Cues from mRNA splicing prevent default Argonaute silencing in C. elegans

SUMMARY

In animals, Argonaute small-RNA pathways scan germline transcripts to silence self-replicating genetic elements. However, little is known about how endogenous gene expression is recognized and licensed. Here, we show that the presence of introns and, by inference, the process of mRNA splicing prevents default Argonaute-mediated silencing in the C. elegans germline. The silencing of intronless genes is initiated independently of the piRNA pathway but nevertheless engages multiple components of the downstream amplification and maintenance mechanisms that mediate transgenerational silencing, including both nuclear and cytoplasmic members of the worm-specific Argonaute gene family (WAGOs). Small RNAs amplified from intronless mRNAs can trans-silence cognate intron-containing genes. Interestingly, a second, small RNA-independent cis-acting mode of silencing also acts on intronless mRNAs. Our findings suggest that cues put in place during mRNA splicing license germline gene expression and provide evidence for a splicing-dependent and dsRNA- and piRNA-independent mechanism that can program Argonaute silencing.

In brief

Organisms distinguish between foreign and endogenous genes and establish heritable patterns of gene expression. Makeyeva et al. now show that the presence of introns and, by inference, mRNA splicing protects C. elegans genes against multiple silencing mechanisms, including a small RNA-mediated pathway triggered by default, independently of known primary Argonautes.

Graphical Abstract

INTRODUCTION

The importance of guided search in biology is now widely appreciated thanks to the discoveries of RNA interference (RNAi) and CRISPR-Cas systems (Barrangou et al., 2007; Fire et al., 1998; Jansen et al., 2002). In both systems, Argonautes and Cas proteins become programmed with 20- to 30-nucleotide (nt) guide RNAs and use the sequence information in their guides to find and regulate genetic targets (Brouns et al., 2008; Elbashir et al., 2001; Hammond et al., 2001; Marraffini and Sontheimer, 2008). One obvious ancestral function of these guided-search mechanisms is to defend against pathogenic nucleic acids, such as viruses and other mobile genetic elements, and several insights have been made into how pathogenic activity leads to programming of the respective search mechanisms (see reviews by Ding, 2010; Hille et al., 2018; Malone and Hannon, 2009). Antiviral Argonautes, for example, are programmed by accessory factors that recognize and process virus-derived, cytoplasmic double-stranded (ds)RNA into short interfering (si)RNA guides (Bernstein et al., 2001; reviewed in Wilson and Doudna 2013). Interestingly, dsRNA-initiated programming of Argonautes has been co-opted to regulate endogenous gene expression by processing genomically encoded stem-loop RNAs into guides called micro-RNAs (miRNAs) (reviewed in Carthew and Sontheimer, 2009; Lee et al., 1993).

In animal germlines, transposon surveillance is mediated by the Piwi Argonaute system, which also silences transgenes and a number of other cellular genes (see review by Ozata et al., 2019). However, PIWI-interacting (pi)RNA guides are not processed from dsRNA, but from single-stranded precursors transcribed by RNA polymerase II (Aravin et al., 2006; Brennecke et al., 2007; Girard et al., 2006; Grivna et al., 2006; Gu et al., 2012; Lau et al., 2006; Ruby et al., 2006). Several components of the machinery that processes piRNA precursors are conserved, but the organization of precursor genes differs widely among organisms and the precise cues that cause piRNA precursors to be recognized by the Piwi Argonaute programming machinery are not understood.

The C. elegans germline is an excellent model for the study of mechanisms of Argonaute programming. In addition to the canonical RDE-1 Argonaute system, which detects foreign dsRNA (Steiner et al., 2009; Tabara et al., 1999; Yigit et al., 2006), the C. elegans germline expresses multiple Argonaute systems that collectively engage nearly all germline mRNAs (Batista et al., 2008; Claycomb et al., 2009; Conine et al., 2010, 2013; Das et al., 2008; Gent et al., 2010; Gu et al., 2009; Han et al., 2009; Vasale et al., 2010). The C. elegans Piwi, PRG-1, engages tens of thousands of 21-nt 5′ U piRNA species (also called 21U-RNAs) (Batista et al., 2008; Gu et al., 2012; Ruby et al., 2006). PRG-1 can initiate heritable silencing on its targets by recruiting RNA-dependent RNA polymerase (RdRP) which amplifies 22-nt antisense RNAs that most frequently initiate with a 5′ G residue (22G-RNAs) and are loaded onto members of an expanded group of worm Argonautes, or WAGOs (worm-specific Argonaute gene family) (Ashe et al., 2012; Bagijn et al., 2012; Das et al., 2008; Gu et al., 2009; Luteijn et al., 2012; Shirayama et al., 2012). A second RdRP system programs the Argonaute CSR-1 with 22G-RNAs that target germline-expressed mRNAs and are thought to protect them from piRNA-dependent induction of WAGO silencing (Seth et al., 2013; Wedeles et al., 2013). The WAGO and CSR-1 Argonautes and their respective repertoires of 22G-RNAs are inherited transgenerationally via sperm and egg (Conine et al., 2013; Phillips et al., 2015; Seth et al., 2013; Shirayama et al., 2012).

In many organisms, mRNA splicing communicates with downstream events in mRNA expression, including mRNA 3′-end formation (Chiou et al., 1991; Cooke et al., 1999; Nesic et al., 1993; Niwa et al., 1990), nuclear export (Luo et al., 2001; Valencia et al., 2008), and mRNA translation on the ribosome (Braddock et al., 1994; Lu and Cullen, 2003; Matsumoto et al., 1998; Nott et al., 2004, 2003). Genetic studies in C. elegans have identified conserved components of the splicing machinery required for Argonaute-mediated silencing in C. elegans (Akay et al., 2017; Jiao et al., 2019; Newman et al., 2018). However, whether these factors directly or indirectly participate in silencing remains elusive.

Here we explore the relationship between splicing and Argonaute surveillance in the germline of the nematode C. elegans. We show that genes lacking introns, including endogenous genes from which introns were removed by precision genome editing, become default targets for Argonaute-mediated silencing and that the resulting amplified small RNAs can act in trans to silence cognate intron-containing genes. Moreover, intron-containing regions of an endogenous gene can also act in trans to protect homologous regions of an intronless transgene from small RNA targeting. We show that the small RNA-mediated arm of intronless silencing depends on the WAGO pathway but is not initiated by piRNAs.

Interestingly, intronless genes failed to express in the germline even when Argonaute small RNA pathways were disarmed by mutation, indicating that a small RNA-independent, cis-acting pathway acts in parallel to silence intronless genes in the germline. Together, our findings support a model in which RNA splicing, and/or other splicing-dependent RNA-processing mechanisms, impart signals on nascent transcripts that prevent their default recognition as templates for Argonaute guide-RNA programming.

