Abstract
Gene regulatory networks (GRNs) that direct animal embryogenesis must respond to varying environmental and physiological conditions to ensure robust construction of organ systems. While GRNs are evolutionarily modified by natural genomic variation, the roles of epigenetic processes in shaping plasticity of GRN architecture are not well-understood. The endoderm GRN in C. elegans is initiated by the maternally supplied SKN-1/Nrf2 bZIP transcription factor; however, the requirement for SKN-1 in endoderm specification varies widely among distinct C. elegans wild isolates owing to rapid developmental system drift driven by accumulation of cryptic genetic variants. We report here that heritable epigenetic factors that are stimulated by transient developmental diapause also underlie cryptic variation in the requirement for SKN-1 in endoderm development. This epigenetic memory is inherited from the maternal germline, apparently through a nuclear, rather than cytoplasmic, signal, resulting in a parent-of-origin effect (POE), in which the phenotype of the progeny resembles that of the maternal founder. The occurrence and persistence of POE varies between different parental pairs, perduring for at least ten generations in one pair. This long-perduring POE requires piwi-piRNA function and the germline nuclear RNAi pathway, as well as MET-2 and SET-32, which direct histone H3K9 trimethylation and drive heritable epigenetic modification. Such non-genetic cryptic variation between wild isolates may provide a resource of additional phenotypic diversity through which adaptation may facilitate evolutionary change and shape developmental regulatory systems.
Introduction
The “Modern Synthesis” of the early 20th Century articulated how biological traits shaped by Darwinian forces result from random mutations following the rules of Mendelian inheritance (1). Since that formulation, it has become clear that non-genetic heritable mechanisms can underlie substantial differences in traits between individuals (reviewed in ref. 2). Extensive epigenetic reprogramming occurs in germline and in gamete pronuclei after fertilization to maintain the totipotent state of the zygote. In mammals, disruption of this process often leads to lethal consequences (reviewed in refs. 3, 4). In C. elegans, aberrant reprogramming of epigenetic memory can result in transgenerational accumulation of inappropriate epigenetic marks and a progressive sterile mortal germline (Mrt) phenotype (5, 6). In many cases, the Mrt phenotype is exacerbated by heat stress, demonstrating that environmental factors may influence epigenetic reprogramming in the germline, and that these epigenetic modifications may be passed to subsequent generations (7, 8). Interestingly, C. elegans wild isolates, each carrying a unique haplotype, exhibit variation in the temperature-induced Mrt phenotype, suggesting differential stress response and distinct epigenetic landscapes in natural populations of the species (9).
Many of the documented instances of epigenetic inheritance in mammals are parental or intergenerational effects (less than three generations for female transmission and two generations for male transmission), which can be attributed to direct exposure of the developing embryos in utero to the triggers that alter epigenetic states (2). Parental traumatic experience can trigger heritable behavioral changes and nutritional status of the parents can cause metabolic remodeling in the offspring, which often lasts for one or two generations (10–13). Epidemiological analyses on different human cohorts demonstrate that paternal grandfather’s food access during pre-puberty period affects the mortality of the grandsons (14–16), revealing the potential for long-term transgenerational epigenetic inheritance (TEI) that is induced by environmental conditions.
Studies over the past decade on C. elegans have provided convincing evidence for TEI that persists for at least three generations. Small RNAs are prime candidates for mediators of epigenetic memory, as their expression undergoes only minimal reprogramming in the germline and embryos (17, 18). Primary siRNAs, processed from exogenous dsRNAs or endogenous small RNAs by DICER, are loaded onto the RNA-induced silencing complex (RISC) and mediate degradation of mRNA targets, thereby silencing gene expression. In addition, primary siRNAs, including PIWI-interacting RNA (piRNA – 21U) in the germline, can guide RNA-dependent RNA polymerases (RdRPs) to particular target mRNAs and then amplify silencing signals by producing an abundance of secondary 22G siRNAs. In the germline, HRDE-1 binds to these secondary siRNAs and localizes to the nucleus, where it recruits NRDE-1/2/4 to the nascent transcripts and genomic sites targeted by the small RNAs. This complex then inhibits RNA polymerase II elongation and directs deposition of repressive H3K9me3 marks on the corresponding genomic loci, mediated by histone methyltransferases MET-2 (H3K9me1/2), SET-23 (H3K9me1/2/3) and SET-32 (H3K9me3). Amplification of secondary siRNAs by RdRPs prevents loss of epigenetic memory over multiple generations and, therefore, may permit long-term heritable epigenetic responses (reviewed in refs. 19, 20).
