Main text

RNAs released into circulation can act as intercellular messages that are used for gene regulation in distant cells. Examples include secretion of small RNAs within exosomes in response to pathogenic fungal infection in Arabidopsis¹, virus-like proteins with their coding mRNAs in developing Drosophila² and mice³, microRNAs from adipose tissue in mice⁴ and small RNAs from the epididymis in mice^5,6,7,8. Such extracellular RNAs have also been detected in humans, but their roles in gene regulation remain unclear despite their use as diagnostic markers for diseases (reviewed in ref.⁹). Furthermore, the recent development of double-stranded RNA (dsRNA)-based drugs (reviewed in ref.¹⁰; refs.^11,12) that can silence genes of matching sequence through RNA interference¹³ has heightened interest in understanding the import of dsRNA into cells. A conserved dsRNA-selective importer, SID-1^14,15,16, is required for the import of extracellular dsRNA into the cytosol of any cell in the nematode C. elegans. This entry into the cytosol is distinct from and can follow the uptake of dsRNA into cells, which can rely on other receptors (e.g., SID-2 for uptake from the intestinal lumen^17,18). SID-1 has two homologs in mammals – SIDT1 and SIDT2. Although similar cytosolic entry of dsRNA through these mammalian homologs of SID-1 is supported by studies in mice reporting entry of viral dsRNA through SIDT2¹⁹, enhanced dsRNA uptake when SIDT1 is overexpressed in vitro²⁰, and uptake of ingested dsRNA into cells through SIDT1²¹, alternative roles for SIDT1 and/or SIDT2 in the uptake of cholesterol have also been proposed²².

Secretion of dsRNA from C. elegans tissues expressing dsRNA from transgenes has been inferred based upon the SID-1-dependent silencing of matching genes in other tissues^14,23. Secreted dsRNA from neurons can silence genes of matching sequence in most somatic cells²⁴ and within the germline²⁵. Extracellular dsRNA delivered into parental circulation by injection or ingestion also enters the germline and can cause silencing of matching genes in progeny^{13,26,27,28,29}. In every case, the entry of dsRNA into the cytosol dictates when and where the processing of extracellular dsRNA can begin. Such intergenerational transport of RNA is an attractive mechanism for explaining endogenous, gene-specific effects in progeny that could occur in response to changes in somatic tissues of parents. However, which conditions induce transport of dsRNA into the germline, when during development this transport occurs, and what regulatory consequences ensue in progeny upon uptake of extracellular dsRNA from parents are all unknown. Despite this lack of knowledge, the analysis of transgenerational gene silencing triggered by dsRNA has revealed that a class of small RNAs called 22G RNA made using the mRNA targeted by dsRNA and bound by Argonaute proteins in the germline^30,31 is necessary for observing silencing in every generation.

Timed dsRNA injection into animals or timed dsRNA ingestion by animals has been the only way to study mechanisms of dsRNA transport throughout C elegans development since no tools to induce dsRNA secretion from cells have been developed. Injection of dsRNA into adult sid-1(-) animals demonstrated that extracellular dsRNA can be directly transmitted to progeny without entry into the cytosol^28,29. This intergenerational transmission of dsRNA in sid-1(-) adult animals requires the yolk receptor RME-2³² and is independent of parental 22G RNA production because dsRNA cannot enter the cytosol in the parent. In further support of the initial intergenerational transport of extracellular dsRNA or dsRNA-derived signals not requiring 22G RNAs, gfp-dsRNA from parents that lack gfp sequences are transported to progeny (Fig. 1E in ref.²⁸). Thus, the silencing signals transported from parent to progeny include the extracellular dsRNA and its derived silencing signals, independent of 22G RNAs. However, 22G RNAs are required for silencing and they may be used for intergenerational transport in subsequent generations.

Fig. 1.

All dsRNAs, regardless of length, have been assumed to be equivalent substrates for entry into the cytosol. This assumption is supported by the uptake of a variety of dsRNA substrates when C. elegans SID-1 is overexpressed in heterologous Drosophila S2 cells¹⁶. Yet, two key observations suggest that dsRNA can take multiple routes in vivo before SID-1-dependent entry into the cytosol. One, even in the presence of SID-1, dsRNA ingested during early adulthood requires RME-2 to cause silencing in progeny²⁸. Two, 50-bp fluorescently-labeled dsRNA requires RME-2 for entry from parental circulation into oocytes²⁸. Establishing the contexts for the use of different modes of dsRNA transport is crucial for understanding the processes regulated by endogenous extracellular dsRNA.

Here, we dissect the intergenerational transport of extracellular dsRNA and discover a role for this mechanism in modulating RNA regulation within the germline. Extracellular dsRNA is transported with developmental and substrate specificity from parental circulation to progeny, and its release from neurons can be enhanced using light-induced oxidative damage. Blocking dsRNA import into the cytosol of all cells revealed heritable changes in gene expression and led to the identification of sid-1-dependent gene-1 (sdg-1). The sdg-1 coding sequence is located within a retrotransposon that is targeted by RNA silencing in the germline. Yet, the SDG-1 protein colocalizes with regulators of RNA silencing in perinuclear granules within the germline and dynamically enters the nucleus in proximal oocytes and in cells of developing embryos. Measurements of sdg-1 expression using native mRNA, a translational reporter, or a transcriptional reporter reveal that expression is easily perturbed in different mutants that impact dsRNA-mediated gene regulation. Expression varies between the two gonad arms of wild-type animals, and different mutant isolates can show an increase or decrease in expression, indicative of a loss of buffered gene expression within the germline. However, consistent with an overall role for SDG-1 (and potentially other SDGs) in promoting RNA silencing, either loss of SID-1 or overexpression of SDG-1 enhances piRNA-mediated silencing within the germline initiated by mating. Therefore, we propose that the import of extracellular dsRNA into the germline tunes intracellular pathways that cause heritable RNA silencing.

Requirements for the entry of extracellular dsRNA into the germline change during development

A convenient method for the delivery of extracellular dsRNA into C. elegans at various times during larval development is the expression of dsRNA in the bacteria that the animals ingest as food²⁶. To determine when ingested dsRNA can enter the germline and cause silencing, we exposed developing animals with a ubiquitously expressed protein (GTBP-1) tagged with GFP to bacteria that express gfp-dsRNA. Silencing was detectable within the germline from the second larval stage (L2) onwards (Fig. 1a and Extended Data Fig. 1a), but either exposure to ingested dsRNA beyond the fourth larval stage (L4) (Fig. 1b) or injection of dsRNA into the 1-day old adult germline (Extended Data Fig. 1b) was required to observe silencing in the germline of 3-day old adults. Combined with the need for exposure to dsRNA after the L4 stage^28,29 for silencing in progeny, even for just 24 hours (Extended Data Fig. 1c), these observations suggest that heritable RNA silencing is not effectively initiated during early development of the germline despite detectable silencing within the germline. One possible explanation for this observation could be that both RNAs derived from the imported dsRNA and downstream silencing signals are continually diluted by the proliferation of germ cells. Heritable silencing by dsRNA ingested from the L4 stage to the first day of adulthood likely relies on entry of dsRNA into the proximal germline because silencing of a somatic gene in progeny after parental ingestion of dsRNA during this period required RME-2 (Extended Data Fig. 1c), which is enriched in the proximal germline (Extended Data Fig. 1d and ref.³²), and some gtbp-1::gfp animals exposed to gfp-dsRNA up to the first day of adulthood showed more silencing in the proximal germline (Extended Data Fig. 1e).

Thus, these results reveal three periods of germline development that can be broadly distinguished based on the response to ingested dsRNA: (1) from the first larval to the fourth larval stage when exposure to dsRNA does not result in maximal silencing within the germline in adults (Fig. 1b); (2) from the fourth larval stage to early adulthood when entry of dsRNA primarily occurs in the proximal germline through RME-2 (Extended Data Fig. 1c,e); and (3) later adulthood when germline silencing by ingested dsRNA is maximal (Fig. 1b) and ingested dsRNA can effectively silence progeny independent of RME-2 (Extended Data Fig. 1c and ref.²⁹). These differences in the entry of ingested dsRNA into cells and/or subsequent silencing could be driven by a variety of changes during development. These include changes in the uptake of dsRNA into the intestine, distribution of dsRNA to other tissues from the intestine, import of dsRNA into the germline, and availability of RNA silencing factors within the germline.

Oxidative damage in neurons expressing dsRNA enhances silencing in the germline by neuronal dsRNA

Another approach for delivering extracellular dsRNA into the germline that better mimics dsRNA transport between cells is the secretion of dsRNA from neurons²⁵. However, the extent of such secretion throughout development is unpredictable. To modulate the secretion of dsRNA from somatic cells into parental circulation during development, we adapted an optogenetic approach for damaging somatic cells³³. Specifically, we generated animals that express the mini singlet oxygen generator (miniSOG) protein in neurons and exposed them to blue light. While animals expressing miniSOG from a single-copy transgene did not show an appreciable defect when compared with wild-type animals, those expressing miniSOG from a multi-copy transgene were paralyzed (Extended Data Fig. 2a,b, top) and had visibly damaged neurons (Extended Data Fig. 2b, bottom). Using this system, we induced oxidative damage in the neurons of animals that expressed dsRNA under the control of a neuronal promoter and evaluated silencing of target genes with matching sequence expressed in other tissues (Fig. 2a). By exposing animals to blue light for 60 minutes at different times during development (Extended Data Fig. 2c), we observed SID-1-dependent enhancement in the silencing of the hypodermal gene bli-1 at the adult stage by neuronal bli-1-dsRNA, with maximal silencing when oxidative damage occurred during mid-to-late larval development (Extended Data Fig. 2d, light exposure every 6 hours from 42 to 66 hours post L4-stage of parent; Extended Data Fig. 2e, ∼2-fold increase from 14.9% to 29.1% in a background with enhanced RNA interference (eri-1(-)) and ∼6-fold increase from ∼1.6% to ∼9.8% in a wild-type background). A similar period of maximal SID-1-dependent enhancement of silencing was also observed when neurons expressing gfp-dsRNA were damaged and silencing of a two-gene operon that expresses two fluorescent proteins, mCherry::H2B and GFP::H2B, in the germline was measured (Fig. 2b-d, and Extended Data Fig. 2f – 48 to 60 hours post L4-stage of parent; sid-1(-) allele (jam80[nonsense]) is depicted in Extended Data Fig. 3). While silencing of gfp::h2b was observed throughout the germline, silencing of the other cistron mCherry::h2b was also observed sometimes, albeit restricted to specific regions of the germline. Silencing of mCherry::h2b was most frequent in the proximal germline and was not observed in any other region without silencing in the proximal germline (proximal germline – 57%, distal germline – 47%, sperm – 29%, Fig. 2d), likely due to reduction of mCherry::h2b::gfp::h2b pre-mRNA³⁴ in those regions. Consistently, the silencing of both gfp::h2b and mCherry::h2b was eliminated in the absence of the nuclear Argonaute HRDE-1 (Extended Data Fig. 2g). The pattern of mCherry::h2b silencing is similar to the spatial pattern observed for the RME-2-dependent entry of dsRNA delivered into parental circulation²⁸ and is consistent with the pattern of target mRNA degradation in the germline by extracellular dsRNA³⁵. However, silencing of gfp::h2b in the germline by neuronal dsRNA did not show a detectable dependence on RME-2 (Extended Data Fig. 2h - difficulty in obtaining transgenic animals that also lack RME-2 resulted in a low sample size for this experiment).

Fig. 2.

Thus, by modulating the secretion of dsRNA from somatic cells for the first time, we gained two insights into the intercellular transport of dsRNA: (1) oxidative damage of neurons during particular periods in development increases the amount of dsRNA and/or changes the kinds of dsRNA in circulation either because of specific enhancement of secretion or nonspecific spillage; and (2) there is a preference for silencing by neuronal dsRNA in the proximal germline. However, the temporal and/or spatial preferences observed for silencing by both neuronal and ingested dsRNA could be because of unknown characteristics of the exported neuronal dsRNA or ingested dsRNA expressed from bacteria (e.g., modifications, lengths, structures, etc.) that influence import or subsequent silencing. This hypothesis is supported by the different genetic requirements reported for silencing by neuronal gfp-dsRNA compared to other sources of gfp-dsRNA²⁴. Alternatively, these preferences could reflect universal constraints for silencing using any extracellular dsRNA (e.g. expression patterns of factors that promote the import or processing of dsRNA).

Extracellular dsRNA in parental circulation can be transported through multiple routes to cause silencing in progeny

While the characteristics of extracellular dsRNA imported into the germline from ingested bacteria or from neurons are unknown, delivery of chemically defined dsRNA into the extracellular space in C. elegans can be accomplished using microinjection^13,28. We examined differences, if any, in the entry of in vitro transcribed dsRNA from the extracellular body cavity into the germline during the L4 and adult stages as evidenced by silencing in the progeny of injected animals. Silencing was comparable regardless of whether wild-type or rme-2(-) parents were injected as L4-staged or adult animals (Fig. 3a and Extended Data Fig. 4a, left; also reported for adults in ref.²⁹). However, some dependence on RME-2 for silencing in progeny was discernable when lower concentrations of dsRNA were used (Extended Data Fig. 4a, right). This result and previous results demonstrating proximal germline-specific silencing in contexts where silencing is less robust (Extended Data Fig. 1e and Fig. 2d) are consistent with RME-2-dependent uptake of dsRNA being a route for extracellular dsRNA to enter the germline. The difference in parental RME-2 requirement for silencing in progeny observed for dsRNA ingested (Extended Data Fig. 1c) or injected (Extended Data Fig. 4a) at the L4 stage could similarly reflect the accumulation of different amounts of dsRNA in parental circulation (e.g., more upon injection than upon ingestion), and/or different kinds of dsRNA (e.g., because of modifications in bacteria or upon transit through the intestine). However, these possibilities could not be easily distinguished because sensitive northern blotting³⁶ revealed that both bacterial and in vitro transcribed dsRNA consist of a complex mix of dsRNAs (Extended Data Fig. 4b-d; consistent with refs.^37,38), hereafter called mixed dsRNA.

