ABSTRACT
Double-stranded RNA (dsRNA) can cause specific gene silencing upon ingestion in many invertebrates and is being developed as a pesticide to target essential genes in animal pests. Silencing by ingested dsRNA is best understood in the worm C. elegans, where ingested dsRNA is recruited into the RNA interference (RNAi) pathway by the dsRNA-binding protein RDE-4 for eventual gene silencing by Argonaute proteins. Although silencing is thought to rely on both the ingested dsRNA and on additional dsRNA-derived RNAs called mobile RNAs that are transported between cells, the specific forms of RNA that enter different tissues and the proteins they engage to cause silencing are unclear. We found that when RDE-4 was expressed at high levels within a somatic tissue, silencing by ingested dsRNA could occur in rde-4(-) somatic tissues but not in the rde-4(-) germline. Consistent with the spread of dsRNA-derived mobile RNAs between cells, silencing was more efficient in rde-4(-) cells located near rde-4(+) cells. Surprisingly, silencing by mobile RNAs derived from ingested dsRNA could bypass the requirement for a nuclear Argonaute that both ingested dsRNA and mobile RNAs derived from neuronal dsRNA require for silencing the same target gene. Furthermore, silencing by mobile RNAs could bypass inhibition of silencing within a tissue that can be caused by expression from repetitive DNA in that tissue. These results identify possible mechanisms that animals can use to evade RNAi and suggest that mobile RNA derivatives of ingested dsRNA can bypass these resistance mechanisms and cause gene silencing.
INTRODUCTION
Killing animals by feeding them double-stranded RNA (dsRNA) that matches an essential gene is a powerful approach to control animal pests. For example, expression of long dsRNA in potato plants was recently used to control populations of Colorado potato beetle that feed on these plants (Zhang et al. 2015). This approach to pest control relies on the ability of many insects and parasitic nematodes to ingest long dsRNA and use it to silence genes of matching sequence (Baum et al. 2007; Mao et al. 2007; reviewed in Koch and Kogel 2014) through RNA interference (RNAi). However, the mechanisms of gene silencing by ingested dsRNA are not well understood, making it difficult to anticipate resistance mechanisms and therefore design effective drugs.
Silencing of genes by feeding animals long dsRNA was first demonstrated in the nematode C. elegans (Timmons and Fire 1998) and it remains the best animal model for understanding this process called feeding RNAi. In C. elegans, ingested dsRNA enters the animal through the intestine and can be delivered into the fluid-filled body cavity that all tissues are exposed to without entry into the cytosol of intestinal cells (Jose et al. 2009; Calixto et al. 2010; Hinas et al. 2012). Entry into the cytosol of any cell requires a dsRNA-selective importer SID-1 (Winston et al. 2002) – a conserved protein with homologs in many insects (Tomoyasu et al. 2008). Upon entry into cells, silencing by dsRNA is thought to occur through the canonical RNAi pathway (reviewed in Grishok 2013). Long dsRNA is first bound by the dsRNA-binding protein RDE-4, which recruits the endonuclease DCR-1 to generate short dsRNAs (Tabara et al. 2002; Parker et al. 2006). One strand of this short dsRNA duplex is used as a guide by the primary Argonaute RDE-1 to identify mRNAs of matching sequence and to recruit RNA-dependent RNA polymerases (RdRPs) to the mRNA. RdRPs then synthesize numerous secondary small RNAs that are used for potent gene silencing within the cytosol by cytosolic Argonautes and/or within the nucleus by nuclear Argonautes. In addition to using long dsRNA for silencing, cells can also generate mobile RNAs (RNAs that can move between cells) derived upon processing of ingested dsRNA within muscle cells that express the dsRNA-binding protein RDE-4 (Jose et al. 2011; Blanchard et al. 2011) or the putative nucleotidyltransferase MUT-2 (Jose et al. 2011). Whether other tissues can make such dsRNA-derived mobile RNAs and whether both ingested long dsRNAs and its mobile RNA derivatives require the same proteins for silencing a gene is unknown.
Here, we show that high levels of RDE-4 in muscle cells, intestinal cells, hypodermal cells, or neurons can all enable generation of mobile RNAs derived from ingested dsRNA in C. elegans and can silence other somatic tissues but not the germline. Furthermore, silencing in somatic tissues by these mobile RNAs can be distinguished from silencing by ingested dsRNA. Specifically, unlike ingested dsRNA, their mobile RNA derivatives can bypass the requirement for the nuclear Argonaute NRDE-3 to silence the hypodermal gene bli-1 and can overcome the inhibition of silencing within a tissue caused by expression from repetitive transgenes in that tissue.
MATERIALS AND METHODS
Worm Strains
All strains were cultured on Nematode Growth Medium (NGM) plates seeded with 100 μl OP50 at 20°C and mutant combinations were generated using standard methods (Brenner 1974). All strains used are listed in Supplemental Material.
