Refining and Redefining the Heterogeneity of DLBCL

Diffuse large B-cell lymphoma (DLBCL) represents a heterogenous collection of lymphomas with varied prognoses and clinical outcomes. Up to 60 percent of patients are cured with upfront chemotherapy, leaving 40 percent of patients having not responded to or relapsed after first-line treatment. The international prognostic index (IPI) score, which is based on certain patient and clinical characteristics, can differentiate between low-, intermediate-, and high-risk groups.¹  Gene-expression profiling of DLBCL identifies three distinct subgroups — germinal center B cell–like (GCB), activated B cell–like (ABC), and type 3, all with different outcomes.^2,3  GCB lymphomas have a significantly improved prognosis over ABC and type 3 lymphomas. This is interesting because another high-risk factor, lymphomas that harbor translocations involving the MYC gene on chromosome 8 as well as a second gene (most commonly BCL2 on chromosome 18), or so-called “double-hit lymphomas,” are almost exclusively of the GCB subtype.⁴  This distinction, then, between GCB and ABC subtypes, is incomplete. Dr. Bjoern Chapuy and colleagues analyzed 304 primary, newly diagnosed DLBCL biopsies by whole-exome sequencing, with an expanded bait set to also identify structural variants and somatic copy number alterations. This analysis led to refinement of our current understanding of the genetic and molecular subtypes of DLBCL and has implications for the design of treatment trials in this collection of diseases.

Dr. Chapuy and colleagues identified nearly 100 candidate cancer genes that are recurrently mutated in DLBCL, as well as the most commonly rearranged genes and areas of chromosomal copy number gain or loss, leading to copy number gain or loss of identifiable driver genes. On average, tumors in this series had alterations in 17 driver genes, with a total of 158 distinct alterations identified. Using non-negative matrix factorization consensus clustering, five distinct genetic subtypes of DLBCL were identified (C1-C5). A sixth subtype, C0, had no detectable alterations. C1 and C5 were largely composed of ABC DLBCLs but differed from each other with respect to prognosis and genetic alterations. C5 lymphomas had frequent gains in chromosome 18q with increased expression of BCL2 and MALT1, as well as mutations in CD79 and MYD88. This genetic signature was found in three of four patients with primary testicular lymphoma and one of two patients with primary central nervous system lymphoma, consistent with a predilection for extranodal involvement. This subtype had the predicted outcome for ABC DLBCL, with only approximately 40 percent of patients with long-term progression free survival (PFS) following R-CHOP chemotherapy (rituximab + cyclophosphamide, doxorubicin, vincristine, and prednisolone). C1 lymphomas, alternately, did quite a bit better with R-CHOP than C5 lymphomas and other ABC subtype lymphomas; long-term PFS was seen in nearly 70 percent of these patients. These lymphomas frequently had structural variants in BCL6 and mutations in NOTCH2; they also exhibited alterations predicted to lead to immune escape, such as mutations in BM2, CD70, and FAS, and rearrangements involving PDL1 and PDL2. This genomic signature was reminiscent of marginal zone lymphoma, perhaps indicating that these lymphomas represented transformation from a previously unrecognized marginal zone lymphoma, or that they arose from an extrafollicular or marginal zone origin. C3 and C4 lymphomas, alternately, were largely GCB-like DLBCL. C4 lymphomas had a relatively good prognosis, as expected for GCB lymphomas, with nearly 70 percent of patients remaining alive and free of relapse with prolonged follow-up. These lymphomas had mutations in core histone genes as well as in genes involved in immune evasion, the B cell receptor/PI3 kinase pathway, NFKB pathway, and JAK/STAT pathway. C3 lymphomas meanwhile, did relatively poorly compared with other GCB lymphomas. These lymphomas were associated with alterations involving BCL2 and mutations in chromatin remodeling genes such as EZH2 and CREBBP. These mutations are frequently seen in follicular lymphoma, suggesting that like C1, C3 lymphomas may represent transformation from a previously unrecognized indolent (follicular) lymphoma. Finally, C2 lymphomas, which were a mix of both GCB and ABC lymphomas, had frequent alterations in TP53 and/or frequent loss of chromosome 17p. As expected, these lymphomas did poorly, with less than 40 percent of patients being alive and free of progression with prolonged follow-up. Double-hit lymphomas were found in both C2 and C3 lymphoma subgroups.

This analysis is being reported simultaneously with that of the group from the National Institutes of Health (NIH), who identified four distinct genomic DLBCL subgroups using exome and transcriptome sequencing, array-based DNA copy number analysis,  and targeted amplicon resequencing of 372 genes: MCD, BN2, N1, and EZB.⁵  Some of their subgroups, such as the MCD subgroup and EZB subgroups with mutations in MYD88 and CD79B, and alterations in EZH2 and BCL2, correspond with the C5 and C3 subgroups in the Dr. Chapuy study, respectively. Interestingly, while both the MCD and C5 subgroups did similarly poorly with R-CHOP chemotherapy, the EZB and C3 subgroups, while genetically similar, seemed to have divergent outcomes. The EZB subgroup had superior, and relatively good, outcomes compared with the C3 subgroup. The NIH group also identified a group of lymphomas with recurrent alterations in BCL6 and NOTCH2, the BN2 subgroup, which was similar to the C1 subgroup. The BN2 subgroup, however, included both ABC and GCB lymphomas as opposed to the largely ABC-enriched C1 subgroup. These groups were similar, however, with respect to outcome following R-CHOP chemotherapy. Finally, the NIH group identified a fourth subgroup, the N1 subgroup defined by recurrent mutations in NOTCH1, that was nearly universally classified as the ABC subtype and carried the worst prognosis of all the subgroups identified.