RNA interference by expression of short-interfering RNAs and hairpin RNAs in mammalian cells

Abstract

Duplexes of 21-nt RNAs, known as short-interfering RNAs (siRNAs), efficiently inhibit gene expression by RNA interference (RNAi) when introduced into mammalian cells. We show that siRNAs can be synthesized by in vitro transcription with T7 RNA polymerase, providing an economical alternative to chemical synthesis of siRNAs. By using this method, we show that short hairpin siRNAs can function like siRNA duplexes to inhibit gene expression in a sequence-specific manner. Further, we find that hairpin siRNAs or siRNAs expressed from an RNA polymerase III vector based on the mouse U6 RNA promoter can effectively inhibit gene expression in mammalian cells. U6-driven hairpin siRNAs dramatically reduced the expression of a neuron-specific beta-tubulin protein during the neuronal differentiation of mouse P19 cells, demonstrating that this approach should be useful for studies of differentiation and neurogenesis. We also observe that mismatches within hairpin siRNAs can increase the strand selectivity of a hairpin siRNA, which may reduce self-targeting of vectors expressing siRNAs. Use of hairpin siRNA expression vectors for RNAi should provide a rapid and versatile method for assessing gene function in mammalian cells, and may have applications in gene therapy.

Figure 1

RNA interference using 21-nt siRNAs synthesized by in vitro transcription. (A) Sequences and expected duplexes for siRNAs targeted to GFP. Both DhGFP1 strands were chemically synthesized, whereas other siRNA strands were synthesized by in vitro transcription with T7 RNA polymerase. GFP5m1 contains a two-base mismatch with the GFP target. Nucleotides corresponding to the antisense strand of GFP are in bold; nucleotides mismatched with the target are lowercase. (B) An example of a DNA oligonucleotide template for T7 transcription. (C) GFP fluorescence was effectively reduced by cotransfection of either the DhGFP1 or GFP5 siRNAs with a GFP expression vector but not by the GFP5m1 siRNA. (D) siRNA inhibition of luciferase activity from vectors with and without GFP sequences inserted into the 3′ untranslated region of luciferase (luc, luciferase; pA, SV40 polyadenylation site). siRNAs synthesized either chemically or by in vitro transcription were similarly effective at inhibiting luciferase when GFP sequences were present in the luciferase mRNA, whereas the mismatched GFP5m1 siRNA did not inhibit effectively. The no-siRNA control (none) is set to 100% for each set of transfections. Data are averaged from three experiments with SE indicated.

Figure 2

RNA interference using hairpin siRNAs synthesized by in vitro transcription. (A) Sequences and expected structures for the hairpin siRNAs to GFP (notation as in Fig. 1). GFP5HP1m2 and GFP5HP1m3 contain single-base mismatches with the sense and antisense strands of GFP respectively, whereas GFP5HP1m1 contains a two-base mismatch identical to GFP5m1 (see Fig. 1A). (B–D) Hairpin siRNA inhibition of luciferase activity (see legend for Fig. 1D). (B) CS2+luc was not inhibited by hairpin siRNAs. (C) GFP5HP1 and GFP5HP1S inhibited luciferase from both sense and antisense targets. The GFP5HP1m1 hairpin did not effectively inhibit luciferase activity from vectors containing either strand of GFP in the luciferase mRNA, whereas GFP5HP1m2 and GFP5HP1m3 reduced inhibition only for the mismatched strand. (D) Denaturation (dn) of the GFP5 siRNA reduced inhibition of a luciferase-GFP target, whereas denaturation of GFP5HP1 did not significantly alter inhibition.

Figure 3

RNA interference with neuronal β-tubulin using in vitro-synthesized siRNAs and hairpin siRNAs. (A) Sequences and expected structures for the siRNAs and hairpin siRNAs against neuronal β-tubulin (notation as in Fig. 1). (B) GFP fluorescence and neuronal β-tubulin expression detected by indirect immunofluorescence in mouse P19 cells 4 days after cotransfection with biCS2+MASH1/GFP and various siRNAs. GFP5 reduced GFP expression to undetectable in most cells without altering neuronal β-tubulin (NT), whereas BT4 and BT4HP1 reduced the number of neuronal β-tubulin-expressing cells without altering GFP expression. The mismatched siRNA BT4HP1m1 had no effect on GFP or neuronal β-tubulin. (C) Number of cells per field of view expressing detectable neuronal β-tubulin or the HuC/HuD neuronal RNA-binding proteins detected by indirect immunofluorescence after cotransfection of biCS2+MASH1/GFP and BT4, BT4HP1, or BT4HP1m1 siRNAs. An average from three fields is shown for each transfection. Neuronal β-tubulin and HuC/HuD were scored in parallel transfections, and cell numbers were normalized to the number of GFP-expressing cells in each field to control for transfection efficiency. Data are from a representative experiment.

Figure 4

RNA interference using siRNAs and hairpin siRNAs expressed in mouse P19 cells from a U6 RNA polymerase III promoter. (A) An example of the transcribed region of a mouse U6 promoter siRNA vector (U6-BT4as). The first nucleotide of the U6 transcript corresponds to the first nucleotide of the siRNA, whereas the siRNA terminates at a stretch of 5 T residues in the vector. (B) Sequences for the siRNAs and hairpin siRNAs to neuronal β-tubulin synthesized from the U6 vector. Expected RNA duplexes are shown for the hairpin siRNAs and for pairs of single-strand siRNAs (notation as in Fig. 1). (C) GFP fluorescence and indirect immunofluorescence for neuronal β-tubulin (NT) 4 days after cotransfection of the indicated U6 vectors and biCS2+MASH1/GFP. (D) Number of cells with detectable neuronal β-tubulin and HuC/HuD after cotransfection of biCS2+MASH1/GFP and various U6 vectors (using parallel transfections; details as in Fig. 3). The expression of either siRNA hairpin reduces the number of neuronal β-tubulin positive cells about 100-fold, whereas cotransfection of two vectors expressing individual siRNA strands reduces the number of neuronal β-tubulin cells about 5-fold. HuC/D expression is not altered by the hairpin siRNAs.