Section 2 -

Molecular Profiling Studies of Lymphoma Wing C. Chan

Introduction Cancer is typically initiated by a genetic alteration in a cell and this change predispose the cell to undergo further genetic alterations [1]. Subsequent secondary abnormalities contribute to the development of a malignant tumor. These cumulative genetic abnormalities alter the profile of transcripts in the neoplastic cells. Thus, each malignant tumor clone has its own molecular profile or signature of transcript that can be determined experimentally by microarray analysis. The characteristics of a tumor and its clinical behavior are determined by the unique set of genetic lesions in the tumor cells and the functional alterations are reflected by the unique gene expression signature of the tumor. Therefore, gene expression profiling can be exploited to improve not only the diagnosis and classification but also the prediction of treatment response and outcome of lymphomas. In the past six years, the microarray technology has been used extensively to investigate various kinds of human tumors including lymphomas that will be the focus of discussion in this presentation. Principle of Microarray Analysis A microarray consists of an ordered array of DNA probes on a solid support. The DNA probes attached onto the solid surface can be synthesized in situ or spotted as cDNAs or oligonucleotides prepared off-line [2, 3]. For spotted microarrays, the sample of interest (test sample) is generally measured against a reference standard to obtain relative expression levels that can be compared across experiments. A single DNA microarray experiment can measure the expression of thousand of genes simultaneously. Comparison of data across different array platforms is still challenging. However, while comparing the expression of individual genes may give discrepant results, the comparison of large groups of genes serving specific functions (signatures) is much more robust. Data Management and Analysis : Each microarray experiment generates thousands of measurements that need to be processed and analyzed. Image processing includes the accurate measurement of fluorescence from a very small surface area, proper background subtraction and data normalization. Many analytical tools are currently available to analyze the massive amount of information generated. [4, 5, 6, 7, 8] Gene expression data are frequently presented in the form of a matrix with the list of genes on one axis and samples on the other axis using one of clustering programs. Hierarchical clustering [5] started by associating pairs of genes with the most similar pattern of expression across samples and then successively combining this initial clusters into larger clusters until all the genes are grouped together onto a single dendrogram. Tumor samples can be similarly clustered according to their overall similarity in gene expression profiles. In supervised clustering, certain investigator-defined parameters are used to guide the clustering. There are certain pitfalls and limitations in microarray analysis [3]. Genes that are expressed at low levels are generally not reliably and reproducibly measured by current microarray technology. The cyanine dyes commonly used do not give equivalent direct labeling. Thousands of parameters are being measured in each tumor while the number of tumor samples is often quite limited making statistical assessment difficult. Because of this discrepancy in dimensionality, validation of the analytical conclusions becomes very important. Different analytical tools may be applied to assess the validity of the conclusions. Another frequently used approach is to divide the patient samples into a test set and a validation set. The validity of the conclusions drawn from the test set will then be examined on the validation set. One can also examine clinical or biological features of the cases to determine if there are meaningful correlations with the microarray findings. Gene Expression Profiling in Lymphoma: Illustrative Examples The potential of gene expression profiling in class prediction (classifying tumors into currently defined categories) and class discovery (finding new tumor types that are biologically meaningful) will be illustrated using some of the studies on diffuse large B-cell lymphoma (DLBCL). In one of earliest studies of DLBCL [7], it was demonstrated that samples from three lymphoproliferative disorders included in the study largely segregated into distinct clusters indicating that gene expression profiling is able to classify cases into existing categories. The investigators noticed that if they supervised the analysis of DLBCLs using a set of genes that are preferentially expressed by normal germinal center (GC) B-cells, two broad groups were delineated. One group expressed many of the genes in the GC-B-cell associated signature, while the other expressed few of these genes but, instead, expressed many of the genes up-regulated in peripheral blood B-cells activated by mitogenic stimuli: the activated B-cell (ABC) signature. Hence, two subgroups of DLBCL that appear to be biologically distinctive can be defined (class discovery). A significantly better overall survival was associated with the group of lymphoma with the GC-B cell-like profile and this association appeared to be independent of clinical risk factors.A subsequent larger study of 240 cases confirms the validity of this subclassification of DLBCL [9] There is an unusual type of DLBCL that presents in the anterior mediastinum (PMBCL) occurring preferentially in young female patients. The tumor tends to have prominent stromal fibrosis with tumor cells typically having moderately abundant pale cytoplasm and no detectable immunoglobulin expression. Whether this represents a unique type of DLBCL has been debated. Two recent studies have demonstrated that this tumor shows a distinct gene expression profile that can differentiate it from the GCB and ABC type of DLBCL. [10, 11] Interestingly, there is a substantial similarity of the gene expression profile of PMBCL with Hodgkin lymphoma suggesting that there may be some shared biological properties between these two tumor types [10, 11] Since the unique gene expression profile of a tumor is determined by the intricate interaction of the genetic abnormalities present, it is anticipated that genetic abnormalities with a major influence on the biology of the tumor, will segregate with tumor subsets defined by gene expression profiling. Recent studies have demonstrated unique profiles of genetic aberrations associated with the three subtypes of DLBCL. [12, 13, 