piRNAs Coordinate poly(UG) Tailing to Prevent Aberrant and Perpetual Gene Silencing

Summary

Noncoding RNAs have emerged as mediators of transgenerational epigenetic inheritance (TEI) in a number of organisms. A robust example of RNA-directed TEI is the inheritance of gene silencing states following RNA interference (RNAi) in the metazoan C. elegans. During RNAi inheritance, gene silencing is transmitted by a self-perpetuating cascade of siRNA-directed poly(UG) tailing of mRNA fragments (pUGylation), followed by siRNA synthesis from poly(UG)-tailed mRNA templates (termed pUG RNA/siRNA cycling). Despite the self-perpetuating nature of pUG RNA/siRNA cycling, RNAi inheritance is finite, suggesting that systems likely exist to prevent indefinite RNAi-triggered gene silencing. Here we show that, in the absence of Piwi-interacting RNAs (piRNAs), an animal-specific class of small noncoding RNA, RNAi-based gene silencing can become essentially permanent, lasting at near 100% penetrance for more than five years and hundreds of generations. This perpetual gene silencing is mediated by continuous pUG RNA/siRNA cycling. Further, we find that piRNAs coordinate endogenous RNAi pathways to prevent germline-expressed genes, which are not normally subjected to TEI, from entering a state of continual and irreversible epigenetic silencing also mediated by continuous maintenance of pUG RNA/siRNA cycling. Together, our results show that one function of C. elegans piRNAs is to insulate germline-expressed genes from aberrant and runaway inactivation by the pUG RNA/siRNA epigenetic inheritance system.

eTOC Blurb

Shukla et al. show that piRNAs limit intergenerational RNAi-induced gene silencing in C. elegans. In the absence of piRNAs, mRNAs targeted by RNAi, as well as endogenous germline-expressed mRNAs, undergo aberrant and perpetual gene silencing driven by a self-perpetuating silencing loop of poly(UG)-tailed RNAs and small RNAs.

Introduction

Small noncoding RNAs, such as PIWI-interacting RNAs (piRNAs), microRNAs (miRNAs), and small interfering RNAs (siRNAs) are key regulators of gene expression in eukaryotes. Small RNAs are bound by Argonaute proteins and, together, these ribonucleoprotein complexes target and silence complementary mRNAs^1,2. piRNAs are an animal-specific class of genomically encoded and germline-expressed small noncoding RNA, which are bound by the PIWI clade of Argonaute proteins and are essential for germ cell function in many species of animals. One widely conserved function of piRNAs is to silence mobile genetic elements termed transposons^3,4.

The functions of some piRNAs, however, remain enigmatic. For instance, most mammalian piRNAs do not exhibit complementarity to transposable elements^5–8 and loss of Miwi, one of three PIWI proteins encoded by the mouse genome, results in male sterility, but does not cause transposon mobilization⁹. Thus mammalian piRNAs have important biological functions, in addition to transposon silencing. Similarly, the biological functions of most C. elegans piRNAs (also known as 21U-RNAs¹⁰) are mysterious. C. elegans piRNAs are bound and stabilized by PRG-1, one of two PIWI clade Argonaute proteins encoded by the C. elegans genome^11–16. Surprisingly, fewer than 1% of transposon families are upregulated in a prg-1 mutant¹⁶, which lacks all piRNAs^11–16, indicating that transposon silencing is not the sole or possibly even major function of C. elegans piRNAs.

While C. elegans piRNAs do not play a vital role in transposon silencing, they do repress the expression of a number of germline-expressed mRNAs^11,13–17, as well as germline-expressed transgenes^18–20. Current models posit that PRG-1/piRNA complexes drive gene and transgene silencing by binding complementary mRNAs and recruiting RNA-dependent RNA Polymerases (RdRPs)^14,17. RdRPs use target mRNAs as templates to synthesize 22G-siRNAs, which are antisense to mRNA templates, 22nt in length, and begin with a guanosine²¹. These piRNA-dependent 22G-siRNAs are bound by a worm-specific clade of Argonaute proteins (WAGOs) and, together, these ribonucleoprotein complexes mediate target mRNA silencing²¹. Interestingly, PRG-1 and its bound piRNAs interact with >16,000 mostly germline-expressed mRNAs^22,23. However, fewer than 100 of these mRNAs undergo piRNA-dependent gene silencing¹⁶, indicating that PRG-1 and piRNAs do not silence most of the mRNAs to which they are bound. Indeed, several recent studies show that C. elegans piRNAs can promote the expression of some mRNAs by preventing aberrant siRNA-mediated gene silencing^15,16,24,25. Why some genes become aberrantly silenced in a prg-1 mutant and how PRG-1 and piRNAs normally prevent this silencing is a mystery.

PRG-1/piRNA-induced silencing of some loci, such as transgenes, can be heritable^18,19. This heritable silencing is initiated by PRG-1 and piRNAs; however, once established, it can be maintained in the absence of PRG-1 and piRNAs for many generations^18–20. Such heritable gene silencing, in the absence of initiating triggers, is an example of transgenerational epigenetic inheritance (TEI), which in C. elegans is also known as RNA-induced epigenetic silencing (RNAe)^18,20. piRNA-induced transgenerational silencing correlates with both the heritable expression of histone 3 lysine 9 trimethylation (H3K9me3) at genomic loci undergoing RNAi inheritance and of 22G-siRNAs antisense to mRNAs transcribed from these loci. The process also depends on nuclear factors, such as the nuclear-localized WAGO HRDE-1, the HP1 homolog HPL-2, and histone methyltransferases^18,19, suggesting that 22G-siRNAs and repressive post-translational histone modifications likely contribute to the heritable gene silencing initiated by piRNAs.

Double-stranded RNAs (dsRNAs) can also initiate TEI in C. elegans^{18,19,26–31} via a conserved gene silencing program termed RNA interference (RNAi)²⁶. In C. elegans, RNAi begins when dsRNA is processed into siRNAs, which are bound by the Argonaute RDE-1^32,33. Together RDE-1/siRNA complexes bind target mRNAs via Watson-Crick base-pairing between the mRNA and the siRNA, resulting in target mRNA cleavage by the ribonuclease RDE-8³⁴. RdRPs then use fragmented mRNAs to generate 22G-siRNAs, which, like piRNA-directed 22G-siRNAs, are bound by WAGOs to carry out gene silencing^21,35. RNAi-directed silencing of germline-expressed mRNAs can be inherited (termed RNAi inheritance) and, like piRNA-directed heritable silencing of transgenes, RNAi inheritance depends on HRDE-1 and is correlated with the inheritance of 22G-siRNAs and repressive repressive histone modifications, such as H3K9me3^18,19,29 and H3K27me3³⁶. The convergence of RNAi-directed and piRNA-directed TEI on a common set of downstream gene silencing effector proteins, such as RdRPs and WAGOs (like HRDE-1), suggests that these two TEI pathways are related.