We have uncovered natural epigenetic variation in the gene regulatory network (GRN) that directs development of the embryonic endoderm in C. elegans. The maternal SKN-1/Nrf2 transcription factor activates the mesendoderm GRN in the EMS blastomere at the four-cell stage. EMS subsequently divides to produce the E founder cell, which gives rise exclusively to the intestine, and its anterior sister, the MS founder cell, which produces much of the mesoderm. A triply redundant (Wnt, MAP kinase, and Src) extracellular signal sent from the neighboring P2 blastomere is received by EMS and acts in parallel with SKN-1 to activate endoderm development in the E lineage. In the laboratory N2 strain, elimination of maternal SKN-1 function causes fully penetrant embryonic lethality and a partially penetrant loss of gut: while the majority of the E cells in embryos lacking SKN-1 adopt the mesoectodermal fate of the normal C founder cell, ~30% undergo normal gut differentiation as a result of this parallel signaling input (SI Appendix, Fig. S1) (reviewed in refs. 21–23).
We recently found that the requirement for SKN-1 shows widespread natural variation across genetically distinct C. elegans wild isolates. While removal of SKN-1 in some of the isotypes causes loss of intestine in virtually 100% of embryos, in other isotypes a majority of the embryos differentiate endoderm. Thus, the architecture of the early stages in the endoderm GRN appears to have undergone rapid change during C. elegans evolution (24).
We report here that, although much of the variation in SKN-1 requirement results from genetic differences between the wild isolates (24), it is also determined in part by cryptic, stably heritable epigenetic variation. This effect is uncovered from reciprocal crosses between wild isotypes with quantitatively different phenotypes. This parent-of-origin effect (POE) is transmitted exclusively through the maternal germline. When mothers experience dauer diapause, an alternative developmental stage in C. elegans that confers resistance to environmental insults and longevity, the POE appears to be transmitted through the maternal nucleus, rather than cytoplasmic factors, and can persist through many generations. We further show that this stress-induced POE requires factors that direct H3K9 methylation and the nuclear RNAi machinery. These findings reveal that heritable epigenetic states underlie differences between natural wild isolates and can influence developmental plasticity in early embryos. Such cryptic epigenetic variation provides a potential resource upon which natural selection might act, thus contributing to evolution of GRN architecture (25).
Results
Transgenerational parent-of-origin effect alters the SKN-1 dependence of endoderm formation
The requirement for SKN-1 in endoderm specification varies dramatically across C. elegans isotypes (24). Depending on the isotype tested, between 0.9% and 60% of arrested skn-1(RNAi) embryos undergo gut differentiation when maternal SKN-1 function is eliminated. The behavior of each isotype is quantitatively highly reproducible, showing low variability through many generations, when analyzed by different laboratories and researchers, and from independent lines established from different founder animals (24).
While performing crosses between isotypes with quantitatively different SKN-1 requirements, we found that the outcomes differed depending on the sex of the parent in reciprocal crosses (SI Appendix, Fig. S2). We initially tested two isotypes in which we observed dramatically different skn-1(RNAi) phenotypes: the laboratory N2 strain (29 ±0.4% sd with gut; n = 1320) and the wild isolate JU1491 (1.2 ± 0.4% sd; n = 1228) (SI Appendix, Fig. S3A, p < 0.001). These results agree well with our previous findings (24). Consistent with variation at the level of maternal components, we found that in reciprocal crosses (i.e., male N2 × JU1491 hermaphrodite, and vice-versa), the quantitative requirement for SKN-1 reliably followed that of the maternal line (Fig. 1A). Unexpectedly, however, we found that this non-reciprocality persisted in subsequent generations: the average phenotype of F2 and F3 embryos continued to follow more closely the behavior of their grandmothers and great-grandmothers than their paternal ancestors (Fig. 1B), despite the fact that, with the exception of the mitochondrial genome, the F1 progeny genotypes should be identical regardless of the sex of the founder P0. Thus, these two strains showed a strong parent-of-origin effect (POE) that persists through multiple generations.
Dauer diapause stimulates long-term transgenerational POE through the maternal line
As epigenetic inheritance can be environmentally triggered, it was conceivable that the POE we observed might be influenced by the experience of the parents. Indeed, we found that POE was seen only when the P0 parents had been starved and experienced an extended period (~2 weeks) of dauer diapause with the N2 × JU1491 crosses. In contrast, the progeny of P0’s that were continuously well fed showed an intermediate average phenotype that was not significantly different between descendants of reciprocal crosses (Fig. 1B), consistent with the known multigenic characteristic of the phenotype (24).