14] Construction of Predictors for Survival In the construction of prognosticators, clinical data are used to supervise the analysis. Typically, the patients are divided into a training and validation set and the training set is used to identify genes or signatures that should be included in the predictor. An outcome predictor score using the Cox proportional hazard model will then be constructed and tested on the validation set. In the series of DLBCL reported by Rosenwald and coworkers [9], four gene expression signatures were found to predict outcome. Each of these signatures consists of hundreds of genes but a signature can be easily represented by a few genes without losing its predictive power. Shipp and coworker, using a different approach, have identified a 13 gene predictor for their series of DLBCL. [15] For mantle cell lymphomas, a group of genes that are associated with cell proliferation appeared to be the major determinant of prognosis [16]. The major molecular predictors of survival in follicular lymphoma (FL) appear to be related to the stromal response representing the composition and function of tumor infiltrating immune cells [17]. Thus the determinants of survival differ significantly in the three types of lymphoma studied. Interestingly, in both DLBCL and FL, significant components of the prognosticators reflect tumor/host interaction. This observation indicates that it is essential to obtain the stromal signature in tumors in constructing outcome predictors. These studies also indicate that gene expression profiling can indeed provide a new and more biologically relevant approach to predicting survival. Perspective: Reliable diagnostic gene expression signatures have been delineated for the majority of B-cell lymphomas accounting for about 85% of all cases. The remaining lymphoma classes are uncommon, and to study these tumors, a well coordinated international effort will be necessary to obtain sufficient samples for study. Molecular prognosticators have been constructed for DLBCL, FL and mantle cell lymphoma. However, the current prognosticators need to be further validated and refined. There are likely to be determinants of prognosis that are not yet discovered. It is also likely that certain biological variables are not readily measured by gene expression profiling alone. Array based comparative genomic hybridization [18]or FISH analysis of specific loci will provide additional valuable information. Abnormalities in specific genes such as mutations and methylation may also provide complementary information. Integrating all these information will result in a better model for predicting survival and will also help us understand the mechanisms underlying the difference in clinical and biologic behavior. It is hoped that novel agents will be developed based on the molecular targets identified from microarray experiments [19]. When novel, mechanism-based therapies become available, it will be essential to obtain the relevant molecular information on each tumor. One can envision the development of a diagnostic microarray containing all the information that is relevant to treatment decisions. Every tumor could be examined by the microarray at diagnosis to guide the appropriate therapeutic interventions. Comprehensive molecular diagnostics and individualized treatment may become a reality in the not-too-distant future. Reference: Pearson PL, Van der Luijt RB. The genetic analysis of cancer. J Intern Med. 1998;243:413-417 Chan WC. DNA microarray analysis: What can it tell us about the biology of lymphoid neoplasm? American Society of Hematologists. Blood 2001:205-209 Chan WC, Staudt LM. Gene expression profiling in lymphoid malignancies in "Expression Profiling of Human Tumors: Diagnostic and Research Applications.", Humana Press. 2003 Ermolaeva O, Rastogi M, Pruitt KD, Schuler GD, Bittner ML, Chen Y, et al. Data management and analysis for gene expression arrays. Nat Genet. 1998;20:19-23 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A. 1998;95:14863-14868 Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science. 1999;286:531-537 Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature. 2000;403:503-511 Hastie T, Tibshirani R, Eisen MB, Alizadeh A, Levy R, Staudt L, Chan WC et al. 'Gene shaving' as a method for identifying distinct sets of genes with similar expression patterns. Genome Biol. 2000;1:RESEARCH0003 Rosenwald A, Wright G, Chan WC, Connors JM, Campo E, Fisher RI et al. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med. 2002;346:1937-1947 Rosenwald A, Wright G, Leroy K, Yu X, Gaulard P, Gascoyne RD, et al. Molecular diagnosis of primary mediastinal B cell lymphoma identifies a clinically favorable subgroup of diffuse large B cell lymphoma related to Hodgkin lymphoma. J Exp Med. 2003;198:851-862 Savage KJ, Monti S, Kutok JL, Cattoretti G, Neuberg D, De Leval L, et al. The molecular signature of mediastinal large B-cell lymphoma differs from that of other diffuse large B-cell lymphomas and shares features with classical Hodgkin lymphoma. Blood. 2003;102:3871-3879 Huang JZ, Sanger WG, Greiner TC, Staudt LM, Weisenburger DD, Pickering DL, et al. The t(14;18) defines a unique subset of diffuse large B-cell lymphoma with a germinal center B-cell gene expression profile. Blood. 2002;99:2285-2290 Iqbal J, Sanger WG, Horsman DE, Rosenwald A, Pickering DL, Dave B, et al. BCL2 Translocation Defines a Unique Tumor Subset within the Germinal Center B-Cell-Like Diffuse Large B-Cell Lymphoma. Am J Pathol. 2004;165:159-166 Bea S, Zettl A, Wright G, Salaverria I, Jehn P, Moreno V, et al. Diffuse large B-cell lymphoma subgroups have distinct genetic profiles that influence tumor biology and improve gene-expression-based survival prediction. Blood. 2005;106:3183-3190 Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RC, et al. Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med. 2002;8:68-74 Rosenwald A, Wright G, Wiestner A, Chan WC, Connors JM, Campo E, et al. The proliferation gene expression signature is a quantitative integrator of oncogenic events that predicts survival in mantle cell lymphoma. Cancer Cell. 2003;3:185-197 Dave SS, Wright G, Tan B, Rosenwald A, Gascoyne RD, Chan WC, et al. Prediction of survival in follicular lymphoma based on molecular features of tumor-infiltrating immune cells. N Engl J Med. 2004;351:2159-2169 Ishkanian AS, Malloff CA, Watson SK, DeLeeuw RJ, Chi B, Coe BP, et al. A tiling resolution DNA microarray with complete coverage of the human genome. Nat Genet. 2004;36:299-303 Ebert BL, Golub TR. Genomic approaches to hematologic malignancies. Blood. 2004;104:923-932