At A Glance
Name: Roderick Beijersbergen
Position: Fellow, Netherlands Cancer Institute
Background: Postdoc, Whitehead Institute — 1996-1999; PhD, Netherlands Cancer Institute — 1995; Msc, biomedical sciences, Leiden University — 1992
Since he works so closely with Rene Bernards at the Netherlands Cancer Institute, it is no surprise that Roderick Beijersbergen has included RNA interference in his own research. Recently, he spoke with RNAi News about what he is up to.
How did you get involved with RNA interference?
In the lab next door, of Rene Bernards, is where Reuven Agami and Thijn Brummelkamp developed the pSuper [RNAi] system, where they were able to produce short-hairpin RNAs from an expression vector that allowed us to have long-term suppression of gene expression.
My research in my lab is focused on control of telomerase expression in mammalian cells. I was setting up cell-based assays and screens to identify regulators of telomerase gene expression during the process of immortalization and transformation of human cells. What I needed for that was a system in which I could have long-term inactivation of genes that would normally repress telomerase gene expression [so I could] find these repressors in the process of transformation [and] immortalization.
That’s when … we got together and said, “What we need as a screening tool to do these kinds of genetic screens is a large collection of shRNA vectors that will allow us to do cell-based assays based on long-term phenotypes, rather than just … synthetic siRNAs that were [used to] screen for short-term phenotypes.”
Can you talk a little bit about what you have going on now?
From that point, we entered into [a project focused on] building up a large collection of shRNA vectors that were aimed at a large number of human genes, [in order to be] able to perform genetic screens with gene inactivation. First we started out making a large collection that was targeting 8,000 human genes; this was a gene set that we selected based on their relevance for cancer — cell cycle control and other important processes that are studied especially within this institute that are relevant for cancer like apoptosis, DNA damage checkpoints, and drug responses.
We had that collection … and with that we developed a whole high-throughput system to be able to maintain and use this large collection of vectors. With these vectors, we started to build up screening systems to be able to use these libraries; while we were doing that, we went back to our favorite systems where we, for example, try to elucidate upstream and downstream pathways of the p53 tumor-suppressor gene. In addition, we’re interested in how cells respond to oxidative stress [and] how cells escape or become resistant to certain types of drugs, especially anti-cancer drugs, so we study Taxol and other chemotherapeutic agents. We are also interested in cell-cycle control, not only for the G1S but also the G2M transition, so we do mitotic-checkpoint screens.
We’re really still building up a whole expertise in using knockdown libraries in various aspects of not only cancer but in other areas in cell biology, to be able to identify novel genes in multiple processes and pathways.
Can you talk about some of your recent findings?
One of the things we recently did was enter into a … pharmacogenomic … screen … with these knockdown libraries. One thing we developed was a system called barcode screening. This actually allows us to screen large, complex pools of shRNA-expression vectors in one [cell] population. It’s based on [the fact] that a certain type of stress will give a selection in a cell population against those cells that are sensitized towards a certain drug, and there will be a selection for those cells that are resistant to the drug.
So … when you compare two populations of cells, both contain the same knockdown vectors, and one has received a stress, we will see that there is either an increase or decrease in the relative abundance of certain shRNA vectors that are associated with a specific phenotype. With that we can now recover … the shRNA-expressing cassettes from the genomic DNA isolated from both populations of cells. We hybridize that to microarrays, and we can quantify the relative abundance of each shRNA in a very complex population. In this way we are able to screen with our human library through 24,000 vectors in, more or less, one tissue culture dish.
We also can do screens now very fast. Normal screens for identifying shRNAs would take us up to six months, but now we can do the whole process within three weeks.
The great advantage of [the barcode] system is that it not only allows you to identify vectors that are associated with what we call a proliferative advantage, but we also hope to be able to use this system to identify vectors that either have a lethal phenotype or actually have a synthetic lethal phenotype. With synthetic lethal, we hope to identify, for example, genes that are needed in cells that contain a mutated p53 gene or have a mutated ras oncogene or have an amplified c-myc gene. In this way, we hope to identify those genes that have to be activated in order for tumor cells to be able to survive in the background of another mutation.
These genes, of course, will be excellent drug targets [in order] to have a tumor-specific effect inactivation of these genes in tumor cells … without affecting normal cells.
The barcode system is now mostly used to rapidly identify knockdown vectors in functional screens; so genes that when they are inactivated give a selective advantage for cells under a specific type of stress or treatment such as drugs. We recently used a novel drug that is designed by Roche called Nutlin.
