At a Glance:
Name: Laurent Daviet
Position: Head of target discovery, Hybrigenics, since 1999
Background: Postdoc, Victor Dzau’s laboratory, Harvard Medical School, 1997-1998; Postdoc, Institut Chochlin de Genetique Moleculare, 1995-1996; PhD in molecular biology and biochemistry, University of Lyon, 1995
This month, a research group at the Paris-based biotechnology company Hybrigenics published a protein-protein interaction map of the fruit fly model organism in the journal Genome Research. The interaction map contains many homologues of proteins known to be involved in human cancer.
ProteoMonitor caught up with Laurent Daviet, the leader of the study, and the head of target dicovery at Hybrigenics, to find out about his background, and to find out more about the work being done at his company.
What were you doing before you joined Hybrigenics? Were you always involved in doing yeast two-hybrid studies?
I got my PhD in 1995 from Lyon, France. Then I did a first postdoc in Paris where I started to do two-hybrid. That’s where I met the people who were involved later in funding Hybrigenics. So they were all researchers in different institutes in Paris — the Pasteur Institute mostly, and the Institut Curie also. Then I moved to the US, to Boston. There I was also interested in two-hybrid, in trying to understand a signaling pathway regulated by a receptor.
What was the signaling pathway that you were studying using yeast two-hybrid analysis?
So it was the Angitensin II receptor pathway in human. It’s one of the major receptors that regulates blood pressure. I used two-hybrid to try to understand the signaling pathway that is regulated by this receptor in the laboratory of Victor Dzau at Harvard Medical School. I did that for two and a half years. During that time, the people I had been working with in Paris were actually funding Hybrigenics, so they asked me to join them. And that’s what I did in 2000.
I was first involved as a project leader in the development of two-hybrid platforms. Hybrigenics from the beginning was really a technology platform company. I was involved in building the yeast two-hybrid technology platform, mostly in higher eukaryotes. Before I joined Hybrigenics, they did a lot already with microorganisms. They published in Nature in 2001 the first genome-wide interaction map of Helicobacter pylori.
I was more involved in mammalian proteomics approaches. I was in charge of this work that published recently in the March 1st issue of Genome Research, on Drosophila. This work was done together with the Institute Curie — a cancer institute in Paris. It was a big collaboration where we looked at using our technology platform to look at signaling pathways that are important in human cancer. We did that in Drosophila as a model organism, but we also did it in humans, although this study is not published yet.
The simple idea was to do the pathway analysis first in a model organism, where most of these pathways were conserved, and then also in human. By doing that, when we can fish out interactions that are conserved in the two species, then we are much more confident that they are real. Because Drosophila has a reduced genome compared to humans, we can analyze it more exhaustively, and we can also infer protein-protein interactions that we find in Drosophila that we may not find in humans, because to reach the same exhaustivity in humans in more difficult because of the larger number of genes.
Now we are working on this comparative analysis between Drosophila and humans — that’s still in progress. We are developing specific tools especially to find orthologues between the two species. That’s currently the most difficult part that we’re dealing with.
When you joined Hybrigenics, were there a lot of other companies that had similar platforms of using yeast two-hybrid for protein-interaction analysis?
Yes, sure. Mostly in the US. The biggest one is Curagen, and there’s similar work being done in Salt Lake City by Myriad.
How did you try to differentiate your company from other yeast two-hybrid-based companies?
The way we did it is first, we really believe that the bait design — the protein that you are going to use as a bait in your two-hybrid — has to be designed in an appropriate way to be optimized for two-hybrid screening. That’s what we did in our Drosophila paper. You can start with very basic things, like if you have a membrane protein, you may choose not to use the full-length protein — that’s what people are mostly using in large-scale approach. I think that with the full-length membrane protein, there’s very little chance that it folds properly in yeast and gives any interesting results. So the bait design was really important for us. For that we collaborated with a number of different labs that were specialists on each protein that we were analyzing, and they were designing the most appropriate bait for interaction screening. [The bait] can be also mutants with well-defined biochemical and functional properties that we know are the most suitable for the interaction screening.
