Non-membrane-bound organelles such as nucleoli, processing bodies, Cajal bodies and germ granules form by the spontaneous self-assembly of specific proteins and RNAs. How these biomolecular condensates form and interact is poorly understood. Here we identify two proteins, ZNFX-1 and WAGO-4, that localize to Caenorhabditis elegans germ granules (P granules) in early germline blastomeres. Later in germline development, ZNFX-1 and WAGO-4 separate from P granules to define an independent liquid-like condensate that we term the Z granule. In adult germ cells, Z granules assemble into ordered tri-condensate assemblages with P granules and Mutator foci, which we term PZM granules. Finally, we show that one biological function of ZNFX-1 and WAGO-4 is to interact with silencing RNAs in the C. elegans germline to direct transgenerational epigenetic inheritance. We speculate that the temporal and spatial ordering of liquid droplet organelles may help cells to organize and coordinate the complex RNA processing pathways that underlie gene-regulatory systems, such as RNA-directed transgenerational epigenetic inheritance.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank members of the Kennedy laboratory for discussions. We thank T. Ishidate and C. Mello for sharing unpublished data. We thank H. Y. Mak for sharing strains. Some strains were provided by the CGC (P40 OD010440). Some strains were provided by the MITANI Laboratory. This work was supported by the National Institutes of Health, RO1 GM088289 (S.K.). B.D.F. and A.S. were supported by NSF graduate research fellowships.Reviewer information
Nature thanks A. Pasquinelli and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
a, A genetic screen was conducted to identify components of the C. elegans RNAi inheritance machinery. The screen contained several filters (see below) to remove known RNAi inheritance factor. Factors defective for RNAi inheritance are also defective for nuclear RNAi6. Therefore, our screen began with selections for mutant alleles that disrupt nuclear RNAi. Two selections were developed for nuclear RNAi mutants. First, the lin-15b and lin-15a genes are transcribed as a polycistronic message that is spliced within the nucleus into lin-15b and lin-15a mRNAs31. Animals containing mutations in both lin-15b and lin-15a exhibit a multivulva (Muv) phenotype32,33. RNAi targeting lin-15b (in eri-1(−) animals) silences lin-15b and lin-15a co-transcriptionally, thus inducing a Muv phenotype34. The previously identified nuclear RNAi factors are required for lin-15b RNAi-induced co-transcriptional silencing of lin-15a and, therefore, for lin-15b RNAi-induced Muv. A second assay for nuclear RNAi is lir-1 RNAi. lir-1 RNAi is lethal because lir-1 is in an operon with lin-26, and co-transcriptional silencing of lin-26 by lir-1 RNAi causes lethality34. Nuclear RNAi defective (NRDE) animals do not die in response to lir-1 RNAi because they fail to silence lin-2634. Previous genetic screens have used suppression of lir-1 RNAi to find factors required for nuclear RNAi. These screens have reached saturations: we have identified several alleles in all the nrde genes using this approach. Unpublished work from the laboratory shows, however, that hypomorphic alleles of the nrde genes will often block lin-15b RNAi-induced Muv and yet still die in response to lir-1 RNAi. We interpret these data to mean that survival from lir-1 RNAi is a much stronger selection for nuclear RNAi mutants than a failure to form Muv in response to lin-15b RNAi. That is, factors that contribute to nuclear RNAi, but are not 100% required for nuclear RNAi, would not be identified by lir-1 RNAi suppression screens. Therefore, our screen looked for suppressors of lin-15b RNAi, which did not suppress lir-1 RNAi, because this screen might identify genes missed in our previous genetic screens. Step 1, identify factors required for nuclear RNAi. eri-1(mg366) animals were mutagenized with EMS. F2 progeny were exposed to bacteria expressing lin-15b dsRNA. Non-Muv animals were kept as candidate novel nuclear RNAi factors. Step 2, discard known nuclear RNAi factors. We probably know all non-essential genes that can mutate to suppress lir-1 RNAi. Therefore, we discarded mutants that suppressed lir-1 RNAi as these alleles are probably known nuclear RNAi factors. Mutants that did not suppress lir-1 may contain mutations in factors important, but not essential, for nuclear RNAi. Step 3, identify mutations that suppress RNAi inheritance. The last filter in our screen was to identify mutant alleles that disrupted RNAi inheritance. We subjected remaining mutant animals to dpy-11 RNAi, which causes animals to become Dumpy (Dpy). Progeny of animals exposed to dpy-11 dsRNA inherit dpy-11 silencing and are Dpy35. RNAi inheritance mutants become Dpy in response to dpy-11 RNAi; however, the progeny of these animals fail to inherit dpy-11 silencing, and, therefore, are not Dpy. Thus, any of our mutant animals that became Dpy in response to dpy-11 RNAi, but whose progeny were not Dpy, were kept for further analysis. Finally, only one mutant was kept from each pool (pools were maintained as independent populations throughout the screen). b, c, Independent alleles of znfx-1 and wago-4 are (as expected) defective for lin-15b RNAi and not defective for lir-1 RNAi. Data are mean ± s.d. of more than three biologically independent samples.
