At A Glance
Name: Michael Tainsky
Position: Director, program in molecular biology and genetics, Karmanos Cancer Institute, Wayne State University, since 1998.
Background: Professor, tumor biology, University of Texas, M.D. Anderson Cancer Center, 1985-98.
Senior staff fellow, laboratory of molecular oncology, division of cancer etiology, NCI, 1981-85.
Scientist, NCI-Frederick Cancer Research Facility, 1977-81.
PhD in biochemistry, Cornell University, 1977.
BA in chemistry, New York University, 1971.
You have a different approach to finding ovarian cancer biomarkers than that which is being pursued by others ...
What I decided to do was create the test from the ground up, and in a way that it could be applicable to standard diagnostic platforms. We look at auto-antibodies in the serum of patients [that] react with ovarian biomarkers — amino acid sequences from a library of ovarian cancer genes. We use a biopanning technique to isolate recombinant clones.
Can you explain biopanning?
We clone amino acid sequences — or the genes for them of course — that react with antibodies that are present in the patients’ serum, but not in healthy women’s serum. By doing that cycle after cycle, about five times, we find we can enrich at each cycle to the point where the majority of the clones we have will react with antibodies in the patient’s serum but not in the healthy women’s serum. The goal there is to then validate [these clones] as markers, using protein microarrays. So we take individual clones that we biopanned, and we check them for reactions with healthy women’s serum and with patients’ serum. We’ve tested probably 380 sera by now — that’s probably two-thirds healthy women and one-third cancer patients. Among the cancer patients, about one-third of them are stage-one ovarian cancer, because that’s the place we’d like to diagnose it — when it’s curable with a scalpel.
So these are women who were already diagnosed.
Correct. Right now we’ve completed proof of principle to say our method can be used to detect a woman with cancer — even stage 1A — and it’s not fooled by a woman who’s healthy. For example, it’s possible that some of these markers might react with other cancers, or [with] other conditions. So the next proof-of-principle types of experiments — and those are in progress — are women with autoimmune diseases, women with other cancers, benign gynecologic conditions, pelvic inflammatory diseases. We want to make sure that we can discriminate all of these things.
The way we do that is by amassing a huge database of reactivity against thousands of clones on a microarray with these different sera that are characterized from their origin — whether the woman is healthy or had cancer. So it’s basically a fancy immunoassay on a chip. The antigens are on the chip, and the antibodies come from the serum. We use two-color fluorescents to normalize every spot, and then check for the presence of immunoglobulins in the patient serum as the signal. We then put the pattern into a neural network, and it distills out the pattern that is indicative of cancer and the one that is indicative of being healthy, and we use that combined pattern to determine who has ovarian cancer and who does not. So there are two tiers of this. We select differentially between clones that react with women with cancer and their antibodies, versus healthy women. Then the computerized chip analysis — the reactivity of those different kinds of sera on the chips — really distills out the best markers among them.
What specificity and sensitivity are you seeing so far?
Current numbers are in the 93rd percentile for each. That’s on a small panel of 480 markers. We’re now working up 3,200 new markers on a chip to increase the specificity and sensitivity. The first experiment is just a proof of principle — [to show that we] could we do it in a small panel of easily-obtained markers. So we’re robotically spotting microarrays, we interrogate them, put the data in databases, and then the computer scientists, led by Sorin Draghici, help us identify the best markers. Many of them turn out to be tumor antigens. I can’t give you an idea of what they are — that’s significant intellectual property. But some of them are known tumor antigens. Many of them are not. Many of them are sequences where there’s no IP. So that’s the very good news.
So do you know what each of the markers is?
Sure. We just sequence the clone. It’s a quick PCR.
You’ve started doing validation type experiments — what have you been seeing so far, in terms of whether they react with other diseases?
Those are just being collected now. A very short panel said that there’s minimal reactivity. There doesn’t appear to be reactivity with most other cancers. Occasionally we see some, but the good news is, because we have thousands of markers, any marker that reacts with, say, a breast cancer patient, we can pull that marker out of the panel. We have no shortage of markers.
This is really a discovery platform. The implementation platform we would expect will be a standard laboratory ELISA test for serum antibodies. What I like about it is it’s kind of like the ‘people’s test.’ It doesn’t require fancy technology. I’m not looking to market a machine. I’m using off-the-shelf technology for all the chip work. We really devised this so that it can be put into any hospital — any lab that can do an immunoassay — tomorrow, as soon as we get the right productizable substrate. No fancy proprietary technology here.
So that’s one of the potential problems with other ovarian cancer diagnostic approaches?
Yeah. The companies we’ve been talking to already make immunoassays for hospital use. Every time they deliver a new assay, they have to retrain people on how to get that particular kit to work. Even when you just peel open a box, open a bottle of new reagent, there’s a learning curve. So compare that to the learning curve involved in processing samples from mass spec.
[Also], as a sample preparation, antibodies in serum are extremely stable. We’ve used serum banks that are decades old that work fine. We’ve done torture tests on blood samples — things like taking two blood samples, putting one in a 37-degree incubator over the weekend, and the other one processing as normal, and then comparing them. They are perfectly similar. IgG is sort of the reinforcing bar of serum. It’s not a labile small peptide, the way they’re going for with SELDI.
So you would say this is easier to use and more reproducible?
Sure. It’s less finicky in terms of sample preparation. Think about the normal blood test. You go to the doctor and he puts [your blood] in a tube, it sits on a bench until someone gets the chance to put it in the fridge, then it goes into a box, and the box if it’s in the summer it’s hot, and in the winter it’s cold. Will IgGs care? Absolutely not. Small hydrolysable peptides? You got a problem.
So I’m very comfortable with our position. I’m not going to publish it until it’s ready though.
When would you envision having this test on the market?
I don’t know. We have a research plan and a grant that encompasses three years of work. But the bottom line is, it’s an immunoassay. It’s translatable immediately. My own guess is that we could have a first-generation research-grade test out in under a year.
Are there any companies you’re talking with right now for commercialization?
Yes. It’s all the people you know. No surprises.
Who are your funders?
I have some NIH money and about $500,000 per year from the state of Michigan. There was also the Gail Purtan Ovarian Cancer Research Fund. When I proposed this as a concept before we had a dime from anyone, they funded our work for [about] two years.
There are two different approaches to the project, one funded by Michigan and one by the NCI. The chips are slightly different — one uses directly the antigens on the phage as we clone them, the other is to produce recombinant proteins and peptides on chips. By doing it in those two ways, we find out the best way to do it.
How did you get involved with this project?
I come from the position of a cancer geneticist. I was very frustrated [about] patients who had BRCA1 or 2 mutations, and we had adequate screening methods for breast cancer, but nothing for high-risk patients for their risk of ovarian cancer. An ideal test is noninvasive, a simple blood test, it should have multi-parameters — one marker isn’t good, lots of markers takes into account the heterogeneity of the disease. So what we decided to do was get a big panel, and then I got the idea of cloning for auto-antigens — it was known that auto-antigens exist in cancer, but they never really had any diagnostic utility. And it’s just clear you want a big panel of markers. And so I went to the chip. The chip allows you to cast a huge net and find the needles in the haystack that are the ideal markers.
There are a lot of small clones involved — you’re not looking for a full-length clone — you want a nice small epitope, preferably where there’s no IP. And that’s really the case — [for] probably 60 percent of our markers it would be impossible for there to be any IP on them.
Around when did you start the project?
My little epiphany [was in] July 1999. Last June I got my first [NIH] grant. So it took about four years.