Fig. 3.

In contrast, when synthesized gfp-dsRNA of a defined length (50 bp) with a fluorescent label was injected into circulation in adult animals, no entry into the germline was observed in the absence of RME-2²⁸. We found that silencing of unc-22 in progeny by similarly synthesized but unlabeled 50-bp unc-22-dsRNA with a 5’ OH delivered into parental circulation also showed a strong requirement for RME-2, unlike mixed dsRNA (Fig. 3a). Further comparison between the two forms of dsRNA revealed that silencing in progeny by 50-bp dsRNA injected into parental circulation was detectably less efficient in somatic cells (Fig. 3b and Extended Data Fig. 5a,b, left), even when ∼14X more 50-bp dsRNA was delivered into parental circulation (Extended Data Fig. 5b, right), and was also less efficient in the germline (Fig. 3b and Extended Data Fig. 5a,c). Efficient silencing in response to added dsRNA requires nuclear Argonaute proteins: NRDE-3 in somatic cells³⁴ and HRDE-1 in the germline³⁰. Both 50-bp dsRNA and mixed dsRNA relied on the HRDE-1 for silencing within the germline (Extended Data Fig. 5a,c) and could silence independent of NRDE-3 in somatic cells (Extended Data Fig. 5a,c). Therefore, the observed difference in the extent of silencing cannot be attributed to differential engagement of these Argonautes, but rather could be the result of differences in the ability of each type of dsRNA to bind to upstream factors in the RNA interference pathway (e.g. RDE-4), differences in the stability of each type of dsRNA, and/or differences in the intergenerational transport of each type of dsRNA. In addition to the diversity of RNA lengths observed in mixed dsRNA, another known feature that could distinguish dsRNA transcribed in bacteria or in vitro from synthesized 50-bp dsRNA is the presence of 5’ triphosphates on the transcribed dsRNA species instead of the 5’ OH present in synthesized 50-bp dsRNA. In support of the impact of 5’ phosphates on transport and/or silencing, addition of 5’ monophosphates to synthesized 50-bp dsRNA injected into parental circulation reduced the dependence on RME-2 for silencing in progeny (Extended Data Fig. 4e,f), potentially by enhancing the ability of synthesized dsRNA to be imported by other dsRNA importers (e.g. SID-1) in the absence of RME-2. Thus, the requirements for dsRNA entry into the germline and subsequent silencing vary for different lengths and/or chemical forms of dsRNA (see summary in Extended Data Table 1).

Fluorescently labeled 50-bp dsRNA delivered into parental circulation localized within intestinal cells in progeny (Fig. 3c, top left), as has been observed for vitellogenin proteins³⁹ and fluorescent dyes⁴⁰. Accumulation of fluorescently labeled dsRNA was also detected at the apical membrane of the intestine, which could reflect exocytosis of dsRNA into the lumen of developing intestinal cells. However, separation of the fluorescent label from dsRNA catalyzed by cellular enzymes cannot be excluded. Therefore, to dissect differences, if any, between the transport of unlabeled short dsRNA (synthesized 50-bp with 5’OH) and mixed dsRNA (mixture transcribed in vitro using ∼1 kb DNA template) we injected unc-22-dsRNA into animals with mutations in genes that play roles in the import of dsRNA. We found that maternal SID-1 was required for silencing by short dsRNA in progeny (Fig. 3c, bottom, left bars), suggesting that the SID-1-dependent entry of short dsRNA into the cytosol likely occurs in the injected parent or during early development in progeny. Uptake of dsRNA from the intestinal lumen requires SID-2, a transmembrane protein located in the apical membranes of intestinal cells^17,18. We found that SID-2 was not required for most silencing in progeny by short or mixed dsRNA injected into parental circulation (Fig. 3c, top right and bottom). Exit of dsRNA from intracellular vesicles requires SID-5, a transmembrane protein located in endolysosomal membranes⁴¹. Silencing in wild-type animals was comparable to silencing in sid-5(-) animals (Fig. 3c, top right). However, when animals that lacked SID-1 were injected, SID-5 was required in progeny for silencing by mixed dsRNA from parental circulation (Fig. 3c, bottom, right bars; as also reported in ref.²⁹). Since dsRNA is expected to be present in vesicles upon entry through RME-2 in the absence of SID-1^28,29, this observation suggests that SID-5 is required for the release of mixed dsRNA from inherited vesicles in progeny.

In summary, injected extracellular dsRNA can enter the germline in parents and be transmitted to progeny through two routes with different substrate selectivity. One route is preferentially used by short dsRNA and relies on RME-2-mediated endocytosis of dsRNA into oocytes, where early exit from vesicles is required for silencing in progeny as evidenced by the need for maternal SID-1 (Fig. 3d, blue). The other route appears to exclude short dsRNA but allows mixed dsRNA entry into the cytosol in the parental germline through SID-1 and exit from inherited vesicles in progeny through a process that requires both zygotic SID-1 and SID-5 (Fig. 3d, grey; ref.²⁹).

Expression of SID-1 is consistent with a role in the intergenerational transport of extracellular dsRNA

All routes of dsRNA transport deduced using the experimental addition of dsRNA ultimately require SID-1 for entry into the cytosol. The proposed model (Fig. 3d) for dsRNA transport into the germline and to progeny suggests the expression pattern of SID-1 is likely to include the germline. We used Cas9-mediated genome editing to insert a piRNA-resistant mCherry sequence^42,43 into the sid-1 coding sequence (Fig. 4a) to observe the endogenous expression pattern of SID-1::mCherry. This fusion protein was detectably functional in contrast to other attempts at tagging SID-1 (see Technical comments in Methods). Fluorescence from SID-1::mCherry was visible in the proximal and distal regions of the adult germline (Fig. 4b). Expression also progressively increased during development with tissue-specific enrichment in the developing embryo (Fig. 4c, left), becoming ubiquitous in hatched L1 larvae (Fig. 4c, middle,d). SID-1::mCherry was not easily detectable in the germline during later larval development (Fig. 4c, middle and right). In combination with the expression of RME-2 observed in the proximal germline (Extended Data Fig. 1d; ref³²), this expression pattern of SID-1 is consistent with the entry of most dsRNA from circulation of adult animals into the germline, followed by activity of transport mechanisms in developing embryos and early larvae that inherit parental dsRNA.

Fig. 4.

To determine if acute, induced expression rather than developmental expression of SID-1 can be sufficient for the import of dsRNA into the germline, we engineered the endogenous sid-1 gene to transcribe a fusion transcript with an aptamer-regulated ribozyme (Extended Data Fig. 6a, left) that cleaves itself when not bound to tetracycline (Extended Data Fig. 6a, right) (based on ref.⁴⁴). Exposing these animals to tetracycline enabled silencing by dsRNA in somatic tissues (hypodermis: Extended Data Fig. 6b, left; body-wall muscles: Extended Data Fig. 6b, right), indicative of stabilization of sid-1 mRNA, production of SID-1 protein, and subsequent dsRNA import in somatic cells. However, such tetracycline-induced silencing was not detectable in the germline (Extended Data Fig. 6c-f). Yet, similar tagging of the ubiquitously expressed gene gtbp-1::gfp resulted in detectable rescue of expression within the germline by tetracycline (Extended Data Fig. 6g). A possible explanation for the poor rescue of SID-1 activity within the germline is that post-transcriptional mechanisms targeting sid-1 mRNA in the germline but not in somatic cells interfere with tetracycline-dependent stabilization of the sid-1 transcript (e.g., piRNA-based regulation of sid-1 mRNA^45,46), or that acute stabilization of the sid-1 transcript does not override developmental regulation of SID-1 translation.

Additional attempts to tag the SID-1 protein guided by structure and to modulate sid-1 transcripts guided by post-transcriptional regulatory interactions could improve control of dsRNA transport between cells. Nevertheless, the developmentally regulated expression observed for both SID-1 and RME-2 in the germline is consistent with intergenerational or transgenerational effects of endogenous dsRNA from parental circulation after development of the adult germline.

Temporary loss of sid-1 causes a transgenerational increase in the levels of mRNA from two germline genes

To understand how the dsRNA importer SID-1 might be used in endogenous gene regulation across generations, we searched for sid-1-dependent changes in gene expression that could be heritable (Fig. 5, Extended Data Fig. 3 and Extended Data Fig. 7). To control for genetic background (see Technical comments in Methods), we used Cas9-mediated genome editing to delete the entire sid-1 coding sequence or introduce a nonsense mutation in cohorts of the same wild-type animals. By comparing polyA+ RNA from this wild type with that of the newly generated sid-1(jam113[deletion]) (Fig. 5a,b and Extended Data Fig. 7a) or sid-1(jam80[nonsense]) (Fig. 5a-c) animals, we found that 26 genes were significantly (q < 0.05) misregulated in sid-1(jam113[deletion]) (Extended Data Fig. 7b) and 6 in sid-1(jam80[nonsense]) (Fig. 5d, top), both including sid-1 (a list of significantly altered genes is in Extended Data Table 2). The most upregulated gene in sid-1(jam113[deletion]), F14F9.5 (Extended Data Fig. 7b), which is located immediately 3’ to sid-1 in the genome, was only misregulated in the deletion mutant and not in the nonsense mutant (Extended Data Fig. 7c, left). Both mutants, however, were equally defective for silencing by ingested dsRNA (Fig. 5b). This observation suggests that while both mutations result in loss of SID-1 protein, the deletion of sid-1 also changes local regulatory sequences (potentially explaining upregulation of the neighboring gene F14F9.5) and eliminates sid-1 mRNA, which could participate in RNA-based regulatory interactions within the germline^45,46. Nevertheless, we could detect two genes that were upregulated in both sid-1(jam113[deletion]) and sid-1(jam80[nonsense]) animals (red in Fig. 5d, top, and Extended Data Fig. 7b): the identical loci W09B7.2/F07B7.2 (Extended Data Fig. 7c, middle), and Y102A5C.36 (Extended Data Fig. 7c, right). Intriguingly, another gene cls-3 also changed in both mutants (Extended Data Table 2) but in different directions (∼3.4-fold decrease in the sid-1(jam80[nonsense]) mutant but a ∼5.8-fold increase in the sid-1(jam113[deletion]) mutant), suggesting that the direction of change in expression can vary. Conservatively, we began by analyzing only the two genes with mRNA levels that changed in the same direction in both sid-1 mutants. Both W09B7.2/F07B7.2 and Y102A5C.36 have been reported⁴⁷ to be expressed within the germline (Extended Data Fig. 7d, left) and regulated by endogenous small RNAs (Extended Data Fig. 7d, middle and right). Spliced mRNA levels measured at a later generation using RT-qPCR demonstrated that both transcripts were upregulated in sid-1(jam80[nonsense]) animals compared to wild-type animals as expected (Fig. 5e), but no upregulation was detectable in sid-1(jam113[deletion]) animals (Extended Data Fig. 7e). This difference between the two sid-1 mutants could reflect increased variation in expression (as was observed for cls-3 using RNAseq) or could reflect complex effects caused by deletion of sid-1 DNA (e.g., F14F9.5 overexpression, loss of sid-1 mRNA, etc.) that could be independent of SID-1 protein function.

Fig. 5.

To determine if changes in W09B7.2/F07B7.2 and Y102A5C.36 expression were heritable, we reverted the sid-1 nonsense mutation to wild-type sequence using Cas9-mediated genome editing. This immediately restored silencing by ingested dsRNA (Fig. 5b) with concomitant recovery of sid-1 mRNA to wild-type levels (Fig. 5e, top). In contrast, changes in both W09B7.2/F07B7.2 and Y102A5C.36 expression persisted (Fig. 5d, bottom) even after one year of passaging the reverted animals (sid-1(jam86[revertant])) (i.e., after >100 generations, Fig. 5e, middle and bottom). Thus, the sid-1-dependent accumulation of mRNA from these two germline genes persisted for many generations, likely through mechanisms that maintain heritable epigenetic changes. We hereafter refer to these sid-1-dependent genes (sdg) that show heritable epigenetic changes in response to temporary loss of SID-1 as sdg-1 (W09B7.2/F07B7.2) and sdg-2 (Y102A5C.36).

The sid-1-dependent gene sdg-1 is affected by many factors that regulate RNA silencing in the germline

To determine if expression of sdg-1 and sdg-2 is regulated by other proteins that play a role in RNA silencing within the germline, we examined 21 published datasets^30,46,48-58 that reported changes that depend on such proteins. For each dataset, we determined if the lists of genes reported as showing significant changes in mutants compared to the respective wild types included sdg-1 and/or sdg-2. This analysis revealed that changes in mRNA and/or antisense small RNAs of sdg-1 were detected in 20 of the 21 datasets while changes in sdg-2 were observed in 9 of 21 (Fig. 5f). When detected, changes in sdg-2 were in the same direction as changes in sdg-1, suggestive of similar regulation of both genes.

