The Enhancer Competition Model to Explain Reciprocal Imprinting
The Igf2, and H19 genes (open boxes), their transcriptional orientation, and chromosomal transcription site (horizontal arrows). Two enhancers which lie downstream 3' of the H19 gene are indicated by the closed circles, and the arrows leading from them indicate which of the genes is transcribed on the maternal and paternal chromosomes. Allele-specific methylation sites are indicated by the CH3 symbols.
Embryological and genetic studies conducted over the past 15 years have suggested that the male and female genomes of mice are not inherited in an equivalent state. The reason for this nonequivalence is now clear: there are several genes that are inherited in silent states from either mothers or fathers. Two such genes are insulin-like growth factor 2 (Igf2), a fetal-specific growth factor expressed almost exclusively from the paternal chromosome, and H19, an unusual gene that codes for a 2.5-kb RNA and whose expression is exclusively dictated by the maternal chromosome. The genes lie within 90 kb of DNA of one another and are co-expressed throughout dev-elopment. This led us to propose that the genes are reciprocally imprinted through competition for a common set of regulatory elements, presumably enhancers (Fig. 1). The bias toward Igf2 on the paternal chromosome is set up by sperm-specific DNA methylation of the promoter and structural H19 gene, which acts to suppress its transcription. We propose that on the maternal chromosome, which is apparently unmethylated at either gene, the H19 gene is favored because it is closer to key regulatory elements and/or it has an inherently strong-er promoter. To test this model, mutations were generated in the H19 gene itself, as well as in the presumptive regulatory elements that govern expression of H19 and Igf2.
Q: What was the starting point for this work?
A: The starting point for this work was not a fascination with imprinting, but a fascination with a mysterious gene. We had cloned the H19 gene in the course of trying to understand the regulation of another gene, a-fetoprotein. H19 bore no resemblance to any other gene that had been described before. Although it was transcribed and processed through the classical mRNA pathway, it lacked an open-reading frame of any length. Normally, that would have been enough to discourage us from pursuing it further because the most likely conclusion was that it coded for a pseudogene. However, two things kept us in the game: the fact that the RNA was transcribed at a very high rate during embryogenesis and the fact that the RNA structure (but no open-reading frame) was conserved in evolution. Finally, the observations that led us to H19's imprinting were that extra copies of the gene were lethal in mice, suggesting that its dosage was being carefully controlled, and that it mapped to an imprinted region of the mouse genome.
Q: Which findings have been most surprising to you or to other scientists?
A: The first surprise to us was that the H19 gene was imprinted in the opposite direction from its nearest neighbor, Igf2. The paternal-specific expression of Igf2 had just been uncovered by Argiris Efstratiadis, Tom DeChiara, and Elizabeth Robertson at Columbia University in New York City, and we had guessed that if H19 was imprinted, it would be imprinted in the same direction. The fact that it was not, immediately eliminated the X-chromosome-inactivation model, whereby a chromosomal domain is simply shut down. The reciprocal nature of the imprinting of Igf2 and H19 led directly to the enhancer-competition model.
Q: What were the greatest stumbling blocks, and what new observations, techniques, reagents, or insights helped you to get past them?
A: The general problem in studying imprinted genes is devising ways to examine the expression of each allele separately. Until our work with H19, the only two imprinted genes that had been identified were uncovered by mutations. In the case of Igf2 the mutation was generated on purpose using homologous recombination. For the other gene, the mannose-6-phosphate receptor, Denise Barlow and her colleagues in Vienna took advantage of a spontaneous mutation in the form of a large deletion. But for H19, there was no useful mutation available and, therefore, we used intraspecific and interspecific F1 hybrids between species of mice that had been separated over an evolutionary time scale -- 3 to 6 million years. This was enough time for point mutations to arise in the H19 gene, and these mutations could be detected by a sensitive RNA-protection assay in which even single base differences between species could be detected. This method, as well as polymerase chain reaction - based methods that exploit the same heterogeneity, are now the standard methods in the field for following the specific expression of the maternal and paternal alleles of a gene in a wild-type animal.
Q: Do you see any potential areas where this research might provide insight to clinical scientists?
A: A great deal of important work has implicated defects in imprinted genes in human disorders. The best studied of these are the Prader-Willi and Angelmann syndromes, which are tightly linked on human chromosome 15. What is fascinating about these disorders is that Prader-Willi is inherited paternally, whereas Angelmann is inherited from mothers; thus, they behave as another pair of reciprocally imprinted genes. Recently, Ute Francke's group at Stanford in Stanford, Calif. and Arthur Beaudet's group at Baylor in Houston have identified nonprotein-coding RNAs in the region, implying that many of the features that we have discovered at Igf2/H19 may also exist at this locus. Furthermore, the Igf2/H19 region on human chromosome 11 has been implicated in Beckwith-Weidemann syndrome. Both the sporadic and inherited forms of this syndrome display parental bias, suggesting the gain-of-function of a growth-promoting gene, possibly Igf2.
Q: How are you following up on this work, and what questions would you ultimately like to answer?
A: The most important immediate question that our work raises is whether noncoding genes such as H19 function through a purely transcriptional mechanism, as implied in Figure 1, or whether the product of the gene plays a role in this or any other biological process. Excitement about H19 itself was increased with the discovery of another nonprotein-coding RNA, Xist, which maps to the X-chromosome-inactivation center. This site is required in cis for an X-chromosome to be inactivated. Unlike any other gene on the X chromosome, Xist is exclusively transcribed from the inactive X chromosome, and this has led to much speculation about its role in X-chromosome-inactivation.
The long-term puzzle I would like to solve is the function of imprinting. From a genetic point of view, imprinting is a dangerous process because it renders the organism functionally hemizygous for the imprinted gene. One has to assume that there is a compensatory benefit. The most interesting model for what this might be was proposed by David Haig, an evolutionary biologist and currently a Fellow at Harvard University in Cambridge, Mass. He suggested that imprinting is the consequence of a tug-of-war between the male and female genomes for the preferential inheritance of their genes in the next generation. In nonmonogamous species in which the embryo is consuming maternal resources, the male's interest is best served by having his immediate progeny consume those resources at the expense of the mother, with whom he will not likely mate again. The mother, on the other hand, must conserve her resources in any one litter if she is to reproduce again. We are testing this model by examining the evolution of imprinting within mammals, concentrating on marsupials and monotremes, which have very different reproductive strategies from each other.