Balancing loci
Integrated transgenes expressing gfp were used to balance mutations in heterozygous animals. Progeny of heterozygous animals were scored as homozygous mutants if they lacked both copies of the transgene. The rde-4(ne301) allele on Chr III was balanced by juIs73. About 99% (153/155) progeny of rde-4(ne301)/juIs73 that lacked fluorescence were found to be homozygous rde-4(ne301) animals either by Sanger sequencing (96) or by resistance to pos-1 RNAi (59).
Transgenesis
To make strains, N2 gDNA was used (unless otherwise specified) as a template to amplify promoter or gene regions. To amplify gfp to be used as a coinjection marker a plasmid containing cytoplasmic gfp was used as a template (unless otherwise specified). All PCR reactions were performed with Phusion Polymerase (New England Biolabs-NEB), unless otherwise specified, according to the manufacturer’s recommendations and the final fusion products were purified using PCR Purification Kit (Qiagen).
Plasmids
The plasmid pJM6 (made by Julia Marré, Jose lab) was used to make Si[Pnas-9::rde-4(+)::rde-4 3’UTR]. The nas-9 promoter (Pnas-9) was amplified using primers P49 and P50 and rde-4(+)::rde-4 3’UTR was amplified using primers P54 and P4. The two PCR products were used as templates to generate the Pnas-9::rde-4(+)::rde-4 3’UTR fusion product with primers P52 and P53. This fused product was purified (QIAquick PCR Purification Kit, Qiagen) and cloned into pCFJ151 using the Sbfl and Spel restriction enzymes (NEB) to generate pJM6.
The plasmid pYC13 (made by Yun Choi, Jose lab) is a derivative of pUC::unc-119_sgRNA with a different sgRNA (gift from John Calarco, Addgene plasmid # 46169)) All other plasmids were as described earlier (pHC448 (Jose et al. 2011), pPD95.75 (gift from Andrew Fire, Addgene plasmid # 1494), pBH34.21 (Harfe and Fire 1998), pCFJ151 (Semple et al. 2012), pCFJ601 (Semple et al. 2012), pMA122 (Semple et al. 2012), pGH8 (Semple et al. 2012), pCFJ90 (Semple et al. 2012), pCFJ104 (Semple et al. 2012), pL4440 (Timmons and Fire 1998), pHC183 (Jose et al. 2009) and pGC306 (a gift from Jane Hubbard, Addgene plasmid # 19658)).
Genome editing
To generate bli-1 null mutants, Cas9-based genome editing using a co-conversion strategy was used (Arribere et al. 2014). Guide RNA for bli-1 was amplified from pYC13 using primer P56 and P57 and guide RNA for co-conversion of dpy-10 was amplified from pYC13 using P58 and P57. The amplified guides were purified (PCR Purification Kit, Qiagen) and tested in vitro for cutting efficiency (Cas9, NEB). For injection into animals, homology template for repair (repair template) was amplified from N2 gDNA using Phusion polymerase and gene specific primers. P59 and P60 was used to amplify a region immediately outside the 5’ region of bli-1 and P61 and P62 was used to amplify a region immediately outside the 3’ region of bli-1 using Phusion Polymerase (NEB). The two PCR products were used as templates to generate the repair template with primers P63 and P64 using Phusion Polymerase (NEB) and the fused product was purified (PCR Purification Kit, Qiagen). Homology template for dpy-10 was a singlestranded DNA oligo (P65). Animals were injected with 3.5pmol/μl of bli-1 guide RNA, 2.4pmol/μl of dpy-10 guide RNA, 0.06pmol/μl of bli-1 homology repair template, 0.6pmol/μl of dpy-10 homology repair template and 1.6pmol/μl of Cas-9 protein (NEB). Resulting progeny animals were analyzed as described in Figure S6.
Feeding RNAi
RNAi experiments were performed on RNAi plates (NGM plates supplemented with 1 mM IPTG (Omega Bio-Tek) and 25μg/ml Carbenicillin (MP Biochemicals)) at 20°C.
One generation or F1-only feeding RNAi
A single L4 or young adult (1 day older than L4) animal (P0) was placed on an RNAi plate seeded with 5μl of OP50 E. coli and allowed to lay eggs. After 1 day, by when most of the OP50 E. coli was eaten, the P0 animal was removed, leaving the F1 progeny. 100μl of an overnight culture of RNAi food (E. coli which express dsRNA against a gene) was added to the plate. Two or three days later, the F1 animals were scored for gene silencing by measuring gene-specific defects (See Table S1). All RNAi E. coli clones were from the Ahringer library (Kamath et al. 2003) and generously supplied by Iqbal Hamza, with the exception of DsRed and unc-54 RNAi, which was made by cloning a fragment of DsRed and unc-54 DNA, respectively, into pL4440 and transforming into HT115(DE3) E. coli cells. Control RNAi by feeding E. coli containing the empty dsRNA-expression vector (pL4440), which does not produce dsRNA against any gene, was done in parallel with all RNAi assays.
