|Home | About | Journals | Submit | Contact Us | Français|
In plants and mammals, small RNAs indirectly mediate epigenetic inheritance by specifying cytosine methylation. We found that small RNAs themselves serve as vectors for epigenetic information. Crosses between Drosophila strains that differ in the presence of a particular transposon can produce sterile progeny, a phenomenon called hybrid dysgenesis. This phenotype manifests itself only if the transposon is paternally inherited, suggesting maternal transmission of a factor that maintains fertility. In both P- and I-element–mediated hybrid dysgenesis models, daughters show a markedly different content of Piwi-interacting RNAs (piRNAs) targeting each element, depending on their parents of origin. Such differences persist from fertilization through adulthood. This indicates that maternally deposited piRNAs are important for mounting an effective silencing response and that a lack of maternal piRNA inheritance underlies hybrid dysgenesis.
In Drosophila melanogaster, the progeny of intercrosses between wild-caught males and laboratory-strain females are sterile because of defects in gametogenesis, whereas the genetically identical progeny of wild-caught females and laboratory-strain males remain fertile (1–3). This phenomenon, known as hybrid dysgenesis, was attributed to the mobilization in dysgenic progeny of P-element or I-element transposons, which were present in wild-caught flies but absent from laboratory strains (4–9). The disparity in outcomes, depending on the parent of transposon origin, indicated the existence of cytoplasmically inherited determinants of the phenotype, which must be transmitted through the maternal germ line (8, 9).
The control of mobile elements in germ cells depends heavily on a small RNA-based immune system, composed of Piwi-family proteins (Piwi, Aubergine, and AGO3) and piRNAs (10, 11). Both Piwi and Aubergine (Aub) are deposited into developing oocytes and accumulate in the pole plasm (12, 13), implying possible transfer of maternal piRNAs into the germ lines of their progeny. We therefore asked whether maternally deposited small RNAs might affect transposon suppression in a heritable fashion and whether piRNAs might be the maternal suppressor of hybrid dysgenesis.
We first focused on I-element–induced hybrid dysgenesis (14). A cross of inducer males carrying active I-elements (designated “I” in Fig. 1) to reactive females devoid of active I-elements (designated “R”) yielded dysgenic daughters (termed SF; Fig. 1 and fig. S1) (5, 6, 8, 15). These were sterile, despite normal ovarian morphology. Sterility correlated with I-element expression in SF ovaries (Fig. 1B) (16). The reciprocal cross of inducer females to reactive males yielded fertile progeny (termed RSF; Fig. 1 and fig. S1).
We sequenced 18- to 29-nucleotide RNAs from the ovaries of inducer (w1118) and reactive (wk) strains, 0- to 2-hour embryos from dysgenic and nondysgenic crosses, and ovaries from SF and RSF daughters (Fig. 1A and fig. S1A). Both parental ovary and early embryo libraries contained similarly complex small RNA populations (fig. S1A). This indicates that small RNAs were maternally deposited, because the zygotic genome remains inactive during most of the period that we analyzed.
In Drosophila, piRNAs loaded into the Piwi, Aub, and AGO3 proteins exhibit distinctive features (17, 18). piRNAs occupying Piwi and Aub are predominantly antisense to transposons and contain a 5′ terminal uridine residue (1U; fig. S2A). In contrast, AGO3 harbors mainly sense piRNAs with a strong bias for adenosine at position 10 (10A; fig. S2B). On the basis of these characteristics, we could infer the binding partner(s) for many small RNAs within our sequenced populations. A comparison of small RNAs in mothers and embryos indicated robust maternal inheritance for the Aub/Piwi pool and substantial but weaker deposition of AGO3-bound piRNAs (fig S2, A and B). This observation is consistent with the degree of maternal deposition of the corresponding Piwi-family proteins (figs. S3 and S4) (12, 13).
Patterns of ovarian piRNAs targeting individual D. melanogaster transposons showed marked similarity between inducer and reactive strains (Fig. 2A). However, there were notable differences (Fig. 2, B and C). The I-element exhibited the lowest piRNA count in the reactive strain and the greatest disparity (21-fold) between strains (Fig. 2, B and C, and fig. S5A). Less pronounced differences were noted for tirant and gypsy12, which were more heavily targeted in the inducer strain, and for 1731 and micropia, which were more heavily targeted in the reactive strain (Fig. 2, B and C). These differences were mirrored in corresponding embryonic libraries, with reactive mothers depositing 12-fold fewer I-element piRNAs than inducer mothers (Fig. 2B). This supports the hypothesis that piRNAs correspond to the maternally transmitted phenotypic determinant noted in many studies of hybrid dysgenesis.
The outcome of an inducer-reactive (I-R) dysgenic cross manifests itself not in embryos but in adults. We therefore asked whether differences in maternally deposited piRNAs continued to influence adult piRNA profiles 2 weeks after fertilization. Consistent with their being genetically identical, SF and RSF daughters had virtually identical piRNA levels targeting nearly all transposons (Fig. 2, B and C). Thus, piRNA profiles for many elements had adjusted to a stable equilibrium during the course of germline development. As an example, for 1731 a ninefold difference in piRNA levels between mothers had equalized in progeny (fig. S5B). In RSF females, I-element piRNAs dropped twofold as compared with their inducer mothers (fig. S5B). This paralleled the overall reduction in active I-element load as the inducer genome was diluted by that of the reactive strain. However, limits on the adaptability of the system are seen in SF daughters for the I-element and, to a lesser degree, for tirant and gypsy12 (Fig. 2, B and C, and fig. S5). Though SF daughters contained ~1.6-fold more I-element piRNAs than their reactive mothers, these were still sevenfold less abundant than in RSF daughters. This deficit ultimately results in de-repression of the paternally inherited, active I-elements and in sterility (Fig. 1, A and B).
