Secondary siRNAs Result from Unprimed RNA Synthesis and Form a Distinct Class

See allHide authors and affiliations

Science  12 Jan 2007:
Vol. 315, Issue 5809, pp. 244-247
DOI: 10.1126/science.1136699


In Caenorhabditis elegans, an effective RNA interference (RNAi) response requires the production of secondary short interfering RNAs (siRNAs) by RNA-directed RNA polymerases (RdRPs). We cloned secondary siRNAs from transgenic C. elegans lines expressing a single 22-nucleotide primary siRNA. Several secondary siRNAs start a few nucleotides downstream of the primary siRNA, indicating that non–RISC (RNA-induced silencing complex)–cleaved mRNAs are substrates for secondary siRNA production. In lines expressing primary siRNAs with single-nucleotide mismatches, secondary siRNAs do not carry the mismatch but contain the nucleotide complementary to the mRNA. We infer that RdRPs perform unprimed RNA synthesis. Secondary siRNAs are only of antisense polarity, carry 5′ di- or triphosphates, and are only in the minority associated with RDE-1, the RNAi-specific Argonaute protein. Therefore, secondary siRNAs represent a distinct class of small RNAs. Their biogenesis depends on RdRPs, and we propose that each secondary siRNA is an individual RdRP product.

RNAi is triggered by double-stranded RNA (dsRNA) molecules that are cleaved into 20- to 25-nucleotide (nt) siRNAs, which guide cleavage of homologous RNA molecules (1). Amplification of RNAi is inferred from the spreading of RNA silencing to regions outside the inducer sequence (transitive RNAi) (24) and involves the production of secondary siRNAs for which RdRPs have been implicated in nematodes, plants, and fungi (5). In vitro, RdRPs can perform both primed and unprimed RNA synthesis (6, 7). In plants, the predominant route appears to be unprimed (5′ to 3′) RNA synthesis starting at the 3′ end of targeted transcripts and secondary siRNA production by DICER cleavages (4, 8). Endogenous secondary siRNAs (trans-acting siRNAs, produced after miRNA-mediated mRNA cleavage) occur in a 21-nt register and correspond to both the sense and the antisense strand (912). Secondary siRNAs in C. elegans differ from those in plants in two aspects: The secondary siRNAs are only found upstream of the initial dsRNA trigger (3, 13), and they are only of antisense polarity (14, 3). Whereas trigger-derived primary siRNAs are bound by RDE-1, the secondary siRNAs are bound by a different set of Argonaute proteins (15). It is unknown how secondary siRNA production in C. elegans proceeds—whether RdRP-dependent cRNA synthesis starts primed on a primary siRNA or unprimed at the primary target site. Although the size of these secondary siRNAs (20 to 25 nt) (3) suggests that they are DICER products, their single polarity is difficult to explain.

To obtain a single and well-defined entry point of transitive RNAi, we constitutively expressed a single 22-nt primary siRNA against the unc-22 body wall muscle gene from a miRNA-like transgene (fig. S1A). Worms transgenic for this 22si-gene show a twitching phenotype indicative of unc-22 silencing, and unc-22 mRNA expression is reduced by 50% (fig. S1B). Transitive RNAi in 22si lines has the characteristic C. elegans pattern, because secondary siRNAs occur only upstream of the primary target site and are only of antisense polarity (Fig. 1A). The 22si twitching phenotype is dependent on rde (RNAi-deficient) genes, such as rde-1, and not on miRNA-implicated genes, such as alg-1 or alg-2 (Table 1). Apparently, the 22siRNA functions as an siRNA, even though it is processed from a primary miRNA-like transcript (fig. S1A), and accordingly (16), 22siRNA-mediated RNAi depends on matching of nucleotides 10, 11, and 12 to the target (fig. S1, A and C to E). Secondary siRNAs are absent in mutant backgrounds that result in nonphenotypic nematodes (Table 1), confirming that RNAi amplification is required for a 22si twitching phenotype. The primary 22siRNA was detected in all mutant backgrounds tested (Table 1), including rde-1 and rde-4 that were previously shown to be deficient in producing primary siRNAs (17, 18). Size fractionation experiments revealed that in a wild-type (WT) background, the 22siRNA occurs in a complex of 100 to 150 kD that represents 22siRNA bound to RDE-1 (Fig. 1B), which is presumably the active silencing complex. miRNA complexes are larger (250 to 500 kD) (Fig. 1B) (19). In rde-1 and rde-4 mutants, the 22siRNA occurs in a 500-to 550-kD complex and cofractionates with DCR-1 (Fig. 1B). The passenger strand (the strand of an siRNA-duplex not incorporated into RISC), which is below detection in a WT background, accumulates in rde-1 and rde-4 mutants and cofractionates with the guide strand (the RISC-incorporated siRNA strand) (Fig. 1B). We infer that, in the absence of RDE-1 or RDE-4, the 22siRNA duplex remains in the DICER complex and the active silencing complex fails to be generated.

