Objective: Hierarchical clusterings of diffuse large B-cell lymphoma (DLBCL) based on gene expression signatures have previously been used to classify DLBCL into Germinal Center B-cell (GCB) and Activated B-cell (ABC) types. To examine if it was feasible to perform a cross-platform validation on the Affymetrix HG-U133A oligonucleotide arrays and improve the classification, we determined the expression profiles of pretreatment, diagnostic samples from 52 primary nodal DLBCL.
Methods and results: First, three previously published gene lists were converted to the HG-U133A probe sets and used for hierarchical clustering. In this way, three subtypes, including the GCB type (n = 20), the ABC type (n = 25) and an intermediate group, Type-3 (n = 5), were distinguished. The CD10 and Bcl-6 expression as well as t(14;18) translocation were prevalent, but not exclusive to the GCB type. By contrast, MUM1 was only expressed in the ABC and in Type-3 samples. The 5-year survival was similar between the groups, but GCB patients showed a better initial response to CHOP or CHOP-like regimens than the remaining patients and tended to have less advanced disease and lower IPI scores. As a next step, an improved set of classifier genes was generated by analysis of 34 patients that were consistently classified as GCB or ABC in the above analyses. Seventy-eight genes were selected and demonstrated on two previously published data sets (Shipp et al. Nat Med 2002;8:68-74 and Houldsworth et al. Blood 2004;103:1862-1868) to exhibit a higher specificity than the original gene lists.
Conclusion: We conclude that gene expression profiling with Affymetrix Genechips is efficient to distinguish between GCB and ABC types of DLBCL and that these are likely to represent separate biological entities. The Genechip platform is highly standardised and therefore useful for future prospective investigations to establish the value of gene expression profiling in the clinical management of DLBCL.