GNN - Genome News Network  
  Home | About | Topics
What Gene Chips Can Do
Can doctors use DNA microarrays to diagnose leukemias in children?
Edward R. Winstead

Featured article.

One of medicine's success stories is the treatment of children with leukemia. Nearly 80 percent of children with acute lymphoblastic leukemia, or ALL, respond to treatment and remain free of the cancer for the rest of their lives. If doctors know which subtype of ALL they are dealing with, they can design the most appropriate therapy based on their experience and diagnosis.

Bone marrow aspirates from pediatric patients with acute lymphoblastic leukemia. View larger

The method of diagnosing ALL subtypes, however, is expensive and requires the expertise of more specialists than are found at most hospitals. Doctors have long hoped for a shortcut that could provide reliable and accurate diagnosis. Now, a new study suggests that they will soon be able to obtain a reliable diagnosis by analyzing the activity of genes in a patient's cells.

The researchers used gene chips to analyze cells from 360 ALL patients who were treated at the same hospital in Tennessee and had extensive medical histories. Based on an analysis of gene expression patterns, the researchers correctly identified 96 percent of the subtypes. The results confirm other recent reports that gene chips can detect molecular differences that characterize clinically distinct forms of childhood leukemia.

"We knew there would be differences because we know that there are distinct subtypes of leukemia," says James R. Downing, who led the study at St. Jude Children's Research Hospital in Memphis. "The surprise was how different the gene expression profiles were."

The researchers identified clear differences between the gene expression profiles of patients who relapsed and those who remained in continuous complete remission. "Although these data will need to be validated in prospective studies, these findings raise the expectation that, in the future, this type of analysis will be used to make therapeutic decisions," the researchers write in Cancer Cell.

At St. Jude, four different laboratories and as many as ten specialists are involved in making an ALL diagnosis. Downing and his colleagues undertook the study a year ago to see if gene expression profiling could replace the other diagnostics and provide biological insights about the disease.

Cluster diagram of gene expression profiles of 29 pediatric ALL samples.

The researchers developed new algorithms to analyze the gene expression data and classify patients depending on their risk for failing therapy. The number of genes required for classifying patients varied among the subtypes. A single gene was sufficient to achieve 100 percent accuracy for two ALL subtypes.

The current method of predicting how a patient will respond to treatment does not identify many patients who are likely to become toxic in response to drugs. The hope is that gene expression profiling will enhance the ability to identify patients who are most at risk for developing complications during therapy, such as secondary malignancies and infections.

The most intense chemotherapy, which may cause complications or even a secondary disease, can be reserved for individuals who are likely to resist treatment.

"Within the genetic subtypes we were able to identify through gene expression profiling those patients who were going to fail therapy," says Downing. "This was a retrospective study using extensive medical histories of patients who were all treated at St. Jude Children's Research Hospital. We know whether a particular patient is cured or not."

Larger studies are needed to correlate expression patterns with treatment outcomes. But the preliminary data are encouraging, says Ching-Hon Pui, a clinician at St. Jude who also participated in the study. "It is very exciting to think that you may be able to predict very precisely whether a patient will develop toxicity from the treatment," he says.

"This is just the beginning," says Pui. "We're studying a fraction of the genes in the human genome. When higher density chips become available these sorts of studies will become even more informative."

The chips in the study contained 12,600 genes—perhaps a third of the human genome. It is likely that a custom 'ALL' gene chip will be developed in the next few years. An ALL chip would include genes whose expression patterns characterize the disease subtypes and provide information about the pathways involved in ALL.

The researchers mined the raw expression data for insights into the biology of ALL. As with other expression profiling studies, they identified genes that may be important to the disease. The raw data have been posted on a Web site and are available to researchers around the world.

"The Web site has had a thousand hits in the last two weeks, and these researchers are sure to develop insights we did not," says Downing. His group undertook the study with the goal of finishing in rapid fashion so the data would be a resource for researchers working on leukemia and other cancers.

Multi-dimensional scaling plot showing the distinct clustering of leukemia subgroups in gene space. View larger

The new results confirm and expand other recent studies of ALL. In December 2001, researchers used gene chips to show that some leukemia patients have a characteristic gene expression profile and constitute a new ALL subtype. Stanley J. Korsmeyer, of the Dana-Farber Cancer Institute in Boston, and colleagues called the new subtype mixed-lineage leukemia or MLL. The study appeared in Nature Genetics.

The study by Korsmeyer and colleagues was confirmed by the St. Jude team using a large sample of patients. In addition, they generalized the finding for all the known ALL subtypes, showing that each has characteristic gene expression patterns. Like the Boston group, they also identified another potential new subtype. About 25 percent of ALL cases have not been classified, and new subtypes are likely to be reported as technology improves, according to Pui.

"It would be a tremendous advance for the field if we could do away with the current diagnostic procedure and use a single platform to diagnose ALL subtypes," says Downing. "It appears that we can, and the studies to confirm our results will now go forward."

To view the study's supplemental data visit

See related GNN articles
»Making predictions about cancer outcomes
»New type of leukemia identified

. . .

Yeoh, E.-J. et al. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell 1, 133-143 (March 2002).
Armstrong, S.A. et al. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 30, 41-47 (January 2002).

Back to GNN Home Page