Interactions between the non-seed region of siRNA and RNA-binding RLC/RISC proteins, Ago and TRBP, in mammalian cells

Abstract

Small interfering RNA (siRNA)-based RNA interference (RNAi) is widely used for target gene silencing in various organisms. We previously showed that 8-nt-long 5' proximal nucleotides, which include seed sequence (positions 2-8 from the 5' end of guide strand), and the complementary sequence of the passenger strand are capable of being simultaneously replaced with cognate deoxyribonucleotides without any substantial loss of gene silencing. In the present study, examination was made of RNA requirements in the non-seed region of siRNA. The non-seed region of siRNA was found to be subdivided into four domains, in which two nucleotide pairs (positions 13 and 14) were replaceable with cognate deoxyribonucleotides without reducing RNAi activity. However, RNA sequences at positions 9-12 and 15-18 were essential for effective gene silencing, and these two double-stranded RNA cores are required for binding of the trans-activation response RNA-binding protein (TRBP). The terminal RNA (positions 19-21) provided Argonaute protein binding sites. Argonaute binding was enhanced by the presence of RNAs at positions 15-18. Knockdown experiments showed that, unlike Argonaute and TRBP, Dicer was dispensable for RNAi. Based on these observations, we discuss possible RNA/protein and protein/protein interactions in RNA-induced silencing complex formation.

Figure 1.

DNA-substitution–dependent changes in target gene silencing due to DNA-modified siRNA (A) and asymmetric RISC formation in DNA-modified siRNA-dependent RNAi (B and C). (A) Effects of dsDNA replacement on RNA silencing activity are shown in the upper panel. The red circle indicates ribonucleotide, whereas blue one indicates deoxyribonucleotide. In each siRNA with or without DNA modification, passenger and guide strands are shown on the top and bottom, respectively. The RNA silencing activity of siRNA with <9 bp DNA substitution from the 5′ end of guide strand is essentially identical to that of nonmodified siRNA (49). Structure of active siRNA with a dsDNA seed duplex, ranging from position 2 to 8, is shown in the lower half. All ribonucleotides in the seed duplex are replaceable with cognate deoxyribonucleotides without any substantial loss of RNAi activity (49). Most, if not all, nucleotides in the remaining (nucleotide position 9–21) should be ribonucleotides, possibly due to binding of RNA-recognition proteins in RLC/RISC. Note that, because of the asymmetry in structure, virtually no off-target effect is associated with siRNA with a dsDNA seed duplex (49). (B) and (C), respectively, show the RISC formation in RNAi due to siRNA with a dsDNA seed duplex and that due to nonmodified siRNA. (B) In the former, the guide strand is recruited to Ago2 to generate a functional RISC but the passenger strand may not be recruited to Ago2 because of asymmetry in structure. (C) In contrast, both guide and passenger strands of the nonmodified siRNA can be recruited to Ago2. Note that the nonmodified siRNA is structural symmetric.

Figure 2.

Effects of 2-bp-long DNA substitutions in the non-seed region of siRNA (siLuc2-153) with a dsDNA seed duplex on RNAi activity as determined by luc reporter assays. Except for DNA/RNA difference, the nucleotide sequences of DNA-modified siRNAs used here were identical to that of siLuc2-153. Red circle, ribonucleotide; blue circle, deoxyribonucleotide. (A) Both guide and passenger strands were simultaneously replaced with DNA. (B) DNA replacement was carried out only in the guide strand. (C) DNA replacement was carried out only in the passenger strand. RNAi was assayed using human HeLa, mouse E14TG2a, Chinese hamster CHO-K1 and Drosophila S2 cells using a dual-luciferase reporter assay system with nonmodified or DNA-substituted siRNAs at 50 nM. However, in S2 RNAi (B), the siRNA concentration used was 5 nM. The dotted line in each graph indicates the level of RNAi due to the parental modified siRNA. Lane 1, siRNA with a dsDNA seed duplex. Lane 2–lane 13, siRNA with a dsDNA-modified seed duplex and two additionally DNA-substituted base pairs in the non-seed-duplex region.

Figure 3.