RNAs transcribed in the germline can be recognized as they exit the nuclear pores by piRNAs bound to the Argonaute PRG-1, which recruits them for regulation by antisense small RNAs called 22G RNA made by proteins organized within perinuclear germ granules (reviewed in ref.⁵⁹). Interaction with piRNAs was detected for RNA from sid-1, sdg-1, and sdg-2, and the control gene tbb-2 using crosslinking, ligation, and sequencing of hybrids⁵³ (Fig. 5f), consistent with their germline expression. Depletion of downstream 22G RNAs targeting both sid-1 and sdg-1 was detectable upon loss of the germ granule component MUT-16⁵¹ (Fig. 5f). Both genes were among the top 500 genes targeted by 22G RNAs bound by the secondary Argonaute HRDE-1/WAGO-9³⁰ (Fig. 5f), suggesting similar downregulation of both genes using 22G RNAs. Furthermore, multiple datasets support downregulation of sdg-1 within the germline by HRDE-1/WAGO-9-bound 22G RNAs in the absence of PRG-1. One, loss of HRDE-1/WAGO-9 increased sdg-1 RNA in whole animals⁵⁸ (Fig. 5f) and in dissected gonads⁵⁷ (Fig. 5f). Two, loss of PRG-1 decreased sdg-1 RNA (Fig. 5f) and increased 22G RNAs that are antisense to sdg-1 (Fig. 5f) in dissected gonads⁵². Three, although animals that lack PRG-1 become progressively sterile, the increase in sdg-1 22G RNA persisted in near-sterile animals (Fig. 5f, near-sterile in ref.⁵²), and this increase was eliminated upon additional loss of HRDE-1/WAGO-9 (Fig. 5f, near-sterile in ref.⁵²).

As expected for sid-1-dependent downregulation of sdg-1, multiple datasets reveal an inverse relationship between the two genes. In animals lacking PRG-1, sid-1 RNA levels increased and sid-1 22G RNAs decreased⁵⁵ (Fig. 5f), but both sdg-1 RNA and sdg-2 RNA levels decreased along with an increase in 22G RNAs^51-56 (Fig. 5f). This inverse relationship between sid-1 and sdg-1 RNA regulation is also observed when many components of germ granules are mutated as indicated by changes in 22G RNA upon loss of the embryonic P granule components MEG-3/-4⁴⁶ (Fig. 5f), the PRG-1 interactor DEPS-1⁵¹ (Fig. 5f), or the Z granule component ZSP-1⁵⁰ (Fig. 5f; also known as PID-2⁶⁰).

In addition to the above studies, pioneering studies that used microarrays identified sdg-1 as upregulated in animals lacking the germ granule component DEPS-1⁴⁹ (Fig. 5f) and in animals lacking the dsRNA-binding protein RDE-4 (Fig. 5f; second-most upregulated in ref.⁴⁸), which recruits dsRNA imported through SID-1 and other intracellular dsRNA for processing and eventual gene silencing. Animals that lack RDE-4 show a ∼47.5-fold increase in sdg-1 RNA⁴⁸. A reduction in RDE-4 activity could also contribute to the ∼11.6-fold increase in sdg-1 RNA seen in deps-1(-) animals because these animals also show a ∼3.2-fold decrease in rde-4 RNA⁴⁹ (one of 13 downregulated genes). These observations support the idea that appropriate regulation of sdg-1 RNA requires both piRNA-mediated processes that act via germ granule components such as DEPS-1 and dsRNA-mediated processes that use SID-1 and RDE-4.

In summary, the levels of sdg-1 RNA are detectably regulated by the dsRNA-selective importer SID-1, the dsRNA-binding protein RDE-4, and the piRNA-binding Argonaute PRG-1. Presence of dsRNA-mediated regulation or loss of piRNA-mediated regulation enhances MUT-16-dependent production of secondary small RNAs that bind the secondary Argonaute HRDE-1/WAGO-9. Consistent with downregulation of these sid-1-dependent transcripts by SID-1, disruption of many components of germ granules results in opposite effects on these transcripts and sid-1 RNA. Intriguingly, a search of protein interaction studies revealed that the SDG-1 protein is among the interactors of two germ granule components: PID-2 by immunoprecipitation⁶⁰ (also known as ZSP-1⁵⁰) and DEPS-1 by proximity labeling⁶¹. Thus, one possibility suggested by these observations is that reduction of sdg-1 RNA via SID-1 alters the amount of SDG-1 protein, which could interact with components of germ granules to mediate RNA regulation within the germline of wild-type animals.

Regulation of sdg-1 RNA is susceptible to epigenetic changes that last for many generations

SDG-1 is encoded by a gene located within a retrotransposon (Extended Data Fig. 8a) that is within a duplicated ∼40 kb region and has two recognizable paralogs (Extended Data Fig. 8b). To facilitate analysis of SDG-1 expression, we tagged both loci that encode SDG-1 with mCherry coding sequences lacking piRNA-binding sites^42,43 (mCherryΔpi) (Extended Data Fig. 8c,d). This tagging resulted in the expression of sdg-1::mCherryΔpi mRNA being ∼16-fold higher than sdg-1 mRNA (Extended Data Fig. 8e), potentially because of the reduction in the overall density of piRNA-binding sites per transcript, the additional introns included in mCherryΔpi (based on refs.^62,63), and/or other unknown factors. Fluorescence from SDG-1::mCherry was observed in the germline of adult animals (Fig. 6a). However, animals showed variation in SDG-1::mCherry expression between their two gonad arms (Fig. 6a, middle shows bright anterior (20% of animals) and right shows bright posterior (6% of animals)). A contributing feature for the observed stochasticity could be the location of sdg-1 within a duplicated region (Extended Data Fig. 8a), as suggested by similar stochastic RNA silencing of multi-copy transgenes but not single-copy transgenes⁶⁴. Despite this variation, unbiased passaging of self-progeny for more than 18 generations continuously preserved SDG-1::mCherry expression in an otherwise wild-type background (Fig. 6b). In contrast, mating, which can perturb RNA regulation within the germline in cross progeny⁴³, caused dramatic changes in sdg-1 expression that persisted in descendants (Fig. 6c). Mating animals that express SDG-1::mCherry with wild-type animals resulted in heritable changes along lineages that express sdg-1::mCherryΔpi mRNA or that express sdg-1 mRNA (Fig. 6c and Extended Data Fig. 9a). This discovery of mating-induced perturbation in gene expression raises caution in interpreting past studies (summarized in Fig. 5f) where changes in multiple independent isolates were not examined. Nevertheless, when we used genetic crosses to determine the impact of mutations on sdg-1 expression, we observed reduced SDG-1::mCherry fluorescence in mutants predicted to have reduced levels of intracellular dsRNA (Fig. 6d, sid-2(-), sid-5(-)) or reduced processing of intracellular dsRNA (Fig. 6d, eri-1(-)). In contrast, we observed an increase in SDG-1::mCherry fluorescence in animals lacking MUT-16 (Fig. 6d). Finally, animals lacking RME-2, which lack the ability to import many maternal factors (e.g., lipids, proteins, RNAs, etc.), also showed an increase in SDG-1::mCherry fluorescence (Fig. 6d).

Fig. 6.

To avoid mating-induced perturbations of RNA regulation within the germline, we used Cas9-mediated genome editing to introduce mutations into animals that express SDG-1::mCherry in an otherwise wild-type background. Use of this approach to mutate a control gene with no known roles in RNA regulation within the germline resulted in similar levels of SDG-1::mCherry fluorescence in multiple isolates of animals with and without the mutation (Extended Data Fig. 9b). In contrast, mutating sid-1 using Cas9-mediated genome editing caused a range of expression levels in different isolates when compared with sid-1(+) animals (Fig. 6e - 6 isolates lower, 2 isolates comparable, and 1 isolate higher), which differs from the increase in sdg-1 mRNA observed upon SID-1 loss in the single isolate examined earlier (Fig. 5). One explanation for these observations is that in the absence of SID-1, the levels of SDG-1 vary in different isolates because the sdg-1 gene becomes prone to heritable epigenetic changes, resulting in either high or low expression states being stabilized in different isolates. Alternatively, it is possible that the ∼16-fold overexpression of sdg-1::mCherryΔpi mRNA perturbs RNA-mediated epigenetic regulation, potentially indicative of a role for the SDG-1 protein in such regulation. Mutating sid-3 also lowered the levels of SDG-1::mCherry in one isolate, but caused no detectable change in another (Fig. 6f). While both isolates with loss of RDE-1 showed lower levels of SDG-1::mCherry, both isolates with loss of the germ granule component DEPS-1 showed higher levels of SDG-1::mCherry (Fig. 6f). These experiments also reveal that in most isolates of animals expected to have reduced levels of intracellular dsRNA (sid-3(-), sid-1(-)) or reduced processing of intracellular dsRNA (rde-1(-)), SDG-1::mCherry fluorescence is reduced. In contrast, isolates lacking DEPS-1 showed increased expression of SDG-1::mCherry.

Collectively, the observations on SDG-1 expression using mutants suggest that the uptake and processing of intracellular dsRNA (mediated by SID-1, SID-2, SID-3, SID-5, RDE-1, and ERI-1) and the function of germ granules (mediated by MUT-16 and DEPS-1) are both necessary to maintain intermediate levels of SDG-1 expression across generations. Once the levels of SDG-1::mCherry were reduced upon loss of SID-1, downregulation persisted across generations even after restoration of wild-type SID-1 (Fig. 6g), just as the upregulation of untagged sdg-1 mRNA also persisted (Fig. 5). Despite >100 generations of such persistent silencing, the expression of SDG-1::mCherry could be restored by mutating deps-1 (Fig. 6g), implicating small RNA-based regulation and germ granules in the maintenance of new epigenetic states established upon loss of SID-1.

SID-1-dependent genes, including SDG-1, could alter RNA-mediated regulation in the germline

Since SDG-1 interacts with regulators of RNA silencing (ZSP-1/PID-2⁶⁰ and DEPS-1⁶¹), loss of SID-1 could both block the entry of extracellular dsRNA into the cytosol and change intracellular RNA regulation through SID-1-dependent genes such as SDG-1. In support of this possibility, intracellular delivery of dsRNA through injection into the syncytial germline of sid-1(-) animals showed a weak defect in silencing (Fig. 2D in ref.²⁸ and Fig. 1 in ref.²⁹). To examine if changes in the levels of SDG-1 alone could account for such a defect in silencing by intracellular dsRNA in the germline, we exposed independently-generated sdg-1 deletion animals and animals that overexpress sdg-1 (i.e., with sdg-1::mCherryΔpi) to dsRNA matching the germline gene pos-1 for ∼16 hours. This short exposure to pos-1-dsRNA caused intermediate levels of silencing in wild-type animals and comparable intermediate silencing in sdg-1 mutant and sdg-1 overexpressing animals (Fig. 7a), suggesting that changes in sdg-1 levels alone are not sufficient to cause a defect in silencing by dsRNA that is detectable using this assay. Alternatively, the previously described defect in silencing by intracellular dsRNA in sid-1(-) animals^28,29 could be through the promotion of competing piRNA-mediated gene regulation in the absence of SID-1. This notion that dsRNA-mediated and piRNA-mediated gene regulation compete in the germline is supported by a study that demonstrated that loss of PRG-1 enhances heritable RNA silencing by dsRNA⁶⁵. To test if loss of SID-1 or a sid-1-dependent gene enhances piRNA-mediated silencing, we used an experimental system which initiates piRNA-mediated silencing of the two-gene operon described in Fig. 2 through mating, independent of externally provided dsRNA⁴³. We found that sid-1(-) animals exhibited enhanced mating-induced silencing (Fig. 7b, top right: ∼50% off in sid-1(+) vs 100% off in sid-1(qt9[nonsense])), while animals lacking sdg-1 showed a small reduction in mating-induced silencing that was not statistically significant (Fig. 7b, bottom right, ∼40% off in sdg-1(+) vs ∼30% off in sdg-1(jam232[deletion])). Taken together, these results support the model that loss of SID-1 weakly inhibits silencing by intracellular dsRNA^28,29 but enhances silencing by piRNAs within the germline, potentially through the action of multiple altered SID-1-dependent genes that collectively promote piRNA-mediated gene regulation.

Fig. 7.

RNA regulation within the germline using piRNAs relies on phase-separated granules (reviewed in ref.⁶⁶). To determine if the identification of SDG-1 as an interactor of the Z-granule component PID-2⁶⁰/ZSP-1⁵⁰ and potentially of the P-granule-adjacent protein DEPS-1⁶¹ could be seen as colocalization with Z-granules in vivo, we examined the localization of SDG-1::mCherry within the cytoplasm at high resolution using AiryScan imaging⁶⁷. SDG-1::mCherry was enriched in perinuclear foci in many animals (Fig. 7c, top; 7 of 9 animals) and these sites of enrichment colocalized with the Z-granule marker GFP::ZNFX-1^68,69 (Fig. 7c, bottom; 100% colocalized in 10 of 12 animals with enrichment). Time-course imaging revealed re-localization of SDG-1::mCherry into the nucleus from the cytoplasm in the -1 oocyte, which will be the first to be fertilized (Fig. 7d, left and Extended Data Movie 1), along with subsequent exclusion from the nucleus before the maternal and paternal pronuclei meet (Fig. 7d, right and Extended Data Movie 2). Such dynamic entry into the nucleus followed by exclusion from the nucleus also occurred during early cell divisions in the developing embryo (Extended Data Movies 3,4). The timing of nuclear entry of SDG-1::mCherry coincides with the nuclear envelope breakdown events that occur during fertilization and early development⁷⁰. The sdg-1 coding sequence was required for regulated nuclear entry because deletion of the sdg-1 open reading frame in sdg-1::mCherryΔpi animals resulted in mCherry expression throughout the germline in both the cytoplasm and nuclei (Fig. 7e). Nuclear localization of the SDG-1 protein in the -1 oocyte is like that of the essential Argonaute CSR-1b⁷¹, thought to play a role in protecting transcripts from silencing. Exposure to ingested dsRNA did not alter the expression of sdg-1p::mCherryΔpi[sdg-1(Δ)]::sdg-1 3’ UTR (Fig. 7f) but loss of rde-4 perturbed expression such that one isolate showed loss of expression while the other showed enhanced expression (Fig. 7g). These results suggest that while the sdg-1 open reading frame is required for exclusion from germline nuclei, it is not required for the response of the sdg-1 gene to changes in intracellular dsRNA. Together, these observations on the levels and localizations of SDG-1 raise the possibility that this protein is actively regulated by extracellular and intracellular dsRNA with a role in heritable RNA silencing, potentially with Z granules, and additional cell-cycle coupled roles during early development, potentially through nucleocytoplasmic shuttling.

