Dual-targeting siRNAs

Abstract

We have developed an algorithm for the prediction of dual-targeting short interfering RNAs (siRNAs) in which both strands are deliberately designed to separately target different mRNA transcripts with complete complementarity. An advantage of this approach versus the use of two separate duplexes is that only two strands, as opposed to four, are competing for entry into the RNA-induced silencing complex. We chose to design our dual-targeting siRNAs as Dicer substrate 25/27mer siRNAs, since design features resembling pre-microRNAs (miRNAs) can be introduced for Dicer processing. Seven different dual-targeting siRNAs targeting genes that are potential targets in cancer therapy have been developed including Bcl2, Stat3, CCND1, BIRC5, and MYC. The dual-targeting siRNAs have been characterized for dual target knockdown in three different cell lines (HEK293, HCT116, and PC3), where they were as effective as their corresponding single-targeting siRNAs in target knockdown. The algorithm developed in this study should prove to be useful for predicting dual-targeting siRNAs in a variety of different targets and is available from http://demo1.interagon.com/DualTargeting/.

FIGURE 1.

Analysis of dual-targeting siRNAs of the first generation. (A) Quantitative real-time PCR performed for dual-targeting siRNAs against BCL6 (black bars) and STAT3 (gray bars) mRNA extracted from HEK293 cells 48 h after transfection. (B) Quantitative real-time PCR performed for dual-targeting siRNAs against Bcl6 (black bars), STAT3 (light gray bars), MYC (dark gray bars), and Bcl2 (open bars) mRNA extracted from HEK293 cells 48 h after transfection.

FIGURE 2.

RISC entry, potential for miRNA-like regulation, and sequence similarity to highly effective standard siRNAs explain observed siRNA efficacy. Observed relative target RNA levels for first generation candidate dual-targeting siRNAs (Fig. 1) plotted against features related to siRNA efficacy. Lines are linear regression models; numbers in the lower-left corners show the proportion of variability in the data explained by the model (r²) and the model's P-value (p). Features are (A,B) target site accessibility ([A] target site accessibility and [B] target site's mean probability of having unpaired nucleotides); (C,D) duplex thermodynamics ([C] duplex stability and [D] target strand stability); (E,F) sequence similarity to highly effective standard siRNAs (scores from [E] GPboost and [F] Reynolds algorithms); (G–K) difference in the duplex ends' thermodynamic stability (ΔΔG; ΔΔG computed by considering 1 nt [G] to 5 nt [K] at the ends of the 19mer siRNA strands); (L) RISC entry, which indicates whether the strand is favored (1), unfavored (−1), or undecided (0) for RISC uptake; calculations were based on ΔΔG 5 nt; (M) potential for miRNA-like regulation (miRNA score); and (N) the sum of the RISC entry, GPboost, and miRNA score values.

FIGURE 3.

Dicer processing of dual-targeting siRNAs. Dicer substrates are processed into smaller molecules by recombinant Dicer. The indicated siRNAs (A) B2S-4543-27, MB2-2019-27, sykB2-700-27, (B) E2D1-733-27, ED1-2880-27, and (C) E2B2-688-27 and surD2-2049-27 were radioactively 5′ labeled either at the first antisense strand (1) or at the second antisense strand (2) and incubated in the presence (+) or absence (−) of recombinant human Dicer. The sequence shown was incubated with either NaOH for the alkali ladder or with RNase T1 under denaturing conditions for the G-ladder. The products were resolved on a 20% denaturing polyacrylamide gel.

FIGURE 4.

Validation of the dual-targeting rational design algorithm. Quantitative real-time PCR performed for dual-targeting siRNAs B2S-4543-27, MB2-2019-27, and sykB2-700-27 (A) against BCL2 (black bars), STAT3 (medium gray bars), MYC (light gray bars), and SYK (open bars), E2B2-688-27 and E2D1-733 (B) against CCNE2 (light gray bars), BCL2 (black bars), CCND1 (medium gray bars), surD2-2049-27 and ED1-2880-27 (C) against BIRC5 (survivin) (black bars), and CCND2 (medium gray bars), EGFR (dark gray bars), and CCND1 (light gray bars). The mRNA was extracted from HEK293, HCT116, and PC3 cells 48 h after transfection. Experiments were performed in independent triplicates.

FIGURE 5.

Influence of dual-targeting siRNAs on the interferon response. Real-time PCR was performed for IFNB1 (black bars), CDKL2 (light gray bars), and OAS1 (dark gray bars) mRNAs extracted from HEK293 cells 48 h after transfection. As a positive control, HEK293 cells were incubated with 5 nM RNA obtained after in vitro T7-transcription or 50 ng poly I:C for 24 h. Data were normalized to RPLP0 and are shown relative to the mock transfected control. Experiments were done in triplicate.

FIGURE 6.

Comparison of dual-targeting siRNAs E2D1-733-27 and E2B2-688-27 with their single-targeting equivalents. HEK293 cells were transfected with the dual-targeting E2D1-733-27 and E2B2-688-27 and their single-targeting equivalents E2D1-E2-27, E2D1-D1-27, E2B2-E2-27, and E2B2-B2-27. Real-time PCR was performed with RNA extracted 48 h after transfection. RPLP0 was used as an internal control. Bcl2 is shown in black, Cyclin E2 in medium gray, and Cyclin D1 in light gray. Experiments were performed in independent triplicates.