© BSIP/SCIENCE SOURCEIt used to be that finding genes that are important for cellular processes in mammalian cells required randomly disrupting those genes by chemical mutation or by insertion of a transposon into the genome. But then, about 10 years ago, RNA interference (RNAi) came along. The technique involves designing and deploying short strands of RNA that bind to gene transcripts with complementary sequences and stop their translation. The advent of RNAi made it possible to systematically disrupt nearly every single gene in mammalian cells using high-throughput screens.

“For mammalian systems this has opened up a whole new realm of interrogating gene function,” says Scott Martin, team leader of RNAi screening at the National Center for Advancing Translational Sciences (NCATS) at the National Institutes of Health. RNAi screens have uncovered genes required for many key cellular processes—such as cell proliferation, the maintenance of stem cells, and apoptosis—and for...

Even though RNAi screening techniques have improved in recent years, with libraries stocked with more RNA probes for each gene that are more effective at knocking down their targets, there are pitfalls. “Off-target effects are the major limitation,” Martin says. You have to take steps to weed out RNAis that silence genes other than the one they’re meant to target, and to be sure that the genes they target are really involved in the cellular process of interest. In addition, there can be challenges from the get-go in choosing the most appropriate library and setting up your screen.

The Scientist asked RNAi experts to weigh in on how to make the most of this potentially powerful technique.


To silence genes in mammalian cells, researchers deliver bits of double-stranded RNA, 21–25 nucleotides in length, dubbed short interfering RNA (siRNA), into cells using a method called transfection, in which the siRNAs form complexes with lipid vesicles that are taken up by cells via endocytosis or membrane fusion. The strands separate, and one of them gets incorporated into a cellular protein complex called the RNA-induced silencing complex (RISC). RISC then binds and cleaves transcripts that have a stretch of sequence that is complementary to the incorporated siRNA strand, thereby blocking production of the protein they encode. Alternatively, siRNA can be delivered into cells using viruses whose DNA has been engineered to express a short hairpin RNA, or shRNA. This shRNA is then converted into siRNA after the hairpin structure connecting the two RNA strands gets cleaved by the ribonuclease Dicer.

So the first step in conducting an RNAi screen is figuring out whether to use an siRNA or an shRNA library. One of the deciding factors is the kind of cells you are working with. Although most cell types, including immortalized cell lines, transfect well, there are some cell types—cells that grow only in suspension, such as immune cells, and certain primary cells, which are cultured directly from living tissue—that do not. Although there are tricks for getting siRNA molecules into hard-to-transfect cells, such as trying different lipid reagents or, in place of lipid carriers, using electrical pulses to create transient pores in the membrane (electroporation), some cell types are simply tough nuts to crack. In these cases, you may have to turn to the shRNA approach.

The choice between siRNA and shRNA could also come down to whether you need long-lasting gene silencing to see your phenotype. After transfection, siRNA molecules get degraded or diluted as cells divide, generally lasting only 5 or 6 days, says Abraham Brass, an assistant professor of microbiology and physiological systems at University of Massachusetts Medical School. This could be enough time to disrupt processes such as cell cycle control, apoptosis, and viral replication, which Brass studies using siRNA screens. However, more-sustained knockdown might be required to study processes such as senescence and differentiation, which can take weeks to complete. shRNA can produce steady-state levels of siRNA indefinitely because the viruses that deliver the shRNAs are typically retroviruses, specifically lentiviruses, mainly derived from HIV. That means the DNA gets incorporated into the cell genome and continues to express the shRNA as the cells divide. (Although the viruses are engineered so that they do not propagate in cells, you still have to work with them in a Biosafety Level 2 cell-culture hood.)