The other thing is the coverage of the screening — When you do a screen against a library, you can choose to pick out very few clones. That’s what Curagen did in its recent Drosophila study — they were analyzing 10 or 12 clones per screen. Well, we cover the library extensively, to saturation, to make sure that we get any interactor, even RNA interactors that are present in the library. The process is very different, and we had to adapt the technology to reach this goal. We have proprietary mating technology that we use for the yeast — we don’t transform them the way people do it most often. We mate them. We have a bait population and a prey population expressing the library, and to cover the complexity of the library, we have this specific mating protocol that we’ve developed.
Besides that, we’ve also developed bioinformatics tools to analyze the screen, to put what we call a confidence score on each interaction based on a statistical analysis of the result. One of the major drawbacks of the two-hybrid is the false positive, and we have a way to label them using the same library. We can figure which domain or which proteins are ‘sticky,’ and stick to any bait that we use. We have this ability to tag proteins that are sticky. And we have different categories of interactions ranging from high-confidence interactions to “sticky proteins”.
We have also developed a database where we store the results and make all the process of pre-identification, calculating this scoring for each interaction. That’s also where we’re different from our competitors. Our database is controlled by a laboratory management system that can handle more than 2,000 screens a year.
Right now we proposed to put our technology out on a service basis. So if somebody has a bait and wants to use our technology to do the screening, we have a number of libraries that have been made in-house that are high-complexity libraries. We have very strict quality control on the libraries. That’s open to the outside as a service. So if you have a bait, you can send us the bait and we will apply our internal technology to the bait to find partners.
Do other companies have their own statistical scoring methods?
Recently, Curagen published something, but it was much less sophisticated than what we did. So it was like a training set. They looked in their database for interactions that they knew were true because they were published, basically, and then after that they tried to find parameters to discriminate these true interactions. They applied these parameters to the whole dataset to get high- and low-confidence measures.
The evaluation that we have is much more subtle. Because we do this saturating screen, we really have this statistical analysis that can not be made if you picked out just a few clones from each screen.
Aside from the Drosophila and human, are there other organisms that you are working with to investigate protein-protein interactions?
For our internal program, we mostly work on Drosophila and human. We have other organisms that we may screen, but we do that only for very specific proteins that we are interested in. The point mostly is to validate interaction, and Drosophila was really interesting for that because you can manipulate them genetically much more easily than mammals. So it’s a tool for the functional validation of interactions.
If you want to do a transgenic, the time is much less in flies than in mice.
Are you collaborating with any pharmaceutical companies for drug development?
Right now the drug development is being done internally, but we are collaborating more in the target-identification and validation level with other companies.
What is the next step in terms of validation after you identify a number of potential drug targets?
First is this protein-interaction map. We try also to combine data from the protein-interaction map with expression data. We generate that internally, and we also work with companies that provide data about gene expression under different conditions.
Also, we have been developing a cell-based, high-throughput screen mostly based on RNA interference. So we have designed cell bait based on pathway-based screens, and we can basically interfere with siRNA to look quickly at the functional level. That’s what we are interested in — to find out if altering the function of these genes has any effect on phenotypes.
We have mostly oncology maps, but we have also interaction maps for cell differentiation, cell-cycle progression and cell migration. We have also pathway readouts, so that you may look at, for example, what happens to the Ras pathway if I silence some specific genes using siRNA.
Then we move quite quickly to developing drug assays for compound screening. We try to move to the drug discovery phase early in the target-initiation process.
Aside from your work on cancer pathways with the Insitut Curie, what other pathways are you working on?
One of the other major pathways that we’re working on now is the ubiquitin protein pathway. We have data that shows that protein stability is regulated by modulating their ubiquitation. That is very interesting to us, so we are really focusing on that class of targets right now.
What other projects is Hybrigenics working on?
Another thing is we’re trying to figure out is if there are links between pathways. When you have a map with thousands of protein-protein interactions, figuring out this [linkage] map is not a trivial task. We’re interested in how different pathways talk to each other. A map is the first step to identifying this link to what we think is functionally important.
What is the next step after you have validated drug targets using siRNA studies?
First we will go in animal models — that’s the next step for sure. We’ve started to do mouse studies — toxicology and so on — for the first compound that we identified. That’s something that we are really coming into right now, with collaborators from the outside. We collaborate with a company in Paris that is specialized in xenograph mouse models.
In terms of testing in humans, that’s something we’d like to do down the road for sure. We still have all the toxicology studies to do, so it’s still too early to speak about clinical trials in humans, but that’s the future goal for sure.