Maintenance of 22G-siRNA expression over generations after dsRNA-triggered TEI depends upon a recently discovered noncoding RNA modification³⁷. mRNAs targeted by RNAi and cleaved by the RNAi machinery, such as the endonuclease RDE-8³⁴, are modified with perfectly alternating 3’ uridine (U) and guanosine (G) repeats (termed poly(UG) or pUG tails) by the ribonucleotidyltransferase MUT-2/RDE-3^37,38. pUG tails recruit RdRPs, which then use pUG RNAs as templates to generate 22G-siRNAs³⁷. Generationally repeated rounds of pUG RNA-templated 22G-siRNA synthesis and 22G-siRNA-directed mRNA pUGylation (termed pUG RNA/siRNA cycling) appear to be the mechanism by which gene silencing memories are propagated across generations in C. elegans³⁷. Although not yet tested, piRNA-directed TEI, which also requires RDE-3¹⁸, is also likely to depend upon pUG RNA/siRNA cycling. The self-perpetuating nature of the pUG RNA/siRNA cycling pathway hints that biological systems likely exist to limit and prevent this potentially dangerous pathway from mistargeting essential genes for generationally stable silencing. In support of this idea, RNAi inheritance is usually finite and recent studies have shown that the generational perdurance of RNAi inheritance is under genetic control^{19,29,39–41}. For instance, the methyltransferase MET-2⁴⁰ and the chromodomain protein HERI-1⁴¹ limit the number of generations that RNAi is inherited in C. elegans. Interestingly, HERI-1 is physically recruited to the chromatin of genes undergoing RNAi inheritance, suggesting that HERI-1 may play a direct role in limiting TEI⁴¹. Little else is known about how C. elegans might regulate and focus its potent, self-perpetuating, and potentially dangerous TEI pathways.

Here we show that piRNAs regulate the accuracy and duration of pUG RNA/siRNA-based transgenerational gene silencing in the C. elegans germline. In the absence of piRNAs, RNAi-initiated pUG RNA/siRNA cycling becomes essentially permanent. Endogenous pUG RNA/siRNA cycling becomes disorganized, with genes normally expressed in the germline becoming inappropriate targets for poly(UG) tailing and undergoing gene silencing. We conclude that one function of C. elegans piRNAs is to protect germline-expressed genes from aberrant and runaway gene silencing by the pUG RNA/siRNA epigenetic inheritance pathway.

Results

Identification of mutations that cause RNAi to become essentially permanent.

We previously conducted a forward genetic screen (Figure 1A) to identify factors that normally limit the duration of dsRNA-triggered TEI (RNAi inheritance) in C. elegans⁴¹. For this screen, we mutagenized animals harboring: (1) oma-1(zu405), a temperature-sensitive, gain-of-function mutation in the germline-expressed gene oma-1 that causes embryonic arrest, unless oma-1 is silenced by RNAi⁴²; and (2) a pie-1::gfp::h2b transgene, which encodes a GFP::H2B fusion protein driven by a germline-expressed promoter (hereafter referred to as gfp). In wild-type animals, dsRNA-induced silencing of oma-1 or gfp lasts for four to ten generations after initiating dsRNA triggers are removed^{19,28,29,39,41}. Our screen identified 20 Heritable enhancers of RNAi (Heri) mutants that exhibited oma-1 and gfp RNAi inheritance for seven or more generations longer than non-mutagenized controls⁴¹. oma-1 and gfp expression was eventually restored within 20 generations after RNAi in 18 of our 20 mutants⁴¹. However, two mutants, gg531 and gg540, were unique in that, more than five years, and hundreds of generations, after gfp and oma-1 were initially targeted for silencing by RNAi, 100% of animals in these lineages continued to silence both gfp (Figure 1B) and oma-1 (Figure 1C). Data presented below will show that the silencing of gfp and oma-1 observed in gg531 and gg540 animals is epigenetic in nature. Henceforth, we refer to this remarkably stable epigenetic inheritance as “perpetual silencing.” We conclude that systems exist to prevent dsRNA-initiated epigenetic inheritance from becoming essentially permanent.

Figure 1. RNAi-mediated gene silencing can become perpetual in C. elegans.

(A) oma-1(zu405); gfp animals were mutagenized with ethyl methanesulfonate (EMS) and fed bacteria expressing oma-1 and gfp dsRNA (RNAi). In wild-type animals, oma-1 and gfp RNAi can induce heritable silencing for several generations after dsRNA triggers are removed^{19,28,29,39,41}. heritable enhancers of RNAi (heri) mutations that extended both oma-1 and gfp RNAi inheritance by seven or more generations were identified⁴¹. Whereas most Heri mutants re-expressed oma-1 and gfp over the next 20 generations, gg531 and gg540 animals continue to inherit oma-1 and gfp RNAi-mediated gene silencing at one hundred percent penetrance more than five years and hundreds of generations after RNAi.

(B) Fluorescence micrographs (63X oil objective) showing gfp expression in oocyte nuclei of animals of the indicated genotypes five years after gg531 and gg540 animals were exposed to gfp RNAi. >100 animals per genotype were scored using the 20X objective of a fluorescent microscope (Zeiss) and the % animals expressing gfp is indicated. If any GFP expression was detected, animals were scored as expressing gfp.

(C) oma-1(zu405) animals lay viable progeny at 15°C (permissive temperature), but lay arrested embryos at 20°C (nonpermissive temperature) unless oma-1 is silenced by RNAi⁴². Therefore, the percentage (%) of oma-1(zu405) embryos that hatch [(# of hatched embryos / # embryos laid) × 100] at the nonpermissive temperature is a readout of oma-1 silencing. For each replicate, % hatched embryos was measured and averaged for 6 individual animals of each of the indicated genotypes five years after gg531 and gg540 animals were exposed to oma-1 RNAi. Error bars are standard deviation (s.d.) of the mean of the three biological replicates.

See also Figure S1.

PRG-1 limits RNAi inheritance.

Whole-genome sequencing of gg531 and gg540 animals identified independent nonsense mutations in the gene prg-1 (Figure 2A), which encodes one of two C. elegans PIWI clade Argonaute proteins^11–13. Previous studies have shown that PRG-1 binds C. elegans piRNAs^11–13 and, together, this complex can initiate transgenerational silencing of germline-expressed genes and transgenes via a mechanism involving piRNA-directed RdRP-based synthesis of 22G-siRNAs^18,19. Our results suggest that, surprisingly, PRG-1 and piRNAs can also restrict transgenerational gene silencing initiated by exogenous dsRNA triggers. The following lines of evidence support this idea. First, we used genetic crosses to remove the silenced gfp allele from prg-1(gg531) animals and to reintroduce an expressed allele of gfp. We then tested these gfp expressing prg-1(gg531) animals for gfp RNAi inheritance and confirmed that these animals exhibited enhanced gfp RNAi inheritance (Figure S1A). Second, we asked whether animals harboring an independently isolated deletion allele, tm872, of prg-1 exhibited an enhanced RNAi inheritance phenotype. tm872 is a 640bp deletion in prg-1 that removes most of the MID domain and part of the PIWI domain (Figure 2A), two conserved domains found in Argonaute proteins², and is, therefore, thought to represent a null allele of prg-1¹³. Indeed, the generational perdurance of RNAi-mediated gene silencing was dramatically enhanced in prg-1(tm872) animals (Figure 2B and Figure S1B). Together, these data confirm that loss of PRG-1 can enhance the generational perdurance of RNAi-induced gene silencing.

