Era of Hope 2011 Plenary Sessions

Session Title: Morning Session – Roads Less Traveled: New Perspectives in Breast Cancer Research

Title of Presentation: Single Molecule and Single Cell Analysis for Cancer Diagnostics

Stephen Elledge, PhD, Harvard Medical School: I’d like to introduce the next speaker David Walt. And the title of his talk is Single Cell and Single Molecule Analysis.

David Walt, PhD, Tufts University School of Medicine: Good morning. I’d also like to thank the organizers as—as well as the advocates. I’m—this is new to—new to me. I’m not a cancer researcher. This is the first—I haven't yet got my funding even from DoD BCRP but I’m about to this week I’ve been told. So I’m coming from outside the field and this has genuinely been an inspiration.

So the goal of sensitive detection for breast cancer which is why I’m going to talk about single molecules and single cells, there’s three things that I can think of. One is routine screening and routine screening equates to early detection. Therapeutic efficacy that is for patients who are on therapy, one would like to know is—is the therapy actually doing anything. And the third is for recurrence monitoring—recurrence monitoring which relates to the third one but it’s not—not—to the first one but it’s not exactly the same.

And so the—right now when we look at the analysis of proteins in the blood, the serum proteome, amount—the number of IVD tests in the ranges of concentration that we’re talking about here, millimolar and micromolar down to picomolar is on the order of about 120 different proteins. And—and the idea here is that one really would like to extend that range down to the proteins that we know are present in blood but we just can't measure. And that equates to on the order of another—another 2,300 proteins. And so I like to think of this as kind of the—the proteome iceberg. What we’re measuring now are the things that are just at the tip of the iceberg but what we’d like to be able to do is measure those things that—that appear that—that we can't presently measure in—in blood.

And so the technology that we’ve developed in my laboratory that I’m going to apply to breast cancer is—is really depicted here. And this is kind of the way that we do clinical analysis today; the way we quantify things is we just look at the intensity of the color. So this is a typical immunoassay for a protein called an ELISA, and it goes from at low concentrations you have a low color; at high concentrations you have a darker color. And what we’ve developed is a way that allows us to literally count the number of molecules that are present. And what it does is two things; number one, it’s much more accurate. You can count the molecules instead of sort of looking at an aggregate signal, and the second is that it’s much more sensitive. That is you can see things that you couldn’t see before.

And so this is my technical slide, so I apologize to—to those of you that—that are not technical but for those of you that—that want—want to know what the technology is about, basically what we do is we take a blood sample. We throw in a bunch of beads into the solution. We capture proteins on those beads that are present in the blood and those proteins that get captured on the blood then we form a sandwich immunoassay so we just a detection antibody that has an enzyme on it. We then take those beads and the beads have either one molecule or zero molecules on them. So that’s how—how this thing—thing actually works. So that allows us to go to the single molecule level—so we have many more beads than we have molecules in solution. We then take those beads and put them into those little micro-wells, capture the beads in the micro-wells, seal the micro-wells with—with just putting some oil over them capturing a sub-straight solution in there and then when those enzymes that are attached to those beads that indicate that there’s a protein attached to them, catalyze the reaction, we get little dots where a protein molecule is. And we just count those numbers of dots and those numbers of dots tell us how many protein molecules are in the solution.

So I’ve founded a company here in Cambridge; it’s called Quanterix and I have to disclose I’m Director of both Illumina and Quanterix. I’m the founder of both—I’m the scientific founder of both of those companies. I know a lot of people were talking about the Illumina Technology. Thanks for using it. But I—I do want to digress briefly because there was a comment last night about putting things in the public domain. And—and what I would—what—what I would tell you is that if some things are not patented then they will never get in the public domain because nobody will use them, nobody is going to invest in them to make the—the—to essentially build a company that then makes money building those instruments.

So putting things in the public domain is great but sometimes it’s a barrier to actually getting the things that we all use in—into the public domain.

So I want to give you two examples of the use of this technology, one for cancer and one for neurological disorders and then—and then extend that conceptually to breast cancer and show you what we’re—what we’re planning to do. So the idea that we had in using this and this was done in conjunction with my colleagues at Quanterix is that prostate cancer surveillance, when men have a radical prostatectomy the problem right now is that the existing tests do not have the sensitivity to allow one to measure PSA in the blood. And PSA is a horrible marker for screening; it—there’s a lot of false positives, there’s a lot of false negatives, but primary false positives. But for men who have had surgery PSA goes down within about 3 months to a level that’s—that’s unmeasurable. But you don’t know what’s happening in this dead zone.

If this is the concentration at which one can detect, if the tumor now is—is growing, if there’s a metastasis, the problem is that you don’t get biochemical recurrence until you get over the level that’s able to be measured clinically. And so what we’ve done is develop—use this technology that we’ve developed, the single molecule measurement to actually drop the sensitivity down way below the present methods. And let me show you; this—this just—this is sort of in the present concentration regime, the cut-off now is—is right around here for the ultra-sensitive test and our test goes down and then continues to go down so it—it goes—goes down roughly 1,000-fold more sensitive than the existing tests.