RESULTS

Intron removal prevents gene expression in the germline but not the soma

To investigate how mRNA splicing affects Argonaute small RNA surveillance in C. elegans, we first compared the expression of intron-containing cdk-1::gfp and intronless cdk-1*::gfp* transgenes (Figure 1A). Aside from the presence or absence of introns, the transgenes have the same regulatory sequences and were inserted into identical locations on chromosome II using Mos1-mediated single-copy insertion (MosSCI) (Frøkjaer-Jensen et al., 2008). The intron-containing version of cdk-1::gfp was robustly expressed in both the soma and germline (Figures 1B and 1C) (Shirayama et al., 2012). By contrast, the intronless cdk-1*::gfp* transgene was weakly expressed in the soma (Figure 1B) and completely silenced in the germline of three independent strains for over 10 generations (Figure 1C), suggesting that introns are essential for germline expression of cdk-1::gfp.

Figure 1. Intron removal leads to germline gene silencing that acts in trans.

(A) Schematic structures of intron-containing cdk-1::gfp and intronless cdk-1*::gfp* transgenes. (Gray boxes, cdk-1 coding sequence; green boxes, gfp coding sequence; V-shaped lines, introns).

(B and C) Fluorescence (upper) and DIC micrographs (lower) of (B) L3/L4-stage vulva in representative cdk-1::gfp, cdk-1*::gfp*, and wild-type (WT) larvae, and (C) adult oocytes in the indicated strains. Scale bars, 20 μm.

(D and E) Genetic crosses and representative epifluorescence images testing the ability of intronless cdk-1*::gfp* to trans-silence (D) cdk-1::gfp or (E) oma-1::gfp. Percent GFP+ (ON) or GFP− (OFF) worms and the number of worms analyzed in each generation is indicated. In (D), control (left) and test (right) crosses are shown. WT indicates nontransgenic worms. Images show presence (ON) or absence (OFF) of CDK-1::GFP signal in oocyte nuclei. In (E), oma-1::gfp is gradually trans-silenced by the intronless cdk-1*::gfp* transgene. Images show presence or absence of OMA-1::GFP signal in cytoplasm of oocytes. White dashed circles indicate location of nuclei in GFP–oocytes.

Intronless gene silencing involves small RNA-dependent and -independent pathways

Transgenes that are heritably silenced by small RNAs can trans-silence homologous transgenes (Shirayama et al., 2012). When crossed with nontransgenic wild-type worms, cdk-1::gfp transgenic animals produced offspring that exhibited bright and easily detected GFP signals (Figure 1D). By contrast, when crossed with worms bearing the cdk-1*::gfp* transgene, cdk-1::gfp transgenic animals produced offspring that were completely silenced (Figure 1D). Moreover, the cdk-1::gfp transgene remained silent in subsequent generations, even after segregating away the intronless allele (Figure 1D). A previous study showed that the oma-1::gfp transgene is resistant to transitive silencing (Seth et al., 2013). However, although the oma-1::gfp transgene was active in F1 progeny after crossing with worms bearing the cdk-1*::gfp* transgene, OMA-1::GFP expression gradually declined and ultimately became fully silenced over the course of five generations when propagated in the presence of the intronless cdk-1*::gfp* transgene (Figure 1E).

Trans-silencing of homologous transgenes by the intronless cdk-1*::gfp* transgene suggests that silencing is maintained by the RDE-3 and WAGO Argonaute-dependent small RNA effector pathway (Ashe et al., 2012; Bagijn et al., 2012; Shirayama et al., 2012). Small RNA-sequencing analyses suggested this to be the case. Compared with the active intron-containing cdk-1::gfp transgene, which produced few antisense small RNAs (Figure 2A) (Shirayama et al., 2012), the cdk-1*::gfp* transgene was robustly targeted by small RNAs that map antisense to the gfp* sequences of the transgene (Figure 2B).

Figure 2. Intronless cdk-1*::gfp* is silenced by a default WAGO-dependent small RNA pathway and by a cis-acting small RNA-independent pathway.

(A and B) Plots showing the density of antisense small RNAs mapping along the coding regions of (A) cdk-1::gfp or (B) cdk-1*::gfp*. Positions of exon junctions in cdk-1::gfp (and corresponding positions in intronless cdk-1*::gfp*) are indicated by broken vertical lines in gene cartoons. The height of each bar represents the number of reads that begin at that position per million total reads.

(C and D) Genetic analyses and representative epifluorescence images testing the role of RDE-3 in intronless cdk-1*::gfp* silencing. Percentage of worms with GFP+ (ON) or GFP− (OFF) germ cells and the number of worms analyzed in each generation is indicated. In (C), intronless cdk-1*::gfp* remained silenced when crossed with an (C) rde-3(ne3370) null mutant. Crosses reveal that cdk-1*::gfp* does not trans-silence cdk-1::gfp in the rde-3 mutant. In (D), CRISPR was used to delete rde-3 (ne4865) in worms in which oma-1::gfp is trans-silenced by cdk-1*::gfp*. OMA-1::GFP signal was restored after homozygosing rde-3(ne4865). CRISPR gene-editing events were validated by genomic PCR and Sanger sequencing.

(E) Density of antisense small RNAs mapping along cdk-1*::gfp* in the rde-3(ne3370) mutant.

(F) Intronless cdk-1*::gfp* was silenced when introduced into a prg-1(tm872) null mutant. Crosses reveal that cdk-1*::gfp* trans-silences cdk-1::gfp in the prg-1 mutant.

(G) Density of antisense small RNAs mapping along cdk-1*::gfp* in the prg-1(tm872) mutant.

RDE-3 is required for the production of WAGO-dependent small RNAs and, thus, for heritable silencing by the WAGO effector pathway (Chen et al., 2005; Shirayama et al., 2012). To ask if silencing of cdk-1*::gfp* requires RDE-3 activity, we crossed the intronless transgene into an rde-3(ne3370) null mutant. We found that cdk-1*::gfp* remained silent—even over multiple generations—in the homozygous rde-3(ne3370) null mutant background (n = 50) (Figure 2C). To ask if transitive silencing induced by cdk-1*::gfp* requires RDE-3, we crossed rde-3(ne3370) bearing the intronless transgene into a second homozygous rde-3(ne3370) strain expressing the intron-containing cdk-1::gfp transgene. Consistent with a requirement for RDE-3 activity, trans-silencing failed to occur in the rde-3(ne3370)-mutant background, even after propagating the transgenes together for several generations (Figure 2C). To ask if RDE-3 is required to maintain the silencing of a transitively silenced intron-containing oma-1::gfp transgene, we used precision genome editing to delete rde-3 in a >5^th generation silenced strain (Figure 2D). We found that 100% of the F2 rde-3 homozygous progeny generated in three independent CRISPR lines exhibited bright GFP expression in proximal oocytes indicative of reactivation of the transitively silenced gene (Figure 2D). CRISPR editing in this (Figure 2D) and all subsequent experiments was confirmed by genomic PCR and Sanger sequencing. As expected, 22G-RNAs antisense to gfp* failed to accumulate in the rde-3 mutants (Figure 2E). Thus, although intronless cdk-1*::gfp* is targeted by an RDE-3-dependent small RNA pathway, it also appears to be silenced by a cis-acting mechanism, independent of small RNAs.