Effects of DNA substitutions in the non-seed-duplex subdomains A, B, C and T on RNAi activity. Except for DNA/RNA difference, the nucleotide sequences of DNA-modified siRNAs were identical to that of siLuc2-153. Red circle, ribonucleotide; blue circle, deoxyribonucleotide. Seed duplex, nucleotide position 2–8; subdomain A, position 9–12; subdomain B, position 13–14; subdomain C, position 15–18; subdomain T, position 19–21. RNAi activity was assayed using HeLa, E14TG2a, CHO-K1 and S2 cells, using a dual-luciferase reporter assay system. The concentration of siRNA was 50 nM. Lane 1, nonmodified siRNA; lane 2, TCBA-RNA (the entire non-seed duplex is RNA); lane 3, TCA-RNA; lane 4, CA-RNA; lane 5, A-RNA; lane 6, TC-RNA; lane 7, C-RNA; lane 8, T-RNA; lane 9, RNA in the guide strand non-seed region; lane 10, RNA in the passenger non-seed region; lane 11, siDNA.

Figure 4.

Effects of Ago2, Dcr, TRBP and PACT knockdown on luc RNAi due to DNA-modified or nonmodified siRNA. Except for DNA/RNA difference, the nucleotide sequences of DNA-modified siRNAs were identical to that of siLuc2-153 or siLuc-36 as indicated. Red circle, ribonucleotide; blue circle, deoxyribonucleotide. (A) Knockdown efficiency of siAgo2, siDcr, siTRBP or siPACT was examined by real-time reverse transcriptase-PCR in HeLa cells. Changes in expression level of each mRNA normalized with that in the cells treated with siControl were measured following the addition of 25 nM siAgo2, siDcr, siTRBP or siPACT. (B and C) Relative change in IC₅₀s of Ago2, Dcr, TRBP and PACT knockdown of siLuc2-153 (B) and siLuc-36 (C). Relative IC₅₀ was normalized using that for RNAi due to control siRNA, siGY441. IC₅₀ values were calculated using the dual-luciferase reporter assay data in Supplementary Figures S3 and S4. P-values were determined by Student’s t-test (**P < 0.01, *P < 0.05).

Figure 5.

EMSA profiles of DNA-substituted siRNAs associated with recombinant PAZ domain of human AGO1. EMSA experiments were carried out using recombinant human AGO1 PAZ domain with ³²P-labeled authentic siLuc-36 or its DNA-substituted derivatives. Relative fractions of bound siRNA are shown by numerals on the left. Lane 1, nonmodified siRNA; lane 2, TCA-RNA; lane 3, CA-RNA; lane 4, A-RNA; lane 5, TC-RNA; lane 6, C-RNA; lane 7, T-RNA; lane 8, RNA in the guide strand non-seed region; lane 9, RNA in the passenger non-seed region; lane 10, 10-nt 3′ protrusion of passenger strand; lane 11, 5-nt 5′ protrusion of passenger strand; lane 12, siDNA.

Figure 6.

EMSA of DNA-substituted siRNAs associated with recombinant human TRBP protein. Except for DNA/RNA difference, the nucleotide sequence of DNA-modified siRNA was identical to that of siLuc2-153. Red circle, ribonucleotide; blue circle, deoxyribonucleotide. (A and B) EMSA experiments. Purified recombinant TRBP and ³²P-labeled siRNA with DNA substitutions (0.5 nM) were incubated with increasing amounts of wild-type TRBP protein, as indicated. In both (A) and (B), seed duplex region was replaced with dsDNA. In (B), dsDNA substitution was also carried out in subdomain B. (C and D) A bound fraction (%) of siRNA was plotted against the input concentration of TRBP. Note that the binding activities of DNA-modified siRNAs were significantly lower than that of the nonmodified siRNA. Structures of DNA modifications are shown in the right margin (C and D). In (C), observed TRBP concentrations giving 50% binding of nonmodified siRNA and siRNA with a dsDNA seed duplex are also shown in the right margin.

Figure 7.

Presumed subdomain structure of siRNA (A) and possible models for functional RISC formation (B) A deduced subdomain structure of siRNA is shown in (A). Subdomain A is required for TRBP binding. Subdomain B is a region replaceable with DNA without a substantial loss of RNAi activity. Subdomain C is not only required for TRBP binding but also serves as an enhancer for Ago2/siRNA binding. Subdomain T provides Ago2 binding. (B) Possible case of RISC formation is depicted. A model consists of three steps. Pre-binding step includes an unidentified factor, ‘X’, which is presumed to be capable of binding to TRBP and recruiting it to siRNA. Note that one of the possible candidates of ‘X’ might be Ago. Dcr is totally dispensable or redundant in function to ‘X’. Although in some cases, RNAi due to siRNA with a dsDNA seed duplex may include neither ‘X’ nor Dcr. In pre-RISC, DNA-modified siRNA duplex is recognized with the PAZ domain of Ago and, in the mature RISC, consisting of Ago2 and the guide strand, is formed.