SHORT AND SWEET: Silencing gene activity in mammalian cells requires the use of small double-stranded RNAs, typically 21–25 nucleotides in length. Different types of small RNAs can be used to silence gene function by RNA interference (RNAi) and the choice depends on both the mammalian cell type being studied and the length of time for which the gene of interest needs to be silenced. Short interfering RNAs (siRNAs) tranfected directly into target cells (top) produce transient silencing, lasting only 5 or 6 days. For longer-lasting silencing, short hairpin RNAs (shRNAs) are the molecules of choice. These RNA double strands are connected at one end by a short hairpin sequence (bottom). A DNA plasmid containing the shRNA sequence can be used to directly transfect a target cell or can be packaged into a virus and incorporated into the target cell genome after infection. Stable integration and expression results in long-lasting gene silencing.THE SCIENTIST STAFFAnother key distinction between the two kinds of libraries is how the screens for each are carried out. An siRNA screen is conducted in an arrayed format: researchers transfect cells in multiwell plates, each containing siRNA molecules designed to silence a single gene, and look for the wells that show a phenotype. Because it takes about 200 384-well plates to do a human whole-genome screen (each gene is targeted by several siRNAs, and each siRNA is tested in duplicate), and because siRNA libraries cost at least $100,000, most labs work with a screening facility, which either does the screen for you or provides the libraries and the automated equipment for you to do it yourself in the facility. ICCB-Longwood (ICCB-L) Screening Facility at Harvard Medical School charges in-house researchers about $12,500 to do a whole-genome siRNA screen using their library and equipment ($17,000 for researchers from other institutions).

Although shRNA screens can be conducted in an arrayed format as well, they can also be conducted in a much simpler setup known as a pooled screen. Here, all the cells are exposed to a mix of about 100,000 different shRNAs in their viral vectors. The number of viruses used is kept low enough so that most of the cells take up only one virus and thus express only one type of shRNA, eliciting knockdown of just one gene. A prerequisite for doing a pooled screen is being able to seek out cells in the mix that either “grow or glow” in response to gene silencing, says Stephen Elledge, a professor of genetics at Harvard Medical School who developed the approach. If gene silencing causes cells to survive (“grow”) or to express a protein that can be separated with fluorescence-activated cell-sorting (“glow”), researchers can separate cells that express particular phenotypes from those that don’t, and then sequence the region of the genome that contains the shRNA in the different cell populations to identify shRNAs that could be responsible for the phenotype. Deep sequencing at a core facility is easier and cheaper than ever, says David Root, director of The RNAi Consortium and The RNAi Platform at the Broad Institute.


UNSOUND SILENCING: Even if an siRNA experiment can be reproduced with high accuracy, the results could still reflect off-target effects. Here, two replicates of an experiment using the same siRNAs targeting the same genes lead to similar phenotypes compared with negative control siRNAs (left). But in a validation experiment, two batches of siRNAs targeting different segments of the same genes have very different effects, suggesting the phenotypes may not be due to silencing the intended genes (right).COURTESY OF EUGEN BUEHLER NIH NCATSThere are many commercially available libraries for doing siRNA screens, including ones that cover the whole human and mouse genomes and ones that target sets of genes, but your choice will probably come down to what the screening facility you work with owns. Some facilities, including the ICCB-L and Duke University’s RNAi Screening Facility, offer siRNA libraries that are arrayed in two different configurations. Both include several different siRNA molecules per gene transcript, each of which is complementary to different regions of the transcript. But in one array setup, all the siRNAs targeting a single transcript are combined in the same well, whereas in the other, they are separated into distinct wells. The former is more efficient because fewer wells have to be screened, says Caroline Shamu, director of ICCB-L. However, after you pick the wells that have a phenotype, you then have to look at the effect of each of the siRNA types in individual wells to show that the phenotype is not due to off-target effects of just one of the siRNA types.

A good alternative to a whole-genome library could be focusing on a smaller set of genes that you think could be important for the biological process you are studying, such as those predicted to encode a kinase or phosphatase, or involved in DNA damage repair. “Most people are interested in these focused libraries,” says Hakim Djaballah, director of the High-Throughput Drug Screening Facility at Memorial Sloan-Kettering Cancer Center. The screens can typically be done with only a few plates, compared to hundreds for a genome-wide screen, so some researchers may even be able to get away with doing the transfection and assaying cells for a phenotype in their own lab, although they would still need to get the library from a facility.

