Diffuse Large B-Cell Lymphomas with a Molecular PMBCL Expression Signature Represent a Distinct Molecular Subtype Associated with Poor Clinical Outcome

Introduction: The recently developed DLBCL90 NanoString assay robustly distinguishes primary mediastinal large B-cell lymphoma (PMBCL) from diffuse large B-cell lymphoma (DLBCL), as well as cell-of-origin (COO) subtypes of DLBCL (ABC, GCB, unclassified) and cases with a Double-Hit (DHIT) signature (Ennishi D., JCO 2019). When this assay was applied to biopsies from 343 patients with de novo DLBCL uniformly treated with R-CHOP, nineteen of these cases had a molecular PMBCL signature (mPMBCL), despite the fact that they were diagnosed as DLBCL based on their morphology, immunophenotype and clinical features. Here, we aimed to comprehensively characterize the molecular and clinicopathologic features of these mPMBCL cases.

Methods: Survival estimates were calculated using Kaplan-Meier analysis, using time to progression (TTP) and disease specific survival (DSS) as endpoints. We applied whole-exome sequencing, copy number analysis (SNP6.0) and RNAseq to identify somatic mutations, copy number aberrations and differentially expressed genes, respectively. FISH was applied to assess the presence of rearrangements affecting MYC, BCL2 and BCL6. We used data previously obtained within our centre from a PMBCL cohort (n=73) to compare mPMBCL with "bona fide" PMBCL (PMBCL) tumors (Mottok et al, Blood 2019).

Results: Median age at diagnosis was significantly higher for mPMBCL compared to PMBCL (62 vs 37 years, p<.001). Only 3 out of 19 mPMBCL presented with an anterior mediastinal mass, a clinical hallmark of PMBCL. Importantly, these cases demonstrated extensive extrathoracic disease, which is unusual for PMBCL. Pleural and/or pericardial effusion was seen in only 11% of mPMBCL compared to 50% of PMBCL (p<.001). Bone marrow involvement was seen in 21% of mPMBCL and not observed in any PMBCL (p<.001). The mPMBCL had a trend toward worse outcomes compared to R-CHOP-treated PMBCL (n=61) (2 year TTP: 62% vs 77%, log-rank p=.12; DSS: 74% vs 84%, p=.10). By COO assignment of the DLBCL90 assay, 16 out of 19 mPMBCL were categorized as GCB-DLBCL. Compared to the other GCB-DLBCL in our DLBCL cohort (excluding DHIT-signature-positive cases), the mPMBCL had inferior outcomes (2-year TTP: 62% vs 87%, log-rank p=.03; DSS: 72% vs 89%, p=.02). There were no significant differences in baseline clinical characteristics between mPMBCL and GCB-DLBCL.

Comparison of the mutational landscape of mPMBCL to PMBCL demonstrated perturbations in the central hallmarks of PMBCL pathogenesis: JAK/STAT signaling, NF-ĸB pathway activation and immune evasion. Genomic aberrations affecting JAK-STAT signaling were shared between mPMBCL and PMBCL, with SNVs or indels affecting IL4R, STAT6 and SOCS1 found in 37%, 37%, and 89% of mPMBCL and 36%, 40% and 69% of PMBCL, respectively. Moreover, copy number analysis revealed JAK2 amplifications in 44% of mPMBCL (71% of PMBCL) and differential gene expression analysis showed increased levels of CD274 (PDL1), PDCD1LG2 (PDL2) and genes belonging to the JAK-STAT-signaling network in mPMBCL. In contrast, these genetic aberrations were rarely observed in a recent whole-exome sequencing study on 304 primary DLBCL tumors (Chapuy B., Nat.Med. 2018). Mutations were also observed in NF-ĸB pathways but the patterns of mutations were distinct between mPMBCL (BIRC3 and BTK) and PMBCL (NKBIE) suggesting convergent biology with alternative mechanisms of pathway dysregulation. Similarly, mPMBCL harbored different mutations (CD83) implicated in immune evasion compared with PMBCL (B2M).

Finally, we compared the mutational landscape of mPMBCL with recently described genetically-defined subgroups of DLBCL. Interestingly, a large majority of mPMBCL harbored at least one of the mutations characteristic of "Cluster 4" (incl. CD83, HIST1H1E, SGK1), a subset of DLBCLs defined by Chapuy et al that predominantly includes GCB-DLBCLs.

Conclusion: We have identified and characterized a subgroup of DLBCL that expresses the PMBCL gene expression signature. Similar to bona fide PMBCL, these tumors are characterized by genomic aberrations that affect JAK-STAT, NF-ĸB signaling and immune response. However, our data suggest that dysregulation of the latter two pathways is established through distinct evolutionary modes that are reflected in differential mutation patterns and anatomical and clinical presentations. Our findings provide potential novel therapeutic avenues for this subset of lymphoma.