Image: ASH ALIZADEH et al.

DNA MICROARRAY ANALYSIS reveals patterns of gene expression as patterns of fluorescent dots. In the data above, the researchers identified two clear patterns, marked by orange and blue bands, indicating that there are two distinct forms of diffuse large-cell B-cell lymphoma (DLBCL): GC B-like DLBCL (orange) and Activated B-like DLBCL (blue).

Scientists working on the Human Genome Project have catalogued countless gene sequences over the past decade; far fewer researchers have focused on what individual genes actually do and how they contribute to disease.

But one such scientist, Patrick O. Brown of Stanford University, has recently met with great success--at least as far as cancer patients are concerned. Using a technique to analyze gene expression, Brown and his colleagues have found that the most common form of non-Hodgkin's lymphoma--diffuse large-cell B-cell lymphoma (DLBCL)--is in fact two distinct diseases.

The results, which appeared in the February 3 issue of Nature, should improve diagnostic tests and make possible the development of more specific treatments for these lymphomas, as well as for other cancers. "This work shows that the molecular portrait of a tumor that we get from DNA microarray analysis can actually be interpreted as a much clearer, more detailed picture of the tumor's biology," Brown says, "and that the new things we see in this picture can really make a difference for the patient."

Brown devised his method of DNA microarray analysis five years ago, using it then to study levels of gene expression in the plant Arabidopsis thaliana. The basic idea is simple: a special robot imprints dotted arrays of some 20,000 microscopic DNA samples--corresponding to particular genes--on a glass slide. Scientists then wash the array with fluorescently labeled messenger RNA from the cells in question. The result: brighter dots show up where genes are expressed at higher levels.

Image: ASH ALIZADEH et al.

SUBCLASSES of DLBCL express different sets of genes involved in B-cell development--and follow very different courses.

The findings on A. thaliana matched those obtained by way of Northern blot testing, the traditional and much slower approach--and so Brown and his colleagues quickly moved on to bigger tasks. In 1998, they used the same approach to identify 500-odd genes activated when yeast forms spores. It was a much greater number than anyone expected. And whereas scientists had assumed that sporulation took place during four stages, Brown found that it actually involved seven.

Work on the lymphochip started soon thereafter. First, the group designed the special microarray, which contained nearly 18,000 genes preferentially expressed in lymphoma, as well as DNA known or suspected to play a role in normal immune system functions. In DLBCL, B lymphocytes--important components of the body's defense system--proliferate uncontrollably and rapidly cripple a patient's immunity. The entire chip was the size of two postage stamps side by side.

Next, the scientists doused the lymphochip with gene transcripts from both normal and cancerous B cells and discovered many different patterns of gene expression. Among the tumors, however, the variation most often occurred within a large group of genes responsible for directing a particular stage of B cell development.

"Statistical analysis of the data revealed the presence of groups of tumors with similar gene expressions," said co-author Michael Eisen, formerly of Stanford and now at Lawrence Berkeley National Laboratory. "We noticed that one of these groups of tumors expressed a large group of genes normally expressed at a specific stage of B cell development, while the remaining tumors expressed a set of genes characteristic of a later developmental stage."

That two distinct tumor profiles emerged was, in fact, of little surprise. Clinicians have long recognized at least two major courses DLBCL can take. Brown and his colleagues went further to match which tumor group was behind which prognosis: they found that patients with tumors expressing genes involved in later stages of B-cell development in general fared much worse after standard chemotherapy.

There may well be other subtypes of DLBCL, Brown points out. Indeed, only 40 percent of the 25,000 cases that occur each year respond well to treatment. But all subtypes--of all cancers, perhaps--stand to benefit from further gene expression analysis. "The differences we are seeing are at the level of the molecules and regulatory systems that are the targets of present and future anticancer drugs," Brown says. And the more clearly scientists can see this bull's-eye, the better they can take aim.