March 11, 2002
Gene profiles expose cancer cells' weak points
Scott Armstrong, MD, PhD, Stanley Korsmeyer, MD, and Stephen Sallan, MD, study data.
In their deadly quest to grow and spread, cancer cells are the ultimate stealth weapon. They hide from the body's immune defenses, hijack new blood supplies, and often evade the most toxic drugs thrown against them.
Cancer cells' outward appearance can camouflage their true nature. In some types of cancer, one patient's tumor may look just like another's under the microscope, yet in one person the cancer proves curable, while in the other it is lethal. And no one can predict which is which — an uncertainty that bedevils patients and doctors alike and hampers effective treatment.
That's why there's a continual search for better methods of cancer diagnosis, which "stands at the crux of everything we do for cancer patients," says Sridhar Ramaswamy, MD, of DFCI's Adult Oncology Department. "Even now, in the genetic age, this remains a problem because current diagnostic methods don't detect all the complexities that determine a cancer cell's behavior."
Sometimes conventional methods can't even determine which tissue or organ gave rise to the cancer, further hampering diagnosis. But now a revolutionary system of cancer diagnosis and classification is taking shape, based not on cells' appearance or where in the body they originated, but on their genes. Using new technology and information from the human genome project, scientists have begun to sort out cells by differences in the activity of their genes. The new cancer-identifying technique records a cell's gene expression profile, or "signature," by determining which of its thousands of genes are active, and which are silent.
Just in the past year, a crop of reports by scientists at Dana-Farber and collaborators at the Whitehead Institute for Biomedical Research in Cambridge, Mass., has demonstrated how gene profiling is already yielding important new insights into a variety of cancers. Any cell's behavior is dictated by the specific off-and-on pattern of its genes, just as a chord on the piano is determined by which keys are sounded and which are silent. Genes that are "on," or expressed, are helping to make proteins that operate the cell, and genes that are "off" are not manufacturing proteins. A human cell contains at least 30,000 genes, scientists believe, each able to make one or more proteins.
Cancer, according to researchers, is fundamentally a disease of genes and the proteins they make. Too much of a certain protein, caused by a gene stuck in the "on" position like a malfunctioning switch, can prod a cell into runaway growth. Conversely, a normally growth-suppressing gene that is stuck in the "off" position can allow cells to grow chaotically and form a tumor.
Other abnormal genes help a tumor obtain a new blood supply in the body, break through tissues and spread to other organs. Therefore, the gene profile of a cancer cell compared with that of a normal cell may reveal genes that are altered or abnormal in the cancer cell — and are at the root of the malignancy.
Usually it's a large number of subtle gene changes that add up to cancer. But, on occasion, the profile spots a single bad gene acting dramatically to turn the cell cancerous. Identifying such genes in different cancers, it is hoped, will lead to new drugs specifically designed to target those genes.
These advances mark the beginning of a revolutionary new system of cancer classification. "The dream is that when someone comes into the hospital," says Ramaswamy, "instead of people stroking their chins, having conferences, and sending the slides to other experts, the diagnosis will be made on the genetic profile of the cancer rather than solely by microscopic appearance."
He headed a team that last year reported it had obtained the gene profiles of 14 types of cancer cells taken from tumors. They said they could tell these 14 varieties apart with 90 percent success using the gene profiles.
Robert Weinberg, PhD, of the Whitehead Institute, a pioneering cancer researcher who made landmark discoveries of how normal genes become cancer-causing "oncogenes," is enthusiastic about the new findings flowing from gene-profiling laboratories.
"These studies hold the promise of being able to classify tumors in ways that are much more useful than the tumor markers used previously," Weinberg explains. "They will tell us whether tumors that have traditionally been lumped together as one type should be divided into a number of distinct subtypes, and whether these subtypes have different prognoses and respond differently to specific therapies."
Spin-off of the human genome project
Two developments have fueled the gene-profiling momentum of the past year or two. First, the decade-long human genome project has virtually finished deciphering the genetic code of all the human cells — including many abnormalities implicated in cancer. Second, the invention of gene "micro-arrays" or "gene chips" makes it possible to quickly identify thousands of genes in a cell and determine their degree of activity.
Colors representing gene activity (red is high, blue is low)
show that MLL has a distinct pattern from both ALL and AML. Each
vertical column of squares represents a tumor sample. Each
horizontal row of squares represents the activity of one gene.
