Temporally restricted autonomous RNAi by germline-injected dsRNA

SID-1 is required to transport dsRNA to the germline, as shown by injecting dsRNA into specific tissues or the pseudocoelom (PC) (Winston 2002; Winston et al. 2002). However, while unc-22 dsRNA injection into any tissue in wild-type animals resulted in unc-22 silenced twitching progeny, unc-22 dsRNA injected into the syncytial germline of a sid-1 mutant, unexpectedly, did not result in any affected progeny (Winston 2002; Winston et al. 2002). To further investigate this result, we injected unc-22 dsRNA directly into the syncytial germline of wild-type and sid-1 mutant adult hermaphrodites, and, every 2 hr after injection, collected their self-progeny. As expected, and consistent with previous observations of systemic silencing, the proportion of twitching progeny from injected wild-type animals quickly rose to 100%, and was sustained for the duration of the experiment (Figure 1B). In contrast to previous observations, we found that injecting dsRNA directly into the syncytial gonads of sid-1^−/− animals produced strongly twitching progeny. However, this was only true of embryos laid within the first ∼18 hr after injection (Figure 1A). The timing of peak silencing varied between injected P₀ animals, but always reached 100% (Figure S1 in File S1). Furthermore, when a single gonad arm was injected in sid-1 mutant animals, a maximum of only 50% twitching progeny was produced (Figure 1C and Figure S1 in File S1). The subsequent decrease in fraction of silenced F₁ embryos to zero, and the animal-to-animal variability when production of silenced progeny ceased suggests that the injected unc-22 dsRNA is rapidly depleted. In previous experiments (Winston 2002; Winston et al. 2002), the injected hermaphrodites were allowed to recover for up to 24 hr before progeny were collected and scored for silencing. The apparent rapid depletion of syncytial germline injected dsRNA explains the past failure to detect sid-1-independent silencing.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

sid-1-dependent and -independent silencing in progeny of dsRNA injected parents. (A–D) Time course of fraction of progeny with the Unc-22 phenotype laid after unc-22 dsRNA gonad injection into wild-type or sid-1 mutant hermaphrodites. (E) Fraction of progeny with Unc-22 phenotype following unc-22 dsRNA gonad or PC injection into sid-1^−/− mutant hermaphrodites crossed to wild-type or sid-1 mutant males. Error bars in (A–D) represent SE from two experiments with 10 injected hermaphrodites each. Error bars in (E) represent SD from 4, 6, and 3 injected hermaphrodites, left to right. ** P < 0.01 by Welch’s t-test.

Maternal SID-1-independent systemic spread and perdurance of parental RNAi

DsRNA injections into a single gonad of wild-type hermaphrodites produced 100% twitching progeny, clearly showing that gonad-injected dsRNA is mobile between gonad arms (Figure 1D). Additionally, in wild-type animals, dsRNA-induced silencing is not limited by the bolus of injected dsRNA, but persists indefinitely in the injected hermaphrodites. Interestingly, injecting unc-22 dsRNA into a single gonad arm of sid-1 mutant hermaphrodites and then crossing them to wild-type males produced nearly 100% twitching heterozygous cross progeny (Figure 1E). This indicates that maternal sid-1 activity is not required for gonad to gonad transfer of dsRNA. It also indicates that maternal sid-1 activity is not required for long-term persistence of the silencing signal. The gonad to gonad dsRNA transfer likely occurs via the PC fluid, which may also be the conduit to the reservoir of dsRNA that persists indefinitely after injection. However, both sid-1-independent gonad to gonad dsRNA transfer and long-term persistence of silencing are only apparent when SID-1 is expressed in the embryo. These new observations, like previous observations following injection into both gonad arms or the PC space, indicate the presence of a maternally active, sid-1-independent dsRNA transport pathway from the PC to the germline that subsequently requires sid-1 in the progeny to realize silencing (Winston 2002; Winston et al. 2002).

