Systematic analysis of off-target effects in an RNAi screen reveals microRNAs affecting sensitivity to TRAIL-induced apoptosis

Abstract

Background: RNA inhibition by siRNAs is a frequently used approach to identify genes required for specific biological processes. However RNAi screening using siRNAs is hampered by non-specific or off target effects of the siRNAs, making it difficult to separate genuine hits from false positives. It is thought that many of the off-target effects seen in RNAi experiments are due to siRNAs acting as microRNAs (miRNAs), causing a reduction in gene expression of unintended targets via matches to the 6 or 7 nt 'seed' sequence. We have conducted a careful examination of off-target effects during an siRNA screen for novel regulators of the TRAIL apoptosis induction pathway(s).

Results: We identified 3 hexamers and 3 heptamer seed sequences that appeared multiple times in the top twenty siRNAs in the TRAIL apoptosis screen. Using a novel statistical enrichment approach, we systematically identified a further 17 hexamer and 13 heptamer seed sequences enriched in high scoring siRNAs. The presence of one of these seeds sequences (which could explain 6 of 8 confirmed off-target effects) is sufficient to elicit a phenotype. Three of these seed sequences appear in the human miRNAs miR-26a, miR-145 and miR-384. Transfection of mimics of these miRNAs protects several cell types from TRAIL-induced cell death.

Conclusions: We have demonstrated a role for miR-26a, miR-145 and miR-26a in TRAIL-induced apoptosis. Further these results show that RNAi screening enriches for siRNAs with relevant off-target effects. Some of these effects can be identified by the over-representation of certain seed sequences in high-scoring siRNAs and we demonstrate the usefulness of such systematic analysis of enriched seed sequences.

Figure 1

Knock-down of many genes previously associated with the TRAIL pathway reduces sensitivity to TRAIL-induced cell death. HeLa Cells were transfected with siRNAs targeting a selection of genes previously associated with TRAIL induced apoptosis. After 48 hrs the sensitivity of cells to treatment with 1 μg/ml TRAIL was measured (see Materials and Methods). Error bars represent 1 standard deviation (n = 3), * indicates significantly different from negative control (t-test on log-transformed data, Bonferroni correct p < 0.05).

Figure 2

A screen for new modulators of the TRAIL-induced apoptosis pathway. 12,190 siRNAs targeting 6,095 genes were screened in duplicate for genes affecting sensitivity to TRAIL induced pathway (see materials and methods). a) Distribution of scores for sample siRNAs. Survival after treatment was median normalised by plate, a z-like score calculated using median and median absolute deviation and the minimum of two replicates was selected for each siRNA (see methods and [45]). This score represents a normalized measure of the effect of a particular siRNA compared to the majority of siRNAs and is robust to outliers. b) Boxplot of scores for each of groups of controls used in the screen. NoT = Not Transfected c) Induction of Caspase-3/7 activity after 6 hrs treatment with 0.5 μg/ml TRAIL after knock-down of confirmed hit genes with multiple siRNAs compared to induction in cells transfected with negative control siRNA. Error bars represent 1 standard deviation (n = 3) all results are significant from negative control (P < 0.05, Students' t-test).

Figure 3

The seed sequence ACTTGA protects cells from TRAIL-induced cell death. HeLa cells were transfected with siRNAs targeting Caspase-8 (siCasp8), Luciferase (siGL2) or siRNAs where the siGL2 sequence had been altered to contain the seed sequence ACTTGA (siGL2+seed) or the seed sequence ATCTGA (siGL2+mutseed). After 48 hrs the sensitivity to treatment with 0.5 μg/ml TRAIL was measured. Results are shown relative to cells not treated with TRAIL. Error bars represent 1 standard deviation (n = 3 biological replicates each with 3 technical replicates). Difference between siGL2-seed and siGL2-mutseed is significant (P < 2.5 × 10-3, t-test on log-transformed data).

Figure 4

miRNAs containing enriched seeds protect a several cell types from TRAIL induced cell death. Cells were transfected with miRNA mimics with same sequence as naturally occurring miRNAs that contain seed sequences enriched in high-scoring siRNAs, an miRNA mimic with a sequence occurring in no human miRNA, or an siRNA targeting Caspase-8 as a positive control. After 48 hrs the sensitivity to treatment with 0.5 μg/ml TRAIL was measured. Error bars represent 1 standard deviation (n = 3 bioligical replicates each with 3 technical replicates, * represents results significantly different from the negative control (P < 0.05, students' T-test on log-transformed data), ** are significantly different after Bonferroni correction.