RNAi screens for functional genomics typically look for loss- or gain-of-function phenotypes. They currently have many applications, including target discovery and validation, lead identification and optimization, mechanism of action discovery, predictive toxicology, and biomarker identification.
While reagents are gaining ground, technical difficulties remain in how best to perform and analyze these assays.
“RNAi reagents are being used ubiquitously,” says David Root, Ph.D., director of the RNAi Platform and project leader of the RNAi Consortium (TRC) at the Broad Institute. Dr. Root was among a handful of scientists and corporate partners who created one of the early genome-wide libraries of shRNA constructs (called TRC1). As of 2011, TRC2 has expanded the library (to 300,000 shRNAs) and has measured the knockdown performance of 100,000 of those constructs.
To perform RNAi screens, researchers can choose among a growing variety of reagents, including siRNA, shRNA, miRNA, and lncRNA constructs, as well as pooled libraries and constructs coupled to inducers. Synthetic siRNA has a transient effect, and it is eventually degraded. shRNA can be introduced into a cell within a lentiviral backbone, which allows the construct to be incorporated into the host cell’s genome and then stably expressed. This means the gene is always expressing the shRNA.
“Our philosophy has always been ‘siRNA if you can, shRNA if you must,’” says Christophe Echeverri, Ph.D., CEO and CSO of Cenix BioScience USA. “If the chosen cell system allows for good target knockdown by siRNA transfection, and the timeline needed for the experiments and assays is compatible with the duration of the siRNA effects, then it’ll almost certainly be an siRNA-based study.”
Louise Baskin, senior product manager at Thermo Fisher Scientific, reports that siRNA is one of Thermo’s most popular RNAi reagents, especially in the SMARTpool format. This reagent combines four different siRNAs to the same gene. “SMARTpool reagents are favored by a lot of screeners because it gives them fewer wells per gene—fewer plates to manage, less transfection agent to use.”
shRNA screens have their specific advantages, however. “With always-on gene expression,” Baskin says, “one advantage is that researchers can create stable cell lines and perform, for instance, two-week-long assays.”
RNAi screens can be targeted or genome-wide. Thermo Fisher sells a variety of constructs, from gene family subsets to entire human or mouse genomes. Kinases and ubiquitin ligases are popular, Baskin says. “People can work either based on gene families, gene pathways, areas of interest like DNA damage response,” or they can perform “larger, unbiased investigations” that interrogate the whole genome.
While screens of individually arrayed shRNAs are still very much being used, pooled shRNA screening has become a more widespread approach. While the process sounds complicated, “in many ways, it’s easier than arrayed screening,” Dr. Root says. Pooled screens transduce an entire library of shRNA viruses into a single population of cells instead of one shRNA per microtiter plate well.
In this stochastic process, each shRNA infects only one cell. Scientists treat the cells according to the specific phenotype being tested for. (For instance, in a cancer screen, a specific drug might be added that causes some cancer cells to die while others survive.) Then scientists “deconvolute” the pool, identifying the “hit” shRNAs (carried by surviving cells) by means of qPCR followed by deep sequencing.
Pooled screens are often used as a first-pass look at a large pool of target genes. Not only are they significantly less expensive, they are also very efficient. “You don’t need robots to do the work,” Dr. Root says, so barrier-to-entry is low for small labs.
Some of the drawbacks are in the complexity of sorting hits from nonhits. “While selection-based RNAi screens using pooled, virally delivered shRNAs are cheaper and in some ways technically easier to run, we still find these to offer a narrower range of applicability, and to represent more opaque “black boxes” in terms of experimental control, troubleshooting, and hit confidence levels,” Dr. Echeverri says.
While pooled screens were initially being done by only a few labs with the necessary expertise, “what we’ve found is that if you instruct other people in doing it, they can,” Dr. Root says. However, he adds, “There are pitfalls. A lot of labs are doing it now, but it’s not a kit.”
Sigma Life Science spent about a year and a half developing a deconvolution protocol, which the company now offers as an in-house service, says Shawn Shafer, Ph.D., market segment manager of functional genomics at Sigma-Aldrich. Dr. Shafer advises his customers to narrow their focus before performing a pooled screen, as that leads to more actionable hits.
