Personalized medicine may soon hold true to its promise for cancer detection and diagnosis . . . so skeptics beware.
You love your doctor. Admittedly, he is not the best doctor in the world, but he knows you. He’s taken the time and asked the questions and considered your answers and nodded his head, and for you, this is invaluable. And in the future, this focus on the individual, this increasingly deep insight into the concerned patient in the office and not the diseased human being at large, may very well save your life.
A life-saver with a broad understanding of what it will take to gain that insight is William Hait, MD, PhD, president-elect of the American Association for Cancer Research. Hait points out that personalized medicine will take two forms: diagnosis and prognosis. And for early cancer diagnosis, steps forward have been slow. He first considers imaging techniques. “There have been some improvements—PET scanning, digital mammography, breast MRI, and helical CT. So, early detection is getting better,” he says, trying to be optimistic. “But it would be hard to point to a true breakthrough technology in screening for early diagnosis.”
Lacking a good imaging screen, might there be early detection with biomarkers? Medical meetings are replete with such posters. But here too, only the promise is as yet
|The Thousand-Dollar Question
Oncology will take the greatest stride toward personalized medicine with genetics—be it nature or nurture, germline or somatic—mutations are the key; get the sequence, spot the trouble, and plot your way around. It is both a simple concept and a profound challenge. Yes, the map is drawn with only the four letters; A, C, G, and T, but it takes piles of money and weeks of time just to get the thing unfolded.
To move forward, low-cost, high-throughput sequencing is essential, and Stanley N. Lapidus, president and CEO of Helicos BioSciences, Cambridge, Mass., thinks he has found the way: True Single Molecule Sequencing. “The advantage of doing single molecules is it allows you to do direct sequencing of the DNA from the tumor. The prep is simple; the reagents needed are minimal; and because of the single molecule design, the density can be quite high—we can put a million strands of DNA on a square millimeter.”
The device, called Heliscope, captures a multitude of single-stranded, enzymatically-clipped molecules of antisense DNA on the proprietary surface of a flow cell. DNA monomers labeled with a fluorophore, are then sequentially added, via DNA polymerase, to the bond sequence. If a complementary base is found, the newly paired base is detected and its position digitally recorded. As the technology is optimized, says Lapidus, maximum output will approach a billion bases an hour. “That’s a rate that would allow for the complete sequencing of a tumor cell in a single day. Right now, that job would take a hundred sequencers working for six months.” And it wouldn’t be cheap.
in hand. “We’re in the dark ages,” says Hait. “Take the big cancers: The number one killer of men and women is lung, and we don’t have a marker for lung. The most prevalent malignancy is breast, and we don’t have a good marker for breast. Prostate is the most prevalent cancer in men, but PSA is an imperfect tool. Ovarian cancer and CA125—this is not a useful screen to detect disease at an early stage.” Pancreas—no marker; non-Hodgkin’s lymphoma—no marker; they’re there, he insists, he knows they’ll be found, but discovery will take more time.
For use in prognosis, the science of biomarkers is more advanced, and this has everything to do with the advent of targeted therapies and their cost. “These medications are very expensive, so it would be really good to know who’s going to respond and who’s not.” It’s not like, say, treating with an ACE inhibitor, where nearly all patients will experience a change in blood pressure. For agents like Tarceva, or Iressa, or Herceptin, there’s only a minority of patients who will respond—as low as 10%. So, the goal is to identify that 10% before you treat. “There’s a huge interest in this because of the finances,” says Hait. “It’s a perfect storm of confluent interests—the patient would like to know, the doctor would like to know, the insurance company, pharma, everyone wants to know.”
Finally, as he scans the diagnostic horizon, Hait sees the signal fire that is the predictive nature of genetics, and specifically, Single Nucleotide Polymorphisms (so-called SNPs, promounced “snips.”). “It’s going to be these individual subtle variations between us that, even though we’re 99% genetically identical, you can still tell me from you and you from your brother because of these polymorphisms—and these subtle variations are the new frontier of determining cancer susceptibilities.” In the not too distant future, your family doctor will, on the day you first meet, take a blood sample, pinpoint your variations, and present you with a map of your personal world of risk.
While there is always hope for the future, if one is diagnosed with advanced lung cancer today, your chemotherapy treatment will be limited to one of four regimens, all of which contain a platinum-based drug as a backbone. “But it’s a random choice. . . . ,” says Anil Potti, MD, assistant professor of medicine at the Duke University Institute for Genome Sciences and Policy, Durham, N.C. “. . . random in the sense that all these regimens have been shown to be equal in large clinical trials. So currently, doctors choose that which they’re most comfortable with.”
