Off-target effects of siRNA specific for GFP

Abstract

Background: Gene knock down by RNAi is a highly effective approach to silence gene expression in experimental as well as therapeutic settings. However, this widely used methodology entails serious pitfalls, especially concerning specificity of the RNAi molecules.

Results: We tested the most widely used control siRNA directed against GFP for off-target effects and found that it deregulates in addition to GFP a set of endogenous target genes. The off-target effects were dependent on the amount of GFP siRNA transfected and were detected in a variety of cell lines. Since the respective siRNA molecule specific for GFP is widely used as negative control for RNAi experiments, we studied the complete set of off-target genes of this molecule by genome-wide expression profiling. The detected modulated mRNAs had target sequences homologous to the siRNA as small as 8 basepairs in size. However, we found no restriction of sequence homology to 3'UTR of target genes.

Conclusion: We can show that even siRNAs without a physiological target have sequence-specific off-target effects in mammalian cells. Furthermore, our analysis defines the off-target genes affected by the siRNA that is commonly used as negative control and directed against GFP. Since off-target effects can hardly be avoided, the best strategy is to identify false positives and exclude them from the results. To this end, we provide the set of false positive genes deregulated by the commonly used GFP siRNA as a reference resource for future siRNA experiments.

Figure 1

A siRNA directed against GFP specifically targets endogenous genes in human cells. (a) After transfection of the control siRNA directed against GFP, underrepresentation of the off-target genes CYLD and SOAT was detected by expression-microarrays 48 h post-transfection in HEK and HeLa cells. Normalization was performed as described in Additional file 5. (b) Verification of underrepresentation of CYLD and SOAT in HEK and HeLa cells by Real-Time PCR using the same mRNA as template. Expression of genes was normalized to the median of 2 housekeeping genes (PGK, DCTN2).

Figure 2

Off-target effects recur in 3/4 different cell lines and are independent of preparation protocols. (a) After transfection of the siRNA directed against GFP, expression of the genes CYLD and SOAT was measured with qPCR in different cell lines after 48 h. As negative controls, a siRNA directed against a candidate tumour supressor gene (siRNA1) was transfected and a mock transfection was performed. Values are normalized to mock and the median of 2 housekeeping genes. (b) 2 different siRNA molecules with the same target sequence but of different methodological origin were transfected, one generated by in vitro transcription (see Methods) and a second siRNA was obtained by oligonucleotide synthesis (Table 1). After transfection, the expression of CYLD and SOAT was measured in HEK and HeLa cells and normalized expression after mock transfection and the median of 2 housekeeping genes (PGK, DCTN2).

Figure 3

Degree of off-target knock down of CYLD and SOAT correlates with the concentration of the transfected siRNA. (a) Different concentrations of the siRNA directed against GFP were transfected in HeLa and (b) HEK cells. Expression of specific off-target genes (CYLD, SOAT), non-target genes used as negative control (LMNB1, C13ORF1) and genes unspecifically upregulated by antiviral cellular response (OAS1) were measured by qPCR 48 h post-transfection. Expression values were normalized to the average of 2 housekeeping genes (PGK, DCTN2). Correlation coefficients between the amount of siRNA transfected and the corresponding mRNA levels of the genes measured are given.

Figure 4

mRNAs deregulated by GFP siRNA contain short sequences perfectly matched to the sense or antisense strand of GFP siRNA. In order to shed light on the molecular mechanism of sequence-specific off-target effects, we analyzed the mRNAs modulated after transfection of GFP siRNA for sequence homology to the transfected siRNA molecules. The number of perfect matches with a homology length of at least 8 bp was significantly higher using the proper GFP siRNA sequences as compared to the shuffled siRNA sequences.