The Paradoxical Efficacy of KDM6 Inhibition in Germinal Centre B-Cell Lymphomas

Introduction

Germinal centre (GC) lymphomas Follicular Lymphoma (FL) and Diffuse Large B Cell Lymphoma (DLBCL) are strikingly dependent on mutations affecting chromatin regulators. The majority of FL tumours harbour multiple mutated chromatin-regulatory genes, including loss-of-function mutations in KMT2D (80%), CREBBP (60%) and gain-of-function mutations in EZH2 (25%), whilst these are also frequently mutated within the GC B-cell (GCB) subtype of DLBCL. The FDA approval of EZH2-inhibitor Tazemetostat for relapsed/refractory FL exemplifies the potential and the need to broaden our search for novel avenues targeting epigenetic reprogramming in GC lymphomas. Hence, in this study, we sought to identify novel epigenetic vulnerabilities and biological interactions in GC B cell lymphomas.

Methods and Results:

25 compounds targeting 18 histone binding/modifying enzymes were tested across 12 GCB-DLBCL cell lines harbouring differing degrees of epigenetic dysregulation. The two most potent cytotoxic compounds (GSK-J4 and KDOBA67) both target the H3K27me3 demethylase, KDM6. The KDM6 family consists of KDM6A (UTX) and KDM6B (JMJD3), both of which demethylate H3K27me3/H3K27me2 to lower methylation states, and KDM6C (UTY) which has negligible enzymatic activity. Indeed, KDM6 inhibition (KDM6i) with GSK-J4 increased levels of the inhibitory chromatin mark H3K27me3 and simultaneously reduced H3K27me1/me2, while we observed no global changes in H3K4me3.

The induction of apoptosis in the cell lines following either inhibition of the 'eraser' KDM6 (leading to increased H3K27me3) or the opposing 'writer' EZH2 (lowering H3K27me3) appears paradoxical at first glance, given the EZH2 gain-of-function mutations seen in FL/DLBCL and known efficacy of EZH2 inhibitors. However, RNA-sequencing analysis of GSK-J4-treated cells revealed no significant overlap with published EZH2 mutation or EZH2 inhibition signatures from GC lymphoma models, implying that KDM6i-regulated H3K27me3 peaks do not overlay gene promoters targeted by EZH2 in the GC reaction. The most significant gene regulation by KDM6i was a significant up-regulation of 6 members of the metallothionein (MT) family within 4-hours of treatment, which was sustained for 72-hours (MT1E, MT1F, MT1G, MT1H, MT1X and MT2A at 4 hours log2FoldChange 5.1-12.1, average p adj 3.4E-06). These metal-binding proteins cluster on chromosome 16 and their up-regulation correlated with up-regulation of the zinc transporter SLC30A1, suggesting a potential biological relationship between KDM6 enzymes and zinc homeostasis, with both isoforms containing a zinc-binding domain. Targeted validation across 6 DLBCL cell lines also showed MT up-regulation correlated with sensitivity to GSK-J4, whilst siRNA-mediated knockdown of KDM6A or KDM6B individually recapitulated the up-regulation of MT genes, albeit to a lower extent. ChIP-qPCR profiling at the MT loci for H3K27me3 and H3K4me3 revealed no significant changes at 4 or 24 hours; considering that expression of MT genes increase within 60 minutes of drug exposure, suggesting these changes may be induced by an indirect mechanism of KDM6A and/or KDM6B activity. Using ENCODE data of 4 cell lines included in our series (DOHH-2, OCI-LY-1, SU-DHL-6, KARPAS-422), we observe a link between GSK-J4 sensitivity, up-regulation of MT genes and significantly lower basal enrichment for H3K27ac at the MT loci. This distinction may reflect altered 3D chromatin conformation that render only sensitive cell lines permissive to regulation by KDM6i, with H3K27ac having recently been shown to modulate enhancer positioning and interactions in DLBCL cell lines.

Conclusion:

Screening epigenetic-targeting compounds in GC lymphomas has identified KDM6 as a particularly promising candidate target. Overall, we describe up-regulation of MT genes as one potential mechanism of action, which correlates with response in DLBCL cell lines and suggests that KDM6 members may regulate H3K27me3 in a manner different from EZH2. In our current model we propose, that in GC-lymphomas KDM6A may have tumour suppressive roles whilst KDM6B is required for cell survival. Given KDM6A is frequently lost in cancer and Kabuki syndrome, a congenital developmental disorder caused by germline mutations in KMT2D or KDM6A, our results highlight the importance of establishing what roles the KDM6 family play in lymphoma pathogenesis.