We sought to determine whether this environmentally triggered POE extends to other wild isolates. Isolates MY16 (in which only 2.2 ±1% sd (n = 1169) of skn-1(−) embryos make endoderm) and JU1172 (40 ± 3% sd; n = 1491) (SI Appendix, Fig. S3B, p < 0.001) show widely different quantitative phenotypes (24). Consistent with our previous findings, in reciprocal crosses of MY16 and JU1172, we observed a strong maternal effect in the requirement for SKN-1: F1 embryos from mated skn-1(RNAi) mothers followed the maternal phenotype (Fig. 1C). In control experiments with well-fed founder P0 worms, this maternal effect quickly dissipated and was not detectable in F2 embryos (Fig. 1C and D). However, when the parental worms experienced dauer diapause, the average skn-1(RNAi) phenotype of their descendants reliably followed that characteristic of the maternal line through at least ten generations (Fig. 1C and D), a strongly perduring effect.
While dauer development enhances POE, we found that it is not an absolute requirement in all cases. Specifically crosses of JU1491 and JU1172 revealed weak POE even without the diapause trigger, although the effect was stimulated by dauer development (Fig. 2C). This observation suggests that cryptic epigenetic differences between some natural isolates may exist even in the absence of an environmental or physiological trigger. Finally, we found that this effect does not appear to be general to all isotype pairs that show very different phenotypes: for example, diapause-induced POE was not detectable with N2 and MY16 (SI Appendix, Fig. S4).
As expected for successful crosses, in all cases ~50% of the F1 offspring were males (SI Appendix, Fig. S5A and D; one-sample t-test, p > 0.05). Further, cultures established from at least eight randomly selected F1s of successful crosses (with ~50% F1 males) all showed POE (SI Appendix, Fig. S5C), thus ruling out the possibility that the maternal-line bias of the phenotype might result from frequent selfing.
To distinguish between the paternal and maternal contributions to the POE, we starved either the P0 male or hermaphrodite and traced the POE for five generations following reciprocal crosses. These experiments demonstrated that the diapause-induced POE is inherited exclusively through the maternal germline (Fig. 1E; SI Appendix, Fig. S5B). This stable non-reciprocality cannot be explained by long-perduring maternal factors in the cytoplasm: each animal produces ~250 progeny and after five generations, this would result in a dilution factor of ~1011.
Heritability of POE is associated with the maternal nucleus, not heritable mitochondrial or cytoplasmic factors
Dauer larvae and post-dauer adults exhibit a metabolic shift which may reflect changes in mitochondrial function (26–29). Further, starvation has been shown to impact mitochondrial structure and function (30). Thus, the observed maternally directed POE results might arise from differences between the mitochondrial genome sequences in the two strains (31) or might be driven by other cytoplasmically inherited factors. Indeed, wild isolates MY16 and JU1172 contain 13 single nucleotide polymorphisms (SNPs) in mitochondrial protein coding genes (SI Appendix, Table S1), which could alter energy metabolism and stress responses (32, 33). To test whether the POE we observed is attributable to maternal inheritance of mitochondria with particular genomic characteristics, we performed reciprocal crosses in which progeny were repeatedly backcrossed to the paternal strain to obtain lines with primarily the MY16 nuclear genome and mitochondria from the JU1172 line and vice-versa (Fig. 2A). While a strong POE was initially observed in F2 skn-1(RNAi) embryos, this effect was rapidly eliminated as more paternal nuclear DNA was introduced. By the F5 generation, the phenotype was indistinguishable from that of the respective paternal strain (Fig. 2B), suggesting that POE is attributable to the nuclear, rather than mitochondrial, genome. Moreover, the transgenerational POE observed with JU1491 and JU1172 (Fig. 2C) cannot result from variation in mitochondrial DNA, as these two strains carry identical mitotypes. Collectively, our results suggest that the POE we observe is not likely to be caused by mitochondrial inheritance.
