Improving Diagnosis and Treatment of Lymphomas with Gene Expression Profiling

Joseph M. Connors, M.D.

Dr. Connors chairs the Lymphoma Tumor Group and Research Ethics Board, is the Acting Leader of the Provincial Systemic Therapy Program, and is the Acting Head of the Division of Medical Oncology at the University of British Columbia.

Over the past half century, enhanced understanding of lymphoproliferative neoplasms has been driven by improvements in understanding their basic biology. Fundamentally new insights into these diseases have emerged from the application of novel technologies. In the area of cytogenetics, we have moved from the low-resolution observations of G banding to the fine detail of fluorescent in situ hybridization (FISH) and spectral karyotyping (SKY)¹ allowing demonstration of multiple specific diagnostic translocations in routine hematopathologic practice. This, in turn, permits precise diagnosis of a variety of lymphomas, for example, mantle cell lymphoma with its t(11;14) and upregulation of BCL1 resulting in inappropriate production of cyclin D1. Starting with sheep red blood cell rosettes and taking advantage of the enormously useful monoclonal antibodies generated from hybridoma technology, we now have panels of well-characterized immunohistochemical reagents permitting accurate assignment of cell lineage and providing surface antigen-based profiles that help us reliably distinguish over thirty unique subtypes of lymphoproliferative neoplasms. The most recent major step forward in the application of evolving technology to understanding lymphomas is DNA microarray-based gene expression profiling². For the first time we have a tool that allows us to appreciate the richness of whole genome gene expression patterns, instead of having to focus on one or a few up- or downregulated genes.

Recently, four major demonstrations of how gene expression profiling can enhance our understanding of specific lymphomas have been provided. Examining each is instructive. Major new insights into the biology of lymphomas have resulted from the use of microarray gene expression technology to elucidate their underlying biology and to identify novel pathways for therapeutic intervention. These studies have substantially improved our understanding of four of the most common B cell lymphomas: diffuse large B cell lymphoma (DLBCL), mantle cell lymphoma, primary mediastinal large B cell lymphoma (PMBCL), and follicular lymphoma.

Lymphomas display enormous variation in their natural history and response to treatment, usually associated with differing histologic diagnoses or easily measured clinical factors such as lactate dehydrogenase level or patient performance status. However, even with apparently equal tumor burdens at presentation, some patients are cured and others succumb to lymphoma. These differences in outcome must reflect intrinsic gene expression variations but have been difficult to characterize with single gene expression studies (essentially, immunohistochemical staining or, occasionally, quantitative PCR). Microarray-based gene expression studies, with their ability to characterize whole families of gene expression pathways, provide novel insights into the basic biology of each lymphoma that cannot be gathered in any other way.

The key insights provided thus far by gene expression studies have differed from one lymphoma to another. In DLBCL, the most frequently encountered type of non-Hodgkin lymphoma, constituting almost 40 percent of the lymphomas seen in North America, a fundamental distinction has been found separating cases into germinal center B cell (GCB) and activated B cell (ABC) type lymphomas with the former preserving many of the gene expression patterns of normal germinal center B cells and displaying at least twice as good a prognosis when compared to the latter, which confers a prognosis much worse than that of the GCB type (five-year overall survival 60 percent versus 35 percent for the comparison of GCB with ABC types)³. In contrast to these findings in DLBCL, which show that what we thought of as a single entity has at least two biologically distinct subtypes, the gene expression profiling of mantle cell lymphoma confirms that this entity is quite homogeneous in terms of the over- and underactivated gene expression patterns. However, prognosis is dramatically affected by the number and type of proliferation-associated genes that are overactive⁴. The quartile of patients with the lowest proliferation score had a five-year survival of over 70 percent while the quartile with the highest score had a median survival of less than one year and a five-year overall survival of 10 percent⁴. These data from DLBCL and mantle cell lymphoma indicate that the potential uses of gene expression profiling extend beyond histologic subtyping to clinically relevant assignment of prognosis.