Image courtesy of
Nature
Genetics and S.A. Armstrong et al. For a high resolution
image, click
here.
The recent series of publications by Dana-Farber scientists all involve the work of Todd Golub, MD, a pioneer in cancer gene expression profiling at DFCI and the Whitehead Institute. Golub was the first scientist, in 1999, to show that the method could distinguish two types of cancer that closely resemble each other (see Winter/Spring 2000 Paths of Progress). The flurry of subsequent papers "is gratifying," he says, "because all of the work coming out now is pushing the envelope a little bit further compared to our first paper."
One of the most talked-about new reports, whose lead author is Scott Armstrong, MD, PhD, of Pediatric Oncology, describes how gene chip studies of cancer cells enabled investigators to identify a unique form of leukemia. The scientists have even given it a name, "mixed lineage leukemia," or MLL. Previously, this was thought to be just a baffling, particularly hard-to-treat subtype of acute lymphoblastic leukemia, or ALL.
The study highlighted a particular gene stuck in the "on" position that seems to be key to the disease. Fortunately, pharmaceutical researchers had previously come up with a drug that inhibits the culprit gene (known as flt-3), and the Dana-Farber contingent says this drug kills MLL cells in the laboratory. Eventually, they hope to test it against MLL in infants, who often die of the rare cancer.
In January, a team led by Margaret Shipp, MD, of Adult Oncology, reported in Nature Genetics that it used gene profiling to distinguish between identical-looking samples of a lymphoma that is unpredictable: it can be cured in about 40 percent of cases, while it is lethal in 60 percent. It's a common form of lymphoma called diffuse large B-cell lymphoma, or DLBCL. "We took tumors from 58 patients that looked exactly the same under the microscope," she says, "and predicted on the basis of molecular profiles which could be cured with standard therapy and which could not." The prediction was strikingly accurate.
Again, as in the MLL discovery, scrutiny of gene patterns in DLBCL identified a misbehaving gene and an errant protein in the more dangerous type of the disease. Shipp views the wayward protein as a potentially valuable new target for better drugs than are now available. She learned that one such drug had already been developed for another disease, and the researchers hope to begin a clinical trial this year.
Matthew Meyerson, MD, PhD, and colleague Arindam Bhattachorjee, PhD
Another paper, from the lab of Matthew Meyerson, MD, PhD, of Adult Oncology, used gene profiles to distinguish several different types of lung cancer. In most cases, the technique sorted the cancers into the same categories used in traditional diagnosis. But, in one case, the genetic snapshots broke new ground: they identified a type of lung cancer that had a worse prognosis for survival than a similar-appearing cancer. This information could be useful in designing treatment plans.
Gaining an edge on the toughest cancers
Gene profiling may also be headed for a role in differentiating some brain tumors in children. John Y.H. Kim, MD, PhD, of Pediatric Oncology is an author of a paper recently published in the journal Nature reporting that genetic snapshots helped specialists predict which tumors of a type called medulloblastoma would be the most life-threatening.
"These tumors were poorly distinguished in the past, but now we've shown they have very distinct molecular profiles that let us predict [the tumors' effects] more accurately," says Scott Pomeroy, MD, of Children's Hospital Boston, who led the large research team from several institutions. It's an important step, he says; with precise prediction, doctors can reserve the most potent treatment — which has the worst side effects — for children with the most aggressive tumors.
In prostate cancer, physicians are frustrated by an inability to identify which men who've had their prostate glands removed are more likely to have a recurrence of the disease. Dana-Farber scientists William Sellers, MD, and Phillip Febbo, MD, both of Adult Oncology, are working with Golub to apply gene profiling to this problem, hypothesizing that more-aggressive tumors might have genetic signatures differing from less-aggressive ones.
Golub, the gene-profiling pioneer, cautions that this process isn't yet being used clinically: more experiments in different kinds of cancers are necessary to validate its reliability. And gene profiling will probably be used in combination with conventional methods, he adds. Still, the excitement is palpable.
Chief of Staff Stephen E. Sallan, MD, believes the promise of gene profiling in designing new drugs will be fulfilled before long. "It is clear," he comments, "that this systematic and quantitative approach to identifying genetic targets for novel therapies will soon become the 'gold standard' in cancer research."
(Paths of Progress, Winter/Spring 2002)