Intronless gene silencing does not require the piRNA pathway

The silencing of foreign sequences by the RDE-3/WAGO-dependent small RNA pathway is initiated by the PIWI Argonaute PRG-1 and piRNAs (Ashe et al., 2012; Bagijn et al., 2012; Lee et al., 2012; Shirayama et al., 2012). We therefore tested if PRG-1 initiates silencing of the intronless cdk-1*::gfp*. Using MosSCI, we inserted the cdk-1*::gfp* transgene directly into a prg-1(tm872) null mutant (at the same chromosomal site used above) and obtained four independent insertion lines (Frøkjaer-Jensen et al., 2008). None of these expressed the intronless cdk-1*::gfp* transgene in the germline (Figure 2F), indicating that PRG-1 is not required for silencing of the intronless cdk-1*::gfp* transgene. Moreover, in the complete absence of PRG-1, we found that the intronless cdk-1*::gfp* transgene was still able to trans-silence an intron-containing cdk-1::gfp transgene (100%, n = 20) (Figure 2F). As in wild type, trans-silencing in the prg-1 mutant background was maintained for multiple generations after segregation away from the intronless allele (Figure 2F). These results indicate that PRG-1 is not required to initiate small RNA-dependent silencing induced by the intronless cdk-1*::gfp* transgene. Indeed, small RNAs antisense to the gfp* region of the intronless cdk-1*::gfp* transgene accumulated to higher levels in prg-1 mutants than in wild-type worms (Figure 2G), suggesting that the piRNA pathway competes with the intronless silencing pathway for RDE-3 or other small-RNA silencing components. Thus, piRNAs do not initiate small RNA-mediated silencing of the intronless cdk-1*::gfp* transgene, nor are piRNAs required for cis-silencing of the intronless transgene.

Endogenous genes are sensitive to intronless silencing

The finding that 22G-RNAs are limited to the gfp sequences implies a difference of some kind between endogenous and foreign sequences in the transgene. To test if endogenous sequences can elicit intronless silencing, we used CRISPR to first delete the majority of the coding region—including the introns—of the nonessential oma-1 gene and then to insert either the intron-containing oma-1 gene or the intronless oma-1* cDNA (Figure 3A). To assess the function of the edited oma-1 alleles, we used RNAi to knock down oma-2. The oma-1 and oma-2 genes are redundantly required for oocyte maturation (Detwiler et al., 2001; Shimada et al., 2002). In worms with the restored, intron-containing oma-1 gene, we found that oma-2(RNAi) resulted in viable and fertile progeny (n = 41), indicating that the oma-1 gene, when restored with introns, is functional. By contrast, oma-2(RNAi) resulted in sterile progeny (n = 39), in all three strains edited with intronless oma-1*, indicating that the intronless allele is silenced. qRT-PCR analysis supported oma-1* silencing and showed that the mRNA produced by the intron-containing allele was about 90 times more abundant compared with the mRNA produced by the oma-1* allele (Figure S1A). Small RNA sequencing revealed a marked accumulation of 22G-RNAs targeting oma-1* (Figures 3B and 3C). The majority of 22G-RNAs mapped to the 3′-untranslated region immediately after the stop codon with additional peaks in the 5′ half of the coding region (Figure 3C). As was true for the cdk-1*::gfp* transgene, the 22G-RNAs targeting oma-1* were absent in rde-3 mutant worms and were increased in prg-1 mutant worms (Figures 3D and 3E). Thus, an intronless oma-1* allele at the endogenous oma-1 locus triggers an RDE-3-dependent response that is independent of piRNAs.

Figure 3. Endogenous genes are sensitive to intronless silencing.

(A) Schematic outlining the removal of endogenous oma-1 coding region and insertion of intron-containing oma-1 or intronless oma-1* alleles. CRISPR was used to modify the endogenous oma-1 locus. An in-frame 3xflag sequence was first inserted at the 3′ end of the oma-1 gene. The oma-1 gene from the middle of exon 1 to the end of exon 6 was replaced with a short stuffer containing a new CRISPR guide site (marked with a black dashed rectangle) to allow subsequent insertion of the wild-type oma-1 sequence with or without (oma-1*) introns. Silencing of intronless oma-1* was confirmed by oma-2(RNAi). To test whether intronless oma-1* is also silenced by the cis-silencing pathway, CRISPR was used to delete rde-3 and then oma-2. The oma-2 genotype of F2 progeny (n = 48) was determined by PCR. The percentage of each genotype is indicated, and the percentage of fertile or sterile worms is indicated. Every CRISPR gene editing event was validated by genomic PCR and Sanger sequencing. (B and C) Plots showing the density of antisense small RNAs mapping along the coding regions of (B) oma-1::3xflag or (C) oma-1*::3xflag. Positions of exon junctions in oma-1::3xflag (and corresponding locations in intronless oma-1*) are indicated by broken vertical lines in gene cartoon. The height of each bar represents the number of reads that begin at that position per million total reads.

(D and E) Plots showing the density of antisense small RNAs mapping along oma-1*::3xflag in (D) rde-3(ne4871) or (E) prg-1(ne4844) mutants.

To determine if oma-1* is silenced by a small RNA-independent mechanism, we used CRISPR HDR to precisely delete the oma-2 gene in the rde-3 mutant background. In this experiment, RNAi knockdown of oma-2 could not be used because rde-3 mutants are RNAi deficient. As a control, we also deleted oma-2 in a strain where oma-1 was restored with introns (Figure 3A). For each genetic background, three independent oma-2 deletions were generated and were confirmed by genomic PCR and Sanger sequencing. In strains where oma-1 was restored with introns, F2 animals homozygous for the oma-2 deletion were 100% fertile (Figure 3A). By contrast, in strains restored with intronless oma-1, 100% of the F2 worms homozygous for the oma-2 deletions were sterile (Figure 3A). Thus, the intronless oma-1* allele generates RDE-3-dependent small RNAs but is also silenced by an additional rde-3-independent mechanism. In agreement with this conclusion, although qRT-PCR analysis showed that oma-1* mRNA levels were slightly increased in rde-3 mutants, they were nevertheless more than 50-fold lower than the levels of the mRNA produced by the intron-containing oma-1 gene (Figure S1B).

Regulatory elements of a naturally intronless gene can partially suppress intronless silencing

Replication-dependent histones comprise a group of intronless genes that are robustly expressed in dividing somatic and germline tissues (Keall et al., 2007; Pettitt et al., 2002; Robbins and Borun, 1967). To explore the sensitivity of a histone gene to intronless silencing, we used CRISPR to insert gfp coding sequences—with or without introns—immediately after the start codon in the endogenous his-61 locus (Figure 4A). Both types of the tagged histone gene were robustly expressed in somatic tissues (including embryos) (Figure 4A). Surprisingly, however, intronless gfp*::his-61—but not the intron-containing gfp::his-61—was completely silenced throughout the germline, including in oocytes, and in mitotic and pachytene nuclei of the adult germline (Figure 4A). Congruently, qRT-PCR analysis of dissected gonads showed that gfp*::his-61 mRNA level was about 30 times lower compared with the cognate intron-containing allele (Figure S1C). Small RNA sequencing revealed an abundant accumulation of 22G-RNAs targeting the gfp sequences of gfp*::his-61 (Figure 4B, upper right panel). Moreover, as with other intronless transgenes, the accumulation of gfp* 22G-RNAs depended on RDE-3 but not PRG-1 (Figure 4B).

