If a mother sometimes has offspring by more than one father and if genes in the offspring are active in acquiring resources from maternal tissues, theory predicts that alleles at some loci in the offspring will evolve different patterns of gene expression depending on the gene’s parent of origin (genomic imprinting). The criteria for the evolution of imprinting are satisfied in many seed plants, and imprinting has been reported from the endosperm of angiosperm seeds. This paper’s purpose is to show that imprinting phenomena in endosperm can provide a coherent explanation of some failures of experimental crosses, and of the prevalence of pseudogamy among apomictic angiosperms. As a consequence of imprinting, seed development comes to depend on a particular ratio of maternal and paternal genomes in endosperm. This ratio is normally two maternal genomes to each paternal genome. Imprinting probably accounts for the failure of crosses between diploids and their autotetraploids, because the 2m: 1p ratio is disturbed in such crosses. Imprinting may also account for the breakdown of endosperm in crosses between related species, if the expression of maternal and paternal genomes in endosperm is out of balance. When a cross fails because of such an imbalance, the reciprocal cross will have the opposite imbalance and a complementary phenotype would be expected. The embryological evidence is consistent with this prediction. For example, many incompatible crosses show delayed wall formation in one direction of the cross, but precocious wall formation in the other direction. Typically, seed development can be classified as showing ‘paternal excess’ or ‘maternal excess’. Paternal excess is associated with unusually vigorous early growth of the endosperm, and maternal excess with the opposite. This pattern is consistent with natural selection on paternal gene expression favouring larger seeds. Genetic evidence from maize confirms an association between paternal gene expression and larger kernel size, and maternal gene expression and smaller kernel size. Genomic imprinting creates a requirement for both maternal and paternal genomes in imprinted tissues. In mammals, imprinting is expressed in derivatives of the zygote. The requirement for a paternal genome has constituted a block to the evolution of parthenogenesis, because all the genes in a parthenogenetic embryo are maternal. In angiosperms, imprinting is primarily expressed in the endosperm rather than the embryo. If the effects of imprinting in the embryo are small, an asexually produced embryo can develop, provided that it is associated with a viable endosperm. Many a Pom^ ts are pseudogamous. That is, the endosperm is fertilized and contains maternal and paternal genes embryo is asexual and contains maternal genes only. Thus, the division of labour between the embryo the endosperm during development of the seed can be seen as a preadaptation for apomixis. Some apomicts are autonomous. That is, the embryo and the endosperm both develop without fertilization. Genomic imprinting in endosperm would seem to constitute a barrier to the evolution of autonomous apomixis. Thus, there is a problem, not previously appreciated, in understanding how autonomous apomixis is possible
The Rab Geranylgeranyl Transferase Beta Subunit Is Essential for Embryo and Seed Development in Arabidopsis thaliana