Figure 2. PRG-1 prevents perpetual RNAi inheritance.

(A) Schematic of the prg-1 gene, which encodes a PIWI clade Argonaute protein, indicating: (1) conserved PAZ, MID and PIWI domains; and (2) nature and locations of prg-1 alleles used in this study. prg-1(gg531) and prg-1(gg540) are two nonsense alleles obtained from the forward genetic screen reported in Perales et al., 2018. prg-1(tm872) is an independently isolated partial deletion in prg-1¹³. prg-1 coordinates were obtained from Wormbase release WS279.

(B) gfp RNAi inheritance assay performed on animals of the indicated genotypes. For each replicate, >50 animals per genotype were scored in each generation and the % of animals expressing gfp is shown. Error bars represent s.d. of the mean for three biological replicates.

(C) 5–6 prg-1(tm872) animals (represented by each point) in which gfp was either re-expressed or still silenced were singled 27 generations after gfp dsRNA treatment from each of the three biological replicates shown in panel (B). Lineages were maintained for five additional generations and then gfp expression was scored.

See also Figures S1 and S2.

Of note, some, but not all, prg-1(tm872) animals continued to inherit gfp silencing 27 generations after dsRNA treatment (Figure 2B). Given that the prg-1 mutants we identified in our genetic screen continue to show 100% penetrant silencing five years after RNAi, we wondered whether, following gfp RNAi, a subset of prg-1 mutant animals enters a state of perpetual silencing in which they, and all of their progeny, exhibit fully penetrant gene silencing. To test this idea, we isolated individual prg-1(tm872); gfp animals in which gfp expression was either restored or still silenced 27 generations after gfp RNAi (Figure 2B), propagated lineages established from these individuals for an additional five generations and then scored gfp expression in each of the lineages (Figure 2C). This analysis showed that 27 generations after RNAi, individuals had entered one of two epigenetic states: either they and all of their progeny expressed gfp or they and all their progeny did not (Figure 2C and Figure S1C). For unknown reasons, the percentage of prg-1 animals entering the perpetually silenced state varied in different experiments (Figure 2B, Figure S1A and Figure S1B); however, in all cases, those animals that entered the perpetually silent state never exited (Figure 2C and Figure S1C). We conclude that, in the absence of PRG-1, dsRNA triggers one of two epigenetic states: finite gene silencing, in which gene expression is eventually restored after several generations, or perpetual gene silencing, in which targets of dsRNA are silenced forever.

We next asked if PRG-1 inhibits the initiation or maintenance of the RNAi-induced perpetually silenced state. We crossed prg-1(tm872) animals in which gfp was silenced 13 generations after gfp RNAi to wild-type animals. We then monitored gfp expression in the prg-1(+) and prg-1(tm872) progeny of this cross, which were all homozygous for gfp. Only prg-1(tm872) progeny continued to silence gfp 15 generations after outcross (Figure 3A), indicating that PRG-1 inhibits maintenance of perpetual RNAi-induced gene silencing. To ask if PRG-1 inhibits the initiation of perpetual silencing, we treated gfp expressing prg-1(gg531/+) heterozygous animals, which are wild-type for PRG-1 activity¹⁹, with gfp RNAi and scored the inheritance of gfp silencing in the prg-1(+) or prg-1(gg531) homozygous lineages of this cross. After 15 generations, some prg-1(gg531), but no prg-1(+), lineages were still inheriting gfp silencing (Figure 3B), indicating that PRG-1 does not inhibit the initiation of perpetual silencing, and confirming that PRG-1 inhibits the maintenance of gfp silencing after gfp RNAi. Taken together, these data show that PRG-1 acts in inheriting generations to prevent RNAi-initiated gene silencing from becoming essentially permanent.

Figure 3. PRG-1 antagonizes the maintenance of heritable silencing.

(A) prg-1(tm872); gfp animals were fed gfp dsRNA. 13 generations (gen) after dsRNA treatment, three animals in which gfp was re-expressed and three animals in which gfp was still silenced were singled and independently crossed to prg-1(+) (e.g. wild-type) males. Lineages (represented by each point) that were homozygous for gfp were established from one prg-1(+) and one prg(tm872) F₂ progeny from each of the crosses, maintained for 15 generations, and then scored for gfp expression. 30–80 animals were scored per lineage.

(B) gfp expressing animals of the indicated genotypes were exposed to gfp dsRNA (RNAi gen). Lineages (represented by each point) were established from singled F₂ progeny of dsRNA-treated animals and gfp expression was scored in these lineages 15 generations after dsRNA treatment (inheriting gen). 50 animals were scored per lineage.

C. elegans PRG-1 possesses endonuclease (Slicer) activity in vitro¹⁴, whose biological function is not yet known, but which is dependent on an evolutionarily conserved DDH motif (catalytic triad)^2,14. CRISPR/Cas9 mediated mutation of the catalytic triad, in a manner that disrupted PRG-1 Slicer activity in vitro¹⁴ (Figure S2A), did not result in perpetual gene silencing after RNAi (Figure S2B). Thus, Slicer activity is not required for PRG-1 to limit RNAi inheritance and, therefore, the biological purpose of this activity remains enigmatic. Further, whereas our data show that PRG-1 suppresses perpetual silencing triggered by exogenous dsRNA, the other C. elegans PIWI clade Argonaute, PRG-2, whose function is not yet known^11–13, did not have a role in limiting RNAi inheritance (Figure S2C).

Perpetually silenced alleles are paramutagenic.

Paramutation is an epigenetic gene silencing phenomenon, first documented in plants⁴³, whereby the epigenetic state of one allele is transmitted to another allele of the same gene. A related process occurs in the C. elegans germline, where it is also known as RNA-induced epigenetic silencing (RNAe)^18,20. While conducting genetic crosses with gg531 and gg540 animals, we noticed that the perpetually silenced oma-1 and gfp alleles in gg531 and gg540 animals exhibited paramutagenic properties. For instance, when we crossed prg-1(gg531) animals harboring perpetually silenced gfp or oma-1 alleles to animals harboring expressed alleles of gfp or oma-1 (identified by linked mutations in dpy-10 or dpy-20, respectively), expressed alleles were converted to the silenced state (Figure 4A and Figure S3A). This trans-silencing was highly penetrant, specifically when transmitted via the female germline (Figure S3A), and was maintained perpetually in lineages lacking PRG-1 (Figure 4A). We conclude that genes undergoing perpetual silencing in prg-1 mutant animals are paramutagenic, emphasizing the irreversible and continuous nature of RNAi silencing that can occur in the absence of PRG-1.

Figure 4. Perpetually silenced alleles are paramutagenic.

(A) prg-1(gg531) hermaphrodites harboring a perpetually silenced allele of gfp (gfp^OFF) were crossed to prg-1(+) males harboring an expressed allele of gfp that was marked by a tightly linked dpy-10(e128) mutation (dpy gfp^ON). Lineages (represented by each point) were established from singled F₂ progeny of the indicated genotypes and gfp expression was scored in the indicated generations. 30–100 animals were scored per lineage in each generation. Error bars are s.d. of the mean.