Figure 4. Regulatory elements of a naturally intronless gene can partially suppress intronless silencing.

(A) Schematic gene structures of intron-containing gfp::his-61 (left) and intronless gfp*::his-61 (right), and representative epifluorescence images analyzing GFP expression in embryos (soma), oocytes, or distal tips of dissected gonads (outlined with white dashes in gfp*::his-61). CRISPR was used to delete rde-3(ne4848) to test the role of RDE-3 in silencing intronless gfp*::his-61. In the resulting rde-3 mutant, epifluorescence (upper) and DIC (lower) imaging of a representative gonadal arm shows GFP signal in the distal germline. GFP signal gradually decreases in the pachytene region and is absent in oocytes. Percent GFP+ (ON) or GFP− (OFF) worms and the number of worms analyzed is indicated. Every CRISPR gene-editing event was validated by somatic GFP expression (in cases of gfp integration), genomic PCR and Sanger sequencing.

(B) Density of antisense small RNAs mapping along gfp::his-61 (top left) or gfp*::his-61 (remaining plots). Small RNAs targeting intronless gfp*::his-61 were examined in wild-type (top right), rde-3(ne4848) (bottom left), and prg-1(ne4766) (bottom right).

(C) Diagram illustrating replacement of his-61 with an in-frame SV40 nuclear localization signal (nls), and deletion of rde-3(ne4850). Representative epifluorescence images of gfp*::nls in rde-3(ne4850) or wild-type oocytes (white dashed circles indicate GFP–nuclei). After deleting rde-3(ne4850), GFP signal was visible in the oocytes and in the entire germline (fluorescence and DIC micrographs at the bottom). Percent GFP+ (ON) or GFP− (OFF) worms and the number of worms analyzed is indicated. Every CRISPR gene editing event was validated by genomic PCR and Sanger sequencing.

Interestingly, whereas the intronless cdk-1*::gfp* and oma-1* genes remained completely silenced in rde-3-mutant germlines, we observed that gfp*::his-61 was expressed in the distal mitotic region in rde-3 germlines (Figure 4A, bottom panel). This result suggests that intronless gfp*::his-61 is less sensitive to or can partially bypass the cis-silencing pathway.

Appending the gfp open reading frame (ORF) approximately triples the open reading frame of the his-61 gene. We wondered if this added length might overwhelm the ability of his-61 regulatory sequences to prevent intronless silencing. To explore this possibility, we used CRISPR to replace the his-61 open reading frame with the 28-amino-acid SV40 nuclear localization signal (NLS), shortening the gene by 25% (Figure 4C). This gfp*::nls gene, driven at the endogenous locus by the his-61 promoter and UTR, remained silent throughout the germline (Figure 4C). However, upon the inactivation of rde-3, GFP was visible throughout the germline, including within oocyte nuclei (Figure 4C, middle and bottom panels). Thus, shortening the intronless gene to approximately twice the length of the endogenous histone gene was sufficient to prevent cis-acting silencing throughout the germline but did not prevent small RNA-dependent silencing of gfp*::nls.

Intronless silencing requires WAGO Argonautes but is initiated independently of known primary Argonautes

Desilencing of gfp*::his-61 in the distal germline of rde-3 mutants suggested its use as an assay to quickly test whether other known small-RNA pathway factors promote intronless silencing. To perform this assay, we used CRISPR editing to either mutate known silencing factors in the gfp*::his-61 strain or introduce gfp* into the his-61 locus in pre-existing homozygous RNA-silencing mutant strains. The de novo introduction of gfp* was not only much faster than crossing, especially when strains bearing multiple mutants were required, but also had the advantage of enabling the assay (in principle) to identify factors required for both the initiation and maintenance of silencing. We validated this assay by inserting intronless gfp* into the his-61 locus of an rde-3(ne4852) mutant and confirmed that GFP was visible in the distal gonads of all four independently generated strains (Figure 5B). RDE-3 is known to promote the formation of RNA templates used by the partially redundant cellular RdRPs RRF-1 and EGO-1 to amplify silencing signals (Aoki et al., 2007; Chen et al., 2005; Gu et al., 2009; Shukla et al., 2020); rrf-1 mutants are viable and fertile, but EGO-1 is essential for fertility and embryo viability. We therefore depleted EGO-1 activity using an auxin-inducible degradation system (ego-1::degron; see STAR Methods) (Zhang et al., 2015). Whereas gfp*::his-61 was silenced in rrf-1 and ego-1::degron single mutants, it was expressed in the distal germline of the ego-1::degron, rrf-1 double mutant in the presence of auxin (Figure 5B). Mutations in the remaining cellular RdRPs, RRF-2 (of unknown function) and RRF-3 (required for 26G RNA production), did not affect gfp*::his-61 expression (Figure 5B). Consistent with the idea that RRF-3 and 26G RNAs do not promote intronless silencing, we did not detect 26-nt small RNAs targeting the intronless reporter in our sequencing data (Figure S2).

Figure 5. Small RNA-mediated intronless silencing is initiated independently of known primary Argonautes.

(A) Assay to test if known small RNA factors are required for intronless silencing. Intronless gfp*::his-61 fusion was generated via CRISPR in small RNA pathway mutants (e.g., rde-3) and animals were examined for germline GFP signal (especially in the distal germline).

(B) List of RdRP, Argonaute, and RNAi co-factor mutants tested in this analysis and percent GFP+ worms in each mutant. Numbers in parentheses indicate the total number of worms examined from the number of independent lines. Every CRISPR gene-editing event for gfp* integration at his-61 locus was validated with phenotypic observation of somatic GFP expression, genomic PCR, and Sanger sequencing analyses. Mutants denoted with a double asterisk were generated by CRISPR in the gfp*::his-61 background, followed by genomic PCR and Sanger sequencing analyses.

The transitive nature of intronless silencing and the involvement of both RDE-3 and RdRPs suggests the involvement of downstream WAGO Argonautes (Gu et al., 2009; Yigit et al., 2006). To test WAGOs directly, we introduced gfp* into the his-61 locus in WAGO single- and multiple-mutant strains. The gfp*::his-61 allele was silenced in wago-1 and wago-9/hrde-1 single mutants, but was de-silenced in the distal zone in the double mutant and in a previously constructed strain bearing mutations in all 12 wago genes (Gu et al., 2009; Yigit et al., 2006) (Figure 5B). These findings suggest that multiple WAGOs, including a predominantly nuclear family member (WAGO-9/HRDE-1) and a predominantly cytoplasmic member (WAGO-1), contribute to intronless silencing (Figure 5B) (Ashe et al., 2012; Buckley et al., 2012; Gu et al., 2009; Shirayama et al., 2012).