a, Animals expressing a pie-1::gfp::h2b transgene were exposed to gfp dsRNA4. The percentage of the P0, F1 and F2 progeny of the indicated genotypes expressing GFP was quantified. Data represent scoring of at least 80 animals in each generation and for each genotype. Note, the gfp reporter transgene used in this study is a multi-copy version of the single copy version used in Fig. 1b. Note that some RNAi inheritance can be seen in znfx-1 mutant animals using this reporter transgene. Thus, in some cases, some RNAi inheritance can occur in the absence of ZNFX-1. b, Animals of the indicated genotypes were exposed to dpy-11 dsRNA. The F1 progeny of these animals were grown in the absence of dpy-11 dsRNA, and were scored for Dpy phenotypes. Data are mean ± s.d. of more than three biologically independent samples. Consistent with the idea that ZNFX-1 (and NRDE-2) is required specifically for inheritance, znfx-1 mutant animals exposed directly to dpy-11 dsRNA are Dpy (data not shown). c, zu405ts is a temperature-sensitive (ts) lethal (embryonic arrest at 20 °C) allele of oma-136. oma-1 RNAi suppresses oma-1(zu405ts) lethality, and this effect is heritable5,6. Animals of the indicated genotypes were exposed to oma-1 dsRNA and the fertility of the progeny of these animals was scored over generations. Data show that znfx-1 mutant animals are defective for oma-1 RNAi inheritance. Data are mean ± s.d. of three biologically independent samples.
Extended Data Fig. 3 CRISPR–Cas9-epitope tagged genes used in this study, with one exception, produce functional proteins and are expressed at or near wild-type levels.
a–e, The addition of epitope tags by CRISPR–Cas9-mediated gene conversion of znfx-1 or wago-4 did not affect function of tagged proteins in these RNAi inheritance. dpy-11 RNAi inheritance assays in which the progeny of animals exposed to dpy-11 dsRNA are visually scored for the inheritance of Dpy phenotypes. The indicated epitope-tagged proteins are functional in this RNAi inheritance assay. n = 3 biologically independent samples; data are mean ± s.d. f, g, pgl-1 mutant animals show a temperature-sensitive (25 °C) sterile phenotype. The addition of epitope tags by CRISPR–Cas9-mediated gene conversion to the pgl-1 locus did not affect PGL-1 function as these animals were fertile. L4 animals were singled from 20 °C to 25 °C and brood sizes were scored. pgl-1::tagrfp (n = 6 animals) and pgl-1::mcardinal; tagrfp::znfx-1; mut-16::gfp (n = 15 animals). h, mut-16(−) animals are defective for pos-1 RNAi. Embryos of the indicated genotype were grown on pos-1 dsRNA-expressing bacteria. Six L4 animals were picked to pos-1 dsRNA-expressing bacteria and laid eggs overnight. Unhatched embryos and hatched animals were scored. The addition of gfp to the mut-16 locus did not affect MUT-16 function. Data are mean ± s.d. of three biologically independent samples. i–k, In some cases, Flag::GFP::WAGO-4-expressing animals are defective in RNAi inheritance, indicating that Flag::GFP::WAGO-4 is not fully functional. i, Animals of the indicated genotypes were exposed to dpy-11 dsRNA and F1 progeny were grown in the absence of dpy-11 dsRNA. The percentage of Dpy animals is shown. At least 150 animals of each genotype were scored. Thus, 3×Flag::GFP::WAGO-4 is not functional for dpy-11 inheritance. n = 3 biologically independent samples. j, 3×Flag::GFP::WAGO-4 is functional during oma-1 RNAi inheritance. See Extended Data Fig. 2c for details of the oma-1 RNAi inheritance assay. n = 3 biologically independent samples; data are mean ± s.d. In Fig. 2d, both wago-4 and znfx-1 are shown to exhibit an Mrt phenotype at 25 °C. Here, 3xflag::gfp::wago-4 animals are not Mrt, indicating that 3×Flag::GFP::WAGO-4 is capable of promoting germline immortality. n = 3 biologically independent samples; data are mean ± s.d. i–o, CRISPR tags did not seem to affect gene expression. To address the possibility that epitope tagging of the genes used in this study changed gene expression levels, we isolated total RNA from animals of the indicated genotypes and used qRT–PCR to quantify indicated mRNA levels. Primers target exon–intron junctions. Early stop or deletion alleles for each of these loci were used as controls. wago-4(tm1019) and znfx-1(gg561) are deletions and primers were located within deleted regions. pgl-1(bn101) and mut-16 (pk710) are nonsense alleles. A decrease in the mRNA levels of these mutants is probably due to nonsense-mediated decay. n = 3 biologically independent samples; data are mean ± s.d.