Discussion

Our analyses suggest a model for the impact of dsRNA from parental circulation on descendants (Fig. 7h). Extracellular dsRNA can accumulate in parental circulation through regulated secretion from neurons (e.g., oxidative damage promotes accumulation) and potentially other tissues. Uptake into the germline is RME-2-dependent for some forms of dsRNA (blue in Fig. 7h), but RME-2-independent for other forms of dsRNA (grey in Fig. 7h). While all forms of dsRNA require SID-1 for the entry of dsRNA into the cytosol in progeny, RME-2-dependent dsRNA also require SID-5. Such dsRNA from parental circulation along with other intracellular dsRNA are processed with the help of the dsRNA-binding protein RDE-4 in progeny (Fig. 7h, bottom). This processing of dsRNA regulates the SID-1-dependent gene sdg-1 by reducing variation between animals in the levels of sdg-1 mRNA, although the dsRNA sequences likely need not match sdg-1 because no such dsRNA was reported in the published dsRNAome⁷² and because this regulation can occur in the absence of the sdg-1 open reading frame. The sdg-1 gene is located within a retrotransposon that is also targeted by heritable RNA silencing, but the SDG-1 protein colocalizes with perinuclear granules called Z granules, which are required for heritable RNA silencing. Consistent with the SDG-1 protein promoting heritable RNA silencing, the sdg-1::mCherryΔpi gene is highly susceptible to mating-induced silencing, potentially owing to the ∼16-fold higher levels of SDG-1::mCherry compared with that of SDG-1 in wild-type animals. In agreement with this proposal, SID-1 limits heritable RNA silencing because loss of SID-1 enhances mating-induced silencing (Fig. 7b, top). Since the sdg-1 gene is located within a retrotransposon that is targeted by heritable RNA silencing, this mechanism for regulating the regulators of heritable RNA silencing such as SDG-1 (and potentially other SDGs) reveals a strategy for tuning an autoregulatory loop for heritable RNA silencing by using competing dsRNA processing. Intriguingly, SDG-1 becomes enriched within nuclei upon nuclear envelope breakdown during fertilization and during early cell divisions in embryos (Fig. 7h, bottom right) with active exclusion from nuclei after each reformation of the nuclear envelope, suggestive of additional roles for this retrotransposon-encoded protein.

While the physiological conditions that promote secretion of dsRNA are not known, the discovery that oxidative damage of neurons can enhance the secretion of dsRNA suggests that disruption of cell structures by oxidative damage (e.g., membrane integrity) or initiation of cellular processes that repair oxidative damage (e.g., through ejection of damaged macromolecules⁷³) also promote the release of dsRNA. Pathologies of the central nervous system in humans, including cancer, stroke, multiple sclerosis, neurodegenerative disease, and brain injury, have been associated with extracellular RNAs detected in circulation (reviewed in ref.⁷⁴), although their origins and regulatory consequences, if any, remain unknown. The gene regulatory effects of neuronal dsRNA released upon oxidative damage of neurons provide convenient readouts that can be analyzed to understand neuronal damage and its consequences in animals.

The trafficking of extracellular dsRNA from parent to progeny has spatial specificity, as evidenced by more silencing within the proximal germline (Fig. 2), temporal specificity, as evidenced by the need for dsRNA beyond the fourth larval stage^28,29 (Fig. 1), and substrate specificity, as evidenced by the differential requirements for 50-bp dsRNA with 5’ OH versus a mix of longer dsRNAs with 5’ triphosphates (Fig. 3). One possible explanation for these constraints could be that proteins mediating dsRNA transport or subsequent silencing differ in their availability during development and in their affinities for different substrates. For example, SID-1, which was not detected in the developing larval germline but was detected in the adult germline (Fig. 4), has an extracellular domain that binds dsRNA⁷⁵ and could prefer dsRNA molecules with 5’ phosphates and/or long dsRNA. Although the selectivity uncovered here could apply to all dsRNA delivered into the extracellular space of C. elegans from any source (see Extended Data Table 1 for constraints on intergenerational dsRNA transport), the chemistry of the delivered dsRNA could be modified by yet unidentified enzymes in vivo to overcome these requirements. Tracking labeled dsRNA with diverse chemistries from parental circulation to progeny could allow correlation of differences observed in progeny silencing to differences in intergenerational trafficking. The germline is a major site of dsRNA import in C. elegans as evidenced by three key observations: the expression of SID-1 in the germline (Fig. 4), heritable misregulation of germline genes in sid-1(-) animals (Fig. 5 and Fig. 6), and accumulation of fluorescently-labeled dsRNA from the extracellular space in the germline^28,29. As a result, sid-1(-) animals could have a defect in the germline that is detectable only under conditions that promote dsRNA transport (e.g., oxidative damage). Multiple physiological defects in the germline and soma of sid-1(-) animals have been reported but have not been widely reproduced, have only been characterized within single generations, or have not been attributed to any specific sid-1-dependent gene(s). These include defects in animals with some misfolded proteins in the endoplasmic reticulum⁷⁶, in animals exiting the dauer stage^77,78, in animals exposed to pathogenic P. aeruginosa^79,80,81, in animals exposed to odor⁸², in intestinal cells that develop in the presence of a multi-copy transgene⁸³, and in animals that overexpress α-synuclein⁸⁴. RNA-seq experiments in this study and comparisons to those of previous studies suggest that genetic background-dependent changes can obscure genuine sid-1-dependent changes (see Extended Data Fig. 7 and Technical comments in Methods), raising caution in the interpretation of putative sid-1-dependent defects. Comparing multiple sid-1 mutants generated using genome editing with animals in which the mutated sequence has been reverted to wild-type sequence in the same genetic background could provide a firmer basis for the identification of additional sid-1-dependent processes.

Genes expressed within the germline are likely regulated by positive feedback loops that are required to continually produce factors for maintaining germline immortality and for preserving form and function across generations^85,86. Thus, germline genes could be particularly vulnerable to heritable epigenetic changes, where deviations in the expression levels of a gene that is regulated by or is part of such feedback loops has the potential to become permanent in descendants. Our analysis of sdg-1 expression suggests that it is part of a regulatory architecture that is susceptible to heritable epigenetic changes through the perturbation of RNA regulation (Fig. 5, Fig. 6, and Fig. 7). Such architectures within the germline could be exploited by ‘selfish’ genetic elements such as retrotransposons to persist across evolution if one of these elements also include genes encoding a regulator. In support of a wider use of such a strategy, a paralog of SDG-1, ZK262.8 (Extended Data Fig. 8b), is also encoded by a gene located within a retrotransposon and its loss along with that of the miRNA-associated Argonaute ALG-2 was reported to be synthetic lethal⁸⁷. To buffer against heritable epigenetic changes, levels of gene expression would need to be maintained within a particular range for a certain regulatory context. Given the association of SDG-1 protein with germ granule components ZSP-1/PID-2 and DEPS-1, and the maintenance of heritable changes in sdg-1::mCherryΔpi expression by DEPS-1, buffering against changes in gene expression could involve both RNA- and protein-based regulation that tunes the function of perinuclear granules. We therefore speculate that one role for extracellular RNAs that enter germ cells in other systems (e.g., tRNA fragments in mammals^5,6,8) could be to similarly buffer against heritable changes in gene expression.

Acknowledgements

We thank Sindhuja Devanapally for data on silencing by neuronal dsRNA in hrde-1(-) animals and mating-induced silencing in sid-1(-) animals; Daphne Knudsen for the generation of mut-16(jam148[nonsense]); Mary Chey, Samiha Tasnim, and Daphne Knudsen for comments on the manuscript; the Caenorhabditis elegans Genetic Stock Center, the Seydoux laboratory (Johns Hopkins University) and the Hunter laboratory (Harvard University) for strains; Quentin Gaudry for help in creating our optogenetics apparatus; the Andrews laboratory for use of their Nikon Eclipse Ti spinning disk confocal microscope; Amy Beaven and the Imaging Core Facility for temporary use of a Leica TCS SP8 DLS microscope with HyVolution and the Zeiss LSM 980 AiryScan 2 Laser Scanning Confocal microscope (supported by grant 1S10OD025223-01A1 from the NIH); Lanelle Edwards, Rex Ledesma, Carlos Machado, and Omega Bioservices for help with RNA sequencing and analysis. Funding: This work was supported by UMD CMNS Dean’s Matching Award for “Training Program in Cell and Molecular Biology” T32GM080201 to N.S. and in part by National Institutes of Health Grants R01GM111457 and R01GM124356, and National Science Foundation Grant 2120895 to A.M.J.

Author contributions

N.S., A.S., and A.M.J designed the research. N.S., A.S., A.L.Y., W.M.C., J.A.M., and A.M.J. performed all experiments, and collected and analyzed data. N.S., A.S., and A.M.J. prepared the manuscript with contributions from all authors.

Data Availability

Source data are provided with this paper. All data are available in the manuscript, the supplementary materials, or on figshare. RNA-seq data has been deposited to Gene Expression Omnibus (GEO) with the accession number GSE185385.

Code Availability

All code is available within the manuscript, the supplementary materials, or at AntonyJose-Lab/Shugarts_et_al_2023 on github.

Declaration of interests

The authors declare no competing interests.

Methods

Strains and oligonucleotides

All strains (listed in Extended Data Table 3) were cultured on Nematode Growth Medium (NGM) plates seeded with 100 μl of OP50 E. coli at 20°C and strains made through mating were generated using standard methods⁸⁸. Oligonucleotides used are in Extended Data Table 4 (for genotyping sid-1(qt9): P1-P2, ttTi5605: P3-P5, eri-1(mg366): P6-P7, sid-1(tm2700): P8-P10, hrde-1(tm1200): P11-P13, nrde-3(tm1116): P14-P16, and rde-4(ne301): P156-P157). Strains made through mating existing mutant strains and genotyping using the above primers are listed below.

To create gtbp-1::gfp animals with hrde-1(tm1200) in the background

AMJ577²⁵ was crossed with JH3197 males to create AMJ1220 and one other independent isolate.

To create gtbp-1::gfp animals with nrde-3(tm1116) in the background

JH3197 was crossed with WM156 males to create AMJ1383.

Transgenesis

Animals were transformed with plasmids and/or PCR products using microinjection⁸⁹ to generate extrachromosomal arrays or single-copy transgenes. All plasmids were purified from bacterial culture using QIAprep Spin Miniprep Kit (Qiagen) and all PCR products were generated with Phusion^® High-Fidelity DNA Polymerase (New England BioLabs) and purified using NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel).

To express sid-1::DsRed in the muscle from an integrated array

pAJ53a (myo-3p::sid1::DsRed::unc-54 3’UTR, made by AMJ while in Hunter Lab, Harvard University) was generated by amplifying part of sid-1 cDNA from pHC355²³ with primers P27 and P18, DsRed and unc-54 3’UTR from pHC183¹⁴ with primers P17 and P30, fusing the fragments using PCR with primers P30 and P31, and then cloning the fusion product into the pHC355 vector backbone using the restriction enzymes NruI and EagI. pAJ53a (40 ng/μl) was then injected into HC196 and animals expressing DsRed were isolated. AMJ3 was isolated as a spontaneous integrant. AMJ3 males were then crossed with AMJ308 hermaphrodites to generate AMJ327.

To express sid-1::DsRed in the germline from a single-copy transgene: The mex-5

promoter was amplified from pJA252 (Addgene #21512) using the primers P19 and P20. The sid-1 gene was amplified from N2 genomic DNA using the primers P21 and P22. The DsRed gene was amplified from pAJ53a (myo-3p::sid-1(+)::DsRed::unc-54 3’UTR; made by AMJ and Tessa Kaplan while in Hunter Lab, Harvard University) using the primers P23 and P24. The sid-1 3’UTR was amplified using the primers P25 and P26. Using NEBuilder^® HiFi DNA Assembly (New England BioLabs), these four amplicons were placed into pCFJ151 (Addgene #19330) digested with AflII (New England BioLabs) and treated with CIP (New England BioLabs) to generate pJM10. pJM10 (50 ng/μl) and the coinjection markers pCFJ601 (50 ng/μl), pMA122 (10 ng/μl), pGH8 (10 ng/μl), pCFJ90 (2.5 ng/μl), and pCFJ104 (5 ng/μl) (plasmids described in ref.⁹⁰) were injected into the germline of adult EG4322 animals. One transgenic line was isolated s described previously⁹⁰ and crossed with HC196 males to generate AMJ576. The integration of mex-5p::sid-1(+)::DsRed::sid-1 3’UTR in AMJ576 was verified by genotyping with primers P3-P5 and Sanger sequencing of the insertion.

To express sid-1::gfp in the muscle from an extrachromosomal array

pTK2 (myo-3p::sid-1::gfp, made by AMJ and Tessa Kaplan while in Hunter Lab, Harvard University) was constructed by amplifying part of sid-1 cDNA from pHC355²³ with primers P27 and P28, gfp and unc-54 3’UTR from pPD95.75 (Addgene #1494) using primers P29 and P30, and then fusing the fragments using PCR with primers P30 and P31 and cloning the product into the pHC355 vector backbone using the restriction enzymes NruI and EagI. pTK2 (10 ng/μl) was injected into HC196 and animals expressing GFP were isolated as AMJ706.

