DNA vectors that express short hairpin RNAs (shRNAs) from RNA polymerase III (Pol III) promoters are a promising new tool to reduce gene expression in mammalian cells. shRNAs are processed to small interfering RNAs (siRNAs) of 21 nucleotides (nt) that guide the cleavage of the cognate mRNA by the RNA-induced silencing complex. Although siRNAs are thought to be too short to induce interferon expression, we report here that a substantial number of shRNA vectors can trigger an interferon response.
MORF4L1 and MORF4L2 are closely related genes involved in chromatin regulation1,
2. We inserted oligonucleotides encoding MORF4L1 and MORF4L2 shRNAs into the pSUPER vector3 and transferred the Pol III promoter−shRNA cassettes into a lentiviral vector4 (Fig. 1a). Quantitative PCR showed that viruses targeting MORF4L1 (vector pAB319) and MORF4L2 (pAB322) reduced the level of the corresponding transcript to 3% of the normal level in human lung fibroblasts (HLFs). Microarray analysis of cells infected with pAB319 and pAB322 showed induction of many interferon target genes (Supplementary Table 1 online). To determine the cause, we carried out quantitative PCR for mRNA encoding 2'5'-oligoadenylate synthetase (OAS1), a classic interferon target gene, on cells expressing each shRNA separately. OAS1 expression was induced >50-fold by the pAB319 MORF4L1 vector alone, whereas the pAB322 MORF4L2 vector alone had no effect (Fig. 1b). Together, the two vectors induced OAS1 expression >500-fold. To distinguish between a MORF4L1-specific effect and an 'off-target' effect, we tested several other MORF4L1 shRNA vectors. Despite reducing MORF4L1 expression by up to 95%, none induced OAS1 (Fig. 1c). To further test if OAS1 induction was a consequence of MORF4L1 inactivation per se, we transduced HLFs with a lentivirus expressing a MORF4L1 cDNA with a mutated shRNA cleavage site. This mutated cDNA restored MORF4L1 expression to wild-type levels in the presence of the pAB319 vector but did not prevent OAS1 induction (Fig. 1d).
(a) Schematic diagram showing the structure of the shRNA lentiviral vector. (b−f) Quantitative PCR assays. (b−d) Lentiviral infection of HLFs. (b) OAS1 is induced by the vector that targets MORF4L1 (pAB319). (c) Induction of OAS1 does not correlate with MORF4L1 silencing (shRNA 1 is pAB319; the sequence of all five shRNAs is given in the Supplementary Note online). (d) OAS1 induction by MORF4L1 shRNA is not suppressed by restoration of MORF4L1 expression with an uncleavable MORF4L1 cDNA. shRNA: pAB319; cDNA: pAB349 (the cleavage site contains 6 mismatches). (e) Transfection of HLFs with the MORF4L1 siRNA encoded by pAB319 does not induce OAS1. (f) Transfection of HeLa cells with plasmid vectors induces OAS1. The MORF4L1 and MORF4L2 plasmids encode the shRNAs expressed by pAB319 and pAB322. The reduction in MORF4L1 mRNA with the MORF4L2 plasmid is expected from the sequence. (g) Northern blotting for shRNAs targeting MORF4L1 and MORF4L2 of RNA extracted from HLFs infected with lentiviral vectors shows the presence of only the correctly processed 21-nt siRNAs (mock, no virus; blank, pAB303; MORF4L1, pAB319; MORF4L2, pAB322).
We next tested whether an interferon response could be induced by a chemically synthesized siRNA duplex corresponding to the shRNA species produced by the lentiviral vector pAB319. Transfection of HLFs with this siRNA duplex reduced the level of MORF4L1 mRNA by 80% but did not induce OAS1 (Fig. 1e). At this level of silencing, there was also no induction of OAS1 by the lentiviral vector pAB319, so we cannot rule out the possibility that a subset of siRNAs may directly trigger an interferon response.
To test whether the interferon induction was specific to lentiviral vectors, we transfected HeLa cells with pSUPER-based plasmid vectors expressing shRNA targeting LMNA, MORF4L1 and MORF4L2. All of the plasmid vectors including the Bluescript control and pSUPER induced OAS1, consistent with reports that transfection of plasmid-based expression vectors can induce an interferon response5. The pSUPER vector expressing the MORF4L1 shRNA consistently induced higher levels of OAS1 expression than the other plasmids (Fig. 1f), indicating that this shRNA species can induce OAS1 when expressed from both plasmid and lentiviral systems.
In addition to pAB319, we observed OAS1 induction with 7 of 23 U6 promoter/lentivirus shRNA constructs targeting different genes. Compared with the H1 vectors, the orientation of the Pol III cassette was reversed and a shorter termination sequence was present in the U6 vectors (see Supplementary Note online). The greater frequency of OAS1 induction with the U6 vectors could be explained by the formation of long double-stranded RNA (dsRNA) molecules between transcripts from the truncated LTR promoter and the U6 promoter. To rule this out, we inverted the orientation of the U6 cassette using a Gateway lentiviral construct for two of the OAS1-inducing U6 vectors (see Supplementary Note online). Both shRNAs still induced OAS1, indicating that the effect does not depend on the orientation of the Pol III cassette (data not shown).
The magnitude of OAS1 induction was greater at higher multiplicities of infection with all of the vectors tested. Together with the observation that the MORF4L2 vector was able to potentiate OAS1 induction by pAB319 (Fig. 1b), the dependence on vector dose is consistent with a model in which shRNAs compete for processing to siRNA, and the accumulation of unprocessed or aberrantly processed Pol III transcripts triggers interferon expression. Northern blotting of cells expressing the pAB319 and pAB322 vectors detected only the presence of the correctly processed siRNAs (Fig. 1g), but this assay may not be sensitive enough to detect the small amounts of dsRNA that are sufficient to trigger an interferon response.
In conclusion, we show that a commonly used shRNA construct can induce an interferon response. Although this may not be a concern in initial screens with shRNA banks, we recommend testing for interferon induction before attributing a particular response to the gene targeted. One simple precaution to limit the risk of inducing an interferon response is to use the lowest effective dose of shRNA vector. Finally, we note that many commonly used tumor cells have a defective interferon response6, which may explain why these effects have not previously been reported.
Accession numbers. The GEO accession numbers for the microarray data are GSM3891 and GSM3892.
Acknowledgments We thank P. Gonczy for critical reading of the manuscript, T. Brummelkamp for supplying pSUPER, the Swiss Institute for Experimental Cancer Research microarray and bioinformatics core facilities for technical assistance and advice on data analysis and the Swiss National Science Foundation NCCR Molecular Oncology Programme and MEDIC Foundation for financial support.
Competing interests statement:
The authors declare that they have no competing financial interests.