To further assess whether nuclear or cytoplasmic/mitochondrial factors underlie the observed POE, we took advantage of a genetic system that generates germlines containing a nucleus derived fully from the paternal line and maternally derived cytoplasm (including mitochondria). In zygotes overexpressing GPR-1 (N2GPR-1 OE) (34, 35), which is required to modulate microtubule-based pulling forces (36), excessive pulling forces cause the maternal and paternal pronuclei to be drawn to opposite poles before nuclear envelope breakdown. This effect generates mosaic embryos in which the anterior daughter (AB) inherits only the maternal chromosomes, while the posterior (P1) receives only the paternal chromosomes. These non-Mendelian events can be scored with the appropriate fluorescent markers (Fig. 2D; SI Appendix, Fig. S6) (35). We found that 72% (n = 230) of the viable F1 progeny from crosses of N2GPR-1 OE hermaphrodites with JU1491 males contained an exclusively paternally derived P1 lineage. If cytoplasmic maternal factors were responsible for the observed diapause-induced POE, the effect would be expected to follow the cytoplasm of the founder P0 worms in the F1 hybrids. In contrast, however, we found that the skn-1(RNAi) phenotypes of F2 and F3 descendants of F1 mosaic animals (those with an N2-derived AB and JU1491-derived P1; SI Appendix, Fig. S6; see Materials and Methods) were indistinguishable from that of the JU1491 strain, regardless of the feeding status of the parents (Fig. 2D’). This finding suggests that the diapause-induced POE is associated with heritable changes in the nucleus, not heritable cytoplasmic maternal factors, including the mitochondrial genome.
POE is not the result of competition in fitness or maternal incompatibility
Parental age has been shown to affect progeny phenotypes in C. elegans and other organisms (37–40). To test the possibility that the POE is influenced by differences in maternal age, we synchronized day-one adults (Fig. 1C) and day-two adults (Fig. 1B, D and E; Fig. 2C). We detected a strong POE in all cases. Moreover, despite large variation in the skn-1(RNAi) phenotype that arises from genetic variation, we observed POE in F5 cultures that were established from very late broods (the last few progeny) produced by senescent F1 animals (Fig. 3A). These findings indicate that parental age does not contribute substantially to the POE observed.
The differences in SKN-1 requirement seen as the result of the POE might reflect maternal incompatibility, which favors particular genetic regions as a result of lethality or slow growth (41, 42). If such regions included those known to influence the requirement for SKN-1 in the endoderm GRN (24), there could be selection for the trait following recombination of the two parental genomes. We note that such a possibility would also require that any such selection is triggered only after starvation and dauer development for the cases in which we observed such an essential requirement. Further, we observed strong diapause-induced POE in embryos from skn-1 RNAi-treated F1 heterozygous mothers, whose genotypes would be identical in the two reciprocal crosses. Thus, the effect at this stage is not attributable to maternal incompatibility resulting in selection against particular allelic combinations that arise by recombination (Fig. 1B–D).
To further investigate whether POE might be driven by genetic incompatibility that is environmentally triggered by starvation/dauer development, we also characterized lethality and fecundity of F2 progeny from the reciprocal crosses. Two mechanisms involving selfish genetic elements that result in maternal incompatibilities were previously described in C. elegans: the peel-1/zeel-1 (41) and sup-35/pha-1 (42) toxin/antidote systems. The wild isolate JU1172 does not carry the paternal selfish peel-1/zeel-1 element (41). When mated, MY16 sperm deliver PEEL-1 toxin, causing F2 embryos that are homozygous for the JU1172 zeel-1 haplotype (~25%) to arrest. We found that, indeed, crosses between JU1172 and MY16 are associated with embryonic lethality. However, although this lethality was slightly lower (~13-16%) when JU1172 was the paternal strain compared to the reciprocal crosses (20-26%; Fig. 3B), the difference is insufficient to explain the strong POE we have observed. Furthermore, this effect does not change appreciably regardless of the experience of the P0 (fed or starved/dauer), which is not consistent with selection induced by this experience. As both MY16 and JU1172 harbor the active sup-35/pha-1 maternal toxin/antidote element (42), the lethality may reflect an unidentified maternal-effect toxin in the MY16 strain. In addition, the progeny of crosses between MY16 and JU1172 in either direction both showed somewhat reduced fecundity/viability, presumably owing to genomic incompatibility between the two strains (Fig. 3C). However, the parental origin of the P0s did not influence the degree of larval lethality or sterility in the F2 animals (Fig. 3C, Fisher-exact test p > 0.05). Together these results indicate that genetic incompatibility alone cannot account for the strong POE we observed. Rather, POE appears to result from perduring epigenetic inheritance reflecting the experience of the original founding parents of the cross.
