may be involved in antagonist coevolution concerning the cellular localization of Igf2r 22. Additionally, imprinted genes were found to have fewer and smaller introns than other genes, suggesting that imprinted genes may have lost their introns due to selection for rapid transcription, under antagonistic coevolution 7.

The above data suggest that imprinted genes could be involved in an arms race due to intragenomic conflict. However, the results are not overwhelming. Further, there appear to be no empirical attempts to examine the molecular evolution or population genetics of naturally imprinted plant genes. Yet, imprinted plant genes may evolve quite differently from imprinted animal genes. Besides the lack of male-driven evolution in plants, there are other important differences. Chakrborty 3 showed that at imprinted loci, the population genetics of recessive mutations is similar to that of dominant or partially dominant mutations at non-imprinted loci. However, in plants, unlike animals, many genes are expressed hemizygously in gametophytes. Thus, there may be little difference in the evolution of imprinted and non-imprinted loci in plants. Another difference between plants may be in the effect of methylation on the rate at which mutations are generated. Methylated cytosines are 10-20 times more mutable than unmethylated cytosines 8. Because imprinting in plants occurs primarily in the endosperm, which does not contribute to the germ line, mutation should be generated at the same rate as in non-imprinted genes. In imprinted mamallian genes, however, it was found that mutation rates from C to T were positively correlated with methylation density, and that methylation-induced mutational patterns could explain some of the variation in Ks across imprinted genes16. Finally, the high level of mating-system variation among plant species provides the opportunity to study the effect of selfing rate on the population genetics of imprinted genes. Thus, comparison of sequence evolution in imprinted plant genes to that of imprinted animal genes is likely to be informative.

We are currently examining the population genetics and sequence evolution of several imprinted plant genes involved in seed growth.