At A Glance
Name: David Stokoe
Position: Assistant professor, Cancer Research Institute, University of California, San Francisco
Background: Assistant research biochemist, UCSF Cancer Research Institute • 1997-1998; Postdoc, Onyx Pharmaceuticals • 1994-1997; PhD, biochemistry, University of Dundee • 1993; BSc, biochemistry, University of Bristol • 1990
After obtaining conflicting results from gene-silencing experiments using antisense and RNA inteference, David Stokoe began comparing the effects of the two technologies using microarrays in order to explain the different outcomes. The results of this work were recently published in the Biochemical Journal.
Last week, Stokoe spoke with RNAi News about his research and antisense versus RNAi.
Could you give some background on what you do in your lab and what you guys focus on?
We’re generally interested in deregulated signal transduction in cancer.
One pathway that’s strongly implicated in being deregulated in cancer cells is that initiated by PI3-kinase, or phosphoinositide 3-kinase. This was initially implicated through mutations in PTEN, which is a tumor-suppressor protein that was isolated in 1997. The following year, PTEN was shown to be a lipid phosphatase that antagonizes PI3-kinase activity. But more recently, within the last year or two, PI3-kinase catalytic subunit has been shown to be mutated in quite a high frequency of tumors, especially breast cancers.
I’ve been interested in how PI3-kinase transmits its signals for awhile now, and about seven years ago we were looking at how PI3-kinase activates [the PI3-kinase effector protein] PKB. We found that this enzyme, which was subsequently termed phosphoinositide-dependent kinase [or PDK-1], was involved in phosphorylating and activating PKB in response to PI3-kinase activity.
So my lab’s efforts over the past few years have been to look at what happens if you knock down or knock out PDK-1 activity. We were using antisense oligonucleotides initially, and published some results on this a few years ago. Then, when RNAi emerged as a promising technology, we also tested some RNAi oligos that we had synthesized.
The [Biochemical Journal] study emerged from confusing results that we got trying to compare the effects of antisense oligonucleotides to RNAi • we were using the short oligos for these experiments.
Could you give some details on how you started doing the experiments and what you found?
As I said, we had been using antisense oligonucleotides, and they worked very effectively • they were third-generation oligonucleotides from Isis Pharmaceuticals. We screened a number [of the oligos] to pick ones that showed effective knockdown, and we came to the conclusion that the knock down of PDK-1 caused cell death. When we tried to confirm some of these results with RNAi, we again showed an inhibition of [cell] proliferation, but we didn’t see any signs of cell death.
This got us into what the differences were [between antisense and RNAi] and why they were occurring. We analyzed several pathways that could account for the cell death by the antisense oligos, [and] found that at least some of those effects were dependent on the transfection reagents and not necessarily on the oligonucleotides themselves. [We also] found some effects occurring with the RNAi oligos that weren’t occurring with the antisense oligos even thought the knockdown was similar. In the end, we got so frustrated that we figured maybe the best way to figure out what was really going on in the cell when we treated the cell with these oligonucleotides is to use microarrays for a global analysis.
I think this is the first time somebody actually compared the two technologies side-by-side in this manner. We’ve certainly seen extensive literature on profiling antisense oligonucleotides by microarrays and RNAi oligonucleotides by microarrays, but it’s kind of nice to compare the two. I think this is the only way that we could come to the conclusion that we did: that the oligonucleotides we’d been using do seem to give you a gene-array signature, but it can be more indicative of the chemistry of the backbone of the oligonucleotide itself, rather then being a sequence-specific effect in response to the genes you are actually targeting.
I think this is going to be an ongoing problem that will presumably be solved by picking the best possible sequences for knockdown that you could use, at the lowest possible concentrations. There’ll always be this potential for non-specific backbone signature, but by picking the most sequence-specific and most potent oligonucleotides you can reduce that to a minimum.
