In 2011, several big pharma companies dealt serious blows to RNA interference (RNAi) technology by abandoning or reducing their programs. But while RNAi may be down, it is far from out.
The discovery of RNAi electrified the field of biology and earned scientists the 2006 Nobel Prize in Medicine. Though the initial fever pitch excitement for RNAi therapeutics has dimmed in some circles, academic and biotech scientists continue to doggedly pursue the challenges that vex the field. Despite setbacks, many clinical trials are under way with more planned.
Keystone Symposia’s “Nucleic Acid Therapeutics” and CHI’s “High Content Analysis” meetings featured new technical approaches and disease-relevant applications for RNAi. Presentations described an allele-specific inhibition treatment for Huntington disease, optimizing high-content RNAi screens, deciphering mechanisms of biothreat pathogens, and use of improved lipid nanoparticles to deliver the RNAi payload.
Genetic illnesses, such as Huntington disease, often affect only one allele. The challenge for RNAi therapeutics is to specifically knock down only the mutant, but not the wild-type, allele.
“We are pursuing an allele-selective approach for inhibition that uses specially constructed single-stranded RNAs that act through the RNAi pathway,” said Dongbo Yu, M.D./Ph.D. student in the laboratory of David R. Corey, Ph.D., professor, pharmacology and biochemistry, University of Texas Southwestern Medical Center.
“Huntington disease is a devastating neurodegenerative disorder that presents in mid-life, and symptoms worsen until death. It is caused by extended runs of a trinucleotide repeating unit (CAG) within the first exon of the huntingtin gene. Normal alleles have 10–17 copies, but mutant alleles have ~35–100 CAG repeats,” Yu explained.
“Our lab previously identified several different classes of oligonucleotides that can specifically inhibit expression of the mutant huntingtin gene. These were either single- or double-stranded designs. However, each type had its own specific drawbacks such as low specificity, limited stability, and biodistribution and/or nuclease sensitivity.”
According to Yu, Isis Pharmaceuticals developed the novel single-stranded RNA (ssRNA) design and provided compounds to the Corey Lab.
“The hybrid chemistry developed by Isis combines the desired properties of antisense oligonucleotides and duplex RNAs into a single novel type of small RNA. We applied this new tool to selective inhibition of huntingtin expression.”
Yu and colleagues provided proof of concept in experiments that used patient-derived heterozygous fibroblast cell lines treated with these strategically mismatched ssRNAs.
“We found that the huntingtin gene was potently knocked down across a broad range of concentrations without significantly affecting the wild-type allele. These results represent the most potent and allele-selective inhibitors identified to date. They also suggest that there are unexplored RNAi pathways that remain to be characterized.”
Dr. Corey predicted that ssRNAs have the potential to be a major breakthrough.
“They combine the advantages of antisense oligonucleotides (e.g., simplicity of using one strand and the potential for good biodistribution) with the potency of duplex RNAs that function through the RNAi complex. Over the next five years we may see ssRNAs emerge as a competing strategy for clinical gene silencing that will challenge antisense oligonucleotides and duplex RNAs.”
The Corey Lab will continue collaborating with partners at Isis and the laboratory of Donald Cleveland, Ph.D., at the University of California, San Diego. “We hope these studies will lead to new RNAi-based therapeutic treatments for Huntington disease,” Dr. Corey said.
High-Content RNAi Screens
High-content screening for RNAi can provide a rich source of high-dimensional phenotypic data to explore the effects of knockdown in a variety of ways that move beyond simple threshold-based selection methods.
The process, however, is not without its challenges, such as the need for a scalable and robust data-management infrastructure and the problem of identifying a subset of biologically meaningful features from the raw, high-dimensional data.
“Extracting functional relationships from high-content RNAi screens definitely is a challenge,” noted Rajarshi Guha, Ph.D., informatics scientist, NIH Center for Translational Therapeutics.
“In the high-content environment, microscopy images of individual cells from each well treated with siRNAs must be evaluated. We use commercial siRNA libraries containing four different siRNAs per gene.
“The use of robotics integrated with image-analysis tools allows for the automated evaluation of various parameters. But manual inspection of the automated analysis is vital, to ensure that the results make sense in the context of the biology being studied.
Dr. Guha and colleagues have developed a framework for hit selection that employs a random forest classification model that can identify phenotypic changes such as cell number, size of their nuclei, cellular shape, and characterization of intracellular changes.
“The random forest method is a machine-learning algorithm, and is an extension of the decision tree method that asks a series of yes/no questions to label observations. The nice thing about decision trees and random forests is that they are somewhat interpretable, compared to many other methods that are black boxes.”
Dr. Guha explained that once phenotypic classifications are made, which are usually coarse-grained, the next step is fine tuning.
“We developed a series of tiered models in order to refine groups into finer populations, allowing us to apply gene ontology (GO) enrichment analysis methods to provide functional annotations and hence prioritize genes for follow-up analysis.”
The take-home message Dr. Guha delivered is that high-content RNAi screening provides datasets that are much richer compared to those obtained from reporter systems.
“It is also becoming increasingly important to characterize multiple phenotypic parameters rather than to focus only on one. Using machine-learning tools and high-content generated functional screens can lead to a much better understanding of the biology and associated phenotypic features produced by RNAi.”