Intensity of L4 animals fed dsRNA against gfp in was measured using Image J (NIH). Animals with an intensity of >6000(a.u.) in a fixed area within the gut immediately posterior to the pharynx were considered not silenced. Based on this criteria, 91.7% of wild-type animals and a 100% of animals expressing DsRed in the muscle (Ex[Pmyo-3::DsRed]) showed silencing. In these silenced animals intensity of gfp was measured from below the pharynx to the end of the vulva and the background intensity for the same area was measured for each animal. The intensity of gfp for the area after background subtraction was plotted for each worm (Figure 4E).
Two generations or P0 & F1 Feeding RNAi
The experiments in Figures 4–7 (except Figure 4A, Figure 6D) and Figures S1, S3, S4, and S7 (except Figure S4A) were performed by feeding both the P0 and F1 generations, as described earlier (Jose et al. 2009). Control RNAi was done in parallel with all RNAi assays. Three or four days after P0 animals were subjected to RNAi, the F1 animals were scored for gene silencing by measuring gene-specific defects (See Table S1).
No difference in gene silencing was observed between F1-only feeding RNAi and P0 & F1 feeding RNAi for rde-4 mutants with tissue-specific rescue. For each RNAi experiment testing rde-4 function, feeding of N2 and WM49 was performed alongside as controls. In the case of Figure 7E (rde-4(-); Is[myo-3::rde-4(+)::rde-4 3’UTR; Ex[Pnas-9::gfp::unc-54 3’UTR]) and Figure 5 (act-5 feeding RNAi of rde-4(-); Si[nas-9::rde-4(+)::rde-4 3’UTR; Ex[Pnas-9::gfp::unc-54 3’UTR]), L4 animals were scored a day after control L4 animals because these animals grew slower than control animals.
Mosaic Analyses
Animals with an integrated Psur-5::sur-5::gfp in an rde-4(-) background that in addition have rde-4(+) DNA in some cells marked by nuclear-localized DsRed expression (a single extrachromosomal array with Psur-5::rde-4(+) and Psur-5::DsRed) were analyzed. Animals were subjected to F1 only RNAi feeding of bacteria expressing no dsRNA (control RNAi) or bacteria expressing dsRNA against gfp (gfp RNAi). The resulting animals were imaged and those with dim gfp expression or no detectable gfp expression in at least one intestinal nucleus were scored as silenced (Figure 3).
Scoring defects
For RNAi treatments, the proportions of animals that displayed the reported mutant defects upon RNAi (see Table S1) were scored as “fraction silenced”.
For bli-1 defects upon RNAi as well as upon Cas9-based genome editing, the pattern of blister formation was scored. Each animal was partitioned into eight roughly equal sections (a to h) as shown in Figure 6A with the vulva being the mid point of the animal. Sections with >50% of their length covered by a blister were marked black and sections with a discontinuous blister were marked grey. Animals that did not follow the anterior more than posterior and dorsal more than ventral susceptibility pattern (a > b > c > d > e > f > g > h) were culled as variants for each genotype and the relative aggregate blister formation in each section among worms with altered susceptibility (Figure 6C, Figure S6E,G,H and S7B,D) were computed using a score of black = 1.0 and grey = 0.5 for each section of every worm. The computed values for each section in all worms of a strain were summed and normalized to the value of the highest section for that strain. To compare multiple strains, these values for each strain were multiplied by the fraction of worms that showed a blister in that strain. Using these measures of normalized relative aggregate blister formation among animals with variant susceptibility, we generated heat maps (Pavlidis et al. 2003), where black indicates highest frequency of blisters and white indicates the lowest frequency of blisters among the sections of all strains that are being compared.
Microscopy
Animals were immobilized in 5μl of 3mM levamisole, mounted on slides, and imaged using an AZ100 microscope (Nikon) at a fixed magnification under non-saturating conditions. Images being compared on any figure were adjusted identically using photoshop (levels adjustment) for display.
ImageJ (NIH) was used to generate merged images (Figure 3D, Figure 4D, and Figure S4G). The LUT was set from 0 (white) to 127 (magenta) for the red channel and from 0 (white) to 127 (green) for the green channel. One channel was then overlayed on the other with 50% opacity.
Semiquantitative RT-PCR
RNA from each strain was isolated from 50 L4-staged animals as described earlier (Devanapally et al. 2015). Primer, P66 was used to reverse transcribe the sense strand of rde-4 and P67 was used to reverse transcribe the sense strand of tbb-2. The resulting cDNA was used as a template for PCR (30 cycles for both rde-4 and tbb-2) using Taq polymerase and gene-specific primers (P68 and P69 for rde-4 and P70 and P71 for tbb-2). Intensity of the bands were quantified as described previously (Devanapally et al. 2015). The level of rde-4(+) mRNA in wild type was set as 1.0 and that in other strains with the ne301 mutation were reported relative to that of wild type after subtracting the level of rde-4(ne301) mRNA in WM49 (0.3 in Figure S5).