Maintenance of POE involves the nuclear RNAi pathway and histone H3K9 trimethylation
The findings noted above suggest that POE is mediated through nuclear signals. The nuclear RNAi pathway has been implicated in a number of examples of TEI (7). To assess whether this pathway might be involved in transmitting the POE we have observed, we analyzed the F4 progeny of reciprocal crosses of post-dauer P0’s in which nrde-4 was knocked down by RNAi (strategy shown in Fig. 4A). While a strong POE was observed in the control animals containing functional NRDE-4, nrde-4(RNAi) abrogated the POE in the F5 embryos (Fig. 4B): i.e., the requirement for SKN-1 was not significantly different in the descendants of reciprocal crosses. It was conceivable that this effect might simply reflect a direct role for NRDE-4 in the requirement for SKN-1 per se. However, we found that the skn-1(RNAi) phenotypes of the MY16 and JU1172 isolates treated with nrde-4 RNAi were indistinguishable from those treated with control RNAi (Fisher-Exact test p > 0.05; Fig. 4C). Thus, these findings implicate NRDE-4, and hence the nuclear RNAi pathway, in the POE process.
Gene silencing though the nuclear RNAi pathway that results in TEI is mediated through the Piwi-encoding homologue, prg-1, and piRNAs that trigger biosynthesis of secondary 22G RNAs; (a second homologue, prg-2, is likely to be a pseudogene) (43–46). We knocked down PRG-1/2 in F4 animals from MY16 and JU1172 reciprocal crosses and found that, in contrast to their siblings treated with control RNAi, POE was abrogated in the F5 embryos (Fig. 4B). Control experiments demonstrated that prg-1/2 RNAi does not affect skn-1 RNAi efficacy in either parent line (p > 0.05; Fig. 4C). Loss of nuclear RNAi factors lowers the efficacy of RNAi targeting of nuclear-localized RNAs (47–49); however, maternal skn-1 mRNA in the early embryos is localized in the cytoplasm, and the silencing effect of skn-1 RNAi would be expected to depend primarily on RISC in the cytoplasm (50).
NRDE-4 is required for the recruitment of NRDE-1 to the targeted loci and subsequent deposition of the repressive H3K9me3 mark, which results in gene silencing (49). Furthermore, H3K9me3 has been implicated in transgenerational silencing of transgenes or endogenous loci mediated by exogenous RNAi (43, 51–53). These observations, and our findings that knockdown of nrde-4 abolishes POE, led us to hypothesize that H3K9 methylation might function as a mediator of POE. Indeed, we found that treating F4 animals that showed POE with RNAi against met-2 or set-32, in contrast to their control siblings, eliminated POE in the F5 embryos (Fig. 4B). Although loss of MET-2 has been shown to enhance RNAi sensitivity (52), we found that neither met-2 RNAi nor set-32 RNAi significantly modifies the skn-1(RNAi) phenotypes of MY16 and JU1172 wild isolates (p > 0.05; Fig. 4C). Thus, the loss of POE in the F5 generation with met-2 or set-32 RNAi is not attributable to modified RNAi response. Collectively, these results suggest that, in response to dauer diapause, piRNAs in the germline direct histone methylation through the nuclear RNAi pathway, thereby maintaining POE across generations (Fig. 4D).
Discussion
While massive epigenetic reprogramming ensures totipotency of the germline during animal development, some epigenetic marks escape erasure, leading to stable epigenetic inheritance that can persist through many generations. Such long-term epigenetic inheritance has the potential to provide a source of cryptic variation upon which evolutionary processes might act; however, little is known about natural epigenetic variation within a species, how it is influenced by environmental conditions, and the degree to which it influences GRN plasticity. In this study, we report four major findings that reveal cryptic natural epigenetic variation and it mechanistic action in a core embryonic GRN in C. elegans: 1) dauer diapause can trigger POE that alters the output of the endoderm GRN. 2) This effect is transmitted through the maternal germline across multiple generations apparently through nuclear signals. 3) Different combinations of wild isolates exhibit variation in their capacity for establishing and maintaining these transgenerational epigenetic states. 4) This POE requires components of the piRNA-nuclear RNAi pathway and H3K9 trimethylation. These findings indicate that maintenance of an acquired epigenetic state in response to environmental stimulus can confer substantial plasticity to a core developmental program and may provide additional natural variation that may be subject to evolutionary selection.