Though active I-elements are confined to inducer strains, all D. melanogaster strains contain ancestral I-related fragments, (6, 19–21). These are typically found in heterochromatin and exhibit 80 to 95% identity to the modern I-element. Such fragments have been proposed to mediate adaptation to and suppression of I-elements in inducer strains (15, 22–24). To understand the lack of adaptation to I-elements in SF daughters, we probed the nature of interaction between active and ancestral I-element sequences.
The ping-pong model of piRNA biogenesis and silencing describes a Slicer-dependent amplification cycle between active transposons and transposon fragments resident in piRNA clusters (17, 18). Sequence features fitting this model were obvious in piRNAs from the inducer strain (fig. S6). More than half of all I-element piRNAs deviated from the modern sequence and, therefore, must have originated from ancestral fragments (Fig. 3A). Of those piRNAs, the overwhelming majority (90%) were antisense. As a whole, sense-oriented species showed a strong tendency (~78%) to originate from modern active I-element copies (Fig. 3B), whereas 63% of antisense species must have originated from heterochromatic fragments.
Because the reactive strain lacks active I-element copies, no ping-pong amplification occurred, and piRNAs mapping to the sense or antisense strand showed no distinguishing pattern of matching to the active element (Fig. 3B). Despite the lack of an efficient silencing response in SF daughters, we still observed a clear trend for sense piRNAs to originate from active I-element copies, which were paternally transmitted (Fig. 3B). This is consistent with the adaptive system having begun to mount a response in SF daughters, although they ultimately failed to silence the element.
In both reactive and inducer strains, the 42AB cluster represents a major source of piRNAs targeting a variety of mobile elements (fig. S8) (17). Of all heterochromatic I-fragments present in the sequenced melanogaster strain, seven lie within 42AB, and we verified the existence of all in both our reactive and inducer strains (fig. S9). None of the other heterochromatic I-fragments lie within the remaining 19 most active clusters. In the inducer strain, the majority of heterochromatin-derived I-element piRNAs arose from ancestral fragments within 42AB, including many of the most abundant species (Fig. 3, A and C). Despite a more than 20-fold difference in their relative levels, I-element piRNAs matching heterochromatic fragments were also derived from 42AB in the reactive strain (Fig. 3C).
We sought to test whether the role of maternally inherited small RNAs in transposon silencing was general. In P-strain/M-strain (P-M) hybrid dysgenesis, crosses between males containing P-elements (P-strains) and females devoid of such elements (M-strains) yield sterile progeny with severe gonadal atrophy (GD sterility) (3, 9). We examined small RNAs from Harwich (Har), a P-strain containing 30 to 50 P-element copies, and w1118, here serving as an M-strain. Harwich showed strong maternal deposition of P-element piRNAs, whereas both M-strain mothers and their 0- to 2-hour embryos lacked such species (Fig. 4A). This contrasted with I- and F-element piRNAs, which were abundant in parents and embryos from both strains. Crosses between Harwich males and w1118 females yielded dysgenic (GD) progeny. Because of the impact of severe gonadal atrophy (Fig. 4B and table S1), we normalized the daughter library using piRNAs targeting the F-element, a transposon exhibiting consistent profiles in all strains examined. Clearly, dysgenic daughters lacked prominent P-element piRNAs and signatures of the ping-pong amplification cycle (Fig. 4B).
Lerik-P(1A) (designated Lk) contains two P-elements in the X-TAS piRNA cluster, and Nasr’Allah-P(1A) (designated NA) contains a single 5′ truncated insertion at the same locus (25, 26). Lk and NA both produce and maternally deposit P-element piRNAs (Fig. 4A), with these species in the NA strain precisely corresponding to the extent of its only P-fragment (Fig. 4A) (25). Unlike w1118, Lk and NA mothers were able to produce fertile offspring with Harwich. This result correlated with robust piRNA production in daughter ovaries and with a strong signature of the ping-pong amplification cycle.
The 5′ end of the P-element largely lacked piRNAs, particularly antisense species, in both dysgenic flies and fertile NA-Harwich progeny (Fig. 4B). NA does not deposit maternal piRNAs corresponding to this region, because of the truncation of its P-element in X-TAS. Thus, maternal piRNAs are important for potent piRNA generation in daughters, even when the P-element is being effectively silenced by piRNAs matching other parts of the transposon.
piRNA clusters have been envisioned as a genetic reservoir of transposon resistance, with immunity being determined by the content of these loci (17). Our data indicate that the content of piRNA clusters alone is insufficient to provide resistance to at least some elements within a single generation. Instead, maternally inherited small RNAs appear to be essential to prime the resistance system at each generation to achieve full immunity (see also 27).
In the I-R system, environmental factors influence the severity of the phenotype in a dysgenic cross (28) in a manner linked to the expression level of ancestral I-fragments (29). Rearing of reactive mothers at elevated temperature or increases in maternal age raise the proportion of fertile progeny. These observations suggest that the experience of the mother translates into a dominant effect on progeny. Our data suggest that this experience may be transmitted through variations in maternally deposited small RNA populations. Thus, transmission of instructive piRNA populations, shaped by both genetic and environmental factors, may provide a previously unknown mechanism for epigenetic inheritance.