Statistical Analyses
Error bars in all cases indicate 95% confidence intervals for single proportions calculated using Wilson’s estimates with a continuity correction (Method 4 in Newcombe et al. 1998). Significance of differences between two strains or conditions was determined using pooled Wilson’s estimates.
Data Availability
All strains are available upon request.
RESULTS
Tissue-specific rescue of rde-4 enables silencing in somatic cells that lack rde-4 expression but not in germline cells that lack rde-4 expression
Evaluating the effects of mobile RNAs derived from ingested dsRNA requires approaches that restrict RNA processing to specific tissues. A common approach used for such restriction of a process is tissue-specific rescue of a mutant by driving expression of the corresponding wild-type gene under the control of a tissue-specific promoter. One caveat of this approach is that expression from a tissue-specific promoter may not be sufficiently tissue-specific and could result in low or even undetectable levels of expression in unintended tissues that are nevertheless sufficient to provide function. Thus, to interpret the results from a tissue-specific rescue experiment, we required that the promoters used be well-characterized and that the effect – in this case gene silencing – be tissue-restricted when using the promoter in at least one scenario. The tissue-specific rescue of rde-1 has been demonstrated to cause tissue-restricted gene silencing in response to feeding RNAi (Qadota et al. 2007; Jose et al. 2011). Therefore, we required that a promoter enable tissue-specific gene silencing when used for tissue-specific rescue of rde-1 before using it for tissue-specific rescue of rde-4.
We generated animals with tissue-specific rescues of rde-4 by expressing rde-4(+) from a repetitive transgene under the control of tissue-specific promoters and assessed silencing in these animals in response to feeding RNAi against genes that function in different tissues. Specifically, we generated animals with rescue of rde-4 in body-wall muscles, intestine, hypodermis, or neurons and fed them dsRNA against target genes expressed in multiple tissues (Figure 1A, top). In all cases, ingested dsRNA enabled gene silencing in somatic tissues with rde-4(+) expression and in other somatic tissues that are expected to lack rde-4 expression. Consistent with restricted expression from the tissue-specific promoters used, silencing of endogenous genes in the corresponding rde-1(-) animals with rde-1(+) expression under the same tissue-specific promoter was restricted to the tissue that expressed rde-1(+) (Figure 1A, bottom; Qadota et al. 2007; Jose et al. 2011). However, some silencing of ubiquitously expressed green fluorescent protein (gfp) in non-muscle tissues was observed even in rde-1(-) animals with rde-1(+) expressed in body-wall muscles (Pmyo-3) (Figure S1), reflecting either modest silencing of this exogenous gene by RDE-1-dependent mobile RNAs or weak misexpression in non-muscle tissues from the muscle-specific promoter. In general, these results suggest that unlike in the case of animals with tissue-specific rde-1 rescue, in animals with tissue-specific rde-4 rescue, expression of the wild-type gene in a somatic tissue is not required for silencing of an endogenous gene expressed in that tissue by feeding RNAi. In contrast, silencing of a gene expressed within the rde-4(-) germline was not detected when rde-4(+) was expressed in any somatic tissue (Figure 1A, top, grey bars). This difference between silencing of somatic genes and of germline genes could result from either tissue-specific or target gene-specific differences. To eliminate target differences, we examined the silencing of gfp expressed from the same promoter (Pgtbp-1::gtbp-1::gfp) in both somatic tissues and in the germline. In animals with rde-4 rescue in a somatic tissue, silencing in response to ingested dsRNA occurred within multiple somatic tissues but not within the germline (Figure 1B&C and Figure S2). Thus, either the germline does not import sufficient amounts of mobile RNAs produced after the processing of ingested dsRNA in a somatic tissue by RDE-4 or the germline RNAi machinery is unable to recognize and use such mobile RNA derivatives of ingested dsRNA.
In summary, analyses of animals with tissue-specific rescue of rde-4 reveal that the generation of mobile RNAs derived from ingested dsRNA can occur in multiple somatic tissues and that these mobile RNAs can silence genes in other somatic tissues but not in the germline.