Fig. 1.

Expression of a 22-nt unc-22 siRNA from miRNA-like transgene results in unc-22 transitive RNAi that is dependent on rde-1, rde-4. (A) Secondary siRNAs in 22si transgenic lines occur only for the region upstream of the primary siRNA and are only of antisense polarity, as analyzed by RPAs with 500-nt single-stranded probes. (B) Analyses of size fractions. Immunoprecipitation (IP), RNA-IP, primer extension (PE) with 18-nt end-labeled oligonucleotides, Western blot analysis, and DICER activity assay were used to determine protein and RNA composition in the various gel filtration fractions. Arrows indicate the 22-nt products generated by primer extension or DICER activity. Empty spaces indicate that the size fraction was not tested.

Table 1.

Genetic requirements for 22si function.

BackgroundPredicted/known functionunc-22 phenotypeSecondary siRNAsView inline22si siRNAView inline22si in size fractionationView inline
WT WT Yes Yes Yes RDE-1 peak
dcr-1 dsRNase miRNAs + siRNAs YesView inlineView inline NDView inline ND ND
alg-1 Argonaute miRNA Yes ND Yes RDE-1 peak
alg-2 Argonaute miRNA Yes ND Yes RDE-1 peak
tsn-1 RISC component Drosophila Yes Yes Yes RDE-1 peak
vig-1 RISC component Drosophila Yes ND ND ND
rde-1 Argonaute siRNAs No No Yes DICER peak
rde-4 RNAi (DICER cofactor) No ND Yes DICER peak
rde-3 RNAi (nucleotidyltransferase) No No Yes RDE-1 peak
rrf-1 Somatic RNAi (RdRP) No No Yes RDE-1 peak
rde-2 Germline RNAi Yes ND Yes RDE-1 peak
mut-15 RNAi No No Yes RDE-1 peak
mut-16 Germline RNAi No No Yes RDE-1 peak
rrf-3 Suppressor RNAi (RdRP) SevereView inline Yes Yes RDE-1 peak
eri-1 Suppressor RNAi (siRNase) Severe Yes Yes RDE-1 peak
rrf-1;eri-1 RNAi + suppressor RNAi No ND Yes RDE-1 peak
rrf-1;rrf-3 RNAi + suppressor RNAi No ND Yes ND
rde-1;eri-1 RNAi + suppressor RNAi No ND ND ND
rde-1;HA::Rde-1;FLAG::Rde-4 RNAi + rescue Yes ND Yes RDE-1 peak
  • View inline* Determined by RPAs.

  • View inline Determined by primer extension analysis.

  • View inline Representative patterns in Fig. 1D.

  • View inline§ Maternal contribution may influence analysis.

  • View inline Some nonphenotypic unc-22 nematodes were observed when dcr-1—specific RNAi (feeding assay) was performed on 22si;eri-1 worms.

  • View inline Not determined.

  • View inline# Nematodes are paralyzed, which represents a severe unc-22 phenotype compared to twitching.

