November 15, 2014 (Vol. 34, No. 20)
Greg Crowther, Ph.D.
When targeting genes, drug developers may opt for knockdown via CRISPR or silencing via RNAi. These weapons, however, may fit different battle plans.
Recent progress in probing gene function via the RNAi and CRISPR methods were a strong theme of the Discovery On Target conference, which took place last month in Boston. Both methods enable researchers to impair the function of a targeted gene.
With RNAi (short for RNA interference), molecules such as small inhibitory RNA (siRNA) or short hairpin RNA (shRNA) are used to induce degradation of a gene’s mRNA transcripts. In contrast, CRISPR (an acronym for clustered regularly interspersed short palindromic repeats, part of a bacterial defense against phages) exploits a gene-editing mechanism to mutate the gene itself, thus knocking it out entirely rather than knocking down its expression.
Scientists interviewed by GEN indicated that neither method is likely to render the other obsolete anytime soon because each has distinct advantages and limitations.
Shawn Shafer, Ph.D., functional genomics market segment manager for Sigma-Aldrich, wondered whether RNAi’s days were numbered after scientists at the Broad Institute published two landmark CRISPR papers in January. “I’ve revised my thinking since then,” he said. “If the ultimate goal is the creation of new drugs, RNAi may be preferable to CRISPR for validating new drug targets because small molecule drugs only reduce a gene’s function somewhat, as RNAi does, rather than knocking it out entirely, as CRISPR does.”
Anja Smith, Ph.D., director of R&D for GE Healthcare Dharmacon, agreed that RNAi remains preferable for certain applications. “If you were to do genome-scale arrayed screens,” she commented, “I would argue that the efficiency is not as high [for CRISPR] as in RNAi knockdown.”
On the other hand, for in-depth exploration of individual genes, “CRISPR can do things that RNAi cannot do,” adds Alex Amiet, senior product manager at GE Healthcare Dharmacon. These CRISPR-only applications include creating tagged fusion proteins and introducing single-nucleotide polymorphisms (SNPs).
Interfering with Obesity
Many conference presentations expressed enthusiasm for ongoing work involving RNAi or CRISPR or both. One such presentation was “Identification of Novel Anti-Obesity Genes in Primary Human Adipocytes using RNAi Screening” by David Fischer, Ph.D., senior director of biological sciences at BioFocus (now a subsidiary of Charles River Laboratories).
For this work, GlaxoSmithKline hired BioFocus to screen 100 genes previously implicated in human obesity. Using primary human adipocytes (fat cells), BioFocus targeted each gene with six distinct shRNAs and characterized phenotypes with assays quantifying lipolysis, adipokine release, mitochondrial activity, lipid droplet formation, and expression of 32 other genes of interest.
According to Dr. Fischer, BioFocus’s RNAi assets include its adenoviral platform for delivering shRNAs to primary human cells. These cells are derived directly from human tissue, so they approximate conditions in vivo better than immortalized cell lines. By using primary cells, drug development projects can “touch base with clinical reality at the beginning rather than at the end,” suggested Dr. Fischer.
Dr. Fischer also believes that a thorough characterization of phenotypes, as achieved with multiple assays, is important in identifying the best possible drug targets. “Sometimes it’s a combination of those readouts that make those [particular] targets the most exciting targets,” he added.
Improving shRNA Design
As exemplified by Dr. Fischer’s study, careful RNAi work generally targets each gene with several distinct RNA sequences, which should have similar effects. Furthermore, each inhibitory RNA should be as specific and as potent as possible.
These goals were the subject of the presentation “Sensor-Based shRNA-mir Reagents for More Effective RNAi Screening” by Gwen Fewell, Ph.D., co-founder and chief commercial officer of TransOMIC Technologies, the exclusive provider of new-generation shRNA collections developed by Gregory J. Hannon, Ph.D., and colleagues at Cold Spring Harbor Laboratory (CSHL).
The development of these shRNA collections was inspired by performance issues with early-generation shRNA designs. “On a gene-by-gene basis,” recalled Dr. Fewell, “one had to test several shRNA’s per target to find one that worked effectively to provide knockdown and relevant phenotypes.
“Confidence in screen hits is also impacted when one or two shRNA’s per gene show up as hits while the majority do not. This increases the potential for false-positive and false-negative results.”
shRNAs were once prepared according to siRNA design rules. The rules were not entirely appropriate for shRNAs, which, unlike siRNAs, require processing by the enzymes Drosha and Dicer. Optimal processing occurs only when the primary microRNA scaffold contains certain conserved domains, as reported last year by David P. Bartel, Ph.D., and colleagues at MIT.
The new shRNA-specific shERWOOD algorithm was developed from a sensor-assay based study of over 250,000 shRNA sequences by Simon Knott, Ph.D., a research investigator in Dr. Hannon’s lab at CSHL. (Dr. Knott shared this work in his conference presentation, “A Computational Algorithm to Predict shRNA Potency.”)
TransOMIC’s new shRNA collections incorporate advances in design that, according to Dr. Fewell, improve specificity, small RNA processing, and knockdown. These designs, Dr. Fewell noted, have been incorporated into several pooled shRNA formats whose alternative promoters and vectors enable screening of cell lines that previously were not amenable to RNAi.