And so this was just the first sort of proof of concept that we could use this test to measure things much more sensitively in blood than—than had ever been measured before. These are normal patients, their normal PSA levels; these are men who have been 8 years out from their—from their surgery and this is the detection limit for—for those men. The ultrasensitive test actually is around 10 picograms per ml and so all of these men would have detection—would come back after their PSA test 8 years saying nondetectable. We can quantify every one of these men and the difference between this man and this man is 1,000-fold. So I’m not sure if that’s clinically relevant but it certainly makes a lot of difference if you can put a number on something than if you can just say nondetectable.

So we then decided to do a study that—that actually looked at outcomes and so these were again done with bank samples. It was a longitudinal retrospective design. There are 11 patients that recurred, 20 non-recurring post radical prostatectomy patients, and so here—here are the results. This was done over a period of time. These samples were—were taken 3 to 6 months, 6 to 12 months, and 12 to 18 months after their—after their surgery and what you can see is that these men, the ones in red all recurred and—and they all had levels that—higher levels. These men in black did not occur. This was a 5-year retrospective study and—and so this is the conventional test right here at the—at the top if I can get the mouse moving here—this is the conventional PSA test. This is the limit of detection here and this is the ultrasensitive right here at 10 and then the—our method actually is able to quantify everybody below here. And so what you can see using the Kaplan Meier time to biochemical recurrence what we had—what we were able to do is with a 1-month, with a 3-month to 6-month Nadir sample—that is one sample taken 3 to 6 months after their radical prostatectomy we could—we had 100% predictive ability on whether those men would be able to recur and all the men that were above that cutoff level at three picograms per ml actually recurred within 5 years. And so this—this is what high sensitivity measurements actually are capable of.

And so we—we still need to do larger studies to validate these findings but—but it’s able to classify men as very low-risk which is a psychological reduction in—in stress and—and so again we—these need to be—need to be validated with larger numbers.

The second example is brain hypoxia; this is in conjunction with [Ki Blenau] at the University of Gothenburg. He’s banked a bunch of samples from patients who had cardiac arrest and the problem with cardiac arrest is that it—it deprives your brain of oxygen and so—so your brain becomes hypoxic and the big problem is that one cannot predict what the cognitive outcome of those patients is going to be. So—so you have—somebody has a cardiac arrest. They—their brain is—is damaged to some extent but no one knows whether they’ll be able to recover or not.

And so what we—what we did is we measured two different proteins, serum tau level and A-Beta 42; these are proteins that are present in the brain that are also associated with Alzheimer’s disease. This is a good outcome for a patient if they have—these two markers have a particular trajectory. Here is a bad outcome. You can see that one of these proteins goes up. Again these are very low concentrations. They cannot be measured using conventional tests. And so using—using this method just to show you the—the sort of bottom line here is that we can predict with 100% sensitivity and 100% specificity what the cognitive outcome will be of those patients 6 months down the road using a blood test measured over the first 4 days after—after their heart attack.

And so again this is not breast cancer but I think it—it shows you the power of being able to measure things in blood at a much more sensitive level.

So here’s what I think is the future of breast cancer screening. And—and I could not find anything on the web that really showed what—what we’re trying to accomplish here because this tumor is—is too large. Dr. Begley in his talk last night said look, a millimeter cubed is a million cells. A centimeter cubed is a billion cells. We’d like to catch this earlier. And so we think we have a method that would allow us to detect a nascent tumor at the million-cell level—that is a millimeter, a millimeter cube and so this is definitely too large for—for us. We don’t want to detect things at that level; we want to detect things way before that. So a millimeter cube would be incredibly small compared to this—this kind of—this kind of diagram. And the reason that it’s—it’s a problem detecting that in the blood is because there’s just not enough stuff being put out by that tumor. It just gets diluted by the rest of your body fluid. And so if one could take a blood test and then put it into an instrument one would have the opportunity to routinely screen for breast cancer and—and—and one could imagine things like this. You just continually get screened, so these are five consecutive blood tests. And the green you see an individual that has none of—no marker that’s elevated. This person in red might begin to have a nascent tumor that’s growing and the person in blue as we heard Dr. Lee talk about maybe the immune system takes care of that and you don’t have to worry about it after—because the immune system simply takes care of that nascent tumor and it just disappears over time. You don’t need to look for it anymore.

But this is what screening enables you to do and I would—I would argue that this is a lot easier if you go to your doctor every year, all women get this screen from—from early childhood all the way up so they get baseline levels of these kinds—these kinds of things. So that’s I think what—what we’re hoping to accomplish here.

