College of American Pathologists
Printable Version

  Feature Story


cap today

Genes pointing the way in lymphoma prognosis

T&S completion compared against T&S collection relative to start of surgery

September 2002
Vida Foubister

The diagnosis of diffuse large-B-cell lymphoma, the most common type of lymphoma in adults, might soon become more informative for patients and clinicians.

Researchers have found gene expression patterns in these tumors that can be used to identify different subsets of the disease and predict patient outcome.

Until recently, physicians have relied on the clinically based international prognostic index to predict which patients with diffuse large-B-cell lymphoma will be among the 40 percent or so who survive long term.

"This large subgroup is thought to contain several variants that one hadn't been able to subclassify consistently by histopathology or immunological phenotyping," says H. Konrad Müller-Hermelink, MD, head of the Department of Pathology, University of Würzburg, Germany.

But the work of Dr. Müller-Hermelink and his colleagues, published in the June 20 issue of the New England Journal of Medicine, has the potential to change that. "By analyzing gene expression data, three different types of lymphoma can be unequivocally defined," he says. "This is not a subjective interpretation of any pathologist, but rather a definite biological difference."

Furthermore, "it was found that a very limited number of genes could give specific information on the prognosis of patients," Dr. Müller-Hermelink says.

Gene-expression subgroups

Building on an earlier study of 40 patients, which identified two subgroups of diffuse large-B-cell lymphoma, the authors used DNA microarrays and hierarchical clustering to analyze the expression of more than 7,000 genes in 240 patients.

"We found the two subgroups that we had previously identified as well as a new subgroup that was only evident with this larger number of tumor biopsies," says Louis Staudt, MD, PhD, section chief, Metabolism Branch, Center for Cancer Research, National Cancer Institute, Bethesda, Md.

The three gene-expression subgroups include the germinal-center B-cell-like lymphoma, which expresses high levels of genes characteristic of germinal-center B-cell-like lymphoma and normal germinal-center B cells; activated B-cell-like lymphoma, which expresses genes characteristic of mitogenically activated blood B cells; and the new subgroup, type 3 diffuse large-B-cell lymphoma, which has a heterogeneous gene expression that suggests it includes more than one subtype of lymphoma.

"We believe that the large subgroups we discovered will most likely have a bearing on many of the therapies [that are used] because they are like different diseases," says Dr. Staudt. "They are as different from one another as breast cancer is from lung cancer. They differ by the expression of thousands of genes."

Germinal-center B-cell-like large-cell lymphoma had the best prognosis, with a 60 percent five-year patient survival rate after anthracycline-based chemotherapy. It exhibited two oncogenic events not seen in the other subgroups: t(14;18) translocation of the bcl-2 gene and amplification of the c-rel locus on chromosome 2p. Patients in the type 3 subgroup had a survival rate of 39 percent; those in the activated B-cell-like subgroup had the poorest prognosis, with a survival rate of 35 percent.

These outcomes are partly due to the differences in the biology of the tumors, says study coauthor Wing Chan, MD, professor of pathology, University of Nebraska Medical Center, Omaha. "But grouping patients into these three subgroups does not tell you everything. There are still some patients in each of the subgroups that survive very well and some that have poor survival."

Gene expression signatures

In their most recent study, the authors also identified individual genes whose expression helps to determine survival. "The predictors we found are useful for each one of the patients, regardless of which subgroup they are in," says Dr. Chan.

A preliminary analysis of 7,399 DNA microarray features identified 670 that were significantly associated with a good or bad outcome. Using hierarchical clustering, the genes associated with outcome were combined into four gene-expression signatures. Because genes within the same signature are known to perform similar functions, 16 genes that were highly variable in expression were selected for use in an outcome predictor model.

"The important point is that we grouped the genes into functional categories that identified different biological features of the tumors that influenced outcome," Dr. Staudt says.

The No. 1 category was proliferation. "Genes expressed in highly proliferating cells identified patients who responded poorly to chemotherapy," he adds.

As with the three original subgroups, the cell of tumor origin was found to predict survival. Genes characteristic of normal germinal-center B cells correlated with favorable patient outcomes.

The last two categories of predictive genes suggest that immune response to the tumor is important for a curative outcome after chemotherapy. Genes that reflected the infiltration of the lymph node with host immune cells and a stroma cell reaction to the tumor predicted a favorable patient outcome.

A second immune response-related gene-expression signature reflects the major-histocompatibility-complex class II proteins, which stimulate immune cells by presenting them with antigens. Patients who lost expression of these genes had a poor outcome. "Our hypothesis is that those tumors are becoming invisible to the immune system," Dr. Staudt says.

