The international team of researchers, led by Dr. Jeffrey Trent, NHGRI's scientific director and head of the NHGRI Cancer Genetics laboratory, used a new technique called gene-expression profiling to differentiate between breast tumors that were caused by inherited genetic changes and those that were not. Based on a laboratory method that uses a type of DNA chip called a microarray, the researchers simultaneously assessed how active some 6,000 genes within breast cancer cells were, altogether making nearly 250,000 measurements. The study revealed clear differences in the patterns of gene activity in breast tumors, patterns that can be as unique as a fingerprint, pinpointing into which group a woman's cancer belongs.
"This powerful new technology gives us a snapshot of exactly which genes are active in a tumor cell," Trent says. "Over the last few decades, scientists have made important progress in understanding the molecular origins of cancer by studying one gene at a time. Now we can look at thousands, and even tens of thousands of genes as they interact to produce a tumor. This capability will have important implications for both diagnosis and treatment."
Genetics of Breast Cancer
Earlier insights tended to give a fairly static view of the genetic changes that produced a tumor. The current study demonstrates the power of a dynamic analysis that shows the interaction of many genes within a cell's metabolic pathways. In what may become a new research and diagnostic trend, the team has established that it can differentiate between hereditary and non-hereditary breast cancer by studying thousands of genes simultaneously.
Physicians have long known that breast cancer can run in families. Approximately 5 to 10 percent of breast tumors are hereditary; the remaining cases are caused by genetic changes that occur during a woman's life and are commonly called sporadic. In the mid 1990s, scientists identified mutations in genes now called BRCA1 and BRCA2 that are the major cause of the hereditary form of the disease. Women inheriting these mutations have a 40 to 85 percent lifetime risk of developing breast cancer, as well as an increased risk of ovarian cancer.
Telling the tumor types apart is not easy with traditional techniques. "When you look at these tumors under a microscope, based on their shapes and other features, it is very difficult to tell which tumors are caused by BRCA1, and virtually impossible to distinguish cancer caused by BRCA2 from those caused by non-inherited mutations," says another of the study's senior authors, Dr. Ake Borg of the University of Lund. Furthermore, the BRCA1 and BRCA2 genes are very large, and searching through them for mutations is both complicated and expensive.
To determine whether gene profiling could tell the difference between sporadic tumors and those with BRCA1 or BRCA2 mutations, the research team examined samples of tumors that had been surgically removed from 22 breast cancer patients at the University of Lund in Lund, Sweden. Fifteen of the women were known to have hereditary breast cancer based on family history and other analyses. Of the 15, seven had mutations in BRCA1; eight had mutations in BRCA2. The control group was comprised of eight women with sporadic cases that had no evidence of any family history of the disease.
When the team examined the gene-expression profiles for the 22 patients using the microarray analyses, they were able to quickly and accurately differentiate the BRCA1 from the BRCA2 inherited changes as well as the non-inherited genetic changes.
"We were surprised that we could separate the groups as well as we could," says Ingrid Hedenfalk, one of the two lead authors of the study and a doctoral student visiting NHGRI from the University of Lund. Placing patients in the proper diagnostic category can be critical to a successful treatment. For example, women shown to be carrying BRCA1 or BRCA2 mutations are at high risk for second breast cancers, and for ovarian cancer, and thus need very close follow-up.
The team's approach relied on a cutting edge technology called a DNA microarray, or sometimes, a gene chip. The microarray itself is merely a glass slide with tiny dots of DNA from different genes arranged in a grid-like array. Using microarrays, the activity of thousands of known genes can be quickly tested in a cell sample.
To perform the analysis, scientists isolate a special collection of molecules from the test cells called messenger RNA or mRNA. These molecules are produced by active genes and indicate which of the estimated 30,000 genes are "turned on" in a cell. The information in the messenger RNAs can be converted into a form of DNA called complementary DNA or cDNA kind of like copying a music CD to digital audiotape; same information, just in a different form. A fluorescent label can be added to the cDNA made from the test cells. A robot then systematically deposits a sample of the cDNA onto the glass slide. If the sample contains cDNA that matches a particular gene on the microarray, the cDNA sticks to that spot on the array, like two pieces of Velcro coming together. Unmatched cDNA is washed away.
Automated computerized detectors then measure the amount of fluorescence for each spot. The brighter the fluorescence, the more cDNA has attached to the known gene, and the more active the gene must be in that tissue sample. The result appears as a pattern of spots that are either bright or dark, showing whether the known gene is active or inactive in the tumor cell.
"Our gene expression profiling technology reveals a pattern, a kind of fingerprint for each tumor type," Dr. Trent says. "The fingerprint shows us key genes involved in tumor development and progression."
Because of the number of gene samples, and the number of patients tested, the study produced almost a quarter million data points. To handle the huge amounts of information generated by this process, the researchers turned to computer-based statistical analyses. With the help of statisticians from the NIH's National Cancer Institute, Agilent Laboratories, and Texas A & M University, the researchers were able to show that the differences between the groups of genes from the cancer cells were, indeed, statistically significant.