(B) prg-1(gg531); gfp^OFF hermaphrodites were crossed to prde-1(mj207); gfp^ON males. F₂ animals were singled, genotyped for prg-1(gg531) and prde-1(mj207) and lineages of the indicated genotypes were established from these animals. gfp expression was scored after the indicated number of generations. 30–90 animals were scored per generation for each genotype. Error bars are s.d. of the mean.

See also Figure S3.

piRNAs limit RNAi inheritance.

PRG-1 is a PIWI clade Argonaute protein that binds piRNAs in the C. elegans germline^11–13. Therefore, the perpetual silencing and paramutation we observe in animals lacking PRG-1 is likely due to the loss of piRNA function in the germline. In support of this idea, one of the twenty mutant strains identified by our genetic screen harbored a nonsense mutation, gg530, in the prde-1 gene (Figure S3B), which encodes a nuclear-localized protein required for the production and/or stability of some piRNA precursor transcripts⁴⁴. To further test the idea that PRDE-1, and, therefore, piRNAs prevent the maintenance of perpetual silencing, we asked whether animals harboring mj207, an independently isolated null allele of prde-1⁴⁴, also exhibited enhanced RNAi inheritance. We conducted gfp RNAi on gfp and prde-1(mj207); gfp animals and found that the loss of PRDE-1 had a complex and subtle effect on gfp RNAi inheritance (Figure S3C). Nine generations after dsRNA exposure, the loss of PRDE-1 caused a statistically significant enhancement of gfp RNAi inheritance (Figure S3C). Additionally, some prde-1(mj207) animals, but no control animals, continued to silence gfp 27 generations after RNAi (Figure S3C). The more subtle enhancement of gfp RNAi inheritance in prde-1(mj207) animals, as compared to prg-1 mutant animals, may be due to the fact that, unlike PRG-1, PRDE-1 is necessary for the production of only some, but not all, piRNAs in the C. elegans germline⁴⁴. Importantly, a related analysis that used paramutation to initiate gene silencing revealed a more dramatic and clearer role for PRDE-1 in the long-term maintenance of perpetual silencing states. We crossed prg-1(gg531) animals harboring a perpetually silenced gfp allele to gfp expressing prde-1(mj207) animals and isolated: (1) prg-1(+); prde-1(+), (2) prg-1(gg531) or (3) prde-1(mj207) progeny from this cross, all of which were homozygous for gfp (Figure 4B). We then monitored gfp expression over generations in these lineages and, as expected based upon results described above (Figure 4A and Figure S4A), paramutation-induced gene silencing was perpetual in prg-1(gg531) but not in prg-1(+); prde-1(+) lineages (Figure 4B). prde-1(mj207) lineages behaved like prg-1(gg531) lineages in that they maintained gfp silencing for far more generations than prg-1(+); prde-1(+) lineages (Figure 4B). We conclude that PRDE-1, like PRG-1, limits the generational perdurance of RNAi inheritance, suggesting that piRNAs prevent perpetual gene silencing in the C. elegans germline.

Perpetual silencing is driven by continuous siRNA production and co-transcriptional gene silencing.

Data in this section show that the perpetual silencing of RNAi-targeted genes in animals lacking PRG-1 is due to continuous maintenance of silencing pathways normally induced, but finite, after RNAi in wild-type animals. First, RNAi inheritance in wild-type animals is correlated with the heritable expression of 22G-siRNAs antisense to genes undergoing transgenerational silencing, which, in the case of oma-1 RNAi inheritance, persist for 4–5 generations^19,29. TaqMan-based siRNA quantification showed that oma-1 (Figure 5A and Figure S4A) and gfp (Figure S4B) siRNAs were still expressed hundreds of generations after gfp and oma-1 RNAi in both prg-1(gg531) and prg-1(gg540) animals. Second, during RNAi inheritance, genes are subjected to co-transcriptional gene silencing (cTGS), which, in the case of oma-1 RNAi inheritance, persists for 4–5 generations^19,29. qRT-PCR-based quantification of nascent oma-1 pre-mRNA levels showed that oma-1 remained co-transcriptionally silenced in prg-1(gg531) and prg-1(gg540) animals hundreds of generations after oma-1 had been targeted for silencing by RNAi (Figure 5B). Third, HRDE-1, a germline-expressed and nuclear-localized AGO, which is required for RNAi inheritance in wild-type animals^18,19,29, was required for perpetual gene silencing of gfp (Figure S5B) and oma-1 (Figure 5C) in prg-1(gg531) and prg-1(gg540) animals. Finally, DEPS-1⁴⁵ and ZNFX-1⁴⁵, two factors that are known to localize to germline condensates termed P granules and Mutator foci, respectively, and that are also known to contribute to RNAi inheritance in wild-type animals, were needed to maintain perpetual silencing of gfp and oma-1 in prg-1(gg531) and prg-1(gg540) animals (Figure 5D and Figure 5E). Together, these data show that the normally finite mechanisms underlying RNAi inheritance in the C. elegans germline become perpetual in the absence of PRG-1 and piRNAs.

Figure 5. Perpetual silencing is co-transcriptional and depends on known HRDE factors.

(A) Taqman-based qRT-PCR was used to quantify the expression of two different oma-1 siRNAs (probes 1 and 2) in the following animals, all harboring oma-1(zu405) and gfp: (1) prg-1(+) animals +/− oma-1 RNAi, (2) prg-1(gg531) animals, and (3) prg-1(gg540) animals, one year after prg-1(gg531) and prg-1(gg540) animals had been treated with oma-1 RNAi.

(B) qRT-PCR using two primer sets (primer set 1 and 2) was used to quantify oma-1 pre-mRNA and mRNA levels in animals of the indicated genotypes, two years after prg-1(gg531) and prg-1(gg540) animals had been treated with oma-1 RNAi. Animals homozygous for tm1396, a 1515bp deletion in the oma-1 gene (oma-1Δ), served as a negative control. oma-1 pre-mRNA and mRNA levels were normalized to the germline-expressed gene nos-3.

(C) CRISPR/Cas9 was used to create a 2885bp deletion (Δ) that removes conserved domains in the hrde-1 gene (Figure S5A) in prg-1(gg531) and prg-1(gg540) mutants. oma-1 silencing was quantified as in Figure 1C for the indicated genotypes in three biological replicates.

(D) Animals of the indicated genotypes were fed bacteria expressing vector control (−) or dsRNA targeting the gene znfx-1 (+) for two generations and gfp expression was then scored. Error bars represent s.d. of the mean of three biological replicates. Three biological replicates are shown, with 50 animals scored per treatment for each genotype.

(E) Control (oma-1(zu405); gfp) animals and prg-1(gg531) and prg-1(gg540) mutants were fed bacteria expressing empty vector control or dsRNA targeting the gene deps-1 for two generations and then gfp expression was scored.

(A-E) Error bars represent s.d. of the mean for three biological replicates.

See also Figures S4 and S5.