In an evaluation of individual genes associated with outcome that were not included in these four signatures, one, BMP6, was found to increase the predictive power of the model. Its expression, which was included in the final outcome predictor, was associated with a poor outcome.

By adding together the expression of these 17 genes, each patient was assigned an outcome predictor score that ranged from -1.7 to 2.4. Each unit increase in score increased the relative risk of death by a factor of 2.7.

Patients were ranked based on their score and divided into one of four quartiles that were found to have five-year survival rates of 73, 71, 34, and 15 percent, respectively.

This work found that the expression of a limited number of genes could be combined into a predictive model that gives specific information about the prognosis of every patient.

"The clinical utility would be, on the one hand, identifying half the patients who have a very favorable chance of being cured by conventional chemotherapy with a five-year survival rate of between 71 and 73 percent," says Dr. Staudt. "On the other hand, fully one quarter of patients were in a very poor risk group that had only a 15 percent chance of being cured by current therapy."

Based on this model, it might be possible to identify those patients who are the best candidates for alternative treatments, such as bone marrow transplantation or therapeutic clinical trials. "This is a method that works," says Dr. Staudt. "Our gene-expression predictor reflects aspects of the biology of the tumor that we can understand and perhaps target therapeutically."

International prognostic index

The international prognostic index, which commonly has been used to predict outcome, measures such parameters as patient age, stage of disease upon diagnosis, total tumor bulk as measured by serum lactate dehydrogenase, performance status of patient upon diagnosis, and whether the tumor has spread beyond the lymph nodes.

The gene-expression-based method and the international prognostic index were found to be independent predictors of prognosis. "You can combine both of the predictors and do an even better job of predicting the outcome of the individual patient than using either alone," Dr. Staudt says.

Though the gene-expression outcome predictor is not yet available commercially, it would be easy to translate it into a clinical test, he adds. "If this could be applied, it would be like any other blood test you would get in the hospital." A patient would be given a score, for example 2.7, that, based on a database of patients with this type of lymphoma, predicts a 70 percent chance of being cured.

Several platforms could be used to measure these individual gene-expression parameters, including a reverse transcriptase/polymerase chain reaction assay or DNA microarray assay. Another option would be to look at the proteins expressed by these predictor genes using an immunochemical assay or monoclonal antibody chip.

"What is so exciting about this gene array data is that by gene expression you have a very objective possibility to define and classify the type of tumor not only by its histological features but also by objective biological criteria," Dr. Müller-Hermelink says. "In addition, these prognostic genes and signatures have been defined so that in the future it appears to be possible to state an individual prognostic index for every patient that has such a tumor. This will have implications for treatment strategies and treatment of these tumors."

In an editorial accompanying the New England Journal study, Ian Magrath, DSc, president of the International Network for Cancer Treatment and Research, Brussels, Belgium, states that "diagnostic precision is an essential foundation for effective therapy, since individual diseases might be expected, in general, to require individual therapeutic approaches."

Yet he also notes that DNA microarray analysis cannot identify all of the factors that determine treatment outcome. For example, the final concentration of active drug in tumor cells is also influenced by genes expressed by normal tissues, circulating enzymes, and anatomical or physiological factors. Microarrays are also not likely to be useful in identifying variations in the activity of drug metabolism or DNA repair enzymes or factors that influence tumor volume.

Future work

The goal of this work, however, isn't just to improve clinicians' ability to predict patient outcome.

"Finding a predictor is just the first step," says Dr. Chan. "We want to find out why it is predictive, what is the molecular mechanism behind that. Once we understand the mechanism, we can design therapeutics to improve the treatment and outcome of patients." Dr. Chan and his colleagues at the Lymphoma/Leukemia Molecular Profiling Project, a consortium of institutions that collaborated on the New England Journal study, are also analyzing gene expression in other types of lymphoma, including Mantle cell and follicular lymphomas.

Another line of research involves correlating gene expression to genomic alterations within tumors. "We have cytogenetic and genomic data already available for many of these tumors," Dr. Müller-Hermelink says.

A clinical trial involving diffuse large-B-cell lymphoma patients who have relapsed is also slated to begin later this year at the NCI and other institutions. It will target the nuclear factor kB signaling pathway, which has been shown to prevent tumor cell death after chemotherapy in activated B-cell-like lymphomas.

The gene expression patterns of patient tumors will be profiled before treatment with an inhibitor of NF-kB and chemotherapy. "We will then use the profiles, after the fact, to find out who responded and who didn't," says the NCI's Dr. Staudt.

"Because we're using a new therapy in a clinical trial," he adds, "by definition we would not be able to predict with certainty which genes would be most important for dictating response. The actual set of genes will change with each change in therapy. We may find a new signature."

Vida Foubister is a writer in Mamaroneck, NY.