The EZH2 histone methyltransferase is the enzymatic core of the Polycomb repressor 2 (PRC2 complex), is highly upregulated in germinal center (GC) B cells and is targeted by gain-of-function somatic mutations that enhance its ability to trimethylate histone 3 lysine 27 in diffuse large B cell lymphomas (DLBCLs) and follicular lymphomas (FLs). We explored the significance and mechanism of action of EZH2 in normal GC development and lymphomagenesis. We observed that EZH2-conditional knockout mice and mice exposed to the novel EZH2-specific inhibitor GSK503 both completely failed to form GCs or high affinity antibodies. Using ChIP-seq, sequential QChIP, RNA-seq and functional assays we demonstrated that EZH2 mediates the GC phenotype through de novo formation of bivalently marked chromatin domains (characterized by overlapping H3K27me3 repressive mark with the H3K4me3 activation mark) at the promoters of target genes involved in cell cycle regulation (e.g. CDKN1A) and in GC exit and terminal differentiation program (e.g. IRF4 and PRDM1). Notably, mutant EZH2 caused hyper-repression of these bivalent genes through increased H3K27me3, which we showed is causal to the mutant EZH2 phenotype. Mice engineered to conditionally express lymphoma-associated EZH2Y641F exhibited aberrant suppression of bivalent gene expression leading to increased proliferation, blockade of terminal differentiation, and massive GC hyperplasia. Transcriptional profiles of human DLBCL patients revealed that those with mutant EZH2 display a unique signature consisting of silencing of GC bivalent genes, suggesting that mutant EZH2 contributes to human lymphomagenesis through paralysis of bivalent chromatin domains.

This scenario is reminiscent of the role of the transcriptional repressor BCL6, which is also required for GC formation. BCL6 also represses CDKN1A, IRF4 and PRDM1 and is required to maintain the proliferation and survival of DLBCL cells. Notably BCL6 represses its targets by associating with BCoR, which forms a variant of Polycomb repressor 1 (PRC1) complex. We hypothesized that EZH2 and BCL6 cooperate to mediate the GC B-cell phenotype and when aberrantly active may cooperate to form GC-derived B-cell lymphomas. Using ChIP-seq studies we found that the target promoters of BCL6-BCoR complex (but not promoters with BCL6 complexes lacking BCoR) significantly overlap with EZH2 bivalent promoter genes in primary human GC B cells and lymphoma cells (Hypergeometric test, p=1.5x10-26). Treatment of DLBCL cells with EZH2 or BCL6 inhibitors or siRNA partially derepressed these genes indicating that both factors cooperate and are required to mediate full repression of these crucial loci. To determine whether EZH2 and BCL6 cooperate to generate GC-derived lymphomas, we transduced bone marrow of IµHABCL6 mice (which mimic BCL6 translocations in DLBCL) with retrovirus encoding mutant EZH2Y641F or GFP alone, and transplanted them into lethally irradiated recipients. Only EZH2Y641F/BCL6 mice showed an accelerated lethal phenotype (log-rank test, p=0.007), with reduced median survival (EZH2Y641F: 309 days, empty vector: 453 days). Serial bone marrow transplantation resulted in even further increased lethality (log-rank test, p=0.004; median survival EZH2Y641F: 127 days, empty vector: 169 days). Given the oncogenic cooperation between BCL6 and EZH2, we hypothesized that rational combinatorial therapy with BCL6 and EZH2 inhibitors might synergistically kill DLBCLs. Indeed, by combining the EZH2 inhibitor GSK343 and the RI-BPI, a drug that inhibits BCL6 by abrogating its interaction with BCoR, we observed a potent synergistic effect on the inhibition of DLBCL cell lines proliferation. The combination of these two inhibitors in mice bearing DLBCL xenografts accordingly suppressed tumor growth more effectively than either agent alone. Finally, the combination also yielded further killing of primary human DLBCL cells growth in a co-culture system that we developed for testing primary human specimens.

In summary we identified the first epigenetic mechanism of lymphomagenesis involving aberrant repression of GC-specific bivalent domains by EZH2 (PRC2) in cooperation with BCL6-BCoR (PRC1) complexes, as well as a rational epigenetic-based and molecular targeted therapeutic approach with the potential to eradicate lymphomas without harming normal tissues.


Creasy:GlaxoSmithKline: Employment.

Author notes


Asterisk with author names denotes non-ASH members.