a, znfx-1 is required in inheriting generations for RNAi inheritance to occur. In brief, we initiated gene silencing in znfx-1/+ heterozygous animals and scored the +/+ and −/− progeny for their ability to inherit gene silencing. Progeny containing at least one wild-type copy of znfx-1 were capable of inheriting gene silencing, whereas −/− progeny were not. More specifically, znfx-1(gg575) +/− animals that express the pie-1::gfp::h2b transgene17 exposed to gfp dsRNA, and progeny from F1 to F3 generations were scored. Micrographs of GFP fluorescence in oocytes are shown. To identify cross progeny, the following strategy was used via CRISPR. pie-1::gfp::h2b was marked by dpy-10 (cn64) (dpy-10 is approximately 0.77 cM from pie-1::gfp::h2b). dpy-10 (cn64)/+ animals are Dpy Rol and dpy-10 (cn64) homozygous animals are Dpy. znfx-1 genotypes were inferred based upon wild-type, Dpy and Rol phenotypes; n > 30 animals. b, znfx-1 is sufficient in inheriting generations for RNAi inheritance to occur. We initiated gene silencing in znfx-1(−/−) animals, introduced a wild-type copy of znfx-1 to progeny (via mating), and scored znfx-1/+ cross-progeny for inheritance. The data show that znfx-1(+/−) progeny, from parents that lack a wild-type copy of znfx-1, were able to inherit silencing. znfx-1(gg575) was marked by dpy-10(cn64) (dpy-10 is approximately 1.09 cM from znfx-1). dpy-10(cn64)/+ animals are Dpy Rol and dpy-10(cn64) homozygous animals are Dpy. znfx-1 genotypes was inferred based upon wild-type, Dpy and Rol phenotypes. n > 20 animals. c, Additional biochemical evidence that ZNFX-1 acts in inheriting generations to promote inheritance. The nuclear RNAi factor NRDE-2 binds to pre-mRNA of genes undergoing heritable silencing6. When znfx-1(−) animals were exposed directly to oma-1 dsRNA, NRDE-2 bound to the oma-1 pre-mRNA at wild-type levels. However, in progeny of znfx-1(−) mutant animals NRDE-2 failed to bind oma-1 pre-mRNA. Animals expressing NRDE-2::3×Flag were treated with +/ oma-1 RNAi. Extracts were generated from these animals as well as the progeny of these animals (which were not treated directly with oma-1 RNAi). NRDE-2::3×Flag was immunoprecipitated with an anti-Flag antibody and NRDE-2 co-precipitating oma-1 pre-mRNA was quantified by qRT–PCR with exon–intron primer sets designed to detect unspliced RNAs (pre-mRNAs) of the oma-1 gene as well as a control germline expressed pre-mRNA gld-2. hrde-1 allele tm1200 and znfx-1 allele gg561 were used. Data are mean ± s.d. of the ratio of signals ± oma-1 RNAi; n = 3 biological replicates.