  • To determine the repertoire of secondary siRNAs generated during 22siRNA silencing, we cloned small RNAs by using a linker-independent protocol and sequence-selected secondary siRNAs by using biotinylated unc-22–specific oligonucleotides. Two different biotin oligonucleotides were used; oligo 1, which selects for secondary siRNAs in the 75-nt region immediately upstream of the primary siRNA (Fig. 2A); and oligo 2, which selects for primary and secondary siRNAs in a 50-nt region around the 22si target site (Fig. 2B). Consistent with ribonuclease protection assay (RPA) analyses (Fig. 1A), the cloned secondary siRNAs are of antisense polarity only and have a predominant length of 21 to 22 nt (table S1). Their production starts a few nucleotides downstream of the primary siRNA target site (Fig. 2B; table S2, column 1), which shows that non–RISC-cleaved mRNAs are among the substrates for secondary siRNA production. Because no further downstream secondary siRNAs are detected in RPA analyses (using 500-nt probes) (Fig. 1A), we infer that RdRP recruitment is restricted to the region around the primary siRNA. The generation of secondary siRNAs is a regulated process because similar sets of secondary siRNAs were cloned (Fig. 2A; table S1) in independent 22si worm lines and cloning experiments. RNA gel blots detecting secondary siRNAs showed that most of the 22si signal is RdRP-independent and represents primary siRNAs (fig. S2A). The secondary siRNAs appear to be in a phased register, suggesting that secondary siRNA production occurs via consecutive steps: either by progressive cleavages on a cRNA (by, e.g., DICER) or by multiple rounds of amplification (secondary siRNAs invoking tertiary siRNA production).

    Fig. 2.

    Secondary siRNA sequences, frequencies, and properties. (A) Schematic representation of secondary siRNAs cloned from three independent 22si lines with biotinylated oligo 1 (bold sequence). Secondary siRNAs are of antisense polarity. The thickness of the lines representing the secondary siRNAs indicates the number of clones. (B) Secondary siRNAs cloned from 22si line 3 in a eri-1 background with biotinylated oligo 2 (bold sequence). (C) Migration of untreated, dephosphorylated [shrimp alkaline phosphatase (SAP)–treated], monophosphorylated [SAP- and T4 polynucleotide kinase (PNK)–treated] and NAIO4-reacted, β-eliminated RNAs. (D) RNA gel blot analysis of linker ligation assays with a 5′ linker and T4 RNA ligase. M, RNA marker. Input consists of gel-isolated 15- to 30-nt RNA fraction or RNAs isolated from IPs with a hemagglutin (HA)–specific antibody on 22si;HA::RDE-1;FLAG:: RDE-4 nematodes. (E) RNA gel blot analysis of capping assay.

    DICER products have 5′-monophosphates and 2′,3′-hydroxy termini. miR-58 and 22si (which are both cleaved by DICER from an miRNA precursor) and the secondary siRNAs carry terminal 2',3′-hydroxyl groups, because NaIO4-reacted, β-eliminated RNAs (20) migrate ∼1 nt faster than the untreated RNAs (Fig. 2C and fig. S2B). Like miR-58 and 22siRNA, the secondary siRNAs contain 5′-phosphates, because alkaline phosphatase–treated RNAs migrate about half a nucleotide slower than untreated RNAs (Fig. 2C and fig. S2B). In contrast to miR-58 and 22si molecules, untreated secondary siRNAs migrate somewhat slower than their 5′-monophosphorylated derivatives (Fig. 2C and fig. S2B), suggesting that secondary siRNAs carry a different number of 5′-phosphates. Absence of a 5′-monophosphate was also inferred from ligation experiments with T4 RNA ligase, which requires a 5′-monophosphate; the 5′ linker successfully ligated to miR-58 and 22siRNA but failed to ligate to secondary siRNAs (Fig. 2D). Secondary siRNAs have 5′ di- or triphosphate because they can be capped with vaccinia virus capping enzyme, as deduced from a 2-nt slower migration (Fig. 2E). A minority of 22siRNAs can be capped and represent secondary siRNAs of the same sequence as the primary siRNA (Fig. 2E). From these findings we infer that it is unlikely that secondary siRNAs in C. elegans are products of consecutive DICER cleavages, although we cannot exclude that they are modified after dicing. Alternatively, they either are 21- or 22-nt cRNAs (short RdRP products), or they are cleaved from longer cRNAs by an alternative nuclease or by an alternative way of dicing.

