College of American Pathologists
Printable Version

  What microarrays say about lymphomas


cap today



October 2006

Feature Story

William Check, PhD

How many molecules does it take to distinguish Burkitt’s lymphoma from diffuse large-B-cell lymphoma? Between 2,500 and 2,600, according to one recent article.

The current World Health Organization classification already goes beyond morphology to include molecular data such as immunophenotyping, which measures expression of CD antigens, and evaluation of c-myc and bcl-2 translocations. Even so, differentiating Burkitt’s from diffuse large-B-cell lymphoma remains imperfect, as the authors of this recent article noted. They pointed out that the characteristic t(8;14) translocation of Burkitt’s lymphoma is also found in five to 10 percent of cases of diffuse large-B-cell lymphoma; since the latter malignancy is 20 times more prevalent, an aggressive lymphoid malignancy with this translocation has at best a 50:50 chance of being Burkitt’s based on the translocation alone. As a result, they write, "[A] lymphoma with a t(8;14) translocation can present a diagnostic problem" (Dave SS, et al. N Engl J Med. 2006; 354:2431-2442).

Distinguishing between these two lymphomas is important because Burkitt’s requires more aggressive chemotherapy. Accordingly, this international group of investigators, who compose the Lymphoma/ Leukemia Molecular Profiling Project, based in the laboratory of Louis M. Staudt, MD, PhD, of the National Cancer Institute, set out to determine whether molecular diagnosis, in the form of gene expression profiling with oligonucleotide microarrays, could improve differential diagnosis of these two lymphoid neoplasms. Profiling was pursued with two arrays: an Affymetrix U133 array, representing more than 39,000 transcripts derived from about 33,000 human genes; and a custom array containing a selected subset of 2,524 genes that have been useful in the diagnosis of lymphoma and genes known to be important in oncology and immunology. The researchers concluded that "[T]he molecular classifier of Burkitt’s lymphoma based on gene expression provides a quantitative and reproducible diagnosis of Burkitt’s lymphoma that is superior to the best current diagnostic methods." Fortunately, the full-scale Affymetrix array was not necessary; the smaller array was diagnostically accurate.

A number of complex comparisons led to the conclusion that an array can beat a panel of expert pathologists in this application. For instance, 26 cases originally called Burkitt’s or Burkitt-like lymphoma at the contributing center had been reclassified on review by the panel to diffuse large-B-cell lymphoma or high-grade lymphoma, not otherwise specified. However, microarray analysis showed that eight of these 26 cases had the gene-expression profile of Burkitt’s lymphoma, which the authors interpreted as "suggesting they represent cases of Burkitt’s lymphoma that are difficult to diagnose by current methods." Outcome data were available for seven of the eight discrepant cases. Five patients treated with standard regimens had not survived beyond two years, while one of two who had received intensive regimens lived more than five years after diagnosis.

Using gene expression microarrays to characterize lymphomas has been an ongoing goal of the Profiling Project and other international teams of investigators. An initial accomplishment was the division of diffuse large-B-cell lymphoma into two categories, one that expressed genes characteristic of germinal center B cells (germinal center B-cell-like) and a second that expressed genes normally induced during in vitro activation of peripheral blood B cells (activated B-cell-like diffuse large-B-cell lymphoma) (Alizadeh AA, et al. Nature. 2000;403:503-511). Later a third category was added, primary mediastinal diffuse large-B-cell lymphoma. Microarray profiles of these categories of lymphoma predicted survival after chemotherapy (Rosenwald A, et al. N Engl J Med. 2002;346:1937-1947). Specific chromosomal alterations were found to be associated with significant changes in gene-expression signatures and to reinforce their prognostic value (Bea S, et al. Blood. 2005;106:3183-3190). Similar findings have emerged with regard to mantle cell lymphoma (Rosenwald A, et al. Cancer Cell. 2003;3:185-197).

"Typical cases of Burkitt’s lymphoma are not that difficult to diagnose," says Wing C. Chan, MD, co-chair with Dr. Staudt of the Profiling Project and Amelia and Austin Vickery professor of pathology and co-director of the Center for Lymphoma and Leukemia Research, University of Nebraska Medical Center, Omaha. "But," he adds, "there are cases that have an atypical morphology or immunophenotype that present a diagnostic problem." That was the reason for the profiling studies, he says, "to see if that would be a way to define Burkitt’s lymphoma more accurately." In Dr. Chan’s view, the data showed that molecular profile is "excellent" and can objectively identify Burkitt’s lymphoma. "For typical cases it is not so necessary to have a molecular profile," he says. "But for cases with ambiguous morphology or phenotype or cytogenetics, a molecular profile will provide additional help."

