he paper at #3 this month has all the hallmarks of a real turning point. A new technology has proved its worth in the diagnosis of human disease, and offers real hope of a breakthrough in therapy. A team under the joint leadership of Howard Hughes Medical Institute researcher Patrick O. Brown of Stanford University and Louis M. Staudt of the National Cancer Institute in Bethesda used a DNA chip to distinguish two kinds of cancer that had previously been treated as a single entity.

The cancer in question is diffuse large B-cell lymphoma (DLBCL), the most common form of non-Hodgkin's lymphoma, which affects the B cells of the immune system. Clinically, the symptoms can vary quite widely, as can the outcome. Broadly speaking, about 40% of patients respond well to conventional therapy and survive well. But for the rest the disease progresses rapidly, with much poorer survival after treatment. Unfortunately for physicians, and even more so for people with DLBCL, there has been no way to predict into which class any particular patient will fall. Until now.

Brown originally developed the DNA chip in 1995, to explore Arabidopsis, the model organism of choice for plant biologists. A robot places several thousand tiny dots of probe DNA, each representing a gene, on a glass plate. The DNA array is then exposed to an extract of copies of all the DNA that is active in a particular cell at a particular time. Sequences from the cell that have a counterpart on the chip find it and remain stuck to the probe after washing. Each sequence from the cell is marked with a fluorescent dye, which is measured by a scanning laser microscope and converted into a digital image of how active each clone on the chip is in the target cell. Researchers can thus very quickly gather huge amounts of data about which genes are active in which cells.

Staudt's group has long been interested in lymphomas, and worked with Brown to develop a specialized "Lymphochip." Each is about the size of two small postage stamps side by side and contains a careful selection of about 18,000 DNA probes derived from diverse cells of the immune system. Using DNA from patients with DLBCL and other kinds of lymphoma the team rapidly amassed almost 1.8 million data points.

Smart software converts the chip images into a pattern of activity that is effectively a DNA profile of the target cell. Even smarter software then sorts the DNA profiles to create another kind of grid, in which each row represents one gene on the DNA chip, and neighboring rows share equal levels of activity, while each column is the gene profile of a particular cell. The exciting part is that when the results are arranged like this, bands of color jump out of the picture. Those bands are "signatures" of gene expression.

Because the Lymphochip includes genes of known function it is possible to label the signatures. Thus the normal B-cell signature contains genes active in the differentiation of B cells and genes that code for components of B-cell receptors. In the search for differences in gene signatures among B-cell lymphomas, the team looked specifically at samples from a variety of purified B and T cells at various phases of differentiation and activation. A signature emerged called the "germinal center B cell signature," consisting of hundreds of genes that are expressed only in B cells undergoing the transition to activation. These genes were not expressed in either resting B cells or activated B cells or T cells.

Overall, the average five-year survival of the patients was 52%, but whereas 76% of those with germinal center B-cell-type tumors were still alive after five years, only 16% of those with activated B-cell-type tumors survived that long.

Of course there could be further sub-categories too. And there is remaining clinical heterogeneity. Some of the "better" group died quickly, and some of the "worse" group survived well. Staudt tells Science Watch that although there are as yet no proven better therapies for "poor-prognosis" DLBCL, the group is profiling many different lymphomas and hopes that this will suggest new therapeutic targets. So there is still plenty to be done. But the DNA chip as a diagnostic tool has without doubt arrived.