Basic Requirements
RNAi experiments have four basic requirements:

•An effective RNAi “trigger”: a specific dsRNA that targets a particular gene transcript to induce the RNAi pathway. In mammalian cultured cells, RNAi is typically induced by siRNA or a short hairpin RNA (shRNA).

•An efficient delivery method for the trigger: For mammalian cultured cells, the two most commonly used methods are transfection and electroporation. For many immortalized cell lines, transfection with a lipid- or amine-based reagent is the preferred option. Electroporation is used for delivery into primary cells and suspension cells for which standard transfection methodologies are problematic.

•A robust phenotypic assay for the RNAi effect: Phenotypic assays are procedures that provide a quantitative measure of any state or condition of the cells in a population. Developing an assay with a high signal-to-noise ratio that truly measures the cellular phenotype is critical to the success of screening experiments using RNAi tools.

•Positive and negative controls: Controls are needed both for siRNA delivery and for the phenotypic assay.

Mammalian RNAi Experiments
In mammalian cultured cells, RNAi is typically induced by the use of siRNAs. There are two general methods for producing siRNAs in cultured cells: delivery of synthetic siRNAs, and introduction of a DNA construct that expresses short hairpin RNA sequences (shRNA) that are processed to siRNAs within the cell (Figure 3). Figure 4 below summarizes the procedures and capabilities of each method (see also Downward 2004 [1], Silva et al. 2004 [59], Sachse et al. 2005 [3]).

Figure 3. Three Ways to Trigger the RNAi Pathway. (1) In non-mammalian systems, the RNAi pathway commences when double-stranded RNA (dsRNA; usually longer than 30 bp) is introduced into cells. In mammalian systems, RNAi can be triggered by DNA based expression vectors designed to express short hairpin RNA (shRNA) molecules (2) or by synthetic short interfering RNA (siRNA) molecules (3) . In each case, gene silencing results from destruction of mRNA that is complementary to the input siRNA (3) or the siRNA molecules created by Dicer cleavage of longer dsRNA (1) or shRNA (2) molecules.  Dicer=cytoplasmic nuclease; RISC=RNA-induced silencing complex; mRNA=messenger RNA.

Delivery of Synthetic siRNAs
Synthetic siRNAs are typically 21 bp double-stranded RNA molecules with dinucleotide 3' overhangs. Synthetic siRNAs are thought to load directly into RISC upon introduction into the cytoplasm by transfection or electroporation. The RNAi effect achieved is transient, lasting typically for 3–7 days.

Careful design of the siRNA is required to maximize silencing of the target gene while minimizing off-target effects. With improved understanding of the mechanism of RNAi, algorithms for designing effective siRNAs have improved concomitantly, decreasing the number of sequences that need to be tested to elicit gene-specific silencing. Nevertheless, good experimental design dictates independently testing multiple distinct siRNA sequences targeting the same transcript to confirm that observed gene silencing is due to specific rather than off-target effects.

The various methods for producing synthetic siRNAs are described below:

Chemical Synthesis
Chemical synthesis is the preferred and most widely used siRNA preparation method for transient experiments in cultured cells. Chemically synthesized siRNAs are easy to use and to transfect in many cell types, and are readily adaptable to high throughput screening methods.

In Vitro Transcription
siRNAs can be readily prepared by in vitro transcription (IVT) of synthetic DNA oligonucleotide templates, but this method requires more hands-on time from the researcher and is subject to higher variability in yield and quality of the products. Higher toxicity has been observed empirically in some cell lines using siRNAs synthesized by IVT.

In Vitro Transcription of Long dsRNA Followed By Cleavage With Dicer or RNase III
Long dsRNAs are prepared by IVT using a template that typically encodes a 200–1000 nt region of the target mRNA. The dsRNA is then digested in vitro with Dicer or RNase III to produce a population of endonuclease-prepared siRNAs (esiRNAs) or siRNA-like molecules, respectively. The product mixture, containing many different siRNAs, is then used to test gene knockdown, bypassing the testing steps involved in selecting an individual effective siRNA sequence. This method requires significant hands-on time and has complex scale-up requirements.

