Germ Granules Coordinate RNA-Based Epigenetic Inheritance Pathways

Abstract

Germ granules are biomolecular condensates that promote germ cell totipotency in animals. In C. elegans, MEG-3 and MEG-4 function redundantly to assemble germ granules in germline blastomeres. Here, we show that meg-3/4 mutant animals exhibit defects in RNA interference (RNAi) that are transgenerationally disconnected from the meg-3/4 genotype. Similar non-Mendelian inheritance is associated with other mutations disrupting germ granule formation, indicating that loss of germ granules is the likely cause of the observed disconnects between genotype and phenotype. meg-3/4 animals produce aberrant siRNAs that are propagated for ≅10 generations in wild-type descendants of meg-3/4 ancestors. Aberrant siRNAs inappropriately and heritably silence germline-expressed genes including the RNAi gene sid-1, suggesting that transgenerational silencing of sid-1 underlies inherited defects in RNAi. We conclude that one function of germ granules is to organize RNA-based epigenetic inheritance pathways and that germ granule loss has consequences that persist for many generations.

Keywords: Transgenerational epigenetic inheritance; endo-siRNA; germ granules; non-Mendelian inheritance; non-coding RNA; phase separation.

Figure 1.. Transgenerational disconnect between meg-3/4 genotype and phenotype.

(a) L2 larvae of the indicated genotypes were fed either bacteria expressing dsRNAs derived from the pos-1 gene, which is required for embryonic viability, or bacteria containing the control vector, L4440. When the animals became adults, they were allowed to lay broods, and % hatching embryos was scored. Black dots represent individual broods (n = 18). Error bars represent +/− standard deviations of the mean (gray bars). (b) Top panel: schematic of genetic crosses. We first marked meg-3(tm4259) meg-4(ax2026) with dpy-3(e27) (dpy-3 is 0.1 cM from meg-3 and 0.8 cM from meg-4). Then, meg-3(tm4259) meg-4(ax2026) dpy-3(e27) animals were crossed to wild-type males, and heterozygous progeny were identified by PCR-based genotyping of both meg-3 and meg-4. As a mutant control cross, meg-3(tm4259) meg-4(ax2026) dpy-3(e27) animals were crossed to meg-3(tm4259) meg-4(ax2026) males. Lines that were either homozygous (control cross) or heterozygous for meg-3/4 mutations were maintained for 22 generations (data presented later in the paper will clarify why this was necessary for this experiment). All lines remained heterozygous for dpy-3(e27). The progeny of homozygous or heterozygous meg-3/4 lines were isolated, and meg-3 and meg-4 were genotyped by PCR. These progeny were used to establish 6 lineages that were maintained for 14 generations under normal growth conditions. Bottom panel: at the indicated generations, animals from each lineage were exposed to pos-1 RNAi and % hatching embryos was scored. Black dots represent individual lineages, colored bars represent the median value of % viable progeny. (c) A meg-3(tm4259) meg-4(ax2026) dpy-3(e27) chromosome was maintained in a heterozygous state for ≅45 generations in animals that were homozygous for two fluorescent protein markers: pgl-1::rfp (Wan et al. 2018), which marks P granules (magenta), and gfp::h2b (Ashe et al. 2012), which marks chromatin (green). PGL-1::RFP and GFP::H2B were visualized in first-generation meg-3/4 animals (indicated by Dpy phenotype) and their meg-3/4(+) and ++/meg-3/4 siblings (indicated by non-Dpy phenotype). Fluorescent micrographs of three embryos in the uterus of one adult are shown. Arrows indicate P granules. The percentage of F₁ adults containing embryos with normal PGL-1::RFP expression and the number of adults scored are indicated. Scale bar, 10 microns. See also Figure S1.

Figure 2.. Ancestral loss of P granules is associated with phenotypic hangovers.

(a) Schematic of genetic crosses used to generate animals scored for RNAi responsiveness in (b). Note: for this experiment, meg-3(tm4259) meg-4(ax2026) animals had been maintained in a homozygous state for dozens of generations prior to outcross. (b) 15 lineages were established from F₂ progeny and were maintained for 24 generations under normal growth conditions. At the indicated generations, animals from each lineage were exposed to pos-1 RNAi and % hatching embryos was scored. Black dots represent individual lineages, colored bars represent the median value of % viable progeny. (c) deps-1(bn124) animals were crossed to wild-type males and descendants of the crosses were scored for pos-1 RNAi sensitivity as described in (b). (d) rde-4(ne301) animals were crossed to wild-type males, and wild-type and rde-4 progeny were scored for pos-1 RNAi sensitivity. See also Figure S2.

Figure 3.. MEG-3/4 help organize endo-siRNA pathways.