Somatic RDE-4 in parents cannot enable feeding RNAi in rde-4(-) progeny
Repetitive transgenes can sometimes result in expression within the germline despite the use of somatic promoters (e.g. heat shock promoter (Sheth et al. 2010)). Such misexpression within the germline and subsequent delivery of RDE-4 into progeny would complicate interpretation of silencing in animals with tissue-specific rescue of rde-4. Therefore, to test if such parental rescue is detectable, we expressed rde-4(+) in the body-wall muscle, hypodermis, or intestine of rde-4(-) animals and fed only their rde-4(-) progeny dsRNA against genes expressed in each of these tissues (Figure 2A, Figure S3A-D). We observed that in most cases rde-4(-) progeny were unable to silence the target gene upon feeding RNAi. The only exception was unc-22 silencing in rde-4(-) progeny of parents expressing rde-4(+) under the body-wall muscle promoter myo-3. The extent of this silencing did not correlate with the rate of transmission of the extrachromosomal Pmyo-3::rde-4(+) array (Figure S3E) and silencing was also detected in rde-4(-) progeny of animals with an integrated Pmyo-3::rde-4(+) array (Figure 2B). One explanation for this observation could be that the myo-3 promoter is misexpressed at low levels within the germline and consequently, maternally deposited rde-4 protein or mRNA enables silencing of unc-22 when larvae are fed dsRNA. In sum, our results suggest that, with the possible exception of rescue using a myo-3 promoter and silencing of unc-22, tissue-specific rescue of rde-4 in somatic tissues of parents does not enable silencing of somatic genes in rde-4(-) progeny that ingest dsRNA. Therefore, cases of silencing in rde-4(-) tissues in response to tissue-specific expression of rde-4(+) (Figure 1) are likely due to the production of mobile RNA derivatives of ingested dsRNAs and not due to inheritance of parental RDE-4 protein by rde-4(-) tissues.
Silencing can occur in rde-4(-) cells when rde-4 mosaic animals ingest dsRNA
Effects of mobile RNAs derived from ingested dsRNA can also be evaluated using mosaic analysis (Yochem and Herman 2003). In this approach, mosaic animals that result from the loss of rescuing DNA during mitosis are examined. Perdurance of protein or mRNA after the loss of rescuing DNA in ancestral cells can complicate the interpretation of effects in descendent cells, which is a different caveat compared to that for tissue-specific rescue. Therefore, to complement our analysis using tissue-specific rescue, we examined silencing by ingested dsRNA in rde-4(-) mosaic animals. Specifically, we co-expressed rde-4(+) and DsRed under the control of the sur-5 promoter, which drives expression in all somatic cells, and examined silencing of gfp upon feeding RNAi of an integrated Psur-5::sur-5::gfp transgene in mosaic animals. This results in animals with an extrachromosomal array where cells that have the DNA for rde-4(+) expression are marked with DsRed expression. In all 19 mosaic animals examined, silencing was observed in rde-4(+) as well as in rde-4(-) cells (Figure 3D). Taken together with the results from tissue-specific rescue of rde-4, these results are consistent with the hypothesis that some derivatives of ingested dsRNA generated in somatic tissues with rde-4(+) expression are transported between cells and are thus mobile.
Expression of a repetitive transgene in a tissue can inhibit silencing in that tissue
Intriguingly, we observed several cases of reduced silencing specifically within cells that express the rde gene in rde mosaic animals and in animals with tissue-specific rescue of rde. In rde-4 mosaic animals, silencing was less efficient in cells that showed expression of rde-4(+) than in cells that lacked expression of rde-4(+) (e.g. the single binucleated intestinal cell in Figure 3D). In animals with tissue-specific rescue of rde-4, silencing of gfp (Figure 4A,B) or unc-54 (Figure S4B) within body-wall muscles that express rde-4(+) and silencing of bli-1 or dpy-7 within hypodermal cells that express rde-4(+) (Figure S4C,D) were not detectable. In animals with tissue-specific rescue of rde-1, silencing of gfp (Figure S1B,C) or unc-54 (Figure S4E) in body-wall muscles that express rde-1(+) was reduced.
To test if these cases of reduced silencing could be explained by insufficient levels of rde expression, we overexpressed rde-4(+) in the hypodermis of wild-type animals and measured silencing of bli-1 and of dpy-7 by ingested dsRNA (Figure 4C, second set of bars). Animals with the extrachromosomal rde-4(+) transgene failed to detectably silence bli-1 or dpy-7. Thus, the observed lack of silencing is not because there is insufficient rde-4(+) expression, but because expression of rde-4(+) within a tissue from a repetitive transgene inhibits feeding RNAi in that tissue. To test if the inhibition was because of high levels of rde-4(+), co-suppression (Napoli et al. 1999) of rde-4, or expression from a repetitive transgene, we expressed gfp from a repetitive transgene in the hypodermis and examined silencing of bli-1 and of dpy-7 by feeding RNAi. No silencing was detected in the presence of gfp expression (Figure 4C, fourth panel), suggesting that inhibition of silencing is the result of expression from a repetitive transgene and not because of rde-4 expression or co-suppression. Similar reduction of silencing was also observed in body-wall muscles when we expressed DsRed from repetitive transgenes in body-wall muscles (Figure 4D-F). Consistently, tissue-specific expression of rde-4(+) in the hypodermis from a single-copy transgene enabled silencing of both dpy-7 and bli-1 by feeding RNAi (Figure 4G). This silencing was also inhibited by expression of gfp in the hypodermis from a repetitive transgene (Figure 4G). Thus, expression from a repetitive transgene and not high levels of rde-4(+) is the likely reason for the observed inhibition of feeding RNAi in rde-4 mosaic animals and in animals with tissue-specific rde-4 rescue.