sid-1-independent dsRNA transmission requires RME-2

The above results, showing sid-1 independent delivery of dsRNA from mothers to embryos and a sid-1-dependent step in the embryos for effective RNAi is similar to feeding RNAi, where the uptake and transport of ingested dsRNA is separable. In feeding RNAi, the intestinal lumen-localized transmembrane protein SID-2 is first required for endocytosis of ingested dsRNA into gut cells, and SID-1 is subsequently required for effective RNAi within intestinal cells likely to release dsRNA from endosomes (Winston et al. 2007; McEwan et al. 2012). A candidate receptor for dsRNA endocytosis in the germline is RME-2, an LDL receptor superfamily homolog that functions in oocytes as a yolk/lipoprotein receptor. In C. elegans, yolk is synthesized in the intestine, exported to the PC, and then taken up by oocytes via receptor-mediated endocytosis (Sharrock et al. 1990; Grant and Hirsh 1999). Thus, we hypothesized that dsRNA that exits injected gonads, or is injected into the PC space, may bind to either yolk protein, another RME-2 substrate, or directly to RME-2 for endocytosis into oocytes. To test this hypothesis, we examined rme-2 mutants, which do not take up any yolk (Grant and Hirsh 1999). We injected unc-22 dsRNA into the PC of rme-2^−/−; sid-1^−/− double mutant hermaphrodites and then crossed these animals to wild-type males. The heterozygous cross progeny did not twitch, showing that, in the absence of maternal sid-1, RME-2 is required for parental RNAi in response to PC injected dsRNA (Figure 2A). While this work was in progress, it was reported that parental RNAi by feeding RNAi also requires RME-2 (Marré et al. 2016). Both of these findings support the hypothesis that RME-2 may endocytose dsRNA into oocytes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Maternal RME-2-dependent inherited silencing. (A) Fraction of progeny sensitive to unc-22 silencing among the self-progeny or indicated cross-progeny of hermaphrodites PC-injected with unc-22 dsRNA; n = 6, 5, 3, 4, and 12 injected hermaphrodites respectively. (B) Fraction of progeny sensitive to unc-22 silencing after wild-type parents were exposed to feeding RNAi at the given periods of time after hatching; n = 30 treated parents for each group. (C) Sensitivity to unc-22 feeding RNAi in progeny after treating (i) wild-type parents as adults, (ii) rme-2 mutant parents as adults, or (iii) rme-2 mutant parents as L4 larvae. Because rme-2 mutants have severely reduced fecundity, the results from each individual parent are presented separately for clarity, with silenced progeny represented in black bars and nonsilenced progeny in gray bars. All error bars represent SD. **** P < 0.00001 by t-test. n.s., not significant.

SID-1 is sufficient in the germline to transmit maternal dsRNA to progeny

The results reported by Marré et al. (2016) suggest that RME-2 is essential for parental RNAi. We tested this explicitly by injecting unc-22 dsRNA into the PC of rme-2 single mutants. We found that the progeny twitched (Figure 2A). This result indicates that RME-2 is not required for dsRNA uptake in sid-1+ mothers, and further suggest that SID-1 may directly transport PC dsRNA to oocytes and embryos.

This finding of asymmetric redundancy between maternal SID-1 and RME-2, where sid-1 is required and rme-2 is required only in the absence of sid-1, contrasts with the report that rme-2 single mutants are defective for inherited RNAi initiated by ingesting unc-22 dsRNA (Marré et al. 2016). Although RME-2 is most abundantly expressed in oocytes, it is possible that rme-2 is also required for feeding RNAi. It is also possible that their negative result may be explained by the difference in potency between PC injection of concentrated in vitro transcribed dsRNA as described above and the feeding RNAi used by Marré et al. (2016). To address these possibilities, we repeated the feeding RNAi assay as described, i.e., exposing L4 hermaphrodites to unc-22 dsRNA-expressing bacteria for 1 day before washing and transferring to control bacteria. By these methods, we detected robust twitching in the fed animals, but failed to detect inherited RNAi among the progeny of even wild-type parents. This indicates that rme-2 is not required for feeding RNAi, but the lack of parental RNAi was perplexing. Further analysis revealed a striking dependency on maternal developmental stage for inherited silencing. We split a batch of freshly hatched wild-type larvae into three populations, and exposed each to unc-22 RNAi food on either day 1 (L1/L2), day 2 (L3/L4), or day 3 (adult) after hatching, thoroughly washing off residual bacteria each day. After a final wash, animals were allowed to lay F₁ progeny on non-RNAi food. We found that only animals exposed to unc-22 food as adults efficiently produced twitching progeny (Figure 2B), even though the P₀ animals fed on days 1 and 2 still exhibited a strong Unc-22 twitching phenotype immediately after RNAi exposure. We then used these modified feeding RNAi conditions, which are effective for parental RNAi in wild-type animals, to determine whether rme-2 is required for parental RNAi. We found that the progeny of wild-type and rme-2 adults placed on unc-22 food show similar proportions of twitching progeny (Figure 2C). Note that rme-2 hermaphrodites are less fecund that wild-type hermaphrodites (Grant and Hirsh 1999). This result shows that rme-2 is not required in sid-1+ hermaphrodites for parental RNAi, even when the dsRNA is introduced by ingestion. Because inherited RNAi by dsRNA feeding requires adult exposure (Figure 2, B and C), it is reasonable to assume that the slightly slower development of rme-2 mutants differentially limited the period of adult exposure compared to wild-type animals when initially placed on the RNAi food as L4 larvae. Thus, the slowed development of rme-2 mutants may have led Marré et al. (2016) to the erroneous conclusion that rme-2 is required for parental RNAi. Indeed, exposing wild-type or rme-2 L4 animals to unc-22 RNAi food for 24 hr produces few twitching progeny (Figure 2, B and C). In summary, independent of the means of dsRNA delivery, in sid-1+ animals rme-2 is not required for transport of dsRNA into the germline.