We were very naïve when we started these experiments, and we also started them in the very early days of RNAi, when there weren’t very many good algorithms to try and predict what would be effective sequences to use. I think since then, there’ve been several excellent publications that we talk about in the paper to help with design strategies. There still seems to be a somewhat empirical element to this, from some presentations I’ve seen from the Dharmacon group, Khvorova et al.
If you can afford to sequence 60 oligonucleotides across your target gene, you can get one or two that are really hyper-functional and knock down the sequences at sub-nanomolar concentrations. These will include the ones you’re going to want to pick for your future experiments. Hopefully, over time, different labs will make this effort to try and pick optimal oligonucleotides empirically, and then better design strategies will emerge.
I think you’re really stacking the odds against yourself if you use antisense or RNAi oligonucleotides themselves at greater than 5 nanomolar [concentrations]. At concentrations greater than that, you’re almost certain to have non-specific effects that may or may not obscure the phenotypes you’re looking for.
[RNAI] is an incredibly rapidly growing area. New strategies [are appearing] using longer siRNA oligonucleotides that can actually be cleaved by Dicer. Recent publications in Nature Biotechnology seem to show [these can have] much more potent effect, and sequences that were not showing any phenotype when presented as 21-mers, when presented in the context of a 27-mer in the same region now do seem to show silencing.
I think it’s just a question of following the literature, using the best strategies that are emerging even though they’re a moving target.
What about your work? Do you still use both antisense and RNAi in your experiments?
I think we’re moving towards RNAi. Initially, my impressions were that there wasn’t very much in the way of differences between the two, and there have been some publications from Isis supporting that.
[Laughs] But my impression is that if you design the RNAi effectively, then you’re getting effect at much lower concentrations compared with antisense. Initially, we were using similar concentrations and doing the comparisons because, quite frankly, those were the concentrations we needed for the RNAi sequences that we had designed to be effective. But … we’ll be using the long RNAi sequences, and getting as sub-nanomolar as possible is going to be the way to go in the future.
What’s the status of the technology in your lab? Are you applying it to the same pathway still, [and] are you exploring other things?
We’ve always just been wanting to use these things as tools. We never wanted to get into the rapidly moving area of whatever mechanisms are underlying the RNAi pathways. So we’ll just follow the literature, be trying some of these longer 27-mers now • these sound very promising.
My ideal would be to do these sorts of experiments with a consortia-type approach where different labs that have interests in different pathways can really put the time into developing RNAi sequences that work very effectively for, say, five or six genes they’re interested in, and then by pooling all of this knowledge into some central database, then eventually there’ll be all the genome covered with very effective RNAi [oligonucleotides] that have been validated by people who are experts in those particular areas. Right now, the danger is that people are doing a lot of experiments with sub-optimal RNAi [oligonucleotides]. It’s very dangerous to take too much from the analysis of the phenotypes when that’s the case.
The big problem, of course, is that these are very expensive experiments to do thoroughly. I guess the price is coming down all the time, but I still think you’re going to want to synthesize 16 or 20 [RNAi oligonucleotides], even using the latest design strategies, to find the ones that are really optimal.
Is [the consortium] just an idea, or something you’re actually pursuing, talking to people about?
I understand that it is something being organized by the [National Cancer Institute], [which has] gotten some experts on board. How it’s going to work practically • I have no idea.
Is this through [the NCI’s] Gene Silencing Section with Natasha Caplen (see RNAi News, 1/30/2004, and 6/11/2004)?
Yes. I’ve just been peripherally introduced to it by Joe Gray, who’s at Lawrence Berkeley National Lab, and who’s involved in this.
This kind of came up because there are companies that are selling genome-wide RNAi libraries, and my take on these is that they’re almost useless. Unless the RNAi [oligos] for each individual gene have been thoroughly validated by people interested in that gene and pathway, I think the potential for off-target effects is just going to be overwhelming. But in five years time, if we can have a similar library where every single gene is covered by three or four oligonucleotides that have individually been validated and shown to be effective at sub-nanomolar concentrations, then we’re really in business.