Nutlin is a compound that interferes with the MDM2-p53 interaction, thereby stabilizing p53. What is thought is that when you stabilize p53, [its] downstream targets are activated and these cells either undergo cell-cycle arrest or they enter apoptosis. So this compound would actually activate p53 in cells, resulting in cell arrest or cell death. We were interested in whether this drug was only acting on p53 or whether there were other genes essential for the cytotoxic response.
We did what we call a barcode screen, where we introduce our whole shRNA library collection — 24,000 vectors. We treated [a] population [of cells] with Nutlin, so p53 [is activated] and [the cells] undergo cell-cycle arrest. Only those cells that are resistant will grow out. What we found in that screen is that p53 itself was essential, which we expected — so p53 inactivation results in resistance against this drug. We also found that inactivation of another gene product … also resulted in the rescue of these cells and they were able to grow in the presence of this drug. This showed us that it’s not only p53 stabilization itself that is essential for the cytotoxic effect of this drug, but also upstream signals that impinge on p53 … have to activate p53 in order to get the downstream cytotoxic effects.
With this simple screen, we can get a much better insight in how different types of drugs [act] within the cell.
Did you do this work in conjunction or on behalf of Roche?
No. It was totally independent. We submitted a paper on this … to Nature Medicine.
Are there companies that have contacted you and wanted to get your help?
We have some collaborations with companies that are interested in the type of work we’re doing. We’re setting up collaborations with them where we develop our systems to [allow them] to use them.
Are these big pharmas? Can you comment on which companies?
The company that we actually work with to design and build our shRNA library was Merck, which is still a partner in the development of these technologies.
Are they using the technology in-house?
Both. They’re developing the technology in-house, and they make use of our expertise. We divide some of the development of technologies among the two of us.
Is this Merck or Rosetta?
It started out as collaboration with Rosetta, but now of course it’s Merck [following the acquisition of Rosetta].
You talked about the pSuper library starting out with 8,000 shRNAs. What’s the collection up to now?
We’re now building a mouse collection, and this will target 15,000 mouse genes. At our institute we have large effort in generating mouse models — we have a large collection of different conditional knockout mice for tumor-suppressor genes and a large number of transgenic models. There are, of course, cell lines available from those models in which we could very nicely do genetic experiments. So, there was a lot of demand for [a] collection for mouse genes, so we designed and built a collection targeting the mouse genome.
That’s what we’re doing at this moment. Over the next year we’ll probably start expanding the human collection as well. We have some big projects going on where there’ll be at least a few thousand genes added to the human collection.
The 8,000 genes were all related to cancer, right?
In the broad sense; it’s more like related to important biological aspects of cells. So there are also genes in there that are involved in metabolism, and processes like ubiquitination, de-ubiquitination, protesomal degradation pathways, gene sets involved in interferon gamma responses, phosphatases, kinases — the whole kinome is in there. So it’s really … broad.
We didn’t include unknown genes. Most of them are known genes, except for a few genes that were mapped as disease genes without a clear function. But most of the genes have been annotated for their function.
Aside from the mouse work, are there any other projects you have going on in the RNAi field?
One of the things is there is a lot of debate in the field about the off-target effects of siRNAs. So that’s something — we’re trying to get a better feeling for what actually the off-target effects are. Because the barcode system we’re using [allows] us to rapidly screen through 24,000 different shRNAs in one phenotypical readout, we’re actually in quite a good position to do many of those experiments and compare them to each other. So we hope to get a feeling of what kind of problems can be associated with these types of screens by doing a lot of them and comparing their results.
We also hope to get a better insight into … the design rules for these shRNAs in a hairpin-based vector. A lot of the design rules now are focused on the generation of synthetic siRNAs, and they seem to be pretty good. But we think there are different rules [for having] a good shRNA that produced as a hairpin in cells [than those] for synthetic siRNAs.
We’re trying to get better design algorithms and design rules for small-hairpin RNAs to be able to efficiently knock down any gene of interest.
Can you talk about what may be some of the differences between shRNA and synthetic siRNAs that lead you to believe the design rules would be so different?
One thing we know is that when we apply the rules that are generally used for siRNAs, they are not guaranteed to work for shRNAs.
A big difference between siRNAs and shRNAs, to my understanding, is that siRNAs are rapidly saturating the system, so you cannot actually introduce different siRNAs in the same cell to knock down multiple genes.
With small-hairpin RNAs, you can introduce several different expression cassettes for these hairpin RNAs within one cell, and you can inactivate expression of multiple genes within one cell. So it seems there is also a difference in how shRNA are brought into the silencing complex and how they are unable to saturate this process in cells.