Professor, molecular genetics/microbiology
Duke University Medical Center
At A Glance
Name: Bryan Cullen
Position: Professor, molecular genetics/microbiology Duke University Medical Center
Background: PhD, microbiology, University of Medicine and Dentistry 1984
MSc, virology, University of Birmingham 1974
BSc, biological sciences, University of Warwick 1973
Starting his career in industry, working as a lab head for the now-defunct Roche Institute of Molecular Biology, Bryan Cullen ultimately came to Duke University in 1987 as an assistant investigator and never left. Now, as director of the university's Center for Virology, he "uses viruses as model systems to study important phenomena in human cells," as he puts it.
One of those phenomena is RNAi, and Cullen recently spoke with RNAi News about his work.
Let's start with a breakdown of your lab and its overall focus.
In terms of RNA interference, we started out with the idea of using RNA interference as a way of inhibiting HIV-1 and other viruses, and with the idea that RNA interference is an important antiviral defense mechanism. That led us into a lot of papers that we've published over the last two or three years on how exactly microRNAs are made.
The reason we got into that [stemmed from an] idea that we came up with originally back in a paper published in 2002 in Molecular Cell, [which was that a way to express] small interfering RNAs stably was to take a natural human microRNA and the one we chose was mir-30 and then modify the stem of the stem loop in mir-30 so that it would now make a novel, artificial microRNA. The idea was to then use pol-II promoters to drive expression of the siRNA. Obviously, the more we knew about how authentic microRNAs were made and … the recognition signals that allowed them to be accurately processed by the cell, the more easily we'd be able to make an artificial microRNA that we could then use to target any gene we were interested in. That kind of took on a life of its own, and we became fascinated by the whole process of how microRNAs are made regardless of the utility of what we were learning. We had a paper earlier this year in The EMBO Journal where we defined for the first time the structural parameters of primary microRNA transcripts and how they're recognized by Drosha.
I think our major contribution essentially is in publishing a number of manuscripts looking at how microRNAs are made. Our secondary one is then using that information to design more effective ways of expressing artificial microRNAs using RNA polymerase II, and that's what really led to my invitation to write a News & Views [for Nature Genetics this month] on these articles that have come out. …
What's happened is that, obviously the RNA interference business started out with siRNAs, and that's still, in terms of drug development, the way to go. But in terms of doing things in the laboratory, doing screens, siRNAs are not necessarily the most efficient way to approach this problem. The first thing that was developed was the small hairpin RNA approach … and that is [an imitation in many ways] of a pre-microRNA, one of the intermediates in microRNA biogenesis. What we had done in our original 2002 paper and in three or four follow-up papers was show that you can actually make a primary microRNA that has an artificial stem, which allows you to knock down a target gene. We were really the only people doing this, and we were essentially throwing rocks into the water: it was generating a few ripples but not a lot of waves.
[A significant effect] wasn't until [Greg] Hannon's people [at Cold Spring Harbor Laboratory] decided to pay attention to our work. They initially started out by changing their shRNA design … to incorporate some of the information we had generated … to optimize their shRNAs to look more like microRNAs. They decided to base [their changes] on our work because they used mir-30 as a template, which is the one we had always chosen to work with. What they found was if they redesigned their short hairpin RNAs to look more like microRNAs they got better expression [although] they were still running them off of pol-III promoters.
What I think was the surprising result for them … was they found when they incorporated these pol-III-driven cassettes into retroviral vectors, the pol-II promoter, which is present in retroviral vectors, was actually driving more microRNA biogenesis than the pol-III promoter was. They could actually take the pol-III promoter out of the system and get very nice levels of microRNA expression. That's because they had all of the cis-acting sequences in place that allow the microRNA to be excised from any RNA transcript, and it's position-independent and promoter-independent. They now took that information and, as we had done to some degree previously but they did a better job, showed that you could use pol-II to drive these microRNA expression cassettes that make artificial microRNAs that could get very high levels of artificial microRNA expression; they call them shRNA-mir. Perhaps more importantly from a scientific perspective, they showed that they could get very nice regulated expression of the microRNA using inducible or tissue-specific pol-II promoters.
I think this probably hasn't got a lot of bearing in terms of clinical treatment, but … it generates a lot of tools [that] I think will be important for future use of RNAi interference in various forms of basic and applied research.
How come you don't see possibilities in the therapeutic setting?
If you're going to deliver one of these cassettes, you're going to have to do it using a gene therapy-type approach. Gene therapy has been promising for a long time, but it still hasn't [resulted in] any treatment. There are lots and lots of clinical trials that are always ongoing with gene therapy, but none of them have led to a treatment for any disease that is used on a routine basis in a clinical setting. And there are ongoing problems with gene therapy; of course the adenovirus approach led to that death at [the University of Pennsylvania in 1999], and there have been a couple of cases with retroviruses where there has been onset of T-cell leukemias.