RNAi in Xenograft Tumor Models
Dr. Fischer’s desire for more realistic models of disease was echoed in the talk “Loss-of-Function Genetic Screens to Find Genes Regulating Cell Responses and Identify Potential Drug Targets” by Paul Diehl, Ph.D., director of commercial development at Cellecta.
Cellecta aims to do pooled screens of complex shRNA libraries directly in tumors produced by implanting cancer cells into immunocompromised mice (xenografts). For the purpose of identifying genes required for tumor progression, the xenograft model of cancer should be superior to cancer cells grown in culture.
shRNA screens in xenografts, however, are not without complications. “When you take [tumor] cells and inject them into the mouse, only a small percentage of the cells actually go to form the tumor,” Dr. Diehl explained. “So you’re already getting a skewing of the population, and it can be difficult to isolate the effects of the shRNA versus the random noise in the background of the tumor.”
One possible solution is to run experiments with a huge number of mice. Cellecta has pioneered an alternative tactic: barcoding the tumor’s founder cells.
With each shRNA-containing cell containing a uniquely identifiable DNA sequence (that is, a barcode), any shRNA effects on cell death can be distinguished from cell death expected from the xenografting procedure. While the data analysis is complex, the basic idea is to see whether any particular shRNAs lead to lower-than-expected survival of the founder cells.
Cellecta is also in the process of developing human genome CRISPR libraries, and comparing screening results obtained with these to those obtained with pooled shRNA. Agreeing with others’ views that RNAi and CRISPR will likely be complementary technologies, Dr. Diehl anticipates that Cellecta’s clients will start screening with both shRNA and CRISPR, or will screen with one method and then validate hits with the other method.
“[Clients have] invested a lot in the shRNA technology, and now they’re ramping up the CRISPR technology and doing both,” informed Dr. Diehl. “Because there’s not a lot of additional investment you need to do to bring CRISPR online if you already have shRNA up and running.”
The Devil Is in the Details
Like Cellecta, GE Healthcare Dharmacon aims to remain a leader in RNAi technology while building its CRISPR capabilities. “It’s too early and overly simplistic to say definitively that one will make the other obsolete or supplant the other,” contended Amiet. The choice of whether to use RNAi or CRISPR or both, she said, will often hinge on the details of the experimental questions.
For large-scale phenotypic screening of many genes, Amiet and Dr. Smith find that RNAi remains more suitable than CRISPR. They report that RNAi generally achieves a 75% or greater knockdown in nearly all cells of the targeted population, whereas CRISPR may create a knockout only in a much smaller percentage of cells. CRISPR-induced phenotypes may thus be difficult to observe unless cells containing the knockout are isolated and expanded. This clonal isolation may not be feasible in an arrayed, high-throughput screening context where hundreds or thousands of genes are being queried.
Such concerns notwithstanding, GE Healthcare Dharmacon has just released its first CRISPR product: a CRISPR-Cas9 gene engineering platform, known as Edit-R, which eliminates the need for cloning and sequencing of target sequences.
On the RNAi side, the company continues its efforts to help researchers determine whether any observed effects are specific to the gene of interest (that is, “on target”). Through collaborations and close partnership with members of the RNAi Global Initiative, GE Healthcare Dharmacon is preparing specialized control reagents for use in follow-up experiments to determine the role of the siRNA seed region in causing nonspecific phenotypes due to off-targeting.
CRISPR, ZFNs, and TALENs
Dr. Shafer of Sigma-Aldrich, presenter of “The State of CRISPR Technology” at the Discovery on Target conference, noted that RNAi is not the only technique to which CRISPR should be compared. Since CRISPR is a genome-editing tool, it may be weighed against ZFNs (zinc finger nucleases) and TALENs (transcription activator-like effector nucleases), which also allow genome editing.
“The great advantage that CRISPR has over ZFNs and TALENs is that those two tools are protein-to-DNA interactions, and CRISPR is a RNA-to-DNA interaction,” emphasized Dr. Shafer. “So [with CRISPR] it’s much easier to make that template because it’s simply a 17- to 20-base pair RNA that you clone into a vector and then express.”
Regarding CRISPR versus RNAi, Dr. Shafer pointed out that some genes must be knocked out completely for their importance to be understood. “What researchers have run into,” he said, “is that some proteins, particularly enzymes and transcription factors, still retain all of their functionality even if they’re present at 1% of their typical amount. It’s not until you knock out that gene completely that you can tell what the disease looks like in the absence of that signaling pathway.”
In these situations, knockouts via CRISPR may be more informative than knockdowns via RNAi.
Dr. Shafer concluded that CRISPR’s potential for off-target effects is still being investigated. As with RNAi, carefully designed sequences may eliminate most potential problems. Most genes can be knocked out with a mutation anywhere in the first two-thirds or so of their coding region; given these wide ranges in which to find usable sequences, Dr. Shafer estimated that 90% of genes can be knocked out with CRISPR without sacrificing specificity.