Consistent with these findings, small RNA sequencing studies on wago-9/hrde-1 and its nuclear RNAi co-factor nrde-4 (Burkhart et al., 2011; Guang et al., 2010, 2008) revealed that many 22G-RNAs targeting gfp were missing in both mutants (for unknown reasons, these depleted 22G-RNAs mapped primarily to the second half of gfp (Figure S3). Together these findings indicate that intronless silencing engages both the nuclear and cytoplasmic arms of the WAGO pathway.

WAGO-dependent silencing can be initiated by and/or function together with several “primary” Argonautes, including PRG-1, RDE-1, ERGO-1, and redundant ALG-3/ALG-4 (Bagijn et al., 2012; Conine et al., 2010; Corrêa et al., 2010; Das et al., 2008; Vasale et al., 2010; Yigit et al., 2006). Because PRG-1 is not required for intronless gene silencing, we were particularly interested to test whether silencing depends on any of the other known primary Argonautes. In addition, we tested the prg-1 homolog prg-2 which is predicted to be a pseudogene. We found that gfp*::his-61 was silenced when introduced by CRISPR in all of these Argonaute mutant strains (Figure 5B). For each edited strain, we confirmed proper gfp* insertion by genomic PCR and sequencing. Thus, none of the primary Argonautes that are known to engage WAGO-dependent silencing are required for intronless gene silencinsg.

Trans-acting signals from endogenous genes with intact splicing protect against small RNA silencing

In our oma-1* replacement experiments above, we were intrigued by the RDE-3-dependent accumulation of 22G-RNAs targeting the oma-1 coding sequence (Figure 3C). Typically, WAGO-dependent 22G-RNAs only accumulate antisense to foreign sequences of a silenced transgene (Seth et al., 2013, 2018; Shirayama et al., 2012). For example, when we directly inserted an intronless oma-1*::gfp* transgene into chromosome II using MosSCI (Frøkjaer-Jensen et al., 2008), 22G-RNA accumulation was limited to the gfp portion of the transgene (Figure 6A). In this and all of the previously described MosSCI experiments, the endogenous intron-containing locus was present. We therefore wondered if trans-acting signals emanating from the existing endogenous intron-containing locus prevent the accumulation of 22G-RNAs targeting the cognate sequences in the silenced transgene. To test this possibility, we deleted endogenous oma-1 by CRISPR and then inserted the intronless oma-1*::gfp* into chromosome II using MosSCI (Frøkjaer-Jensen et al., 2008) (Figure 6B). Strikingly, in the absence of the endogenous locus, antisense 22G-RNAs targeting both the oma-1 and gfp portions of the intronless transgene accumulated to equally high levels (Figure 6B). Thus, the endogenous, spliced allele of oma-1 communicates with cognate sequences from oma-1 transgenes, preventing them from templating small RNA-silencing signals.

Figure 6. Trans-acting signals from endogenous loci protect against small RNA-mediated silencing.

(A and B) Plots showing the density of antisense small RNAs mapping along intronless oma-1*::gfp* transgene in the (A) wild-type or (B) oma-1 deletion worms. The schematic in (B) outlines the deletion of endogenous oma-1 (including promoter and 30 UTR) on chromosome IV (LGIV), single-copy insertion of intronless oma-1*::gfp* on LGII, and small RNA sequencing. oma-1 deletion at the endogenous locus was validated by genomic PCR and Sanger sequencing analyses. (C) Model figure. Introns/splicing protect transcripts from default WAGO-dependent and cis-silencing. (Top) Factors associated with splicing (orange ovals) counteract silencing cues (red octagons) deposited on pre-mRNA by default. (Bottom) Silencing cues on unspliced transcript recruit RdRP, which makes small RNAs that guide Argonaute-mediated trans-silencing of cognate intron-containing genes (dashed arrow). In addition, possibly in response to the same cues, unspliced transcripts are silenced in cis, for example, by disrupting mRNA processing or promoting export to nuage where RNAs are sequestered and used for small-RNA templating.

DISCUSSION

Genetic studies in plants, animals, and fungi have identified mRNA-splicing components as factors required for RNA silencing by Argonautes (Akay et al., 2017; Bayne et al., 2008; Czech et al., 2013; Handler et al., 2013; Herr et al., 2006). Moreover, phylogenetic studies suggest that genes encoding spliceosome and RNA-silencing components tend to be retained together in eukaryotic genomes (Tabach et al., 2013), consistent with a functional relationship. The essential role of mRNA splicing in gene expression, however, has made it difficult to determine whether splicing factors directly participate in Argonaute-mediated silencing or indirectly promote surveillance by regulating the expression of the silencing machinery—e.g., guide RNAs, Argonautes, or other essential co-factors (Goriaux et al., 2014; Kallgren et al., 2014; Zhang et al., 2014). Removing the introns from reporter genes, rather than impairing splicing by genetic perturbations, has allowed us to circumvent many of these issues and to uncover surprising and robust interactions between intronless mRNAs and Argonaute surveillance pathways.

We have shown that the presence of functional introns— and thus mRNA splicing—prevents the default targeting of C. elegans germline transcripts by a piRNA-independent Argonaute pathway. Our findings imply that nascent transcripts are marked in ways that are interpreted by the Argonaute surveillance machinery. For example, positive factors that actively license expression might be deposited on intron-containing transcripts. Alternatively, negative signals that would otherwise trigger default recognition by the silencing machinery might be actively removed in response to splicing. Of course, some combination of positive and negative cues could also be involved.

Because nearly all germline transcripts template the production of 22G-RNAs in C. elegans (Claycomb et al., 2009; Gu et al., 2009; Ruby et al., 2006), RdRPs might be recruited by default to all mRNAs, and cues from splicing might help determine which Argonaute system is loaded: either the CSR-1 Argonaute, which appears to protect expression (Claycomb et al., 2009; Conine et al., 2013; Seth et al., 2013; Wedeles et al., 2013), or the WAGO Argonautes that promote silencing (Bagijn et al., 2012; Gu et al., 2009; Shirayama et al., 2012; Vasale et al., 2010). Loading of the CSR-1 Argonaute with 22G-RNAs templated from spliced mRNAs could account for trans-acting signals that promote expression of cognate genes, while the loading of WAGO Argonautes with 22G-RNAs templated from intronless RNAs could promote silencing. Surveillance by the piRNA system could fit into this picture by functioning to overcome either CSR-1 protection or other positive signals put in place by splicing (Figure 6C). Indeed, a recent study showed that a piRNA sensor with three introns accumulated fewer WAGO 22G-RNAs than did an otherwise identical sensor with a single intron, strongly suggesting that genes with more introns are less prone to piRNA silencing (Akay et al., 2017). In addition, the type, number, and location of introns have been shown to influence gene expression in the C. elegans germline (Aljohani et al., 2020). A previous study suggested that mRNAs with weak splicing signals are retained on the spliceosome and become targets of small RNA silencing (Dumesic et al., 2013; Newman et al., 2018). Perhaps the removal of all introns from a gene causes it to express a nascent transcript where the only remaining splicing signals would, by definition, be nonoptimal. In the absence of strong signals, therefore, weak splicing signals might abound and invariably result in default retention of the RNA in unresolved spliceosomes.

