Epigenetic rewiring of BCL6 drives responses and unveils synthetic dependencies in large B cell lymphoma

Diffuse large B cell lymphoma (DLBCL) is a clinically and molecularly heterogeneous malignancy that can be classified into germinal center (GCB) and activated B cell (ABC) subtypes. Clinically, these subsets are often classified by immunohistochemistry (IHC) with CD10 and BCL6 expression being characteristic of GCB and IRF4 (MUM1) expression being characteristic of ABC. Efforts to inhibit signaling pathways active in ABC DLBCL by addition of targeted therapies to frontline chemoimmunotherapy have failed to meet their primary endpoints, and mechanisms of resistance to targeted therapies in DLBCL have not been extensively explored in clinically relevant models. Targeting BCL6 has recently emerged as an exciting therapeutic direction, with multiple BCL6 degraders now being tested in early phase clinical trials. BCL6-targeting transcriptional/epigenetic chemical inducers of proximity (TCIP) are an innovative approach to recruit transcriptional coactivators that flip BCL6 from a transcriptional repressor to an activator and induce the expression of its target genes. We thoroughly tested the efficacy of a TCIP that recruits BRD4 to BCL6-bound site (TCIP1) using our extensive PDX repository, evaluating models with uniformly high (n=5), heterogeneous (n=6) or negative (n=2) expression of BCL6 by IHC. PDX models with high BCL6 expression all responded rapidly to TCIP1, with eradication of tumors in 4/5 models including those from CAR T refractory tumors with p53 mutation. As expected, BCL6 negative models showed no response. Interestingly, BCL6 heterogeneous models also showed no response compared to vehicle control. Mechanistic studies by RNA-sequencing following short term in vivo exposure to TCIP1, its constituent components (BCL6 BTB binder; BRD4 binder) or vehicle control in BCL6 high (n=3) or heterogeneous (n=3) models revealed a selective up-regulation of IRF4 activity in BCL6 heterogeneous models. Evaluation by IHC showed that, at baseline, BCL6 heterogenous models consisted of mixed populations of BCL6+IRF4- and BCL6-IRF4+ cells that were polarized to a uniformly BCL6-IRF4+ state under TCIP1 pressure, then returned to heterogeneous states following secondary implantation. IRF4 can be indirectly targeted using lenalidomide (len), thus we evaluated the activity of TCIP1+len combination in two BCL6 heterogeneous models. Single agent TCIP1 or len treatment resulted in no significant reduction in tumor volume compared to vehicle control. However, TCIP+len drove significant in vivo responses. In conclusion, we present the first data implicating epigenetic mosaicism and plasticity as a resistance mechanism to targeted therapy in DLBCL. Using a large array of PDX models we show impressive efficacy for TCIP1 in BCL6 high DLBCL and identify a rational strategy to overcome escape via epigenetic plasticity using a combination of TCIP1 and lenalidomide to target polar epigenetic states.