Professor, department of physiology
Tufts University School of Medicine
At A Glance
Name: Daniel Jay
Position: Professor, department of physiology, Tufts University School of Medicine, since 1998.
Background: Associate and assistant professor, department of cellular and developmental biology, Harvard University, 1989-1998.
Postdoc, department of neurobiology, Harvard University Medical School, 1985-1988.
PhD, Harvard University, 1985.
Tufts professor Daniel Jay was recently awarded a $140,000 grant from the National Institutes of Health to develop "proteome signatures" in lymphomas (see chart). ProteoMonitor caught up with Jay to find out about what the signatures involve, and what led to the current work.
What is your background in terms of proteomics?
I was trained as a membrane protein biochemist at Harvard - that was my PhD. For my postdoc I was a junior fellow in the Harvard Society of Fellows, which was sort of an independent postdoc position, and for that I developed this technology CALI, or Chromophore Assisted Laser Inactivation, which is sort of the foundation of the work we're doing now.
CALI is a means to acutely and locally inactivate protein function in situ. It basically knocks down protein function. It uses specific antibodies that are tagged with a dye or chromophore, and laser light that is absorbed by that dye. So what you do is you generate very short-lived free radicals, so it's kind of a highly localized photosensitizer.
So basically, you bind the antigen to an antibody attached to a photosensitive molecule. Then when you subject it to light, you destroy the bound antigen, but other proteins in the cell aren't affected. And we had used that as a means to identify what proteins did in cells for many years.
We got into the proteomics work because after a talk that I gave at the National Cancer Institute, they encouraged me to apply our technologies to cancer. This was always around the time that proteomics and high-throughput methods were emerging. So the first idea was to apply this CALI method in a high throughput fashion to identify proteins involved or required for cancer cell invasion and metastasis.
So basically, instead of doing one antibody and one protein and one hypothesis at a time, we were doing large scale antibody libraries directed against potentially every protein on the surface of a cancer cell. Then we would use this CALI approach, or its latest incarnation, called FALI - Fluorophore Assisted Light Inactivation.
How, specifically, did you apply your technology to cancer research?
In collaboration with a company called Xerion Pharmaceuticals, based in Germany, we used a phage display library directed against one particular cancer cell. It was fibrosarcoma - it's an HT1080 cell line, which is sort of a model system for invasion.
That was around 1998. That worked out great. We have about 3,000 antibodies. We've done about 10 percent of them, and the first couple resulted in nice papers - there was one in Nature Cell Biology last June. And we identified two [cancer proteins] that I can tell you about so far - one is an extracellular HSP90, a molecular chaperone. We showed that it had a role in invasion.
What we did is this FALI-based screen. Anything that showed up as a positive - i.e. a significant change in invasion - we then used that to immunoprecipitate and mass spec identify the cognate antigen.
Are there other groups working with this method?
There are a few, including a couple that are looking to combine this with molecular genetics. There's maybe 20 or 30 papers out there. The Nature Cell Biology paper states that it's a very generalizable method. The original idea was the identification of new targets that might be disease relevant.
What's the reason the method hasn't become more popular?
That's a good question. I guess at the time, when we first started it, it was a little physics intensive. Since then, I think there are obviously competing technologies like RNAi that were developed around the same time. Also, this technology was originally mostly applied to questions of neural development, because that was my area before cancer, and I think maybe because it was more focused on that, only people who were thinking along those lines were using it.
So this method doesn't really involve mass spec?
Well, the inactivation process doesn't, but what it does is provide an antibody for mass spec identification. So the mass spec kind of comes in afterwards, because we believe this function-first screen is an expedited route to finding targets.
What did you do after you developed the CALI method as a postdoc?
After my postdoc, I was a professor at Harvard in the molecular and cellular biology department, and then in 1998, I moved to Tufts medical school. I'm currently in the physiology department here.
At Harvard, the research was all on questions of neural development. When I moved to Tufts - that was around the time that I started to work on cancer.
For the neural development, we were primarily interested in questions of axon guidance growth cone motility, so we looked at a lot of proteins that were found at the tips of growing axons, at the growth cones. We didn't identify new proteins, but we did show what the functions of these proteins were in the cells. For example, there are several myosin isoforms, and we showed specific cellular roles for each for them.
