Molecular predictors of central nervous system (CNS) recurrence in diffuse large B-cell lymphoma (DLBCL).

Abstract

e20082 Background: CNS recurrence is a devastating outcome after treatment of DLBCL. Clinical features poorly predict CNS recurrence risk or the need for prophylactic therapy. We examined whether specific mutational profiles in DLBCL may correlate with the risk of CNS recurrence. Methods: We evaluated DLBCL tumors from patients (pts) with either isolated CNS or systemic (without CNS) recurrence using a 592 gene next generation sequencing (NGS) assay (Illumina NextSeq platform, average coverage depth > 750x; Caris Life Sciences). 3 common molecular subtypes of DLBCL were identified using a simplified classification: (1) MCD subtype ( MYD88L265P or > 2 other mutations in CD79B, PIM1, ETV6, BTG1, TBL1XR1, or PRDM1) , (2) subtype characterized by TP53 mutations (± complex karyotype with multiple structural variants), and (3) GCB subtype characterized by ≥2 mutations in BCL2, CREBBP, EZH2, KMT2D, TNFRSF14, GNA13, MEF2B, or PTEN. We compared prevalence of these subtypes between our groups and unselected DLBCL datasets by Chapuy et al. ( Nat Med; n = 135) and Reddy et al. ( Cell, 2018; n = 1001). We additionally examined clinicopathologic data, including cell of origin, MYC, BCL2 and/or BCL6 rearrangements, and complex karyotype. Results: The study included 26 cases of DLBCL with CNS-only recurrence (n = 13) or systemic-only recurrence (n = 13). We observed no significant difference between these groups in any clinicopathologic characteristics, and no significant difference for mutations in any specific gene (using Fisher’s exact test adjusted for multiple testing). The MCD subtype was more frequent among pts with CNS recurrence (46%) versus systemic recurrence (31%), and was statistically significantly enriched in our CNS recurrence group compared with the unselected DLBCL datasets by Reddy (18%, P= 0.017) or Chapuy (19%, P= 0.030). In contrast, we observed no difference in CNS or systemic recurrence for the TP53 subtype (23% in both groups) or the GCB subtype (15% in both groups), and these proportions were not significantly different compared with the unselected DLBCL datasets ( P> 0.05). Conclusions: In-depth molecular classification of DLBCL using single nucleotide, structural chromosomal, and copy number variants is unfeasible in current clinical practice. Our data demonstrate that a clinically meaningful molecular signature predicting future CNS recurrence could be designed from standard NGS assays using a simplified classification. With further validation, this signature may prove useful for selecting pts for CNS-directed prophylaxis.