Journal of the National Cancer Institute Advance Access originally published online on November 11, 2008
JNCI Journal of the National Cancer Institute 2008 100(22):1566-1569; doi:10.1093/jnci/djn424
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© Oxford University Press 2008.
NEWS |
Superhighway or Blind Alley? The Cancer Genome Atlas Releases First Results
Three years ago, when the National Cancer Institute proposed its monumental $1.5 billion effort to catalog the genomic changes involved in cancer, the goal was to improve the ability to diagnose, treat, and prevent cancer. Now the first results of The Cancer Genome Atlas project (TCGA) have arrived, and they reinforce the emerging realization that the more researchers investigate the genetics of cancer, the farther away new therapeutic applications seem to get. In response, some researchers say that it's time to step back and look at cancer holistically, searching not for the individual genetic mutations but instead for unifying principles that underlie the disease process.
"This idea of digging deep into the genome gives you a lot of information, and it's an important tool, but it's really only one tool," said Lynn Hlatky, Ph.D., director of the Center of Cancer Systems Biology at Tufts University in Boston. "If your goal is making progress therapeutically, I think it's not the best investment to dig this deep until you have a larger overriding principle that really works for you. You need unifying principles that are predictive."
TCGA was intended to catalog all the genetic alterations involved in cancer and provide a roadmap for understanding how cancer develops and spreads. But the genetic complexity of cancer is making it more difficult to sort out which genetic changes actually cause disease. The NCI's $100 million pilot project—examining genetic changes in glioblastoma multiforme, the most common adult brain cancer—revealed a genetic train wreck of gross genomic alterations and mutations and signaling gone awry. But perhaps most surprisingly, the brain tumor study, which appeared in the journals Science and Nature in the first week of September, revealed new mutations associated with chemotherapeutic treatment for glioblastoma multiforme. The finding reveals that the array of mutations shifts with treatment, making genetic characterization a moving target.
Parallel studies still under way that are looking at genetic changes in ovarian and lung cancers, have not yet been published.
In the brain cancer project, the investigators studied 206 previously collected primary tumor samples, including 21 samples collected after treatment. Of these, the investigators selected 91 tumor samples for mutation analysis in 601 genes known to be important in cancer. They found 453 mutations in 223 genes, one-third of which had multiple mutations. Moreover, the tumors had an array of genetic differences among them, with most mutations occurring in only a few cases.
Lynda Chin, M.D., the co-principal investigator of TCGA's center at Harvard Medical School, acknowledges the challenges of evaluating so many data but says that newly developed computational tools are proving that finding potentially clinically relevant information is possible.
"Among the complexity of the genome, the question is: can we identify things that are of value, that are not just noise?" she said. "I think the answer is yes. I think the data now show, even with today's technology, we are able to detect these biologically important events, and they are already changing the way we think about cancer."
Chin, scientific director of the Belfer Cancer Genomics Center at Dana-Farber Cancer Institute, points to the unexpected revelation, described in TCGA research network's Sept. 4, 2008, Nature article, that tumors from patients with recurrent glioblastoma multiforme develop genetic resistance to temozolomide, a common chemotherapy treatment for treating the disease. Genomic analysis of recurrent glioblastoma multiforme revealed that patients had developed a secondary mutation in a key DNA mismatch repair gene that allowed the tumor cells to evade the killing action of temozolomide. That information, she said, will allow researchers to test whether calcium channel blockers, a class of compounds recently found to inhibit the growth cells with mismatch repair defects, could prevent the emergence of drug resistance.
The research also reveals a pattern of gene mutations in a network of biochemical pathways, some of which were known to be involved in glioblastoma development, but with new insight into mutations in genes, such as PIK3R1, that were not previously known to be important in the disease. Before this study, scientists had identified mutations in the catalytic domain, p110a, which help drive tumor growth. Now investigators are reporting new mutations that probably interfere with the regulatory region of PI3K, a popular target for drug companies that are developing new cancer treatments.
"There are at least 10 drug companies developing inhibitors to PI3 kinase, and they all want to know how to pick patients who will respond," said Chin. "They say, ‘Perhaps it should be patients with mutations in the catalytic domain?’ But now we know the cancer can activate the enzyme by mutating not only the catalytic domain, the p110a, but also the regulatory domain."
The finding adds another level of complexity to the effort to predict who will respond to PI3K inhibitors, she said. It suggests that looking at mutations in the catalytic domain are not enough. Some cancers appear to find other ways to activate this crucial growth pathway.
The complexity of the problem is discouraging to those investigators looking to find the next targeted therapy for cancer. Right now, "we need 20 different drugs against 20 different targets," said Chin. "We need to get to that point of having a collection of effective drugs and knowing when to use which combination based on a genetic profile of each patient. Obviously, we are not there yet."
Holistic Cancer Systems
For some researchers, simply cataloging the panoply of genetic alterations in cancer cells seems a dubious exercise. "Genetics has been oversold," said Garry Nolan, Ph.D., associate professor of molecular pharmacology at Stanford University School of Medicine. "And the backlash, I think, is what we need to prepare for. To the extent that people can be convinced that a holistic approach is what we need, the better off we'll be."