(a) gfp::h2b; pgl-1::rfp; meg-3/4 dpy-3 animals that had been maintained in the homozygous state for dozens of generations were crossed to gfp::h2b; pgl-1::rfp males (WT). In the F₂ generation, meg-3/4(+) and ++/meg-3/4 adults (indicated by non-Dpy phenotype) or meg-3/4 homozygous adults (indicated by Dpy phenotype) were imaged. Fluorescent micrographs of three embryos in the uterus of one adult are shown. Arrows indicate P granules. The percentage of F₂ adults containing embryos with normal PGL-1::RFP expression and the number of adults scored are indicated. Scale bar, 10 microns. (b) Loss of HRDE-1 or NRDE-2 suppresses the RNAi defect associated with meg-3/4. Individual animals of the indicated genotypes were scored for pos-1 RNAi sensitivity as described in Figure 1A (n = 18). Error bars represent +/− standard deviations of the mean (gray bars). WT, wild type. (c) Volcano plot showing log2 fold change in the # of endo-siRNAs targeting each C. elegans gene in meg-3(tm4259) meg-4(ax2026) animals relative to wild type on the x-axis and the −log10 adjusted p-value on the y-axis. Dots shown in red indicate endo-siRNA pools that were significantly different between genotypes. rde-11 and sid-1 endo-siRNA pools are labeled for reasons that will become clear in Figure 5. (d) The number of endo-siRNAs targeting all MEG-3/4-regulated genes normalized to the total number of small RNAs sequenced from each sample are shown for each replicate of wild type and meg-3/4. k=1000. (e) Overlap between the list of MEG-3/4-regulated genes and published lists of genes targeted by CSR-1-bound endo-siRNAs (Claycomb et al. 2009), genes in the WAGO class (Gu et al. 2009), and genes targeted by HRDE-1-bound endo-siRNAs (Buckley et al. 2012). Numbers of genes overlapping and non-overlapping between lists are indicated. See also Figure S3 and Table S1.

Figure 4.. Wild-type descendants of meg-3/4 inherit aberrant endo-siRNA populations.

(a) Z scores of the levels of endo-siRNAs targeting each MEG-3/4-regulated gene. Biological replicates were plotted side-by-side for wild type (WT), meg-3(tm4259) meg-4(ax2026) (meg-3/4), and wild-type animals descending from meg-3/4 mutant animals (P₀) for the indicated number of generations (F₃-F₂₅). Genes were sorted into groups via hierarchical clustering (complete linkage method). The three major clusters of endo-siRNAs are shown and are indicated as Class 1–3. (b) Biplot of the top two principal components (PC1 and PC2) determined by principal component analysis. Points represent individual small RNA sequencing libraries, which are color-coded by genotype and by generation. Two biological replicates were analyzed for each genotype and each generation. Each PC represents a source of variation, and the distribution of points along a given PC axis indicates the degree of correlation between libraries for that PC. Note: PC1 explains most of the variation in the sequencing data. Prior to PCA, read counts were subjected to a regularized log transformation using the rlog function in DESeq2 (Love et al. 2014). See also Figure S4.

Figure 5.. Ancestral loss of P granules is associated with heritable changes in sid-1 expression.

(a) endo-siRNAs sequenced from animals of the indicated genotypes (and generations after outcross) that map to the sid-1 locus. A schematic of the sid-1 locus is shown below. Counts were normalized to total number of reads. (b) Top panel: endo-siRNA reads mapping to the sid-1 locus were quantified in replicates and across generations during Rde hangovers. Counts were normalized using the median of ratios method (DESeq2). Replicates are shown in red and blue. k=1000. Bottom panel: qRT-PCR was used to quantify sid-1 mRNA in the samples shown in the top panel. sid-1 mRNA values are shown relative to the mRNA values of nos-3, a germline-expressed gene. (c) Adults of the indicated genotype were injected with 200 ng/ul of pos-1 dsRNA in either one or both arms of the germline (n ≥ 9 animals per condition). After 18–22 hours of recovery, animals were allowed to lay eggs and % hatching embryos was scored. Some of the variability observed in meg-3/4 and sid-1 animals may be due to the fact that responses of sid-1 mutants to dsRNA injection change with time post-injection, and the timescale of this change varies from animal to animal(Wang & Hunter 2017). Error bars represent +/− standard deviations of the mean (gray bars). **, p-value < 0.01; ***, p-value < 0.001 (Student’s t-test). WT, wild type. (d) Model for initiation of Rde hangovers: P granules coordinate and organize endogenous small interfering RNA (endo-siRNA) biogenesis by concentrating endo-siRNA pathway factors (blue ovals) such as RNA-dependent RNA polymerases (RdRPs) and Argonautes (AGOs) together with the appropriate mRNAs. In the absence of MEG-3/4 and therefore P granules, endo-siRNA pathway factors engage the “wrong” pool of mRNAs (long red lines) and thereby initiate the production of aberrant endo-siRNAs (short red lines). (e) Model to explain transgenerational disconnects between meg-3/4 genotype and phenotype. Upon the loss of MEG-3/4, germ cells begin to produce slightly higher-than-normal levels of endo-siRNAs that target one or more genes required for experimental RNAi (RNAi gene-x). Due to the heritable nature of the endo-siRNA pathway, endo-siRNAs targeting RNAi gene-x accumulate slowly over generations in the absence of MEG-3/4 (lag). After 5–9 generations of aberrant endo-siRNA production/inheritance, these endo-siRNAs reach a level that causes silencing of RNAi gene-x; hence the defect in RNAi. After re-introduction of MEG-3/4, aberrant endo-siRNA pools continue to propagate across generations and continue to silence RNAi gene-x in genetically wild-type animals (hangover). Over the course of ten generations, genetic systems reassert their control over endo-siRNA biogenesis, returning endo-siRNA pools to normal. See also Figure S5.