Perpetual pUG RNA/siRNA cycling underlies perpetual gene silencing.

RDE-3 is a ribonucleotidyltransferase required for RNAi^46,47 and RNAi inheritance³⁷ in C. elegans. RDE-3 adds poly(UG) or pUG tails to mRNAs targeted for silencing by RNAi (Figure 6A)³⁷. These pUG tails recruit RdRPs, such as RRF-1, to pUG RNAs, which then serve as templates for antisense siRNA synthesis by RdRPs during RNAi and RNAi inheritance³⁷. Generationally repeated sense/antisense cycles of pUG RNA-mediated siRNA biogenesis coupled with siRNA-directed mRNA pUGylation (pUG RNA/siRNA cycling) appear to mediate RNAi inheritance in the C. elegans germline (Figure 6A)³⁷. However, despite the self-perpetuating nature of pUG RNA/siRNA cycling, the process typically only lasts a finite number of generations in wild-type animals³⁷. The following data show that the perpetual silencing of oma-1 and gfp in prg-1(gg531) and prg-1(gg540) mutants is due to the maintenance of pUG RNA/siRNA cycling. First, we detected oma-1 and gfp pUG RNAs in both prg-1 mutant strains identified by our genetic screen, hundreds of generations after oma-1 and gfp had been targeted for silencing by dsRNA (Figure 6B). These non-templated pUG tails were added to oma-1 and gfp mRNA fragments (Figure S6A) at sites reminiscent of the pUGylation sites previously observed after RNAi in wild-type animals³⁷. Second, pUG RNA expression correlated with perpetual gene silencing. For instance, 3/3 prg-1(tm872) lineages that continued to silence gfp 33 generations after gfp RNAi (Figure 2C) expressed gfp pUG RNAs, while 3/3 lineages no longer silencing gfp did not (Figure 6C). Third, the poly(UG) polymerase RDE-3 was required for perpetual silencing. A 484bp deletion introduced into the rde-3 gene in prg-1(gg531) animals using CRISPR/Cas9 (Figure S6B) was sufficient to abrogate gfp and oma-1 pUG RNA expression (Figure 6B), and to halt perpetual gfp (Figure 6D) and oma-1 silencing (Figure 6E). We conclude that perpetual silencing of RNAi targets in animals lacking piRNAs is driven by continuous maintenance of the pUG RNA/siRNA pathway, indicating that PRG-1 and piRNAs normally limit the generational perdurance of pUG RNA/siRNA cycles.

Figure 6. Perpetual silencing is driven by perpetual pUG RNA/siRNA cycling.

(A) Model summarizing pUG RNA/siRNA cycling³⁷. siRNAs derived from exogenous dsRNAs induce the cleavage of target mRNAs³⁴. 5’ cleavage products are then modified with poly(UG) tails by the ribonucleotidyltransferase RDE-3. pUG RNAs serve as templates for 22G siRNA synthesis by RdRPs. A worm-specific clade of Argonaute proteins (termed WAGOs) binds 22G siRNAs and, together, this complex targets complementary mRNAs for: 1) transcriptional and translational silencing^29,35,65,66 (not shown here), and 2) cleavage and subsequent de novo pUGylation. Cycles of pUG RNA-based siRNA production and siRNA-directed mRNA pUGylation maintain silencing over time and across generations. Despite the self-perpetuating nature of pUG RNA/siRNA cycling, for unknown reasons, cycling lasts for a finite number of generations in wild-type animals.

(B) Three biological replicates of pUG PCR (see STAR Methods) to detect oma-1 and gfp pUG RNAs were performed on total RNA isolated from animals of the indicated genotypes. A deletion (Δ) was introduced in the rde-3 gene (Figure S6B) in prg-1(gg531) animals using CRISPR/Cas9.

(C) gfp pUG PCR was performed on total RNA isolated from gfp expressing and gfp silenced lineages that were established in Figure 2C. Animals were collected 33 generations after gfp dsRNA treatment.

(B-C) gsa-1, which has an 18nt genomically encoded pUG repeat in its 3’UTR, serves as a loading control.

(D) Fluorescence micrographs showing gfp expression in the oocytes of oma-1(zu405); gfp animals vs. prg-1(gg531) mutants with or without rde-3 deletion (Figure S6B). >100 animals of each genotype were scored.

(E) oma-1 silencing was measured in biological triplicate by quantifying the % embryos hatched at 20°C for animals of the indicated genotypes as described in Figure 1C.

See also Figures S6 and S7.

piRNAs prevent pUG RNA-based perpetual silencing of germline-expressed genes.

Recent studies have found that one function of C. elegans piRNAs is to coordinate endogenous 22G-siRNA systems in the germline^15,16,24,25 and that, in the absence of this coordination, some germline-expressed genes, most strikingly the replication-dependent histones, undergo 22G-siRNA-dependent aberrant gene silencing^15,16. We wondered if the mechanism underlying these previously documented cases of aberrant gene silencing in prg-1 mutants^15,16 might be aberrant pUG RNA/siRNA cycling. his-10/14/26, his-11/15/44 and his-12/16/43 are three sets of three nearly identical replication-dependent histones genes known to be subjected to aberrant silencing in prg-1 mutants^15,16. We first confirmed that mRNAs encoded by these genes were aberrantly silenced in prg-1(gg531) animals (Figure 7A). We next asked if this aberrant silencing was associated with aberrant histone pUG RNA production. Indeed, we detected his-10/14/26, his-11/15/44, and his-12/16/43 pUG RNAs in prg-1(gg531) animals, but not in prg-1(+) control animals (Figure 7B). Like pUG RNAs produced in response to dsRNA, histone pUG RNAs consisted of 5’ fragments of histone mRNAs modified with non-templated pUG tails (Figure S6C). Deletion of rde-3 in prg-1(gg531) animals (Figure S6B) restored histone gene expression to near wild-type levels (Figure 7A) and abolished histone pUG RNA expression (Figure 7B). Expression profiling and small RNA-seq studies have identified a number of other genes, in addition to the replication-dependent histones, whose expression is downregulated in prg-1 mutants^15,16. We wondered if aberrant silencing of these genes might also be explained by aberrant mRNA pUGylation. We tested this idea for the predicted protein-coding gene r03d7.2, which becomes downregulated in prg-1 mutants (Figure 7C)^15,16, and found that r03d7.2 mRNAs were, indeed, pUGylated (Figure 7D) and silenced (Figure 7C) in an RDE-3-dependent manner in prg-1(gg531), but not prg-1(+), animals. We observed similar results for 2 out of 3 additional protein-coding genes previously reported to be downregulated in prg-1 mutants^15,16. To confirm that this pUGylation induced aberrant silencing was a general phenomenon resulting from loss of PRG-1, we also analyzed prg-1(tm872) animals for mRNA and pUG RNA expression patterns^11,13. As expected, prg-1(tm872) exhibited aberrant silencing of histone and r03d7.2 mRNAs (Figure S6D) and produced aberrant histone and r03d7.2 pUG RNAs (Figure S6E). We conclude that the downregulation of some predicted protein-coding mRNAs in prg-1 mutants is mediated by aberrant and perpetual pUG RNA/siRNA cycling.