One problem in making firm conclusions from the data, Dr. Chan acknowledges, is that there is no gold standard to evaluate the findings. Criteria in the study included cytogenetics, pathologist review, phenotyping, and response to treatment. "There were a number of criteria that we could correlate, but no gold standard," he says. Corresponding data correlated well. But, he says, "None of them were perfect. The closest to a gold standard is the diagnosis of a panel of experienced hematopathologists using all the available information."

Practically, Dr. Chan sees a 1,000- to 2,000-gene array as an adjunct to traditional pathologic workup, perhaps substituting for many ancillary investigations. "One thousand genes on a chip is really simple," he says. "Since the diagnostic algorithm is based on gene expression signatures, the easiest thing is to go directly to a small microarray. It would be more work to transform the signatures into RT-PCR and even more work to transform them into immunohistochemistry. We would need to make antibodies and to show that gene expression data can be translated into protein data." And possibly the immunohistochemistry assays would need to be quantified. Dr. Chan sees the microarray format as being well within the capability of most large medical center laboratories with an established molecular diagnostic section. "Can we eventually convert into an IHC format?" Dr. Chan asks. "That could be a possibility. We would need to reduce the number of markers. Immunostains are the format most familiar to pathologists."

"I think this is important news," says Nancy Lee Harris, MD, Austin L. Vickery professor of pathology at Harvard Medical School and Massachusetts General Hospital, who wrote the editorial accompanying this article and a similar report on the same topic from a German group (Hummel M, et al. N. Engl J Med. 2006;354:2419-2430). "Its practical significance is less immediately obvious than its conceptual significance."

Conceptually, Dr. Harris told CAP TODAY, "This work demonstrates that there is a characteristic signature for Burkitt’s lymphoma and that two completely independent research groups using very dissimilar methods came up with essentially the same overall result—that it is possible to identify a gene expression signature that clearly distinguishes Burkitt’s lymphoma from other high-grade B-cell lymphomas." Not enough information is presented in the articles to know whether the gene expression signature the two groups identified includes the same genes. "I think they both found downregulation of NF-κB-related pathways in Burkitt’s cases," Dr. Harris says. "But the Hummel group did not talk much about specific genes, so it is hard to know how much overlap there is."

A second point of conceptual significance is that "There has been some debate about whether the defining feature of Burkitt’s is this myc gene rearrangement and whether other lymphomas with the myc gene rearrangement become Burkitt’s. I think these studies show this is not the case." Diffuse large-B-cell lymphomas carrying the myc translocation did not have the signature of Burkitt’s lymphoma and, based on the limited clinical followup presented, they did not seem to behave like Burkitt’s.

In Dr. Harris’ view, these two articles represent the beginning of the proof that cases identified by microarray profiling have the biological behavior of traditional Burkitt’s lymphoma. However, she cautions, "The clinical information in both of these studies is limited. It suggests that the Burkitt’s signature may be important in defining a clinical entity that responds to a particular therapy, but I think a prospective study would be required to clearly demonstrate that."

In addition to whether a myc translocation makes something a Burkitt’s lymphoma, another question has been whether a bcl-2 translocation makes a lymphoma into something other than Burkitt’s. "This was left a bit ambiguous in both of these studies," Dr. Harris says, "particularly in the Dave paper." She points to cases the Profiling Project studied that had both myc and bcl-2 translocations that otherwise had a signature very similar to the Burkitt’s type. "They tended to include those in the Burkitt’s category," Dr. Harris says. "But there is some evidence that cases that have bcl-2 or both translocations behave quite differently from classical Burkitt’s lymphoma." In her mind there is still a question of what to do with classification and treatment of cases that have a bcl-2 translocation. "I think at the least that pathologists should look for bcl-2 translocations and mention their presence in the report as a possible bad prognostic factor."

Practically, Dr. Harris says, "These studies show that it is possible to backtrack from microarrays and immunohistochemistry to identify some currently used and potentially additional antigens that could be included in a panel to distinguish Burkitt’s lymphoma from large-B-cell lymphoma." In her editorial is a table listing these antigens and how they could be used. These studies show, she says, how myc rearrangements can be used in a more sophisticated way in differential diagnosis (Harris NL, Horning SJ. N Engl J Med. 2006;354:2495-2498). For instance, the German group showed that myc rearrangement of the immunoglobulin genes is more common in Burkitt’s lymphoma. In other lymphomas, myc translocation is more often associated with non-immunoglobulin genes.