Extended Data Fig. 5 WAGO-4 is an Argonaute that localizes to the peri-nucleus and is required for RNAi inheritance.
a, oma-1(zu405) is a temperature-sensitive lethal (embryonic arrest at 20 °C) allele of oma-1. oma-1 RNAi suppresses oma-1(zu405) lethality and this effect is heritable5. Animals of the indicated genotypes were exposed to oma-1 dsRNA, and F1 to F5 progeny were grown in the absence of oma-1 dsRNA. Number of viable progeny of P0 (directly exposed to oma-1 RNAi) and inheriting generations (F1 to F6, grown in the absence of oma-1 RNAi) were scored (20 °C). Data are mean ± s.d. of three biologically independent samples. b, Animals of the indicated genotypes and expressing a pie-1::gfp::h2b transgene were exposed to gfp dsRNA17. Micrographs of animals +/− gfp RNAi as well as the F1 progeny of these animals are shown. The percentage of animals expressing GFP is indicated, and represent the scoring of at least 90 animals in each generation and for each genotype. c, We used CRISPR–Cas9 to append a gfp tag upstream of the predicted wago-4 atg start codon. Top, fluorescent micrographs of gfp::wago-4 in 2-cell, 4-cell and ~300-cell embryos. Bottom, fluorescent micrograph of the germline of an adult gfp::wago-4 animal. Images are representative of more than three animals at each lifestage.
Extended Data Fig. 6 Visualization of Z granule formation with antibodies targeting PGL-1 (P granule) and HA::ZNFX-1 (Z granule).
To control for possible artefacts caused by fluorescent epitopes, we conducted immunofluorescence on HA::ZNFX-1-expressing animals using anti-PGL-1 (K76 Developmental Studies Hybridoma Bank) and anti-HA (Abcam ab9110) antibodies. a, Anti-PGL-1 and anti-HA signals colocalized in the P2 blastomeres of 4-cell embryos. b, Anti-PGL-1 and anti-HA signals were adjacent, yet distinct in, in pachytene germ cells. No PGL-1 or HA::ZNFX-1 signal was detected in pgl-1(bn101) animals, which do not express PGL-1 or HA::ZNFX-1, establishing that immunofluorescent signals were specific. c, Magnification of foci from a and b. Images in a–c are representative of three independent animals at each life stage. Scale bars, 1 μm (a), 1 μm (b) and 0.5 μm (c).
Extended Data Fig. 7 Z granules independently form liquid-like condensates that do not colocalize with P bodies but localize adjacent to EGO-1 foci, and can be physically and temporally dissociated from P granules.
a, During oocyte maturation, ZNFX-1 foci detach from the nuclei, assume spherical shapes and move away from the nucleus; this behaviour is consistent with Z foci being liquid-like condensates. In addition, the data show that Z foci can exist at developmental stages during which P granules are no longer visible, indicating that Z foci can be temporally separated from P granules. Image is of maturing oocytes of animals expressing the indicated fluorescent proteins. Long arrows indicate oocytes that contain Z granules, but not P granules. Image representative of more than three animals. b, Z foci exhibit properties reminiscent of liquid droplets. Left, GFP::ZNFX-1-expressing animals were subjected to FRAP (see Methods) and fluorescence was monitored in bleached area over indicated time. Data are normalized to a non-bleached control granule from the same sample. Data are mean ± s.e.m. of n = 7 individual granules from 7 animals. Right, heat maps showing recovery of ZNFX-1 fluorescence in a representative bleached Z granule. c, Z foci do not colocalize with other known liquid droplets. GFP::ZNFX-1 does not colocalize with markers of processing bodies. PATR-1 and DCAP-1 localize to processing bodies15. Fluorescent micrographs of somatic blastomeres of embryos expressing the indicated fluorescent proteins. ZNFX-1 does not colocalize with markers of processing bodies in these cells. Images are representative of more than three independent animals. d, Z foci in adult pachytene germ cells do not colocalize with EGO-1. ZNFX-1 foci form adjacent to EGO-1 foci. Fluorescent micrographs of a single pachytene germ cell nucleus from animals expressing GFP::ZNFX-1 and TagRFP::EGO-1. A 3D render of a representative foci is shown below. Images are representative of three independent animals. Scale bars, 0.5 μm. e, Z foci can be physically separated from P granules. Gonads were isolated from animals expressing GFP::ZNFX-1 and PGL-1::TagRFP and subjected to shearing force as described8. Time-lapse imaging at 10-s intervals is shown. A PGL-1-labelled P granule detaching from the nucleus and flowing throughout the cytoplasm is shown (large arrow). ZNFX-1-labelled Z granules remain immobile (small arrow). Physical shearing was induced as previously described8. In brief, GFP::ZNFX-1 and PGL-1::TagRFP adults were dissected to extrude gonads. Isolated gonads were squeezed between two coverslips to generate shearing force. Coverslips were then mounted on a slide and imaged immediately with a spinning disc confocal. Z stacks were acquired every 10 s. Images are representative of four independent animals.
