Controlling elements are wild cards in the epigenomic deck
The genomes of multicellular eukaryotes are loaded with retrotransposons, parasites that propagate by transcription of their genomes, reverse transcription, and insertion of a new copy into the host genome (1, 2). Waves of replication have formed families of mostly fragmentary or otherwise degenerate retroelements (1, 2). Epigenetic silencing suppresses retrotransposon activity, keeping them from wreaking genetic and epigenetic havoc in their hosts (3, 4), but active suppression must be maintained (2). Discussions of their biological role have focused on genetics: disruption of genomic structure, and adaptation to form regulatory elements and parts of proteins. Less obvious, but possibly far more important, is their ability to disrupt normal patterns of transcription. McClintock observed that DNA transposons in maize could reversibly alter the expression of genes in the general vicinity of their insertion sites, and so termed them “controlling elements” (5). Retrotransposons also have this ability: Depending on their epigenetic state, they may either lie quietly without interfering in affairs or seize control and dramatically change patterns of gene expression. In this issue of PNAS, Kano et al. (6) describe retrotransposon-controlling elements in the murine dactylin gene. Both appear to effect the dactylaplasia phenotype only when epigenetically active, and their activity is regulated by an unlinked modifier. This finding neatly illustrates some properties of controlling elements and promises in time to give new insights into the mechanisms by which they are kept silent (or not).
Controlling elements create “transcriptional interference,” in which an inserted or reactivated promoter either alters the activity of a nearby promoter or itself transcribes a gene (7). Importantly, this property is separable from the purely genetic effects of an element's insertion. The variety of their effects tells us something about how little we understand the mechanisms of gene regulation in complex genomes. Because controlling elements are epigenetically regulated, they may exhibit mosaic and heritable states of activity. Plant geneticists have developed a large body of work on controlling elements (5, 8), but two examples from animals are particularly well studied.
In the murine agouti viable yellow (A vy) allele, an intracisternal A-particle (IAP) retrotransposon is inserted 100 kb upstream of agouti (Fig. 1 A) (9). Agouti is normally expressed only in one phase of the hair follicle cycle; its product binds to melanocortin receptors and gives fur its agouti pattern (10). When silent, the IAP has little or no effect on the normal pattern of agouti expression, but when it is active, a cryptic promoter in its 5′ LTR transcribes agouti in a constitutive pattern (9), producing a syndrome of yellow fur, obesity, type II diabetes, and cancer (10). Epigenetic mosaicism for activity of the IAP causes the phenotypes of isogenic A vy mice to vary from fully affected through a spectrum of mosaic intermediates to completely unaffected (11). The maternal epigenetic state of A vy is weakly heritable, indicating some tendency for germline stability, but a more prominent feature is germline epigenetic instability: Mice of any A vy phenotype will have offspring with a spectrum of phenotypes (11, 12). Thus, A vy mice illustrate key features of controlling elements: abrogation of normal transcriptional controls, epigenetic variation and inheritance, the requirement for an active transcription state, and the ability to act at a distance from the affected gene.
Controlling elements can interfere with normal gene expression. (A) In the murine A vy allele, a controlling element activates a gene. The A vy allele of agouti contains an inserted IAP retrotransposon. (Upper) When the IAP is active, antisense transcription from its 5′ LTR (red arrow) reads through the agouti gene; the IAP's LTR has a constitutive pattern of activity, resulting in ectopic expression of the agouti gene product. (Lower) When the IAP is silent, agouti is expressed from the hair-cycle-specific promoters 1B and 1C. (B) In the Drosophila y 2 allele, an active controlling element silences a gene. The y 2 allele of yellow contains an inserted gypsy retrotransposon. (Upper) When gypsy is active, it interferes with the action of upstream enhancers (green circles) and prevents expression of yellow in tissues where those enhancers are required for expression. (Lower) When gypsy is silent, there is no interference with yellow expression. Mutations of su(Hw) and mod(mdg4), which act on gypsy, can result in the loss or extension of gypsy's effect. (C) The murine Dac 1J and Dac 2J alleles each contain an insertion of a MusD retrotransposon in the dactylin locus. (Upper) The Dac 1J insertion is 10 kb upstream of the dactylin coding exons. When the MusD is epigenetically active (top), dactylaplasia results without any apparent effect on dactylin expression; when the MusD is silent (bottom), the phenotype is wild type. (Lower) The Dac 2J insertion in intron 5 of dactylin is associated with absence of dactylin mRNA. When the MusD is active (top), dactylaplasia is present; when the MusD is silent (bottom), the phenotype is wild type.