Figure 7. piRNAs coordinate the germline mRNA pUGylation system.

(A, C, E) qRT-PCR was performed in biological triplicate to quantify mRNA expression of the indicated genes using total RNA isolated from animals of the indicated genotypes. Results were normalized to the germline-expressed nos-3 mRNA. Error bars are s.d. of the mean for three biological replicates.

(B, D, F) Gene-specific pUG PCR assays (see STAR Methods) were performed in biological triplicate to detect pUGylation of the indicated mRNAs. gsa-1 is a loading control. *The same RNA samples were used for panels b, d and f, so gsa-1 loading control is identical in these panels.

See also Figures S6 and S7.

Finally, while some genes are known to undergo aberrant gene silencing in prg-1 mutants, other genes are upregulated^15,16. We wondered if upregulation of these genes might be due to loss of mRNA pUGylation and, therefore, be indicative of a global disorganization of the mRNA pUGylation system in the absence of piRNAs. We tested this idea for the bath-13 gene, which is upregulated in prg-1 mutants^15,16. We first confirmed that bath-13 mRNA levels are also upregulated in prg-1(gg531) mutants (Figure 7E). Interestingly, we found that bath-13 mRNA levels were similarly elevated in prg-1(gg531) and prg-1(gg531); rde-3Δ double mutant animals, suggesting that prg-1 and rde-3 act in the same genetic pathway to inhibit bath-13 expression (Figure 7E). We then asked if bath-13 mRNA upregulation in prg-1 mutants correlated with a loss of bath-13 mRNA pUGylation. Indeed, we found that bath-13 mRNAs are pUGylated in prg-1(+) animals and that this pUGylation is lost in prg-1(gg531) animals (Figure 7F) and in prg-1(tm872) animals (Figure S6E). Taken together, these data suggest that PRG-1 and piRNAs coordinate gene expression programs in the C. elegans germline by focusing the activity of the poly(UG) polymerase RDE-3 activity towards the appropriate mRNAs.

Discussion

Here we show that in the absence of PRG-1 and piRNAs, dsRNA-triggered TEI, which normally lasts for only a finite number of generations, can become essentially permanent. This perpetual silencing, whose fidelity of transmission approaches that of DNA-based inheritance, is driven by perpetual pUG RNA/siRNA cycling. Further, we find that, in the absence of piRNAs, the endogenous mRNA pUGylation system becomes disorganized, with some natural targets of RDE-3 escaping modification, and other mRNAs, including the replication-dependent histone RNAs, becoming novel substrates for RDE-3 (Figure S7). Together, these results show that piRNAs prevent germline-expressed mRNAs from entering self-perpetuating pUG RNA/siRNA cycles, thus protecting these mRNAs from undergoing perpetual and runaway transgenerational silencing. These data, along with recently published reports^15,16, indicate that the Argonaute PRG-1 possesses pro-expression functions in the C. elegans germline, as has been shown previously for the germline-expressed Argonaute CSR-1^48–52. The lack of aberrant silencing of CSR-1 target mRNAs in the absence of PRG-1¹⁶ hints that the mechanism by which PRG-1 and CSR-1 promote gene expression may be distinct.

How PRG-1/piRNAs regulate the targets of pUGylation in C. elegans remains a mystery, but here we propose some models that might explain this novel function of PRG-1/piRNAs. P granules are biomolecular condensates that form in C. elegans germ cells^53,54 and promote germ cell fate specification by transmitting maternal mRNAs and epigenetic factors to the germline precursor cell during embryonic development⁵⁵. The PRG-1/piRNA complex binds thousands of germline-expressed transcripts, and many of these interactions likely occur in P granules, where PRG-1 localizes^11,13. Recent work showed that binding by PRG-1/piRNAs may serve to sequester mRNAs to P granules, preventing these mRNAs from interacting with silencing machinery in the cytoplasm⁵⁶. Indeed, our work suggests this sequestration may prevent some transcripts from interacting with and, thus, being pUGylated by, RDE-3, which localizes to Mutator foci, distinct cytoplasmic perinuclear germline condensates required for 22G-siRNA amplification⁴⁷ (Fig. S11b). Interestingly, a related model has been proposed for piRNA function in Drosophila where piRNAs promote germline specification by binding to and entrapping maternal mRNAs inside germ granules, which are biomolecular condensates constituting part of the fly germplasm⁵⁷. Thus, the regulation of germline gene expression programs via sequestration of mRNAs within biomolecular condensates may be a conserved, direct, and ancient function of the animal piRNAs. Alternatively, PRG-1/piRNAs may coordinate the pUGylation system indirectly by competing with other RNAi pathways, such as the exogenous RNAi pathway, for the same set of downstream silencing factors (RdRPs and WAGOs)^19,35,58,59. According to this indirect “competition” model, in the absence of piRNAs, shared downstream silencing factors are freed up to maintain aberrant pUG RNA/siRNA cycles. Related competition models have been proposed to explain why mutations in the ERI/DICER complex enhance exo-RNAi^60–62, why endo-RNAi mutants show enhanced miRNA silencing⁶³, and why animals lacking the MET-2 methyltransferase show prolonged RNAi inheritance⁴⁰. Of note, it is also possible that a complex interplay between mRNA sequestration, competition between RNAi pathways and other PRG-1/piRNA-dependent phenomenon combine to explain how PRG-1/piRNAs coordinate mRNA pUGylation in C. elegans.

Recent studies show that ≅1000 mRNAs are repressed in prg-1 mutant germlines and that ≅100 of these are aberrantly targeted by siRNAs^15,16. The data suggest that the aberrant silencing of germline expressed genes that occurs in the absence of piRNAs may be relatively widespread. Why some germline mRNAs, but not others, become novel targets of aberrant pUGylation in the absence of piRNAs is not known. Whereas exogenous dsRNA is likely the trigger that induces perpetual silencing of mRNAs targeted by RNAi when PRG-1/piRNAs are absent, the molecular triggers initiating aberrant silencing of germline-expressed transcripts, such as the histone mRNAs, appear to be more complex. It was previously proposed that the lack of a poly(A) tail, which is characteristic of eukaryotic replication-dependent histone mRNAs, might predispose histone mRNAs to aberrant silencing in piRNA mutants¹⁵. However, we find that mRNAs that are thought to be polyadenylated, such as r03d7.2, can also be subjected to irreversible pUGylation and silencing (Figure 7D) indicating that the molecular signals triggering aberrant silencing in the absence of piRNAs remain to be identified. Interestingly, we find that 27 generations after experimental RNAi, some prg-1(−) lineages show perpetual silencing of gfp, while others do not (Figure 2B). These results hint that a molecular threshold may exist that must be reached for any given mRNA to enter stable pUG RNA/siRNA cycling. The molecular nature of this threshold is unknown. Possibilities include the amount of siRNA or pUG RNA maternally deposited into embryos, where only a specific amount of inherited silencing RNA is capable of initiating stable pUG RNA/siRNA cycling. According to this “threshold” model, some mRNAs, such as the histone mRNAs, may be more likely to produce the numbers of siRNAs or pUG RNAs that are needed to reach pUG RNA/siRNA cycling thresholds in the absence of piRNAs. While different mRNAs may differ in their propensity to enter states of perpetual silencing, because pUG RNA/siRNA cycling is self-perpetuating, once the “event horizon” threshold of gene silencing is reached, gene silencing is complete and irreversible in all cases. While different mRNAs may differ in their propensity to enter states of perpetual silencing, because pUG RNA/siRNA cycling is self-perpetuating, once the event horizon is reached, gene silencing is complete and irreversible in all cases. One question that arises from these findings is whether expressed mRNAs ever spontaneously cross the “event horizon” and enter pUG RNA/siRNA cycling in wild-type animals, thereby creating epialleles, which can be stably inherited over generations. Given the heritable nature of epigenetic states in C. elegans, it is enticing to also consider if such epialleles provide phenotypic variation that allows C. elegans populations to adapt rapidly to changing environments. In support of this idea, a recent study suggests that spontaneous siRNA-maintained epimutations arise at a rate 25x higher than DNA mutations in C. elegans populations⁶⁴.