To express PH::miniSOG in neurons from an extrachromosomal array

pNMS03 (rgef-1p::PH::miniSOG::unc-54 3’UTR) was generated by amplifying the vector backbone of pHC337 excluding the gfp-dsRNA hairpin sequence using primers P35 and P36, and assembling it with PH::miniSOG(Q103L) amplified from pCZGY2851 (gift from Andrew Chisholm) with primers P33 and P34 using NEBuilder^® HiFi DNA Assembly (New England BioLabs). pNMS03 (40 ng/μl) was injected into N2 animals with pHC448⁹¹ (myo-2p::DsRed2::unc-54 3’UTR; 40 ng/μl) as a coinjection marker to create AMJ837 and two other isolates.

pNMS03 (40 ng/μl) was also injected into N2 animals with PCR products forming rgef-1p::DsRed (40 ng/μl) generated previously⁹¹ as a coinjection marker to create AMJ936 and two other isolates.

To express PH::miniSOG in neurons from a single-copy transgene

pNMS05 (rgef-1p::PH::miniSOG::unc-54 3’UTR with ttTi5605 homology arms and Cbr-unc-119(+)) was generated by amplifying the transgene rgef-1p::PH::miniSOG::unc-54 3’UTR from pNMS03 with primers P37 and P38 containing AvrII restriction sites and cloning the fragment into pCFJ151 after AvrII (New England BioLabs) digestion. pNMS05 (50 ng/μl) and the coinjection markers pCFJ601 (50 ng/μl), pMA122 (10 ng/μl), pGH8 (10 ng/μl), pCFJ90 (2.5 ng/μl), and pCFJ104 (5 ng/μl) (plasmids described in ref.⁹⁰) were injected into the germline of adult EG4322 animals. One transgenic line was isolated as described previously⁹⁰ and designated as AMJ1019. The integration of rgef-1p::PH::miniSOG::unc-54 3’UTR in AMJ1019 was verified by genotyping with primers P3-P5 and Sanger sequencing of the insertion.

To express PH::miniSOG with bli-1-dsRNA in neurons from an extrachromosomal array

pNMS03 (40 ng/μl) was injected with rgef-1p::bli-1-sense (40 ng/μl) and rgef-1p::bli-1-antisense (40 ng/μl) PCR products generated previously⁹² into GR1373 animals with pHC448 (myo-2p::DsRed2::unc-54 3’UTR) as a coinjection marker (40 ng/μl) to create AMJ1007 and one other independent isolate. AMJ1007 was crossed with HC731 males to create AMJ1108 and two other isolates. AMJ1108 was crossed with HC196 males to create AMJ1114 and one other isolate. AMJ1007 was crossed with N2 males to create AMJ1123 and one other isolate.AMJ1123 males were crossed with 3X outcrossed FX02700 (designated as AMJ1153) to create AMJ1151 and two other isolates. AMJ1151 was crossed with GR1373 males to create AMJ1173 and two other isolates.

To express PH::miniSOG with gfp-dsRNA in neurons from an extrachromosomal array

pNMS03 (40 ng/μl) and pHC337 (rgef-1p::gfp-dsRNA::unc-54 3’UTR; 40 ng/μl) were injected into AMJ819⁴³ with pHC448 (myo-2p::DsRed2::unc-54 3’UTR; 40 ng/μl) as a coinjection marker to create AMJ1009 and one other independent isolate. AMJ1009 was crossed with N2 males to create AMJ1134. AMJ1159 was crossed with AMJ1134 males to create AMJ1312 and two other isolates.

All other transgenes were generated previously (ccIs4251¹³; oxSi487⁹⁰; tmIs1005⁹³; jamEx140²⁵; qtEx136²⁴).

Cas9-mediated genome editing

Genome editing was performed by injecting nuclear-localized Cas9 (PNA Bio) preincubated at 37°C for 10 min with either a single-guide RNA (sgRNA) generated by in vitro transcription (SP6 RNA Polymerase, New England BioLabs) or hybridized crRNA/tracrRNA (IDT), as well as an oligonucleotide or PCR-amplified homology repair template, into the C. elegans distal gonad. Screening for plates with successfully edited F1 animals was performed using either dpy-10 co-CRISPR^94,95 or the pRF4 plasmid used as a co-injection marker⁹⁶. All plasmids were purified from bacterial culture using QIAprep Spin Miniprep Kit (Qiagen) and all PCR products were generated with Phusion^® High-Fidelity DNA Polymerase (New England BioLabs) and purified using NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel). Alleles generated by genome editing are schematized in Fig. 4a (sid-1), Fig. 5a (sid-1), Extended Data Fig. 1d (rme-2), Extended Data Fig. 3 (deps-1, mut-16, sid-2, rme-2, sid-1, rde-1, sid-5, and sid-3), Extended Data Fig. 6a (sid-1), and Extended Data Fig. 8c (W09B7.2/F07B7.2 (sdg-1)), and oligonucleotides used are in Extended Data Table 4.

To delete the rme-2 coding sequence

Two sgRNAs targeting the start and end of the rme-2 coding sequence were in vitro transcribed from a SP6 transcription template amplified from pDD162 (Addgene #47549) using primers P42 (start sgRNA) or P43 (end sgRNA) as forward primers and P40 as a universal reverse primer. An sgRNA targeting dpy-10 for co-CRISPR was also in vitro transcribed using a similar template amplified from pDD162 with primers P39 and P40. All sgRNAs were purified using organic extraction, were precipitated using ethanol, and resuspended in water prior to injection. Injection into HC196 with all sgRNAs, Cas9 and the homology repair templates for rme-2 (P44) and dpy-10 (P41), and screening for edited animals were performed as described above. Genotyping for rme-2(deletion) was performed using a triplex PCR with primers P45-P47 to isolate AMJ1120 and one other isolate and the rme-2 deletion was verified using Sanger sequencing. AMJ1120 was crossed with N2 males to isolate AMJ1131. AMJ1131 males were crossed with EG6787 and AMJ471 hermaphrodites to isolate AMJ1146 and AMJ1204 animals, respectively.

To delete the sid-1 coding sequence

Injection of crRNAs targeting the start (P59) and end (P52) of the sid-1 coding sequence (IDT), tracrRNA, Cas9, a sid-1(deletion) homology repair template (P60) and pRF4 into N2 and AMJ1372, and subsequent screening were performed as described above. Genotyping for sid-1(deletion) was performed using triplex PCR with primers P8, P54 and P55 to isolate AMJ1324 and one other independent isolate from N2 and AMJ1479-AMJ1482 from AMJ1372. The sid-1 deletion was verified by Sanger sequencing in all isolates. AMJ1324 was crossed with AMJ1131 males to create AMJ1366.

To delete the sid-2 coding sequence

Injection of crRNAs targeting the start (P71) and end (P72) of the sid-2 coding sequence (IDT), tracrRNA, Cas9, a sid-2(deletion) homology repair template (P73) and pRF4 into N2, and subsequent screening were performed as described above. Genotyping for sid-2(deletion) was performed using triplex PCR with primers P74-P76 to isolate AMJ1368 and one other independent isolate. The sid-2 deletion was verified by Sanger sequencing in both isolates. AMJ1368 was crossed with AMJ1324 males to create AMJ1380.

To delete the sid-5 coding sequence

Injection of crRNAs targeting the start (P61) and end (P62) of the sid-5 coding sequence (IDT), tracrRNA, Cas9, a sid-5(deletion) homology repair template (P63) and pRF4 into N2, and subsequent screening were performed as described above. Genotyping for sid-5(deletion) was performed using duplex PCR with primers P64 and P65 to isolate AMJ1332 and three other independent isolates. The sid-5 deletion was verified by Sanger sequencing in all four isolates. AMJ1332 was crossed with AMJ1324 males to create AMJ1367.

To introduce a nonsense mutation into sid-1 coding sequence

An sgRNA was designed to introduce into sid-1 a nonsense mutation mimicking the qt9 allele¹⁴. This sgRNA was in vitro transcribed from a SP6 transcription template amplified from pDD162 (Addgene #47549) using primers P48 and P40. An sgRNA targeting dpy-10 for co-CRISPR was also in vitro transcribed using a similar template amplified from pDD162 with primers P39 and P40. Both sgRNAs were purified using organic extraction and were precipitated using ethanol prior to injection. Both sgRNAs, Cas9 and the homology repair templates for sid-1(nonsense) (P49) that includes a missense mutation (S155P) and nonsense mutation (R156*) downstream that would prevent recutting of edited DNA by Cas9, and for dpy-10 (P41) were injected into N2. Screening for edited animals was performed as described above. Genotyping for sid-1(nonsense) was performed using a duplex PCR with primers P1 and P2 followed by restriction digestion with HpyCH4V to isolate AMJ1159. The nonsense mutation was confirmed using Sanger sequencing. AMJ1159 males were crossed with AMJ581²⁵ to create AMJ1504 and two other independent isolates.

Injection of a crRNA with the same target sequence (P88) (IDT) as the sgRNA described above, tracrRNA, Cas9, the same sid-1(nonsense) homology repair template (P49) and pRF4 into N2 and AMJ1372 and subsequent screening were performed as described above. Genotyping for sid-1(nonsense) was performed using duplex PCR with primers P1 and P2 followed by restriction digestion with HpyCH4V to distinguish AMJ1399 from N2, and AMJ1389 and AMJ1442-AMJ1446 from AMJ1372. The nonsense mutation was verified using Sanger sequencing in all isolates.

To revert the mutation in sid-1(nonsense) animals

An sgRNA was designed to revert the nonsense mutation described above back to wild-type sid-1 sequence. The sgRNA was in vitro transcribed from a SP6 transcription template amplified from pDD162 (Addgene #47549) using primers P50 and P40. An sgRNA targeting dpy-10 for co-CRISPR was also in vitro transcribed using a similar template amplified from pDD162 with primers P39 and P40. Both sgRNAs were purified using organic extraction and were precipitated using ethanol prior to injection. Injection into AMJ1159 with both sgRNAs, Cas9 and the homology repair template for sid-1(revertant) (P51), which also reverted the missense mutation (S155P) and nonsense mutation (R156*) downstream of sid-1(nonsense) to wild-type sequence, and dpy-10 (P41). Screening for edited animals was performed as described above. Genotyping for sid-1(revertant) was performed using duplex PCR with primers P1 and P2 followed by restriction digestion with HpyCH4V to isolate AMJ1217 and two other independent isolates. The revertant was verified using Sanger sequencing in all isolates.

Injection of a crRNA with the same target sequence (P93) (IDT) as the sgRNA described above, tracrRNA, Cas9, a sid-1(revertant) homology repair template (P51) and pRF4 into AMJ1389 and AMJ1399, and subsequent screening were performed as described above.

Genotyping for sid-1(revertant) was performed using duplex PCR with primers P1 and P2 followed by restriction digestion with HpyCH4V to distinguish AMJ1412 and AMJ1413 from AMJ1389, and AMJ1405-AMJ1410 from AMJ1399. The revertant was verified using Sanger sequencing in all isolates.

To tag W09B7.2/F07B7.2 with mCherry

Injection of a crRNA with the target sequence listed as P80 (IDT), tracrRNA, Cas9, an mCherry sequence lacking piRNA binding sites amplified using primers P81 and P82 from pSD6⁴³ as a homology repair template with homology arms to the C-terminus of W09B7.2/F07B7.2, and pRF4 into N2, and subsequent screening were performed as described above. Genotyping for identical tags W09B7.2::mCherryΔpi and F07B7.2::mCherryΔpi in isolate AMJ1372 was performed using triplex PCR with primers P79, P83 and P84. Tagging of both loci is evident in Extended Data Fig. 8d. The mCherryΔpi insertion was verified by Sanger sequencing. AMJ1372 hermaphrodites and males generated by heatshock were mated with N2 males and hermaphrodites, respectively, to examine expression in cross progeny and in homozygosed wild-type and W09B7.2/F07B7.2(jam137[mCherryΔpi]) animals across generations in six independent F1 lineages from each cross. See Fig. 6c and Extended Data Fig. 9a for associated data. YY916 males were crossed with AMJ1372 to generate AMJ1662. The 3xflag::gfp::znfx-1 locus was genotyped using primers P153, P154, and P155.

To introduce a nonsense mutation into rde-1 coding sequence

Injection of a crRNA with the target sequence listed as P94 (IDT), tracrRNA, Cas9, a rde-1(nonsense) homology repair template (P95) mimicking rde-1(ne300)⁹⁷, and pRF4 into AMJ1372 and subsequent screening were performed as described above. Genotyping for rde-1(nonsense) was performed using duplex PCR with primers P96 and P97 and restriction digestion with NlaIII to isolate AMJ1447 and AMJ1448. The nonsense mutation was verified by Sanger sequencing for all isolates.

To tag sid-1 with wrmScarlet at the 3’ end

Injection of a crRNA with the target sequence listed as P52 (IDT), tracrRNA, Cas9, a sid-1::wrmScarlet13 homology repair template with the beginning (1) and end (3) but not the middle (2) of the coding sequence⁹⁸ (P53), and pRF4 into N2 and subsequent screening were performed as described above. Genotyping for wrmScarlet13 was performed using duplex PCR with primers P54 and P55 to isolate AMJ1280. The wrmScarlet13 insertion was verified by Sanger sequencing. Subsequent injections were performed into AMJ1280 with a wrmScarlet13 specific crRNA with the target sequence listed as P56 (IDT), a complete wrmScarlet coding sequence amplified from pSEM89 (made in Boulin Lab – gift from Kevin O’Connell) with primers P57 and P58 and the same components as described previously. After similar screening, genotyping for full wrmScarlet insertion was performed using duplex PCR with primers P54 and P55 to isolate AMJ1282 and one other independent isolate. The full wrmScarlet insertion was verified by Sanger sequencing. AMJ1282 was crossed with AMJ577 males to create AMJ1365.