sid-5 functions with sid-1 in the embryo

Pseudocoelomic unc-22 dsRNA transported into oocytes or embryos in the absence of maternal sid-1 requires embryonic sid-1 to silence unc-22. We next asked whether sid-2, sid-3, or sid-5 are also required in the embryo. We first crossed sid-1; sid-2 double mutant hermaphrodites injected with unc-22 dsRNA to either wild-type or sid-2 males, and scored silencing in the cross progeny. If sid-2 activity is required in the embryo, then the cross progeny of only the wild-type males will twitch. If sid-2 activity is not required in the embryo, then the cross progeny of both wild-type and sid-2 males will twitch. Using this method, we determined that neither sid-2 nor sid-3 are required for either maternal uptake or embryonic release (Figure 3, A and B), but we found a striking requirement for sid-5 for embryonic release.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Maternal and zygotic Sid-dependence of inherited silencing. (A) Fraction of unc-22 silenced cross progeny from sid-1; sid-2 double-mutant hermaphrodites first PC-injected with unc-22 dsRNA and then crossed to either wild-type, sid-1; sid-2 double mutant, or sid-2 single mutant males. (B) Fraction of unc-22 silenced cross progeny from sid-1; sid-3 double-mutant hermaphrodites first PC-injected with unc-22 dsRNA and then crossed to wild-type, sid-1; sid-3 double mutant, or sid-3 single mutant males. (C) Schematic of the injection, cross, and unc-22 silencing scoring of cross progeny from sid-1; sid-5 double mutant hermaphrodites first PC-injected with unc-22 dsRNA and then crossed to wild-type males. sid-5 is X-linked, thus hermaphrodite progeny are heterozygous and males are hemizygous. Error bars in (A and B) represent SD from four, seven, and five injected hermaphrodites in (A) and one, three, and three injected hermaphrodites in (B). Three injected hermaphrodites in (C). n.s. = not significant.

SID-5 is an endosome-associated protein required for efficient systemic RNAi (Hinas et al. 2012). In our double mutant rescue experiments, almost no silencing was observed in the next generation if sid-5 is not rescued (Figure 3C), the same effect as if sid-1 is not rescued. Because sid-5 is located on the X-chromosome, crossing a sid-1^−/−; sid-5^−/− hermaphrodite to a wild-type male results in heterozygous sid-5^+/− hermaphrodite progeny and hemizygous sid-5^0/- mutant male progeny. When performing the unc-22 dsRNA PC injection and rescue in this context, we saw that 100% of hermaphrodite progeny but few of the male progeny showed the unc-22 silencing phenotype (Figure 3C). The importance of SID-5 in the context of sid-1-independent transmission of RNA is especially surprising given the previous reports of only weak systemic RNAi defects for sid-5 mutants (Hinas et al. 2012). However, those experiments were performed in the context of fully functional and normally expressed SID-1, in which case there may be alternative routes for RNA to reach the cytoplasm that are less dependent on SID-5.

