The current wave of enthusiasm for siRNA as both a tool for determining gene function and a source of therapeutics stems from the discovery, two years ago, that siRNA could induce RNAi in mammalian cells while side-stepping the fatal interferon response triggered by long dsRNA.
However, new research which suggests that siRNAs might not be as specific as once thought has been cropping up lately; in May, a group at Merck’s Rosetta Inpharmatics unit published a paper in Nature Biotechnology detailing how 16 siRNAs they designed to target the coding region of the IGF1R gene in HeLa cells, and 8 that targeted the MAPK14 gene, actually regulated different genes than the ones targeted for silencing.
Now, experiments conducted by Bryan Williams and colleagues at the Cleveland Clinic Foundation’s Lerner Research Institute — and detailed in the September issue of Nature Cell Biology — indicate that the off-target effects of siRNA include the activation of some genes associated with the interferon system itself.
“We were interested initially in trying to find out why long double-stranded RNAs did not work in mammalian cells,” Williams told RNAi News. Postulating that this failure is due to the upregulation of dsRNA response proteins PKR and RNAse L, the researchers tested for sequence-specific suppression of reporter luciferase activity using vectors that produce 500-bp inverted repeat dsRNA molecules complementary to GL3 or RL mRNA in mouse cell lines with PKR and RNAse L deletions.
According to Williams, the researchers used siRNAs as controls, “going along with the dogma” that they might not cause the non-specific effects seen with long dsRNA.
In the end, the dsRNA showed non-specific toxicity and no specific targeting, leading the researchers to conclude that the dsRNA-responsive pathways in mammalian cells extend beyond PKR and RNAse L to the more than 70 other dsRNA binding proteins in cells, Williams said. But more important was the discovery that the various 21-mer siRNAs, including ones targeting Lamin A/C and GAPDH, resulted in interferon-mediated activation of the Jak-Stat signaling pathway and upregulation of a number of interferon-stimulated genes (ISGs).
“We had already modeled…double-stranded RNA onto the PKR double-stranded RNA binding domain. So we know how [the domains] fold, we know the structure of those domains, and we know they recognize 16-base pairs. These are 21-base pairs, and I thought: ‘Well, maybe [the siRNAs] are not going to activate … when you add them to a cell. It turns out, of course, they do activate.”
So what does all this mean? Well, it seems there is good news and bad news.
The bad news first: In cells where the interferon-response pathway is functional, the phenotypes measured by researchers conducting gene expression research might not be due to the specific knockdown of a target but also the non-specific upregulation of ISGs, Williams said. “I suspect, and can certainly identify, a number of papers in the literature where results are described that are probably not due to the specific knockdown of the targets.”
Williams also cautioned that siRNAs being developed as drugs are likely to have non-specific effects. “Almost all humans have an intact interferon-response pathway — that’s how we survive virus infections … and so this is going to lead to major problems in the development of siRNAs as therapeutics.”
“Double-stranded RNA itself has been developed as a therapeutic agent … and used in patients,” he added. “It has major toxicities … [that] are, to some extent, attributable to the induction of interferon, but there are also toxicities that are associated with the double-stranded RNA molecules themselves. The toxicities can be quite varied and certainly dose-limiting.”
But the good news is that these issues appear to be better characterized as roadblocks rather than dead ends.
Williams noted that, in some cases, non-specific effects probably won’t be a major issue. In a number of cell lines, such as certain tumor cell lines, he said, the interferon pathway is defective. In cases like these, “researchers can be pretty certain that the effects of siRNA are due to the knockdown of their target.” As for toxicities in humans, he said that these might be addressed with chemical modifications to siRNA or the development of stable single-stranded siRNA.
Tom Ulich, senior vice president of R&D at Alnylam, also pointed out that just because an off-target effect appears troublesome in in vitro testing doesn’t mean it will be a problem in vivo.
“Our collective experience as an industry in drug discovery and development teaches us that not all, and indeed many, in vitro phenomena that may raise an initial safety concern are not necessarily a concern in clinical practice,” he said. And while this doesn’t mean that off-target effects observed in preclinical tests can be ignored, Ulich noted that they’re something the industry has always had to deal with.
“SiRNA will follow a well-defined path to the clinic, and it will include the traditional preclinical development hurdles,” he said. “All drug modalities, whether they are traditional small organic molecules or protein constructs or antibodies or peptides or antisense oligos or gene therapy, can exhibit off-target effects. SiRNA will, no doubt, also have the potential to exhibit off-target effects.
“The trick from the drug design standpoint is to design an siRNA for therapeutic use that does not have clinically significant toxicity that’s related to off-target effect[s],” he said.
In terms of using siRNAs for gene function discovery and target validation, Chris Echeverri, CEO and CSO of Cenix BioScience, said that off-target effects are even less of an issue.
“In our experience, if you design your siRNAs correctly, if you select the sequence correctly, it’s an extremely low risk,” he said. “The lesson to be learned is to design your experiments carefully with the proper controls. And one of the key things you can do to protect yourself against the chance that the phenotype you’re seeing is not specific to your target is to use multiple siRNAs against that target.
“Let’s remember that RNAi is still quite a young field. We’re still learning exactly how the pathway works, and we’re going to get some surprises. We should expect some things will be not exactly as we planned,” he added. “One shouldn’t be an alarmist. It’s still early days and data will go in both directions. We have to wait and be patient and see how the data accumulate.
“Last year we went through the problem that RNAi was over-hyped,” he said. “This year we’re going… in the other direction. The reality will probably some be in the middle.”