Abstract
Gene silencing is a significant obstacle to genome engineering and has been proposed to be a non-self response against foreign DNA1,2,3,4. Yet, some foreign genes remain expressed for many generations1,3,4 and some native genes remain silenced for many generations1,5,6. How organisms determine whether a sequence is expressed or silenced is unclear. Here we show that a stably expressed foreign DNA sequence in C. elegans is converted into a stably silenced sequence when males with the foreign DNA mate with wild-type hermaphrodites. This conversion does not occur when the hermaphrodite also has exonic sequences from the foreign DNA. Once initiated, silencing persists for many generations independent of mating and is associated with a DNA-independent signal that can silence other homologous loci in every generation. This mating-induced silencing resembles piRNAmediated silencing because it requires the Argonaute PRG-1 (ref. 7) for initiation and the Argonaute HRDE-1 (ref. 1,5) for maintenance. Loss of HRDE-1 can revive gene expression even after 150 generations. Thus, our results reveal the existence of a mechanism that maintains gene silencing initiated upon ancestral mating. By allowing retention of potentially detrimental sequences acquired through mating, this mechanism could create a reservoir of sequences that contribute to novelty when activated during evolution.
Results
Mating is routinely used to introduce genes, including fluorescent reporters, into different genetic backgrounds and it is generally assumed that gene expression is unaffected by this manipulation. While expression from many transgenes is indeed unaffected by mating (Extended Data Fig. 1), we identified a single-copy transgene that violates this rule during the course of our experiments on gene silencing in the hermaphrodite worm C. elegans8. This transgene9 consists of a bicistronic operon that expresses mCherry and GFP in the germline (Fig. 1a, Extended Data Fig. 2). We observed differences in expression from this transgene depending on the gamete through which the transgene was inherited (Fig. 1b). While progeny inheriting the transgene from the oocyte showed uniform fluorescence, progeny inheriting the transgene from the sperm displayed variation in fluorescence that ranged from bright to undetectable – a measurable difference of ~12.5-fold (Fig. 1c, d). Fluorescence of both proteins was similarly affected in each animal (Extended Data Fig. 3), consistent with co-transcriptional or nuclear silencing of the bicistronic pre-mRNA. This silencing was observed in progeny despite stable expression in all male parents (Extended Data Fig. 2b), suggesting that silencing is initiated within cross progeny and not in male parents. While not all cross progeny showed silencing, silenced cross progeny tended to have silenced self progeny in the next generation (Fig. 1e, Extended Data Fig. 4, also see Genetic Inferences in Methods). Thus, gene expression can be affected by the direction of mating and expression in the next generation can depend on the sibling chosen for propagation by selfing. Because this silencing is distinct from previously reported epigenetic silencing phenomena (see Extended Table 1 and Supplementary Discussion), we refer to it as mating-induced silencing.
Mating-induced silencing was not observed in any descendant of cross progeny that inherited the transgene through both gametes (compare Extended Data Fig. 5a with Fig. 1b). It is possible that the maternal presence of an active, i.e. expressed, transgene (Ta) prevents silencing of the paternally inherited transgene. To test if maternal Ta in the hermaphrodite parent is sufficient for preventing mating-induced silencing, we mated hemizygous Ta hermaphrodites with Ta males and examined silencing in progeny that inherited the transgene only from the male (Fig. 2a). All cross progeny showed stable expression of the paternally inherited transgene (Fig. 2a), suggesting that the transgene was protected from silencing by an inherited maternal signal. Consistently, no silencing was observed in any self-progeny of hemizygous parents despite the expected inheritance of the transgene through hermaphrodite sperm in 50% of progeny in each generation (Extended Data Fig. 5b, also see Genetic Inferences in Methods). Thus, a DNA-independent signal transmitted through oocytes can protect the paternal transgene from mating-induced silencing.