These are issues that still need to be carefully resolved and so I don't think gene therapy is likely to lead to any treatment in the near future. In the longer-distance future, if gene therapy can be made into a usable technique then this kind of RNA interference approach would be extraordinarily appropriate. It's a technical issue.
When last we spoke, you were talking about the work you were doing looking into the characteristics of microRNAs associated with their binding to exportin-5, and you were doing some investigation into Drosha (see RNAi News, 7/30/2004). Can you touch on where you are now in that aspect of your work?
A lot of that work has been published. Exportin-5 binds to what is about a 63- or 65-nucleotide stem loop with a 2-nucleotide overhang. The question is, what is exportin-5 seeing that allows it to recognize this RNA from among all the other RNAs in the cell? Not too surprisingly it turns out it's the actual characteristics of the stem. So exportin-5 likes a short 3' overhang [about] 2-nucleotides [long]. You need a stem of a certain length, which is about 17 base pairs or something like that. That's what it sees it doesn't actually see any sequence specificity.
That was even more dramatic with the Drosha story, which was published in The EMBO Journal with a follow-up paper in JBC, both this year. What we were able to show was that Drosha actually sees a stem of a specific length, and the length is about 30 base pairs long. If it is significantly shorter than 30 base pairs, like 26 or something like that, or significantly longer, say 40, then the stem is no longer recognized by Drosha. That stem has to have a large terminal loop that's unstructured, and the terminal loop has to be bigger than 10 nucleotides. That has to be flanked in turn by single-stranded sequences, so the sequences around the stem themselves cannot be structured. If you present Drosha with something that looks like that, it sits down on the stem loop and measures 22 base pairs away from the junction of the stem and the loop so it's actually sort of a molecular ruler. Then it cleaves 22 base pairs away from the stem/loop junction, which is about two-thirds of the way down the stem. That leaves you then with the pre-microRNA, which is this RNA with a 2-nucletide 3' overhang and about a 22 base pair stem. And it has this large loop, and that's what's recognized by exportin-5; exportin-5 cares about the overhang and the stem, it doesn't care so much about the loop. Drosha cares a lot about the loop. That [pre-microRNA] is transported to the cytoplasm.
Dicer is also a molecular ruler. Dicer binds to the base of the stem; it recognizes a 2-nucleotide overhang on an RNA stem. It's pretty catholic in its tastes it doesn't have to be anything too specific. Just a 2-nucleotide overhang on an RNA stem is fine for Dicer. What is does is then cleave 22 base pairs away remember, Drosha cleaves 22 base pairs down, Dicer cleaves 22 base pairs up. The net result of that is Dicer pretty precisely removes the terminal loop. That explains what is going on it's all RNA recognition.
There's a lot of work to be done still in terms of trying to understand the domain organization of the RNase III enzymes and how they recognize their structures. But I think at the molecular level we have a good grasp now of what is being recognized and it is specific kinds of RNA structures that are being recognized.
Where does that lead you and you lab into what you're focusing on now?
What we're mostly focusing on now are viral microRNAs. We had a paper in press in PNAS back in April or so where we looked a Kaposi's sarcoma herpes virus and showed that that virus makes 11 microRNAs. Tom Tuschl's lab [at Rockefeller University] has been very active in this field also he's had a couple of very nice papers. … We have papers submitted that I obviously can't tell you too much about looking at microRNA expression from a number of viruses and showing that several of the human viruses make microRNAs, and those microRNAs are in some cases highly conserved.
The big thing we're working on now is identifying the exact targets of the viral microRNAs because the viral microRNAs are made by these viruses when they infect cells. The point is that, presumably, the microRNAs are made so that the virus can turn off genes that the host uses to defend itself against viruses. We have preliminary data identifying a target for one of the viral microRNAs that is a protein actually involved in mediating antigen presentation. Given the large number of microRNAs, I think what we're going to see is that a lot of these microRNAs are actually involved in attenuating aspects of the immune system.
From a basic science perspective, it's fascinating. It [also] opens up the possibility of treating viral diseases by blocking their microRNAs. This is kind of the antagomir approach that was published in Nature a couple of weeks ago by the Rockefeller group (see RNAi News, 11/4/2005). If you could in some way specifically target the viral microRNAs, and if these viral microRNAs are critical, then you would be able to treat the viral disease, perhaps even cure it.
The third thing is, these things are presumably microRNAs designed to inactivate human genes, and that offers you the potential to identify naturally designed microRNAs that are a very effective inhibitor of human genes, and potentially those might be genes involved in mediating inflammation or inappropriate immune responses in certain settings. So one of the things a lot of people have thought about with microRNAs or siRNAs is if they could be delivered locally to, say, an arthritic joint or to some location where there is inappropriate inflammation or excessive immune response. You might be able to identify viral microRNAs that are actually targeted against genes that you might want to turn off in a certain setting, so some of the reagents might be quite useful. Those are all pie-in-the-sky [ideas] of course, but they show that we are thinking ahead.