Expression of shRNA from a Plasmid or Viral Vector
Short hairpin RNA (shRNA), expressed from a plasmid or viral vector within the cell, can trigger RNAi. Although vector constructs are more labor intensive to use compared to chemically synthesized siRNAs, and vector-encoded siRNA design rules are not as well established, this method does provide a viable alternative when chemically synthesized siRNAs cannot be used. In addition, viral-based vectors permit delivery by infection, which can be beneficial if your cell system is very difficult or impossible to transfect with siRNAs (e.g., terminally differentiated cells).

If your experimental conditions require strong silencing for more than a week, a vector that can integrate into the host genome is a wise choice. Integration vector-based libraries also offer the potential of selective screens. In this approach, a selective pressure is applied to the cells to screen for RNAi targets that are required for sensitivity to the selection; only cells with constructs expressing shRNA to the required target survive and grow. This strategy has been used in a large-scale RNAi screen for genes involved in p53-dependent proliferation arrest [60].

A refinement of selective screens is aimed at identifying RNAi targets whose silencing would not permit survival of a selective screen—“barcode” screening [60, 61]. A gene-specific sequence, the barcode, is incorporated into each shRNA construct. Pools of shRNA constructs are introduced into cells, and subjected to a selective pressure. The shRNA representation in the pools before and after the selection is compared by amplification of the barcodes present in the pools with different fluorescent labels and hybridization to barcode microarrays. This as-yet not fully validated strategy has promise as a method for identifying genes with potential as therapeutic targets [1].

A major disadvantage of plasmid-based libraries for large-scale screens is the lower transfection efficiency of plasmids compared to oligonucleotides. Synthetic siRNAs must be delivered only to the cytoplasm, whereas plasmid vector constructs must enter the nucleus; viral vectors improve this delivery limitation. In addition, the labor and cost of preparing DNA for transfection from large numbers of plasmids is significant, and replication of bacterial cultures potentially subjects the bacteria carrying the plasmids to selective pressure for recombination events that lead to loss of hairpin sequences [1]. Also, the effective concentration of the shRNA that is expressed inside the cell may not be as readily controlled as that of synthetic siRNAs, which decreases the user’s control over the risk of off-target effects. Some groups report higher variability in effectiveness of silencing by shRNAs than by siRNAs [3].

CLICK HERE for interactive, online version Figure 5. siRNA Decision Tree: A Guide to Designing siRNA Experiments Determining which approach to use for an RNAi experiment in mammalian cultured cells begins with answering a few simple questions. The first question is whether you need long-term silencing. Transfection of siRNAs usually induces strong silencing for 3–7 days; the length of the effect depends on the proliferation rate of the cells and the potency of the siRNA. Most experiments can be completed in this amount of time. Although you can re-transfect cells to extend the RNAi effect, if your experiment requires more than a week of silencing, a vector-based approach may be a better choice.

Non-mammalian RNAi Experiments
As model systems for studying development and cell regulation, the nematode Caenorhabditis elegans and the fruit fly Drosophila melanogaster each offer powerful genetics, a completely sequenced genome, and a significant degree of gene conservation with vertebrates [53, 54]. In C. elegans and Drosophila, long dsRNA (e.g., typically >200 bp) complementary to the target transcript is used to induce RNAi (Figure 3).

RNAi is readily induced by feeding C. elegans with bacteria harboring plasmid constructs engineered to express dsRNA [19, 55] or by soaking the worms in dsRNA [18]. Both strategies are adaptable to large-scale, high throughput screens [56, 57].

In cultured Drosophila cells, large-scale screens using RNAi can be performed via transfection or “bathing” (adding dsRNA directly to the culture medium in which the cells are grown). Transfection is used in cell lines where bathing is not an efficient delivery method or when the screen uses transient transfection of reporter constructs [20]. Bathing has also been used in combination with transfection of reporter constructs [58].

Double-stranded RNA for direct delivery in either system is typically synthesized by bidirectional in vitro transcription of DNA templates representing either genomic fragments or open reading frames [20, 56, 57].

The availability and utility of genome-wide C. elegans and Drosophila RNAi libraries has moved RNAi-based research from the study of single genes to the identification of new genes involved in biological processes of interest. A similar transformation to functional genomics is occurring in mammalian systems now that genomic siRNA libraries are available.