Large functional genomics screens continue to present challenges to the average lab, including how to account for off-target effects and false positives. “These RNAi mechanism reagents are not perfectly specific to the target gene you designed it for,” Dr. Root says.
People have been aware for a long time that one of the mechanisms mediating these off-target effects were seed-based effects—a “seed” region of an siRNA or shRNA construct mediates suppression of an unintended target gene. Finding patterns to these effects is going to improve data analysis, Dr. Root believes.
Comparing across different RNAi reagents for the same gene helps, as they generally all have different seeds, so all won’t suffer from the same off-target effects. Increasingly, researchers have been employing more sophisticated data analyses that look directly for seed-based effects of the RNAi reagent set. “Happily, it looks like these seed-based effects may actually be nearly all of the off-target effects,” Dr. Root says. “Knowing that will make off-target issues easier to deal with.”
Advanced hit selection algorithms have been developed to weed out certain types of false positives, Dr. Echeverri says, “but their effectiveness and breadth of applicability has by and large proven to be somewhat more limited than we had hoped.”
One new tool to address this issue has emerged over the last year, namely the noncleaving C911 controls introduced by Eugen Buehler, informatics group leader at the NIH. These controls are based on altering bases in the middle of the siRNA. “They potentially present a much more ‘actionable’ source of control data to more effectively rule out a large proportion of false positives coming from siRNA studies,” Dr. Echeverri says.
Next-generation screens are poised to increase both the breadth and depth of current experiments. Both Baskin and Dr. Shafer agree that profiling the activity of regulatory noncoding genes—miRNAs and lncRNAs—is becoming more prevalent. While miRNA reagents have been out for a number of years, including products that both mimic and inhibit miRNA expression, lncRNAs are just beginning to gain traction. Thermo Fisher recently launched screening libraries for mouse and human lncRNAs. “We’re seeing interest from people who are already doing very large screens and are interested in augmenting those screens,” Baskin says.
Thermo Fisher’s miRIDIAN miRNA inhibitors are synthesized within an miRNA format, so to speak, with hairpin structures on each end. This gives them greater stability in the cell and therefore, greater potency. Over at Sigma Life Sciences, they recently launched miRNA inhibitor products based on TuD, or “tough decoy,” RNAs.
Mirimus makes what’s called RNAi-GEMMS, which takes the concept of inducible hairpin expression to the level of animal studies. With their engineered mouse models capable of reversible, fluorescence-coupled gene silencing in vivo, these “RNAi mice” enable researchers to determine the effects of gene silencing in live mice, CEO Prem Premsrirut, Ph.D., says.
“This allows researchers to understand the potential toxicities associated with gene silencing so they can either fine-tune the drug to miss those targets or develop prophylactic combination therapy to avoid these toxicities,” Dr. Premsrirut says. Most of the applications for their mouse models are in the areas of target identification, validation, and toxicology assessment.
Another new technique is gene editing, which is an alternate way to inhibit gene function. “It could either be a replacement or a complement to RNAi,” Dr. Root says. The main technologies utilize zinc finger nucleases, TALENs, and CRISPRs, which have the ability to target specific regions of the genome for editing, including adding a methylation tag or chopping out a portion of the gene entirely, for instance. “Once you can recognize very specific sequences in the genome and attach functional units to that recognition molecule, you can start modifying the genome in a specific, targeted way,” Dr. Root says.
Researchers are also starting to combine multiple experiment types, for example, an RNAi screen with a gene-expression profile or with the copy-number changes in a cancer cell line. In another direction, Dr. Root and colleagues are now helping Todd Golub’s lab at the Broad Institute use new 1,000-gene expression signatures with RNAi perturbations in order to better compare the effects of small molecules, shRNA-produced gene knockdowns, or overexpression of genes. Having a library of expression signatures induced by gene knockdowns “changes the game for interpreting RNAi effects,” Dr. Root says.
Dr. Echeverri has been encouraging his drug developer partners to incorporate RNAi-drug modifier strategies. “These datasets identify genes and pathways that modulate the drug effects, yielding precious new insights from target candidates for possible new combination therapies to biomarker predictions,” he says. In the vast majority of cases, they use multiparametric readouts, i.e., multiplexed microscopy-based assays, “to generate richer, more insightful phenotypic characterizations.”