To examine the sequence requirements for the production of the protective signal, we tested whether different homologous sequences could prevent mating-induced silencing. We used genome editing to delete parts of Ta (Pmex-5::mCherry::h2b::tbb-2 3’ utr::gpd-2 operon::gfp::h2b::cye-1 3’ utr with Cbr-unc-119(+) upstream) (Fig. 2b, Extended Data Fig. 2a). Neither deletion of the tbb-2 3’ utr and gfp::h2b sequences (TΔ) nor subsequent deletion of upstream sequences (TΔΔ) and h2b from mCherry::h2b (TΔΔΔ) eliminated the protective signal (Fig. 2b, c). One possible interpretation of these results is that the maternal mCherry sequence can protect paternal gfp::h2b from silencing, potentially at the level of the bicistronic pre-mRNA. However, because mating-induced silencing occurred despite the presence of two identical h2b genes (his-58 and his-66) in the C. elegans genome, we infer that not every homologous maternal gene is capable of protecting Ta from silencing. Consistently, neither a Dendra2::h2b transgene with shared sequences nor gtbp-1::gfp could prevent mating-induced silencing of Ta (Fig. 2b, Fig. 2d). Like maternal Ta, maternal TΔΔΔa also retained the property of transmitting a DNA-independent protective signal (Fig. 2e). Thus, a DNA-independent signal derived from maternal Pmex-5::mCherry::cye-1 3’ utr is sufficient to protect both mCherry and gfp of paternal Ta from mating-induced silencing (Fig. 2f).
Protection from mating-induced silencing and susceptibility to mating-induced silencing could have different sequence requirements. Therefore, we examined all deletion variants (Fig. 2b) by crossing males expressing the variant with hermaphrodites without the corresponding transgene. All variants were silenced (Extended Data Fig. 6, also see Genetic Inferences in Methods), suggesting that elimination of an operon structure, histone sequences, and upstream C. briggsae unc-119 sequences did not eliminate the susceptibility to mating-induced silencing. Thus, a minimal gene that has a mex-5 promoter driving the expression of mCherry with cye-1 3’ utr (Pmex-5::mCherry::cye-1 3’ utr) is susceptible to mating-induced silencing.
To dissect the properties of mating-induced silencing, we examined the interaction of the inactive, i.e. silenced, transgene (Ti) with other homologous sequences. Mating Ti males with Ta hermaphrodites resulted in cross progeny that showed silencing (Fig. 3a, top) and progeny from the reciprocal cross also showed a small increase in silencing (Fig. 3a, bottom). Thus, Ti can silence Ta in trans, especially when Ti is inherited through the sperm. To examine if Ti can silence other homologous loci, we mated Ta or Ti hermaphrodites with males expressing homologous (gfp or mCherry) or non-homologous (rfp) sequences tagged to endogenous genes present at other genomic loci (Fig. 3b, c). Animals with Ti showed silencing of gfp and mCherry, but not rfp (Fig. 3b, c). Interestingly, silencing of the ubiquitously expressed gtbp-1::gfp and gtbp-1::mCherry was restricted to the germline, and undetectable in somatic tissues (Fig. 3b). Thus, Ti can silence homologous genes expressed from different loci within the germline, suggesting that Ti generates a sequence-specific silencing signal that is separable from Ti. We therefore tested if parental presence of Ti could affect the expression of homologous sequences in progeny. We examined progeny of a hemizygous Ti parent that did not inherit Ti but did inherit Ta or a homologous gene from the other parent. Cross progeny showed silencing in both cases (Fig. 3d, e, also see Genetic Inferences in Methods). Thus, mating-induced silencing generates a DNA-independent signal that can be inherited through both gametes and can silence homologous sequences in the germline of progeny (Fig. 3f).
The spread of silencing to other loci was not observed in the absence of matching exonic sequences in Ti (Fig. 3c, e). Because this requirement is characteristic of silencing by antisense small RNAs in C. elegans, we examined whether genes implicated in RNA-mediated silencing also play a role in mating-induced silencing. Specifically, we tested the requirement of the double-stranded RNA (dsRNA) importer SID-1 (ref. 10), the primary Argonaute RDE-1 (ref. 11), the RNA-dependent RNA polymerase RRF-1 ( ref. 12), the somatic secondary Argonaute NRDE-3 (ref. 13), and two germline Argonautes, HRDE-1 (ref. 5) and PRG-1 (ref. 7). To test if each gene is required for initiation, we examined mating-induced silencing in the corresponding mutant backgrounds. Substantial silencing was observed in all cases except in animals that lack the prg-1 gene (Fig. 4a, also see Genetic Inferences in Methods). Thus, initiation requires the germline Argonaute PRG-1 and potentially associated germline small RNAs called piRNAs7. Because the minimal Pmex-5::mCherry::cye-1 3’ utr is still susceptible to mating-induced silencing (Extended Data Fig. 6), it is likely that piRNAs recognize a part of this sequence. Such piRNA-mediated silencing is expected to be stable for many generations14. Consistently, we found that mating-induced silencing persisted for >20 generations without selection (Fig. 4b, Extended Data Fig. 7). The silenced transgene retained the capacity to silence homologous genes in trans even after >200 generations (Extended Data Fig. 8a) although the DNA-independent silencing signal was not detectably inherited for more than one generation (Extended Data Fig. 8b). However, unlike silencing of Ta by mating, silencing of Ta by Ti does not generate a DNA-independent signal (Extended Data Fig. 8c). Therefore, the DNA-independent signal made in every generation does not account for the transgenerational stability of mating-induced silencing.
