RNA interference reduces the expression of a gene without changing its DNA, which makes it fast and reversible and also makes it easy to overinterpret. A knockdown is a reduction, not an ablation; it is never complete, it varies between cells and reagents, and the silencing machinery itself perturbs the transcriptome. Most published siRNA experiments report a phenotype after one or two reagents without demonstrating that the phenotype is caused by the intended target. This post is about the design and validation that let you make a knockdown claim you can stand behind.
How silencing works, briefly
Long double-stranded RNA is processed by Dicer into ~21-nucleotide small interfering RNAs, which are loaded into the RNA-induced silencing complex (RISC). The guide strand pairs with the target mRNA, and Argonaute2 cleaves the transcript, which is then degraded. Endogenous microRNAs share this machinery, which is why exogenous small RNAs can saturate or compete with the endogenous pathway. Short hairpin RNAs (shRNAs) are expressed from a vector as a stem-loop that is processed by Dicer into the same effector species, giving sustained silencing rather than the transient effect of transfected siRNA.
siRNA design: the classical rules and their limits
The original design rules were derived from early empirical studies and remain a useful starting point:
- A 19-nucleotide duplex with 2-nucleotide 3′ overhangs, historically dTdT, which improve stability and are required for efficient RISC loading.
- G+C content of roughly 30–50%; low-complexity and very GC-rich sequences perform poorly.
- An A or U at position 19 (the 5′ end of the antisense strand) and a G or C at position 1 of the sense strand, which biases guide-strand selection.
- Avoid stretches of four or more identical nucleotides, avoid the 5′ and 3′ untranslated regions and sequence near the start codon where regulatory proteins and secondary structure interfere, and avoid regions with strong secondary structure.
- BLAST the guide against the transcriptome to avoid silencing unintended genes; a perfect match elsewhere, or even a 3′-UTR seed match, can produce a confounding phenotype.
Scores and algorithms (the Ui-Tei, Reynolds, and Amarzguioui rules, and later models trained on measured knockdown efficiency) improve the hit rate, but none approaches certainty. The practical strategy is to design or purchase four to five independent siRNAs per gene, screen them for knockdown by RT-qPCR, and use at least two that reduce the target substantially — preferably to less than 20–30% of the control — for the phenotype experiments. A single siRNA whose effect is not reproduced by an independent sequence is not evidence.
shRNA vectors: sustained knockdown
shRNAs are delivered in plasmids or lentiviral vectors under a U6 or H1 promoter. Lentiviral delivery is the standard for hard-to-transfect cells and for experiments requiring weeks of silencing; it integrates into the genome, so a single transduction can produce a stable knockdown line. The trade-offs:
- Stem-loop design: a 19–22 bp stem with a short loop (tetraloop or microRNA-30-style backbone) is processed efficiently; the microRNA-30 backbone often gives better processing and less toxicity than a simple hairpin.
- Expression level: high U6-driven expression can saturate the RISC machinery, compete with endogenous microRNAs, and cause toxicity independent of the target. Titrating the viral dose or using an inducible promoter reduces this.
- Sequence variance: a single shRNA sequence is not guaranteed to work; test several, and ideally validate the phenotype with two shRNAs that target different regions of the transcript.
- Controls: include a non-targeting scrambled shRNA and, where possible, an shRNA targeting a fluorescent reporter, processed and transduced identically. Polyclonal populations are preferable to single clones for avoiding clonal artefacts.
Validating knockdown by qPCR and western blot
Knockdown must be documented at both the transcript and protein level, because mRNA reduction does not always translate into protein reduction and vice versa.
- RT-qPCR: use primers that amplify a region present in all target transcripts and are not affected by alternative splicing. Check that the siRNA targets a region that the primers detect; a knockdown assayed by a primer pair that spans an exon not covered by the siRNA can miss the effect. Include reference genes validated for stability in the experimental system, and report the fold-remaining relative to the non-targeting control.
- Western blot: detect the target protein and confirm the band identity with a positive control and a size marker; quantify relative to a loading control with a linear-range signal. An antibody that is not validated for the target will produce an apparent "knockdown" that reflects a change in something else.
- Time course: report the time point at which silencing is maximal and the duration. siRNA effects peak at 24–72 hours and fade by 5–7 days; shRNA effects take longer to develop and are stable.
- Dose: report the siRNA or viral concentration. Excess siRNA causes off-target effects and toxicity, so the experiment should use the lowest concentration that gives robust knockdown.
Off-target effects and the rescue experiment
RNAi is intrinsically prone to off-target silencing. A partial match between the guide strand and other transcripts, especially a match to the 3′ untranslated region, can reduce those transcripts even without perfect complementarity. In addition, high concentrations of any small RNA activate innate immune sensors such as protein kinase R and the interferon response, particularly in primary cells and in vivo. The result can be a phenotype that has nothing to do with the intended target.
Mitigations that matter:
- Use the lowest effective concentration, and avoid concentrations above 50–100 nM for siRNA.
- Confirm the phenotype with at least two independent siRNAs or shRNAs that share no sequence similarity beyond the target.
- Where genome-wide off-target analysis is warranted, use RNA-seq after knockdown to assess changes in other transcripts, or a method such as photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation to map the RISC-loaded targets.
- Rule out innate immune activation with a control small RNA of scrambled sequence, and by measuring interferon-stimulated genes if the phenotype is unusually broad.
The strongest single control is the rescue: re-express the target from a construct that is refractory to the siRNA. For siRNA, introduce silent mutations in the target site (or use a cDNA from a different species with sufficient sequence divergence); for shRNA, the same principle applies. If re-expression restores the original phenotype, the effect is attributable to loss of the target; if it does not, the phenotype was off-target. This experiment converts a correlative knockdown into a causal claim and is the difference between a solid paper and a weak one.
A final note on interpretation: knockdown and knockout are not interchangeable evidence. A partial reduction tests a different condition from a full null, and embryonic lethality or a dramatic knockout phenotype may not be reproduced by an acute knockdown. Report the residual expression level, use the appropriate method for the question, and avoid extrapolating a knockdown result into a claim about complete loss of function.
References
- Fire A, Xu S, Montgomery MK, et al. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature. 1998;391(6669):806–811.
- Elbashir SM, Harborth J, Lendeckel W, et al. Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature. 2001;411(6836):494–498.
- Ui-Tei K, Naito Y, Takahashi F, et al. Guidelines for the selection of highly effective siRNA sequences for mammalian and chick RNA interference. Nucleic Acids Res. 2004;32(3):936–948.
- Reynolds A, Leake D, Boese Q, et al. Rational siRNA design for RNA interference. Nat Biotechnol. 2004;22(3):326–330.
- Brummelkamp TR, Bernards R, Agami R. A system for stable expression of short interfering RNAs in mammalian cells. Science. 2002;296(5567):550–553.
- Stewart SA, Dykxhoorn DM, Palliser D, et al. Lentivirus-delivered stable gene silencing by RNAi in primary cells. RNA. 2003;9(4):493–501.
- Silva JM, Li MZ, Chang K, et al. Second-generation shRNA libraries covering the mouse and human genomes. Nat Genet. 2005;37(11):1281–1288.
- Jackson AL, Bartz SR, Schelter J, et al. Expression profiling reveals off-target gene regulation by RNAi. Nat Biotechnol. 2003;21(6):635–637.
- Echeverri CJ, Beachy PA, Baum B, et al. Minimizing the risk of reporting false positives in large-scale RNAi screens. Nat Methods. 2006;3(10):777–779.
- Cullen BR. Enhancing and confirming the specificity of RNAi experiments. Nat Methods. 2006;3(9):677–681.