To tag rme-2 with wrmScarlet at the 3’ end

Injection of a crRNA with the target sequence listed as P67 (IDT), tracrRNA, Cas9, a rme-2::wrmScarlet13 homology repair template with the beginning (1) and end (3) but not the middle (2) of the coding sequence⁹⁸ (P69), and pRF4 into N2 and subsequent screening were performed as described above. Genotyping for wrmScarlet13 was performed using duplex PCR with primers P70 and P47 to isolate AMJ1281. The wrmScarlet13 insertion was verified by Sanger sequencing. Subsequent injections were performed into AMJ1281 with a wrmScarlet13 specific crRNA with the target sequence listed as P77 (IDT), a complete wrmScarlet coding sequence amplified from pSEM89 (made in Boulin Lab – gift from Kevin O’Connell) with primers P57 and P58 and the same components as described previously. After similar screening, genotyping for full wrmScarlet insertion was performed using duplex PCR with primers P54 and P55 to isolate AMJ1284 and two other independent isolates. The full wrmScarlet insertion was verified by Sanger sequencing.

To tag sid-1 internally with mCherry

Injection of a crRNA with the target sequence listed as P110 (IDT), tracrRNA, Cas9, an mCherry lacking piRNA binding sites amplified from pSD6⁴³ as a homology repair template with homology arms to exon 4 of sid-1 with primers P111 and P112, and pRF4 into N2 and subsequent screening were performed as described above. Genotyping for mCherryΔpi was performed using triplex PCR with primers P113, P114 and P79 to isolate AMJ1438 and one other isolate from the same lineage. The mCherryΔpi insertion was verified by Sanger sequencing. Subsequent injections were performed into AMJ1438 with a crRNA targeting the 5’-end of mCherryΔpi (P115) (IDT), a homology repair template containing a 45-nt linker sequence (P116) and the same components as described previously. After similar screening, genotyping for the linker insertion was performed using duplex PCR with primers P113 and P79 to isolate AMJ1485 and two other independent isolates. Insertion of the linker was verified by Sanger sequencing in all three isolates.

To introduce a nonsense mutation into sid-3 coding sequence

Injection of a crRNA withthe target sequence listed as P66 (IDT), tracrRNA, Cas9, a sid-3(nonsense) homology repair template (P85) mimicking sid-3(qt31)⁹⁹ and pRF4 into AMJ1372 and subsequent screening were performed as described above. Genotyping for sid-3(nonsense) was performed using duplex PCR with primers P86 and P87, and restriction digestion with StyI to isolate AMJ1449 and AMJ1450. The nonsense mutation was verified by Sanger sequencing for both isolates.

To introduce a nonsense mutation into deps-1 coding sequence

Injection of a crRNA with the target sequence listed as P68 (IDT), tracrRNA, Cas9, a deps-1(nonsense) homology repair template (P137) mimicking deps-1(bn124)⁴⁹ and pRF4 into AMJ1372 and AMJ1412 and subsequent screening were performed as described above. Genotyping for deps-1(nonsense) was performed using allele specific PCR with primers P138 and P139 amplifying the wild-type sequence and primers P140 and P141 amplifying the mutant allele to isolate AMJ1451-AMJ1452 from AMJ1372 and AMJ1574-AMJ1575 from AMJ1412. The nonsense mutation was verified by Sanger sequencing for both isolates.

To insert the tetracycline K4 aptazyme into the 3’UTR of sid-1

njection of a crRNA with the target sequence listed as P52 (IDT), tracrRNA, Cas9, a sid-1::tetracycline-K4-aptazyme homology repair template (P78) and pRF4 into N2 and subsequent screening were performed as described above. Genotyping for insertion of the aptazyme sequence was performed using duplex PCR with primers P54 and P55 to isolate AMJ1323. The aptazyme insertion was verified by Sanger sequencing. AMJ1323 hermaphrodites was crossed with AMJ477²⁴ males to create AMJ1330 and with AMJ471²⁵ males to create AMJ1350. AMJ1323 males were crossed with JH3197 to create AMJ1355.

To insert the tetracycline K4 aptazyme into the 3’UTR of gtbp-1(ax2053[gtbp-1::gfp])

Injection of a crRNA with the target sequence listed as P89 (IDT), tracrRNA, Cas9, a gtbp-1::gfp::tetracycline-K4-aptazyme homology repair template (P90) and pRF4 into JH3197 and subsequent screening were performed as described above. Genotyping for insertion of the aptazyme sequence was performed using duplex PCR with primers P91 and P92 to isolate AMJ1542. The aptazyme insertion was verified by Sanger sequencing.

To introduce a missense mutation into dpy-10 coding sequence

Injection of crRNA with the target sequence listed as P142 (IDT), tracrRNA, Cas9, and a dpy-10(mis) homology repair template (P41) mimicking dpy-10(cn64)¹⁰⁰ into AMJ1372 was performed as described above and heterozygous F1 animals were screened for by passaging “rolling” animals. Animals that appeared wild-type and those that appeared Dpy (homozygous dpy-10(-)) were isolated from three independently edited F1 animals. See Extended Data Fig. 9b for associated data.

To delete the W09B7.2/F07B7.2 coding sequence

Injection of crRNAs targeting the start (P143) and end (P144) of the W09B7.2/F07B7.2 coding sequence (IDT), tracrRNA, Cas9, a W09B7.2/F07B7.2(deletion) homology repair template (P145) and pRF4 into N2, and subsequent screening were performed as described above. Genotyping for W09B7.2/F07B7.2(deletion) was performed using triplex PCR with primers P146-P148 to isolate AMJ1577, AMJ1612, and AMJ1613. Deletion of both W09B7.2/F07B7.2 loci was verified by absence of wild-type band by PCR (see Extended Data Fig. 8d) and Sanger sequencing in all three isolates.

To delete the W09B7.2/F07B7.2 coding sequence from W09B7.2/F07B7.2::mCherryΔpi

Injection of crRNAs targeting the start (P143) and end (P149) of the W09B7.2/F07B7.2 coding sequence (IDT), tracrRNA, Cas9, a W09B7.2/F07B7.2(deletion) homology repair template (P150) and pRF4 into AMJ1372, and subsequent screening were performed as described above. Genotyping for W09B7.2/F07B7.2(deletion) was performed using triplex PCR with primers P148, P151, and P152 to isolate AMJ1615, AMJ1616, and AMJ1617. Deletion of both W09B7.2/F07B7.2 loci was verified by absence of wild-type band by PCR (see Extended Data Fig. 8d) and Sanger sequencing in all three isolates. AMJ1615 was outcrossed 1X with N2 males to generate AMJ1770. Genotyping was performed using primers P158-P160 and expression was verified by widefield microscopy. AMJ1615 was crossed with WM49 males to generate AMJ1766 and AMJ1767. Many of the AMJ1770 and AMJ1766 animals had defective germline morphology and therefore only animals with apparently normal morphology were selected for quantification of sdg-1p::mCherryΔpi expression. Genotyping for sdg-1p::mCherryΔpi was performed using the same primers (P158-P160) and initial expression was verified by widefield microscopy.

Light-induced damage of neurons

Optimizing duration of light exposure

20-30 animals expressing PH::miniSOG in neurons (multi copy, AMJ837; single copy, AMJ1019) were placed on an unseeded NGM plate and exposed to blue light (470 nm wavelength) at a distance of approximately 7.5 cm from an LED (Cree Xlamp XP-E2 Color High Power LED Star – Single 1 UP, LED supply) producing light at a power of ∼2 mW/mm² flashing at a frequency of 2 Hz for different durations of time. Animals were then scored for movement defects immediately after light exposure, OP50 was seeded onto the plate, and animals were scored again 24 h post light exposure (Extended Data Fig. 2a). Wild-type animals were exposed to blue light for the same durations as control. Representative widefield images of unparalyzed (wild type) and paralyzed (coiled, AMJ837) animals were taken using a Nikon AZ100 microscope and Photometrics Cool SNAP HQ² camera (Extended Data Fig. 2b, top). Confocal images of animals expressing PH::miniSOG and DsRed in neurons (AMJ936) with and without 30 minutes of blue light exposure were taken using a Leica TCS SP8 DLS microscope with HyVolution using a 40X oil objective lens. DsRed was excited using a 638 nm laser and fluorescence was collected through a 598 nm emission filter (Extended Data Fig. 2b, bottom). Images were adjusted for display using Fiji¹⁰¹ (NIH).

Silencing by bli-1-dsRNA

Five L4 animals with an extrachromosomal array expressing PH::miniSOG and bli-1-dsRNA in neurons were placed on seeded NGM plates and allowed to lay progeny for 24 h. P0 animals were then removed and F1 progeny were exposed to blue light as described above for 60 min at different time points after initial P0 L4 animals were passaged. 96 h post light exposure F1 progeny with the array were scored for bli-1 silencing (presence of blisters) in gravid adults (Extended Data Fig. 2c, top, d,e).

Silencing by gfp-dsRNA

L4 animals with the mex-5p::mCherry::h2b::gfp::h2b transgene (oxSi487) (Fig. 2b) were mated with L4 male animals with an extrachromosomal array expressing PH::miniSOG and gfp-dsRNA in neurons (Fig. 2a). After 36 h of mating and laying progeny, P0 animals were removed from plates and F1 progeny were exposed to blue light as described above for 60 min at different time points after initial P0 L4 animals were mated. 96 h after mating, F1 cross progeny hermaphrodites with the array were imaged as adults (Extended Data Fig. 2c, bottom) under a coverslip in 10 μl of 3 mM levamisole on a 2% agar pad using a Nikon AZ100 microscope and Photometrics Cool SNAP HQ² camera. A C-HGFI Intensilight Hg Illuminator was used to excite GFP (filter cube: 450 to 490 nm excitation, 495 dichroic, and 500 to 550 nm emission) and mCherry (filter cube: 530 to 560 nm excitation, 570 dichroic, and 590 to 650 nm emission). Animals were scored as bright if fluorescence was easily detectable without adjusting levels, dim if fluorescence could be observed after level was adjusted to saturation, and not detectable if fluorescence was still not observed after level adjustments (Extended Data Fig. 2f). Representative images were adjusted in Adobe Photoshop to identical levels for presentation (Fig. 2b-d).

Sensitive northern blotting

Double-stranded RNA was in vitro transcribed from a PCR amplicon using T7 RNA Polymerase (New England BioLabs) (Extended Data Fig. 4d) or expressed in HT115 E. coli after IPTG induction during exponential growth (Extended Data Fig. 4b,c) and extracted using TRIzol (Fisher Scientific). RNA was then separated by size using fully denaturing formaldehyde polyacrylamide gel electrophoresis (FDF-PAGE)¹⁰² wherein 10 μg RNA samples were heated with formaldehyde to disrupt dsRNA duplexes and run on a 4% denaturing polyacrylamide gel next to 1-kb and 100-bp DNA ladders for size comparison. After migration, the ladder lanes were stained with ethidium bromide and imaged, and the RNA lanes were transferred to a positively charged nitrocellulose membrane using a Trans-Blot^® Turbo^TM Transfer System (Bio-Rad) and crosslinked using 120 mJ/cm² UV radiation. Blots were then exposed to 2.5 pmol of 40-nt HPLC purified DNA oligonucleotides conjugated to digoxigenin (DIG) using the DIG Oligonucleotide Tailing Kit (Roche) hybridized to the nitrocellulose membrane at 60°C overnight (42°C for 2 h for 5S rRNA) in ULTRAhyb^TM buffer (Invitrogen^TM) to probe the sense or antisense strands of unc-22 (Extended Data Fig. 4b,d) or gfp-dsRNA (Extended Data Fig. 4c) at different positions (protocol adapted from ref.³⁶). After hybridization, the membrane was washed and blocked using the DIG Wash and Block Buffer Set (Roche), incubated with Anti-DIG-AP, Fab fragments (Roche) and developed with CSPD (Roche) at 37°C for 15 min. Chemiluminescence from the AP/CSPD reaction was imaged using a LAS-3000 (Fujifilm) or iBright^TM CL1000 (Invitrogen^TM) imager. Blots were compared to ethidium bromide-stained ladders after imaging to visualize fragment size. Blots were stripped using two washes with 5% SDS (Sigma Aldrich) and two washes with 2X SSC (Sigma Aldrich) and the hybridization, blocking and development procedures were repeated for each probe (5S RNA probe: P118; unc-22 probes: P119-P124; gfp probes: P125-P130).

Injection of dsRNA

Injection of synthetic dsRNA

RNA oligonucleotides were purchased from IDT and resuspended in IDT Duplex Buffer (unc-22: P131 and P132; gfp: P133 and P134; fluorescently labeled gfp: P135 and P136). 1 μg each of HPLC purified 50-nt sense and antisense oligonucleotide was diluted to 100-350 ng/μl with IDT Duplex Buffer at a final volume of 10 μl. Alternatively, unc-22 single-stranded RNA was treated with polynucleotide kinase and annealed in equal proportion at a final concentration of ∼97 ng/μl of unc-22-dsRNA in IDT Duplex Buffer (Extended Data Fig. 4e,f). This mixture was heated to 95°C for 1 min and cooled at a rate of 1°C/min to a final temperature of 25°C. The mix was centrifuged at 16500 x g for 20-30 min and loaded into a microinjection needle. Young adult animals were injected 24 h after the L4 stage in the body cavity just beyond the bend of the posterior gonad arm²⁸. Injected animals were recovered with M9 buffer and isolated onto NGM plates and allowed to lay progeny. In cases where animals were mated with N2 males post injection, two adult N2 males were placed on each plate with an injected hermaphrodite.

Injection of in vitro transcribed dsRNA

Templates for transcription were amplified from RNAi vectors using one common primer specific to the T7 promoter sequence (P117). PCR products were purified using column purification (Macherey-Nagel, ref. 740609.50) and subsequently used for transcription by T7 RNA Polymerase (New England BioLabs). Many transcription reactions were pooled and purified using one column to produce concentrated RNA samples. Annealing, centrifugation, and injection into the body cavity of animals staged as L4s (injected between pharynx and anterior intestine) or young adults were performed as described for synthetic dsRNA with identical concentrations unless otherwise indicated in figure legends. In cases where animals were mated with N2 males post injection, two adult N2 males were placed on each plate with an injected hermaphrodite.