Dauer diapause induces persistent epigenetic inheritance
The perduring epigenetic effect that we have observed is triggered in parents that have experienced dauer diapause. Dauer entry and formation require extensive epigenetic remodeling and some of these changes are retained throughout the remainder of development: post-dauer adults contain distinct chromatin architecture and particular pools of small RNA that differ from animals that have not experienced dauer diapause (29, 54). In addition, the progeny of starved animals show increased starvation resistance and lifespan (55–57). Consistent with the model that the effect of ancestral developmental history is carried across generations, we found that dauer diapause leads to TEI that modifies the quantitative SKN-1 input in endoderm development.
The TEI we have observed varies in its long-term perdurance, depending on the wild isolates involved. In crosses between the laboratory strain N2 and wild isolate JU1491, dauer diapause-induced POE lasted for three generations, but was subsequently lost, similar to the transmission dynamics of the silencing effect induced by exogenous RNAi (58, 59). This progressive transgenerational loss in the effect may result from passive dilution of regulatory small RNAs and active restoration of an epigenetic “ground state” over generations, although the detailed mechanisms for such a process are not well understood (59). In contrast, in crosses between the MY16 and JU1172 wild isolates, we observed stable TEI that lasted for at least 10 generations and which conceivably persists longer. Consistent with a recent study that identified genetic determinants of efficient germline maintenance and epigenetic reprogramming among C. elegans wild isolates (9), our results showed that the generational duration of epigenetic inheritance may also be influenced by genetic background, suggesting an interplay between genetics and epigenetics.
The transmission of the silencing effects of exogenous RNAi in C. elegans has provided an excellent paradigm for revealing mechanisms of epigenetic inheritance (19, 20). While inheritance of exogenous RNAi and physiological responses triggered by changing environment share overlapping machinery, our results reveal two key differences between the two processes: 1) we demonstrated that epigenetic memory triggered by dauer diapause is transmitted exclusively through the maternal germline. This contrasts with the inheritance of exogenous RNAi (59) and transgenes (60), which show paternal bias. 2) We found that PRG-1 is required for the maintenance of POE. In contrast, the piwi-piRNA pathway had previously been shown to be required for the initiation, but not maintenance of transgene silencing in the germline. Once established, this silencing state depends on the nuclear RNAi pathway, which promotes deposition of H3K9me3 marks on the transgene (43, 44). Supporting our findings, however, Simon et al. demonstrated that PRG-1 is important for maintaining germline mortality through a mechanism that is independent from its action in transgene silencing. Animals that lack PRG-1 exhibit dysregulation of gene expression and reactivation of transposons and tandem repeats, showing that piRNAs are required to maintain silencing of at least some endogenous loci (61, 62).
Relationship of POE to genomic imprinting in C. elegans
Genomic imprinting is perhaps the best-studied example of epigenetic inheritance. Differential DNA methylation or histone modifications on the two parental chromosomes, established during gametogenesis or post-fertilization, escape epigenetic reprogramming, causing genes to be expressed in a parent-of-origin manner (63, 64). However, in the case of C. elegans, animals that inherit the entire paternal genome are fertile and viable, as we and others have shown (34, 35). This observation reveals that genomic imprinting is not essential for normal development or survival in C. elegans, consistent with an early study in which animals containing individual chromosomes from only one parent were analyzed (65). Nevertheless, the X chromosome of sperm, unlike that of oocytes, is devoid of H3K4me2 activation marks, a pattern that persists through several rounds of cell division cycles during early embryogenesis (66). In addition, the expression of sperm-derived autosomal transgenes is greater than that in oocytes, which may result from differential epigenetic remodeling in sperm and oocyte chromatin upon fertilization (60). While these findings demonstrate the imprinting capacity of C. elegans, endogenous imprinted genes have not yet been reported.
We propose that passage through dauer diapause may induce paternal-specific silencing through deposition of repressive H3K9me3 on paternal loci that affect the endoderm GRN and the SKN-1 requirement (24), leading to the POE we observed. Although imprinting is not essential for viability in C. elegans, its effects may become significant in response to environmental stimuli. For example, maternal dietary restriction elevates vitellogenin oocyte provisioning (67). Both yolk-associated fatty acids and small RNAs, which have been proposed to be associated with yolk particles, promote epigenetic changes in the nucleus, and might thereby direct establishment of parent-of-origin epigenetic marks (68–71). Dauer-favoring conditions also reduce insulin/insulin-like growth factor signaling and enhance starvation stress resistance in the progeny (67). With recent advances in techniques for examining transcriptional regulatory landscapes (64, 72), it will be of interest to identify loci that are responsive to environmental stimuli and that may be differentially imprinted across generations.