Nolan is not surprised by the genetic complexity seen in the TCGA results. In fact, it has been known for a while that even apparently homogeneous cancers are really more than one disease, he pointed out. What's needed is a way to connect mutation to function.
"It's not that the cancer genome atlas isn't important," said Nolan. "It's that it didn't have a built-in mechanism to look at function, to say now that we have this information, how are we going to figure out if a patient has four cancers or only one?"
Nolan's research, which has focused on cancers of the blood, such as lymphoma, indicates that tumor cells that look identical actually have many subtypes, only some of which may be susceptible to treatment. His approach has been to look for overarching themes and network-level unifying elements in cancer cells. "Our approach is to let the cell tell me what the answer is by assaying enough proteins that I know what their functions are in a network," he said. "Our goal is not to identify individual targets but to identify a network state as a desired goal."
Nolan sees the whole cell as a processing center that he can interrogate to identify its state and eventually to use that information to tell which drugs will work for a particular disease. As proof of principle, Nolan and his colleagues published an article in the Oct. 7, 2008, issue of Cancer Cell showing that it is possible to track the progression of juvenile myelomonocytic leukemia by using a flow cytometry technique that measures levels of key signaling protein networks throughout treatment.
Nolan said that he would love to be able to apply his techniques to the primary tumors used in TGCA and to determine how many subtypes are present in each tumor. "Until you can get to that level of detail, you are forever working with averages," he said.
Similarly, in a study of chromosomal alterations in breast and colon cancer, published in the Oct. 16, 2008, issue of Proceedings of the National Academy of Sciences, Victor Velculescu, M.D., Ph.D., Bert Vogelstein, M.D., and their colleagues at the Johns Hopkins Medical Institutions point out that examining genetic alteration in primary tumors, as was done in the TCGA project, will also be contaminated by normal cells within the tumor tissue. The DNA from normal tissue can mask subtle changes in gene copy number, they wrote in the discussion section.
|
In that study, the researchers chose to study primary tumors that had been cultured in the laboratory to eliminate contaminating normal tissue. Their aim was to identify chromosomal segments that are deleted or duplicated in tumor cells. Using this technique, they found 1,077 copy number changes, many of which would not have been found had normal tissue been present.
"It sounds really nice when you are saying ‘Let's study primary tumor sample directly from the patient,’ but in practice there are all sorts of problems with that," said Velculescu, who published a competing article on glioblastoma multiforme genetic analysis at the same time as the TCGA project group. "When you do sequence analyses, you can be confounded by the normal nontumor tissue and DNA that is present within these samples. And it becomes a major problem when it's time to do copy number analyses because the signal gets drowned out by the nontumor DNA that is there."
Targeting the Cancer State
Researchers such as Isaac Kohane, M.D., Ph.D., who contributed bioinformatics expertise to TCGA, believe that the next big shift in the understanding of cancer will come from approaching the disease from a developmental perspective. As director of the informatics program at Children's Hospital in Boston and codirector of the Harvard Medical School Center for Biomedical Informatics, he sees cancer as a "state," in which it is possible to identify meaningful alterations that correspond to developmental pathways.
He and his colleagues have shown in a series of reports that it is possible to correlate individual lung cancers in people to various developmental stages in normal lung tissue and predict how aggressive the tumor will be on the basis of its similarity to a particular developmental stage of normal lung development. The closer the tumor gene profile was to an early developmental stage, the worse the prognosis. Moreover, the gene products enriched in the aggressive tumors were those involved in cell adhesion, providing a potential target for therapeutic intervention.
"Development is a normal, healthful process," Kohane said. "So if we see that cancer looks like that process, this might give us an idea about these unifying pathways that we could then target."
Likewise, Hlatky and her colleagues are investigating how perturbations in the cancer cell's angiogenic lifeline, a unifying requirement for cancer growth, can be exploited as a target common to most solid tumors. In a July 31, 2007, report in Proceedings of the National Academy of Sciences, lead authors Amir Abdollahi, Ph.D., and Peter Huber, Ph.D., of the German Cancer Research Center in Heidelberg, along with Hlatky and her colleagues, described how gene expression shifts when pancreatic cancer cells change from a relatively benign state to an aggressive, invasive state. The group's strategy is to find compounds that could alter the balance toward a quiescent state, thus removing the tumor's ability to spread.
"The genes are simply the players carrying out the story," said Hlatky. "The key is that these tumors really need oxygen—that's a constraint. You either have to be right on a [blood] vessel, or you have to have something that brings these vessels to you, so that was a very good place to think about targeting. Even though there could be redundancies, you know what the tumor needs to do and if you cut that off, that's checkmate to that tumor."
The trouble with TCGA, she said, is that it is stuck at the level of the gene. It's not hypothesis driven and doesn’t address the complexity of the cancer in its environment.
"You could question putting this kind of investment into a single level," she said. "This level is important, but you have to have overriding principles that tell you the dynamics of the tumor and where it is going in order to stop it."
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||