The pUG RNA/siRNA gene silencing pathway is self-perpetuating and, therefore, potentially dangerous. If, for example, any germline mRNA was to mistakenly enter the pathway, it might never exit. Therefore, dedicated systems, such as PRG-1 and piRNAs, exist to limit and/or regulate pUG RNA/siRNA cycling. prg-1 mutants exhibit fertility defects, including reduced brood size, that are exacerbated by elevated temperatures^11–13. The aberrant silencing of essential genes that occurs in the absence of piRNAs has been linked to this sterility^24,25, and genetic perturbations that restore expression to aberrantly silenced genes in prg-1 mutants partially rescue the sterility defects of prg-1 mutants¹⁶. These results, combined with our data showing that aberrant silencing in prg-1 mutants is driven by pUG RNA/siRNA cycling, suggests that some of the germline defects associated with prg-1 are likely due to disorganization of RDE-3-dependent pUGylation, indicating that coordination of pUG RNA/siRNA cycling is an important mechanism by which piRNAs promote germ cell function.

STAR Methods

Resource availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Scott Kennedy (kennedy@genetics.med.harvard.edu).

Materials Availability

All unique/stable reagents generated in this study are available from the Lead Contact.

Data and Code Availability

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and subject details

C. elegans culture and genetics were performed as described previously⁶⁹. Strains were maintained at 20°C on Nematode Growth Medium (NGM) plates and fed OP50 E. coli bacteria, unless otherwise noted. Of note, animals expressing zu405, a gain-of-function, temperature-sensitive mutation in the gene oma-1⁴² were maintained at 15°C. Phenotypic analyses were performed on hermaphrodites. A list of strains used in this study can be found in Table S1.

Method details

RNA interference (RNAi) and RNAi inheritance assays

To perform RNAi experiments, embryos were isolated via hypochlorite treatment (egg prep) of gravid adult hermaphrodites and dropped onto RNAi plates (standard NGM plates with 1mM Isopropyl β-D-1-thiogalactopyranoside and 25ug/ml carbenicillin) seeded with HT115 E. coli bacteria expressing either L4440 (Addgene, #1654) empty vector control or L4440 carrying inserts to trigger the production of dsRNA (RNAi) against a gene of interest. The oma-1 RNAi clone came from the C. elegans RNAi collection (Ahringer lab). The gfp RNAi clone was obtained from the Fire lab. The znfx-1 RNAi clone was a custom clone made for this study.

To measure oma-1(zu405ts) silencing, embryos were isolated via hypochlorite treatment of gravid adult hermaphrodites and dropped onto oma-1 RNAi plates. Animals were grown at 20°C for 2–3 days. 6 larval stage 4 (L4) or adult animals were singled for each strain/genotype and allowed to lay embryos overnight at 20°C. Adults were removed from plates on the next day and the total number of embryos laid was counted. On the following day, the total number of embryos that hatched was counted. % embryos hatched was then calculated as the (# of hatched embryos / # embryos laid) × 100. For oma-1 RNAi inheritance experiments, the first generation of animals were treated with oma-1 RNAi and then each generation, some adults were scored for oma-1(zu405ts) silencing while the remaining population was egg prepped and embryos were dropped onto NGM plates seeded with OP50 E. coli bacteria. This process was repeated every generation for the indicated number of generations.

For gfp RNAi and RNAi inheritance assays, embryos were dropped onto RNAi plates seeded with bacteria expressing either the L4440 empty vector control or gfp dsRNA. Once animals were gravid adults, they were washed off of plates using M9 + Triton X-100 buffer. 50–150 animals were placed onto a microscope slide and gfp expression was scored using the Plan-Apochromat 20 × /0.8 M27 objective on an Axio Observer.Z1 fluorescent microscope (Zeiss). Images were taken with the Plan-Apochromat 63 × /1.4 Oil DIC M27 objective. All image processing was done using Fiji⁶⁸. For gfp RNAi inheritance experiments, each generation, 50–150 gravid adults were scored for gfp expression and the remaining population of gravid adults were egg prepped and embryos were dropped onto NGM plates seeded with OP50 E. coli bacteria. This process was repeated every generation for the indicated number of generations.

Paramutation crosses

Expressed oma-1(zu405ts) and gfp alleles were marked with tightly-linked dpy-20(e1282) and dpy-10(e128), respectively, mutations to allow for differentiation of expressed/naive alleles from silent oma-1(zu405ts) and gfp alleles in prg-1(gg531) and prg-1(gg540) mutants. More specifically, oma-1(zu405ts) and gfp animals were crossed with dpy-20(e1282) and dpy-10(e128) animals, respectively, and 200–300 F₂s were singled and genotyped to identify rare double mutants resulting from a crossover. Paramutation crosses for oma-1(zu405ts) were performed at 15°C using the prg-1(gg540) mutant. Four hermaphrodites were mated per cross (see Figure S4) and 5 F₁ heterozygotes were singled for each mated hermaphrodite. Four F₂s animals homozygous for the naive dpy-20(e1282) oma-1(zu405ts) allele were singled for each F₁ animal and allowed to lay a brood. Four pools of 5 F₃ animals for each F₂ were tested for oma-1(zu405ts) silencing based on the embryonic arrest assay described above. For the gfp paramutation test in Figure 4a, prg-1(gg531) mutants were outcrossed with gfp expressing animals to remove the oma-1(zu405ts) allele, allowing this experiment to be performed at 20°C. Note: because the silenced oma-1(zu405ts) and gfp alleles are paramutagenic in prg-1(gg531) mutants, the wild-type oma-1 allele in the new strain (YY1256) is also silent, as is the gfp allele. Two hermaphrodites were then mated with animals homozygous for the expressed dpy-10(e128) gfp allele. Lineages were established from homozygous dpy-10(e128) gfp F₂ animals, which were genotyped for prg-1(gg531) once they had laid a brood. Gravid adults were egg prepped every generation to maintain these lineages. gfp expression was scored in the F3 and F7 generations using the Plan-Apochromat 20 × /0.8 M27 objective on an Axio Observer.Z1 fluorescent microscope (Zeiss). At least 50 animals were counted for each strain.