The potential role of cryptic epigenetic variation in accelerating evolutionary change
It is clear that in C. elegans, stress responses can be transmitted transgenerationally and influence physiology adaptively in the offspring (45, 55–57). In Arabidopsis, it has been shown that experimentally induced, or naturally occurring epigenetic variations, once stabilized, can be subjected to artificial selection (73, 74), highlighting the potential capacity of TEI to facilitate adaptation and evolution. While TEI is prominent in worms and plants with a short life cycle, and hence environmental conditions may be relatively constant through multiple generations (58, 75), there is evidence that TEI may also occur in mammals with much longer reproductive cycles (76). Epigenetic inheritance may be especially important in organisms with low genetic diversity, such as those, including C. elegans and Arabidopsis, that propagate by self-fertilization. Many C. elegans strains isolated from neighboring locations are near-identical and polymorphism rates are low even among genetically distinct isotypes (77). In such homozygous, genetically non-diverse populations, epigenetic variations may provide a particularly rich resource upon which natural selection may act.
Environmental factors can induce plastic phenotypic changes that are subjected to Darwinian selection. Over time, the phenotypic variants may become genetically fixed, a process known as “genetic assimilation” (25). As the rate of genetic mutations is low in C. elegans (2.1 × 10−8 per nucleotide site per generation)(78), heritable epigenetic variants may act as a buffer to cope with rapid environmental change before adaptive mutations arise. Alterations in epigenetic states can also affect mutation rates and trait evolution (79) and TEI might therefore accelerate the rate of evolution by facilitating genetic assimilation. We propose that epigenetic inheritance affecting SKN-1 dependence may contribute towards the rapid change in the endoderm gene regulatory network architecture that we previously observed among C. elegans wild isolates (24). SKN-1 acts in pleiotropic functions, including mesendoderm specification, oxidative stress and unfolded protein responses, promoting longevity, and modulating metabolism during starvation (reviewed in ref. 80). It is conceivable that SKN-1 is particularly susceptible to plastic changes in its regulatory outputs as a means of adapting to frequently varying environmental conditions. In the wild, C. elegans experience a boom-and-bust cycle and most worms isolated in the wild are present in the dauer stage (81). As we have shown, dauer diapause is associated with strong heritable epigenetic responses (55) that may, therefore, influence developmental plasticity and adaptive evolution in response to the local environment. We believe that our findings may provide among the first example of environmentally-induced heritable epigenetic changes that modulate developmental inputs into an embryonic GRN.
Materials and Methods
C. elegans strains and maintenance
N2 (Bristol, UK), MY16 (Mecklenbeck, Germany), JU1491 (Le Blanc, France), and JU1172 (Concepcion, Chile). JR3336, (elt-2∷GFP) X; (ifb-2∷GFP) IV. PD2227 (35), oxIs322 II; ccTi1594 III. oxIs322 contains [myo-2p∷mCherry∷H2B + myo-3p∷mCherry∷H2B + Cbr-unc-119(+)] II. ccTi1594 contains [mex-5p∷GFP∷gpr-1∷smu-1 3’UTR + Cbr-unc-119(+), III: 680195] III.
Worm strains were maintained as described (82) and all experiments were performed at 20°C unless noted otherwise. To ensure no carry-over of parental stress response, fresh worm stock was obtained from −80°C and maintained in 150mm NGM plate seeded with E. coli OP50 for at least five generations prior to beginning experiments. To avoid genetic drift and lab domestication, a fresh worm stock was obtained every ~30 generations.
To obtain males for crosses, 20–30 L4 hermaphrodites were picked into 7% ethanol solution in microcentrifuge tubes and rotated for an hour (83). Worms were pelleted by centrifugation at 2000 rpm for 30 seconds. They were then transferred to NGM plates seeded with E. coli OP50. F1 male progeny were mated with sibling hermaphrodites to establish male stocks.
Dauer induction and POE assays
The animals were maintained on NGM plates seeded with OP50. Once the cultures became crowded and exhaust their food supply, they were incubated for an extra two weeks at 20°C. The worms were then washed with M9 buffer and incubated in 1% sodium dodecyl sulfate (SDS) for 30-60 minutes with gentle agitation to select for dauer larvae (84). Isolated dauer larvae were then washed with M9 to remove all SDS and allowed to recover overnight on 60 mm NGM plates seeded with OP50.