Extended Data Fig. 8 Quantification of centres and surfaces of fluorescence for Z granules, P granules and Mutator foci.
a, b, Distances between centres (a) and surfaces (b) of the spaces occupied by PGL::mCardinal, TagRFP::ZNFX-1 and MUT-16::GFP were calculated as described in the Methods. Data are mean ± s.d. of 10 granule measurements across 3 independent animals. Distances have been corrected for chromatic shift.
Extended Data Fig. 9 P granule assembly factors contribute to RNAi inheritance, normal Z granule morphology and the ability of ZNFX-1 to bind mRNAs.
a, DEPS-1 is required for P granule formation in adult germ cells11,37,38. deps-1(bn124)/+ animals expressing the pie-1::gfp::h2b transgene17 were exposed to gfp dsRNA. Progeny were grown in the absence of gfp dsRNA for three generations. Fluorescent micrographs show GFP expression in oocytes. The percentage of animals expressing pie-1::gfp::h2b is shown. These data show that DEPS-1 activity is required in inheriting generations to allow for gfp RNAi inheritance. n = 32 animals for P0, n > 100 for F1 to F3. b, MEG-3/4, DEPS-1 and PGL-1 also contribute to P granule formation11,37,38. dpy-11 RNAi causes animals exposed to dpy-11 dsRNA to become Dumpy (Dpy). Progeny of animals exposed to dpy-11 dsRNA inherit dpy-11 silencing and are Dpy35. RNAi inheritance mutants become Dpy in response to dpy-11 RNAi; however, progeny fail to inherit dpy-11 silencing, and, therefore, are not Dpy. Animals of indicated genotypes were exposed to dpy-11 dsRNA. F1 progeny were grown in the absence of dpy-11 dsRNA. (−) indicates non-Dpy; (+) indicates mild Dpy phenotype; (++) indicates strong Dpy. pgl-1, deps-1, and meg-3/4 are defective for dpy-11 RNA inheritance. n = 3 biologically independent samples for each condition. c, Animals of indicated genotypes were exposed to dpy-11 dsRNA and F1 progeny were grown in the absence of dpy-11 dsRNA. Body lengths of F1 animals were measured by Image J. Data are expressed as body length from progeny of dpy-11 RNAi-treated animals divided by the average body length from control animals. The mean of n > 12 animals with P values calculated by Student’s two-tailed t-test is shown. d, In deps-1(bn124) animals, most Z granules are smaller than normal while one Z granule/nucleus becomes enlarged. Images are from pachytene region of germline. Images are representative of more than three animals. e, In deps-1(bn124) animals, ZNFX-1 does not bind RNA. Wild-type or 3×Flag::ZNFX-1-expressing animals were treated with oma-1 dsRNA. ZNFX-1 was immunoprecipitated in RNAi generation with anti-Flag antibodies and co-precipitating RNA was subjected to qRT–PCR to quantify oma-1 mRNA co-precipitating with ZNFX-1 in wild-type or deps-1(bn124) animals. gld-2 is a germline-expressed control mRNA. Data are mean ± s.d. of three biologically independent samples. f–i, Loss of ZNFX-1 or WAGO-4 does not seem to affect the formation of Z granules marked by GFP::WAGO-4 or GFP::ZNFX-1 (f), Mutator foci marked by MUT-16::GFP (g), or P granules marked by PGL-1::RFP (h, i). Note, in late embryonic germline development, PGL-1::TagRFP foci may not be efficiently concentrated into Z2/Z3 in wago-4 mutant (data not shown). Images are representative of more than three animals. All images in d–i were taken using a 60× objective, and scaled to the same size as the other images within a panel. Scale bars, 5 μm (f, i).