Dr. Harris doesn’t see microarrays themselves being incorporated into routine heme/path practice. "I don’t see that happening," she says, "certainly not soon." She wrote that RNA extraction and microarray analysis are "laborious and expensive."

"A big impediment to using diagnostic microarrays in practice is the technical difficulty of preserving mRNA in biopsy specimens and extracting it so that it can be applied to microarrays," she explains. "The number of specimens suitable for these studies when you try to do them retrospectively is not large. Given that most samples are from community hospitals and samples are placed in fixative at variable intervals from the time they are removed from patients, this is a really important technical obstacle."

Dr. Harris believes that these obstacles will be overcome quickly if microarrays turn out to be "incredibly useful." But for now she sees microarrays being more useful as tools for discovering genes and proteins than for actual clinical diagnosis. "Genes and proteins discovered in microarray analysis can be assayed using more traditional and less stringent methods," she says. "Then we won’t have to worry about mRNA extraction." Daniel Jones, MD, PhD, associate professor in hematopathology and director of the Molecular Diagnostics Laboratory at M.D. Anderson Cancer Center, Houston, agrees that both articles are "excellent technically, from a methodological viewpoint.

"These studies represent a proof of principle that one can classify lymphomas by molecular means," he told CAP TODAY. "But they remain proof-of-principle papers, rather than directly addressing whether treatment outcomes will vary depending on molecular subgroup."

Implications for outcome are fuzzy, Dr. Jones says, because cases were likely treated with a variety of regimens unrelated to their molecular profile, an inevitable drawback of a retrospective study.

Both research groups suggested there is a subset of cases whose microarray profile looked like that of Burkitt’s without detectable c-myc. Conversely, a number of cases had c-myc rearrangement but didn’t look like Burkitt’s on profiling. "These complexities had been known, but they have now been shown on a molecular level," Dr. Jones notes. He points out a day-to-day aspect of practice from this work. In his view, "the most important finding" from both articles was to show that cases with a c-myc translocation, when combined with other genetic events like bcl-2 gene rearrangement, can look very different from Burkitt’s lymphoma.

In the Hummel article, "cases that have the c-myc translocation yet don’t look like Burkitt’s still did poorly [with conventional lymphoma therapy]," he says. "The implication is that in those tumors with high-grade histologic features, regardless of their exact resemblance to Burkitt, it is probably worth doing a test for the c-myc translocation. This reduces the complexity of microarray analysis to a single test."

More generally, he notes that both groups started out by investigating selected cases with high-grade features. They demonstrated within high-grade cases that some look like Burkitt’s and at a very high frequency have the genetic signature of Burkitt’s. At the same time, there are cases that don’t look like Burkitt’s that have the Burkitt’s signature and those that don’t look like Burkitt’s and do have c-myc. "All three of those groups could potentially be identified by doing analysis for c-myc translocation," Dr. Jones says.

A "bigger take-home lesson" for lymphoma clinics, he says, is that "the time has arrived to do stratified prospective trials based on c-myc status in large-cell lymphoma comparing aggressive versus more conventional therapy."

As for translating the microarray data into a simpler format, Dr. Jones is skeptical. "One thing, paradoxically, about the lymphoma array story," he notes, "is that as you get more microarray data, enthusiasm for translating it into simpler tests decreases." The reason may be that more data has made people realize there are discordant tumors in every diagnostic subgroup. So, unlike the situation in breast cancer, enthusiasm for marker panels done by RT-PCR or IHC for large-cell lymphomas is declining.

At the same time, Dr. Jones acknowledges, doing whole genome microarray profiles raises the issues of getting fresh tumor biopsies and getting microarrays done in a timely manner. "That might be a bit more ambitious than we can handle at present," he says. "Our laboratory at M.D. Anderson," led by pathologist W. Fraser Symmans, MD, "is going to try that approach in breast cancer, which is more common and for which it will be easier to get a sufficient number of patients." They will prospectively explore whether therapy stratification using the entire Affymetrix array signature can be used to select optimal treatment.

"My main thinking about why lymphoma profiles haven’t taken off," Dr. Jones says, "is that there is not a lot of difference in treatment protocols. Since a very limited range of therapies is being tried, it is hard to convince clinicians that it is worthwhile going through all this trouble with microarray analysis."

As lymphoma clinicians become more "venturesome and creative in integrating targeted therapies," Dr. Jones says, "it will matter a whole lot more."

William Check is a medical writer in Wilmette, Illinois.