The action of the IAP-controlling element on the agouti locus is simple: it transcribes agouti in a new pattern. The gypsy retrotransposon of Drosophila has a more complex mode of action. Gypsy is an LTR retrotransposon; when inserted 5′ of yellow, it blocks the action of upstream enhancers on the yellow promoter (Fig. 1 B) (13). It has similar effects in other loci (14) and may also affect a homologous locus on another chromosome (15, 16). The enhancer-blocking activity of gypsy depends on binding of Su(Hw), which appears to regulate transcription of gypsy, and on the associated factor Mod(mdg4) (16). Mutation of Mod(mdg4) allows gypsy to act as a bidirectional silencer, blocking the action of both proximal and distal enhancers; it also produces variegated yellow expression (17). The continuing investigations of gypsy constitute the most detailed studies of a controlling element's mechanism of action.
Kano et al. (6) have characterized two insertional mutations at the murine dactylin locus, Dac 1J and Dac 2J, that cause dactylaplasia (limb malformation). They find that both Dac 1J and Dac 2J carry insertions of highly similar full-length MusD retrotransposons: in Dac 1J, 10 kb upstream of the dactylin coding region in an antisense orientation; and in Dac 2J, in intron 5 of dactylin in the sense orientation. In mice displaying dactylaplasia, the 5′ LTR of MusD exhibits acetylated histone H3K9 and hypomethylated DNA, and ectopic MusD expression is observed in the apical ectodermal ridge, the major signaling center for the developing limb. In mice without the phenotype, the 5′ LTR carries modifications consistent with epigenetic silence, and ectopic MusD expression is absent. Thus, the MusD insertions in Dac 1J and Dac 2J are necessary but not sufficient to cause dactylaplasia: Epigenetic activity of the inserted MusD is required in addition to the insertion itself.
The MusD elements do not seem to cause dactylaplasia by acting directly on dactylin expression (Fig. 1
C). Dac
1J mutants do not show any change in the dactylin mRNA transcript. Conversely, in Dac
2J mutant embryos, dactylin
Thousands of relatively intact retrotransposons are scattered throughout mammalian genomes.
mRNA is absent regardless of whether the dactylaplasia phenotype is expressed. This raises the possibility that dactylaplasia
is caused by effects of the active MusD elements on another gene. Kano et al. (6) suggest that they affect the expression of a gene involved in limb development (Fgf8) located 70 kb away from dactylin. Fgf8 expression is partially lost in Dac mutants but restored when MusD is silent.
Most intriguing is the finding that the epigenetic state of the MusD insertions is controlled by an unlinked locus, mdac/Mdac. When a single Mdac allele is present, the MusD elements are epigenetically silenced; when only mdac is present, the elements are active, and dactylaplasia is expressed. The nature of the mdac/Mdac locus, and the mechanism of its effect on Dac 1J and Dac 2J, remains to be elucidated. However, other MusD elements do not appear to be responsive to mdac/Mdac, making it unlikely that the modifier has some general effect. The picture is consistent with epigenetic suppression by the product of a single Mdac allele, but a fuller understanding must await further investigation of mdac/Mdac.
The findings of Kano et al. (6) lend support to the possibility that many mammalian retrotransposons can act as controlling elements. Thousands of relatively intact retrotransposons are scattered throughout mammalian genomes. Most genes have, in their immediate vicinity, multiple retrotransposons that may have the potential to act as controlling elements if they escape epigenetic silencing (7). This is one powerful explanation for the elaborate lengths that cells go to maintain retrotransposon silence (3, 4); mistakes in these mechanisms could account for many small but significant phenotypic variations (including human disease) (7). The behavior of controlling elements suggests that retrotransposons provide an enormous reservoir of potential transcriptional regulatory elements in higher eukaryotes: If released from their epigenetic incarceration, they may simply cause havoc, or they may adapt to create new patterns of gene expression.
Footnotes
- ‡To whom correspondence should be addressed. E-mail: dimartin{at}chori.org
-
Author contributions: J.E.C. and D.I.K.M. wrote the paper.
-
The authors declare no conflict of interest.
-
See companion article on page 19034.
- © 2007 by The National Academy of Sciences of the USA