Repetitive transgenes can produce dsRNA (Hellwig and Bass 2008) that might compete with ingested dsRNA for engaging the gene silencing machinery within a cell. Such competition between pathways has been proposed as the reason why enhanced silencing in response to feeding RNAi can occur in animals that lack genes required solely for the processing of endogenous dsRNA (Lee et al. 2006). During endogenous RNAi, the RdRP RRF-3 and the exonuclease ERI-1 produce dsRNA that is processed by the endonuclease DCR-1 and the primary Argonaute ERGO-1 into guide RNAs, which are used to subsequently silence genes of complementary sequence (Figure 4H and reviewed in Billi et al. 2014). Loss of components required solely for the processing of endogenous dsRNA (rrf-3, eri-1, and ergo-1) has been shown to enhance feeding RNAi (Zhuang et al. 2011). Similar enhancement of silencing could potentially relieve the inhibition of feeding RNAi caused by expression from repetitive transgenes. We found that loss of rrf-3 or of eri-1 but not of ergo-1 eliminated inhibition of feeding RNAi by expression from a repetitive transgene within the hypodermis (Figure 4I). Because RRF-3 and ERI-1 are not required for the production of dsRNA from repetitive transgenes (Kim et al. 2005), these results suggest that silencing by feeding RNAi in the presence of expression from a repetitive transgene is enabled by loss of dsRNA production at endogenous loci (see Figure S9 in Blumenfeld and Jose 2016).
Together, our results suggest that expression from some repetitive DNA in a tissue interferes with silencing by ingested dsRNA of some genes within that tissue.
Silencing in tissues that lack rde-4 expression is observed only in animals with high levels of tissue-specific rde-4 expression
Despite inhibition of silencing within tissues that express rde-4(+) from a repetitive transgene, robust silencing in tissues that lack rde-4 was observed in animals with tissue-specific rde-4 rescue and in rde-4 mosaic animals (Figure 1, Figure 3, and Figure S2). To test if similar silencing in tissues that lack rde-4 can occur when rde-4(+) is expressed within a tissue using a single-copy transgene, we subjected animals with tissue-specific rde-4 rescue in the hypodermis to feeding RNAi against genes expressed in the hypodermis (bli-1), the body-wall muscles (unc-22), or the intestine (act-5) and measured silencing. While silencing of bli-1 was comparable to that in wild-type animals, silencing of unc-22 or act-5 was barely detectable (Figure 5, top). This lack of silencing in tissues that lack rde-4 expression could either be because such silencing requires expression of any repetitive transgene within the rde-4(+) tissue or because it requires high levels of rde-4(+) expression. Two observations support the latter possibility: (1) silencing in tissues that lack rde-4 was not detected even when hypodermal expression of rde-4(+) from the single-copy transgene was combined with hypodermal expression of gfp from a repetitive transgene (Figure 5, bottom); and (2) the level of rde-4(+) mRNA in animals with repetitive transgenes was higher than that of animals with the single-copy transgene (Figure S5). Together these results suggest that silencing by mobile RNAs derived from ingested dsRNA in tissues that lack rde-4 requires the high levels of rde-4(+) expression within a tissue that can be achieved through expression from repetitive transgenes.
Site of rde-4 expression dictates pattern of silencing in cells that lack rde-4 expression
When an animal is only scored as silenced versus not silenced in response to feeding RNAi, qualitative differences between animals (e.g. silencing in different subsets of cells) are overlooked. Examination of such differences requires a target gene whose silencing in subsets of cells can be discerned in each animal. We found that null mutants of the hypodermal gene bli-1 result in a fluid-filled sac (“blister”) along the entire worm (Figure S6A-C; Brenner 1974), and that blisters that form upon feeding RNAi in wild-type animals had a different pattern (Figure 6A). Specifically, when the worm is divided into eight sections (Figure 6A, top), anterior sections tended to be more susceptible to silencing when compared with posterior sections, resulting in a stereotyped pattern of relative susceptibility to blister formation upon bli-1 feeding RNAi (Figure 6A, bottom and Figure S6E). This bias in the tendency to form blisters could reflect the graded uptake of dsRNA from the anterior to the posterior in the intestine upon feeding RNAi. These characteristics of blister formation as a result of bli-1 silencing enable examination of qualitative differences, if any, between silencing in wild-type animals and in animals with tissue-specific rde-4 rescue.