Reciprocal crosses were set up using L4s and the animals were allowed to mate overnight. Control experiments using well-fed animals were performed in parallel. Mated hermaphrodites, as indicated by the presence of copulatory plugs (except for crosses involved N2 males which do not deposit plugs), were transferred to a fresh NGM plate to lay eggs for ~5-7 hours and early brood was discarded to avoid contamination of self-progeny. The hermaphrodites were then transferred to a fresh seeded NGM plate to lay eggs overnight. The hermaphrodite (P0) was then removed, leaving the F1s alone. Once the F1 worms reached early or mid-adulthood, they were treated with 15% alkaline hypochlorite solution to obtain F2 embryos which were allowed to hatch on food. This procedure was repeated until the F4 generation (F10 for the experiment shown in Fig. 1B) was obtained. At each generation, L4 worms were used to determine skn-1 RNAi phenotype (SI Appendix, Fig. S2).
For crosses between PD2227 hermaphrodites and JU1491 males, the POE assay was performed as described above. F1 L4s were immobilized on 5% agar pad with 5 mM levamisole diluted in M9 and observed using Nikon Eclipse Ti-E inverted microscope. Mosaic worms that expressed myo-2∷mCherry, but not myo-3∷mCherry and mex-5∷GFP (i.e., PD2227-derived AB and JU1491-derived P1), were recovered on seeded NGM plate in the presence of M9. 20 F2 animals were then randomly selected and observed to ensure no worms expressed fluorescent markers, i.e. JU1491 nuclear genotype (Fig. 2D, SI Appendix, Fig. S6).
Viability and embryonic lethality scoring
To score viability, young hermaphrodites (F1 progeny of the reciprocal crosses) were allowed to lay eggs on an NGM plate seeded with OP50. The next day, newly hatched L1s (F2) were transferred to individual seeded plates. “Larval lethal” was defined as the percentage of worms that arrested as L1s. Worms that reached adulthood but failed to reproduce in five days were scored as sterile. To score embryonic lethality, individual young hermaphrodites (F1) were allowed to lay eggs on an NGM plate seeded with a small drop of OP50 for ~4-8 hours. The hermaphrodites were then removed, leaving the embryos. The fraction of unhatched embryos were counted and scored ~24 hours later. At least two independent broods were scored.
RNAi
Feeding-based RNAi experiments were performed as described (24). RNAi clones were obtained from either the Vidal (85) or Ahringer (86) libraries. RNAi bacterial strains were grown at 37°C in LB containing 50 μg/ml ampicillin. The overnight culture was then diluted 1:10. After four hours of incubation at 37°C, 1 mM IPTG was added and 60-100 μl was seeded onto 35 mm agar plates containing 1 mM IPTG. Seeded plates were allowed to dry and used within five days when kept at room temperature. For skn-1 RNAi, five to 10 L4 animals were placed on RNAi plate. 24 hours later, they were transferred to another RNAi plate to lay eggs for 12 hours. The adults are then removed, leaving the embryos to develop for an extra 5-7 hours. Embryos expressing birefringent gut granules are quantified and imaged on an agar pad using a Nikon Ti-E inverted microscope under dark field with polarized light (SI Appendix, Fig. S1B).
For met-2, set-32, nrde-4 and prg-1/2 RNAi, 15-30 F3 L4 animals showing POE were placed on plates of E. coli containing an empty control vector (L4440) or expressing double stranded RNA. 24 hours later, they were transferred to another RNAi plate to lay eggs for about seven hours. The adults were then removed and the F4 animals were allowed to develop on RNAi bacteria. F4 L4 larvae were used for skn-1 RNAi assay for POE as described above (Fig. 4A).
Statistics and figure preparation
Statistics were performed using R software v3.4.1 (https://www.r-project.org/). Two-sample two-tailed t-tests were used to compare skn-1 RNAi phenotype between two groups, unless stated otherwise. Welch’s t-tests were performed if the variances of the two groups being compared are not equal. Plots were generated using R package ggplot2.
Acknowledgments
We thank Coco Al-Alami for experimental assistance. We thank members of the Rothman lab for helpful advice and feedback. Nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the National Institutes of Health - Office of Research Infrastructure Programs (P40 OD010440). This work was supported by grants from NIH (#1R01HD082347 and # 1R01HD081266).
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.
- 12.
- 13.↵
- 14.↵
- 15.
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.
- 28.
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.
- 39.
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.
- 70.
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