Scoring of gene silencing

For scoring silencing by unc-22 dsRNA, 10-30 L4 animals were passaged into 10 μl of 3 mM levamisole and scored for twitching, observed as rapid movement of the head and/or tail (as in ref.²⁸), 3-4 days after injection for progeny of rme-2(+) parents and 4-5 days after injection for progeny of rme-2(-) parents with no appreciable difference between days in which animals were scored post injection. Weak and strong twitching were scored as in Movies S1-S3 of ref.²⁸. Numbers of silenced animals and total animals scored were summed across all days of scoring and experimental replicates.

When scoring silencing of gfp, animals were either scored by eye in comparison to animals injected with duplex buffer only (i.e. buffer; Extended Data Fig. 1b) or were mounted in 10 μl of 3 mM levamisole on a 2% agar pad and imaged under a coverslip as P0 adults (2 days post injection) or F1 L4s (3 days post P0 injection) using a Nikon AZ100 microscope and Photometrics Cool SNAP HQ² camera. A C-HGFI Intensilight Hg Illuminator was used to excite GFP (filter cube: 450 to 490 nm excitation, 495 dichroic, and 500 to 550 nm emission). Representative images for gfp expression in F1 animals after P0 injection were adjusted to identical levels in Adobe Photoshop for presentation (Fig. 3b and Extended Data Fig. 1e). See “Imaging and quantification of reporters using widefield microscopy” for other methods of scoring gfp expression after imaging.

Imaging of fluorescently labeled dsRNA

Embryos were imaged 22 hours post P0 injection with labeled dsRNA. Laid embryos were picked off plates and placed into 5 μl of 3 mM levamisole on a coverslip for at least 5 minutes before placing on a 2% agarose pad on a slide. Embryos were imaged using the Eclipse Ti Spinning Disk Confocal (Nikon) with the 60X objective lens. Atto 565 was excited using a 561 nm laser and fluorescence was collected through a 415-475 nm and 580-650 nm emission filter. Images were adjusted for display using Fiji¹⁰¹ (NIH).

Feeding RNAi

P0 and F1 feeding

E. coli (HT115) expressing dsRNA was cultured in LB media with 100 μg/μl carbenicillin overnight at 250 rpm. 100 μl of cultured bacteria was then seeded onto RNAi plates and incubated at room temperature for approximately 24 h. L4 animals were passaged onto seeded RNAi plates and progeny were scored for silencing by bacteria expressing dsRNA targeting unc-22 (twitching in levamisole), bli-1 (blisters), pos-1 (dead eggs) or expressing L4440 as an empty vector control.

P0 only feeding

RNAi bacteria were cultured and seeded as described above. L4-stage or young adult-stage (24 h post L4) animals were passaged onto seeded RNAi plates and cultured at 20°C for approximately 24 h. In some cases, fed P0 animals were then scored for silencing as described above and subsequently imaged under widefield microscopy (Fig. 7f). To score unfed progeny, fed P0 animals were picked into 1 ml of M9 buffer and washed four times to remove any residual bacteria (as in ref.²⁸). After washing, animals were resuspended in 200 μl of remaining M9 buffer and placed onto a seeded NGM plate. 1 h later, animals were isolated onto single NGM plates and their progeny were scored for silencing as described above.

Limited P0 only feeding

RNAi bacteria were cultured and seeded as described above. L4-stage animals were passaged onto seeded RNAi plates and cultured at 20°C for approximately 16 h. Animals were then passaged onto NGM plates seeded with E. coli (OP50) and cultured for 1.5 h at room temperature. Animals were then again passaged to new OP50 seeded plates (1 animal on each plate) and progeny (only L3 larvae, L4 larvae and adults) were counted after 4 days of being cultured at 20°C (∼96 hours after moving to new OP50 plates).

F1 only feeding

L4-staged animals were passaged onto RNAi plates seeded with 10 μl of E. coli (OP50). Animals were allowed to develop into adults and lay eggs over 24 h at 20°C and then removed from plates. Plates with eggs were then seeded with RNAi bacteria cultured and seeded as described above and further cultured at 20°C. Hatched progeny were imaged throughout development or as adults 3 days after being staged as L4 animals (day 3 adults).

Tetracycline-induced expression

For animals cultured with OP50 E. coli

81.6 μl of a 500 μM solution of tetracycline in water was added to 4 mL NGM plates previously seeded with OP50 E. coli (at least two days earlier) to create plates with ∼10 μM tetracycline (concentration based on ref.⁴⁴). Volumes of 166.7 μl and 444.4 μl of tetracycline solution were used to create plates with final concentrations of ∼20 μM or ∼50 μM, respectively (see Extended Data Fig. 6d). Control plates were also made by adding the same amount of water to seeded NGM plates without tetracycline. Tetracycline plates and control plates were incubated at room temperature out of direct light overnight to allow any remaining liquid to dry. Animals were passaged to tetracycline or water plates with or without previous injection of 10 μM tetracycline or water into adult gonads. Progeny expressing neuronal unc-22 or gfp-dsRNA were scored for silencing on the first day of adulthood. In the case of silencing of gtbp-1::gfp by neuronal gfp-dsRNA, animals with the array expressing gfp-dsRNA were passaged as L4s onto new tetracycline or water plates to be imaged as day 1 adults. The brood size of animals cultured on OP50 with 10 μM tetracycline or water was scored by staging single L4 animals on NGM plates with 10 μM tetracycline or water and moving animals every 24 h to new 10 μM tetracycline or water plates. Progeny laid on each of the four days were counted after growing to adulthood, continuously cultured under either condition.

For animals cultured on HT115 E. coli expressing dsRNA

Bacteria expressing bli-1-dsRNA, gfp-dsRNA, pos-1-dsRNA or L4440 control vector were cultured overnight to a maximum time of 24 hours (for gfp-dsRNA and L4440 only) and 100 μl of bacteria was seeded onto RNAi plates. Plates were incubated for 1-2 days at room temperature to allow for growth and drying of bacteria. 10 μM tetracycline or water was added to newly seeded plates as described above. After drying of tetracycline and water, P0 animals were added to plates and F1 animals were scored for silencing by bli-1-dsRNA or gfp-dsRNA as adults in the next generation. Silencing by pos-1-dsRNA was scored by measuring the brood of three L4 animals staged on a single RNAi plate with pos-1-dsRNA and 10 μM tetracycline or water. Brood size over four days was measured after moving all P0 animals every 24 h to new 10 μM tetracycline or water plates and scoring adult progeny cultured under either condition.

In all experiments, animals expressing unc-22-dsRNA in neurons were exposed to the same tetracycline and water solutions used and scored for unc-22 silencing as adults as a control for effectiveness of tetracycline (see summary of data in Extended Data Fig. 6b).

Imaging and quantification of reporters using widefield microscopy

All animals and embryos expressing fluorescent reporters were imaged in 10 μl of 3 mM levamisole on a 2% agar pad using a Nikon AZ100 microscope and Photometrics Cool SNAP HQ² camera. A C-HGFI Intensilight Hg Illuminator was used to excite mCherry (filter cube: 530 to 560 nm excitation, 570 nm dichroic, and 590 to 650 nm emission), GFP or other autofluorescent molecules in the green channel (filter cube: 450 to 490 nm excitation, 495 nm dichroic, and 500 to 550 nm emission) and autofluorescent molecules in the blue channel (filter cube: 325 to 375 nm excitation, 400 nm dichroic, 435 to 485 nm emission). Intensity of GFP and mCherry were quantified in Fiji¹⁰¹ (NIH) using the methods described below. Representative images were adjusted in Fiji¹⁰¹ (NIH) and/or Adobe Photoshop to identical levels for presentation (Fig. 4b,c, Fig. 6a, Fig. 7e, Extended Data Fig. 6c-e,g).

For GTBP-1::GFP quantification post dsRNA injection

Somatic gfp expression was quantified between the pharynx and anterior gonad arm by drawing a circle or ventral to dorsal line within the boundaries of the animal (Extended Data Fig. 5a) on a brightfield image, creating a mask, imposing that mask onto the GFP channel image and measuring average intensity or intensity along the line, respectively. To measure background fluorescence, the same circle or a new circle was used to measure average intensity outside of the animal. Germline GFP expression was quantified by freely selecting part of the distal or proximal region of the anterior or posterior gonad arm (Extended Data Fig. 5a) excluding the intestine to avoid intestinal autofluorescence. Selection was performed using a brightfield image, a mask was created and imposed onto the GFP channel image and average intensity was measured. To measure background fluorescence, the same selection boundary was moved outside of the animal and average background intensity was measured. To plot average GFP intensity measured by a circle or free selection, average background intensity was subtracted from GFP intensity for each image and plotted with a box plot (Extended Data Fig. 5c). To plot GFP intensity along the ventral to dorsal axis in the anterior soma, the average intensity in each tenth of the axis was calculated for each animal and plotted with a shaded region representing 95% confidence intervals (Extended Data Fig. 5b, top). To calculate differences in intensity between the interior and exterior of animals, the average intensity of the 0.4-0.6 region of the axis was divided by the average intensity of the 0.1 and 0.9 points of the axis. These values were calculated and shown for each animal as a box plot (Extended Data Fig. 5b, bottom). All plotting was done using custom R scripts.

For GTBP-1::GFP quantification after exposure to dsRNA via feeding or neuronal expression

Animals fed L4440 or gfp-dsRNA for different durations of the P0 and/or F1 generation were scored for silencing in the germline and soma at different stages during the F1 generation (Fig. 1 and Extended Data Fig. 1a). Somatic GFP intensity (a.u.) was quantified in the tail region by drawing a ventral to dorsal line within the boundaries of the animal (Extended Data Fig. 6c,e) on a brightfield image, creating a mask, imposing that mask onto the GFP channel image and measuring average intensity or intensity along the line. To measure background fluorescence, a circle was used to measure average intensity outside of the animal. Germline GFP intensity (a.u.) was measured by free selection of germ cells but avoiding intestinal cells at each stage, selecting a region around the primordial vulva in L2 animals, in one of two extending gonad arms in L3 and L4 animals, in the proximal or distal gonad in young adults, and of eggs in utero in gravid adults. To measure background fluorescence, the same selection or a new selection was used to measure average intensity outside of the animal. To plot average GFP intensity measured by free selection, average background intensity was subtracted from GFP intensity for each image and shown as a box plot (Fig. 1, Extended Data Fig. 1a and Extended Data Fig. 6c,e). All plotting was done using custom R scripts.

Adjustment of fluorescence images of sid-1::mCherryΔpi animals for comparison to images of wild-type animals

Representative images of sid-1(jam195[linker::mCherryΔpi]) and wild type animals at different stages were adjusted to the same maximum and minimum displayed values of intensity using Fiji¹⁰¹ (NIH) to highlight each region of interest below saturation (Fig. 4b,c).

For quantification of SDG-1::mCherry and mCherry expressed under the sdg-1 promoter

Germline mCherry intensity was quantified by freely selecting part of the distal (for Fig. 7f,g) or proximal region of the anterior or posterior gonad arm excluding the intestine to avoid quantifying intestinal autofluorescence. Selection was performed using a brightfield image, a mask was created and imposed onto the mCherry channel image and average intensity was measured. To measure background fluorescence, the same selection boundary was moved outside of the measured gonad arm and average background intensity was measured. To plot average mCherry intensity, average background intensity was subtracted from mCherry intensity for each gonad arm and shown as a box plot using custom R scripts (Fig. 6b-g, Fig. 7f,g, and Extended Data Fig. 9). In Fig. 6c, SDG-1::mCherry intensity measurements, adjusted by subtracting background intensity and intensity measurements made in a wild-type animal lacking mCherry, were normalized to RT-qPCR measurements by multiplying each median intensity value by a conversion factor. This conversion factor was calculated by dividing the median SDG-1::mCherry intensity in AMJ1372 animals by the median relative sdg-1 mRNA level in AMJ1372 RNA samples. All estimated relative sdg-1 expression values were then normalized to those of wild-type animals by dividing all values by the wild-type value.

Imaging and quantification of reporters using confocal microscopy

For the endogenous gene tag sid-1::mCherryΔpi

SID-1::mCherry fluorescence from an L1-staged animal was imaged using LSM 980 Airyscan 2 Laser Scanning Confocal (Zeiss) with a 63X oil objective lens after paralyzing the worm as above. mCherry was excited using a 561 nm laser and fluorescence was collected through a 422-477 nm and 573-627 nm emission filter. For Fig. 4d, after removing noise using a 3D gaussian blur with 2.0 sigma in X, Y, and Z, depth-coded maximum intensity projections of Z-stacks were stitched together for display as described earlier²⁴.

For the endogenous gene tag W09B7.2/F07B7.2::mCherryΔpi

Adult animals were placed in 10 μl of 3 mM levamisole and imaged using the Eclipse Ti Spinning Disk Confocal (Nikon) with a 60X objective lens or the LSM 980 Airyscan 2 Laser Scanning Confocal (Zeiss) with a 63X oil objective lens. GFP was excited using a 488 nm laser and fluorescence was collected through a 499-557 nm and 659-735 nm emission filter, and mCherry fluorescence was excited and collected as described above. Images and movies were adjusted in Fiji¹⁰¹ (NIH) and Adobe Photoshop to identical levels for presentation (Fig. 7c,d and Extended Data Movies 1-4).