Labeled nucleotides can be used to visualize functional transported dsRNA

Efforts to visualize labeled dsRNA to date have been hampered by various technical challenges, primarily the observation that large bulky fluorescent moieties interfere with RNAi and dsRNA transport, and methods that either fail to label 100% of the dsRNA molecules or only label the ends produce false positive results. Further, it is essential to demonstrate that fluorescence remains associated with dsRNA. To address these issues, we synthesized unc-22 dsRNA (500 bp) using 5EU nucleotides, replacing a significant fraction of the normal uridine nucleotides with nucleotides carrying the modified base; 5EU carries a small alkyne modification that allows the RNA to be easily visualized after fixation and labeling through click chemistry (Jao and Salic 2008). 5EU-labeled unc-22 dsRNA injected into a single gonad arm of a wild-type worm results in >85% twitching progeny, indicating that, like unmodified dsRNAs, 5EU dsRNA is both mobile and capable of silencing target genes (Figure 4A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Visualizing 5EU labeled functional dsRNA. (A) Injected 5EU dsRNA injected into only one gonad arm produces >50% affected progeny; n = 8 injected hermaphrodites. (B) Localization of PC injected Cy5::5EU heteroduplex dsRNA. Cy5 fluorescence and 5EU detection colocalize in the pseudocoelom (i) and a coelomocyte (cc) (ii; white arrowheads). (C) Independent localization of PC injected 5EU and Cy5 labeled dsRNA. Images in (B) and (C) represent portions of dissected and partially flattened adult hermaphrodites. Thick dotted lines mark the boundary of the animal, and thinner dotted lines mark structures such as the gonad (gon) or intestinal cells (int) as landmarks for orientation. (D, E) Cy5 and 5EU signal in embryos collected from adults injected with the dsRNA species described in (B) and (C), respectively. The two Cy5 images are overexposed, revealing diffuse autofluorescence and no detectable RNA. Bars, 10 μm.

To determine whether fluorescently labeled 5EU signal represents intact dsRNA and not degraded nucleotides or other irrelevant species, we used a second label and synthesized dsRNA composed of one strand labeled with 5EU and the other strand internally labeled with Cy5-uridine. Although Cy5-labeled dsRNA is not capable of systemic RNAi, it suffices for demonstrating the properties of labeled dsRNA. We injected the Cy5:5EU duplex dsRNA into the PC of adult hermaphrodites, and, several hours later, fixed the animals to visualize both labels (Figure 4B). Much of the signal appears in the pseudocoelom as pools surrounding other tissues. Importantly, both labels were detected in the same locations and at similar intensities. Coinciding labeled RNA was also detected in punctate foci in coelomocytes, providing further evidence for the structure and stability of the injected RNA.

In addition to the Cy5:5EU duplex, we also injected a 1:1 mixture of Cy5-labeled dsRNA with 5EU-labeled dsRNA. We found that, although these mixed dsRNAs were both readily detectable in the PC, their apparent relative abundances were more disparate than the colabeled duplex dsRNA, and, more frequently, one label could be seen without the other (Figure 4C). That is, when not physically bound together, each label appears free to vary independently. Together, these data suggest that 5EU fluorescence represents bona fide dsRNA as opposed to disassociated strands or degraded nucleotides or labels.

We also attempted to detect labeled RNA in embryos from PC-injected hermaphrodites. The Cy5:5EU heteroduplex was not detected above background in embryos (Figure 4D). Furthermore, while we could detect 5EU-labeled RNA punctae in embryos, the coinjected Cy5-labeled RNA was only detected in the PC and coelomocytes (Figure 4E). These observations are consistent with Cy5-containing RNA being impaired for normal trafficking while 5EU-labeled RNA behaves as expected for unmodified RNA. The inability of either the Cy5:5EU heteroduplex or Cy5-RNA to enter embryos suggests there is selectivity in the SID-1 and RME-2-dependent transmission processes. Such selectivity via RME-2 mediated uptake is not expected if dsRNA is simply hitchhiking on yolk proteins.

Labeled dsRNA and VIT-2::GFP largely fail to colocalize

RME-2 was initially described as a yolk receptor (Grant and Hirsh 1999), thus the observation that RME-2 may function in parallel to SID-1 to transport dsRNA into oocytes and embryos suggests an association between RNA and yolk (Marré et al. 2016; Figure 2). We used 5EU-labeled dsRNA to examine interactions between yolk and dsRNA. The Marré et al. study also examined interactions between yolk and a 50 bp end-labeled dsRNA (Marré et al. 2016). However, these studies did not control for interactions between yolk and the dye or control in vivo for integrity of the dye and dsRNA.