In a surprising twist, the scientists found their gene-profiling technique to be so good that it found a previously unrecognized rare event among the classically diagnosed tumors. Among the control tumors that had been originally diagnosed as non-hereditary, one tumor showed a gene-profile that made it look like a hereditary form of the disease caused by mutations in the BRCA1 gene.
After careful ethical advice was obtained, the researchers decided to re-contact the patient to ask for permission to do additional testing. The researchers then sequenced her BRCA1 gene and found no mutations. This puzzling result lead the team to test for a newly discovered mechanism that appears to contribute to the cause of some forms of cancer, even when a critical gene is not mutated. Instead, the gene is somehow abnormally turned off. Scientists call this recently discovered phenomenon gene silencing.
"We discovered that even though the spelling of her BRCA1 gene was normal, it was silenced due to a non-inherited mechanism called methylation," Dr. Trent says. Methylation previously has been identified as a way that a cell may temporarily or reversibly silence the activity of a gene. It remains unclear just what caused the methylation to silence this patient's BRCA1 gene and essentially turn it off. Nevertheless, Dr. Trent says, several recent studies by researchers elsewhere have shown that gene methylation may be involved in a number of different cancers, including breast and colon cancer, and may be a more common mechanism for tumor formation than previously thought.
Although the research team initially tested thousands of genes to find the pattern associated with the hereditary and nonhereditary forms of breast cancer, computational techniques helped them reduce the number of genes needed to a much more manageable number. "In fact," says co-lead author Hedenfalk, "we were able to narrow the number of genes needed to separate the three groups to just 50 or so."
"Not unexpectedly, these critical genes turn out to be involved in various aspects of cancer progression," says Dr. David Duggan, the study's other co-lead author. These aspects include mechanisms controlling the repair of damaged DNA, cell division, cell death, and other important cellular housekeeping functions.
These 50 or 60 genes are obvious candidates for further research to understand the molecular basis of cancer and are potential targets for the development of new drug treatments. "If you were trying to develop a therapeutic that selected hereditary breast cancer," Dr. Trent says, "these genes would be a good place to start."
The study also sheds new light on how breast cancer develops. "Even though BRCA1 and BRCA2 look nothing like each other in terms of their DNA sequence, they can cause the same net effect increased susceptibility to both breast and ovarian cancer," Dr. Borg says. By looking at the kinds of cellular genes that are active in these cancer cells, "our results indicate that these two types of inherited genetic changes use quite different pathways, or series of biochemical steps, to cause the same devastation."
Results from the study may also help researchers find additional genes involved in breast cancer. In August 2000, an NHGRI-led team reported finding evidence for a third hereditary breast cancer gene in certain Nordic families. By using gene expression profiling to study these families, the researchers hope to narrow their search for this, and possibly, other genes involved in breast cancer.
Already, the research is being hailed by leading cancer specialists as a breakthrough in basic science with potentially broad clinical applications.
"This very important research by a superb group begins to better clarify the different functions of these hereditary breast cancer genes BRCA1 and BRCA2," said Dr. Dan Von Hoff, director of the Arizona Cancer Center in Tucson, Arizona. "This pioneering piece of microarray work will enable us to better design prevention and therapy strategies for patients with hereditary breast cancer and, I suspect, for many other cancers as their genetic basis becomes elucidated."
For NHGRI Director Dr. Francis Collins, the discovery provides a glimpse of the future of genome research. "This work is an excellent example of the kind of research that will characterize the next phase of the Human Genome Project, as scientists move from sequencing the entire human genetic code to understanding the functions of genes in health and disease," he says.
Members of the research team include (in order of the study's list of authors) NHGRI's Ingrid Hedenfalk, David Duggan, and Yidong Chen; Michael Radmacher of the National Cancer Institute (NCI); Michael Bittner of NHGRI; Richard Simon of NCI; Paul Meltzer of NHGRI; Barry Gusterson of the University of Glasgow, United Kingdom; Manel Esteller of the Johns Hopkins Oncology Center; Olli-P. Kallioniemi and Benjamin Wilfond of NHGRI; Ake Borg of the University of Lund, Sweden; and Jeffrey Trent, of NHGRI.
Other authors include Mark Raffeld of NCI; Zohar Yakhini and Amir Ben-Dor, both of Agilent Laboratories; Edward Dougherty of Texas A & M University; Juha Kononen of NHGRI; Lukas Bubendorf of NHGRI and the University of Basel, Switzerland; Wilfrid Fehrle and Stefania Pittaluga, both of NCI; Sofia Gruvberger, Niklas Loman, Oskar Johannsson, and Hakan Olsson, all of the University of Lund; and Guido Sauter of the University of Basel.
Their report, "Gene-Expression Profiles in Hereditary Breast Cancer," Hedenfalk,
et. al.,appears in the Feb. 22, 2001, issue of the New England Journal of Medicine, 244:539-548.