Components of the splicing machinery might also play roles in Argonaute surveillance outside of their functions in mRNA splicing. For example, the exon-junction complex (EJC), which is deposited on exons during splicing, promotes mRNA nuclear export (Luo et al., 2001) and signals the ribosome to promote nonsense-mediated decay of mRNAs with premature stop codons (Kim et al., 2001; Le Hir et al., 2001). Similarly, a recent study has implicated components of the U1 snRNP in a process that prevents the premature cleavage and polyadenylation of transcripts within introns (Kaida et al., 2010). If signals from the abortive recruitment of splicing machinery are responsible for the default silencing of intronless mRNAs observed here, then our findings imply that the presence of functional introns (or their successful removal by the splicing machinery) can counteract these signals. Perhaps the successful resolution of splicing recruits an RNA helicase that then translocates along the mRNA to ensure the removal of stalled spliceosomes or other co-transcriptionally deposited signals (orange oval in Figure 6C), liberating the mRNA from influences that would otherwise prevent expression or trigger Argonaute-mediated silencing.

We used RNAi to test whether components of the splicing machinery—including core factors required for spliceosome assembly, activation, and catalysis, as well as components of the exon-junction and transcription-export (TREX) complexes—promote the silencing of gfp*::his-61. RNAi of these factors failed to desilence gfp*::his-61 even in the distal region of the gonad, despite clear phenotypic evidence of successful knockdown in many cases (Figure S4). In the future, it will be interesting to ask whether reporters with some but not all introns removed, or containing weak splice consensus sequences, exhibit increased levels of silencing when splicing is perturbed.

A recent study identified EMB-4, a homolog of the intron-binding helicase Aquarius, as a WAGO-9/HRDE-1-interacting factor that increases the sensitivity of genes with multiple introns to piRNA silencing (Akay et al., 2017). EMB-4 is required for piRNA silencing only when multiple introns are present in the target, a finding consistent with the idea that positive signals from splicing promote resistance to small RNA-mediated silencing. EMB-4 was also enriched in coIPs with the putative protective Argonaute CSR-1 (Tyc et al., 2017). Clearly further investigation is required to understand whether recruitment of EMB-4 during splicing, or more directly as an Argonaute co-factor, enables it to shape the sensitivity of intron-containing mRNAs to piRNA silencing.

We do not know what mechanisms underlie cis-silencing of intronless mRNAs, nor do we know whether or how these mechanisms trigger downstream recognition by the Argonaute small RNA system. Components of the THO complex, required for mRNA processing and export downstream of splicing (reviewed in Katahira, 2012), suppress the expression of a number of transposon families (Zhang et al., 2021). Interestingly, these THO factors appear to promote silencing of unspliced mRNAs independently of the piRNA pathway (Zhang et al., 2021). Previous studies have linked THO components to the expression of unspliced piRNA precursors (Hur et al., 2016; Zhang et al., 2012, 2014), raising the possibility that components of this complex promote intronless gene silencing directly in cis, and also promote the export of intronless mRNAs as piRNA precursors, suggesting that these factors may lie at or below the divergence of the cis- and trans-silencing pathways.

Many naturally intronless genes are nevertheless abundantly expressed, most of them short, including histone and spermatogenesis genes (Miller et al., 2004; Pettitt et al., 2002). Interestingly, the endogenous retro-element CER1 expresses an ~8-kilobase intronless mRNA in the germline, somehow avoiding both piRNA silencing and the intronless silencing mechanism described here (Dennis et al., 2012). Moreover, our intronless reporters are not silenced in the somatic cells of the worm, and mitotic versus meiotic germ cells exhibit different sensitivities to the cis-arm of intronless silencing. Thus, gene-specific and tissue-specific mechanisms can bypass or counteract intronless silencing.

Histone regulatory sequences appeared to partially counteract the cis-arm of intronless silencing. Indeed, they were sufficient to completely bypass cis-silencing when the intronless reporter was comprised of only a gfp*::nls ORF. The histone regulatory sequences, however, did not prevent small RNA-mediated silencing of this transgene. The gfp ORF is approximately twice the size of endogenous his-61, suggesting that regulatory elements, perhaps in the 3′ UTR of his-61, are insufficient to shield the extended ORF from recognition by the intronless small-RNA silencing machinery. The finding that inclusion of introns in gfp protected the fusion mRNA from silencing suggests that splicing and the histone UTR both exert local influences that promote mRNA escape from these surveillance mechanisms.

The finding that histone regulatory regions can bypass cis-silencing but not small RNA-mediated silencing could mean that these mechanisms are initiated independently. Alternatively, it may be that small RNA pathways are simply more sensitive to these cues perhaps due to the amplification of small RNA silencing signals. Regulatory sequences in the histone mRNA 3′ UTR are known to recruit U7 snRNP, which includes core components of the spliceosome (Dominski and Marzluff, 2007), and thus could recruit licensing factors or activities that promote the local removal of default silencing signals in a manner analogous to that envisioned for the successful resolution of splicing (see Model Figure 6B and above discussion).

Our findings provide evidence that mRNA splicing and histone regulatory sequences recruit activities to nascent mRNAs that prevent the default programming of an Argonaute system in the C. elegans germline. Licensing mechanisms of this type could be used as a line of defense against viral gene expression. However, just as eukaryotic cells appear to co-opt an antiviral dsRNA response to regulate endogenous gene expression via miRNAs, it is interesting to contemplate that cells may deploy intronless RNAs in order to program Argonautes as a means to regulate cognate intron-containing mRNAs. This need not involve RdRPs, as many Argonautes engage guide RNAs transcribed by Pol II (Corrêa et al., 2010; Gu et al., 2012). For example, by expressing an intronless antisense transcript from a processed pseudogene, the cell might initiate Argonaute programming that acts in trans to regulate the intron-containing gene from which the pseudogene was processed. Such a mechanism could underlie the recently discovered process of transcriptional compensation (El-Brolosy et al., 2019; Serobyan et al., 2020; Watanabe and Lin, 2014)

Conceivably there are numerous as yet undiscovered cellular mechanisms that program Argonaute silencing and an equal number of mechanisms that enable transcripts to navigate and evade surveillance by Argonaute systems. Understanding this complexity will require evaluating the response of Argonaute systems to alterations in the myriad RNA-binding and -modifying activities that engage transcripts throughout the mRNA life cycle.