RNA sequencing, principal component analysis and differential expression analysis

For analysis of previously generated sid-1(-) alleles

Mixed-stage animals were washed from 10 plates in biological duplicate 5 days after passaging L4-staged animals. Total RNA was extracted from pellets using TRIzol (Fisher Scientific). PolyA+ RNAs were purified and converted to DNA libraries by the University of Maryland Genomics Core using the Illumina TruSeq Library Preparation Kit. FASTQ files were processed¹⁰³ using the command “cutadapt -j 0 -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA -m 20 -q 20 - o cutread.gz fasta1.gz”. Reads were assigned transcript IDs and counted¹⁰⁴ using the command “salmon quant -i celegans.index -l A -r cutread.gz -p 8 – validateMappings -o quant_file”. For conversion of transcript IDs to gene IDs, a table of matching transcript and gene IDs was generated from a GTF file using the command “grep “^[^#]” Caenorhabditis_elegans.Wbcel235.101.gtf | awk ‘{if($3 == “transcript”){print}}’ | awk ‘{print $12,$14}’ | tr - d ‘“;’ > transcript_id_gene_id.tsv”. Conversion was then made using this table with tximport¹⁰⁵ in R, whereafter only genes with more than 0.1 counts per million for at least 2 samples were used in subsequent analyses with pairs of sample types (sid-1(qt9[nonsense]) vs. wild type and sid-1(tm2700[deletion]); tmIs1005[sid-1(+) vs. sid-1(tm2700[deletion])). After normalizing samples using the trimmed mean of M-values method¹⁰⁶, principal component analysis was performed in R by comparing samples based on the 500 genes with the largest standard deviations in their log2-fold change between each set of samples (see Technical comments). Differential expression analysis was performed using limma(voom)¹⁰⁷ in R (example available at AntonyJose-Lab/Shugarts_et_al_2023 on github). Volcano plots of differential expression for all genes compared were plotted using custom R scripts with genes having an adjusted p-value threshold (q-value) less than 0.05 in black and those greater than 0.05 in grey (see Technical comments).

For analysis of newly generated sid-1(-) alleles

Total RNA >200 nt was extracted using RNAzol (Sigma-Aldrich) from 200 μl pellets of mixed-stage animals collected from 6 non-starved but crowded plates in biological triplicate for each strain. PolyA+ RNAs were purified and converted to DNA libraries using the Illumina TruSeq Stranded mRNA Library Preparation Kit. Library quality was assayed using TapeStation (Agilent) and libraries were sequenced using a HiSeq X10 (Illumina) by Omega Bioservices. FASTQ files were processed¹⁰³ using the command “cutadapt -j 0 -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA -A AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT -m 20 -q 20 -o cutread1.gz –p cutread2.gz read1.gz read2.gz”. Reads were assigned transcript IDs and counted¹⁰⁴ using the command “salmon quant -i celegans.index -l A -1 cutread1.gz -2 cutread2.gz -p 8 –validateMappings -o quant_files”. For conversion of transcript IDs to gene IDs, a table of matching transcript and gene IDs was generated as described above. Conversion was then made using this table with tximport¹⁰⁵ in R, whereafter only genes with more than 0.1 counts per million for at least 3 samples were used in subsequent analyses. Normalization, principal component analysis (Fig. 5c and Extended Data Fig. 7a) and differential expression analysis were performed as described above. Volcano plots of differential expression were plotted as described above (Fig. 5d and Extended Data Fig. 7b). Genes that were similarly misregulated in Fig. 5d and Extended Data Fig. 7b are in red.

For analysis of data from Reed et al., 2020

FASTQ files were processed¹⁰³ using the command “cutadapt -j 0 -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA -m 20 -q 20 -o cutread.gz fasta1.gz”. Reads were assigned transcript IDs and counted¹⁰⁴ using the command “salmon quant -i celegans.index -l A -r cutread.gz -p 8 –validateMappings -o quant_file”. For conversion of transcript IDs to gene IDs, a table of matching transcript and gene IDs was generated as described above. Conversion was then made using this table with tximport¹⁰⁵ in R. Normalization and differential expression analysis were performed as described above. Volcano plots of differential expression were plotted as described above with sid-1, sdg-1 (W09B7.2/F07B7.2) and sdg-2 (Y102A5C.36) in red and all other genes in grey (Extended Data Fig. 7d).

Genome mapping and visualization of sequencing reads for sid-1-dependent genes

After RNA sequencing samples were processed as described above, reads were mapped to the C. elegans genome¹⁰⁸ using the command “hisat2 -p 8 -x Celegans98index -1 cutread1.gz -2 cutread2.gz -S sam1”. The SAM file outputs were then converted to BAM files¹⁰⁹ using the command “samtools view -b sam1 | samtools sort -> bam1.bam” and BAM index files were created for visualization using “samtools index bam1.bam”. Reads for the sid-1 and F14F9.5 locus, W09B7.2/F07B7.2 locus, and Y102A5C.36 locus were plotted using custom R scripts and axes were normalized for each sample based on its total mapped reads, calculated using the command “samtools view -c -F 4 bam1.bam” (Extended Data Fig. 7c).

Comparisons with published datasets

Datasets in 21 published studies were collected and compared based on the gene names to identify changes in sid-1, sdg-1, sdg-2 and tbb-2 (control), if reported (Fig. 5f). After standardizing the names across all datasets, the fold-changes reported, if available, were used to plot a heatmap. Cases where fold-changes were not available were set conservatively as log2(fold change) = 2. The R script used is available at AntonyJose-Lab/Shugarts_et_al_2023 on github.

Reverse transcription and quantitative PCR

Total RNA was extracted using TRIzol (Fisher Scientific) from 200 μl pellets of mixed-stage animals collected from 3-6 non-starved but crowded plates in biological triplicate for each strain. The aqueous phase was then washed with an equal amount of chloroform and precipitated overnight on ice with 100 μl of 3 M sodium acetate, 1 ml of 100% ethanol and 10 μg glycogen (Invitrogen). RNA pellets were washed twice with 70% ethanol and resuspended in 22 μl nuclease free water. RNA samples were then Dnase-treated in Dnase buffer (10 mM Tris-HCl pH 8.5, 2.5 mM MgCl2, 0.5 mM CaCl2) with 0.5 U Dnase I (New England BioLabs) at 37°C for 60 minutes followed by heat inactivation at 75°C for 10 minutes. RNA concentration was measured and 1 μg of total RNA was used as input for reverse transcription using 50 U SuperScript III Reverse Transcriptase (Invitrogen) (+RT) or no reverse transcriptase as a negative control (-RT) (RT primers: tbb-2 (P98), sid-1 (P101), W09B7.2/F07B7.2 (P104), Y102A5C.36 (P107)). For qPCR, each +RT biological replicate was assayed in technical triplicate for each gene target, along with a single -RT sample for each corresponding biological replicate using 2 μl cDNA and a no template control (NTC) with the LightCycler® 480 SYBR Green I Master kit (Roche). Ct values were measured with the Bio-Rad C1000 CFX96 Real-Time System and Bio-Rad CFX Software (qPCR primers: tbb-2 (P99 and P100), sid-1 (P102 and P103), W09B7.2/F07B7.2 (P105 and P106), Y102A5C.36 (P108 and P109)). To calculate relative change in mRNA abundance compared to wild type, we calculated log2(2^{(-(gene Ct – tbb-2 Ct))}) using the median of technical replicates for the biological triplicates of each genotype. Ct values were only used if they were lower than corresponding -RT and NTC Ct values. The median value of wild-type biological replicates was then subtracted from the value for each sample to plot calculated values with respect to wild-type levels (Fig. 5e, Extended Data Fig. 7e, Extended Data Fig. 8e and Extended Data Fig. 9a).

BLAST searches and protein alignment

BLAST (NCBI) searches were performed using the W09B7.2/F07B7.2 (SDG-1) amino acid sequence with default parameters and any homologs identified were aligned to SDG-1 using Clustal Omega¹¹⁰ with default parameters. Alignments produced are shown in Extended Data Fig. 8b with residues shared by two proteins (grey highlight) or all three proteins (black highlight) indicated.

Annotation of the Cer9 retrotransposon containing W09B7.2/F07B7.2

The Cer9 retrotransposon containing W09B7.2/F07B7.2 (sdg-1) was annotated using sequence features from UCSC Genome Browser and amino acid sequences obtained from ref.¹¹¹. The 5’ and 3’ LTR sequences were identified using RepeatMasker and were confirmed to have TC and GA dinucleotides at the beginning and end of each sequence, respectively¹¹¹. Amino acid sequences from ref.¹¹¹ corresponding to gag and pol (PR: protease, RT: reverse transcriptase, RH: RNaseH, IN: integrase) elements of Cer9 were used in tblastn (NCBI) searches to determine their positions in the Cer9 retrotransposon sequence that also contains sdg-1.

Mating-induced silencing

Mating-induced silencing was assayed by crossing males with the transgene labeled T (oxSi487) encoding mex-5p::mCherry::h2b::gfp::h2b to hermaphrodites lacking the transgene, both in otherwise wild-type backgrounds or indicated mutant backgrounds. Reciprocal control crosses were performed in parallel where hermaphrodites with T were crossed to males lacking T. Animals were imaged and scored as described for this transgene in the “Light-induced damage of neurons” section.

Technical comments

Making a sid-1 translational reporter

Previous attempts at observing SID-1 localization relied on multi-copy transgenes¹⁴, which can become silenced within the germline¹¹² and could produce a variety of tagged and untagged proteins⁶⁴. When using multi-copy transgenes to express a SID-1 fusion protein tagged at the C-terminus with DsRed or GFP (Extended Data Fig. 10a) under the control of a promoter that drives expression within body-wall muscles, we observed intracellular localization of SID-1::DsRed or SID-1::GFP (Extended Data Fig. 10b) along with rescue of gene silencing by ingested dsRNA in body-wall muscles by both arrays (for SID-1::DsRed – silencing in wild type = 100% (n = 10), sid-1(qt9) = 0% (n = 11), sid-1(qt9); jamIs2[myo-3p::sid-1(+)::DsRed] = 100% (n = 6); for SID-1::GFP – silencing in wild type = 100% (n = 50), sid-1(qt9); jamEx193[myo-3p::sid-1(+)::gfp] = 100% (n = 60)). However, similar tagging to express SID-1 fusion proteins from either a single-copy transgene expressed in the germline (SID-1::DsRed) or the endogenous locus (SID-1::wrmScarlet) did not enable gene silencing by ingested dsRNA (for evaluating function of mex-5p::sid-1(+)::DsRed::sid-1 3’ UTR (jamSi12): silencing of pos-1 in wild-type = 100% (n = 7), sid-1(qt9) = 0% (n = 7), jamSi12; sid-1(qt9) = 0% (n = 15); for evaluating function of sid-1::wrmScarlet(jam117): silencing of pos-1 in wild-type = 100% (n = 8), sid-1(jam80) = 0% (n = 8), sid-1(jam117) = 0% (n = 8)), suggesting that the C-terminal fusions of SID-1 were likely non-functional and that apparent function when using multi-copy transgenes reflects production of untagged variants. In support of our rationale, a recent prediction of SID-1 structure^113,114 suggests that the C-terminus is sequestered (Extended Data Fig. 10c), a feature that may be disrupted by the addition of C-terminal fluorophores, potentially leading to misfolded proteins that are degraded. Consistently, we found that internal tagging of the sid-1 gene using Cas9-mediated genome editing to express SID-1::mCherry (Fig. 4) resulted in a fusion protein with detectable function (percent unc-22 silencing - wild type = 100% (n = 59), sid-1(jam195[sid-1::mCherryΔpi]) = ∼98% (n = 52); percent bli-1 silencing – wild type = ∼87% (n = 833), sid-1(jam195[sid-1::mCherryΔpi]) = ∼0.01% (n = 634)).

RNA sequencing analysis of existing sid-1 mutants

We initially analyzed polyA+ RNAs extracted from wild-type animals, two available sid-1 loss-of-function mutants^14,93 (sid-1(-)) and one available rescue strain where sid-1(-) was rescued with a transgene that overexpresses sid-1(+)⁹³, but found that pairwise comparisons between wild-type and mutant samples with otherwise similar genetic backgrounds did not yield any significantly misregulated genes present in both comparisons (Extended Data Fig. 10d). Strains with similar genotypes (sid-1(+) or sid-1(-)) did not cluster together when using principal component analysis (Extended Data Fig. 10e), suggesting that other differences (e.g., genetic background) obscure or misrepresent differences between sid-1(+) and sid-1(-) animals.

Rationale for inferences

Prior models and assumptions

All dsRNA is trafficked similarly. Entry of dsRNA into the germline can initiate transgenerational RNA silencing of some but not all genes. No SID-1-dependent germline genes are known, suggesting that SID-1 could be used solely in response to viral infection by analogy with roles of other members of RNA interference pathways.

Evidence supporting key conclusions

Temporal selectivity of dsRNA transport was probed using three approaches for delivery of dsRNA (damage-induced release from neurons, ingestion, and injection). Spatial selectivity of dsRNA import and/or subsequent silencing was inferred based on differences in the frequency of patterns of silencing within the germline. Substrate selectivity of dsRNA transport pathways was probed using genetic mutants and dsRNA of different lengths and 5’ chemistry. Diversity of dsRNAs made in bacteria and upon in vitro transcription was visualized using Northern blotting. Analysis of sid-1 mutants generated from the same wild-type cohort and a revertant was used for better control of genetic background, aiding in the identification of sid-1-dependent genes (sdg). Separate measurement of sdg-1 expression in descendants of independently edited isolates, along different lineages after perturbations, and in different gonads within single animals demonstrated stochasticity in gene expression and revealed establishment of different heritable epigenetic states. Co-localization of SDG-1::mCherry in perinuclear foci with the Z-granule marker GFP::ZNFX-1, its reported association with the Z-granule component ZSP-1/PID-2 and DEPS-1, changes in its levels in response to loss of SID-1 or the dsRNA-binding protein RDE-4 and its dynamic nuclear localization similar to CSR-1b was used to propose that SDG-1 plays a role in small RNA regulation while also being modulated by the activity of SID-1 and RDE-4.