Primary mediastinal large B cell lymphoma (PMBCL) has been thought of as a special subtype of DLBCL. Its distinct clinical presentation in younger patients with a female predominance has led to the suspicion that it constitutes a unique entity. However, reliable distinction from DLBCL has remained elusive. Gene expression analysis has settled the question and provided surprising insight into this lymphoma's unique biology, including similarities to Hodgkin lymphoma⁵. Based on a core cluster signature of 46 specific genes, 35 of which are relatively overexpressed in PMBCL and 11 in DLBCL, PMBCL can be quite accurately distinguished from all other types of large B cell lymphomas. This unique gene expression signature includes a number of regulators of T cell activation, but, even more surprisingly, over one-third of the genes that are more highly expressed in PMBCL than in other large B cell lymphomas are also characteristically overexpressed in cells from putative Hodgkin lymphoma cell lines. Of particular note is the overexpression of several genes associated with nuclear factor κ B (NFκB), revealing a pivotal role for this gene complex and suggesting possible points of attack for future therapeutic agents^{5, 6}. In the case of PMBCL, the application of gene expression profiling has provided not only improved accuracy of diagnosis and provocative suggestions for possible targets for new therapeutic agents, but also intriguing linkages to Hodgkin lymphoma. Hodgkin lymphoma is generally considered quite separately from the other lymphomas, but it, in retrospect, shares many characteristics with PMBCL, including presentation in younger patients, frequent involvement of the mediastinal lymphoid tissue, high cure rates, and a high level of durability of responses with very low late relapse rates.

The most recent study of gene expression profiling for lymphoma has concentrated on the second most common of the lymphomas seen in North America, follicular lymphoma (see the Diffusion article, "Immune Response and Clinical Course in Follicular Lymphoma" on page 10 of this issue). Once again, surprises emerged from the analysis⁷. As expected, the gene expression profiling again permitted identification of subgroups with widely different prognoses. Using a signature set of genes uniquely predictive for follicular lymphoma, four quartiles with widely disparate median lengths of survival (13.6, 11.1, 10.8, and 3.9 years) could be distinguished, a much more powerful separation of subgroups than is available with current prognostic models based on clinical or laboratory markers (Figure 1). What was unexpected was the observation that two major components of the follicular lymphoma signature gene set are reflective of specific functions of the infiltrating non-neoplastic cells in the biopsies. The component of the signature labeled the immune-response 1 signature includes selected genes encoding markers of T cell activation and function and others that are highly expressed in macrophages. The immune-response 2 signature includes genes known to be preferentially expressed in macrophages, dendritic cells, or both. Of importance to note is that in both cases, the genes expressed are linked to specific T cell, macrophage, and dendritic cell functions, not just the presence of those cells. Even more intriguing, the overexpression of the immune-response 1 signature correlates with an improved prognosis, but that of the immune-response 2 signature with a poorer outcome. An accurate prediction can only be made when the relative contributions of both signatures are taken into consideration⁷. This analysis of follicular lymphoma gene expression patterns adds another dimension to what can be learned from these studies. In addition to gene expression profiling's virtues of more accurate diagnosis, better prediction of prognosis, and biologic insights suggesting promising therapeutic targets, we can add enhanced appreciation of the contribution of other immune effector cells to the disease process and outcome.

The ultimate utility of any technicologically novel investigative technique takes some time to be realized. Just as it has taken decades to begin to fully realize the immense potential of monoclonal antibody technology, moving from imprecise diagnostic tests through the accumulation of more than 200 well validated current cluster designation markers and widely recognized specific immunophenotypic profiles to potent therapeutic agents such as rituximab and alemtuzumab, we can expect that the full potential of DNA-based gene expression profiling is only beginning to be realized. Validation of a widely commercially available and affordable platform for performance of signature-based gene expression profiling is now being planned by the Leukemia Lymphoma Molecular Profiling Project. Small numbers of surrogate immunohistochemical markers⁸ or markers identified using the real-time polymerase chain reaction (RT-PCR) technique⁹ may allow rapid reproducible identification of such important subtypes of DLBCL as the GCB and ABC variants. Clinical trials incorporating gene expression profiling are underway and promise to prospectively validate the initial observations described in this article (Bruce Cheson, CALGB, personal communication) and test the hypothesis that targeting key gene expression pathways will produce useful therapeutic improvements (for example, anti-NFκB agents for PMBCL). Additional profiling of less common B cell neoplasms such as Burkitt lymphoma and the T cell lymphomas is planned for the near future. DNA-based gene expression profiling of lymphomas is here to stay and our ability to diagnose and treat lymphomas is the better for it.

References:

1. Schrock E, et al. Science 1996;273(5274):494-7.
2. Staudt LM. N Engl J Med 2003;348(18):1777-85.
3. Rosenwald A, et al. N Engl J Med 2002;346(25):1937-47.
4. Rosenwald A, et al. Cancer Cell 2003;3(2):185-97.
5. Rosenwald A, et al. J Exp Med 2003;198(6):851-62.
6. Savage KJ, et al. Blood 2003;102(12):3871-9.
7. Dave SS, et al. N Engl J Med 2004;351(21):2159-2169.
8. Hans CP, et al. Blood 2004;103(1):275-82.
9. Lossos IS, et al. N Engl J Med 2004;350(18):1828-37.

Return to Table of Contents