At Tufts, most of the first years were spent developing the screen. And then since then, we've found this extracellular HSP90, and most recently a polio virus receptor. The other one is an extracellular matrix receptor called CD44 - but that had sort of been shown before. The other two are novel.
You recently received a $140,000 NIH grant for research on "proteome signatures and target validation in lymphomas." What does that project involve?
The lymphoma grant actually takes advantage of the fact that we have this large library, and the idea here is instead of using the library solely for this inactivation strategy, we would take it and use it as a means to develop a surface proteome signature. By that, I mean to use an array of these antibodies in immunocytochemistry to show relative levels of expression or binding of these antibodies to the surfaces of a particular cell type such as a lymphoma. And then, on the basis of that, use the kind of informatics involved in self-organizing maps that are currently used for microarray studies, but apply them to this surface proteome signature - this array of immunocytochemistry signals, essentially.
So we have 500 of these antibodies, and what we're showing is the differential staining patterns on a particular cell type. So the question is, by comparing different cell types - and this is a hypothesis that we have not yet proven, but we are working towards - can we use this signature as a means of proteomic classification of different tumors, or different cancers?
The first line of this grant is to see as a proof of principle if we can use it for molecular classification. But the other thing is that the self-organizing map approach basically defines several or as many as you want of those 500 antibodies as predictor of identity. And on the basis of that, those then become potential candidates for biomarkers for diagnostics. In addition, given their importance in defining different classifications, we can study the loss of their function with the FALI approach. So we can test if any of those particular cell surface proteins would be important for a cancer-relevant process.
Why did you choose lymphomas?
That was in part due to a collaboration with Philip Tsichlis, who is the head of our molecular oncology research institute, because he had well-defined, or genetically-defined lymphomas in mouse models. I should say we may be switching it away from lymphomas. We're thinking of working with breast cancer cells. It's obviously a generalizable approach, and we may switch to different approach based on what's the best proof of principle for these ideas.
What do you do with the important proteins you discover from the initial screen? Do you go on to validate them?
The general strategy in the lab is to based on showing its initial importance, to then try to understand cellular mechanism. Then we can test [the effect of the key proteins] on animal models.
Does the company that you are collaborating with help in developing these key proteins? [Jay co-founded Xerion in 1998.]
They would, except for the fact that they're currently undergoing some financial problems. They may be bought out by another company. Tufts does have a licensing agreement with Xerion, and several of the first proteins discovered by the screen have been patented, or co-patented.
Are any of the proteins you discovered close to being used as the basis for clinical therapies or diagnostics?
No, this is all fairly early. We are applying for grants to sort of follow up some of this work in terms of looking at pre-clinical studies. For example, for HSP90 - we're looking for inhibitors for extracellular HSP90 that wouldn't affect intracellular HSP90, and looking to see if that inhibitor would have an effect in reducing metastasis. But that's at the grant stage.
What other projects are you working on?
We're following up work on brain tumor dispersal because some of the work we found on the screen also affect brain tumor migration. As you probably know, one of the most awful things about brain tumors is they just spread in the brain rapidly, making current therapies ineffective. So for example, polio virus receptor is one of the proteins we found that has a role in brain tumor dispersal, and we are following that up.
We're also following up HSP90, particularly in breast cancer, and we're obviously continuing our screens.
What are your goals for the future?
We would like to ask how comprehensive can such a functionally proteomic screen be? In other words, we have a few thousand antibodies, but is it enough? So understanding how would we test if it is indeed comprehensive, and secondly applying it to other diseases and designing parallel screens that would test for those. And for anything we discover on the screen, really validating its importance in the disease, and studying it in animal models.
Is your current NIH lymphoma grant renewable?
Yes. This is a so-called R21/R33 vehicle for NIH. The first two years are at the current rate or about $140,000 per year, then the grant mechanism reviews whether or not you've reached your milestone, in which case the grant increases markedly, probably to about $600,000 per year for a couple of years. The milestones are basically proofs of principle that this methodology could be used as a molecular classification.