Limitations of the study

There are several limitations of the current study. Foremost, we have not as yet identified a direct link between the splicing machinery and silencing. While the complete absence of introns results in silencing, we have not explored whether the size of exons, or the number of introns are important factors. The genetic mechanism(s) of “cis-silencing” are entirely unknown and may or may not be linked to the same cues that trigger small-RNA silencing. Finally, we used an intronless gfp*::his-61 reporter to facilitate much of the genetic analysis of small RNA silencing described here. Further studies will be required to determine if findings based on this reporter hold true for other intronless genes.

STAR*METHODS

Detailed methods are provided in the online version of this paper and include the following:

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Craig Mello (Craig.Mello@umassmed.edu).

Materials availability

All materials generated in this study are available from the Lead Contact without restrictions.

Data and code availability

• Original and processed small RNA high-throughput sequencing datasets are publically available under the following accession number GEO: GSE178985. Sanger sequencing trace files can be found at Mendeley Data: https://doi.org/10.17632/c8h2f2prch.2.

• This study did not generate a new code, but the scripts used in the study are available from the Lead Contact upon request.

• 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 strains and genetics

All strains in this study were derived from Bristol N2 line and cultured in a 20°C incubator, on normal growth media (NGM) agar plates seeded with E. coli OP50, essentially as described (Brenner, 1974). The majority of the assays were performed on hermaphrodites; males were used solely for crossing to generate additional strains or test transitive silencing as described in corresponding figures. The strains used in this study are listed in Table S1. Details of CRISPR alleles generated in the study, including description of the lesions, guide and donor repair sequences are provided in Table S2. Sanger sequencing trace files of the resulting CRISPR alleles can be found at Mendeley Data repository under Mendeley Data: https://doi.org/10.17632/c8h2f2prch.2.

METHOD DETAILS

Gonad dissections

Gonads were dissected in 0.5mM solution of tetramisole (Sigma-Aldrich) in PBSTw (PBS with 0.1% Tween 20) on Rite-On glass slides (Thermo Scientific), fixed with 2.5% paraformaldehyde in Happy buffer (81mM HEPES pH 6.9, 42mM NaCl, 5mM KCl, 2mM MgCl2, 1mM EGTA) (from personal correspondence with James Priess), covered with cover glass (MedSupply Partners), and directly imaged as described below.

Microscopy

Worms were mounted using a 0.5mM solution of tetramisole (Sigma-Aldrich) in M9 buffer (Brenner, 1974) on Rite-On glass slides (Thermo Scientific) and cover glass (MedSupply Partners). Epifluorescence and differential interference contrast (DIC) microscopy analysis and images were captured using an Axio Imager M2 Microscope (Zeiss), an ORCA-ER digital camera (Hamamatsu) and Zen (Zeiss) software. Images were processed using Fiji/ImageJ software (Schindelin et al., 2012).

MosSCI genome editing

Transgenic lines were generated in EG4322 background following Mos1-mediated single copy insertion (MosSCI) protocol as described (Frøkjaer-Jensen et al., 2008). Briefly, EG4322 animals carrying an unc-119(ed9[null]) III mutation and a Mos1 transposon were microinjected with a combination of plasmids: pCFJ601(carrying a Peft-3::transposase), fluorescent mCherry markers – pGH8 (Prab-3::mCherry), pCFJ104 (Pmyo-3::mCherry) and pCFJ90 (Pmyo-2::mCherry) –, and a repair template (pCCM956 or pCCM957 – a plasmid carrying a transgene to be integrated and an intact Cbr-unc-119(+)). After progeny of the injected animals were starved, unc-119(+) and mCherry-negative animals were selected. Transgenic integration at ttTi5605 site was assessed by PCR genotyping.

CRISPR/Cas9 genome editing

CRISPR lines were generated by Cas9 ribonucleoprotein (RNP) editing (Dokshin et al., 2018; Ghanta and Mello, 2020). For gfp insertion in gfp*::his-61 fusion strains and restored endogenous oma-1 strains, the annealed PCR products with overhangs served as donors (Dokshin et al., 2018). For deletion mutations and short insertions (such as FLAG and degron), commercially synthesized single stranded DNA oligonucleotides served as repair templates (IDT). Genomic PCR and Sanger sequencing were used for the validation of CRISPR editing. Details of CRISPR alleles generated in the study, including description of the lesions, accession links to Sanger sequencing trace files, guide and repair sequences are provided in Table S2.

Small RNA cloning and deep sequencing

Synchronized L1 populations were plated, collected and flash frozen in four volumes of TRI Reagent (Sigma) at the young adult stage. On the day of the experiment, after three freeze-thaw cycles, worm pellets were combined with ceramic spheres (Lysing Matrix D, MP Biomedicals) and homogenized using FastPrep system (MP Biomedicals) for three 20-sec cycles at a speed of 6 m/s. Total RNA was extracted using first a phase separation reagent 1-Bromo-3-chloropropane (BCP, Molecular Research Center), then aqueous phase was transferred to Phase Lock Gel tubes (Quanta Bio) for additional separation with phenol-chloroform. Following isopropanol precipitation and 80% ethanol wash, total RNA was resuspended with water and frozen at −80C. Small RNA isolation and cloning were performed as described (Li et al., 2020; Seth et al., 2018). Briefly, small RNAs, size-selected using mirVana miRNA Isolation Kit (Thermo Fisher Scientific), were treated with a recombinant PIR-1 pyrophosphatase (a generous gift from Dr. Weifeng Gu) to remove the 5’ γ and β phosphates, and ligated to 3’ and 5’ adapters using T4 RNA ligase 2 and T4 RNA ligase 1 (New England Biolabs), respectively. cDNA was then generated using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific) and amplified via PCR. After gel size selection, final libraries were sequenced on an Illumina HiSeq platform at the University of Massachusetts Medical School Deep Sequencing Core Facility.

Data analysis

Following standard removal of adapter sequences using Cutadapt tool (Martin, 2011), reads were mapped using bowtie2 aligner (Langmead and Salzberg, 2012; Langmead et al., 2019), normalized by the number of the reads mapping to the genome (WB272) or transcriptome (WBcel235) and multiplied by 5,000,000. Visualization of antisense 22-nucleotide-long reads starting with a guanine, as well as small RNA length profiles, were generated using ggplot2 (Wickham, 2016). All scripts are available upon request.

RNAi

RNAi was performed by feeding worms E. coli strain HT115(DE3) transformed with the control vector L4440 or a gene-targeting construct from the C. elegans Ahringer or Vidal RNAi libraries (Kamath and Ahringer, 2003; Rual et al., 2004). Frozen bacterial stocks were streaked on ampicillin (100 μg/ml) and tetracycline (10 μg/ml)-containing LB agar plates and grown overnight at 37°C. Then individual colonies were inoculated into LB with ampicillin (1:1000) and grown for 16–20 hrs on a shaker at 37°C. NGM plates containing 1 mM isopropyl β-d-thiogalactoside and 100 μg/ml ampicillin were seeded with the liquid culture (80 μl per plate) and incubated at room temperature for one day. L1 or L3 larvae were plated on RNAi plates and kept at 20°C until the phenotypes of the adults were scored (see Figure S4).