We injected our 5EU-labeled dsRNA into the PC of adults expressing GFP-labeled yolk (vit-2::GFP) and imaged the injected adults as well as isolated embryos. We detected 5EU-labeled dsRNA colocalized with VIT-2::GFP granules at the surface of developing oocytes in adult gonads (Figure 5A). This colocalization is consistent with codependence on oocyte-expressed RME-2 for the uptake of yolk and RNA, as reported previously (Marré et al. 2016). However, interior optical sections contain 5EU punctae not associated with any yolk, suggesting that although yolk and dsRNA may share the same endocytosis receptor, they are either separately endocytosed or sorted to different endosomes soon after import (Figure 5A). Analysis of 5EU-labeled dsRNA and VIT-2::GFP in embryos is consistent with this view. The majority of detectable 5EU fluorescence in 1–2 cell embryos does not colocalize with GFP fluorescence (Figure 5B). The lack of colocalization might reflect sid-1-dependent release of dsRNA from endosomes. However, in sid-1^−/− mutant embryos labeled dsRNA also fails to colocalize with VIT-2::GFP (Figure 5B). This further supports the idea that, although dsRNA and yolk share a common mechanism for endocytosis into oocytes, they likely do not share common endosomal compartments. This contrasts with the observations and conclusions of Marré et al. (2016), who observed extensive colocalization between VIT-2::GFP and their end-labeled dsRNA in oocytes and early embryos, but segregation during later embryogenesis. These observations led them to conclude that mechanisms exist to separate the two components from within a common vesicle (Marré et al. 2016). The difference between the results presented here and those previously reported is most likely related to differing dsRNA labeling strategies (see Discussion).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Colocalization of 5EU-labeled dsRNA and VIT-2::GFP on and within oocytes. (A) PC-injected 5EU-labeled dsRNA colocalized with GFP-labeled yolk at the surface of developing oocytes (white arrowheads), but not intracellularly (notched arrowheads). More proximal oocyte (top) contains more VIT-2::GFP. (B) Maximum z-projections of VIT-2::GFP and 5EU-labeled dsRNA in wild-type and sid-1^−/− embryos show little colocalization. Bars, 10 μm.

dsRNA transits RAB-7-containing vesicles independent of SID-1

Since internalized dsRNA appears to be physically separate from yolk granules, we wondered whether the 5EU punctae might colocalize with the late endosome marker RAB-7 (Feng et al. 1995; Grant and Hirsh 1999). We injected adult hermaphrodites in the PC with 5EU-labeled dsRNA, and visualized the resulting embryos in conjunction with antibody staining against GFP-labeled RAB-7. In wild-type embryos, we occasionally detected dsRNA together with RAB-7-positive vesicles, but the majority of the 5EU punctae were not associated with GFP-RAB-7 (Figure 6A, upper row). In sid-1 mutant embryos, 5EU punctae are also readily detected outside of RAB-7 vesicles, and there does not appear to be significantly more 5EU retained within RAB-7 vesicles (Figure 6A, lower row). Thus, although silencing by inherited dsRNA involves early steps in the endocytotic pathway, the dsRNA appears to transit this system rapidly. Although sid-1 is required for inherited silencing, its activity is not apparent in the localization of 5EU-labeled dsRNA in oocytes or early embryos.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Colocalization of dsRNA and GFP::RAB-7. (A) 5EU foci detected in embryos after pseudocoelomic 5EU-labeled dsRNA injection colocalizes with GFP:: RAB-7 (notched arrowheads) but is also found outside of RAB-7 structures (open arrowheads) in both wild-type (upper row) and sid-1 mutant (lower row) embryos. Bars, 10 μm. (B) Model for three inherited dsRNA transport pathways. (1) DsRNA injected directly into the syncytial germline can silence the resulting progeny without SID-1. Some injected dsRNA exits the gonad to the PC and is then directly or indirectly via yolk (2) endocytosed into developing oocytes via LDL receptor super-family member RME-2, or (3) this PC dsRNA can also be directly transported into oocytes via SID-1.