    Several alternatives exist for initiation of secondary siRNA production: (i) cRNA synthesis occurs primed on 22siRNA molecules; (ii) cRNA synthesis starts unprimed at 22si-RISC–cleaved mRNA; and (iii) RdRPs are recruited to an mRNA targeted by a 22siRNA, and unprimed cRNA synthesis occurs by internal entry around the target site. Occurrence of this third route is supported by the finding that secondary siRNA production occurs a few nucleotides downstream of the primary siRNA. To investigate whether cRNA is generated by elongation of the primary siRNA, we generated two new transgenic lines in which the 22siRNA carries a single mismatch to the unc-22 target mRNA at position 21 or position 11. Replacement of the mismatches within the secondary siRNAs would show unprimed cRNA synthesis, whereas primed synthesis would result in maintenance of the mismatched nucleotide in the secondary siRNA. Both 22si-mm21 and 22si-mm11 transgenic nematodes show a weak twitching phenotype indicative of unc-22 silencing and produce relatively low levels of secondary siRNAs (Fig. 3A), possibly due to the presence of a single-nucleotide mismatch. Sequence analysis of secondary siRNAs that have overlap with the primary siRNA showed that for all these secondary siRNAs, both mm21 and mm11 had been replaced by the nucleotide complementary to the mRNA (Fig. 3, B and C; table S2). Thus, secondary siRNA production occurs only unprimed. The 22siRNA is bound by RDE-1 (Fig. 1B and fig. S3A), which carries the DDH motif needed for RISC cleavage (21). The presence of a mismatch at the presumed RISC cleavage site in the 22si-mm11 primary siRNA did not alter the range of secondary siRNAs that were cloned (Fig. 3C, table S2), suggesting that nematode RNAi does not primarily depend on RISC cleavage of mRNAs. A minority of the secondary siRNAs is loaded into the RNAi-specific argonaute protein RDE-1 (fig. S3, A and B), and most secondary siRNAs may be loaded into secondary Argonaute proteins (15) (fig. S3C). The phased register in which the secondary siRNAs appear to be ordered (Fig. 2A) may arise from multiple amplification rounds (by RDE-1–loaded secondary siRNAs) or from preferred initiation sites of cRNA synthesis.

    Fig. 3.

    Secondary siRNAs result from unprimed cRNA synthesis. (A) RNA gel blot detection of primary and secondary siRNAs in 22si- and 22si-mm10,11,12 lines. (B) Sequence of secondary siRNAs cloned from 22si-mm21 lines. The gray bar indicates mismatch 21. (C) Sequence of secondary siRNAs cloned from 22si-mm11 lines. The gray bar indicates mismatch 11.

    Secondary siRNAs represent a distinct class of 21- to 22-nt small RNAs that can can silence mRNAs in trans (3). They result from unprimed RNA synthesis (Fig. 3, B and C), are only of antisense polarity (Fig. 1A), carry 5′ di- or triphosphates (Fig. 2, C to E), and are mainly loaded into a distinct set of Argonaute proteins (15) (fig. S3). Secondary siRNAs can be produced on non–RISC-cleaved mRNAs presumably by internal entry of the RdRP. A 3′ nonprogressive DICER cleavage may be responsible for generating the secondary siRNA 3′ end and degrading the mRNA (fig. S4). We find that for the 22siRNA, a single-nucleotide mismatch decreases silencing efficiency considerably. Likely, these strict prerequisites for inducing transitive RNAi protect organisms against unintended off-target effects (22, 23).

    Supporting Online Material

    Materials and Methods

    Figs. S1 to S4

    Tables S1 and S2


    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article