Potti is determined to find a better way. Using the NCI-60 panel of cancer cell lines, and a gene chip from Affymetrix, he’s taking a genetic look into drug response. “For a cell line that has been treated with, say, a taxane, some lines
|Sorting It All Out
Prior to diagnosing a cancerous tumor, one has to find it (an image) and sample it (a biopsy). Characterization of the tumor then follows. Speeding this process in the future is the technology of CellSearch, an innovation from Veridex LLC, Warren, N.J. Recently cleared by the FDA to aid in the monitoring of metastatic breast cancer, CellSearch is able to collect and identify the main instigator of cancer mortality, the circulating tumor cell (CTC)—the seed of new metastases.
Robert McCormack, PhD, vice president of clinical and medical affairs at Veridex explains its current application: “The test has two intended uses; it’s prognostic and predictive. Women who at baseline don’t have CTCs in their blood, are on a very favorable prognostic curve as compared to women whose CTCs are elevated.” On the other hand, beyond a certain threshold of rising cell counts, it’s highly likely that the patient is failing treatment. By using the presence of CTCs as response criteria, the oncologist can be tipped off to a failure sooner than radiographic evidence would indicate.
The technology that allows CTCs to be gathered is enabled by magnets. “A tiny iron particle is conjugated to a monoclonal antibody raised against EtCam, a cell adhesion molecule unique to epithelial cells,” says McCormack, explaining further that for adults, 90% or more of cancers they develop are epithelial in origin.
The automation of CellSearch stirs this molecular construct into 7.5ml of the patient’s blood. Unbound cells are washed away. Then a series of stains are applied—again, epithelial-specific—and the remaining pellet is scrutinized by an optics station with a specially configured fluorescent analyzer. Then, the robots step back, and a technician has a look. “The final step requires the human eye” concedes McCormack. “The regulatory environment still wants the human [to sign off] on the result rather than have an artificial intelligence telling us it’s a tumor cell.”
This non-invasive cell sampling technique has further prognostic utility in tumor phenotyping. Though currently only available for research purposes, Veridex has reagent kits to identify tumor cells expressing HER-2/neu, the EGF receptor, and the glycoprotein MUC-1.
are sensitive, some are resistant. So, we take the extremes of these responses and look at the associated gene expression.” With these data you then ask the question, how do the expression patterns of these two cell lines differ? “If you treat a hundred patients with lung cancer we know about only 30% respond to a taxane, even though all cells have a cytoskeleton (the cellular target of a taxane).” Explaining this phenomenon is the multitude of disruptive events that happen in a cancer cell, and this can only be profiled through the analysis of gene expression.
Now back to the patient. “So if you use a genomic approach to pick the second drug, you are going to do much better than a random choice.” Alternately, expression data may tell the researcher that the backbone itself is wrong and the optimum combination is a mix that has yet to be contemplated. “The latter is the more distant opportunity—we have to prove it first in a prospective fashion,” says Potti. But he’s certain of his approach, and thinks that validation is just a few years off. “It’s inevitable that we look at either genomic or proteomic approaches to try to better define the patients that will benefit from chemotherapy, and once identified, those who are most likely to respond to a certain agent rather than a random shot in the dark.”
Helping to shine light on the situation is David Mankoff, MD, PhD, assistant professor of nuclear medicine at the University of Washington Medical Center, Seattle.Mankoff is a specialist in the use of positron emission tomography (PET).
“Let’s say I’m going to try brand X version of treatment,” says Mankoff. “What I’d really like to know as soon as possible is if brand X is a winner, so if it’s not, I can move on to brand Y.” And the healthcare provider can make that determination much sooner with FDG-PET by looking at functional changes rather than trying to detect a change in tumor size.
Using this rapid assessment of response, says Mankoff, “it might be informative to image very brief treatment trials to choose a compound that looks like it has the best chance of working. For example, studies have shown that FDG metabolism in GI stromal tumors changes as early as 48 hours after starting Gleevec.”
Despite this and other successes, Mankoff admits that acceptance of this strategy will take time. “It’s a new technology, a bit of a different paradigm. Oncologists are very used to using information from a biopsy to direct therapy. We’re going to have to design good solid trials to show that imaging is predictive.”
In the meantime, Mankoff is anxious about using the technology with a variety of new probes that are based on the growing list of tumor targets. But he’s concerned that progress may be hampered by regulatory snags because the imaging compounds themselves are considered drugs. “What keeps me awake at night is that we already have on our horizon some of these probes and we need to figure out how to take the next step—how to get them out of animals and into patients.” For investigators like Mankoff, the hope for the future is now.
About the Author
Neil Canavan is a freelance journalist of science and medicine based in New York.
This article was published in Drug Discovery & Development magazine: Vol. 10, No. 3, March, 2007, pp. 22-24.
Filed Under: Drug Discovery