EZH2 Enables the Proliferation of Germinal Center B Cells and DLBCL through a Rb-E2F1 Positive Feedback Loop Involving Repression of CDKN1A

Many B cell lymphomas arise from germinal center (GC) B cells of the humoral immune system, which are unique in their ability to replicate at an accelerated rate, which requires attenuation of replication checkpoints. Upon activation, GC B cells upregulate EZH2, a Polycomb protein that mediates transcriptional repression by trimethylating histone 3 lysine 27 (H3K27me3). Conditional deletion of EZH2 results in failure to form GCs. EZH2 is often highly expressed or affected by somatic gain of function mutations in GC B cell-derived diffuse large B cell lymphoma (DLBCL) and is required to maintain lymphoma cell proliferation and survival. Our previous research identified CDKN1A (p21 ^Cip1) as a direct target of EZH2 in GC B cells and DLBCLs. EZH2 causes promoter H3K27 trimethylation and transcriptional repression of CDKN1A in GC B cells and DLBCL cells. Treatment of DLBCLs with a specific EZH2 inhibitor (GSK343) or EZH2 shRNA caused CDKN1A H3K27me3 demethylation and derepression. Based on these considerations we hypothesized that silencing of CDKN1Athrough H3K27me3 might explain the proliferative GC and DLBCL phenotype.

To test this notion, we crossed GC-specific conditional Cg1Cre;Ezh2^fl/fl mice with Cdkn1a^-/- mice. We assessed GC formation after T cell-dependent immunization in double vs. single Cdkn1a or Ezh2 KO mice. Cdkn1a^-/- mice manifested perfectly normal GC formation, whereas there was complete absence of GCs in Cg1Cre-Ezh2^fl/fl mice. In contrast, Cg1Cre;Ezh2^fl/fl;Cdkn1a^-/- double KO mice exhibited normal GC formation as measured by immunohistochemistry and flow cytometry. While conditional deletion of Ezh2 in GCs abrogates immunoglobulin affinity maturation, the double KO mice manifested normal development of high affinity antibodies after specific antigen exposure (NP-KLH). Cell cycle analysis of double KO mice showed a similar proportion of GC B cells in S phase as WT or Cdkn1a^-/- controls, as measured by BrdU incorporation, indicating that loss of p21 allows progression of cell cycle. These effects were linked to the methyltransferase function of EZH2 since Cdkn1a^-/- also rescued the loss of GCs driven by administration of EZH2 inhibitor observed in WT mice. We observed a similar phenomenon in DLBCL cells since shRNA-mediated depletion of CDKN1A rescued the growth suppressive effect of EZH2 shRNA or specific EZH2 inhibitors. Therefore H3K27me3 and repression of CDKN1Aexplains to a large extent how EZH2 enables GC formation and maintains growth of DLBCL cells.

To further understand the role of EZH2 as a driver of the cell cycle we explored its relation to the G1/2 checkpoint regulated by p21^Cip1. We found that GC B cells from Cg1Cre;Ezh2^fl/fl;Cdkn1a^-/- double KO mice exhibited high levels of phospho Rb by IHC, similar to the levels found in WT or Cdkn1a^-/- control mice. Hyperphosphorylation of Rb induces its inactivation, allowing the release of E2F transcription factors and cell cycle progression. EZH2 was previously shown to be a direct target of E2F1, E2F2 and, to a lesser extent E2F3. Among these we found that E2F1 mRNA and protein expression are especially highly expressed and upregulated in GC B cells vs. naïve B cells. By qChIP we show that E2F1 is bound to the EZH2 promoter in GC-derived DLBCL cell lines. Moreover, E2F1 gene expression is positively correlated with EZH2 (R=0.35, p<0.0001) and moderately inversely correlates with CDKN1A (R=-0.22, p<0.0001) in a cohort of 757 DLBCL patient samples. Therefore, we explored the function of E2F1 in GC formation. We found that E2f1^-/- mice developed reduced number and size of GCs as compared to control mice (E2f1^-/- vs. WT, p<0.01). To determine if this phenotype was due to a lack of induction of EZH2 by E2F1, we transduced bone marrow of E2f1^-/- or WT donor mice with retrovirus encoding EZH2-GFP or GFP alone, transplanted them into lethally irradiated recipients and assessed the GC reaction after immunization. Notably, EZH2 expression successfully rescued E2f1^-/- phenotype (E2f1^-/-+GFP vs.E2f1^-/-+EZH2, p<0.001), indicating that the pRb-E2F1 pathway drives the GC reaction by inducing EZH2.

In summary we identified a positive feedback loop required for GC formation and DLBCL whereby EZH2 controls GC B cell proliferation by suppressing the critical cell cycle checkpoint gene CDKN1A, allowing cell cycle progression with a concomitant phosphorylation of Rb. This causes the release of E2F1, which positively regulates the expression of EZH2.