Silencing of bli-1 in animals with tissue-specific rescue of rde-4 in non-hypodermal tissues was associated with variations in the stereotyped pattern of blister formation (Figure 6B-D). Specifically, rde-4(-) animals with high levels of rde-4(+) expression, assumed based on the high levels of gfp coexpressed from the same promoter, in posterior intestinal cells displayed a higher relative frequency of posterior blisters than did wild-type animals (Figure 6B and Figure S6F,G). To systematically analyze such differences, we culled animals that had a pattern of blister formation that differed from a reference blister susceptibility pattern observed in most wild-type animals (see methods). We found that unlike in wild-type animals, in animals with tissue-specific rde-4 rescue, patterns of blisters that differed from the reference pattern were common (Figure 6C,D). Furthermore, the pattern of variant blister susceptibility differed depending on the tissue in which rde-4(+) was expressed (neurons vs. body-wall muscles) and not on the promoter used (Prgef-1 or Punc-119 for neurons and Pmyo-3 or Punc-54 for body-wall muscles) (Figure 6C,D and Figure S6H). Taken together, these results suggest that dsRNA-derived mobile RNAs generated by expression of rde-4(+) in different tissues cause different patterns of silencing in rde-4(-) tissues: mobile RNAs derived from ingested dsRNA appear to cause silencing in nearby rde-4(-) hypodermal cells.
Ingested dsRNA and their mobile RNA derivatives have different requirements for gene silencing
The ability to examine qualitative differences in silencing of bli-1 provides an opportunity to determine differences, if any, between silencing by ingested dsRNA and silencing by dsRNA-derived mobile RNAs.
To determine the pathway through which silencing of bli-1 occurs in response to ingested dsRNA, we examined silencing in animals that lack genes required for the canonical RNAi pathway (reviewed in Grishok 2013). We found that in addition to the dsRNA-binding protein RDE-4, the primary Argonaute RDE-1, and the RNA-dependent RNA polymerase RRF-1, genes that act in the nuclear RNAi pathway (the nuclear Argonaute NRDE-3 (Guang et al., 2008; Mao et al., 2015) and downstream components NRDE-2, NRDE-1, and NRDE-4) were also required for bli-1 silencing in response to ingested dsRNA (Figure 7A).
To test if mobile RNAs derived from ingested dsRNA also require the nuclear RNAi pathway for silencing bli-1, we examined genetic requirements for blister formation upon bli-1 feeding RNAi in animals with rde-4 rescue in non-hypodermal tissues. While no silencing of bli-1 was detected in the absence of rde-1 or of rrf-1 when rde-4 was rescued in neurons, substantial silencing could be detected in the absence of nrde-3 when rde-4 was rescued in neurons, in body-wall muscles, or in the intestine (Figure 7B,C and Figure S7A). Furthermore, the patterns of blister formation observed in the absence of nrde-3 were different from those observed in the presence of nrde-3 (compare Figure 7C with Figure 6D, also see Figure S7B). These results suggest that bli-1 silencing by mobile RNAs derived from ingested dsRNA can occur independent of NRDE-3, potentially using one of the many other Argonautes in C. elegans (Yigit et al. 2006). Similar nrde-3-independent silencing of bli-1 was also observed upon loss of eri-1 or upon overexpression of rde-4(+) in hypodermal cells from a single-copy transgene (Figure S7C-E). However, expression of bli-1 dsRNA in neurons resulted in bli-1 silencing that depended on both sid-1 and nrde-3 (Figure 7D), suggesting that mobile RNAs derived from ingested dsRNA and mobile RNAs made from neuronal dsRNA have different requirements for silencing bli-1. Thus, unlike silencing by ingested dsRNAs and neuronal mobile RNAs, silencing by mobile RNAs derived from ingested dsRNA is akin to silencing in an enhanced RNAi background or silencing within a tissue with RDE-4 overexpression and can occur independent of the nuclear RNAi pathway.
Because silencing by ingested dsRNA and by ingested dsRNA-derived mobile RNAs appear to have different requirements, we wondered if silencing by both forms of dsRNA were equally susceptible to inhibition by expressed repetitive transgenes. To test for inhibition, we expressed gfp from a repetitive transgene in the hypodermis of wild-type animals or rde-4(-) animals with rde-4(+) in body-wall muscles and examined silencing in response to feeding RNAi of the hypodermal genes dpy-7 or bli-1. Unlike silencing by ingested dsRNA, silencing by their mobile RNA derivatives was not robustly inhibited (Figure 7E).
In summary, our results suggest that silencing of a gene by ingested dsRNA-derived mobile RNAs has different requirements compared to those for silencing of the same gene by ingested dsRNA or by neuronal mobile RNAs (Figure 8A).
DISCUSSION
Our analyses of feeding RNAi in C. elegans reveal that high levels of RDE-4 in one somatic tissue can generate dsRNA-derived mobile RNAs that can silence a gene in other somatic tissues but not in the germline (Figure 8B). Silencing by ingested dsRNA and by dsRNA-derived mobile RNA can be distinguished based on their differential requirement for the nuclear Argonaute NRDE-3 to silence the hypodermal gene bli-1 and their differential susceptibility to inhibition by expression from repetitive DNA (Figure 8A).