To do a pooled shRNA screen, you can order the library yourself, either in the form of viruses ready to add to cells, or as stocks containing DNA plasmids that can be transfected into a special cell line that churns out viruses. There are two kinds of libraries available. One, developed by The RNAi Consortium (TRC) at the Broad Institute, is sold by Sigma-Aldrich in both customizable and off-the-shelf versions. Sigma-Aldrich’s MISSION LentiPlex, an off-the-shelf version of TRC’s library, sells for about $13,000; the company also sells customizable whole-genome libraries and smaller libraries for between $10,000 and $30,000.

The other type of shRNA library, sold by Thermo Fisher Scientific, expresses the shRNA as part of a long transcript that gets processed through a series of steps into the final shRNA. This shRNA version is called shRNAmir, and it was designed to mimic the way a species of RNA called microRNA is expressed and processed in cells. The hope was that shRNAmirs would be processed more efficiently than traditional shRNAs, leading to better gene knockdown, although it is not clear if this is actually the case, says Elledge, who, along with Gregory Hannon at Cold Spring Harbor Laboratory, developed shRNAmir-encoding plasmids. However, shRNA-mir libraries can offer a level of flexibility in gene knockdown that shRNA libraries do not provide: some commercially available library versions of these molecules are designed with a gene promoter that can easily be turned on and off. Conventional shRNA libraries, in contrast, express the shRNAs from a constitutive, or “always on,” promoter. Controlling the timing of knockdown could let you study genes that are essential for cell survival, or manipulate cells in between delivering the shRNA to them and silencing the gene.


“We don’t usually have a shortage of potential gene candidates [from an siRNA screen], fortunately or unfortunately,” says NCATS’s Martin. The hard part is showing that the siRNA associated with that phenotype did indeed silence the gene it was designed to target and not another, unintended gene. Studies suggest that the majority of phenotypes are due to siRNA off-target effects, and that they are mediated through the seven nucleotides between positions 2 and 8 in the siRNA. This stretch, called the seed region, can mimic the activity of naturally occurring microRNA molecules in the cell, binding to and silencing many gene transcripts that have a complementary seven-nucleotide sequence in their 3’-untranslated region. To help weed out these spurious siRNA hits, researchers have developed statistical programs. Eugen Buehler, an informatician at NCATS, developed one such program called common seed analysis, which compares the phenotypic activity between siRNAs that have a given seed sequence and those that don’t (J Biomol Screen, 17:370-78, 2012). Similarly, the Broad Institute’s David Root and John Doench developed approaches to look for hits from pooled shRNA screens that have the same seed sequence.

Once you have narrowed down your potential siRNA hits, it is important to test three to five additional siRNA or shRNA molecules that are designed to target the same gene as your hit, Martin says. If two different RNAi molecules aimed at the same gene produce a similar phenotype, that supports the idea that the effect is real rather than an off-target effect. Another convincing validation technique is the rescue experiment, which involves delivering siRNA along with extra copies of the gene it’s designed to target. The trick is that these copies carry a mutation that prevents the siRNA from binding to them. If the siRNA is doing its job in silencing the native gene, the siRNA-resistant copies should reverse the phenotype. If, on the other hand, siRNA is hitting off-target, the phenotype should remain unaffected. However this type of experiment can be difficult—the siRNA-resistant gene is expressed at high levels, which can affect its function—and many researchers skip this validation step and head straight to studying the biology, Martin says.

The crux of validating any siRNA picked up in a screen is to demonstrate how the silenced gene could be involved in the process you are studying, such as showing that the gene product interacts with proteins in a pathway or localizes to a region of the cell that makes sense. Here, using a library against known genes, such as kinases, could pay off because you can probably find information about the genes in PubMed, Djaballah says. Although there is currently no central repository to check whether your gene of interest has been found in another RNAi screen, there are a couple of places to look. According to Shamu, a growing amount of data from screens is being deposited in PubChem BioAssay (www.ncbi.nlm.nih.gov/pcassay) and in the Genome RNAi database operated by the University of Heidelberg (genomernai.dkfz.de/GenomeRNAi/). 

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?