If maintenance of silencing for many generations relies on an active process, then loss of genes required for such silencing could result in the recovery of gene expression. Full recovery of gene expression was observed when hrde-1 was eliminated even after >150 generations (Fig. 4c, d). Silencing persisted in the absence of every other gene (nrde-3, rde-1, rrf-1, sid-1, and prg-1) that was tested 154 to 165 generations after initiation of mating-induced silencing. Crucially, a subsequent retest of loss of hrde-1 171 generations after initiation also resulted in full recovery of gene expression (Fig. 4c, d, Extended Data Fig. 9, also see Genetic Inferences in Methods). Current understanding of silencing by HRDE-1 suggests that nascent transcripts are recoginized by antisense small RNAs bound to HRDE-1, resulting in the recruitment of histone modifying enzymes that generate H3K9me3 at the locus5. The recovery of expression upon loss of HRDE-1 suggests that none of these events that depend on this Argonaute are transgenerationally stable, but rather silencing is actively established in every generation.
Modern genome engineering enables the precise introduction of any sequence into any genome. This study reveals that the fate of such sequences can change during genetic crosses. In progeny of males with a transgene and hermaphrodites without, piRNA-mediated transgenerational silencing is triggered (also see Supplemental Discussion). At genomic loci where this phenomenon can occur, mating of ancestors hundreds of generations ago could have triggered gene silencing that continues to be maintained.
Methods Summary
All C. elegans strains were generated and maintained by using standard methods15. Animals with the transgene T (oxSi487) were introduced into mutant genetic backgrounds through genetic crosses using transgenic hermaphrodites and mutant males to avoid initiation of mating-induced silencing. Cross progeny from genetic crosses were identified by balancing or marking oxSi487 with recessive mutations dpy-2(e8) unc-4(e120) or dpy-2(e8), respectively. In some crosses, cross progeny were identified by genotyping for oxSi487 transgene using PCR. Genome editing was performed using Cas9 protein and sgRNA16. Silencing of all transgenic strains was measured by imaging under identical nonsaturating conditions using a Nikon AZ100 microscope. Quantification of images was performed using NIS Elements (Nikon) and ImageJ (NIH). Detailed procedures are provided in Supplementary Material.
Author contributions
S.D., S.A., and A.M.J. designed and analyzed experiments. S.D. and S.A. performed experiments. S.D. and A.M.J. wrote the manuscript. All authors edited the manuscript.
Author Information
The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper. Correspondence and requests for materials should be addressed to A.M.J. (amjose@umd.edu)
Supplemental Discussion
Comparison of mating-induced silencing with related epigenetic phenomena
The hallmarks of mating-induced silencing are: (1) silencing is initiated upon inheritance only through the male sperm; (2) once initiated, silencing is stable for many generations; (3) transgenerational silencing is associated with a DNA-independent silencing signal that is made in every generation, can be inherited for one generation, and can silence homologous sequences; and (4) maternal exonic sequences can prevent initiation of silencing. While to our knowledge no other known phenomenon shares all of these hallmarks (Extended Data Table 1), phenomena that share some of these features are highlighted below and can inform future mechanistic studies.
Paramutation refers to meiotically heritable changes in gene expression transferred from one allele (“paramutagenic”) to another allele (“paramutable”) when they interact within a cell (reviewed in 19). In addition to similar heritability, both paramutation20,24,37,38,43 and mating-induced silencing rely on small RNAs to spread silencing from one locus to another homologous locus. However, there are several aspects of paramutation that were found to be different from mating-induced silencing, when tested. First, a paramutagenic allele often requires associated repetitive sequences21,22,23. Second, how a paramutagenic allele first arises remains obscure19. Third, while some alleles are paramutable, others are not, for reasons that are unknown20. The reliability of initiating and also protecting from meiotically heritable silencing at a defined single-copy locus described in this study will be useful in discovering possible shared mechanisms that have remained unclear in the ~60 years since the original discovery of paramutation in maize24.