Mobile RNAs derived from ingested dsRNAs could be chemically modified short dsRNAs
Ingested dsRNA can be transported across the intestine and accumulate in the fluid-filled body cavity without entry into the cytosol of intestinal cells in C. elegans (Jose et al. 2009; Calixto et al. 2010; Hinas et al. 2012). Additionally, dsRNA-derived mobile RNAs made from multiple tissues (intestine, body-wall muscles, hypodermis, and neurons in Figure 1) can be present in the body cavity. Our results reveal three differences between silencing by these two forms of extracellular dsRNA – differential silencing of germline genes, differential requirements for a nuclear Argonaute, and differential susceptibility to inhibition by repetitive transgenes. One explanation for these differences could be quantitative differences in the amounts of ingested dsRNAs and their mobile RNA derivatives delivered into the target tissues. Alternatively, we speculate that cells could distinguish ingested dsRNA from their mobile RNA derivatives.
Three considerations suggest that mobile RNAs derived from ingested dsRNA are chemically modified short dsRNAs. First, the requirement for processing by RDE-4 but not by RDE-1 to generate dsRNA-derived mobile RNAs (Figure 1) suggests that they are short dsRNAs. Second, the inability of dsRNA-derived mobile RNAs to cause silencing within the germline (Figure 1) is consistent with them being short dsRNAs because RNAs need to be longer than 50 bp for entry into the germline and/or for subsequent silencing (Feinberg and Hunter 2003). Third, terminal chemical modification of dsRNA-derived mobile RNAs could be the basis for the ability of these RNAs to bypass the requirement for the Argonaute NRDE-3 (Figure 7B&C) because Argonautes recognize small RNAs by binding to their termini (Czech et al. 2011) and because the putative nucleotidyltransferase MUT-2 is also required for the generation of mobile RNAs from ingested dsRNAs (Jose et al. 2011).
Efficiency of RNAi could be regulated by expression from repetitive DNA
Our discovery that expression from repetitive DNA within a tissue can interfere with silencing by feeding RNAi within that tissue (Figure 4) raises concerns for the use of RNAi to infer the function of a gene. For example, RNAi of a gene in strains that express fluorescent reporters within a tissue from a repetitive transgene could be specifically inhibited in that tissue resulting in differences between the defects observed upon RNAi of a gene and those observed upon mutation of that gene. Thus, inferences from past experiments that used feeding RNAi in the presence of tissue-specific expression from repetitive transgenes may need to be revised.
The efficiency of feeding RNAi differs in different tissues and is a key concern for the application of feeding RNAi to combat animal pests (Koch and Kogel 2014). For example, in C. elegans, genes expressed in neurons are relatively refractory to silencing by feeding RNAi (noted in Tavernarakis et al. 2000). One reason for such reduced silencing could be that neurons have high levels of expression from endogenous repetitive DNA. Consistent with this possibility, both silencing in tissues with expression from repetitive DNA (Figure 4I) and silencing in neurons are enhanced upon loss of the exonuclease ERI-1 (Kennedy et al. 2004) or the RdRP RRF-3 (Simmer et al. 2002). Thus, when feeding RNAi is used against animal pests, expression from endogenous repetitive DNA could cause resistance to silencing.
Ingested dsRNA could engage similar mechanisms in many invertebrates
Feeding RNAi can be used to silence essential genes in many insects and parasitic nematodes (see Jose 2015; Koch and Kogel 2014; Lilley et al. 2012 for reviews). Recent studies suggest that several characteristics of feeding RNAi in these invertebrates are similar to those in C. elegans. First, long dsRNA (>60 bp) is preferentially ingested (Bolognesi et al. 2012) and realization of this preference was crucial for developing plastid expression as an effective strategy to deliver long dsRNA into crop pests (Zhang et al. 2015). Second, dsRNA can be detected in intestinal cells and in internal tissues upon feeding RNAi (Ivashuta et al. 2015). Third, with the exception of dipteran insects, most invertebrates have homologs of the dsRNA importer SID-1 (Tomoyasu et al. 2008). Finally, silencing initiated by feeding RNAi can persist for multiple generations (Abdellatef et al. 2015). These similarities suggest that insights gleaned using the tractable animal model C. elegans are likely to be applicable to many invertebrates - including agronomically important insect and nematode pests.
ACKNOWLEDGEMENTS
We thank Leslie Pick and members of the Jose lab for critical reading of the manuscript; Julia Marre (Jose lab) for generating the plasmid pJM6; Yun Choi (Jose lab) for generating the plasmid pYC13; the Caenorhabditis elegans Genetic stock Center, the Hunter lab (Harvard University), and the Seydoux lab (Johns Hopkins University) for some worm strains and the Hamza lab (University of Maryland) for bacteria that express gfp-dsRNA. Critical comments from anonymous reviewers were crucial in arriving at the working model proposed in this manuscript. This work was supported in part by National Institutes of Health Grant R01GM111457 (to A.M.J.)