RNA hairpins that trigger cancer cell death
September 23, 2010 § 3 Comments
One of the central problems of cancer is that cancer cells look an awful lot like normal cells. It’s easy to kill cancer cells; the hard part is to kill them while not killing (too many of) the cells that the patient would like to keep. All kinds of ways of distinguishing between cancer cells and normal cells have been tried, from general strategies such as attacking cells that are growing rapidly (anti-mitotics such as taxol) to highly targeted strategies, like inhibiting the mutant kinase that’s characteristic of a certain class of cancers (Gleevec). Some of these work, to some degree, but they are far from perfect. A recent paper from the Pierce lab, pointed out to me by Peng Yin, aims to target mutations directly, using small RNA hairpins to detect the presence of sequence changes in messenger RNAs, and directly trigger an apoptotic response (Venkataraman et al. 2010. Selective cell death mediated by small conditional RNAs. Proc. Natl. Acad. Sci. USA, doi: 10.1073/pnas.1006377107).
The strategy Venkataraman et al. use depends on a pathway that evolved to help defend the body against viruses. In a eukaryotic cell, RNA is carefully kept single-stranded by a large array of RNA-binding proteins. The only time the cell expects to see double-stranded RNA (dsRNA) is when a virus has infected it, either because the viral genome consists of dsRNA or because viral replication involves the production of dsRNAs. So when dsRNAs larger than 30 base pairs long are present, a protective pathway involving protein kinase R (PKR) is triggered. Activation of PKR shuts down protein synthesis — frustrating the virus’s attempt to make more copies of itself — and may trigger apoptosis if the infection can’t be cleared.
Doesn’t that sound like a great strategy? [The immune system is really cool. But complicated.] Unfortunately, viruses have evolved a range of ways to block this pathway, and now, like other components of the immune response, it only provides partial protection. Still, cancer cells are not viruses; can we use this pathway in cancer cells instead?
Venkataraman et al. set out to create a system that will detect specific mutations that are common in certain cancers, and use the presence of these mutations to trigger the formation of double-stranded RNA. To do this, they designed two RNA hairpins: one that opens up only in the presence of RNA from the mutant gene, and a second that is triggered to open up by the first. Once triggered, the two hairpins catalyze each other to open, so you end up with a long chain of overlapping hairpin 1/hairpin 2 hybrids. What’s key is that the length of each of the RNA hairpins is only 14 base pairs, so each individual hairpin is too short to activate the PKR response. But once 4 or 5 hairpins have concatenated themselves together, PKR can bind and dimerize.
The little bits that stick out from the hairpins are essential to the strategy. The authors call them “toeholds”, and they are what makes it possible for the hairpins to escape the kinetic trap they’re in. Without a toehold, hairpin 1 (with the binding energy of 14 base-pairs to keep it closed) would remain stable whether its cognate mRNA was present or not — at least at 37 °C. The binding of mRNA to the toehold gets the hybridization started, and makes it possible for mRNA binding to compete off the other side of the hairpin.
The authors designed three separate sets of hairpins, each targeting a different mutation in cancer: a deletion of several exons in the EGF receptor, which is frequent in one of the nastiest cancers around, glioblastoma; a fusion between two genes, tpc and hpr, common in prostate cancer; and the ews/fli1 fusion found in 85% of a family of tumors (including bone tumors and neuroepithelomas) called Ewing’s tumors. First, they looked to see whether mRNAs containing these mutations could selectively trigger the formation of long double-stranded RNAs made of repeated units of these hairpins. They do: each mutant RNA creates large polymers several hundred base-pairs long, but the hairpins are not triggered by wild-type RNA.
Now for the more exciting question: can the hairpins actually activate PKR in cells, and selectively kill cells expressing the correct mutant mRNA? Yes, they can; and the killing efficiency is actually rather dramatic. Over 95% of the target cells are killed in each case, and the non-target cells (cancer cell lines with other mutations) are completely spared. When the few cells that survive the first round of killing are re-grown and re-treated with the RNA hairpins (delivered via transfection), you again get over 95% killing. This is important, because if killing were significantly reduced in the second round this would imply that it is easy for cancer cells to evolve around the PKR response; but there is no evidence that this is the case. Activation of PKR can be seen by Western blot, and can be blocked by a small molecule that inhibits the PKR pathway. And the cell death appears to be via apoptosis, as expected. (Note that PKR activates the extrinsic pathway of apoptosis; the intrinsic pathway is often disrupted in cancer cells via mutations in the p53 network, making the extrinsic pathway potentially a better target.)
So — if it turns out that intracellular RNAs can be useful therapeutics, which is still very much an open question — this approach seems to have the potential to be highly selective and effective. RNA has already been recognized as a potentially attractive targeting agent for anti-cancer therapeutics, since RNA hybridization can detect a single mismatch (= a single point mutation), and therapeutic approaches based on RNA interference (RNAi) are already being explored in several companies. The approach described by Venkataraman et al might be significantly more powerful: whereas RNAi blocks protein production from the offending mRNA, this hairpin-based approach triggers apoptosis, killing the whole cell. On the less positive side, you need to get two RNAs into the cell instead of one, ideally in roughly equal amounts. But since the problem of intracellular delivery in real patients isn’t yet solved, there is no knowing whether two will be harder than one. While we’re still ignorant, let’s by all means be optimistic.
Venkataraman S, Dirks RM, Ueda CT, & Pierce NA (2010). Selective cell death mediated by small conditional RNAs. Proceedings of the National Academy of Sciences of the United States of America PMID: 20823260
Update: I edited the figure to clarify a couple of things.