Okay; so the—the other thing that I wanted to talk about was single cell measurements. Before—before I move to this I do want to urge any of you in this room, any of the scientists who have markers that they think would be specific for breast cancer that they think would appear in serum please e-mail me. I’d be happy to get my—students in my lab looking at that marker and developing assays and seeing if this is going to be done. That’s going to be our job at the beginning of this project is really to find those kinds of markers that really are specific and selected for breast cancer.

So the second part of the talk has to do with single cell measurements and the—the reason that this is an important area is because—and—and we’ve heard—there’s been a few talks on single cell measurements but I want to introduce all of you to the topic in a big sort of broad sense. If you have a population of cells, blue and white cells here, and you take a measurement of those cells, let’s say a tumor sample and you grind it up and make a sort of aggregate sample you’ll get an average value for those blue and white cells that’s right here. But it could be that there’s two different populations of those cells and in fact we know in tumors that there are some cells that are much more aggressive than others. And so the idea is you’d like to identify those individual cells that are—or see whether any of those kinds of really aggressive cells are present.

Here’s just another example of why—because cells turn on at different times. If you just take an aggregate signal you just see this red thing but the—the single cells turn on at different times with different—different kinetic outcomes.

So why do we want to use single cell analysis for breast cancer? There’s heterogeneity in this—in this tumor. What you’d like to do is take those cells, array them and then look for the really bad actor in that—in that needle biopsy. So if you get a needle biopsy and you have a few million cells you’d like to know what is it on a cell-by-cell basis, and not what is it on—in aggregate. So the idea is to then array those cells and then look at the—the individual responses of those cells and—and this just shows you some work that we’ve done in the lab with yeast cells, with E. coli cells. The results of a yeast experiment show that you get very different—these are—these are something for the scientists; these two hybrid experiments, you get very different responses from these different kinds of strains of yeast.

What I’d like to point out though that is in a—in a population where you don’t even expect a signal you get rare cells that are actually putting out pretty strong signals. And so if we just looked at the ensemble—that is the bulk of that—of that sample and—and took that measurement you’d get a measurement was very close to baseline yet there are some cells that have very strong reacting properties in there.

So the goals are to develop methods for genotyping these individual cells, doing gene expression on these individual cells, doing protein levels of these cells, and so we’ve developed a method of single cell genotyping that’s related to the single molecule approach and just to show you some—some results on this, this is with E. coli. We haven't gone to mammalian cells yet. We’ve also done it for yeast but trapping individual cells, capturing their DNA in—in these wells and then genotyping them you can actually get signals—you get genotypes for the individual—the individual cells and you can go through this in—in—at multiple loci on the—on the DNA.

So the idea is—is really to—in summary is to bring single molecule measurements to the clinic to be able to do routine screening for both early detection, therapeutic efficacy, as well as recurrence monitoring and that single cell measurements enable the ability to perhaps stratify tumor biopsies and detect those rare cells that can help guide therapy. I’d like to thank the—my students, the funding agencies, and then DoD BCRP—we hope.

Question: So regarding your request for good markers for breast cancer, one that might be a good place to start would actually be tau for which you have a test already designed. Tau is expressed in up to 50% of invasive ductal carcinomas and so the other question that I have for you is when you looked at the cardiac patients did you also look at normal and do your levels of tau in serum stay reproducibly very low so that you could see an increase in tau?

David Walt: So every patient turns out to have a different starting point with—with both A-beta and—and with tau. And—and so you know getting to personalized medicine, having that baseline for individuals is critical and—and so there’s no sort of—we don’t know yet but there’s no normal range for tau or normal range for A-beta. What we look at is the trajectory of those proteins over time after the—the insult. But—but tau is a great suggestion; thank you.

Question: How much variation do you have in the baseline, 100-fold, 1,000-fold?

David Walt: It’s on the order of about 10-fold.

Question: You mentioned the ability to look at genes on the DNA. I was wondering whether you have thought about gene signatures, looking at transcripts.

David Walt: We—we have; the—the thing about transcripts is that what you have to do with transcripts is you have to take the single cell. You have to break it open and then you have to sort of use the single molecule detection approach on a single cell basis in a multiplex manner. So you—you can't just put the cell in the well and then do gene expression. You have to put—take the cell, break it open, capture the transcripts, deposit those using encoded beads, and then do that. And yeah; yeah we’re doing that.

Question: To have a good breast cancer screening biomarker test, you’re going to probably need to look for more than one biomarker. And I was curious about the multiplexing capabilities of your method.

David Walt: Yeah; so I just alluded to that. We’re at the point where we can use a single molecule method to do a about a 10-plex. We’re working probably to a 20-plex; we—we don’t think that’s going to be a problem. Beyond that we may have—we may have some issues just because of some nonspecific binding issues, but—but—and cross-reactivity but we think we can get to certainly 10 and perhaps 20.