CRISPR/Cas9-mediated genome editing

The CRISPR/Cas9 strategy described previously^45,70,71 was used to generate deletions of rde-3 and hrde-1, as well as to introduce the DDH → DAH mutation¹⁴ in prg-1. Guide RNA plasmids and repair template DNA were prepared as described previously⁴⁵. All guide RNAs were designed using the guide RNA selection tool CRISPOR⁷².

Gene expression quantification using qRT-PCR

Total RNA was extracted from gravid adult animals using TRIzol Reagent (Life Technologies, 15596018). 2ug of total RNA was reverse-transcribed to generate first-strand cDNA using the Superscript III First-Strand Synthesis System (Invitrogen, 18080051) and random hexamers. Note: total RNA was heated with dNTPs and random hexamers to 65°C for 5 mins and immediately chilled on ice before proceeding with remaining cDNA synthesis steps. First-strand cDNA was then treated with RNAse H at 37°C for 20 mins. cDNA was then 1:100 (for oma-1 and histone RNA quantification) or 1:8 (for all other quantifications) and 2ul was used to quantify gene expression using iTaq Universal SYBR Green Supermix (Bio-Rad, 1725120) according to manufacturer’s instructions. qRT-PCRs were performed using the CFX Connect machine (Bio-Rad) and semi-skirted PCR plates (Bio-Rad, 2239441). All qRT-PCR data was normalized to quantification of nos-3, a germline-expressed gene. qRT-PCR primers are listed in Table S2.

Taqman-based small RNA quantification

1ug of Trizol-extracted RNA from gravid adult animals (see above) was used for Taqman assays. Small RNAs were reverse transcribed into cDNA using the Taqman MicroRNA Reverse Transcription Kit (Applied Biosystems, 4366596). oma-1 and gfp small RNAs were then quantified by qRT-PCR using TaqMan Universal Master Mix II, no UNG (Applied Biosystems, 4440040) and custom TaqMan small RNA assays from Applied Biosystems (assay IDs: gfp: CSLJH0V, oma-1 probe 1: CSKAJ9W, oma-1 probe 2: CSLJIF4, oma-1 probe 3: CSMSGMC, oma-1 probe 4: CSN1ESK). qRT-PCRs were performed using the CFX Connect machine (Bio-Rad) and semi-skirted PCR plates (Bio-Rad, 2239441).

pUG RNA detection using pUG PCR.

pUG RNAs were detected using pUG PCR, a PCR-based assay described previously³⁷. Briefly, total RNA was extracted using TRIzol Reagent (Life Technologies, 15596018). 5ug of total RNA and 1pmol of reverse transcription (RT) oligo (Table S2) was used to generate first-strand cDNA using the Superscript III First-Strand Synthesis System (Invitrogen, 18080051). Total RNA, dNTPs and RT oligo were mixed and heated to 65 C for 5 mins and immediately chilled on ice before proceeding with remaining cDNA synthesis steps. 1ul of cDNA was used to perform a first round of PCR (20ul volume) with Taq DNA polymerase (New England BioLabs, M0273) for 20–25 cycles and primers listed in Table S2. These PCRs were then diluted 1:100 and 1ul was used to perform a second round of PCR (50ul volume) for 25–30 cycles using primers listed in Table S2. gsa-1, which has an 18nt long genomically UG repeat in its 3’UTR, served as a control for all pUG PCR analyses. PCR reactions were then run on 1.5–2% agarose gels. Images were acquired using a ChemiDoc MP Imaging System (Bio-Rad). All image processing was done using Fiji⁶⁸. All pUG PCR reactions were sequenced by cutting out lanes of interest from agarose gels and gel extracting the DNA using QIAquick Gel Extraction Kit (Qiagen, 28706). 3ul of gel extracted PCR product was used for TA cloning with the pGEM-T Easy Vector System (Promega, A1360) according to manufacturer’s instructions. Ligation reactions were incubated overnight at 4°C. Transformations were performed with 5-alpha Competent E. coli cells (NEB, C2987H) and plated on LB/ampicillin/IPTG/X-gal plates (according to pGEM-T Easy Vector System manufacturer’s instructions). White colonies were selected on the day next and inoculated in Luria Broth overnight. Liquid cultures were then miniprepped using QIAprep Spin Miniprep Kit (Qiagen, 27106) and plasmid DNA was Sanger sequenced using a universal SP6 primer (5’-CATACGATTTAGGTGACACTATAG-3; Dana-Farber/Harvard Cancer Center DNA Resource Core, Harvard Medical School) or a universal M13 primer (5’-TGTAAAACGACGGCCAGT-3’; Quintarabio, Cambridge, MA).

Quantification and Statistical Analysis

Throughout the results section, all data is represented as mean ± standard deviation, as calculated in Prism 8 (GraphPad; RRID:SCR_002798). Any shown p-values were generated by performing a two-tailed, unpaired Student’s t-test using Prism 8. Sample sizes (n) represent the number of animals analyzed or the number of biological replicates of each strain, as reported within the relevant figure caption.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Escherichia coli (E. coli) OP50	Caenorhabditis Genetics Center (CGC)	RRID:WB-STRAIN: OP50
E. coli HT115(DE3)	CGC	RRID:WB-STRAIN: HT115(DE3)
Ahringer RNAi Libraries in E. coli HT115(DE3)	67	RRID:SCR_017064
gfp RNAi clone	Andy Fire’s laboratory	N/A
TaqMan Small RNA Assays	Applied Biosystems	assay IDs: CSLJH0V, CSKAJ9W, CSLJIF4, CSMSGMC, CSN1ESK
Chemicals, peptides, and recombinant proteins
TRIzol Reagent	Invitrogen	15596018
Critical commercial assays
Superscript III First-Strand Synthesis System	Invitrogen	18080051
iTaq Universal SYBR Green Supermix	Bio-Rad	1725120
Taqman Universal Master Mix II, no UNG	Applied Biosystems	4440040
TaqMan MicroRNA Reverse Transcription kit	Applied Biosystems	4366596
pGEM-T Easy Vector System	Promega	A1360
Experimental models: Organisms/strains
See Table S1
Oligonucleotides
See Table S2
Software and algorithms
Fiji	68
ZEN Digital Imaging for Light Microscopy	Zeiss	RRID:SCR_013672
Prism 8	GraphPad	RRID:SCR_002798
Other
Axio Observer.Z1 fluorescent microscope	Zeiss	N/A
Plan-Apochromat 63 × /1.4 Oil DIC M27 objective	Zeiss	N/A
Plan-Apochromat 20 × /0.8 M27 objective	Zeiss	N/A
ORCA-Flash 4.0 CMOS camera	Hamamatsu	N/A

Highlights.

• Loss of piRNAs leads to perpetual silencing of mRNAs targeted by exogenous dsRNA

• Perpetual gene silencing is driven by poly(UG)-tailed (pUG) RNAs and small RNAs

• pUG RNAs also drive aberrant silencing of endogenous mRNAs in piRNA mutants

• The piRNA pathway insulates mRNAs from perpetual and aberrant silencing