Bacterial clones targeting the following genes were used: oma-2, snr-2, sftb-1, yju-2, F33D11.10, Y65B4A.6, hel-1, mag-1, rnp-4, rnp-2, cyn-13, sel-13, snu-13, rnp-7, mog-1, mut-16, pos-1.

Gonad isolation and total RNA extraction

Ten gonads from N2, gfp::his-61 and gfp*::his-61 young adult worms were dissected in 0.5mM solution of tetramisole (Sigma-Aldrich) in PBSTw on Rite-On glass slides (Thermo Scientific) and incubated first in a PBS-EDTA-ATA buffer (1xPBS, 0.1 mM EDTA, 1mM aurintricarboxylic acid), followed by total RNA extraction using TRI Reagent (Sigma) and BCP (Molecular Research Center), and isopropanol RNA precipitation, as described in the TRI Reagent manual (Applied Biosystems, Manual 9738M Revision D).

RT-qPCR

For qPCR analysis of oma-1 samples, total RNA from the whole adult worms was used, as oma-1 only expresses in the germline. cDNA from both oma-1 and gfp::his-61 total RNA sample sets was synthesized with random hexamer primers using a SuperScript IV First-Strand Synthesis System (Invitrogen). qPCR reactions were set up using Fast SYBR Green Master Mix (Applied Biosystems) as per manufacturer’s protocol, and the qPCR assay was run on the QuantStudio Real-Time PCR system. For normalizing oma-1 transcript level, actin served as an endogenous control. The gfp::his-61 samples were normalized to a housekeeping ama-1 gene (RNA polymerase II subunit). Fold change in reporter expression was compared between samples using two-sided Welch’s t-test. P-values are reported in figure legends.

Auxin-inducible depletion of EGO-1

Auxin treatment was performed as described (Zhang et al., 2015). Indole-3-acetic acid (IAA; Fisher Scientific) was dissolved in ethanol with a final stock concentration of 400 mM. NGM plates containing 500 μM IAA (prepared by adding IAA to NGM agar at ~50°C) were seeded with fresh E.coli OP50 and incubated at RT in complete darkness for 2 days, then stored at 4°C in a light resistant container. Prior to the experiment, plates were warmed up at RT for 1 hr. Then, L1 larvae of degron::ego-1 strains were plated on NGM plates with or without 500 μM IAA, kept at RT and scored for embryonic lethality and germline GFP expression at the adult stage.

QUANTIFICATION AND STATISTICAL ANALYSIS

Exact sample sizes for every assay are described in figures. RT-qPCR data analysis was performed in R with “stat” package, using two-sided Welch’s t-test. Further statistical analysis details, including sample size and p-values, can be found in the Figure S1 legend.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Small RNA sequencing data	This paper	GEO: GSE178985
Sanger sequencing trace files	This paper	Mendeley Data: https://doi.org/10.17632/c8h2f2prch.2
Chemicals and recombinant proteins
Tetramisole hydrochloride	Millipore Sigma	L9756
10X PBS buffer PH 7.4	Thermo Fisher Scientific	AM9625
Rite-On glass slides	Thermo Fisher Scientific	1256820
Microscope slide cover glass	MedSupply Partners	G07–140410
TRI Reagent	Millipore Sigma	T9424
Alt-R® S.p. Cas9 Nuclease V3	Integrated DNA Technologies	1081058
Lysing Matrix D	MP Biomedicals	116913050-CF
Molecular Research Center BCP	Fisher Scientific	NC9551474
Phase Lock Gel tubes (heavy)	Quanta Bio	2302830
Acid-Phenol:Chloroform	Thermo Fisher Scientific	AM9720
Recombinant PIR-1 pyrophosphatase	Gift from Dr. Weifeng Gu, University of California, Riverside	N/A
T4 RNA Ligase 2	New England Biolabs	M0373S
T4 RNA Ligase 1	New England Biolabs	M0204S
SuperScript III Reverse Transcriptase	Thermo Fisher Scientific	18080085
Fast SYBR Green Master Mix	Thermo Fisher Scientific	4385612
Indole-3-acetic acid	Fisher Scientific	AAA1055614
Critical commercial assays
mirVANA™ miRNA isolation kit	Thermo Fisher Scientific	AM1560
SuperScript IV First-Strand Synthesis System	Thermo Fisher Scientific	18091050
Oligonucleotides
Primer: oma-1 Forward: CCGTTTCACC ACTCGATCAC	This paper	Integrated DNA Technologies
Primer: oma-1 Reverse: AAACTCTGAA TCGCGCGAAC	This paper	Integrated DNA Technologies
Primer: ama-1 Forward: AACGTGTGGA TTTCTCTGCG	This paper	Integrated DNA Technologies
Primer: ama-1 Reverse: ACTTGGCACC AGGATACTGTG	This paper	Integrated DNA Technologies
Primer: gfp::his-61 Forward: ACCTGTC CACACAATCTGCC	This paper	Integrated DNA Technologies
Primer: gfp::his-61 Reverse: ACGATGA AGACGACCGACTG	This paper	Integrated DNA Technologies
Primer: actin Forward: TGAAGTGCGAC ATTGATATC	This paper	Integrated DNA Technologies
Primer: actin Reverse: CTTGGAGATCC ACATCTGTT	This paper	Integrated DNA Technologies
CRISPR donors, see Table S3	This paper	N/A
Recombinant DNA
cdk-1p::cdk-1*::gfp*::cdk-1 3’UTR (entirely intronless); Cbr-unc-119(+) for MosSCI insertion into ttTi5605 II	This paper	pCCM956
oma-1p::oma-1*::gfp*::oma-1 3’UTR (entirely intronless); Cbr-unc-119(+) for MosSCI insertion into ttTi5605 II	This paper	pCCM957
Software and algorithms
Fiji/ImageJ	(Schindelin et al., 2012)	https://imagej.net/Welcome
Cutadapt	(Martin, 2011)	https://cutadapt.readthedocs.io
Bowtie2	(Langmead and Salzberg, 2012; Langmead et al., 2019)	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
R	RStudio Team	http://www.rstudio.com/
Ggplot2	(Wickham, 2016)	https://ggplot2.tidyverse.org/
BioRender	BioRender Team	https://biorender.com/
Experimental models: Organisms/strains
nrde-4(ne4679[p.Met12_Glu160del,fsTer21]) IV	Personal correspondence with Dr. Takao Ishidate, University of Massachusetts Medical School, Worcester	WM772
Strains used in this study are listed in Table S1.	This paper	N/A

Highlights.

• Lack of splicing triggers heritable silencing in the C. elegans germline

• Intronless genes are silenced by small RNA-dependent and -independent mechanisms

• Small RNA silencing is initiated independently of dsRNA or piRNA triggers

• A cognate spliced mRNA can prevent default small RNA silencing in trans