The unpredictable silencing that occurs at some single-copy reporter transgenes within the C. elegans germline has been called RNA-induced epigenetic silencing or RNAe1. Some studies of RNAe1,25, but not others (p.94 in (ref. 2)) report genetic requirements for initiation and maintenance that are similar to those for mating-induced silencing – prg-1 only for initiation and hrde-1 only for maintenance. Although transgenes silenced through RNAe are associated with more small RNAs than unsilenced transgenes1, it remains unclear whether this quantitative increase in small RNAs is the cause or consequence of silencing. Nevertheless, a model proposing RNAe as a response to foreign or non-self DNA has emerged1,2,3,25. This model is inadequate because the same sequence can be either silenced or expressed within the germline1 and endogenous genes are subjected to transgenerational silencing through similar PRG-1- and HRDE-1-dependent mechanisms5,6,7,26,27. Furthermore, the features of a transgene that trigger silencing are unknown. Tethering the Argonaute CSR-1 to the nascent transcript28 or adding intronic sequences that are found in native germline-expressed genes4 can increase the frequency of expression of a foreign sequence but does not itself determine whether a sequence is expressed. Thus, despite these efforts, the mechanisms that enable stable expression or silencing of a gene across generations remain unclear.
Unlike RNAe, mating-induced silencing can be predictably initiated and thus provides a reliable assay for evaluating how organisms establish stable expression or silencing of a gene. Our analyses suggest that the decision to express paternal foreign sequences (mCherry and gfp) is re-evaluated in each generation based upon maternal mRNA (Fig. 2). Although mating-induced silencing is not a general property of transgenes (Extended Data Fig. 1), a similar silencing phenomenon with dependence on maternal mRNA has been observed for the endogenous gene fem-1 (ref. 29). However, it is unknown whether this fem-1 silencing also shares the trans silencing properties and genetic requirements of mating-induced silencing.
Taken together, the paradigm of mating-induced silencing established here provides a reliable model to study epigenetic mechansims that dictate expression or silencing of a sequence in every generation in otherwise wild-type animals.
Implications for genetic studies
The field of genetics relies heavily on analyses of animals generated by mating. Our study reveals that the direction of a genetic cross could strongly influence the phenotype of cross progeny. Additionally, because not every sibling from a cross has the same phenotype, the choice of the sibling selected for further manipulation can have a profound effect. Subsequent transgenerational persistence of silencing can make phenotype independent of genotype, resulting in erroneous conclusions. Thus, when using genetic crosses to generate strains both the direction of the genetic cross and choice of the individual cross progeny selected for propagation needs to be controlled for - especially when evaluating epigenetic phenomena. For example, we ensured that every cross was performed with the transgene present in the hermaphrodite to avoid intiating mating-induced silencing in our studies examining silencing by dsRNA from neurons8. Such methodological considerations impelled by this study could impact conclusions drawn from previous studies of epigenetic silencing in C. elegans.
Possible impact on evolution
Our results reveal a mechanism that silences genes in descendants in response to ancestral mating. The transgenerational stability of this gene silencing with the possibility of recovery of expression even after 170 generations (Fig. 4) suggests that this mechanism could be important on an evolutionary time scale. Genes subject to such silencing could survive selection against their expression and yet be expressed in descendants as a result of either environmental changes that alter epigenetic silencing or mutations in the silencing machinery (e.g. in hrde-1). This mechanism thus buffers detrimental genes from selective pressures akin to how chaperones buffer defective proteins from selective pressures30. Many endogenous genes in C. elegans are silenced by HRDE-1 (ref. 1, 5, 27, 31), some of which could have been acquired when a male with the gene mated with a hermaphrodite without the gene. An interesting direction to explore next is to examine whether this mechanism facilitates adaptation.
Acknowledgements
We thank Nathan Shugarts for most of the Sanger sequencing of oxSi487 in Extended Data Fig. 2a; members of the A.M.J. laboratory for critical reading of the manuscript; the Caenorhabditis elegans Genetic Stock Center, the Seydoux laboratory (Johns Hopkins University), the Cohen-Fix laboratory (National Institutes of Health), and the Hunter laboratory (Harvard University) for some worm strains. This work was supported in part by National Institutes of Health Grant R01GM111457 (to A.M.J.).